32.3.9.1 ..
Sf . _. 2
v 253.2:
$3..“an ..
3:
$3.; ..

‘w

‘ . x z.
Fairgéuéf. .u .2} .
“Esau“?

3.... 5:: I:
.5122 2‘12.
5...; .1
It.
0;.
.5
:32ttxzoz
linufivvl‘ ,
“In; I."
. .sl; .3 . a... k. I. :
3.5.? \M’wrzm «z. “a.
r .33.. .NI
.1 E: ‘u! .
V ‘ a? ‘ I:
11.3%; . bur: 3?.
331.2... v
:3?
its:
3....
. 1.4 2,
A5. 7|»
, . , 1:15;..4.
.3. .JiaeLz-UW
29.2.3..- Zé...é.i ta!
.71....1r‘v .12 3 12...?)
7:2..‘uitta. 53gb-
. 31.5.3... . . .
V . a 9.
, 1 s.
. no .51.
A. 1
.. a .rflwnwa
1,315! 325.1... ) . V3.
2.39. .1: 3313..
$2! :xav‘iqii
.31. r.
A 3.
.Slgtpfw
2..
x g I
“ z 37‘!
'flntrnda :1; v
4 I 31.11)!
133.35

Lin
.39

3‘s!
‘0. 0-15:

I 4.10.

 

 

 

 

 

 

 

 

mes;-

3007

This is to certify that the
dissertation entitled

PATTERNS OF BEVERAGE CONSUMPTION ASSOCIATED
WITH ADOLESCENT OBESITY IN THE US.

presented by

Debra Rose Keast

has been accepted towards fulfillment
of the requirements for the

Doctoral degree in Human Nutrition

 

 

’/_

' J“ v.

Major Professor’s Signature
8 wet; ~ 0 e
Date

MSU is an Affirmative Action/Equal Opportunity Institution

. _ _ __ _ _

LIBRARY
Michigan State
University

 

 

 

 

-.-.- -— -~-<-----.-----a-----.o.-~-c-.-.--.-.--. .

—--.---

PLACE IN RETURN BOX to remove this checkout from your record.
TO AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE DATE DUE DATE DUE

 

FEB 17 2009

 

090710

 

 

 

 

 

 

 

 

 

 

 

 

 

2/05 p:/CIRC/DaleDue.indd-p.1

 

 

 

PATTERNS OF BEVERAGE CONSUMPTION
ASSOCIATED WITH ADOLESCENT OBESITY IN THE US.

By

Debra Rose Keast

A DISSERTATION

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY

Department of Food Science and Human Nutrition

2006

ABSTRACT

PATTERNS OF BEVERAGE CONSUMPTION
ASSOCIATED WITH ADOLESCENT OBESITY IN THE US.

By

Debra Rose Keast

The aims of this cross-sectional study were to examine the associations of
beverages with meal occasions; and to examine associations of beverage patterns with
dietary intake, sociodemographic/lifestyle factors, BMI percentile, percent body fat and
overweight/risk for overweight. Subjects were adolescents age 12-16 years (n=1,872) in
the Third National Health and Nutrition Examination Survey (NHANES D], 1988-94).

BMI percentile and overweight/risk for overweight were determined from
measured BMI using CDC Growth Charts, and percentage body fat from two skinfold
measurements. Youth were classified by their ‘typical’ beverage pattern using the ‘share’
method to identify the combination of beverage types (i.e., milk, juice and/or sugar-
sweetened sofi drinks/ fruit drinks) each reported frequently (i.e., beverage frequency
‘share’ 225% total beverage frequency) in a food frequency questionnaire. The
percentage of beverage occasions that were breakfast, lunch, dinner or snack, and mean
nutrient intakes of beverage groups were estimated using 24-hr recall data. Chi-square
tests were used to examine associations of sociodemographic/lifestyle factors with
beverage patterns. Crude and adjusted associations of beverage patterns with body
weight and body fat were examined by estimating least-square (LS) mean BMI percentile

and percentage body fat, as well as the odds ratios for overweight/risk for overweight.

Energy intakes did not differ, but ‘sugar-sweetened beverages’ groups had higher
added sugars and lower calcium intakes than ‘milk’ groups. Youth drank milk and juice
most frequently with breakfast, while sugar-sweetened beverages were consumed most
frequently at the snack occasion. A greater percentage of the ‘milk & juice’ group than
the ‘sugar—sweetened beverages’ group ate breakfast every day (57% vs. 21%), exercised
daily (35% vs. 28%), played on 23 sports teams (37% vs. 16%), and watched TV 51
hour/day (39% vs. 20%), while less watched TV 24 hours/day (34% vs. 10%). Beverage
patterns were also associated with income and other indicators of SES (p<0.01).

Both BMI percentile and percent body fat were lower in the ‘milk & juice’ group
than in the ‘other beverages (diet soft drinks, coffee, tea)’ group, and BMI percentile was
higher in the ‘sugar-sweetened beverages’ group than in the ‘milk & juice’ group, in
unadjusted models. Overweight/risk for overweight was low in the ‘milk & juice’ group
(OR=O.52, 95% CI = 0.28-0.96), and high in the ‘other beverages’ group (OR=2.74, 95%
CI = 1.12—6.68), compared to the ‘milk’ group. Adjusted associations of ‘milk & juice’
with obesity outcomes were significant in most models, except those adjusted for SES,
physical activity or TV watching. ‘Sugar-sweetened beverages’ was not associated with
BMI percentile in models adjusted for SES or TV watching. Effects of ‘other beverages’
were not significant in models adjusted for youth who were trying to lose weight.

This study suggests that the ‘milk & juice’ pattern was correlated with breakfast
and physical activity, while the ‘sugar—sweetened beverages’ pattern was correlated with
snacks and watching TV. SES, lifestyle factors and beverage patterns were inter-related,
and all were associated with adolescent obesity. Results suggest youth eat breakfast

daily, reduce soft drink intake, and limit TV watching to reduce the risk for overweight.

Copyright by
DEBRA ROSE KEAST
2006

Dedicated to my three children, Matthew, Sandra, and Richard, who inspired me as they
have grown through their adolescent years to become awesome, young individuals

ACKNOWLEDGEMENTS

I will always be grateful for the generous support and guidance received from
many people who have collectively helped me achieve this volume of work. Each of my
Guidance Committee members has made a unique contribution to my dissertation
research, and I would like to thank everyone for giving so much of their time and talents:
Drs. Lorraine Weatherspoon, Norman Hord, Victor Fulgoni, Ellen Velie, and my major
Advisor, Dr. Sharon Hoerr. My highest admiration and deepest gratitude is held for Dr.
Hoerr, who has enabled me to grow tremendously and learn so much from her, as she
worked diligently to “transform” me through this process. I am so grateful for this
wonderful program and the community of fellow graduate students, especially those
members of lab groups, past and present. Special thanks go to Dr. Seung-yeon Lee for
her mentoring as she preceded me in completing her Doctorate. I would also like to
thank family, especially my Mom, husband and children, for the encouragement and
support received over the years. Family stays with you, but after I leave this place, I will
always remember the great faculty and friends I’ve made here at Michigan State
University, and I know that I will continue to interact with many as colleagues

throughout my career.

vi

TABLE OF CONTENTS

List of Tables ................................................................................ x
List of Figures ............................................................................... xiii
Key to Symbols or Abbreviations .......................................................... xiv
Chapter One: Introduction ................................................................ 1
Chapter Two: Review of the Literature ................................................. 9
A. Adolescent obesity ........................................................... 10
1. Secular trends in the prevalence of adolescent obesity
in the US ................................................................... 10
2. Pathophysiology ........................................................... 11
3. Risk factors ................................................................ 14
a. Sociodemographic factors ...................................... 14
b. Parental obesity .................................................. 18
c. Physical activity and sedentary behavior .................... 19
(1. Dietary quality .................................................... 21
4. Assessment of adolescent obesity ........................................ 23
a. BMI and CDC growth charts ................................ 24
b. Anthropometric indices of adiposity ......................... 26
B. Beverage intakes of adolescents ................................................. 28
1. Trends in beverage intakes ............................................... 28
2. Beverage intakes in relation to life style and dietary quality ......... 31
C. Beverage intakes and adolescent obesity ........................................ 32
1. Milk ......................................................................... 33
a. Observational studies ............................................ 34
b. Clinical trials ..................................................... 38
2. Fruit juice ................................................................... 46
3. Sugar-sweetened beverages .............................................. 53
4. Low calorie “diet” soft drinks ........................................... 58
5. Coffee and tea ............................................................... 62
6. Alcoholic beverages ........................................................ 68
D. NHANES dataset selected for research .......................................... 73
Chapter Three: Methods ................................................................... 80
A. Study design ......................................................................... 80
B. Subjects/ study participants ...................................................... 80
1. Dataset ...................................................................... 80
a. NHANES interview and examination components .......... 81
b. NHANES dataset for public use ............................... 81
c. Sample design .................................................... 82
2. Analytic sample ............................................................ 83

vii

C. NHANES data collection procedures ........................................... 85

1. Anthropometric assessment methods .................................... 86
a. Weight ............................................................. 86
b. Height .............................................................. 86
c. Skinfold thickness ................................................ 86
2. Dietary assessment methods ............................................... 87
a. The food frequency questionnaire (FFQ) ..................... 87
b. The 24-hour recall dietary interview ........................... 89
c. Assessment of water intakes .................................. 91
D. Variable definition ................................................................. 91
1. Beverage categories and beverage patterns ............................ 91
a. Categorization of beverages reported in 24-hour
recall dietary interviews ....................................... 92
h. Classification of youth by beverage pattern
frequently consumed ........................................... 93
2. Meal occasion and meal location ....................................... 94
3. Dietary quality assessment ............................................... 95
4. Sociodemographic variables ............................................. 98
5. Lifestyle variables ......................................................... 99
a. Physical activity ................................................. 99
b. TV watching ...................................................... 100
6. Overweight and adiposity ................................................ 101
E. Statistical analysis .................................................................. 107
1. Aim 1 ........................................................................ 107
a. Frequency of beverage consumption ......................... 109
b. Prevalence of beverage patterns ............................... 110
c. Frequency of beverage consumption in beverage
pattern groups .................................................... 111
d. Frequency of beverage consumption at meals ................ 114
e. Mean nutrient intake of beverage pattern groups ............ 115
2. Aim 2 ......................................................................... 117
3. Aim 3 ......................................................................... 119
F. Modeling the relationship between beverage patterns and obesity .......... 122
G. Strengths and limitations .......................................................... 128
Chapter Four: The Meal Context of Beverages Consumed by Adolescents
in the US, Beverage Consumption Patterns and Energy Intake ..................... 130
A. Abstract .............................................................................. 130
B. Introduction .......................................................................... 131
C. Methods .............................................................................. 134
D. Results ............................................................................... 143
E. Discussion ........................................................................... 155

Chapter Five: Risk Factors for Overweight Associated with
Patterns of Beverage Consumption Among Adolescents in the US .................. 168
A. Abstract .............................................................................. 168

viii

B. Introduction .......................................................................... 169
C. Methods .............................................................................. 171
D. Results ............................................................................... 177
E. Discussion ........................................................................... 195
Chapter Six: Patterns of Beverage Consumption Associated with
Adolescent Obesity in the US .............................................................. 204
A. Abstract .............................................................................. 204
B. Introduction .......................................................................... 205
C. Methods .............................................................................. 209
D. Results ............................................................................... 217
E. Discussion ........................................................................... 232
Chapter Seven: Conclusion ................................................................ 242
A. Summary of Findings .............................................................. 242
B. Strengths and Limitations ......................................................... 244
C. Implications ......................................................................... 252
D. Recommendations for Future Research ......................................... 256
Bibliography ................................................................................. 260

ix

LIST OF TABLES

Chapter Two
Table 1. Observational studies of associations of body weight or body

fat with dairy consumption or calcium intake ..................................

Table 2. Randomized control trials of effect on body weight or body

fat fiom increased dairy consumption or calcium supplementation .........

Chapter Three
Table 1. Beverage categories and beverage subcategories corresponding
to beverage items in the NHANES 111 food frequency questionnaire

(FFQ) .................................................................................

Table 2. Dietary Reference Intakes (DRI) for adolescents aged 12-16

years .................................................................................

Table 3. Physician’s assessment of sexual maturation and 5-stage
classification of pubic hair (male and female), male genitalia and

female breasts ......................................................................

Chapter Four
Table 1. Beverage categories and beverage subcategories corresponding
to beverage items in the NHANES III food frequency questionnaire

(FFQ) ................................................................................

Table 2. Mean frequency of beverage consumption (# of times/day)
and the beverages’s percentage share of total beverage consumption
(% of total frequency), using food frequency (FFQ) data of adolescents

aged 12-16 years ....................................................................

Table 3. Distribution (% of adolescents aged 12-16 years in categories)
of the frequency (# of times) of consuming beverage (sub)types in
the FFQ, and distribution (% of adolescents aged 12-16 years in
categories) of the beverage (sub)type’s percentage share of total

beverage frequency in the F FQ ...................................................

Table 4. Percentage of adolescents aged 12-16 years classified by
beverage pattern, i.e., the combination of beverages frequently
consumed in the FFQ, based on criteria for frequent consumption
using the ‘frequency’ method (frequency 21 times/day) and

‘share’ method (share 225% of total fiequency) ...............................

' Table 5. Mean beverage frequency (# of times/day) in the FFQ and
24-hr recall among groups of adolescents aged 12-16 years

35

39

88

97

103

136

144

145

147

classified by beverage pattern, i.e., the combination of beverages

frequently consumed in the FFQ, based on criteria for frequent

consumption using the ‘share’ method (share 225% of total

frequency) ........................................................................... 149

Table 6. Mean beverage frequency (# of times/day) in the 24-hr recall
among beverage consumers aged 12-16 years, and percentage
contribution of meal occasions and meal locations to the total
beverage frequency ................................................................ 153

Table 7. Mean daily nutrient intake from total diet, beverages and
non-beverages among groups of adolescents aged 12-16 years

classified by beverage pattern .................................................... 156
Table 8. Dietary Reference Intakes (DRI) and daily nutrient intake in
gender-age groups of adolescents aged 12-16 years .......................... 160
Chapter Five

Table 1. Beverage categories and beverage subcategories corresponding
to beverage items in the NHANES III food frequency questionnaire

(FFQ) ................................................................................ 174

Table 2. Distribution (% of adolescents aged 12-16 years in categories)
of sociodemographic and lifestyle characteristics, and the prevalence
of overweight/risk for overweight among adolescents classified by
these characteristics ................................................................ 178

Table 3. Distribution of sociodemographic and lifestyle characteristics
within beverage pattern groups of adolescents aged 12-16 years ............ 183

Chapter Six
Table 1. Beverage categories and beverage subcategories corresponding
to beverage items in the NHANES 111 food frequency questionnaire

(FF Q) ................................................................................ 211
Table 2. Mean body weight and body fat measurement used to assess

overweight and adiposity among adolescents age 12-16 years classified
by beverage pattern ................................................................ 218

Table 3. Building the linear regression model of the effects of beverage
pattern on BMI percentile in adolescents aged 12-16 years ................. 220

Table 4. Building the linear regression model of the effects of beverage
pattern on percentage body fat in adolescents aged 12-16 years .............. 225

xi

Table 5. Building the logistic regression model of the effects of beverage
pattern on the likelihood of overweight/risk for overweight in
adolescents aged 12-16 years ..................................................... 229

xii

LIST OF FIGURES

Chapter One
Figure 1. Conceptual model for the association of beverage patterns
with obesity .................................................................. 5

Chapter Three
Figure 1. Conceptual model for the association of beverage patterns
with obesity ................................................................. 123

Chapter Five
Figure 1. Conceptual model for the association of beverage patterns
with obesity .................................................................. 201

Chapter Six

Figure 1. Conceptual model for the association of beverage patterns
with obesity ................................................................. 235

xiii

AI

CDC
CSFH
BNH
CVD
DFE
DHHS
DRI
EAR
EGCG
FDA
FFQ
GTE
IONI
MEC

MSJ

MSJ
Msj

mSJ

KEY TO SYMBOLS OR ABBREVIATIONS

American Academy of Pediatrics

Adequate Intake

Agricultural Research Service

Centers for Disease Control and Prevention
Continuing Survey of Food Intake by Individuals
Body Mass Index

Cardiovascular disease

Dietary Folate Equivalent

Department of Health and Human Services
Dietary Reference Intake

Estimated Average Requirement
Epigallocatechin gallate

Food and Drug Administration

Food Frequency Questionnaire

Green tea extract

Institute of Medicine of the National Academy of Sciences
Mobile Examination Center

‘All caloric beverages’ pattern

‘Milk & sugar-sweetened beverages’ pattern
‘Milk & juice’ pattern

‘Milk’ pattern

‘Sugar-sweetened beverages & juice’ pattern

xiv

msJ

msj
NCHS
NFCS
NHANES
NHLBI

NRC

SMA
USDA
WC
WHR

YRBS

‘Sugar-sweetened beverages’ pattern
‘Other beverages & juice’ pattern

‘Other beverages’ pattern

National Center for Health Statistics
Nationwide Food Consumption Surveys
National Health and Nutrition Examination Surveys
National Heart, Lung, and Blood Institute
National Research Council

Retinol Adjusted Equivalent

Sexual maturation

US. Department of Agriculture

Waist circumference

Waist-hip ratio

Youth Risk Behavior Survey

XV

Chapter 1

INTRODUCTION

The prevalence of overweight among adolescents has been steadily increasing
over the past three decades in the US. Among adolescents age 12-19 years, the current
prevalence rate of 17.4% is more than three times higher than the rate of 5.0% observed
30 years ago (Ogden et al., 2002; Hedley et al., 2004; Ogden et al., 2006). Intakes of soft
drinks have also increased dramatically over the same time period (Cavadini et al.,
2000a; French et al., 2003; Nielsen & Popkin, 2004). In 1994-96, 1998, added sugars
intake from sofi drinks and other sugar-sweetened beverages was 52.6% of the total
amount of added sugars consumed by boys, and 48.5% for girls aged 12-17 years, while
the total added sugars intake was 20% of total energy in both boys and girls (Guthrie &
Morton, 2000). Some researchers have suggested that consumption of sugar-sweetened
beverages is related to the high prevalence of overweight in youth (Hamack et al., 2000;
Bray et al., 2004).

Consumption of milk has declined among children and adolescents
simultaneously with the increased intake of sugared beverages (Morton & Guthrie, 1998;
Moshfegh et al., 2000; Enns etal., 2002; Enns et al., 2003). Further, an inverse
relationship between calcium and adiposity has been hypothesized (Zemel et al., 2000),
which suggests that low fat milk intakes of youth might be inversely correlated with body
fat. Age was associated with decreased intakes of milk and increased intakes of soft
drinks, especially in girls (Lytle et al., 2000; Bowman et al., 2002; Forshee & Storey,
2003). If intakes of milk are low and soft drink intakes are high in adolescents compared

to children of younger age groups, then the low calcium intakes combined with excess

calories fi'om intakes of sugar-sweetened beverages might be associated with the
prevalence of overweight in adolescents.

Several prospective studies (Ludwig et al., 2001; Berkey et al., 2004) and
intervention studies (James et al., 2004; Ebbeling et al., 2006) have shown an association
between soft drinks and weight gain in youth, although cross-sectional studies examining
sugar-sweetened beverage consumption in relation to Body Mass Index (BMI) or
overweight status have had inconsistent results (Forshee & Storey, 2003; F orshee et al.,
2004; Boumtje et al., 2005). Cross-sectional studies using the Continuous Survey of
Food Intake by Individuals (CSFII) or National Health and Nutrition Examination Survey
(NHANES) data (Forshee & Storey, 2003; Forshee et al., 2004; Boumtje et al., 2005)
might have been limited by inaccurate assessment of beverage consumption due to under-
reporting (Bandini et al., 1990; Johnson, 2000; Subar et al., 2003) or intra-individual
variation in intake (Beaton et al., 1979; Sempos et al., 1985). The current study also used
NHANES data, but beverage intakes were assessed using a food frequency questionnaire
(F FQ) rather than the 24-hour recall to address intra-individual variation, and a method
for defining beverage patterns was developed to address under-reporting of intakes.

Cross-sectional studies using the CSFII data to examine the association of
beverage intakes with BMI or overweight in adolescents (Forshee & Storey, 2003;
Boumtje et al., 2005) were limited in that self-reported information rather than measured
height and weight were used to compute BMI (Kuczmarski et al., 2001; Brener et al.,
2003). In addition, analyses of children or adolescents that examined BMI in relation to
beverage intakes (Forshee & Storey, 2003; Forshee et al., 2004) were limited, because

BMI was not adjusted for gender and age using the CDC Growth Charts reference data

(Kuczmarski etal., 2000; Bachman et al., 2006). In the current study, BMI percentile
was calculated from measured height and weight using the gender-specific BMI-for-age
percentile reference data provided by the CDC Growth Charts, because criteria based on
the BMI-for-age percentiles were recommended to assess overweight/risk for overweight
in children and adolescents (Kuczmarski et al., 2000). In addition, percentage body fat
was determined from summed triceps and subscapular Skinfold measurements (Slaughter
et al., 1988), because BMI does not accurately assess body composition, e. g., overweight

youth with a high proportion of lean body mass are not overfat (Forbes, 1987).

Conceptual Model

Lifestyle behaviors, such as watching TV and fast food restaurant use, might be
associated with food and beverage choices that lead to excess energy intake (Coon et al.,
2002; French et al., 2001b). TV watching might be associated with physical inactivity as
well (F orshee et al., 2004), and all might contribute to energy imbalance, weight gain,
and obesity in youth (Bray & Champagne, 2005). Soft drinks are frequently consumed
while snacking and watching TV (Coon et al., 2001; Giammattei et al., 2003; Matheson
et al., 2004; Phillips et al., 2004), and with meals eaten at fast food restaurants (French et
al., 2001b; Bowman et al., 2004). Youth who are characterized by beverage patterns that
include sugar-sweetened beverages might be more overweight than other youth because
of the energy intakes contributed by these beverages (Hamack et al., 1999; Troiano et al.,
2000; Bowman, 2002), or because of the energy intakes contributed by snack foods and
fast foods typically consumed with sugar-sweetened beverages (Harnpl et al., 2003; Utter

et al., 2003; Bowman et al., 2004; Phillips et al., 2004). A model is needed to address the

complexity of these relationships.

The conceptual model in Figure 1 shows the beverage consumption pattern being
hypothetically determined by the meal context in which beverages are consumed. If
beverage choice is dependent upon the meal occasion (e. g., breakfast, lunch, dinner,
snack) and meal location (e. g., home or places away from home such as school or a fast
food restaurant), then the beverage pattern can be a marker for the lifestyle variables that
determine the frequency with which that meal context occurs for youth. If milk and
100% mu juice were frequently consumed at breakfast, a habit of eating breakfast
regularly might determine whether youth are classified as milk and 100% mu juice
consumers. If sugar—sweetened beverages were frequently consumed at fast food
restaurants and as snacks, fast food restaurant use and TV watching might determine
whether youth are classified as sugar-sweetened beverage consumers. In turn, youth
classified by beverage patterns associated with fast food restaurant use or snacking
behavior might then have higher energy, fat, sugar, and lower calcium intakes from
beverages and from the total diet, compared to youth classified by beverage patterns
associated with breakfast consumption. Energy intake is the mediator between beverage
patterns and obesity. The model shown in Figure 1 depicts the interrelationships of the

study aims described next.

Specific Aims
The current dissertation research used data from NHANES III, 1988-94, in a

cross-sectional analysis to examine beverage patterns and other potential risk factors for

563° .53 @5933 03.85: he 33:33.3 2: .8.“ .28:— EEnooaoU A 953%

 

 

 

 

/_—\

\

//

25:52:5—
3.55—

9

«

 

 

55%
.31.;

 

 

 

 

 

3—3:— 3.5:”.—

 

 

Aw

 

 

 

 

 

.3332. as is?
153.4 .ufiu—efim
Em 3.5.5.—
32— : we mam .252:
awn—:52..."— .aemwom

a

magma—cw— ...o—Eomv .ow<

 

33.58.50 15:89—

 

F535—
owaaogm

 

union—a3 >H.

 

 

 

 

 

 

 

 

gauged =32
:2330 .32

 

 

 

 

 

 

 

overweight in youth age 12-16 years in the US. The overall goal of this research was to
examine the association of beverage patterns with adolescent obesity (i.e., overweight
and adiposity). The associations of beverage patterns with nutritional, sociodemographic
and lifestyle risk factors for overweight were also examined. The following were the

three (3) specific aims of this study.

Aim 1: To examine the frequency of consuming beverage types, and the percentage of
youth classified into groups by their ‘typical’ beverage pattern (i. e., the combination of
beverage types most frequently reported as ‘usual’ intake using a F F Q); to examine
the association of beverages with meal occasions and meal locations; and to examine
the mean 24-hour recall intake of energy, dietary fat, total sugars, added sugar,
calcium and other nutrients in beverage pattern groups.

Four main types of beverages examined included (1) milk and milk drinks, (2)
100% fruit juice (3) sugar-sweetened beverages (fruit drinks and regular sofi drinks) and
(4) other beverages (diet soft drinks, coffee/tea and alcoholic beverages). Youth were
classified by their ‘typical’ beverage pattern using two different methods to identify the
combination of beverage types (i.e., milk, juice and/or sugar-sweetened beverages) that
were each reported frequently in the F FQ. Criteria to identify the most frequently
consumed beverage types using the ‘frequency’ and ‘share’ methods, respectively, were
frequency 21 time/day, and share 225% of total beverage frequency.

Mean energy, dietary fat, total sugars, added sugar, calcium and other nutrient
intake (in the 24-hour recall) of beverage pattern groups were compared to examine the

association of beverage patterns with energy and nutrient intake. The beverage pattern

groups’ mean intake of water (ounces) and mean intakes of milk, 100% fruit juice, sugar-
sweetened beverages, other beverages (e.g., diet soft drinks, coffee/tea and alcoholic
beverages) (times/day), in both the FF Q and 24-hour recall, were also examined. The 24-
hour recall data were used to examine mean beverage frequency at meals (e.g., breakfast,
lunch, dinner/supper and snacks) consumed at-home or at places away-from-home (e. g.,

school, fast food restaurants).

Aim 2: To examine the distributions of sociodemographic and lifestyle characteristics
that are risk factors for overweight among beverage pattern groups defined above in
Aim 1.

Youth were classified into groups by their ‘typical’ beverage pattern using the
‘share’ method (i.e., share 225% total beverage frequency) to identify the combination of
beverage types (e.g., milk, 100% fruit juice and sugar-sweetened beverages) most
frequently reported as ‘usual’ intake in the FFQ. To identify risk factors for overweight,
the prevalence of overweight/risk for overweight (i.e., BMI 285th percentile of BMI-for-
age) was estimated for youth classified by each sociodemographic/lifestyle characteristic.
To examine the distributions of risk factors for overweight, the percentage of each
beverage pattern group classified by each sociodemographic/lifestyle characteristic was
estimated, and the percentage of each beverage pattern group classified by weight status
(e. g., underweight, normal, at risk for overweight, overweight) was estimated as well.
Sociodemographic factors examined were gender, age, race-ethnicity, region,
urbanization, parental overweight/obesity, household income and other indicators of

soCial economic status (SES). Lifestyle factors included smoking, alcohol use, weight

loss behavior, physical activity (i.e., exercise and team sports participation), TV
watching, frequency of breakfast consumption, and type of milk (e. g., whole, 2%, 1%,

skim) usually consumed.

Aim 3: To examine the association of adolescent obesity with the beverage pattern
groups defined above in Aim I, adjusted for potential covariates of interest.

Indicators of obesity (i.e., overweight and adiposity) examined separately
included BMI percentile, percentage body fat as assessed by Skinfold measurements, and
overweight/risk for overweight (i.e., BMI 285th percentile of BMI-for-age). Youth were
classified into groups by their ‘typical’ beverage pattern using the ‘share’ method (i.e.,
share 225% total beverage frequency) to identify the combination of beverage types (e.g.,
milk, 100% fruit juice and sugar-sweetened beverages) most frequently reported as
‘usual’ intake in the FFQ. In addition to beverage pattern groups, other beverage
variables considered in analyses were fluid ounces of water and type of milk (e. g., whole,
2%, 1%, skim) usually consumed. Covariates considered in analyses were gender, age,
race-ethnicity, stages of sexual maturation, region, urbanization, parental overweight/
obesity, SES indicators, smoking, alcohol use, weight loss behavior, exercise, team sports
participation and TV watching. Covariates were examined separately one at a time to
investigate whether each one might potentially confound the relationship between

beverage patterns and overweight/adiposity.

Chapter 2

REVIEW OF THE LITERATURE

The purpose of this review is to describe the concurrent trends in adolescent
obesity and concomitant factors, such as beverage consumption, and to describe prior
investigations of the hypothesized relationship between beverages and overweight in
youth. The review of the literature begins with a description of how the prevalence of
adolescent obesity in the US. has been increasing since the late 1970’s (Ogden et al.,
2002). Many overweight and obese1 adolescents remain so throughout adulthood.

Obesity-related health conditions affecting young age groups have an increased
lifetime burden, thus an understanding of the risk factors associated with adolescent
obesity is important to public health. Potential risk factors for obesity include socio-
demographic variables, as well as lifestyle variables, such as physical activity and dietary
intake. Beverage intakes are also correlated with lifestyle variables that affect dietary
quality and the risk for obesity, such as TV watching, fast food restaurant use and
breakfast consumption, thus a short review of these factors is included. Next, the types of
beverages consumed by youth are discussed, because each one is either positively or
inversely associated with body weight or body fat. This review concludes by describing
why the dataset that was selected to investigate the relationship between beverage

patterns and obesity in adolescents is appropriate for this analysis. As each section is

 

‘ The terms overweight and obesity are used interchangeably in this review. Overweight is defined by the
Centers for Disease Control and Prevention (CDC) reference criteria for children aged 2-19 years based on
age- and sex-specific percentiles for body mass index (BMI) (Kuczmarski et al., 2000). In a recent report
fi'om the Institute of Medicine (Koplan et al., 2005), “obesity” in children and adolescents referred to a
BMI 2 95‘h percentile. By late adolescence, the 95‘h percentile is approximately 30 kg/m‘, the BMI criteria
used to define obesity in adults age 20 years and older (NHLBI, 1998).

related to the research proposal, justifications for aspects of the analysis plan are provided

as appropriate throughout the literature review.

A. Adolescent obesity
1. Secular trends in the prevalence of adolescent obesity in the US.

The CDC has been tracking the prevalence of overweight among children and
adolescents age 2-5, 6-11 and 12-19 years by monitoring trends in the percentage of
youth whose Body Mass Index (BMI) exceeds the reference percentiles of BMI-for-age.
The subjects of the current study, adolescents age 12-16 years, were a subgroup of the
latter age group of youth for which national survey data is available, therefore this section
describes the trends among those aged 12-19 years.

The prevalence of overweight among adolescents age 12-19 years has been
steadily increasing over the past three decades in the United States. Recent NHANES

data indicated that in 1999-2002, 30.9% of adolescents age 12-19 years had a BMI 2 85th

percentile, and 16.1% had a BMI 295th percentile (Hedley et al., 2004). The latest figures
from the 2003-2004 NHANES were increased to 34.3% of adolescents age 12-19 years
having a BMI 285th percentile, and 17.4% had a BMI 295th percentile (Ogden et al.,
2006). Although there were no differences in the prevalence of BMI 295th in children
and adolescents between 1999-2000 and 2001-2002 (Hedley et al., 2004), significant tests
for trends indicated an increase in the prevalence rates between 1999-2000 and 2003-
2004 (Ogden et al., 2006).

There were no significant differences in the prevalence of overweight in

adolescents between NHANES 1, 1971-74, and NHANES II, 1976-80, but the prevalence

10

of overweight in adolescents doubled between NHANES H, 1976-80, and NHANES III,
1988-94, by increasing from 5.0% to 10.5% (Ogden et al., 2002). The rate increased
again by 5.6% to an overweight prevalence of 16.1% in adolescents between NHANES
III, 1988-94, and NHANES, 1999-2002 (Hedley et al., 2004), which was more than three
times higher than the rate of 5.0% observed in NHANES H, 1976-80. As noted above,
the rate has continued to increase in recent years to an overweight prevalence of 17.4% in
adolescents in NHANES, 2003-2004 (Ogden et al., 2006).

Over time, the distribution of BMI-for-age has become more skewed to the right
for children and adolescents aged 12-19 years (Flegal & Troiano, 2000). This shows that
the heaviest children are becoming heavier, which has important health implications,
because the risk for comorbities increases for children at the highest levels of BMI. Rates
of hypertension and hyperinsulinemia were twice as high for children above the 97th
percentile as for children between the 95th and 97th percentiles of BMI-for-age in the
Bogalusa Heart Study (Freedman et al., 1999). Thus, as the BMI levels become more
extreme in the adolescent population, the rate of obesity-related comorbities will be

higher than before.

2. Pathophysiology
The prevalence of obesity-related health conditions has dramatically increased
among children. Many obesity-related disorders can occur during childhood and
adolescence (Dietz, 1998; Must & Strauss, 1999; Deckelbaum & Williams, 2001; Must &
Anderson, 2003). The most common pathophysiologies are glucose intolerance/ insulin

resistance, and other aberrations of the metabolic syndrome. Other common

11

pathophysiologies are cardiovascular lesions such as hypertension, dyslipidemia, left
ventricular hypertrophy and atherosclerosis (Daniels et al., 2005).

The prevalence of metabolic syndrome was almost 30% among obese adolescents
age 12-19 years (Cook et al., 2003). It is diagnosed when at least three of the five
following symptoms are present: glucose intolerance, abdominal obesity,
hypertriglyceridemia, low high-density lipoprotein (HDL) cholesterol, and high blood
pressure (NHLBI, 2002; Cook et al., 2003). Components of the metabolic syndrome
contribute to the development of atherosclerosis, thus cardiovascular disease is advanced
with obesity in youth as well (Cook et al., 2003). In a population-based sample,
approximately 60% of obese children aged 5-10 years had at least one physiological
CVD risk factor, such as elevated total cholesterol, triglycerides, insulin, or blood
pressure, and 25% had two or more of these CVD risk factors (Freedman et al., 1999).

Increased body fat and insulin resistance are also risk factors for type 2 diabetes in
childhood obesity (Goran et al., 2003). Because the degree of insulin resistance in
children increases with the severity of body fatness (ADA, 2000), more children will
reach the diagnostic threshold for type 2 diabetes as more obese children become even
fatter. A 10-fold increase in the prevalence of type 2 diabetes in children occurred
between 1982 and 1994 (Pinhas-Hamiel etal., 1996). Although the prevalence of type 2
diabetes is only at the rate of 1-2 cases per 1000 in adolescence (Fagot-Campagna et al.,
2001), estimates are that three-fourths of obese adolescents will track to be overweight as
adults (Guo et al., 2002). Because type 2 diabetes is a consequence of adolescent obesity
occurring later in life, the prevalence of type 2 diabetes has increased greatly among

young adults as well (Mokdad et al., 2003). Thus, the burden of developing diabetes at a

12

young age is that major complications of diabetes, such as retinopathy, nephropathy, and
neuropathy, end-stage renal failure, cardiovascular disease, and even death will occur
earlier in life.

Some less common but still serious physiologic disorders of obesity in children
include steatosis and cholelithiasis affecting hepatic function, and proteinuria that can
result from damage to the renal system (Must & Anderson, 2003; Daniels et al., 2005).
Some obese girls will suffer from menstrual abnormalities and other symptoms of
polycystic ovarian syndrome during adolescence (Buggs & Rosefield, 2005).
Pseudotumor cerebri is a rare but serious neurological disorder that can occur in
childhood obesity, and obstructive sleep apnea is another life-threatening condition (Must
& Anderson, 2005; Daniels etal., 2005). Asthma is also exacerbated by childhood
obesity (Gold et al., 2003). Excess body weight during adolescence can cause orthopedic
problems like slipped capital femoral epiphysis and Blount’s disease.

Health professionals have observed that, because of the rapid grth and
development that occurs during puberty, adolescence appears to be a critical time for
obesity onset that persists into adult years (Dietz, 1994; Lawlor & Chaturvedi, 2006), and
up to 75% of overweight adolescents will become obese adults (Guo etal., 2002). Adult
obesity is associated with a myriad of health conditions that potentially shorten the
lifespan especially among those who become obese at a young age. Thus, early
identification and amelioration of easily modifiable factors contributing to childhood
obesity is important to reduce the comorbities that occur both in adolescence and later in

life.

13

3. Risk factors.

Potential confounders of the association between beverages and overweight
include sociodemographic factors, parental obesity and physical activity. The
relationship between beverage patterns and obesity is mediated through dietary factors
that affect energy intake. This section reviews how sociodemographic factors, parental
obesity, physical activity and dietary factors are correlated with the increased prevalence
of childhood and adolescent obesity.

a. Sociodemographic factors.

Overall, both boys and girls have shown similar increased rates of overweight
(Ogden et al., 2002; CDC, 2003a). However, children in certain race-ethnic groups, and
those living in certain regions of the United States had higher rates of obesity than other
groups in the population (Hedley et al., 2004; Strauss & Pollack, 2001). Family
composition, number of family members, household income and other parameters of SES
all potentially influence the risk for overweight in children of those families (Koplan et
a1,2005)

Ethnic groups. Among adolescents age 12-19 years, the prevalence of obesity is
21% among non-Hispanic blacks and 22.5% in Mexican-Americans compared to 13.7%
in non-Hispanic whites (Hedley et al., 2004). Among boys, the prevalence of obesity is
higher in Mexican-Americans (24.7%) than in non-Hispanic blacks (18.7%), although
non-Hispanic black girls (23.6%) have the highest prevalence compared to Mexican-
American girls (19.9%) (Hedley et al., 2004). Ethnic disparities evolved from the
accelerated increase in prevalence of obesity among minority groups. Between 1986 and

1998, there was a 120% increase in the prevalence of obesity in African American and

'14

Hispanic youth, compared to a 50% increase in non-Hispanic whites (Strauss & Pollack,
2001). It is unclear why these differences consistently appear and to what degree they are
driven by the social environment, built environments, income and education of parents
and/or familial food and physical activity preferences and practices (Link & Phelan,
1995; Diez Roux, 2005; Sallis & Glanz, 2006; Gordon-Larsen et al., 2003).

Region. In 1998, the obesity prevalence among youth was 17.1% in southern
states compared to 10.8% in western states (Strauss & Pollack, 2001). BMI determined
from self-reported height and weight of children and adolescents participating in the
Youth Risk Behavior Survey (YRBS) reported by states (CDC, 2004) show statewide
variations in the prevalence of overweight and obesity, but no clear pattern across broad
regions of the US. emerged. Reasons for statewide differences in the prevalence of
adolescent obesity are not entirely clear, but differences between race-ethnic groups
(Hedley et al., 2004), local economies (Finkelstein et al., 2005), climate (Steegrnan,
2005) and even local school curricula and policies (Kubik et al., 2005) have all been
considered as possible contributors to geographic distributions of overweight. After
controlling for race-ethnicity, income, diet, physical activity and television viewing,
when differences between local communities were averaged over four broad regions
(Northeast, Midwest, South and West), the geographic area was not associated with BMI
in either children or adolescents in NHANES 111, 1988-94, but BMI of children in the
West was lower than that of children in the Northeast in CSFII, 1994-96 and 1998
(Storey et al., 2003).

Family characteristics. Marital status, and among married, age of marriage,

determines family composition of households with children (NRC, 2003). The age and

15

gender of the family reference person, whether he/she is employed, and the household
size then potentially influence family meal patterns, which affect overweight in youth
(Anderson et al., 2003; Videon & Manning, 2003). The percentage of mothers with
children under age 18 who work has increased to 72% from 47% in 1975 (U .S.
Department of Labor, 2004). Since 1965, breakfast consumption among adolescents has
continued to decline (Siega-Riz et al., 1998), perhaps because more mothers are working,
and children of working mothers are more likely to skip meals than those whose mothers
are not employed outside the home (Crepinsek & Burstein, 2004). As more women are
working outside the home, home-prepared family meal times occur less frequently. More
meals are eaten away from home also because smaller households (due to delayed
marriage, higher divorce rates and single parenthood) have fewer economies of scale in
home preparation of meals than larger families (Sturm, 2004), and with the availability of
fast food restaurants, the cost of meals eaten away from home are less than in the past.

Family income and SES. An increase in obesity prevalence among African
American youth was greatest for those with the lowest household income (Strauss &
Pollack, 2001). The National Longitudinal Study of Adolescent Health also
demonstrated an inverse relationship between obesity prevalence and family SES
(Goodman, 1999; Goodman et al., 2003). In NHANES III, 1988-1994, an inverse
relationship between family income and obesity was observed in adolescents of white but
not other race-ethnic groups (Troiano & Flegal, 1998; Alaimo et al., 2001; Miech et al.,
2006), but caution interpreting these data was advised because many estimates had a
large (>30%) coefficient of variation (Troiano & Flegal, 1998). Prevalence of

overweight increased among poor (those living in households with income below the

16

poverty line) compared to nonpoor for non-Hispanic white adolescents age 15-17
(p<.01), but not for those age 12-14 years. For non-Hispanic blacks, a trend (p<.1)
towards increased prevalence of overweight among poor compared to nonpoor among
adolescents age 15 -1 7 was reversed among the younger age group in that more nonpoor
than poor adolescents age 12-14 were overweight (p<.1) (Miech et al., 2006).

An increased prevalence of overweight among low-income adolescents may be
due to the quality of their diets, if only less expensive, energy-dense fOOds and beverages
are consumed (Alaimo et al., 2001). More affordable foods and beverages tend to have a
higher energy density because of their fat and sugar content, thus the dietary patterns of
low-income individuals could result in excessive energy intake and weight gain
(Drewnowski & Specter, 2004; Drewnowski & Dannon, 2005). The current study
investigated whether sugar-sweetened beverages were associated with greater energy
intakes and a greater risk for overweight. Sugar-sweetened beverages but not other
energy-dense snack foods such as baked goods, ice cream, chips and candy, were
associated with increased BMI z-scores over 4 years in female adolescents without
controlling for income (Phillips et al., 2004). Dietary factors associated with poverty
among adolescents age 12-17 years in NHANES, 1999-2002, included sugar-sweetened
beverage consumption and breakfast skipping, but percentage of calories from snacks and
percentage of calories eaten away from home did not differ by poverty group (Miech et
al., 2006). If poverty is driving disparities in sugar-sweetened beverage consumption,
breakfast skipping and overweight in adolescence (Miech et al., 2006), then poverty
might explain a positive association between a sugar-sweetened beverage pattern and

overweight and an inverse association between a breakfast beverage pattern and

17

overweight. Relationships between beverage patterns and overweight would then be

mitigated because income was included in the model of the current study.

b. Parental obesity

Both genetics and the family environment can affect a child’s risk for overweight
if one or both parents are obese. Children can inherit obesity susceptibility genes from an
obese parent or parents; or can be exposed, after birth, to diet and activity patterns that
promote obesity. Moreover, recent research suggests that an altered intrauterine
environment may be a third way that parental obesity can affect the child. There are a
number of potential mechanisms by which maternal obesity in pregnancy may promote
offspring obesity (Whitaker & Dietz, 1998; Levin, 2000; Oken & Gillrnan, 2003). For
example, obese mothers are more likely to experience diabetes in pregnancy, and some
evidence suggests that the offspring of mothers who have diabetes in pregnancy may
have an increased risk of developing obesity later in life (Silverman et al., 1998). Given
these mechanisms that may promote intergenerational obesity, prevention efforts should
be targeted towards women of childbearing age as well as their children.

Because parental obesity was found to be a risk factor for overweight in
adolescents age 12-16 years, researchers using the NHANES III dataset investigated
factors that might potentially protect against overweight in this high risk group (Fiore et
al., 2006). Eating breakfast some days or every day was associated with normal body
weight only among those having one or two obese parents, while sports team
participation was associated with normal body weight only among those with two

nonobese parents (F iore et al., 2006). Other results from this study indicated that

18

overweight adolescents might be over-reporting water consumption and under-reporting
energy intake (Fiore et al., 2006). That breakfast consumption was associated with
normal body weight (F iore et al., 2006) suggested that a milk and juice beverage pattern

might be protective against adolescent overweight after controlling for parental obesity.

c. Physical activity and sedentary behavior

Youth have opportunities to be physically active while walking or bicycling to
and fi'om school, attending physical education classes, participating in team sports or
exercising during their free time. Based on the Youth Risk Behavior Survey (YRBS),
one-third of high school students are not engaging in recommended levels of moderate or
vigorous physical activity and 10% report that they are inactive (CDC, 2003b, 2004).
From 1977 to 2001, the percentage of total trips in which children aged 5 to 15 years
walked to school declined from 20.2% to 12.5% (Sturm, 2005b). High school students
who participated daily in physical education classes declined from 42% in 1991 to 25%
in 1995 (DHHS, 1996b) and increased slightly to 28.4% in 2003 (CDC, 2004). A time-
use study showed that children aged 3 to 12 years had less free time available to exercise
at play, as more time was spent participating in organized sports in 1997 than in 1981
(Sturm, 2005a), but there may be less opportunities for adolescents to participate in team
sports (Elkins et al., 2004)

Physical activity levels decline substantially during adolescence, especially in
girls (Kimrn et al., 2002). Enrollment requirements and resources available for physical
education classes are reduced as students move from elementary and middle school

grades to the high school level. Whereas recreational sports teams programs are

19

accessible to children of younger ages in many communities, the increased
competitiveness of sports preclude many adolescents from participating in high school.
As girls in a longitudinal study grew older, a higher BMI was associated with greater
reductions in physical activity (Kim et al., 2002). Among high school students in the
YRBS, overweight boys were less likely to engage in moderate physical activity, and
girls who were either overweight or at risk for overweight were less likely to participate
in sports (Levin et al., 2003). Afier controlling for age, grade and football, sports
participation was associated with lower BMI among Afiican American students of both
genders in low-income, inner city high schools (Elkins et al., 2004). While youth with
lower BMI might be more likely to participate in sports, because physical activity
“enrichment” has effectively reduced percent body fat in higher risk children (Burke et
al., 1998), obesity could be prevented if more opportunities for organized exercise
programs were available to all high school students (Elkins et al., 2004).

Hours spent watching TV is inversely associated with physical activity (Crespo et
al., 2001), thus as a measure of sedentary behavior, TV watching could be correlated with
obesity as well. Trends in TV watching have increased in parallel to the rise in adolescent
obesity. In 1999, 98% of households had a television (Nielsen Media Research, 2000)
compared to 10% in 1950 (Putnam, 1995). There was more than one television in 88% of
homes in 1999 (Roberts et al., 1999) compared to 35% in 1970 (Lyle & Hoffman, 1972).
The percent of homes with three or more televisions increased ten-fold over the same
time period (Rideout et al., 2003). In 2003, 50% of children less than 6 years of age had
three or more televisions in their home, and 36% had a television in their bedrooms

(Rideout et al., 2003). During a typical day 36% of children watched television for one

20

hour or less, 31% watched TV for one to three hours, 16% watched it for three to five
hours, and 17% spent more than 5 hours watching television (Roberts et al., 1999).
However, among adolescents, between 1990 and 2001 there has been a decline in
watching television 3 hours or more, although there was an increase in television
watching for one hour or less (Child Trends, 2002). Children are now using other types of
electronic media as well, including video games and computers (Roberts et al., 1999;
Rideout et al., 2003), that adds to the amount of time in sedentary activities.

Nevertheless, youth spend a significant amount of discretionary time watching television,
and this modifiable sedentary behavior has a significant potential impact on overweight

in youth.

(1. Dietary quality.

Trends in eating patterns, such as increased snacking behavior, increased
consumption of meals away from home, and increased portion sizes contribute to
increased intakes of energy-dense foods and beverages. Energy-dense snacks comprise a
large proportion of total daily calories (J ahns et al., 2001). Trends in away-from-home
food consumption are partly due to greater availability of convenient, shelf stable,
portable foods that are accessible in many places at any time (Food Marketing Institute,
1996, 2003; French et al., 2001a; Sloan, 2003). Away-from-home foods contributed 20%
of children’s total caloric intake in 1977-1978 and 32% in 1994-1996 (Lin et al., 1999).
Foods eaten outside the home decrease the nutritional quality of children’s diets (Lin et
al., 1999; French et al., 2001b). Portion sizes of most foods consumed both at home and

away from home have increased between 1977 and 1996 (Nielsen & Popkin, 2003;

21

Smiciklas-Wright et al., 2003), and increased portion sizes are associated with overeating
(Rolls, 2003).

Total energy intake. Between 1977-1978 and 1994-1996, total calories
consumed by adolescent boys aged 12 to 19 years increased by 243 from 2,523 to 2,766
calories (Enns et al., 2003). At the same time, total calories consumed by adolescent girls
aged 12 to 19 years increased by 123 from 1,787 to 1,910 calories (Enns et al., 2003). No
significant increased trends in energy intake were observed in children aged 6-11 years
between 1977-1978 and 1994-1996, 1998 (Enns et al., 2002).

Total fat consumption. Between 1965 and 1996, the proportion of energy from
total fat consumed by children declined from 39% to 32%, and saturated fat from 15% to
12% (Cavadini et al., 2000a). Children aged 6-11 years in 1994-1996, 1998 consumed
25% of calories from discretionary fat (USDA, 2000; Enns et al., 2002). For adolescents
aged 12-19 years, girls consumed 25% and boys consumed 26% of their calories from
added fat (USDA, 2000; Enns et al., 2002).

Added sugars. Children aged 6-11 in 1994-1996, 1998 consumed 21-23
teaspoons of added sugars in a 1,800-2000 calorie diet, while the Food Guide Pyramid
recommended no more than 6-12 teaspoons of added sugars for a 1,600-2,200 calorie diet
(USDA, 1996; Enns et al., 2002). Recommendations of the MyPyramid personalized diet
plan are expressed as limits on calorie intake from extra fats and sugars. For example, no
more than 290 discretionary calories are recommended for a 2,200 calorie diet (Johnston,
2005). The Institute of Medicine (2002) Dietary Reference Intakes Report suggested an
upper limit of 25% of calories for added sugars to avoid displacement of micronutrients

from the diet, but the DRI should not be interpreted as a dietary guideline (Sibbald,

22

2003). Both the former US Food Guide Pyramid (USDA, 1996) and the World Health
Organization (2003) recommended limiting added sugar to 10% of calories, and this
guideline for added sugar intake was used in the current study.

The theoretical model that was the basis for the current study suggested that the
relationship between beverage patterns and obesity might be mediated through dietary
factors that affect energy intake. For example, a hypothesis of the current study was that
a sugar-sweetened beverage pattern would be associated with overweight, because youth
who consume sugar-sweetened beverages might have greater energy intakes than youth
who consume other beverages. Both beverages and non-beverages contribute energy to
daily intakes. Energy intakes are derived from added sugars in sugar-sweetened
beverages, as well as from the added sugars and dietary fats in foods consumed at meals
accompanied by sugar-sweetened beverages. Energy-dense foods high in fat and sugar,
as well as sugar-sweetened beverages, are often consumed at the snack occasion and
meals eaten at fast food restaurants. Consuming beverages containing added sugar,
energy-dense snack foods, and meals away-from-home are all possibly related to the

likelihood of being overweight.

4. Assessment of adolescent obesity.

The best data to assess the prevalence of overweight and adiposity among youth
in the US. are available from the National Health and Nutrition Examination Surveys
(NHANES) conducted by the National Center for Health Statistics (N CHS). NHANES
anthropometric data included measured, rather than self-reported, height and weight,

from which the Body Mass Index (BMI) was determined and compared to the CDC

23

reference criteria of BMI-for-age. In addition, the current study used Skinfold thickness
measurements to determine percentage body fat. This section will review literature
pertinent to the use of BMI and Skinfold thickness measures to assess overweight and

adiposity in youth.

a. BMI and CDC growth charts.

BMI, defined as weight in kilograms over height in meters squared (kg/m2), is a
surrogate measure of body fatness used to assess the prevalence of overweight in youth.
The Centers for Disease Control and Prevention (CDC) reference criteria for overweight
in children aged 2-19 years is based on age- and sex-specific percentiles for BMI. A
BMI >5th percentile and <85th percentile is considered normal weight for height; the
category for those with a BMI between the 85‘h and 95th percentiles is “at risk for
overweight”; and “overweight” is defined as a BMI 295th percentile (Kuczmarski et al.,
2000). Although pediatricians and other health professionals are encouraged to interpret
the “at risk for overweight” and “overweight” criteria as guidelines used to identify
patients whose BMI warrant further clinical investigation and monitoring at followup
visits, the BMI percentiles are not firnctional definitions of obesity. The criteria used to
define overweight, the percentiles of BMI-for-age, are merely statistics used to describe
the extremes of the distribution of BMI-for-age in the reference population (F legal, 1993;
Trioano & Flegal, 1999).

In a recent report from the Institute of Medicine a BMI 295th percentile was used
to define “obesity” in children and adolescents (Koplan et al., 2005). The term “obesity”

was used by the Institute of Medicine to be consistent with international standards

24

(Lobstein et a1. 2004), and also to remind the readers that addressing the increasing
prevalence of BMI 295‘h percentile is an urgent priority, because obesity is a serious
medical condition. The 95‘h percentile in the BMI distribution among adults is 30 kg/m“,
and at this point in the BMI distribution, mortality from obesity-related health conditions
was found to steadily increase (NHLBI, 1998).

Although only one set of criteria, 25 kg/m2 and 30 kg/m‘, is used to define
overweight and obesity, respectively, in adults of all ages, using BMI as a reference for
children and adolescents is complicated, because BMI varies with age in youth. For this
reason, gender- and age-specific BMI percentiles are used as references for children age 2
to 19 years. By late adolescence, the 95‘h percentile approaches 30 kg/mz, the criteria
used to identify obesity in adults. Cole et a1. (2000) determined which percentiles
corresponded to a BMI of 25 kg/m2 or 30 kg/m2 at age 18 years for each of six countries:
Brazil, Great Britain, Hong Kong, the Netherlands, Singapore, and the United States. For
example, 30 kg/m2 is the 99‘h percentile for 18-year-old males in Great Britain, thus the
99‘‘1 percentile would be used over all age groups of British children to define obesity,
while 30 kg/m2 corresponds to the 97‘h percentile at age 18 years for boys in the US. To
ease international comparisons, Cole et a1. (2000) proposed averaging the six countries’
percentile curves that converged with the adult definition of overweight and obesity at
age 18 years.

The selection of the 85‘h and 95‘h percentiles to define overweight and obesity in
children is arbitrary, and the CDC BMI-for-age percentiles pertain only to the reference
population in the US. Because overweight was already increasing in NHANES III,

1988-94, rather than using the most recent NHANES to revise the grth charts in 2000,

25

CDC based the revised growth references on distributions in national surveys conducted
from 1963 through 1980 for children aged 6 to 19 years (Kuczmarski etal., 2000).
Appropriate use of the reference percentile distributions include monitoring changes in
the prevalence of overweight and obesity over time, or comparing the extent of
overweight across population subgroups.

In the current study, adolescents age 12-16 years were classified by patterns of
beverages consumed, and the prevalence of overweight and obesity in these beverage
pattern groups were compared, to determine whether beverage consumption was

associated with overweight and obesity among adolescents in the US.

b. Anthropometric indices of adiposity.

BMI does not accurately assess body composition, e.g., overweight youth with a
high proportion of lean body mass are not overfat (Forbes, 1987). Intake of calcium from
milk might be related to body composition or adiposity in general (Zemel et al., 2000).
Intake of fructose from soft drinks might be related to the distribution of body fat (Elliot
et al., 2002). Truncal adiposity, i.e. body fat deposited at the waistline, is a component of
the metabolic syndrome (Duncan et al., 2004), and as such, it might be positively
associated with intakes of sugared beverages (Elliot et al., 2002). Alternatively, milk and
dairy foods may protect against the development of metabolic syndrome (Pereira et al.,
2002), or milk may appear to be protective because of the inverse association between
milk and soft drinks (Huang & McCrory, 2005). A strength of the current study was that
the hypothetical relationships between beverages and adiposity were investigated using

the anthropometric data available in NHANES III.

26

 

Skinfold thickness (mm) measurements were used to assess percentage body fat.
Various equations have been developed by regression analyses using skinfold thickness
to predict percentage body fat as measured by underwater weighing (Lohman, 1986). In
the current study, regression equations for adolescents developed by Slaughter et a1.
(1988) were used to predict percentage body fat from the summed triceps and
subscapular skinfold measurements. These equations are the best to use with adolescents,
because separate skinfold equations for predicting body fat in children age 8-18 years
were developed for prepubescent, pubescent, and postpubescent white and black male
adolescents and all female adolescents whose summed tricep and subscapular skinfolds
were 35 mm or less (Slaughter et al., 1988). In order to use the Slaughter et a1. (1988)
equations, males were classified by race and sexual maturation (SMA) stage.

Roche et a1. (1981) conducted a validation study among 6-12 and 13-17 year old
boys and girls (n=261) to determine which of the various anthropometric measurements
had the highest correlation with percentage body fat obtained by underwater weighing.
Measurements included triceps and subscapular skinfolds, weight, relative weight, BMI
(kg/m2), and the ponderal index (kg/m3). Compared to body weight measures, triceps
skinfold had the highest correlation with percentage body fat (from underwater weighing)
in all four groups, although the confidence interval of the correlation coefficient
overlapped substantially with that of subscapular skinfold, as well as that of BMI for girls
(Roche et al., 1981). The Slaughter et a1. (1988) equations were based on summed
triceps and subscapular skinfold measurements. These equations were cross-validated on
samples from the literature, and disagreement was found only with samples using the

Lange caliper, rather than the Harpenden skinfold caliper used to develop the equations

27

(Slaughter et al., 1988). Further cross-validation of the Slaughter skinfold equations for
children and adolescents (J anz et al., 1993) found very high intraclass (reliability)
correlations (ICC = 0.98-0.99), and high validity correlations (rs = 0.79-0.99) compared
to Lohman’s Siri age-adjusted body density equations (Lohman, 1986). Thus, the
Slaughter et al. (198 8) equations are unique in their appropriateness for youth of different

genders, races, stages of maturation as well as in their validity and reliability.

B. Beverage intakes of adolescents
1. Trends in beverage intakes

Trends in national surveys show that in recent years, youth drink much more soda
and less milk than in the past (Cavadini et al., 2000a; Bowman, 2002; Enns et al., 2002;
Enns et al., 2003; French et al., 2003; Nielsen & Popkin, 2004). Some authors have noted
that the trend in obesity prevalence is parallel to the trend in added sugars available in the
food supply (Hamack et al., 2000; Bray et al., 2004). If the added sugars in soft drinks
and other sugar-sweetened beverages consumed by youth contribute an excess of energy
intake compared to energy expenditure, the positive energy balance could lead to
incremental weight gain in adolescence. This section reviews the evidence that trends in
beverage intakes have increased the contribution of beverages to the total energy intake
of youth.

Milk consumption decreased by 37% in adolescent boys and 30% in adolescent
girls between 1977-1978 and 1994 (Cavadini et al., 2000a). In 1994-1996, for
adolescents aged 12-19 years, only 12% of girls and 30% of boys consumed the number

of dairy servings recommended by the Food Guide Pyramid (USDA, 2000). In 1977-

28

1978, children aged 6-11 years consumed 4 times more milk than any other beverage, and
adolescents aged 12-19 years drank 1.5 times more milk than any other beverage. These
data were compared to those collected 20 years later to identify the secular trends in
beverage consumption. In 1994-1996, 1998, children aged 6-11 consumed 1.5 times
more milk than soft drinks, and adolescents consumed twice as much soft drinks as milk
(French et al., 2003).

While total milk intake of adolescents has declined between 1977-1978 and 1994-
96 from 472 to 303 g/day (-36%) for boys, and from 303 to 189 g/day (-38%) for girls,
there has been a greater decrease in whole milk consumption. There was a 61% decrease
in whole milk intake (from 257 to 100 g/day) for boys and a 60% decrease (from 166 to
67 g/day) for girls. Lowfat/skim milk increased by 188% (from 105 to 197 g/day) for
boys, and by 183% (from 66 to 121 g/day) for girls, such that, compared to 1977-78, a
greater share of total milk consumed in 1994-96 was lowfat/skim rather than whole milk
(Enns et al., 2003). Thus, the contribution of milk to the total energy intake of youth has
declined because the fat content of milk consumed by youth has decreased, and total
consumption of milk has declined as well.

Soft drink consumption nearly tripled among adolescent boys from 7 to 22 ounces
per day between 1977-1978 and 1994 (Guthrie & Morton, 2000; French et al., 2003).
Soda was consumed by children as young as 7 months old (Fox et al., 2004). By 14 years
of age, 32% of adolescent girls and 52% of adolescent boys consumed 3 or more 8-ounce
servings of soda daily (Gleason & Suitor, 2001). Analyses of CSFH data have shown
that, from 1977-78 to 1994-96, 1998, energy intakes from sugars added to sugar-

sweetened beverages (soft drinks and fruit drinks) have nearly doubled from 70 Kcal/d

29

(3.9% of total energy) to 136 Kcal/d (6.9% of total energy) among those age 2 years and
older in the US. (Bray et al., 2004). Also using data from CSF H, 1994-96, 1998, added
sugars intakes from sugar-sweetened beverages was 52.6% of the total amount of added
sugars consumed by boys aged 12-17 years, and 48.5% for girls in the same age group
(Guthrie & Morton, 2000). Added sugars intakes from all foods and beverages were 20%
of total energy (Guthrie & Morton, 2000), an amount clearly in excess of the former Food
Guide Pyramid recommendations (6% of total energy at 1600 Kcal, 8.7% at 2200 Kcal,
and 10.1% at 2800 Kcal) (Johnson & Frary, 2001).

In NHANES IH, 1988-1994, beverages contributed 21.6% and 20.5% to the total
energy intake of adolescent boys and girls age 12-19 years. The shares of energy intake
contributed by beverage categories such as milk, sweetened beverages (soft drinks and
fruit drinks), 100% fruit juice, and other beverages (coffee, tea and alcoholic beverages)
were 6.9, 10.2, 1.9 and 2.5%, respectively, for boys, and 6.6, 10.6, 2.0, and 1.2% for girls
(Troiano et al., 2000). Using CSFII data to compare dietary intakes of adults age 18
years and older between 1989-1991 and 1994-1996, of all the food groups assessed,
whole milk and soft drink intakes changed the most during this brief time interval;
although reduced-fat milk intake increased by ~85 g/day, whole milk intake decreased by
~100 g/day and soft drink intake increased by ~90 g/day (Chanmugam et al., 2003).
Because declines in whole milk consumption were nearly offset by increases in reduced-
fat milk intake in adults, the authors asserted that the significant increase in soft drink
intake (p < 0.05) was responsible for most of the increase in energy intake from

beverages between 1989-1991 and 1994-1996 (Chanmugam et al., 2003).

30

For adolescents age 12—19 years, between 1989-91 and 1994-96, total milk intake
decreased by ~75 g/day for boys (-19%; 376 vs. 303 g/day), and girls reduced their milk
intake by ~50 g/day (-21%; 239 vs. 189 g/day), while there was no change in lowfat/skim
milk intakes (boys: 219 vs. 197 g/day; girls: 131 vs. 121 g/day) (Enns etal., 2003). Over
this 5-year period, there was a tremendous increase in fruit drink intakes (boys: 205 vs.
104 g/day; girls: 134 vs. 87 g/day) in addition to the increases in soft drink intakes (boys:
608 vs. 424 g/day; girls: 396 vs. 324 g/day), such that sweetened beverage intakes
increased 154% and 129% for boys and girls, respectively, between 1989-91 and 1994-96
(Enns et al., 2003). Because of these shifts between beverage categories, sugar-
sweetened beverages continue to contribute a large portion of the energy intakes of youth,
which justify the examination of trends in beverage consumption as a possible risk factor

for the concomitant increase in obesity.

2. Beverage intakes in relation to life style and dietary quality
Life-style behaviors, such as watching TV and fast food restaurant use, might be

associated with food and beverage choices that lead to excess energy intake (Coon &
Tucker, 2002; French et al., 2001b), TV watching might be associated with physical
inactivity as well (Forshee et al., 2004), and all might contribute to energy imbalance,
weight gain, and obesity in youth (Bray & Champagne, 2005). Soft drinks are frequently
consumed while snacking and watching TV (Coon et al., 2001; Giammattei et al., 2003;
Matheson et al., 2004; Phillips et al., 2004), and with meals eaten at fast food restaurants
(French et al., 2001b; Bowman et al., 2004). Youth who are characterized by beverage

patterns that include sugar-sweetened beverages might be more overweight than other

31

youth because of the energy intakes contributed by beverages (Hamack et al., 1999;
Troiano et al., 2000; Bowman, 2002), or because of the energy intakes contributed by
snacks and fast foods typically consumed with sugar-sweetened beverages (Hampl et al.,
2003; Utter et al., 2003; Bowman et al., 2004; Phillips etal., 2004).

Just as sugar-sweetened beverages might be associated with TV watching,
snacking and fast food restaurant use, beverages like milk or juice might be associated
with more healthful lifestyle habits such as frequent breakfast consumption and regular
exercise. Due to the need for hydration, drinking water or any beverage, even sugar-
sweetened beverages like Gatorade and other sports drinks might be associated with
exercise-induced sweating or sports team participation, two measurements of physical
activity used in this study. A beverage pattern classification characterized by milk and
juice consumption was expected to be specific to regular breakfast consumers, because
milk was poured on cereal and orange juice was often consumed at breakfast. Milk and
juice consumers were expected to consume more nutrient-dense diets than sweetened
beverage consumers, because of the vitamin and mineral intakes derived from milk, juice,
cereal and other foods consumed at the breakfast meal occasion. Milk is high in vitamins
and minerals like vitamin A, calcium and vitamin D (Miller et al., 2001); juice is high in
vitamin C, folate and potassium (Fellers et al., 1990); and cereal is fortified with many

vitamins and minerals (Subar et al., 1998).

C. Beverage intakes and adolescent obesity

In the current study, combinations of beverage categories were defined to

characterize the beverage consumption pattern consumed most often by youth. The

32

beverage categories included milk, 100% fruit juice, sugar-sweetened beverages (e. g.,
soft drinks and fruit drinks) and other beverages (e. g., low-calorie “diet” soft drinks,
coffee/tea, and alcoholic beverages). This section will review the known relationships
between obesity and each individual beverage category, from which hypotheses

regarding the beverage patterns were generated.

1. Milk

Researchers have hypothesized that dairy products and calcium may be inversely
associated with body weight and body fat (Zemel et al., 2000; Davies et al., 2000;
Heaney et al., 2002; Parikh & Yanovski, 2003; Teegarden, 2003). The proposed
mechanism involves the activation of vitamin D, affecting the intracellular calcium level
to regulate fat metabolism, such that low calcium diets promote lipogenesis, and high
calcium diets affect lipolysis in the adipocyte (Zemel et al., 2000; Zemel, 2001; Zemel,
2002; Zemel & Miller; 2004; Zemel, 20040). A review by Barr (2003) identified 26
studies in the literature that reported dairy product or calcium intakes and also assessed
body weight or body composition. These studies and others recently published are
summarized in Tables 1 & 2. Table 1 presents results of 19 observational studies, and
Table 2 summarizes 37 clinical studies. The purpose of this section is to evaluate
whether observational studies and clinical trials provide evidence for the proposed effect

of dairy product intake or calcium supplementation on body weight and body fatness in

children and adults.

33

a. Observational studies

As shown in Table 1, many case-control, cross-sectional and prospective studies
showed an inverse relationship between dairy or calcium intake and body weight,
although no association or even a positive association was observed in the largest
prospective cohort study. Of 19 observational studies in children and adults, 10 showed a
clear inverse relationship between dairy or calcium intake and body weight, while 4
studies showed an inverse association only in some subpopulations, no association was
observed in 4 studies, and a positive relationship was seen in 1 prospective cohort study.

Ten observational studies investigated the relationship between dairy or calcium
intake and body weight in children and adolescents. An inverse relationship with body
weight was observed in two case-control studies (Tanasescu et al., 2000; Lelovics &
Tamavolgyi, 2004) and in two cross-sectional studies (Tsakalou et al., 2004; Barba et al.,
2005). Two longitudinal studies showed an inverse relationship between dairy or
calcium intake and body fat (Carruth & Skinner , 2001; Skinner et al., 2003). One cross-
sectional study showed no relationship with body weight and an inverse relationship with
body fat (Novotrry et al., 2004), while another cross-sectional study and a prospective
study showed no relationship with either body weight or body fat (Venti et al., 2005;
Phillips et al., 2003). A positive relationship between body weight and frequency of milk
(white and chocolate) consumption was seen in a large prospective study (Berkey et al.,
2005)

Ten of the observational studies in Table 1 sampled adults, mostly women, and
assessed dairy/ calcium intakes as well as body weight and/or body composition. One

cross-sectional study reported no association with body weight (Dicker et al., 2004),

34

 

 

 

 

 

 

 

 

 

288 are 2 £8 a $88
0235 2me E .x. 8226 :3 sausage anus : w 56:5 we asim
amazon 288 as» x is 3 28¢
Dena: 388 E .x. can 2:: 5% 35393 fine .9: E 56.20 .m: 5:50
30:83 on
.58 mm are S
338 E .x. 0% £8 on :23 «as Goon
:23883 oz 435 J? Esme—«U a. Eém 3588* GENE .333 a. nee—EU .m.D 35>
AEomEv—m 56830
gauze 8i be $8 cacao baa Boos 8 53 econ 2.88
533083 02 5’ £5830 barn _m:ouoom-x Ammmuav .3 Ed £8330 mmaam >50>oz
3 as. $83
megs Em as: 0.: 38882 gamma 2 .: SEED 0820 Sense
Afiogmeoo e58 83m
.33 S :a £85... £83
saga Em we ea: 93 3888* if; ESE m .8 5620 be: spam
£868 E88 3980 $83
382.: .m> $25 5830 harm -680 Awnuev .S 3-3 68230 bumqsm 8323
205:8 .m> S 3-5
as: 59% 0% Base defiance See 5 883
882: :23 E30 £58 EEO a. Eém -ommu Ammuev 0:25 £2230 Barnum 383an
258% 30525 oémoaxm ©0502 awaken 36-: Booms—em wefiom Cb
Hemomcem 29305 .8542 “Ea

 

83:: 8338 he eouafismeoo 2:3 5m? 5 boon 8 Emmy? boon mo meozfioOmma Mo 3653 Eeouetomno A 298.

35

 

 

 

 

 

 

 

 

8-8 5 A883 an“ .85 $83
082: 82m .8 ma. 83 a. 08 8883* 3.38 mg .8 SA 88... 8: 8:82
80803 233 $28838
.808 8 0802: “8 ..\c
883 083 :8 8 928 SN .323
.88 88 8888 Gs 883 .983 $83
a 982: .88 8880 0.8 8888* 9&qu a: .083 $8 8: .m.: 83
€050?
a 08.8 .888 8.3 8:26 883 8: 28¢
8888 oz Em 8e 88 8-8 8888* 838 .88 «2: 38... 888 888
$8 8
80803 38803“ .3.“ 80803 mmm .808 Amoomv
a 0.82: .x. .88 .5, .828 8888* 85.8 m3 8 8-8 88¢. 88:0 .888
383 mm
288 €82 .083 8V :88 28$
0802: “a gem 88030 cock _m:ouoom-x EEHE 680808 .5803 .m.D 8.825
28 a 358.8 58 85
m8 5 288 88 .x: E a $83
2,88 Em < 8 .82 08 5 38885 83.28 8 3-8 .8880 .m.: 88m
36 3888 S N;
8 88 .x. < 8880 888 -88 a $ 88888 A83
8033088 oz .808& dam < .88.“ >85 0.: 258%on AwSHE 6890-88 630 .m.D mam—mam
:smom GofioEV 08momxm @0822 8885 08?: 8885 magnum arc
Emanczm .0805 853‘ 8E

.983 _ 88

36

 

Afiwa 32 S,

 

 

 

 

Rom < 886 Sam
.é Baas 3:25 IV $88
893 E q 8328 4 0% S 038385 sens 8:83 do .2300
Afiwm mad— 3,
Box cow-
588 86 0-3 ma? ow Em 2523 $83
85%: r5 < 8:830 < x3; ANV 038385 Awomucv 525 omega .m::v< nooNU «>9:an
8:022:
98% E30 immune
a mags g < .E < E o: » o3: Q83
”£25 Em azoam mace ban 9: 3389.05 @2.?& £85 a. 223 .333. .md 855
Em 30M «0 ESQ
32 E each: £0 48%
2me as: \6 ban 282 E s
gags E < .5 < 48MB 3% ”>83on smug » :2: 5:83 .m: 883 a:
gamed @0625 0:50qu @0502 :wmmua 36-: $8 Bum magnum Cb
Emoqvcm >53an 85=< “gm

.9an _ 2E

37

while another cross-sectional study showed no association with either body weight or
body fat (Venti et al., 2005). Two cross-sectional studies (J acqmain et al., 2003; Loos et
al., 2004) and two prospective studies (Lin et al., 2000; Pereira et al., 2002) reported an
inverse relationship in some subpopulations. Body fat was inversely associated with
calcium in one cross-sectional study (Lovejoy et al., 2001), while body weight was
inversely associated with either dairy or calcium intakes in another cross-sectional study
(Mirmiran et al., 2005), and in two prospective studies (Kabrnova et al., 2004; Ochner et

aL,2004)

b. Clinical trials

As shown in Table 2, although a few weight loss studies involving energy
restriction showed an inverse relationship between dairy/ calcium and body weight/ fat,
most clinical trials showed that neither dairy products nor calcium supplements lowered
body weight or adiposity. Table 2 includes 37 clinical trials of dairy products (n=14),
calcium supplements (n=19), or both (n=4). Indicators of bone health were the primary
endpoints of most of these trials, although weight, BMI and/or body composition was
also assessed. Of 37 randomized trials assessing the effect of dairy products or calcium
supplementation on body weight, 31 showed no effect, only 4 showed weight loss, and
weight gain was the result in 2 studies.

Of the 27 trials of dairy or calcium supplementation in which the treatment did
not involve energy restriction, most studies (n=24) showed no difference in body weight
or composition. An inverse association between calcium and body weight was seen in

only one of these studies; weight loss in the supplemented group was 0.76 vs. 0.32 kg/yr

38

 

 

hmA

 

 

 

 

 

 

 

 

 

m Z $28 52 E < omzmaoaogmoa COGS

e was 3 .E 2 4 .5322 2:2 22.. cm 225 oz Sm .2-~ 23TH; 2%-? .5225 220 =3

32: 2 5

52 SEE am: 883

e E 2 < 22. .232 22%... N2 223 02 23-2 sans ..2 22-22 .323 .m: Em
2.2282 .x. 808 2% 2 82226 2238522822 223:

m2 .95 .E a < .222 258 cm? .228 oz 52 2-~ 82".; 2 8A .8203 .md :52
2223 6 2» 2 2226 2SA 33982 2 68:

m2 Em 2 < .233 222 no Rm 2 .223 oz .82 2-~ sens -2822 .5283 .823. 2259
.2 9.02 883

222 E 2 < 6 Ewe 22+ .225 oz Ba 2-2 5“,; .5803 .md 52m
2.22 2.2 9.83

m2 .3 .282 .2 .E 2 2. 2.o 22: 28+ .225 oz 82 .2-N 82qu .2 a .25 .md 0.23
89.2 5.2 2.52% 88$

m2 .5 22.8 .2 .E 2 < 8 Ewe 82 223 02 Hum .2-~ Sana .2 2-2 .230 302 82.502
2.22 5.2 Sm 2322:

m2 .3 222 .2 .E 2 < 2:2 8:3 use 82 .223 oz 08-2: smug .2 32 .230 .32.: 52220
2.2.2 52 228:

m2 .3 282 .2 .E 2 2. .so Ema SS .223 oz Sm 2._ male .2 : .220 2.: 220
:30M Emoqum omen 0.5.0.0me 1.6M :wEoQ 36-: 38.326 wcfiom A890

8.58% 8822. EE

 

2038582229222 8323

8 28:92:88 .666 @8335 Eat Ham been 8 Emmy? been so Bebe mo $2.5 32:8 2328028122 .N. 038.

39

 

 

 

 

 

 

 

 

 

 

 

3.830808 $.33

mz 522 .2 .2, 2 q 8 22 N .8 22.. 2 2228 oz 52 2-2 2:52 -52 .5225 25:22 .522
.2 3-2 285:

mz 222 .22 2 2. 282 225 8 22 2 2228 oz 82 2-2 22 21.2 .225 25252 2.: 55222.

2222.. ow

.2255 25 .2222 22 .2 5. : 88$

22 52225 .2 .2, 2 < 252 2 808 m 2 2228 oz .82 .22 82”.: -22 .5220 2550 228
2.282» 2202 82 .2 2-2: 22 233

m2 5 .25 8 4 2920020 Em 2 88220 02 08-w_ A8 Nvué .2825 228mm .882. 88.262
82.88 2:8 68: €33

.22 2 .2, 2 2. 222 8 222 2.22 2228 oz 82 .22 22.52 2 as .28 .5232 25.2522
82 252 2223

22 2 .2, 2 < 288 222 82 2228 oz 02-22 5.qu 2 5 .5220 252 52
52 5.8:

m2 2 .E 2 q 808 222 82 2228 oz 02-2: 2855 2 5 .5228 228 52
2 222 2. 22.2 82 228:

.22 .2 .2, 2 < 225 8 222 82 2228 oz 02-22 €82 .2 0.: .220 2.: 222-2
8 2» v. 2-2.2 222 8 2222 22v 228$

.22 252 22 .25 2 < .8 22 3.3 H52 228 oz 82 2-2 222.12 .32: .5225 .25 5226
2:8 538280 mmA 8588808 2: 28$

22 E 2 < 52228 23.2 m 228 oz 55 .2-~ 22 52 -58 .5225 .2222. 2.2220
2832 888:5 omen 98.0-0me Rom 888G 0822-8 38 Sam mafiom €83

5.252 .823. 8.2.2

.5282 m 228

40

 

 

Egan

 

 

 

 

 

 

 

 

5m 55:58; 3A cfiaow 383
m2 E a < 8 3w _ 8328 oz 87% Ezus s SA .5803 302 20m
Amy—38830: m3
$5882 8:
5m 833%3 #83
m2 E 3.8 .E a < 606 3» ~ 8326 oz 83 emmus .8803 .m5 383
a $2: 52 E < Bmsmmonoe 38:
mz 32: E .E a < 33% 6 $8 82 8305 oz 5m 3 sans -qu .8503 .m.: mmwa
8m ESE a. :3:
m2 E a < 608 E» _ 8:65 oz 85 sans wages 5:83 .md ”533
A882, Sm $3:
a £553 22 .58 on s 8&3:
m2 5 $8 .x. .E a < 2: + 8 Rue com 8328 oz 5m 3. Gtua E SM .33? .md $838
Hammozogmoa
.32 88 38:
m2 E a < 8 3m _ 8228 oz 5m I song 292 E 5803 saw $5-33
5m aaflofifimom 63:
a E a < 606 Ewe 82 8328 oz T27 sans » 8N 5:53 .md 38m
Amoafiv
mz E a < 8 fig 87 8228 oz Ex I Song 3502 .8535 «3.50 8:85
«38% “Bantam 80G oSmoaxm 30M @609 36-: 38 Sam wfitom A820

858%

853 ES

.9283 N. 2an

41

 

 

 

 

 

 

 

 

 

6 $8 com 8828 52 828 3253
a 88 88 .E 8 < 6 8&8 82-82 882 8% Eé fins .5803 a 822 .m: 88%
6 888 8: .52 880 383
m2 “8 285 .E 8 < .9 8 8&8 gm ban 88 2a-? Sons .5803 a 822 .md 888,88
6 38 8:
a a 285 8 < .2, no Ewe 282 8:880 82 82o 8 8:26 8888
m2 E 8 4 w: m 882 8% 23-2 sons “58E 2. 822 .md 88%
52 828 8 E88 €83
m2 88 83 .E 8 < 8 Rm 2 8826 8% 23-2 82n5 .8803 .md 888%
82:8 93:8 HUM 0890 Mo Ego a: Avoomv
m Z as.“ 302 £5 E < 8 Eu Bob E805 EEO mo> 23-~_ Smnav .3895 a. 52 ABE? 530m
Accomv
emoao .5 Eo>o Ogom
m2 88 88 .E 8 q 88 I” .2, ma 7 ES 8; 52 Eva .5803 a :22 .md -8522
93:3
. 88 82 88588 E 288 28¢
m2 3? E < wmg t5 N..v + NO aim fl Emu—NU mo> OEum ANWHGV amen—o £0503 ABC Gomfioh
N882 $28
$5: Swo— .mmmE HUM .Rmsmaocogmon 33:
m2 “8 dam .E 8 < 2288 0888 80 2m _ 8:880 8.» 08-8 2 mug .8825 .m.D 82m
258m “£82m omen osmomxm 30M :wmmon 88-: $8 Bum wafiom €33
8.58.2 883~ 882

.9802 N. 288

42

 

 

 

52.6%
8m §E< .830 383
6 «a 309 J? E < w>m m .m> w>m Hv Eman— mo> VELVN Ammuav .5803 um :02 .m.D BEoN
5m 80% 388
a E 38 .3, 5 < 5o 38 8: Em; mew V3-2 sauna .582» a. :22 .md 38$
“38% “angm omoQ oSmoaxm 30M :wmmofl onmmé floofsm wcfiom A393

858m

85.2 ES

.66on m 033

43

in the control group (Recker et al., 1996). Among the 11 trials of dairy products without
energy restriction, one study found an increase in body weight (Barr et al., 2000), and
another found an increase in both body weight and body fat (Lau et al., 2001). Neither
weight nor fat were affected by either dairy or calcium supplementation in 11 of 12 trials
that assessed changes in both weight and body composition (Chan et al., 1995; Cadogan
et al., 1997; Merrilees et al., 2000; Lloyd etal., 1993; Dibba et al., 2000; Dawson-Hughes
et al., 1997; Riggs et a1, 1998; Wosje & Kalkwarf, 2004). None of the 11 trials
conducted in children or adolescents found an effect of dairy (Chan et al., 1995; Cadogan
et al., 1997; Merrilees et al., 2000; Lappe etal., 2004) or calcium supplementation ‘
(Johnston et a1, 1992; Lloyd et al., 1993; Lee et al., 1994; Lee et al., 1995; Bonjour et al.,
1997; Nowson et al., 1997; Dibba et al., 2000) on body weight or body fat.

Ten clinical trials in adults have tested the effect on weight loss of a reduced-
calorie diet that included dairy products or calcium supplements. Three used calcium
supplements (Ricci et al., 1998; Jensen et al., 2001; Shapses et al., 2004), while five
studied the effects of dairy products (Harvey-Berino et al., 2004; Bowen et al., 2004;
Thompson et al., 2005; Zemel et al., 2005a; Zemel et al., 2005b), or dairy products and
supplements (Zemel et al., 2004a; Zemel et al., 2004b). Only three of the 10 trials found
that dairy calcium had a significant association with weight loss (Zemel et al., 2004b;
Zemel et al., 2005a; Zemel eta1., 2005b). Body composition was also assessed in all
except one of these studies (Jensen et al., 2001). Four of these reported significantly
greater fat loss between the high dairy consuming groups and low dairy groups in calorie-

restn'cted overweight or obese adults (Zemel et al., 2004a; Zemel et al., 2004b; Zemel et

al., 2005a; Zemel et al., 2005b). The remaining studies found no difference between high
and low-dairy or calcium supplemented groups in loss of body weight or body fat.

To summarize these findings, the majority of clinical trials have not supported an
inverse relationship between body weight and the calcium in dairy products, in contrast to
the observations of many cross-sectional and prospective studies, but consistent with a
possible positive relationship seen in one large prospective cohort study. Except for
those studies conducted by Zemel and colleagues, the evidence accumulated from
randomized clinical trials indicates that neither dairy products nor calcium supplements
reliably facilitate weight loss or fat loss. Two of the 37 clinical trials demonstrated an
increase in body weight with a dairy treatment (Barr et al., 2000; Lau et al., 2001).
Without energy restriction, supplementing the diet with dairy products may actually lead
to weight gain rather than weight loss, because of the additional calories consumed by the
treatment group.

In observational studies, it was difficult to differentiate influences of correlated
lifestyle factors such as exercise level, soda consumption, total fat and fiber intake
(Novotny et al., 2004). These lifestyle factors, and not the dairy consumption, could be
responsible for the variance in body weight (Huang & McCrory, 2005). In the current
study, beverage patterns that include both soda and milk were distinguished from those
that include soda without milk, or milk without soda, to determine whether milk alone
has any relationship with body weight or body fat after adjusting for physical activity and
other lifestyle factors.

The inverse relationship between milk and soda consumption in cross-sectional

and prospective studies could explain the divergence between findings from

45

observational studies and clinical trials of dairy and calcium supplementation. In
conclusion, the hypothesis that dairy or calcium is inversely related to body weight and
body fatness in children and adults is not supported by evidence from the current

literature review.

2. Fruit Juice

Before the development of technology for the easy extraction and processing of
juices a few decades ago, only water would be offered when children were thirsty, but
fruit juice is now a common staple in children’s diets today (Anonymous, 1995). Most
children enjoy the taste of sweet beverages, and many parents consider 100% fruit juice
to be healthful. A potential problem exists in that fruit juice can be easily over
consumed, suggesting that excess calorie intake from fi'uit juice consumption might lead
to childhood obesity. Obesity in adolescence might result fi'om the persistence of
childhood obesity. In addition, preferences for sugar-sweetened beverages, such as fruit
drinks and carbonated soft drinks, may persist into adolescence from childhood habits
formed when 100% fruit juice rather than water is consistently offered to quench thirst.

Increased intakes of fruit juice consumption (Dennison, 1996) might relate to the
high prevalence of childhood obesity, which was 10.4% in 1999-2000 among children
aged 2-5 years (Ogden et al., 2002). High intakes of fruit juice (212 oz/day) among
preschool children have been associated with overweight in one cross-sectional study
(Dennison et al., 1997) but not in longitudinal studies (Skinner et al., 1999; Alexy et al.,
1999). A recent longitudinal study found that neither fi'uit juice, nor any other beverage

' (including fruit drinks, milk, regular or diet soda) were associated with BMI change

46

among low income pre-school WIC-participating children aged 2-5 years living in North
Dakota (N ewby et al., 2004). A small cross-sectional study (11 = 77) among preschool
children participating in the Atlanta W1C program also found no association between
overweight and fruit juice consumption (Kloeblen-Tarver, 2001). The original research
that asserted that consumption of 100% fruit juice was associated with both short stature
and childhood obesity (Dennison et al., 1997) was controversial, because parents and
nutritionists viewed 100% fruit juice to be more nutritious than other sweetened
beverages (Doucette & Dwyer, 2000; Rampersaud et al., 2003). It is important to resolve
the fruit juice controversy, to understand whether fruit juice is associated with childhood
and adolescent obesity, given the extent to which obesity in adolescence results from the
persistence of childhood obesity, or from gradual weight gain over time.

In the controversial study, intakes of 100% fruit juice assessed using seven-day
food records were associated with measured height and weight in 168 children age 2 and
5 years attending a well child care visit to physicians practicing in rural upstate New
York (Dennison et al., 1997). Nineteen of the children (11%) were consuming at least 12
fluid ounces of fruit juice per day. Short stature (defined as height less than the 20th sex-
specific percentile for age) was significantly more common among these heavy juice
drinkers than among those drinking less juice (42% vs. 14%). Obesity was also more
common among those drinking 212 oz/day juice; 32% had a BMI 290‘h age- and sex-
specific percentile, as compared to 9% of those drinking less juice. Dennison et a1.
(1997) concluded that excess fruit juice was associated with increased risks of both short
stature and obesity in preschool children. In some children, excess fruit juice may have

displaced other, more nutrient-dense foods from the diet, leading to a slowing of growth.

47

In other children, fruit juice may have been consumed in addition to an already-adequate
diet, leading to excessive caloric intake.

Impaired growth due to excessive fruit juice consumption had been previously
implied in a report of eight cases of nonorganic failure to thrive in l-2-year-olds (Smith
& Lifshitz, 1994). Impaired growth could result from the displacement of nutrient-dense
foods, as well as from diarrhea and other gastrointestinal symptoms of carbohydrate
malabsorption, particularly when fruit juices high in both sorbitol (which is
nonabsorbable) and fructose (which may be incompletely absorbed), such as apple juice,
are overconsumed (Lifshitz et al., 1992; Doucette & Dwyer, 2000). Dennison et a1.
(1999) conducted a follow-up study to determine whether the effect of fruit juice
consumption depended on the type of fruit juice consumed. Orange juice was absolved
in this study, as the previously observed association with short stature was specific only
to the consumption of apple juice or grape juice, presumably due to malabsorption of
carbohydrate, and the association with obesity was observed with apple juice intakes only
(Dennison et al., 1999).

The report of excessive fi'uit juice consumption has remained controversial, and
one correspondent wrote, “How can something that causes failure to thrive be associated
with obesity?” (Levine, 1997). Letters to the editor of Pediatrics raised several questions
about the design and interpretation of the Dennison et a1. (1997) study (Dhurandhar,
1997; Kennedy et al., 1997; Stinson, 1997). One criticism of the study was that the 75th
percentile was used as a cut point to define obesity, rather than the 95th or 85th percentiles
used to define “overweight” or “at risk for overweight” today (Kuczmarski et al., 2000),

and mother’s heights were taken into account in the analyses while fathers’ heights were

48

not. The duration of fruit juice intake by the 2-year—olds may have been too brief to
produce detectable effects on height and weight, and the number of children who
consumed _>.12 oz/day fruit juice was small. Most importantly, the cross-sectional design
of the study did not allow for causal inferences, and it was possible that “overweight”
children were drinking more fruit juice while reducing intakes of fruit drinks and soda in
response to their weight, an example of reverse causation. In response, the authors stated
that data had been collected to determine optimal methods of assess children’s dietary
intake, and the relationship between fruit juice intake and grth parameters was
examined only after it was noted that many children seemed to drink excessive amounts
of juice. Thus, some data that might have been relevant (such as father’s heights) were
not collected. The authors agreed that cause-and-effect relationships had not been
established, and recommended that longitudinal studies of fruit juice intake and child
growth should be conducted.

A longitudinal study of 105 children aged 2-3 years found no relation between
daily fruit juice intake and height, BMI or ponderal index (calculated as kg/m3), nor did
they find that fruit juice consumption displaced milk from the diet (Skinner et al., 1999).
A similar study by the same authors followed children as they grew from age 2 to 6 years
(Skinner & Carruth, 2001). This second study also showed no relationship between fruit
juice and growth, but fi'uit juice intake decreased while soda intake increased, as children
grew older (Skinner & Can'uth, 2001). In another longitudinal study of 205 German
children age 305 years, each with four or more repeated measurements, none consuming
12 oz or more hit juice per day were overweight, and BMI and height z-scores were not

related to excessive juice consumption (Alexy et al., 1999).

49

Although negative results from longitudinal studies have not supported the
original observation, the American Academy of Pediatrics still issued a statement
regarding fruit juice consumption (AAP, 2001) in response to the assertions that
excessive consumption of 100% fruit juice is associated with failure to thrive (Smith &
Lifshitz, 1994) and short stature as well as obesity (Dennison et al., 1997) in children.
The AAP advised against introducing fruit juice into the diets of infants less than 6
months old, and recommended limiting 100% fruit juice consumption to 4-6 oz/day for
children 1-6 years, and 8-12 oz/day for children 7-18 years (AAP, 2001). Analyses of
CSFII 1994-96, 1998 data found that mean daily intakes of 100% fruit juice, 4.6 and 3.4
oz for children 6 months-6 years and 7-18 years, respectively were within the
recommended amounts (Rampersaud et al., 2003). Intakes of <6 oz/day were consumed
by 73% of children 6 months-6 years, and <12 oz/day were consumed by 91% and 94%
of those aged 6 months-6 years and 7-18 years, respectively (Rampersaud et al., 2003).
Intakes of sweetened beverages (fruit drinks and carbonated soft drinks) increased, and
milk intakes decreased with age such that at age 5 years and older, mean intakes of either
of the two sweetened beverages exceeded 100% fruit juice intake, and at age 13, children
were drinking more carbonated soft drinks than milk as well (Rampersaud et al., 2003).

Orange juice can be an important source of nutrients in the diets of pre-school
children. In a representative sample of 329 Native American and non-Hispanic white
children aged 1-6 years living in rural Oklahoma, orange juice contributed 46.2% of total
vitamin C intake (Stroehla et al., 2005). Orange juice was the first-ranked dietary source
of vitamin C, with the second-ranked dietary source, sweetened fi'uit drinks, contributing

' 8.7% and oranges, the only solid fruit contributing more than 2%, ranked fifth at 4.2% of

50

vitamin C intake. Orange juice contributed 15.9% of folate, second only to cold cereals
at 17.4%, and orange juice contributed 3.7% of carbohydrate intake. Orange juice was
ranked tenth as a source of carbohydrate, and other beverages that ranked higher included
nondiet soft drinks (rank 1, 8.8%), sweetened fruit drinks (rank 2, 7.7%), whole milk
(rank 4, 5.8%), 1% or 2% milk (rank 5, 4.7%) and 100% apple/ grape juice (rank 8,
4.1%). Although orange juice contributed more vitamin C and folate to the diets of these
children, other beverages frequently consumed contributed more to energy and
carbohydrate intakes (Stroehla et al., 2005).

In a recent prospective study, the association of beverage consumption with
annual weight change was studied using a semiquantitative FF Q in 1,345 children aged 2-
5 years who visited a North Dakota clinic twice within 6-12 months time (Newby et al.,
2004). Because earlier studies had classified consumption of more than 12 oz/day as
high fi'uit juice intake (Dennison et al., 1997; Alexy et al., 1999), fruit juice intakes and
intakes of other beverages were dichotomized in a similar manner, as well as being
modeled as continuous variables. In this sample of children who visited the W1C clinic
between January 1, 1995 and June 30, 1998, about 45% of children consumed 12 oz or
more fi'uit juice per day. Analyses of the association between beverage consumption and
weight change were adjusted for sex, change in height, age and total energy intake at
baseline, birth weight, maternal years of education, race/ethnicity, residence and poverty
level. Energy intake was then excluded because it can be argued that it should not be in
the model if total energy is the mediating variable in the pathway between the exposure
(beverage consumption) and the outcome (changes in weight or BMI). All beverages

were significantly associated with total energy, but there were no significant relations

51

“1-

between weight change and intakes of fruit juice, fruit drinks, milk, regular soda or diet
soda. There was no association between weight change and beverage consumption,
regardless of whether all beverages were considered individually or included in the same
model, and results did not change when energy intake was omitted from the model
(Newby et al., 2004).

Although the majority of studies have shown that neither 100% fruit juice nor
sweetened beverages was associated with obesity in preschool children, some studies
have indicated that sweetened beverage consumption was associated with obesity in
adolescence. Preferences for consumption of sweetened beverages, such as fruit drinks
and carbonated soft drinks, may persist into adolescence due to habits formed in
childhood when 100% fruit juice rather than water is consistently offered to quench thirst.

According to Food and Drug Administration (FDA) regulations, only 100% fruit
juice can be labeled as a fruit juice, and any product containing less than 100% fruit juice
must be labeled with a descriptive term, such as “drink,” “beverage,” or “cocktail” (AAP,
2001). Like carbonated soft drinks, fi'uit drinks often contain added sweeteners, and they
are nutritionally distinct from 100% fruit juice. For example, orange juice is a good
source of folate and potassium, while FDA does not allow fortification of fiuit drinks
with either of these two nutrients (Rampersaud etal., 2003). Nevertheless, there is a
common misperception that fruit drinks are the same as fruit juice, especially among
adolescents.

A concern is that adolescents may believe that they are regularly consuming a
healthful beverage such as orange juice for breakfast, when in actuality they are

' consuming fruit drinks, such as Hi-C and Sunny Delight. This is relevant to the current

52

 

study, because adolescents may have misconstrued a question asked in a FFQ, “How
often do you drink fi'uit juice, such as apple juice or orange juice?” The sequence of the
questions in the F FQ was such that the fi'uit juice question was asked before the one
about fi'uit drink consumption. Thus, the respondent would not realize the mistake until
they are later asked, “How often do you drink beverages such as Hi-C, Tang or Sunny
Delight?” The result of the misperception and misclassification of the exposure is that
fruit juice might be found to be associated with adolescent obesity in the current research,
because fi'uit drinks contain added sweeteners and are often consumed heavily by
adolescents. Thus, the “fruit juice controversy” may not be resolved by the current study

if fruit juices are mistaken for fruit drinks often consumed by teens.

3. Sugar-sweetened beverages

Sugar—sweetened beverages include soft drinks and fruit drinks that contain added
sugars. These beverages are frequently consumed, and contribute a large portion of the
total energy intake of youth. For those youth having an excess of energy intake from
sugar-sweetened beverages compared to energy expenditure, the positive energy balance
could lead to incremental weight gain in adolescence. This section reviews prospective
and cross-sectional studies that have investigated the relationship between sugar-
sweetened beverage intakes and body weight in youth.

Several prospective studies (Ludwig et al., 2001; Mrdjenovic & Levitsky, 2003;
Berkey et al., 2004) and intervention studies (James et al., 2004; Ebbeling et al., 2006)
have investigated the relationship between soft drink intakes and weight gain in youth. In

a 19-month study of children aged 11-12 years (n=548) in Boston-area middle schools,

53

each additional 12-oz serving/day of soft drink was associated with an increase in BMI of
0.18 kg/mz (p=0.03) and a 60% increased obesity risk (p=0.02) (Ludwig et al, 2001).
Mrdjenovic & Levitsky (2003) measured the body weight of children aged 6-13 years
(n=21) at the beginning and end of their 4-8 week stay at a day camp, and found that
those who consumed the highest level of sugar-sweetened beverages (>16 oz/day) gained
1.12i0.7 kg, while those who consumed 6-12 oz/d or 12-16 oz/d sugar-sweetened
beverages gained 0.32-O.48i0.4 kg (differences between the latter two groups were not
statistically significant). In a large 2-year study, offspring of Nurse’s Health Study 11
participants aged 9-14 years (n= 5,321 boys; n=6,871 girls), each additional serving/day
of sugar-sweetened beverages was associated with year-to-year increases in BMI of 0.04
kg/m2 (p=0.01) for boys, and 0.03 kg/m2 (p=0.08) for girls (Berkey etal., 2004).

In a 1-year intervention study of children aged 7-11 years (n=644) in six primary
schools in southwest England, soft drink consumption and the prevalence of overweight
and obesity decreased by 0.6 glasses (150 mL) per day and 0.2% in the treatment group,
while increased soft drink consumption of 0.2 glasses (50 mL) per day was associated
with a 7.5% increased prevalence of overweight and obesity in the control group (James
et al., 2004). A 25-week study of 103 adolescents aged 13-18 years was to provide
noncaloric beverages by home delivery to decrease consumption of sugar-sweetened
beverages in the intervention group. The intervention compared to the control resulted in
a net effect of reducing BMI by —O.75 kg/m2 among the subjects in the upper baseline-
BMI tertile, but the net change in BMI was not statistically significant overall.
Prospective and intervention studies have shown a direct albeit weak association between

'SOft drink intakes and weight gain in youth. However, all except the most recent study

54

1

(Ebbeling et al., 2006) were limited to children aged 14 years and younger, and results
cannot be generalized beyond the respective study populations.

Several cross-sectional studies used the national survey databases and reported
soft drink intake in relation to BMI or overweight status (Troiano et al., 2000; Forshee &
Storey, 2003; Forshee et al., 2004; Boumtje et al., 2005). Descriptive data from a
bivariate analysis of NHANES IH, 1988-94, data, with no adjustments for covariates,
suggested that soft drinks contributed a higher proportion of energy intake for overweight
than nonoverweight adolescents aged 12-19 years (males: 10.3% vs. 7.6%; females: 8.6%
vs. 7.9%). Standard errors for the descriptive statistics or tests for significant differences
were not reported (Troiano et al., 2000). Further analyses of NHANES III data showed
no correlation of soft drinks or fruit drinks with BMI for adolescents aged 12-16 years
(F orshee & Storey, 2003; F orshee et al., 2004). The same researchers also found no clear
relationship between BMI and sugar-sweetened beverage consumption using CSF H,
1994-98, data (Forshee & Storey, 2003). Others using CSFII data in logistic regression
models found no relationship between soft drinks and obesity in adolescents aged 12-18
years, but soft drinks were inversely associated with normal weight and positively
associated with obesity in children aged 5-11 years (Boumtje et al., 2004).

Using both NHANES and CSFH data, Forshee and Storey (2003) controlled for
age, gender and ethnicity and found no relationship between beverage intakes and BMI,
but rather than predicting BMI from beverage intakes, BMI was entered along with the
covariates as an independent variable, and intakes of each beverage category (fluid milk,
regular soft drinks, fruit drinks, 100% fruit juice; g/d assessed using 24—hour recall) were

the dependent variables in separate regression models. In a second study by the same

55

authors, NHANES III data were used, and BMI was the dependent variable in a linear
regression model, while intakes of the various beverage categories (regular soft drinks,
diet soft drinks, fruit drinks, coffee, tea, beer, wine) were included simultaneously in the
model (Forshee et al., 2004). Beverage intakes (times/month) were assessed using FFQ
in one model, and beverage intakes (g/d) were assessed using 24-hour recall in another
model, and both the F P Q and 24-hour recall models also included energy from sources
other than beverages (Kcal/day, assessed using the 24-hour recall). Models were adjusted
using the usual covariates such as gender, age, race—ethnicity, region and household
income (% of poverty threshold), and physical activity. Rather than using the exercise
variable, only sports team participation was used to assess physical activity. In addition,
TV watching (hours per day) was used as a measure of sedentary behavior.

In another study, intake variables simultaneously included in the model were soft
drinks, whole milk, low fat milk, other dairy, grains, vegetables, fruits, meat, fats and
oils, legumes, dietary fiber, cholesterol, sodium, with amounts of food and nutrients
assessed using 24-hour recall and averaged over 2 days (Boumtje et al., 2005). Analyses
were also adjusted for income, race, Hispanic origin, age, gender, region, participation in
food stamps or and school lunch programs, exercise and TV watching. Rather than using
BMI as the dependent variable in a linear regression model, logistic regression was used
to estimate the risk for overweight and obesity based on gender- and age-specific BMI
percentile criteria (Boumtje et al., 2005).

The current NHANES HI study differs from the research previously conducted
using the national survey data sets. Because of the potential for multicollinearity among

covariates in the model, in the current study, intakes of the various beverage categories

56

were not included simultaneously in the model. Instead, the beverage pattern
classification was entered into the model as a dummy variable to assess the independent
effects of consuming different combinations of beverage categories. Usual beverage
patterns were assessed using only the F FQ, and energy intake assessed using the 24-hour
recall was not included in the model. Because the hypothesized relationship between
beverage patterns and obesity is mediated through energy intake, the model would be
“overcontrolled” if energy intake were included as a covariate. Instead, a separate
investigation contrasted mean energy intakes of groups of youth characterized by the
various beverage patterns. If the associations between beverage patterns and BMI were
reported, this would have been consistent with previous research, but, because BMI is not
adjusted for age and gender, indicators of overweight and adiposity included the BMI
percentile and percentage body fat.

It has been hypothesized that intakes of sugar-sweetened beverages are correlated
with adolescent obesity, not only because these beverages are frequently consumed and
contribute a large portion of the total energy intake of youth, but also because sugar
consumed as a liquid seems to promote obesity more consistently than sugar consumed in
the form of solid foods. It has been hypothesized that liquids, rather than solids, are less
likely to trigger satiety mechanisms, and this hypothesis was explored in the “j ellybean”
study (DiMeglio & Mattes, 2000). This was a crossover trial in which subjects were fed
isoenergetic carbohydrate loads as either a soda (a liquid load) or jelly beans (a solid
load) during two 4-wk periods separated by a 4-wk washout. Subjects compensated for
the 450 calories contained in the jellybeans by consuming roughly 450 fewer calories

form other food, but they failed to compensate for the soft drinks, essentially adding 450

57

calories to their previous diet. Significant increases in both body weight and BMI
occurred only during the liquid load period (DiMegglio & Mattes, 2000). These results
and other studies that compared regular soft drinks to low-calorie “diet” soft drinks
(discussed in the next section) suggest that calories in sugar-sweetened beverages “add

on” energy intake rather than displacing other foods in the diet.

4. Low-calorie “diet” soft drinks

In recent decades, intense sweeteners, such as aspartame, have been marketed as
sugar substitutes used to prevent weight gain, but remarkably few long-term studies have
examined the efficacy of aspartame in dietary approaches to weight management.
Consuming foods and beverages containing artificial sweeteners might have a minimal
effect on energy intake and body weight, because dieters might be compensating for the
energy deficits by consuming other calorie-containing foods (Rolls, 1989). Another
concern is that the perception of sweetness even without calories might actually stimulate
the appetite. This section will review clinical trials that investigated the efficacy of
aspartame in weight management. Short-tenn clinical trials that compared sugar-
sweetened beverages with low-calorie “diet” soft drinks are emphasized in this review.
The use of artificial sweeteners in beverages might lower the energy intake that would
occur when calories are provided by sugar-sweetened beverages.

Raben et a1. (2002) reported that subjects whose diets were supplemented with
sucrose (~152 g sucrose/day, 70% from soft drinks) for 10 weeks had significantly higher
energy intakes, body weight, fat mass, and blood pressure; decreases or no changes in

these variables occurred when the subjects diets were supplemented with foods and

58

beverages containing artificially sweetened supplements (0 g sucrose/day). The authors
noted that energy obtained from sweetened beverages appeared to be less satiating than
energy from solid foods (Raben et al., 2002), and others have found that it might be
easier to over-consume energy when drinking liquids than when eating solids (DiMeglio
& Mattes, 2000).

De Castro (1993) investigated the effects of various beverages and foods on
energy displacement and reported that subjects (n = 323) who drank sugar-free soda
consumed significantly less carbohydrate throughout the day than individuals who did
not drink diet soda. Meanwhile, those who drank sugary sodas consumed more
carbohydrate and energy on days when they drank soda than on days when they did not.
After subtracting the energy content of beverages from total energy intakes, energy
intakes from all other foods and beverages on days when soda was consumed did not
differ significantly from total energy intake on days when soda was not consumed. These
data indicate that energy derived from beverages lead to elevated energy intakes and does
not displace energy ingested from other foods (i.e., individuals do not compensate by
reducing food intake).

An investigation on lunchtime beverages consumed by pre-school children (n =
134) found that chocolate milk sweetened with sucrose increased the children’s
lunchtime energy significantly (~65 extra kcal compared with plain milk and chocolate
milk sweetened with aspartame (p <0.05). There was no compensation for energy when
the children selected their afternoon snack 3 hours later (Wilson, 1999). A review of the
literature suggests that caloric compensation might occur only if a test meal is provided

within a short time (0-30 min) following the liquid load ingestion (Almiron-Roig et al.,

59

2003). Timing of the occasion when beverages are consumed in relation to other meals is
thus important. CSFII data show that beverages, especially soft drinks, are frequently
consumed at a mid-meal snack or beverage break (Hampl et al., 2003), and energy
consumed at the next meal would be unaffected if that meal did not take place until hours
later. If there is little energy intake compensation for the calories consumed when
drinking sugar-sweetened beverages, then artificially sweetened beverages might be
efficacious alternatives to prevent the weight gain that might occur when sugar-

sweetened beverages are consumed.

Tordoff & Alleva (1990) conducted a crossover trial in which subjects (21 men
and 9 women) consumed 4 bottles/day (1135 g) of soda sweetened with either high-
fructose corn syrup (I-IFCS) or aspartame, during each of two 3-week periods. Compared
to the 3-week control period in which no soda was given, drinking aspartame-sweetened
beverages decreased body weight of men by 0.47 kg (p < 0.05) but not women. While
drinking sugar-sweetened beverages women gained weight significantly (0.97 kg, p <
0.0 1) and men gained slightly (0.52 kg, NS). These data are difficult to interpret, because
Women lost significantly more weight than men during the control period, and the weight
Changes that occurred during one experimental period might be compensating for those
Changes that occurred during prior periods.

Few long-term studies examined the efficacy of aspartame in dietary approaches
to Wei ght management. Blackburn et a1. (1997) investigated whether adding the use of
aspartame to a multidisciplinary weight-control program would improve weight loss and
long-term body weight control. They randomly assigned 163 women who were obese to

ConSume or to abstain from aspartame-sweetened foods and beverage during 16 wk of a

60

19-wk weight-reduction program (active weight loss), a 1-yr maintenance program, and a
2-yr follow-up period. Women in both groups lost ~10% of their initial body weight (10
kg) during active weight loss. The weight loss was positively correlated with exercise
and eating control among women in both groups. In addition, the weight loss was
positively correlated with aspartame intake (r = 0.32, p < 0.01) among women assigned to
the aspartame group. During maintenance and follow-up, participants in the aspartame
group experienced a 2.6% (2.6 kg) and 4.6% (4.6 kg) regain of initial body weight after
71 and 175 wk, respectively, whereas those in the no-aspartame group gained an average
of 5.4% (5.4 kg) and 9.4% (9.4 kg), respectively.

Although there is little evidence that artificial sweeteners reliably decrease
appetite or reduce energy intake, drinking low-calorie “diet” beverages might possibly
lower the energy intake that would occur if caloric, sweetened drinks were consumed
instead. Implementing the substitution of low-calorie “diet” beverages for sugar-
sweetened beverages as one strategy in a multidisplinary weight-loss program can
improve long-term weight control (Blackburn et al., 1997). Whether weight loss is
aChieved when “diet” beverages are consumed cannot be investigated in a cross-sectional
Study. A positive correlation between “diet” soft drinks and the prevalence of overweight
Observed in a cross-sectional study could easily be explained by reverse causation:
oVerweight youth drinking “diet” soft drinks because they are overweight. In the current
StUdy, the regression model was adjusted to control for those who were trying to lose

Weight, in order to explain any positive relationship between overweight and “diet” soft

drink intake.

61

5. Coffee and Tea

Although coffee and tea are not widely consumed by youth, and alone do not
provide many calories, the added sugar, cream and flavored syrups consumed with these
beverages can be a source of discretionary calories. In addition, both coffee and tea
contain stimulants like caffeine and theobromine, all known to increase metabolic rate
(Acheson et al., 1980; Astrup et al., 1990; Bracco et al., 1995). This section will review
recent studies relevant to their intake and body weight.

Metabolic rate and lipid oxidation are increased by caffeine contained in coffee
and tea (Acheson et al., 1980; Astrup et al., 1990; Bracco etal., 1995), but whether long-

term caffeine use affects body weight is unclear (Astrup et al., 1992). In contrast to black
tea, which is a product of fermentation, green tea and oolong tea are non- and partially
fermented/oxidized products, respectively, containing polyphenols, such as
epigallocatechin gallate (EGCG), which may confer effects in addition to those attributed
to caffeine (St-Onge, 2005). Caffeine acts by inhibiting the degradation of cyclic AMP by
Phosphodiesterase, and by increasing noradrenalin release through antagonism of
adenosine (Dulloo et al., 1992). Tea catechins inhibit the degradation of noradrenalin by
catechol O-methyl-transferase (Borchardt & Huber, 1975). Thus, both caffeine and
EGCG increase the stimulation by noradrenalin of energy and lipid metabolism.

In an in vitro study, green tea extract (GTE) stimulated thermogenesis in brown
adipose tissue, while caffeine alone did not, and this effect of green tea was thus
attributed to the polyphenolic compounds (Dulloo et al., 2000). In an in viva experiment,
GTE with low caffeine and high catechin content elevated hepatic lipid metabolism in

' mice fed a high-fat diet for 1 month (Murase et al., 2002). Oolong tea prevented an

62

increase in body weight, adipose tissue and fatty liver in mice fed a high-fat diet as well
(Han et al., 1999). Young Wistar rats drinking green tea (GT) for 3 weeks had reduced
adipose tissue weight compared to controls drinking water (Ashida et al., 2004). GT
reduced glucose uptake into adipose tissue, while uptake of glucose by skeletal muscle
increased (Ashida et al., 2004). Exercise-induced muscular beta-oxidation activity was
increased by catechin supplementation in mice, and whole-body lipid oxidation as
determined by indirect calorimetry was also higher in the catechin-exercise group
(Murase et al., 2006). Thus, animal models have shown effects of GT on reduced
adipose tissue weight and reduced diet-induced obesity mediated via increased energy
expenditure in various target tissues, while glucose uptake into adipose tissue and
lipogenesis decreased as well.

In a study of 10 men conducted in France using a metabolic chamber, capsules
containing GTE were found to increase 24-hr energy expenditure by 2.8% more than
caffeine alone, and 3.5% more than the placebo (Dulloo et al., 1999). Urinary
norepinephrine increased, and a decreased 24-hr respiratory quotient indicated that the
GTE increased fat oxidation. This was a randomized, double-blind, placebo-controlled,
cross-over trial, but the duration of each treatment was only 1 day, too short of a time
period to assess changes in body weight (Dulloo et al., 1999). An epidemiological study
of tea consumption in Taiwan (n=1,103) found that those who consumed tea for more
than 10 years had lower percent body fat, smaller waist circumference, and decreased
waist-hip ratio (Wu et al., 2003). Most Taiwanese drink either green or oolong tea, but
the type of tea, amount consumed and catechin intake were not assessed in this study (Wu

et al., 2003).

63

In a double-blind, controlled trial, 17 Japanese men drank 1 bottle/day of oolong
tea containing GTE, while 18 men (controls) drank oolong tea without GTE (N agao et
al., 2005). Both groups were on a low-calorie diet, and both beverages contained the
same amount of caffeine, but the GTE group had further reductions than controls in body
weight, BMI, waist circumference, body fat mass, and subcutaneous fat area after the 12-
week treatment (N agao et al., 2005).

In the Netherlands, however, among 46 overweight women, GTE consmned for
12 weeks with a low-energy diet had no effect on weight, BMI, waist-hip ratio and fat
mass (Diepvens et al., 2005). Further, GTE consumed for 13 weeks had no effect on
weight maintenance after 104 overweight and moderately obese male and female subjects
had achieved a 7 .5% weight loss with a 4-week low-calorie diet (Kovacs et al., 2004).
Caffeine intake was 300 mg/d in both the treatment and control groups in the weight loss
study (Diepvens et al., 2005), but caffeine intake from other sources was not controlled in
the weight maintenance study (Kovacs et al., 2004).

GT consumers regaining the most weight were observed to have higher habitual
caffeine intakes, and authors suggested that GT and oolong tea might be effective against
obesity only in low-to-moderate caffeine users (Kovacs et al., 2004). The same group of
investigators confirmed this hypothesis in a subsequent study by repeating the experiment
while controlling for habitual caffeine intake; 4 weeks of a very low-energy diet was
followed by GTE or placebo given during a 13-week maintenance period (Westerterp-
Platenga et al., 2005). Although habitual high-caffeine consumers had greater reductions
in weight, fat mass, and waist circumference on the low-energy diet than habitual low-

Caffeine consumers, only the low-caffeine GT group continued these reductions during

64

the following 13-week period, while the other 3 groups regained what was lost
(Westerterp-Platenga et al., 2005). Whether it was the caffeine or catechins in green tea
that prevented weight gain remains unclear.

Short-term thermogenic effects of caffeine have been reported in studies of lean
and obese men and women, but the metabolic rate and lipid oxidation are less stimulated
by caffeine in obese than in normal-weight subjects (Acheson et al., 1980; Bracco et a1,
1995). The thermogenic effects of caffeine are dose-dependent (Astrup et al., 1990), but
whether these short-term effects of caffeine translate to long-term changes in body weight
are unclear. There was no difference in weight loss after 24—weeks of energy-restriction
for those given caffeine (3 x 200 mg/d) compared to placebo (Astrup etal., 1992).

There have been inconsistencies in results of epidemiological studies of the
association between body weight and caffeine intake fi'om food and beverages. Over a
lZ-year period from 1986 through 1998, increases in caffeine intake, as well as increases
in coffee or tea consumption, were associated with slightly smaller weight gain in 18,417
male and 39,740 female health professionals participating in two large prospective
studies (Lopez-Garcia et al., 2006). A meta-analysis of epidemiological studies has

shown an inverse relationship between coffee consumption and type 2 diabetes risk (van
Dam & Hu, 2005), and some type 2 diabetes risk reduction can be explained by the
hypothesized inverse association between caffeine intake and weight gain. Caffeine
intake explained some weight change in the NHANES I Epidemiologic Followup Study

(NHEF S), and caffeine intake decreased diabetes risk only among those who had

Previously lost weight (Greenberg et al., 2005).

65

A study conducted in northern Italy showed no association between BMI and
coffee consumption (Tavani et al., 1994), while in Tromse, Norway, where boiled coffee
is traditionally brewed, a positive association with BMI was observed (J acobsen &
Thelle, 1987; Wilsgaard et al., 2005). A large prospective study conducted in
Copenhagen, Denmark, found that coffee intake was positively associated with waist
circumference in women (without controlling for BMI), and tea intake was positively
associated with BMI-adjusted changes in waist circumference in men (Halkjaer et al.,
2004)

Although black, unsweetened coffee and tea contribute little to energy intake,
added sugar and cream can transfonn coffee and tea into high-calorie sweetened
beverages. Frequent tea times and coffee breaks of caffeine users could then contribute a
large proportion of total calories consumed throughout the day, especially if high-calorie
baked goods are also consumed at these snack occasions. College women who drank
gourmet coffee beverages had a 206 kcal/d higher energy intake, and 32 g/d higher sugar
intake than those who did not (Shields et al., 2004).

The decision to add sugar to coffee or tea was influenced by past dental health
experiences, education and parents, although the sweet taste of sugar giving immediate
gratification outweighed any negative aspect of sugar consumption in the minds of 16-
year-olds who responded to a questionnaire based on Ajzen & Fishbein’s ‘Theory of
Reasoned Action’ (Freeman & Sheiharn, 1997). Dissatisfaction with their appearance
due to acne or obesity was the main factor that persuaded some British teenagers to stop

sweetening their coffee or tea, reduce added sugar, or use a sugar substitute (Thomson,

66

1977). Using coffee or tea to help control weight would not be effective, if sugar is
added to caffeinated beverages.

Sweet, iced tea is frequently consumed in Southern regions of the United States,
and in a study of eating patterns and obesity in 10-year-old black and white collected in
Bogalusa, Louisiana, the sweetened beverage category consisted of 58% soft drinks, 20%
fruit drinks, 19% tea, and 3% coffee (Nicklas etal., 2003). Soft drinks such as Coca-
Cola, Pepsi, and Mountain Dew contain both added sugars and caffeine. Caffeine is an
addictive substance, and many young people are in the habit of consuming soft drinks
instead of coffee or tea at breakfast time and throughout the day. Glucose-caffeine
“energy drinks” recently marketed would also increase added sugar intakes of consumers
who use these beverages as stimulants (Anonymous, 2006).

The 24-hour recall dietary assessment method was used in the Bogalusa Heart
Study to determine whether the children consumed beverages containing added sugar
(N icklas et al., 2003). The FFQ item used in the current NHANES IH study of
adolescent obesity and beverage consumption is specific with regard to caffeine
contained in coffee and tea. However, the respondents were not asked how often sugar
was added to these beverages. The 24-hour recall data was used to compare mean
energy, caffeine, and added sugars intakes of youth who usually consumed the various
types of beverages (according to the FFQ). The association between BMI and caffeine
intake from coffee and tea would be attenuated, if fi'equent consumers of coffee/tea have

high energy and added sugars intakes.

67

6. Alcoholic Beverages

Alcohol is a calorie-dense food, and as such alcoholic beverages potentially relate
to obesity. Alcohol contains 7 calories per gram, with an energy density approaching that
of fat (9 calories per gram), ahnost twice the 4 calories per gram of carbohydrate or
protein. An alcoholic beverage containing 12-15 grams of alcohol would add about 100
calories from the alcohol alone, and calories from carbohydrate and sugars are found in
beer, wine and mixers as well. Some coolers contain up to 7% alcohol and >250 kcal per
8-fl oz due to added sugars, compared to 104 kcal per 8-fl oz soda (Popkin et al., 2006).
These heavily marketed alcohol-containing beverages are packaged like soft drinks and
targeted towards young adults who may have preferred sweetened beverages in
adolescence (Popkin et al., 2006). Because alcohol is an addictive substance, it is
possible, but not established, that marketing these coolers and other alcoholic beverages
toward young adults might engender a link between sugar-sweetened beverage intakes in
youth and a pattern of alcohol consumption that continues into adulthood.

Alcoholic beverages and sugar-sweetened beverages were both top sources of
energy consumed at snacks by young adults age 19-29 years (Zizza et al., 2001). In some
subpopulations of the US, obese women reported significantly higher energy intakes
than non-obese women due to greater consumption of both alcoholic beverages and
sugar-sweetened beverages (Teufel & Dufour, 1990). The following review of literature
pertains to the association of obesity with alcohol intake in adults, while the current study
was conducted with a sample of adolescents age 12-16 years that includes only a few

under-age drinkers.

68

Although heavy drinkers might substitute alcohol for food and therefore become
emaciated and malnourished, moderate alcohol consumers usually add ethanol calories to
their normal diet and potentially gain weight (Suter et al., 1997). Whether alcohol
consumption leads to weight gain depends on a lack of compensation for liquid calories
at subsequent meals. Unlike other macronutrients, alcohol may bypass satiety
mechanisms mediated by gastrointestinal hormones, such as cholecystokinin or gastrin,
and a low hormonal response could lead to under—compensation for alcohol calories
consumed (Astrup et al., 2002). On the other hand, consumption of high-energy preload
drinks of either alcohol or sweetened carbonated beverages did not result in a
compensatory decrease in the ad libitum intake of energy at a subsequent meal, thus there
may be insufficient compensation for both alcohol and carbohydrate calories when
presented in a liquid form (Mattes, 1996; Poppitt etal., 1996). That liquid calories do not
trip satiety mechanisms was shown in an experiment that compared ad libitum food
intake after ingesting either soda or jelly beans, in which there was compensation only for
calories in a solid form (Di Meglio & Mattes, 2000).

The metabolism of alcohol favors the accumulation of fat over its combustion,
because acetate formed from alcohol replaces fat as a substrate for oxidation, and the
accumulation of reduced NADH suppresses lipid oxidation (Suter et al., 1997). Acute
alcohol improved insulin sensitivity in type 2 diabetic subjects, and this effect may be
due in part to the inhibition of lipolysis, because alcohol also increased lactate and
decreased free fatty acid concentrations without affecting insulin secretion (Avogaro et
al., 2004). However, most cross-sectional studies have shown an inverse association

' between alcohol consumption and BMI in women, while findings are inconsistent in men

69

(Hellerstedt et a1. 1990). For example, in Wisconsin, a strong inverse relationship
between drinking and body weight in women was found, but weight was not different
between drinking and non-drinking men (Hansen & Remington, 1992). Another cross-
sectional study showed an inverse association of beer intake with BMI in women, but no
relationship between beer and BMI was found in men (Bobak et al., 2003). Another
recent study of women found an inverse relationship between alcohol and BMI in cross-
sectional analyses of BMI (Wannamethee et al., 2004).

Prospective analyses, however, resulted in a J -shaped relationship between
alcohol and weight gain, such that consumers of a moderate amount (< 30 g/d) had the
lowest risk of weight gain compared to non-drinkers or heavy consumers, and for heavy
consumers, the risk of weight gain increased with the amount of alcohol consumed
(Wannamethee et al., 2004). The inverse association of alcohol with weight gain
observed between non-drinkers and moderate drinkers may be due to reverse causality, as
women who are more susceptible to weight gain may abstain fi'orn alcohol to avoid the
additional calories (Hellerstedt et al., 1990).

A common belief that alcohol consumption leads to weight gain is epitomized by
the expression 'beer belly' to describe the abdominal obesity observed in men who are
heavy beer consumers. Because alcohol consumption often results in abdominal obesity,
waist circumference (WC) or the waist-hip ratio (WHR) may have a clearer relationship
with alcohol consumption than BMI. A study of male self-defense officials in Japan
showed that alcohol intake was positively and strongly associated with WHR, but not
BMI (Sakurai et al., 1997). A recent cross-sectional study of beer and obesity, also found

that beer was associated with WHR but not BMI, but only in men who were nonsmokers

70

(Bobak et al., 2003). Beer intake was not related to WHR in women, but there was a
weak inverse association with BMI (Bobak et al., 2003). In a prospective study, WC,
WHR and BMI was evaluated in relation to alcohol intake in France, where wine is most
commonly consumed, and it was found that increases in both WC and WHR were
associated with alcohol intake in both men and women (Dallongeville et al., 1998).
Again, an inverse relationship between BMI and alcohol was observed in women, while
no association of BMI with alcohol was seen in men (Dallongeville et al., 1998). A
recent study found a J -shaped relationship in both men and women between WHR and
both total alcohol consumption and wine, such that men and women consuming less than
100 g day had a lower WHR than non-drinkers or those consuming more (Lukasiewics et
al., 2005). The same J -shaped relationship between BMI and both total alcohol
consumption and wine existed only in men (Lukasiewics et al., 2005). In contrast,
another recent study found that beer and spirits associated with increases in WC among
women, but not men (Halkj aer et al., 2004). The Copenhagen City Heart Study showed,
however, that beer and spirits was associated with increases in WC among both men and
women, and no relationship was found with wine (Vadstrup et al., 2003).

Drinking alcohol increased energy intake at meals and snacks, not only by the
amount of energy derived from the alcohol itself, but also because food intake may be
less restrained when alcohol is consumed (Astrup et al., 2002). Theoretically, wine could
be more weight gain inducing, because wine has a higher energy density than beer. Total
energy intake at a meal was higher when red wine was served ad libitum rather than beer
or carbonated soft drinks, because energy intake from the beverage was higher with wine,

and it was not compensated for in the food intake (Buemann et al., 2002). However, a 6-

71

week cross-over trial in which 14 free-living male subjects drank 2 glasses (270 mL) red
wine daily at dinner, there were no differences in body weight, body fat or energy intake
assessed by a 3 d dietary record (Cordain et a1. 1997). BMI and body composition also
did not change in a 4-week crossover trial of 450 mL red wine in 34 men (Beulens et al.,
2006), and in a 10-week cross-over trial of 190 mL red wine consumed 5 days per week
by 20 overweight women (Cordain et al., 2000). Further, moderate intake of red wine at
meals is a component of the traditional Mediterranean diet, and this diet was inversely
associated with BMI and obesity (Schroder et al., 2004), and with a smaller weight gain
in a prospective study (Sanchez-Villegas et al., 2006).

Data collected from 3,327 men participating in the British Regional Heart Study
were used to evaluate the association between adiposity and alcohol consumption at
meals (Wannamethee et al., 2005). At a 5-year follow-up when these men were middle-
aged, it had been shown that heavy alcohol intake (>30 g/d) was associated with weight
gain and increased BMI (Wannamethee & Shaper, 2003). All types of alcoholic
beverages were associated with WC, WHR, BMI and %BF, regardless of whether the
alcoholic beverage was consumed at meals (Wannamethee et al., 2005). Again, effects
were stronger for WC and WHR (measures of central adiposity) than for measures of
general adiposity (BMI, %BF) (Wannamethee et al., 2005). Although beer and spirits
were more strongly related to measures of adiposity than wine (Wannamethee et al.,
2005), in contrast, previous studies had shown an inverse association of wine and BMI
among men of Northern Ireland (Marques-Vidal et al., 2001) and Wisconsin (Hansen &

Remington, 1992).

72

In summary, most prospective studies have found that alcohol consumption is
associated with weight gain, particularly in men, while many cross-sectional studies have
shown an inverse association between alcohol and BMI in women. Measures of central
adiposity, including WC and WHR, are more strongly correlated with alcohol intake than
BMI. A J -shaped relationship may exist such that consumers of a moderate amount of
alcohol gain less weight than abstainers, light drinkers or heavy drinkers. Stronger
effects in men than in women, and stronger effects of beer than wine, can be explained by
the fact that more women than men prefer to drink wine, and moderate amounts are
consumed especially when wine accompanies a meal.

The NHANES includes a series of questionnaire items to assess the frequency of
alcohol consumption and the prevalence of heavy drinking. Estimates of the prevalence
of underage drinking using these data were low, because only a few respondents among
the sample of adolescents age 12—16 years included in the current study reported frequent
consumption of alcoholic beverages. The relationship between alcoholic beverage
consumption and obesity is not relevant to the adolescent population. Should future
studies link sweetened beverage intakes of youth with alcohol intake in adults, then this
review may be more relevant to the current study of the relationship between beverage

patterns and adolescent obesity.

D. NHANES dataset selected for research
The National Health and Nutrition Examination Survey (NHANES) is a series of
nationally representative, cross-sectional surveys conducted by the National Center of

IL'Iealth Statistics (N CHS) that include anthropometric measurements made by physical

73

examination, as well as dietary assessment to monitor the nutritional status of the U.S.
population. The current NHANES is a continuous survey beginning in 1999, and
previous surveys include NHANES III, 1988-1994, NHANES H, 1976-1980 and
NHANES I, 1971-1975. This section compares the NHANES with the series of national
food and nutrient surveys conducted by USDA. Unique features of the NHANES 111,
1988-1994, led us to conclude that this dataset is best for the current study.

National nutrition surveys, such as NHANES, provide critical information to
monitor food consumption, dietary intake, and the related health status of the U.S.
population. It is important to assess trends in energy intake to understand the relationship
of diet to overweight and obesity. A series of national food consumption surveys
conducted since the early 1970s are used to identify changes in dietary intake that have
occurred over the years. Although dietary assessment methods used in the national
surveys have been improved, some inconsistencies in the methods warrant caution in
interpretation of data on trends. Changes in dietary intake methodology include the
automation of data collection with the use of standardized probes to assist respondent’s
recall and improved visual aids for estimating portion-sizes (Briefel et al., 1997; Dwyer
et al., 2003b). These methodological improvements were necessary to address
underreporting of intakes, especially among overweight persons, which have also made it
difficult to assess trends.

NHANES dietary intake data are available from the National Center for Health
Statistics (NCHS) for the years 1971-75 (NHANES I), 1976-80 (NHANES 11), 1982-84
(Hispanic HANES), 1988-94 (NHANES HI), and 1999-2002 (Briefe1& Johnson, 2004).

The U.S. Department of Health and Human Services (DHHS) has been conducting

74

national health examination surveys since 1960, and after the Ten State Nutrition Survey
in 1968-70, the nutrition component was added to the health examination survey to form
the first NHANES (Wotecki, 2003). The USDA Nationwide Food Consumption Surveys
(NF CS) and the Continuing Survey of Food Intake by Individuals (CSF II) were
conducted by the U.S. Department of Agriculture (USDA) in 1977-78 (NF CS), 1985-86
(CSFII), 1987-88 (NFCS), 1989-1991 (CSFH), 1994-1996 (CSFII), and the CSFII
Supplemental Children’s Survey was conducted in 1998 (Tippett etal., 1999). Prior to
the 1977-78 NFCS, USDA had been conducting household food consumption surveys
since the 19303, and both household food use and the intake of individuals were
components of the Nationwide Food Consumption Surveys (NF CS). However, the
household food-use portion of the survey may have contributed to respondent burden,
and it was dropped after the 1987-88 NFCS had a very low response rate. With the
collaboration of DHS and USDA per the National Nutrition Monitoring and Related
Research Act of 1990, and integration of the CSFII dietary component (What We Eat in
America) into the NHANES, dietary intake data is now collected continuously (Madans
et al., 2003). The 24-hour recall information from NHANES 2003-2004 will be available
with the next data release.

Unlike the CSFII, in addition to the assessment of dietary intake, the NHANES
includes anthropometry, blood and urine assessments, and many health status
measurements (Bialostosky et al., 2001; Briefel, 2001). Nutritional status is assessed not
only by dietary intake but also by biochemical, anthropometric, clinical and functional
indicators. Because both dietary and health information is available for the same person,

and a link to vital statistics is planned for the future, one of the uses of the NHANES data

75

is to examine the cross-sectional or longitudinal relationships between food intake,
nutritional status, health status, occurrence of disease, and mortality (Dwyer et al.,
2003a). With each representative sample of the U.S. population continuously providing
data, another use of NHANES is to monitor trends in food intake coupled with trends in
risk factors for disease (Dwyer et al., 2003a).

NHANES dietary assessment methods have always included 24—hr dietary recall
interviews. In addition, food frequency questionnaires (FFQ) have been included in the
survey since 1971. A comprehensive semi-quantified FFQ, known as the food propensity
questionnaire, was deve10ped by NCI, pilot-tested in NHANES 2002, and implemented
in NHANES 2003-04 (Krebs-Smith et al., 2004; Subar et al., 2004). The propensity
questionnaire contains more than 100 questions, and it is mailed after the examination to
be self-administered by the respondent (Subar et al., 2004). The FFQ used in NHANES
I-111 and in the Hispanic HANES was extensive, but only a targeted FFQ on milk, green
leafy vegetables, legumes, fish, shellfish, and alcohol is in the NHANES 1999-2001.
Other diet-related questionnaire content include water intake, salt use, dietary
supplements, medications, food security, special diets, food program and school feeding
program participation, and a history of milk consumption. The Diet and Health
Knowledge Survey (DHKS) ascertained dietary knowledge, attitudes and behaviors of
the main meal planner/preparer in CSFII 1989-91 and adults 20 years and older in CSFII
1994-96, and the DHKS had been considered for future integrated NHANES.

At least two 24-hour recalls for a portion of the sample are needed to estimate

day-to-day variation and remove its effect when estimating usual intakes of groups using

' the methods recommended by the National Academy of Sciences. The usual intake

76

distribution provides estimates of the percentage of the population whose intakes are
above the UL or below the EAR or some other reference amounts, and these estimates are
more representative than intakes that are not adjusted for intra—individual variation. The
C-SIDE software has been developed at Iowa State University to perform these
adjustments (Carriquiry, 1999; Caniquiry, 2003). Because the adjustment does not affect
means, it is not necessary to use NAS method and ISU software when mean intakes of
the group are compared to the AI (Murphy, 2003). Unfortunately, it is not possible to use
the method to estimate the usual intake of an individual, as it is valid only for estimating
the usual intake of groups. Because of an individual’s day-to-day variation in dietary
intake, there is usually an attenuation of associations between diet and health outcomes.
By averaging an individual’s intake on more than one day, however, there will be less
attenuation of associations than with the use of only a single 24-hour recall. Although
two or three days does not accurately capture true, usual intake (Basiotis et al., 1987), the
use of more than a single day will improve estimates.

In the past, the NFCS and 1989-1991 CSFH included three consecutive days of
dietary intake consisting of a 24—h dietary recall followed by a 2-day diet record.
Beginning with the 1994-96 CSFH there were two non-consecutive 24-h recalls. Both
were in-person dietary interviews, with the second conducted 3-10 days after the first but
not on the same day of the week. NHANES I and II had only a single 24-hour recall.
Although NHANES III and NHANES 1999—2001 had a second 24-hour recall for 5% and
10%, respectively, of the samples (there was also a Supplemental Nutrition Survey of
adults 50 years and older in NHANES III), beginning in 2002, 100% of the sample

~ (n=5,000/y) had two 24-h recalls. Day 1 was an in-person interview conducted with 3-

77

dimensional food models, while Day 2 was a telephone follow-up using a food model
booklet. Because the dietary assessment method used to collect intakes have been
consistent on only Day 1 of the various surveys conducted over the years, trends data are
based on only the first day of intake.

The current USDA/ARS Automated Multiple Pass Method has five passes, and it
is based on a triple-pass method that had been used successfully in NHANES III, 1988-
1994, and CSFII, 1994-96, to help the respondent remember foods consumed in the 24-
hour recall period. Automated coding in NHANES III with the Dietary Data Collection
system used the University of Minnesota Nutrition Coordinating Center (N CC) code
generator program, and this enabled NCHS to link the NHANES III intakes to food
composition data for vitamin D and other nutrients available only in the NCC database.
The CSFH and current NHANES use the Food Intake Analysis System of the University
of Texas for automated coding of foods. In the 5-pass method, first, respondents are
asked to list, without interruption, all foods and beverages consumed in 24 h on the day
before the interview (“quick list”). Second, the respondent is asked about consumption
of commonly forgotten foods, such as nonalcoholic and alcoholic beverages, sweets, salty
snacks, fruit, vegetables, cheese, bread and rolls. Third, they are asked to give the time
and name of the meal occasion. The fourth pass is the “detail cycle” in which probes
elicit details about each meal occasion and each food, including method of preparation,
where it was obtained and amount consumed. In the fifth pass the respondent is asked
whether the respondent drank or ate anything else during the 24-hour period. Thus, by

providing a variety of memory cues at different times throughout the interview, the 5-

78

pass method assists the respondent in remembering foods, and it has significantly reduced
underreporting.

In summary, the utility of the NHANES dietary data is that it can be linked with
anthropometric, biochemical, clinical and disease history information. NHANES
includes measured rather than self-reported height and weight, in addition to waist
circumference and skinfold thickness measurements used to assess body composition.
Thus, a cross-sectional analysis of the relationship between dietary factors and obesity
are possible because intake and anthropometric data are available for the same individual.
Further, detailed food intake records can also be linked to the Nutrient Database for
Standard Reference and Pyramid Servings databases to estimate energy intake and
consumption of added sugars (Murphy, 2003). Although only a single 24-hour recall was
available for NHANES I and NHANES II, in NHANES HI and beyond, a second day is
available on at least a portion of the sample to estimate usual intake using the methods
recommended by the National Academy of Sciences. In addition, an extensive FFQ was
available in NHANES III, and this was the main reason why this dataset was selected for
use in this study. Data from the food propensity questionnaire will not be available until

the release of NHANES 2003-2004 dataset (Krebs-Smith et al., 2004; Subar et al., 2004).

79

Chapter 3

METHODS

A. Study design

This study was a cross-sectional, secondary data analysis of beverage intakes and
obesity among U.S. adolescents age 12-16 years. Data fi'om the Third National Health
and Nutrition Examination Survey (NHANES 111, 1988-94) (DHHS, 1994) were used in
these analyses. Aim 1 was to examine the mean frequency of consuming beverage types,
including milk, 100% fruit juice, sugar-sweetened beverages (regular soft drinks and fi'uit
drinks) and other beverages (diet soft drinks, coffee/tea, and alcoholic beverages), and the
prevalence of beverage patterns (i.e., the combination of beverage types most frequently
consumed); to examine the association of beverages with meal occasions and meal
locations; and to examine mean energy, dietary fat, total sugars, added sugar, calcium and
other nutrient intake of beverage pattern groups. Aim 2 was to examine the distributions
of sociodemographic and lifestyle characteristics that are risk factors for overweight
among beverage pattern groups. Aim 3 was to examine the association of obesity
outcomes (BMI percentile, percentage body fat, and overweight/risk for overweight) with
these beverage pattern groups used as “dummy" variables in regression analyses adjusted

for potential covariates of interest.

B. Subjects/ study participants
1. Dataset
The NHANES III, 1988-94, data were used for these analyses, rather than the

more recent NHANES, 1999-2002, data, because the food frequency questionnaire used

80

to assess each youth’s typical beverage pattern is available only in NHANES H1. The
National Center for Health Statistics (NCHS) of the Centers for Disease Control and
Prevention conducts NHANES to obtain nationally representative information on the
health and nutritional status of the U.S. population. The NHANES III survey received

human subject approval from the Centers for Disease Control and Prevention.

a. NHANES interview and examination components

Data pertaining to youth participating in NHANES HI were obtained from
household interview component, and from the health examination, anthropometric and
interview components conducted at a mobile examination center (MEC). Household
interview data were collected using the Household Screener Questionnaire, Family
Questionnaire and Household Youth Questionnaire. The Household Youth Questionnaire
was administered to a proxy respondent, usually the parent or guardian of children and
youths 2 months to 16 years of age. Youth were interviewed in a private setting at the
MEC without a proxy respondent. Interview components conducted at the MEC included
the MEC Youth Questionnaire (ages 8-16 years), the 24-Hour Dietary Recall (all ages)
and Dietary Food Frequency (ages 12—16 years). The Dietary Food Frequency interview
and other interviews of NHANES HI participants age 17 years and older were conducted

using the Household Adult Questionnaire and MEC Adult Questionnaire.

b. NHANES dataset for public use

The NHANES IH dataset that is available for public use can be downloaded from

' the NCHS website, or it can be retrieved from the NHANES III 1988-94 CD-ROM using

81

the NCHS Statistical Export and Tabulation System (NCHS, 1997). Variables pertaining
to the current analysis of adolescents 12-16 years were obtained from three separate files
on the NHANES III, 1988-94, CD-ROM: the Household Youth Data File, the
Examination Data File, and the Dietary Recall Data Files. The Household Youth Data
File contained the individual and family demographic information. The Examination
Data File included the FF Q data for youth 12-16 years, and the physical activity, smoking
and alcohol data from the MEC Youth Questionnaire. In addition, the Examination Data
File included sexual maturation stage data from the physician’s examination for ages 8-
18 years, and for all ages, body measurements, as well as the total nutrient intakes, and
intakes of plain drinking water obtained in the dietary interview. Detailed information
about the individual foods reported by respondents was obtained from the Dietary Recall

Data Files.

c. Sample design

NHANES III was conducted in 2 phases: 1988-1991 and 1991-1994. Subgroups
of the population, such as non-Hispanic blacks, Mexican-Americans, children and the
elderly were over-sampled to allow increased precision in estimates for these groups.
With the use of the sampling weights, participants were a representative sample of the
civilian, non-institutionalized U.S. population age 2 months and over, selected using a
complex, stratified, multistage, probability cluster sample design (DHHS, 1994). The
first stage of the sample design was to select the Primary Sampling Unit (PSU) or survey
location, the second stage was to select areas within the PSU, the third stage was to select

‘ households, and the fourth stage was to select individuals within these households. There

82

were 89 survey locations, and over the six years, 39,695 sample persons were selected,
33,994 (86%) participated in the household interview, and 30,818 of these participants
(78% of the selected sample) were later examined in the mobile examination center

(MEC).

2. Analytic sample

Adolescents age 12-16 years who were interviewed and examined in NHANES
III comprised the analytic sample for this study. This age group was selected partly
because of the availability of dietary data for youth 12 years and older, and partly because
physical activity was measured differently for youth age 8-16 and adults age 17 years and
older. In addition, the TV watching item was not available in NHANES 111, 1988-94, for
those age 17 years and older. The FFQ was not available in the CSFII or more recent
NHANES, 1999-2002, and in NHANES III, 1988—94, the FFQ was not administered to
children younger than 12 years. Although children age 6-11 years often participated with
a proxy in the 24-hour recall dietary interview, the proxy was the primary respondent for
children younger than 12 years.

The NHANES III sample included 2,216 adolescents aged 12-16 years selected
for examination, and of those, 2,079 (93.8% of 2,216) were examined in the MEC. Upon
examination, 19 respondents (0.9% of 2,216) found to be ineligible (16 teenage girls were
pregnant, and three were lactating) were excluded from the current study. Teenage girls
who were pregnant or lactating were excluded from this study, to avoid misclassification
of overweight and obesity, as the expansion of the waistline, and increased body weight

' in childbearing and in the postnatal period are not due to increased adiposity.

83

Because the current study used both 24-hour recall and FFQ data, the analytic
sample was limited to adolescents age 12-16 years who completed a 24-hour recall
dietary interview as well as all beverage items in the FFQ. Among the 2,060 non-
pregnant, non-lactating youth, there were 1,997 youth (90.1% of 2,216; 941 boys and
1,056 girls) who had a complete, reliable 24—hour recall dietary interview, and of these,
97 (4.4% of 2,216) were missing beverage frequency data (79 left at least one of the
beverage items blank, and another 18 gave a “don’t know” response to at least one
beverage item) and another 28 (1.3% of 2,216) were missing data on water intake. The
24-hour recall interview was reliable, but incomplete (missing information for one or
more meals, foods or beverages) for 11 respondents (0.5% of 2,216), and unreliable for
10 respondents (0.5% of 2,216). The 24-hour recall data collected from 20 respondents
(0.9% of 2,216) was lost due to a computer malfunction or file transfer problem, and
there were 22 non-respondents (1.0% of 2,216) to the 24-hour recall dietary interview.
These 22 non-respondents included those who refused, those whose proxy did not know
what they ate, and those who were not interviewed for some other reason. After
excluding those with missing data, the remaining analytic sample of 1,872 youth (84.5%
of 2,216) included 871 boys (46.5% of 1,872) and 1,001 girls (53.5% of 1,872) aged 12-
16 years.

Tests for differences between participants and non-participants, and tests for
differences between the analytic sample and those who were excluded were conducted.
Some characteristics of adolescents were ascertained at the household interview, and
those who were examined in the MEC (participants: 11 =2,079) were compared to those

' who were interviewed but not examined (non-participants: n=137). The household

interview data showed that characteristics of participants and non-participants were not
significantly different (p<0.05). Tests for differences in characteristics ascertained at
either the interview or examination were used to compare the analytic sample (n=l,872)
and those excluded due to ineligibility or missing data (n=207), and no differences were

found (p<0.05).

C. N HANES data collection procedures

This research was a secondary data analysis of NHANES HI data. Therefore, the
data collection procedures were those followed by NHANES staff, while data analysis
procedures were those followed in conducting the research plan of current study. The
NHANES data collection procedures discussed below include anthropometric assessment
methods and dietary assessment methods. Data analysis procedures were steps
undertaken to prepare the data for statistical analysis, such as exporting the variables of
interest from the NHANES III dataset on CD-ROM and merging the data into a SPSS
system file, recoding extreme values reported in the FFQ and recoding continuous
measures into categorical variables. Further data analysis procedures included
determining overweight status and percentage body fat from BMI and skinfold thickness
measures, categorizing beverage items, defining beverage patterns, updating the
NHANES 1H food composition database, and calculating intakes of nutrients that were
appended to the database. The data analysis procedures pertaining to each specific aim
will be discussed in the variable definition and statistical analysis sections that follow this

section on data collection procedures.

85

1. Anthropometric assessment methods
Body measurements included weight, height, and skinfold thickness. Standard
procedures followed for the anthropometric measurements in NHANES IH were
described in the NHANES III Body Measurements (Anthropometry) reference manual

(DHHS, 1998).

a. Weight

Kilogram body weight was measured with accuracy to 2 decimal places using a
Toledo electronic-load cell scale. No adjustment was made for 0.1-0.2 kg weight of
clothing (underpants, a disposable paper examination gown, paper pants, and foam rubber

slippers).

b. Height

Stature was measured using a Holtain Height Stadiometer. Participants stood on
the floorboard of the stadiometer with heels, buttocks, scapulae and the back of the head
placed against the vertical backboard of the stadiometer. The head was positioned in the
Frankfort horizontal plane. Hair ornaments, buns, braids, etc. were removed to obtain an
accurate measurement. The examiner read the measurement to the assistant who
recorded it to the nearest 0.1 cm, and a photograph of the measurement tape was taken to

verify the accuracy of the reading.

c. Skinfold thickness

Measurements of body fat included skinfold thickness at four different anatomic

86

body sites: triceps, subscapular, suprailiac, and thigh. All skinfolds were measured using
Holtain skin calipers. The measurements were taken on the right side of the body. The
fold of skin and underlying subcutaneous adipose tissue was gently grasped between the
examiner’s left thumb and forefingers about 2.0 cm above the place where the
measurement was to be taken. The jaws of the calipers were placed at the marked level,
perpendicular to the length of the fold, and the skinfold thickness was measured to the
nearest 0.1 mm while the fingers continued to hold the skinfold. The actual measurement

was read from the caliper about 1-2 seconds after the caliper tension was released.

2. Dietary assessment methods

a. The food frequency questionnaire (FFQ)

The FFQ was administered to youth aged 12-16 years as part of the interview that
took place in the MEC. The subjects were asked to think about their usual diet over the
past month, and to respond with how often they usually consumed the FFQ items per day,
per week, per month, or not at all. NHANES staff converted the fi'equency data to times
per month, by multiplying times per day by a factor of 30.4, for example: 3 times per day
= round (3 * 30.4) times per month = 91 times per month. Amounts reported per week
were divided by 7 and then multiplied by 30.4 to convert to times per month: 5 times per
week = round (5/7 * 30.4) times per month = 22 times per month.

The FFQ included 12 beverage items, as listed in Table 1. The categories of FFQ
beverage items listed in Table 1 included milk and milk drinks (chocolate and regular),
100% fruit juice (citrus and non-citrus), sugar-sweetened beverages (fruit drinks and

regular soft drinks), and other beverages (diet soft drinks, coffee/tea, and beer/wine/

87

Table 1. Beverage categories and beverage subcategories corresponding to beverage

items in the NHANES III food frequency questionnaire (F F Q)

 

Beverage categories &

beverage subcategories

FF Q beverage items

 

1. Milk and milk drinks,

chocolate & regular

2. 100% mu juice,

citrus & non-citrus

3. Sugar-sweetened bevg.
3a. Fruit drinks
3b. Regular soft drinks
4. Other beverages
4a. Diet soft drinks

4b. Coffee/ tea

4c. Alcoholic

beverages

1a. How often did you have chocolate milk and hot cocoa?

lb. How often did you have milk to drink or on cereal?

Do not count small amounts of milk added to coffee or tea.

2a. How often did you have orange juice, grapefruit juice
and tangerine juice?

2b. Other fruit juices such as grape juice, apple juice,
cranberry juice, and fruit nectars?

3a. How often did you have Hi-C, Tang, Hawaiian Punch,
Koolaid, and other drinks with added vitamin C?

3b. Regular colas and sodas, not diet?

4a. Diet colas, diet sodas, and diet drinks such as Crystal
Light?

4b (1). Regular coffee with caffeine?

4b (2). Regular tea with caffeine?

4c (1). Beer and lite beer?

4c (2). Wine, wine coolers, Sangria, and champagne?

4c (3). Hard liquor such as tequila, gin, vodka, scotch,

rum, whiskey and liqueurs, either alone or mixed?

 

88

liquor). An additional item in the FFQ asked what type of milk was usually consumed:
whole milk, reduced fat (2%), low fat (1%), or skim/nonfat milk. The FFQ fi'uit drinks
item asked specifically about Hi-C, Tang, Hawaiian Punch, Koolaid and other drinks
with added vitamin C. The FFQ coffee/tea items asked specifically about regular coffee
with caffeine, and regular tea with caffeine. An open-ended section of the FFQ permitted
the respondent to list items usually consumed, but not specifically mentioned in the
questionnaire, thus, consumption of fruit drinks without added vitamin C, decaffeinated
coffee or decaffeinated tea may have been reported as an open-ended response. Only the
responses to the FFQ items that were solicited, and not the responses to the open-ended
section of the FFQ were used in this study, because the responses to the open-ended

section were “voluntary and sporadic” (V elie, personal communication).

b. The 24—hour recall dietary interview

The 24-hour recall dietary interviews were conducted in English and Spanish by
bilingual dietary interviewers in a private room to ensure confidentiality. Children aged
12 years and older were permitted to report their own intake without a proxy respondent.
The dietary interviewers contacted other information sources such as care providers and
schools to obtain complete dietary intake data for respondents. Interviewers reviewed
and edited the dietary recall data to ensure that they were as complete as possible. The
interviewers made initial determinations regarding the completeness and reliability of the
dietary recalls, and final determinations were dependent on review by NCHS staff. For
quality control, NCHS and contractor staff were present to observe dietary interviews,

and interviews were also audiotaped for retrospective reviews. The dietary interviewers

89

worked in two-person teams; performing 10-percent cross-check reviews of their
partners' work using printed recall reports, and verification problem forms were
submitted if there were any discrepancies between partners. Extreme values that were
flagged in the editing process were then verified by reviewing the interviewer’s
documentation of any unusual entries, by reviewing the audiotape of the interview to
check the information presented by the respondent, and by reviewing any verification
problem forms submitted by the interviewer’s partner who cross-checked the work.
Finally, newsletters, field memoranda, telephone calls, and staff retraining sessions were
other methods used to maintain quality control during the survey (DHHS, 1996a).

[The dietary interview was conducted while using a microcomputer and the
NHANES III Dietary Data Collection (DDC) System. The protocol of the DDC System
used in NHANES IH, 1988-94, incorporated the three components of the multiple-pass
24-hour recall method: the quick list, detailed description, and review (Johnson et al.,
1996). However, the multiple-pass method was not denoted as such until it was later
described and implemented for the 1994-96 Continuing Survey of Food Intakes by
Individuals (CSFII) (Guenther et al., 1995). USDA continued to validate the multiple-
pass method as an expanded five-step protocol: the quick list, forgotten foods list, time
and occasion, detail cycle, and final probe review (Conway et a1, 2003; Conway et al.,
2004). This method was not used officially by NCHS for dietary data collection until the
CSFII was integrated into the NHANES in January 2002 (Dwyer et al., 2001). However,
it seems that the DDC of the NHANES, 1988-94, was actually a forerunner of the

multiple-pass method.

90

A quick list of foods consumed in the previous 24-hour period was first recorded
in the DDC. After the quick list was entered, the DDC then provided food specification
screens with a series of standardized prompts and choices that allowed the interviewer to
select and enter specific foods, amounts, ingredients and preparation methods. The DDC
prompts helped to determine the level of detail about specific foods necessary for the
computer system to assign numerical codes linking the foods consumed to the nutrient
composition database. The primary source of food composition data for NHANES III
was the U.S. Department of Agriculture (USDA) Survey Nutrient Database; USDA
provided two nutrient files (first in 1993, second in 1995) for use in NHANES HI
(NCHS, 1997). Because the Nutrition Coordinating Center (NCC) of the University of
Minnesota had developed the DDC, the NHANES III survey foods were also cross-
referenced to the NCC nutrient composition files. Although vitamin D was not in the
USDA Survey Nutrient Database, the second release of the NHANES III 24-hour recall
dietary data included vitamin D intakes, because vitamin D was in the NCC nutrient

composition database (N CHS, 1998).

c. Assessment of water intakes

Although water intakes were not ascertained using the 24—hour recall method, the
dietary interview included a question to respondents about how many glasses/cups or
fluid ounces of plain water they usually drank daily. If the response was in terms of
glasses or cups, participants were asked to specify the fluid ounces per glass or cup.

Before release of the public-use data, responses reported in terms of numbers of

91

glasses/cups per day were converted to fluid ounces per day, by multiplying the volume

of water in the glass/cup by the number of glasses/cups consumed per day.

D. Variable definition
1. Beverage categories and beverage patterns

Aim 1 of this study included an examination of the mean frequency of beverage
consumption, and mean nutrient intake of beverage pattern groups, so youth were
classified by the beverage pattern (i.e., combination of beverage types) most fi'equently
consumed, as reported in the FF Q. The 4 main categories (and 7 subcategories) of the 12
FFQ beverage items listed in Table 1 included (1) milk and milk drinks, chocolate and
regular, (2) 100% fruit juice, citrus and non-citrus, (3) sugar-sweetened beverages (3a.
fruit drinks, 3b. regular soft drinks), and (4) other beverages (4a. diet soft drinks, 4b.
coffee/ tea, and 4c. beer/ wine/ liquor). The FF Q chocolate milk and regular milk items
were combined to form the milk category. Citrus fruit juice and non-citrus fruit juice
were combined to form the 100% fruit juice category. The coffee and tea items were
combined, and the alcoholic beverage items (beer, wine and hard liquor) were combined.
For the purposes of defining beverage patterns, fruit drinks and regular soft drinks were

combined to form the sugar-sweetened beverages category.

a. Categorization of beverages reported in 24-hour recall dietary interviews
Aim 1 included an examination of the fi'equency with which beverages were
consumed at meal occasions (breakfast, lunch, dinner, snack) eaten at locations at-home

- and away-from-home. The 24-hour recall data were used, because the FF Q does not

92

specify meal occasion and meal location. The USDA survey food codes for beverages
reported in the 24-hour recall dietary interview were classified into the beverage
categories and subcategories listed in Table 1. The 24-hour recall beverage categories
were identical to those described by the FFQ items, except that the fi'uit drinks category
included fruit drinks without vitamin C, and the coffee/tea subcategory included
decaffeinated coffee and tea. The milk category included milk poured on cereal as well
as milk consumed as a beverage. Milk combined with cereal was identified using the
code for the type of combination known as cereal with additions. Beverage mixtures,
such as coffee with milk, were identified using the code for this type of combination, and
the USDA survey food code for the main component of the beverage mixture was used to

classify the combination into the appropriate category.

h. Classification of youth by beverage pattern frequently consumed

Beverage patterns were defined a priori by identifying all possible combinations
of three categories of beverages: milk (milk & milk drinks), sugar-sweetened beverages
(fruit drinks & regular soft drinks) and juice (100% fruit juice). To classify youth into
beverage pattern groups, the FFQ beverage frequencies were used to identify
combinations of beverages that characterized a youth by his/her ‘typical’ beverage
pattern. Two different methods, the ‘frequency’ method, and the ‘share’ method, were
developed to classify youth by the combination of beverage types most frequently
consumed. For each beverage, the ‘share’ was determined by expressing the beverage
frequency as a percentage of the summed frequency for all beverages. Criteria to identify

the most fiequently consumed beverage types using the ‘frequency’ and ‘share’ methods,

93

respectively, were frequency 21 time/day, and share 225% of total beverage fiequency.
Notations for beverage patterns were combinations of capital letters M, S and J to
represent milk, sugar-sweetened beverages and juice, respectively, consumed at a level
meeting the criteria, and lowercase letters m, s and j represented consumption of less than
the specified amount. The beverage patterns were named according to the eight beverage
combinations: ‘all caloric beverages’ (MSJ), ‘milk & sugar-sweetened beverages’ (MSj),
‘milk & juice’ (MsJ), ‘milk’ (Msj), ‘sugar-sweetened beverages & juice’ (mSJ), ‘sugar—
sweetened beverages’ (mSj), ‘other beverages & juice’ (msJ) and ‘other beverages’ (msj).
Other beverage variables in the model in which beverage patterns were correlated with
body weight or body fat included ounces of water, and type of milk (whole, 2%, 1%,

skim) usually consumed.

2. Meal occasion and meal location

Aim 1 included an examination of the mean frequency of beverage consumption
at meal occasions and at meal locations reported in the 24-hour recall dietary interview.
In the same way that the contribution of food groups to total nutrient intakes were
estimated from the mean nutrient intakes obtained from each food group (Block et al.,
1985a; Block et al., 1985b), mean frequency of beverage consumption at meal occasions
and at meal locations among consumers of the various types of beverages was estimated
to determine the percentage contribution of meal occasions and meal locations to the total
beverage frequency. The meal occasions’ and meal locations’ shares of the total
beverage frequency were estimated to examine the association between meal context and

‘ beverage choice. Hypotheses were that milk or 100% fi'uit juice would be consumed

94

more frequently than sweetened beverages at breakfast and other meals eaten at home,
and sweetened beverages would be consumed more frequently than milk or 100% juice at
the snack occasion and at locations away from home.

Meal occasions were self-reported by the respondent. After he/ she was handed a
card listing the various meal occasions, he/she was asked to name the meals that
comprised the foods and beverages reported during the 24-hour recall interview. The list
of meal occasions included breakfast/brunch, lunch, dinner/ supper, and snack/beverage
break. Spanish translations of meal occasion names were desayuno/ahnuerzo
(breakfast/brunch), comida (lunch), cena/merienda (dinner/supper), and entre
comidas/bebida (snack/beverage break). Meal locations included home, work, school, a
fast food or take-out establishment, a restaurant (with table service) or cafeteria,

someone’s home, transit, and other meal places.

3. Dietary quality assessment

Although youth were classified by their ‘typical’ beverage pattern based on
beverage fiequencies reported in response to the FFQ, because the FFQ in NHANES III
was not designed to assess nutrient intake, the 24-hour recall method was used instead to
estimate mean nutrient intakes of beverage pattern groups. Intakes of food energy (Kcal),
fat (g), sugars (g), caffeine (mg), vitamin A (pg RAE), vitamin C (mg), folate (pg DFE),
calcium (mg), potassium (mg) and vitamin D (pg) fi'om the total diet, from beverages,
and fi'om non-beverages were assessed using the 24-hour recall dietary intake data.

Because of the risk for overweight, intakes of food energy (Kcal) and

macronutrients such as total fat (g), saturated fat (g), carbohydrate (g), total sugars (g),

95

and added sugars (g) are primarily of interest. It’s recommended that the diet contain
<30% calories from total fat, and <10% calories from saturated fat or added sugars.
Intakes of caffeine (mg) from soft drinks, coffee and tea might also be a concern for
teens. Teenagers, especially girls, are at risk for inadequate intake of some key nutrients.
Many of these vitamins and minerals can be obtained from the diet by drinking milk or
juice. Thus, additional nutrients of interest included vitamin A (pg RAE), vitamin C
(mg), folate (pg DFE), calcium (mg), potassium (mg) and vitamin D (pg). The mean and
median of the distribution of daily nutrient intake were compared to Dietary Reference
Intake (DRI) criteria, such as the age- and gender-specific Estimated Average
Requirement (EAR) or Adequate Intake (AI) level shown in Table 2 (Institute of
Medicine (IOM) of the National Academies of Science, 1997; IOM, 1998; IOM, 2000;
IOM, 2001; IOM, 2002).

The primary source of food composition data for NHANES ID was the U.S.
Department of Agriculture (USDA) Survey Nutrient Database; USDA provided two
nutrient files (first in 1993, second in 1995) for use in NHANES III (NCHS, 1997).
Because the Nutrition Coordinating Center (NCC) of the University of Minnesota had
developed the Dietary Data Collection system for NHANES HI, survey foods were also
cross-referenced to the NCC nutrient composition files (NCHS, 1998). Although vitamin
D was not in the USDA Survey Nutrient Database, the second release of the NHANES III
24-hour recall dietary data included vitamin D intakes, because vitamin D was in the
NCC nutrient composition database (NCHS, 1998). The USDA Food and Nutrition
Database for Dietary Studies (FNDDS) (USDA, 2004) was used to access food

composition data for vitamin A (pg RAE), folate (pg DFE), caffeine and total sugars,

96

Table 2. Dietary Reference Intakes (DRI) for adolescents aged 12-16 years

 

Age (v) of Boys Age Iv) of Girls
Nutrient 12-13 14-16 12-13 14-16

 

Estimated Average Rguirement (EAR)

 

Carbohydrate (g) 100 100 100 100
Vitamin A (pg RAE) 445 630 420 485
Vitamin C (mg) 39 63 39 56
Folate (pg DFE) 250 330 250 330
Adequate Intake
Calcium (mg) 1300 1300 1300 1300
Potassium (mg) 4.5 4.7 4.5 4.7
Vitamin D (pg) 5 5 5 5

 

SOURCES: Dietary Reference Intakes for Calcium, Phosphorus, Magnesium,
Vitamin D, and Flouride (1997). Dietary Reference Intakes for Thiarnin, Riboflavin,
Niacin, Vitamin B6, Folate, Vitamin B12, Pantothenic Acid, Biotin, and Choline (1998).
Dietary Reference Intakes for Vitamin C, Vitamin E, Selenium, and Carotenoids (2000).
Dietary Reference Intakes for Vitamin A, Vitamin K, Arsenic, Boron, Chromium,
Copper, Iodine, Iron, Manganese, Molybdenum, Nickel, Silicon, Vanadium, and Zinc
(2001). Dietary Reference Intakes for Energy, Carbohydrates, Fiber, Fat, Protein and
Amino Acids (Macronutrients) (2002). DRI reports from the Institute of Medicine of the

National Academies may be accessed via www.n_ag3.edu.

97

because intakes are either unavailable in the NHANES HI dataset, or if available, units
are inconsistent with DRI recommendations.

The added sugars content of food was obtained from the USDA Pyramid Food
Group Servings Database, Version 2.0 (Cook and Friday, 2004). Added sugars include
sugars and other sweeteners eaten separately or added to foods at the table as well as
sweeteners used as ingredients in prepared and processed foods such as breads, cakes and
other grain desserts, soft drinks, jams and jellies, candies and ice cream. Added sugars
and other sweeteners include white sugar, brown sugar, raw sugar, corn syrup, corn syrup
solids, high fi'uctose corn syrup, malt syrup, maple syrup, pancake syrup, fructose
sweetener, liquid fructose, honey, molasses, anhydrous dextrose, crystal dextrose, and
dextrin eaten separately or used as ingredients in prepared or processed food. Added
sugars do not include naturally occurring sugars such as lactose in milk and fructose in
fruit. In the USDA Pyramid Servings database, a serving of added sugars was defined as
the quantity of sweetener that contains the same amount of total sugars (4.196 grams)
provided by a teaspoon (4.2 grams) of table sugar. Therefore, a serving of added sugars

is equivalent to 4.2 grams of table sugar 0 (Cook and Friday, 2004).

4. Sociodemographic variables
Sociodemographic factors examined were gender, age, race-ethnicity, region,
urbanization, parental overweight/obesity, household income and other indicators of
social economic status (SES). Categories for race-ethnicity included non-Hispanic white,

non-Hispanic black, Hispanic and other race-ethnic groups. Four broad regions of the

98

U.S. included the Northeast, Midwest, South and West. The urban/rural classification
was a metropolitan area having a population 21 million vs. all other areas. The poverty
income ratio was the ratio of household income to the poverty level for household size in
an urban or rural area. Other indicators of SES included the family size and
characteristics of the household head or family reference person (FRP), such as their
gender, age group, marital status, education and employment status. BMI was calculated
from reported weight and height of the youth’s mother and father to determine whether

both or one of two parents were overweight (BMI 225 kg/mz) or obese (BMI 230 kg/mz).

5. Lifestyle variables

Lifestyle factors examined were smoking, alcohol use, weight loss behavior,
frequency of breakfast consumption, type of milk usually consumed, physical activity
and TV watching. The questionnaire that included lifestyle questions, such as those
asking youth whether they currently smoked cigarettes or drank alcohol in the past year,
was administered in a private setting at the mobile examination center (MEC) without a
proxy respondent. Youth were also asked whether they were currently trying to lose
weight and how often they ate breakfast (e. g., every day, some days, rarely, never,
weekends only). The FFQ included a question asking whether whole, 2%, 1% or

skim/nonfat milk was usually consumed.

a. Physical activity
Physical activity was assessed using the following questionnaire items asking

youth about their exercise habits and participation in team sports:

99

1. How many times per week do you play or exercise enough to make you sweat

or breath hard?

2. In the past year, how many sports teams or organized exercise programs have

you been involved in (not including physical education or gym classes)?
The first item above was termed “self-reported exercise-induced sweating” when it was
developed and validated (Ainsworth et al., 1993; Gruner et al., 2002). This measure of
physical activity is used in national surveys, such as CSFH and NHANES, as well as
several epidemiological studies, including the Male Physicians’ study (Lee et al., 1999;
Liu et al., 2000). Responses to a questionnaire item asking, “How often do you exercise
vigorously enough to work up a sweat?” has been shown to be correlated with measures
of physical fitness, such as maximal oxygen uptake and treadmill time during a maximal
exercise test (Siconolfi et a1, 1985; Kohl et al., 1988), and the item has more recently
been validated among patients with coronary artery disease (Gruner et al., 2002). The
sweat frequency question was investigated in relation to the frequency and duration of
different physical activities with known intensity ratings coded in terms of metabolic
equivalents (MET) (Ainsworth et al., 1993). Although there was no association with
light (<4 MET) physical activity, both moderate (4.0-5.9 MET) and intense (26 MET)
physical activity correlated well with self-reported exercise-induced sweating (Gnmer et

aL,2002)

b. TV watching

Youth were asked twice, once in the household and again at the MEC, about time

spent watching TV the previous day, and the average hours/day in two days was used in

100

the current study. Hours of “screen time” watching TV or playing video games has been

shown to be a reliable and valid indicator of sedentary behavior (Robinson, 1999).

6. Overweight and adiposity

Obesity outcome variables defined for the purposes of this study included both
overweight and adiposity (i.e., percentage body fat). To adjust weight for height, the
Body Mass Index (BMI) was calculated from measured height and weight using the
formula: BMI = weight (kg)/ height (m)2. To adjust BMI for age and gender, the SAS
program available from NCHS was run to calculate the youth’s BMI percentile gender-
specific BMI-for-age percentile reference data provided by the 2000 CDC Growth Charts
for boys and girls, 2-20 years, because criteria based on the BMI-for-age percentiles were
recommended to assess overweight/risk for overweight in children and adolescents
(NCHS; Kuczmarski et al., 2000). In addition, percentage body fat was determined from
summed triceps and subscapular skinfold measurements (Slaughter et al., 1988), because
BMI does not accurately assess body composition, e. g., overweight youth with a high
proportion of lean body mass are not overfat (Forbes, 1987).

The BMI percentile and percentage body fat were dependent variables in two
separate linear regression analyses. The BMI percentile was used to classify youth by
weight status (i.e., underweight: BMI <5th percentile, normal body weight: BMI 25th
percentile and BMI <85th percentile, at risk for overweight: BMI 285th percentile and

<95th percentile, and overweight: BMI 295th percentile). The presence of

overweight/risk for overweight (i.e., BMI 285th percentile) was the dependent variable

101

for logistic regression analyses. Underweight youth were excluded from logistic
regression analyses.

Percentage body fat was determined from summed triceps and subscapular
skinfold thickness (mm) measurements using prediction equations for children and
adolescents age 8 to 18 years developed by Slaughter et a1. (1988). There were 35 youth
whose triceps or subscapular skinfold thickness was too large for the calipers. Rather
than excluding these obese adolescents from analyses, percentage body fat was calculated
using the group’s maximum skinfold thickness (mm) measurements in place of the
missing values.

Slaughter et a1. (1988) developed separate skinfold equations for prepubescent,
pubescent, and postpubescent white and black male children/adolescents and all female
children/adolescents whose summed tricep and subscapular skinfolds were 35 mm or
less, and male and female children/adolescents whose summed tricep and subscapular
skinfolds were greater than 35 mm. Adolescent boys age 12-16 years were classified by
race and sexual maturation (SMA) stage to use the Slaughter equations. White and black
male adolescents were classified by the composite pubic hair and genitalia rating as
follows: prepubescent (SMA stages 1.0 - 2.0), pubescent (SMA stages 2.5-3.5) and
postpubescent (SMA stages 4.0-5.0) (Boileau et al., 1984; Slaughter et al., 1984).

The physician’s examination included an assessment of SMA according to the 5-
stages of developing pubic hair (male and female), male genitalia, and female breasts
(Tanner, 1962; DHHS, 1991). The Physician Examiner’s Training Manual (DHHS,
1991) described the 5-stage coding of the sexual maturation stages for each of the three

areas of assessment as shown in Table 3. A composite score was calculated by

102

Table 3. Physician’s assessment of sexual maturation and 5-stage classification of pubic

hair (male and female), male genitalia and female breasts.

 

Stage

Description of sexual maturation stage

 

Pubic haLir (male and female)

Preadolescent. The vellus over the pubis is no further developed than that over
the abdominal wall, i.e., no pubic hair.

Sparse growth of long, slightly pigmented downy hair, straight or only slightly
curled, appearing chiefly at the base of the penis or along the labia.

Considerably darker, coarser, and more curled. The hair spreads sparsely over the
junction of the pubis or mons.

Hair now resembles adult in type, but the area covered by it is still considerably
smaller than in the adult. Still limited to pubis or mons. No spread'to the medial
surface of the thighs.

Adult in quality, quantity and type with distribution in the classically "male" or
"female" pattern, with spread to the medial aspect of the thighs.

Male genitalia

Preadolescent. Testes, scrotum, and penis are about the same size and proportion
as in early childhood.

Enlargement of scrotum and testes. The skin of the scrotum reddens and changes
in texture. There is little or no enlargement of the penis at this stage.

Enlargement of penis (occurs at first mainly in length). There is further growth of

the testes and scrotum.

103

Table 3, cont’d.

 

Stage Description of sexual maturation stage

 

4 Enlargement of penis, with growth in breadth and development of glans. There is
further enlargement of testes and scrotum; there is increased darkening of scrotal
skin.

5 Genitalia is adult-like in size and shape. No further enlargement takes place after
Stage 5 is reached; it seems, on the contrary, that the penis size decreases slightly

from the immediate postadolescent peak.

Female breasts
1 Preadolescent. There is elevation of papilla only.
2 Breast bud stage. Elevation of breast and papilla appear as a small mound. There

is widening and elevation of the areola, with pigmentation.

3 Further enlargement and elevation of breast and areola with no separation of their
contours.

4 Projection of areola and papilla to form a secondary mound above the level of the
breast.

5 Mature stage. Projection of papilla only due to recession of the areola to the

general contour of the breast.

 

SOURCES: Physician Examiner’s Training Manual (DHHS, 1991) and Tanner (1962)

104

averaging the pubic hair rating with the genitalia rating for boys and with the breast
rating for girls (Tanner, 1962). Pubertal status of those missing both SMA ratings was
determined by age (years).

Listed below are the skinfold equations for predicting body fat in prepubescent,
pubescent, and postpubescent white and black male, and all female children and
adolescents age 8-18 years whose summed tricep and subscapular skinfolds were 35 mm
or less (Slaughter et al., 1988).

Prepubescent White Males:

1.21 (triceps + subscapular) - .008 (triceps + subscapular)2 - 1.7

Prepubescent Black Males:

1.21 (triceps + subscapular) - .008 (triceps + subscapular)2 — 3.2

Pubescent White Males:

1.21 (triceps + subscapular) - .008 (triceps + subscapular)2 — 3.4

Pubescent Black Males:

1.21 (triceps + subscapular) - .008 (triceps + subscapular)2 — 5.2

Postpubescent White Males:

1.21 (triceps + subscapular) - .008 (triceps + subscapular)2 — 5.5

Postpubescent Black Males:

1.21 (triceps + subscapular) - .008 (triceps + subscapular)2 — 6.8

All Females: i

1.33 (triceps + subscapular) - .013 (triceps + subscapular)2 — 2.5
The following equations were used for male and female adolescents with a sum of tricep

and subscapular skinfold greater than 35 mm (Slaughter et al., 1988).

105

All Obese Males:

0.783 (triceps + subscapular) + 1.6

All Obese Females:

0.546 (triceps + subscapular) + 9.7

Roche et al., (1981) conducted a validation study among 6-12 and 13-17 year old

boys and girls (n=261) to determine which of the various anthropometric measurements
had the highest correlation with percentage body fat obtained by underwater weighing.
Measurements included triceps and subscapular skinfolds, weight, relative weight, BMI
(kg/m2), and the ponderal index (kg/m3). Compared to body weight measures, triceps
skinfold had the highest correlation with percentage body fat (from underwater weighing)
in all four groups, although the confidence interval of the correlation coefficient
overlapped substantially with that of subscapular skinfold, as well as that of BMI for girls
(Roche etal., 1981). The equations above are based on summed triceps and subscapular
skinfold measurements (Slaughter et al., 1988). These equations were cross-validated on
samples from the literature, and disagreement was found only with samples using the
Lange caliper, rather than the Harpenden skinfold caliper used to develop the equations
(Slaughter et al., 1988). Further cross-validation of the Slaughter skinfold equations for
children and adolescents (J anz et al., 1993) found very high intraclass (reliability)
correlations (ICC = 0.98-0.99), and high validity correlations (rs = 0.79-0.99) compared

to Lohman’s Siri age-adjusted body density equations (Lohman, 1986).

106

E. Statistical analysis

The current dissertation research used data from NHANES 111, 1988-94, in a
cross-sectional analysis to examine beverage patterns and other potential risk factors for
overweight in youth age 12-16 years in the U.S. The Statistical Package for the Social
Sciences, SPSS for Windows, Version 12.0 (SPSS Inc. Chicago, IL, USA) was used to
prepare the dataset for analysis, while all statistical analyses were performed using
SUDAAN version 8 (Research Triangle Institute, Research Triangle Park, NC), to
account for the effect of the complex sampling design while conducting statistical tests.
The six-year MEC-examined sample weights were applied to the calculation of all
population parameters, such as the mean, median, or percentage of the population. The
sample weights account for unequal probability of selection, non-coverage and non-
response and adjust the data to the 1990 U.S. Census figures, therefore the estimates were
representative of adolescents in the 1990 U.S. population. The linearization (Taylor
series) method was used in SUDAAN to account for effects of the clustered sample
design in estimating the variance of population parameters (Shah et al., 1997).

The overall goal of this research was to examine the association of beverage
patterns with adolescent obesity (i.e., overweight and adiposity). The associations of
beverage patterns with nutritional, sociodemographic and lifestyle risk factors for
overweight were also examined. The following were the three (3) specific aims of this

study.

Aim 1: To examine the frequency of consuming beverage types, and the percentage of

youth classified into groups by their ‘typical’ beverage pattern (i.e., the combination of

107

beverage types most frequently reported as ‘usual’ intake using a F F Q); to examine
the association of beverages with meal occasions and meal locations; and to examine
the association of beverage patterns with energy, dietary fat, total sugars, added sugar,
calcium and other nutrient intake.

Four main types of beverages examined included (1) milk and milk drinks, (2)
100% fruit juice (3) sugar-sweetened beverages (fi'uit drinks and regular soft drinks) and
(4) other beverages (diet soft drinks, coffee/tea and alcoholic beverages). Youth were
classified by their ‘typical’ beverage pattern using two different methods to identify the
combination of beverage types (i.e., milk, juice and/or sugar-sweetened beverages) that
were each reported frequently in the F F Q. Criteria to identify the most frequently
consumed beverage types using the ‘frequency’ and ‘share’ methods, respectively, were
frequency 21 time/day, and share 225% of total beverage fi'equency.

Mean energy, dietary fat, total sugars, added sugar, calcium and other nutrient
intake (in the 24-hour recall) of beverage pattern groups were compared to examine the
association of beverage patterns with energy and nutrient intake. The beverage pattern
groups’ mean intake of water (ounces) and mean intakes of milk, 100% fruit juice, sugar-
sweetened beverages, other beverages (e. g., diet soft drinks, coffee/tea and alcoholic
beverages) (times/day), in both the FFQ and 24-hour recall, were also examined. The 24-
hour recall data were used to examine mean beverage fi'equency at meals (e. g., breakfast,
lunch, dinner/supper and snacks) consumed at home or at places away-fiom-home (e.g.,

school, fast food restaurants).

108

a. Frequency of beverage consumption

The FFQ data was used to examine the mean frequency of consuming types of
beverages, and the percentage of youth who frequently consumed each beverage type.
Four (4) main types and 7 subtypes of beverages were examined. The main beverage
types included (1) milk, chocolate and regular, (2) 100% fruit juice, citrus and non-citrus,
(3) sugar-sweetened beverages, and (4) other beverages. Subtypes of (3) sugar-
sweetened beverages were (3a) fruit drinks, and (3b) regular soft drinks, and subtypes of
(4) other beverages were (4a) diet soft drinks, (4b) coffee/ tea, and (4c) beer/ wine/
liquor. Therefore the 7 subtypes of beverages included (1) milk, chocolate and regular,
(2) 100% fruit juice, citrus and non-citrus, (3a) fruit drinks, (3b) regular soft drinks, (4a)
diet soft drinks, (4b) coffee/ tea, and (4c) beer/ wine/ liquor. The FFQ items
corresponding to beverage types (and subtypes) are listed in Table 1.

For each beverage type (and for each subtype of beverages), the mean beverage
frequency was estimated from the frequency of consumption each youth reported in the
FFQ. The mean beverage share was estimated from the beverage’s share of total
consumption determined for each youth by expressing the beverage frequency as a
percentage of the summed frequency for all beverages. Two different methods, the
‘frequency’ method, and the ‘share’ method, were used to classify youth as frequent
beverage consumers. Criteria for frequent beverage consumption using the ‘frequency’
and ‘share’ methods, respectively, were frequency 21 time/day, and share 225% of total
beverage frequency. Therefore, in the ‘frequency’ method, the prevalence of frequent
beverage consumption was defined as the percentage of youth consuming the beverage at

least once per day, and in the ‘share’ method, prevalence of frequent beverage

109

consumption was the percentage of youth having a beverage share that was at least 25%

of the total beverage frequency.

b. Prevalence of beverage patterns

Beverage patterns were defined a priori by identifying all possible combinations
of three types of beverages: milk (milk & milk drinks), sugar—sweetened beverages (fruit
drinks & regular soft drinks) and juice (100% fruit juice). To classify youth into
beverage pattern groups, the FF Q beverage frequencies were used to identify the
combination of beverage types that characterized a youth by his/her ‘typical’ beverage
pattern. Two different methods, the ‘fi'equency’ method, and the ‘share’ method, were
developed to classify youth by the combination of beverage types that were each reported
frequently in the F FQ. For each beverage, the ‘share’ was determined by expressing the
beverage frequency as a percentage of the summed frequency for all beverages. Criteria
to identify the most frequently consumed beverage types using the ‘frequency’ and
‘share’ methods, respectively, were frequency 21 time/day, and share 225% of total
beverage fiequency. Beverage pattern prevalence estimates were the percentage of youth
classified by the combination of beverage types that were each reported frequently, using
the two different criteria of the ‘frequency’ and ‘share’ methods to define frequent
beverage consumption.

Notations for beverage patterns were combinations of capital letters M, S and J to
represent milk, sugar-sweetened beverages and juice, respectively, consumed at a level
meeting the criteria, and lowercase letters m, s and j represented consumption of less than

the specified amount. The beverage patterns were named according to the eight beverage

110

combinations: ‘all caloric beverages’ (MSJ), ‘milk & sugar-sweetened beverages’ (MSj),
‘milk & juice’ (MSJ), ‘milk’ (Msj), ‘sugar-sweetened beverages & juice’ (mSJ), ‘sugar-

sweetened beverages’ (mSj), ‘other beverages & juice’ (msJ) and ‘other beverages’ (msj).

c. Frequency of beverage consumption in beverage pattern groups

Mean beverage fi'equency of consumption in both the F FQ and 24-hour recall
were estimated for beverage pattern groups defined using the ‘share’ method to classify
youth by the combination of beverages that were each reported frequently in the FFQ. To
estimate the 24-hour recall frequency (# times/day) of beverage types and subtypes, the
USDA survey codes for beverages reported in the 24-hour recall dietary interview were
classified into the beverage categories and subcategories listed in Table 1. The 24-hour
recall beverage categories were identical to those described by the FFQ items, except that
the fruit drinks category included fruit drinks without vitamin C, and the coffee/tea
subcategory included decaffeinated coffee and decaffeinated tea. The milk category
included milk poured on cereal as well as milk consumed as a beverage. Milk combined
with cereal was identified using the code for the type of combination known as cereal
with additions. Beverage mixtures, such as coffee with milk, were identified using the
code for this type of combination, and the USDA survey food code for the main
component of the beverage mixture was used to classify the combination into the
appropriate category.

The sample-weighted mean beverage frequency, and standard error of the mean
were estimated using the linearization method of SUDAAN. The 95% confidence

interval of the mean beverage fi'equency was calculated from the standard error adjusted

111

for the design effect using SUDAAN. Because SUDAAN was designed to test only for
significant differences between groups, and paired t-tests for significant differences
between repeated measures (using the FF Q and 24-hour recall) can not be done using
SUDAAN, the 95 % confidence intervals were used to determine whether the mean
beverage frequency in the FFQ and 24-hour recall were significantly different. It was
determined that the mean beverage frequency estimated using the FFQ was not
significantly different from mean beverage fiequency estimated using the 24—hour recall
when the 95% confidence intervals overlap. Paired t-tests conducted using sample-
weighted data in SPSS would have the same result as those tests of the 95% confidence
intervals constructed using SUDAAN, because the sample design would have the same
effect on both measures of mean beverage frequency estimated for the same group
(Herman, personal communication).

The mean beverage frequency of a beverage pattern group was not biased by
intraindividual variation (Guenther et al., 1997). The standard error of the mean was
estimated using SUDAAN to adjust the standard error for the sample design in testing for
significant differences between beverage pattern groups. T-tests for significant
differences in mean beverage frequencies were used to compare beverage pattern groups.
Mean beverage frequency of beverage pattern groups were compared using the
Bonferroni method to adjust statistical significance tests for the number of comparisons.
The Bonferonni method was used to adjust the critical value for determining statistical
significance, as follows: a / k, where or = 0.05 and k = # of comparisons.

Three comparisons of interest were 'caloric beverage' patterns vs. 'other beverage’

patterns, 'milk' patterns vs. 'sugar—sweetened beverage' patterns, and 'juice' patterns vs.

112

'non-juice' patterns. For these comparisons, the 8 beverage patterns were collapsed into
categories, as follows. To compare the ‘caloric beverages’ (K) and ‘other beverages’ (0)
categories, the ‘all caloric beverages’ (MSJ) pattern was combined with the ‘milk &
sugar-sweetened beverages’ (MSj) pattern to form the ‘caloric beverages’ (K) category,
and the ‘other beverages & juice’ (msJ) and ‘other beverages’ (msj) patterns were
combined to form the ‘other beverages’ (0) category. To compare the ‘milk’ (M) and
‘sugar—sweetened beverages’ (S) categories, the ‘milk & juice’ (MSJ) pattern was
combined with the ‘milk’ (Msj) pattern to form the ‘milk’ (M) category, and the ‘sugar-
sweetened beverages & juice’ (mSJ) and ‘sugar—sweetened beverages’ (mSj) patterns
were combined to form the ‘sugar-sweetened beverages’ (S) category. The ‘milk &
juice’ (MSJ) pattern was combined with the ‘sugar—sweetened beverages & juice’ (mSJ)
pattern, and the ‘milk’ (Msj) pattern was combined with the ‘sugar-sweetened beverages’
(mSj) pattern to compare juice (J) and. non-juice (N) categories.

Mean beverage frequency of beverage pattern groups were estimated using both
FFQ and 24-hour recall data to validate the classification of beverage types. It was
expected that the ‘juice’ groups would drink juice more frequently than the ‘non-juice’
groups, and the ‘milk’ groups would drink milk more frequently than ‘sugar-sweetened
beverages’ groups, for example. However, if the ‘sugar-sweetened beverages’ groups
drank fruit drinks more frequently than the ‘milk’ groups in the F FQ, but not in the 24-hr
recall, this might indicate that fruit drinks were misclassified in the FF Q. If the 24-hour
recall intakes of fruit drinks were higher for ‘juice’ groups than ‘non—juice’ groups, then
some misclassification of fruit drink consumers to ‘juice’ groups might have occurred.

The concern is that youth might mistakenly report fi'uit drink fi'equency in response to the

113

100% fruit juice question in the FF Q, because 100% fruit juice is listed before fruit drinks
in the sequence of ordering the FFQ beverage items. Nutritionists who assign food codes
to beverages mentioned in the 24-hour recall interview use probes to solicit specific
details, such as brand names. If the brand name Sunny Delight was given for what the

respondent thought was orange juice, then Sunny Delight would be correctly classified

with other fruit drinks in the 24-hour recall.

d. Frequency of beverage consumption at meals

Mean fi'equency of beverage consumption at meal occasions eaten at-home and
away-from-home were estimated to determine the meal occasions’ and meal locations’
shares of the total beverage frequency. Because the meal occasion and meal location is
not specified in the FFQ, the frequency with which beverages are consumed at specific
meals was determined using 24-hour recall data. Mean 24-hour recall intakes of groups
are not biased by intraindividual variation (Guenther et al.,~1997). Meal occasions
included breakfast/brunch, lunch, dinner/supper, and snack/beverage break. Meal
locations included home, work, school, a fast food or take-out establishment, a restaurant
(with table service) or cafeteria, someone’s home, transit, and other meal places.

In the same way that the contribution of food groups to total nutrient intakes are
estimated fi'om the mean nutrient intakes obtained from each food group (Block et al.,
1985a; Block et al., 1985b), mean frequency of beverage consumption at meal occasions
and at meal locations among consumers of the various types of beverages was estimated
to determine the percentage contribution of meal occasions and meal locations to the total

beverage frequency. The meal occasions’ and meal locations’ shares of the total

114

beverage frequency were estimated to examine the association between meal context and
beverage choice. Hypotheses were that milk or 100% fruit juice would be consumed
more frequently than sweetened beverages at breakfast and other meals eaten at home,
and sweetened beverages would be consumed more frequently than milk or 100% juice at

the snack occasion and at locations away from home.

e. Mean nutrient intake of beverage pattern groups

Beverage pattern groups were defined using the ‘share’ method to classify youth
by the combination of beverages that were each reported frequently in the F F Q, but
because the FF Q in NHANES III was not designed to assess nutrient intake, intakes of
food energy (Kcal), fat (g), sugars (g), caffeine (mg), vitamin A (pg ME), vitamin C
(mg), folate (pg DFE), calcium (mg), potassium (mg) and vitamin D (pg) were assessed
using the 24-hour recall dietary intake data instead. The 24-hour recall intakes of the
beverage pattern groups were used to estimate mean daily nutrient intake; mean nutrient
intake from beverages and from non—beverages; and mean percentage energy from fat,
total sugars, and added sugars. Mean 24-hour recall intake of a beverage pattern group
was not biased by intraindividual variation (Guenther et al., 1997).

The standard error of the mean was estimated using SUDAAN to adjust the
standard error for the sample design in testing for significant differences between
beverage pattern groups. T-tests for significant differences in mean nutrient intake were
used to compare beverage pattern groups. Mean nutrient intakes of beverage pattern
groups were compared using the Bonferroni method to adjust statistical significance tests

for the number of comparisons. The Bonferonni method was used to adjust the critical

115

value for determining statistical significance, as follows: a / k, where or = 0.05 and k = #
of comparisons.

Three comparisons of interest were 'caloric beverage' patterns vs. 'other beverage'
patterns, 'milk' patterns vs. 'sugar-sweetened beverage' patterns, and 'juice' patterns vs.
'non-juice' patterns. For these comparisons, the 8 beverage patterns were collapsed into
categories, as follows. To compare the ‘caloric beverages’ (K) and ‘other beverages’ (0)
categories, the ‘all caloric beverages’ (MSJ) pattern was combined with the ‘milk &
sugar-sweetened beverages’ (MSj) pattern to form the ‘caloric beverages’ (K) category,
and the ‘other beverages & juice’ (msJ) and ‘other beverages’ (msj) patterns were
combined to form the ‘other beverages’ (0) category. To compare the ‘milk’ (M) and
‘sugar-sweetened beverages’ (S) categories, the ‘milk & juice’ (MsJ) pattern was
combined with the ‘milk’ (Msj) pattern to form the ‘milk’ (M) category, and the ‘sugar-
sweetened beverages & juice’ (mSJ) and ‘sugar-sweetened beverages’ (mSj) patterns
were combined to form the ‘sugar—sweetened beverages’ (S) category. The ‘milk &
juice’ (MsJ) pattern was combined with the ‘sugar-sweetened beverages & juice’ (mSJ)
pattern, and the ‘milk’ (Msj) pattern was combined with the ‘sugar-sweetened beverages’
(mSj) pattern to compare juice (J) and. non-juice (N) categories.

Mean and median nutrient intake of boys/ girls age 12-13 and 14-16 years were
estimated to distinguish gender-age groups having different DRI levels. The mean and
median of the distribution of daily nutrient intake were compared to Dietary Reference
Intake (DRI) criteria, such as the age- and gender-specific Estimated Average

Requirement (EAR) or Adequate Intake (AI) level shown in Table 2 (Institute of

116

Medicine (IOM) of the National Academies of Science, 1997; IOM, 1998; IOM, 2000;

IOM, 2001; IOM, 2002).

Aim 2: To examine the distributions of sociodemographic and lifestyle characteristics
that are risk factors for overweight among beverage pattern groups defined above in
Aim 1.

Youth were classified into groups by their ‘typical’ beverage pattern using the
‘share’ method (i.e., share 225% total beverage frequency) to identify the combination of
beverage types (e. g., milk, 100% mu juice and sugar-sweetened beverages) most
frequently reported as ‘usual’ intake in the F F Q. To identify risk factors for overweight,
the prevalence of overweight/risk for overweight (i.e., BMI 285th percentile of BMI-for-
age) was estimated for youth classified by each sociodemographic/lifestyle characteristic.
To examine the distributions of risk factors for overweight, the percentage of each
beverage pattern group classified by each sociodemographic/lifestyle characteristic was
estimated, and the percentage of each beverage pattern group classified by weight status
(e.g., underweight, normal, at risk for overweight, overweight) was estimated as well.

Sociodemographic factors examined were gender, age, race-ethnicity, region,
urbanization, parental overweight/obesity, household income and other indicators of
social economic status (SES). Lifestyle factors included smoking, alcohol use, weight
loss behavior, physical activity (i.e., exercise and team sports participation), TV
watching, frequency of breakfast consumption, and type of milk (e. g., whole, 2%, 1%,

skim) usually consumed.

117

Categories for race-ethnicity included non-Hispanic white, non-Hispanic black,
Hispanic and other race-ethnic groups. Four broad regions of the U.S. included the
Northeast, Midwest, South and West. The urban/rural classification was a metropolitan
area having a population 21 million vs. all other areas. The poverty income ratio was the
ratio of household income to the poverty level for household size in an urban or rural
area. Other indicators of SES included the family size and characteristics of the
household head or family reference person (FRP), such as their gender, age group,
marital status, education and employment status. BMI was calculated from reported
weight and height of the youth’s mother and father to determine whether both or one of
two parents were overweight (BMI 225 kymz) or obese (BMI 230 kg/mz).

The survey included lifestyle questions, such as those asking youth whether they
currently smoked cigarettes or drank alcohol in the past year. Youth were also asked
whether they were currently trying to lose weight and how often they ate breakfast (e.g.,
every day, some days, rarely, never, weekends only). The FFQ included a question
asking whether whole, 2%, 1% or skim/nonfat milk was usually consumed. Physical
activity was assessed using the following questionnaire items asking youth about their
exercise habits and participation in team sports:

1. How many times per week do you play or exercise enough to make you sweat

or breathe hard?

2. In the past year, how many sports teams or organized exercise programs have

you been involved in (not including physical education or gym classes)?

118

Youth were asked twice, once in the household and again at the MEC, about time spent
watching TV the previous day, and the average hours/day in two days was used in the
current study.

The prevalence of overweight/risk for overweight (BMI 285th percentile) was
estimated for youth classified by categories of each sociodemographic/lifestyle factor,
and the percentage distribution of each sociodemographic/lifestyle factor was estimated
for all youth and for each beverage pattern group. The percentage of each beverage
pattern group classified by weight status (e. g., underweight, normal, at risk for
overweight, overweight) was estimated as well. The Chi-square test for the association
of categorical variables was performed to determine whether the observed and expected

distributions were different (p<0.05).

Aim 3: To weamine the association of adolescent obesity with the beverage pattern
groups defined above in Aim I, adjusted for potential covariates of interest.

Indicators of obesity (i.e., overweight and adiposity) examined separately
included BMI percentile, percentage body fat as assessed by skinfold measurements, and
overweight/risk for overweight (i.e., BMI 285th percentile of BMI-for—age). Youth were
classified into groups by their ‘typical’ beverage pattern using the ‘share’ method (i.e.,
share 225% total beverage frequency) to identify the combination of beverage types (e.g.,
milk, sugar-sweetened beverages and/or 100% fruit juice) most frequently reported as
‘usual’ intake in the FFQ. In addition to beverage pattern groups, other beverage
variables considered in analyses were fluid ounces of water and type of milk (e. g., whole,

2%, 1%, skim) usually consumed. Covariates considered in analyses were gender, age,

119

race-ethnicity, sexual maturation stage, region, urbanization, parental overweight/
obesity, SES indicators, smoking, alcohol use, weight loss behavior, exercise, team sports
participation and TV watching. Covariates were examined separately one at a time to
investigate whether each one might potentially confound the relationship between
beverage patterns and overweight/ adiposity.

The association of obesity outcomes with beverage pattern groups was examined
after controlling for covariates using regression analysis. Multiple regression analysis
was used to implement a multiple analysis of variance (MAN OVA) with categorical
variables (i.e., dummy variables) in a general linear model. The p-value of the Wald F
test for the treatment effect was interpreted to test for a significant association between
any one of the beverage pattern groups and obesity. Because the p-value to test whether
the effect of a particular beverage pattern group was significant indicates only whether
the beta coefficient was different from zero, Bonferroni contrasts of least square (LS)
means were performed to indicate whether the ‘milk & juice’ beverage pattern group was
significantly different from each of the other seven beverage pattern groups. LS mean
BMI percentile and LS mean body fat (%) were adjusted for effects of covariates. The
95% confidence intervals for odds ratios were calculated and interpreted to indicate
whether a beverage pattern group’s risks were significantly different from 1.0. Odds
ratios for overweight/risk for overweight were adjusted for effects of covariates as well.

In addition to examining the independent effects on obesity of beverage patterns,
the independent effects of sexual maturation, and confounding variables were also
examined. A true confounder must have independent effects on both beverage patterns

and obesity, and it might not be necessary to include a potential confounder in the final

120

model if it is not a true confounder. For each potential confounder, it was first
determined whether it had an effect on obesity by including it alone in the regression
model. Beverage patterns were then added to the model.

Each covariate was added individually to the model that included only the
beverage pattern, amount of water and type of milk usually consumed (Model 1).
Covariates were examined separately one at a time to investigate whether each one might
potentially confound the relationship between beverage patterns and overweight]
adiposity. If the effect of beverage patterns was no longer significant with the inclusion
of the potential confounder in the model then it was concluded that this covariate was
indeed confounding the relationship between beverage patterns and overweight/adiposity.
Covariates not found to be confounders were added to Model 1 together to form an
adjusted model (Model 2), and the remaining covariates were each added individually
one at a time to the adjusted model (Model 2).

Whether an effect of TV watching on obesity was mediated through both
beverage patterns and physical activity was investigated. This was done by first
determining whether TV watching has an effect on obesity without including physical
activity in the model. The effect of physical activity was also determined without
including TV watching in the model. Then, both TV watching and physical activity were
included in the model together, to examine whether the effect of TV watching on obesity
was associated with the level of physical activity. If the effect of physical activity was
diminished, then it was concluded that TV watching explained the level of physical
activity. If TV watching was still significant after including physical activity in the

model then it was concluded that TV watching has an effect independent of physical

121

activity. An independent effect of TV watching after controlling for physical activity
might be due to an association between TV watching and beverage patterns. It was then
determined whether an effect of beverage patterns is diminished when TV watching is
included in the model. This would indicate that the effect of TV watching on obesity was
partially mediated through beverage patterns. Whether TV was significant after
including both beverage patterns and physical activity in the model would indicate
whether the effect of TV watching on obesity was mediated fully through beverage
patterns and physical activity. If that were the case, it might be ‘over-controlling’ the
model if TV watching were included.

Whether physical activity is an effect modifier was also investigated. Physical
activity could potentially affect both beverage patterns and obesity. Youth who exercise
frequently would require beverages for hydration, and may choose to drink sports
beverages, which are classified with sugar-sweetened beverages. Consumption of sports
beverages may not necessarily be associated with body fatness, because youth who

exercise frequently would have more lean than fat mass.

F. Modeling the relationship between beverage patterns and obesity

Results of this study provide much needed data regarding interrelationships
between sociodemographic/ lifestyle factors, beverage patterns, dietary intakes, and
adolescent obesity in the U.S. The following description of aims is to show how they
relate to the conceptual model shown in Figure 1.

As shown in Figure 1, energy intake is hypothetically the mediator between

beverage patterns and adolescent obesity, and Aim 1 was to investigate whether beverage

122

banana 5? 2:03am unseat:— ue .8338me of .5.“ Eve:— _a:Eoo:eU A 0.5!...—

 

 

8.8

.2.

 

 

u

f—fl

\

//

25%—Sawm—

s

 

«

3.5.5— A - I I»...

 

 

23:8.
.eeaEm

 

 

 

 

 

AV

 

 

3.35 .90....“—

 

 

 

 

a£>a=oa 8.: Emma?
.383. .uumu—efim
~55 magnum
a2— mflue mam .0885 mm:
:«Augauam 639%
£223. 4.659 an...
mug—gum

 

Fatah
owaeogm

 

wages; E

 

 

 

 

 

 

 

 

 

e833 .8:
:33on :62

 

 

 

 

 

 

123

.835 $85 53, £2.8th owm8>on mo 833088 of 8885 8 98 A8382 308
28 5886 E25 “8:80 :88 85 £5 mowfigon .«o 223883 85 8.588 8 me? ~ 83.

 

8.3:— magnum

 

,4

 

 

 

 

Eaten
engagem—

 

 

.Pu=3 J charm

 

. net—39H .32
Ema-39¢ =32

 

 

 

124

inseam Emma owmeo>on 32.85 05 muons bucoeotmp @025ng 203 £32,330
pom 890$ v3.2 233.:— pSw oEdmuonopomoOm 05 350:? oEEeBop 8 33 N 5?.

6.33 .— 95w?—

 

 

 

 

 

 

 

  

 

 

Serum
0335m—

 

 

 

.S_>«._on 9.8— 3385
3.28:. age—cam
Em 939:5
can: mm .«o mam .0585 mm
:snugfism £33m
9:35—3— :o—Eow due.
Ewen—.850 15:85

 

 

 

 

ass. 3585 W

 

 

unseat >9

125

.8888 90 888.58 38809

89 @8888 $8890 53> 8.8th 89889 90 82888.8 85 88888 8 mm? m 83..

 

 

 

 

 

6.88 .— 95w:—

 

 

 

9.9.53.
83939 1 . «8:883 >9.

 

 

 

 

 

 

 

4‘

 

 

 

F8389
owaugom

 

 

 

.5352. 83 Emma?
.38.? .8228
En 328.39
88: 9:99: mam .8885 9:9
5.195859 .8339
£225....— 585 .84.

 

 

126

patterns (i.e., the types of beverages most frequently consumed) were associated with
energy intake by examining the mean energy intake of beverage pattern groups. The
associations of the meal context (e.g., meal occasion and meal location) with beverages
were also examined in Aim 1. The beverage pattern is hypothetically determined by the
meal context in which beverages are consumed, and the beverage pattern can be a marker
for the lifestyle variables that detemiine the frequency with which that meal context
occurs for youth.

Aim 2 was to investigate the association of beverage patterns with
sociodemographic and lifestyle risk factors for overweight by examining the distribution
of these factors among beverage pattern groups. Chi-square tests of association between
beverage categorical variables are based on the differences between the observed and
expected distributions. The sociodemographic and lifestyle factors were examined to
investigate whether and how they affected the classification of youth to the various
beverage pattern groups. Sociodemographic and lifestyle factors examined included
those listed in Figure 1 as potential confounders of the relationship between beverage
patterns and obesity, as well as exercise, team sports participation, TV watching and the
frequency of breakfast consumption.

Aim 3 was to examine the association of beverage patterns with adolescent
obesity. The beverage pattern groups were used as “dummy" variables in regression
analyses. Obesity outcomes included as dependent variables in linear regression models
were the BMI percentile and percentage body fat. The dependent variable in a logistic
regression model was overweight/risk for overweight defined by the 85th percentile of the

reference BMI-for-age distribution (Kuczmarski et al., 2000). The associations of obesity

127

outcomes with beverage pattern groups were adjusted for physical activity (i.e., exercise
and team sports), TV watching, sexual maturation, and potential confounders. As shown
in Figure 1, not only the independent effects on obesity of beverage patterns, but also the
independent effects on obesity of TV watching, physical activity, sexual maturation and
confounding variables were determined. Regression models were used to investigate
whether an effect of TV watching on obesity was mediated through both beverage
patterns and physical activity. Whether potential confounders led to a spurious
correlation between beverage patterns and obesity was also examined. The relationships
of beverage patterns with sociodemographic and lifestyle factors examined in Aim 2
provided insight regarding the covariates that affected the association of beverage

patterns with obesity outcomes in Aim 3.

G. Strengths and Limitations

Strengths of an analysis of NHANES III data are that a multitude of health and
nutrition data are provided for a large national sample, and results are representative of
the U.S. population. Using NHANES, measured height and weight, not self-reported data
as in CSFII, were used to compute BMI. An advantage of NHANES is that there is not
just BMI, but a variety of anthropometric data available, including skinfold thickness
measures used to compute percentage body fat in adolescents.

A comprehensive food frequency questionnaire available only in NHANES III
was used to assess usual beverage consumption, but the FFQ was not validated or
quantified to estimate nutn'ent intakes. Use of the FFQ increases the precision of

population parameter estimates by reducing the potential for error due to intraindividual

128

in dietary intake. In classification of youth by beverage patterns, the frequencies of each
beverage consumed were converted to percentages of the total beverage frequency to
address the potential for bias due to under- or over-reporting. Classification of youth into
mutually exclusive beverage pattern groups addressed the potential for bias due to
multicollinearity among the single beverage items. Classification of youth into beverage
pattern groups also allowed for the use of the 24-hour recall data, because 24-hour recall
intakes of groups are not biased by intraindividual variation (Guenther et al., 1997).

The most important limitation of this study is that cross-sectional data lack
temporality, and cannot be used to make causal inferences. Interpretation of results of
this study would be difficult when an unexpected finding can be explained by reverse
causality. For example, if low calorie “diet” soft drinks were positively associated with
body weight, there would be two possible explanations to consider, i.e., “diet” soft drinks
promote weight gain (contradictory to what was expected), or overweight youth drink
“diet’ beverages because they are trying to lose weight (reverse causality). Cross-
sectional data cannot be used to determine which of the two explanations would be

responsible for a positive association between “diet’ soft drinks and obesity.

129

Chapter 4

THE MEAL CONTEXT OF BEVERAGES CON SUMED BY ADOLESCENTS IN

THE U.S., BEVERAGE CONSUMPTION PATTERNS AND ENERGY INTAKE
ABSTRACT
Objective To examine the meal context in which beverages were consumed; and to
examine associations of beverage patterns (i.e., the types of beverages most frequently
consumed) with energy and nutrient intake.
Design Cross-sectional analysis of data from the Third National Health and Nutrition
Examination Survey (NHANES IH, 1988-94). Both food frequency questionnaire (F PO)
and 24-hour recall data were used to examine how often beverages were consumed. The
meal context of beverage intake was examined using 24-hour recall data, and youth were
classified by beverage pattern using FFQ data. Energy and nutrient intakes from the total
diet, beverages and non-beverages were assessed using 24-hr recall data.
Subjects A U.S. population-based sample of adolescents age 12-16 years (n=1,872).
Statistical analysis Sample-weighted means, medians and standard errors were
estimated using SUDAAN software. Descriptive statistics included the mean frequency
of consuming each beverage type, and the percentage contribution of meal occasions and
meal locations to the total beverage frequency. The Bonferroni method was used to test
for differences in mean energy and nutrient intake of beverage pattern groups (p<0.05).
Results Types of beverages consumed by adolescents, and their percentage share of the
total beverage frequency, included milk (39%), sugar-sweetened fruit drinks/soft drinks
(30%), juice (20%), and other beverages (e.g., diet soft drinks, coffee and tea) (11%).

Milk and juice were consumed most frequently at the breakfast meal, while sugar-

130

sweetened beverages were consumed most frequently at the snack occasion. Soft drinks
were consumed at fast food restaurants more frequently than any other beverage.
Comparisons of ‘milk’ and ‘sugar-sweetened beverages’ patterns showed that fat intake
from beverages was greater for ‘milk’ groups, and added sugar intake from beverages
was greater for ‘sugar-sweetened beverages’ groups. Daily fat and energy intakes from
the total diet did not differ, but calcium, potassium and vitamin D intakes were higher in
‘milk’ groups, and percent calories from added sugars were higher in ‘sugar—sweetened
beverages’ groups. Daily energy, calcitun and vitamin D intakes were lower, and caffeine
intakes were higher in ‘other beverages’ groups than in ‘caloric beverages’ groups.
Implications Because beverage choice was associated with the meal context, the
beverage pattern could be used to indicate how often youth consume different types of
meals (e.g., breakfast, snacks, meals eaten at fast food restaurants). Higher added sugar
and lower calcium intakes in ‘sugar-sweetened beverages’ groups than in ‘milk’ groups
suggest that the beverage pattern could potentially be a risk factor for overweight.
Consumption of ‘other beverages’ patterns might put youth at risk for inadequate

calcium, potassium or vitamin D intakes.

INTRODUCTION

Trends in national surveys have shown a dramatic increase in the intakes of soft drinks by
youth in the U.S. (Cavadini et al., 2000a; Bowman, 2002; Enns et al., 2002; Enns et al.,
2003; French et al., 2003; Nielsen & Popkin, 2004). Some authors have noted that the
trend in obesity prevalence parallels the trend in added sugars available in the food

supply (Hamack et al., 2000; Bray et al., 2004). If the added sugars in soft drinks and

131

other sugar-sweetened beverages consumed by youth contribute an excess of energy
intake compared to energy expenditure, the positive energy balance could lead to
incremental weight gain in adolescence. Soft drinks are frequently consumed while
snacking and watching TV (Coon et al., 2001; Giammattei et al., 2003; Matheson et al.,
2004; Phillips et al., 2004), and with meals eaten at fast food restaurants (French et al.,
2001b; Paeratakul et al., 2003; Bowman et al., 2004). Caloric intakes from energy-dense
snack foods (Hampl et al., 2003; Phillips et al., 2004) and fast foods (Bowman et al.,
2004; Paeratakul et al., 2003; Schmidt et al., 2005) are then added to calories contributed
by soft drinks (Hamack et al., 1999; Troiano et al., 2000). Thus, youth who drink soft
drinks are expected to have higher energy intakes than youth who do not.

Intakes of soft drinks increase with age, while milk intakes decline during
adolescence, especially in girls (Lytle et al., 2000; Bowman et al., 2002). This inverse
relationship between intakes of soft drinks and milk might have negative implications for
nutritional status (Hamack et al., 1999). Girls might reduce intakes of milk during
adolescence because they think milk is fattening (Clark, 1999; AAP, 1999), but if they
drank low-fat milk, the calcium in it might be beneficial for weight control (Zemel et al.,
2000). The decline in milk intake might relate to the increased prevalence of breakfast
skipping in adolescence (Siega-Riz et al., 1998), an additional negative influence on
nutritional status (Rampersaud et al., 2005). Youth consume milk more often if they had
cereal with milk for breakfast (Ortega et al., 1998).

If youth consumed both milk and orange juice frequently at breakfast, then those
youth who regularly eat breakfast might be distinguished from breakfast skippers by

identifying those who do or do not consume milk and juice in combination. In addition

132

to milk and juice, cereal and other foods consumed at the breakfast meal occasion would
i then contribute to vitamin and mineral intakes of milk and juice consumers. Milk is high
in vitamins and minerals like vitamin A, calcium and vitamin D (Miller et al., 2001);
juice is high in vitamin C, folate and potassium (Fellers et al., 1990); and cereal is
fortified with many vitamins and minerals (Subar et al., 1998).

l’revious research examined nutrient intakes of youth who did or did not drink
sugared-sweetened beverages and milk in combination (Bowman et al., 2002). The
current study of beverage patterns also considered the independent nutritional effects of
drinking juice. A milk and juice pattern was expected to be associated with a nutrient-
dense diet, if milk and juice were frequently consumed at breakfast. Sugar-sweetened
beverage patterns were expected to be associated with energy intake from both beverages
and non-beverages, if energy-dense snack foods or fast foods were eaten frequently with
sugar-sweetened beverages.

The aims of this study were to examine the meal context in which 12-16-year-old
adolescents in the U.S. consumed beverages; and to examine associations of beverage
patterns (i.e., the types of beverages most frequently consumed) with energy and nutrient
intake. How often youth consumed milk, 100% fruit juice, sugar-sweetened fruit
drinks/soft drinks, and other beverages (e. g., diet soft drinks, coffee/tea and alcoholic
beverages) (times/day) at meals (e. g., breakfast, lunch, dinner/supper and snacks) eaten at
home or at places away-from-home (e. g., school, fast food restaurants) were examined.
Youth were classified by their ‘typical’ beverage pattern according to the combination of

beverage types that were each reported frequently as ‘usual’ intake in the FFQ. Mean

133

food energy, dietary fat, total sugars, added sugar, calcium and other nutrient intakes of

beverage pattern groups were compared.

METHODS

Dataset

The Third National Health and Nutrition Examination Survey (NHANES HI, 1988-1994)
data were used for analyses, rather than the more recent 1999-2002 NHANES data,
because the dietary food frequency questionnaire used to assess the youth’s usual
beverage pattern is available only in NHANES IH. The National Center for Health
Statistics (N CHS) of the Centers for Disease Control and Prevention (CDC) conducts
NHANES to obtain nationally representative data on the nutritional and health status of
the civilian, non-institutionalized U.S. population. Participants were selected using a
complex, stratified, multistage, probability cluster sample design (DHHS, 1994). The
survey included interviews conducted in the household and at. a mobile examination

center (MEC).

Subjects

Adolescents age 12-16 years who were interviewed and examined in NHANES III
comprised the analytic sample for this study. The NHANES III sample included 2,216
adolescents aged 12-16 years selected for examination, and, of those, 2,079 (93.8% of
2,216) were examined in the MEC. Upon examination, 19 respondents (0.9% of 2,216)
found to be ineligible (l 6 teenage girls were pregnant, and three were lactating) were

excluded from the current study. Because the current study used both 24-hour recall and

134

F FQ data, the analytical sample was limited to adolescents age 12-16 years who
completed a 24—hour recall dietary interview as well as all beverage items in the FFQ.
There were 1,997 youth (90.1% of 2,216) who had a complete, reliable 24—hour recall
dietary interview, and of these, 97 (4.4% of 2,216) were missing beverage frequency data
and another 28 (1.3% of 2,216) were missing data on water intake. After excluding those
with missing data, the remaining analytic sample of 1,872 youth (84.5% of 2,216)

included 871 boys (46.5% of 1,872) and 1,001 girls (53.5% of 1,872) aged 12-16 years.

Dietary assessment methods
The FFQ and 24-hour recall dietary interviews were administered to youth at the MEC.

Food frequency questionnaire

Twelve (12) beverage items in the FFQ are listed in Table l. The subjects were
asked to think about their usual diet over the past month, and to respond with how often
they usually drank the beverages per day, per week, per month, or not at all. NHANES
staff converted the frequency data to times per month, by multiplying times per day by a
factor of 30.4, for example: 3 times per day = round (3 * 30.4) times per month = 91
times per month. Amounts reported per week were divided by 7 and then multiplied by
30.4 to convert to times per month: 5 times per week = round (5/7 * 30.4) times per
month = 22 times per month. In the current study, beverage frequency (times per month)
divided by 30.4 converted the data to provide intakes in terms of times per day.

Four main categories (and seven subcategories) of the 12 FF Q beverage items
listed in Table 1 included (1) milk and milk drinks, chocolate and regular, (2) 100% fruit

juice, citrus and non-citrus, (3) sugar-sweetened beverages (3a. fruit drinks, 3b. regular

135

Table 1. Beverage categories and beverage subcategories corresponding to beverage

items in the NHANES III food frequency questionnaire (F FQ)

 

Beverage categories &

beverage subcategories

FFQ beverage items

 

1. Milk and milk drinks,

chocolate & regular

2. 100% fi'uit juice,

citrus & non-citrus

3. Sugar-sweetened bevg.
3a. Fruit drinks
3b. Regular soft drinks
4. Other beverages
4a. Diet soft drinks

4b. Coffee/ tea

4c. Alcoholic

beverages

1a. How often did you have chocolate milk and hot cocoa?

1b. How often did you have milk to drink or on cereal?

Do not count small amounts of milk added to coffee or tea.

2a. How often did you have orange juice, grapefruit juice
and tangerine juice?

2b. Other fi'uit juices such as grape juice, apple juice,
cranberry juice, and fruit nectars?

3a. How often did you have Hi-C, Tang, Hawaiian Punch,
Koolaid, and other drinks with added vitamin C?

3b. Regular colas and sodas, not diet?

4a. Diet colas, diet sodas, and diet drinks such as Crystal
Light?

4b (1). Regular coffee with caffeine?

4b (2). Regular tea with caffeine?

4c (1). Beer and lite beer?

4c (2). Wine, wine coolers, Sangria, and champagne?

4c (3). Hard liquor such as tequila, gin, vodka, scotch,

rum, whiskey and liqueurs, either alone or mixed?

 

136

soft drinks) and (4) other beverages (4a. diet soft drinks, 4b. coffee/tea and 4c. beer/
wine/liquor). Regular milk included both milk to drink and milk on cereal. An
additional item in the F FQ asked what type of milk was usually consumed: whole, 2%,
1% or skim/nonfat milk.

Beverage patterns were defined a priori by identifying all possible combinations
of three types of beverages: milk (milk & milk drinks), sugar-sweetened beverages (fruit
drinks & regular soft drinks) and juice (100% mu juice). To classify youth into
beverage pattern groups, the FF Q beverage frequencies were used to identify the
combination of beverage types that characterized a youth by his/her ‘typical’ beverage
pattern. Two different methods, the ‘frequency’ method, and the ‘share’ method, were
developed to classify youth by the combination of beverage types that were each reported
frequently in the FF Q. For each beverage, the ‘share’ was determined by expressing the
beverage frequency as a percentage of the summed frequency for all beverages. Criteria
to identify the most frequently consumed beverage types using the ‘frequency’ and
‘share’ methods, respectively, were frequency 21 time/day, and share 225% of total
beverage frequency. Notations for beverage patterns were combinations of capital letters
M, S and J to represent milk, sugar-sweetened beverages and juice, respectively,
consumed at a level meeting the criteria, and lowercase letters m, s and j represented
consumption of less than the specified amount. The beverage patterns were named
according to the eight beverage combinations: ‘all caloric beverages’ (MSJ), ‘milk &
sugar-sweetened beverages’ (MSj), ‘milk & juice’ (MSJ), ‘milk’ (Msj), ‘sugar—sweetened
beverages & juice’ (mSJ), ‘sugar-sweetened beverages’ (mSj), ‘other beverages & juice’

(msJ) and ‘other beverages’ (msj).

137

24-hour recall dietary interview

The 24-hour recall data were used to examine beverages consumed at meal
occasions eaten at locations at-home and away-from home, because the FFQ does not
specify meal occasion and meal location. Respondents reported the meal occasion (e. g.,
breakfast/brunch, lunch, diner/supper, snack/beverage break) and place where the meal
was eaten, and Spanish translations were available as needed. Meal locations included
home, work, school, a fast food or take-out establishment, a restaurant (with table
service) or cafeteria, someone’s home, transit, and other meal places.

The USDA survey codes for beverages reported in the 24-hour recall dietary
interview were classified into the beverage categories and subcategories listed in Table 1.
The 24-hour recall beverage categories were identical to those described by the FFQ
items, except that the fi'uit drinks category included fi'uit drinks without vitamin C, and
the coffee/tea subcategory included decaffeinated coffee and decaffeinated tea. The milk
category included milk poured on cereal as well as milk consumed as a beverage. Milk
combined with cereal was identified using the code for the type of combination known as
cereal with additions. Beverage mixtures, such as coffee with milk, were identified using
the code for this type of combination, and the USDA survey food code for the main
component of the beverage mixture was used to classify the combination into the
appropriate category.

Assessment of water intakfi

Although water intakes were not ascertained using the 24-hour recall method, the
dietary interview included a question to respondents about how many glasses/cups or

fluid ounces of plain water they usually drank in a 24-hour period. If the response was in

138

terms of glasses or cups, participants specified the fluid ounces per glass or cup, and
numbers of glasses/cups were converted to fluid ounces.

Nutrient intake assessment

Although youth were classified by their ‘typical’ beverage pattern based on
beverage frequencies reported in response to the F F O, because the NHANES HI F FQ was
not designed to assess nutrient intake, the 24-hour recall method was used instead to
estimate mean nutrient intakes of beverage pattern groups. Intakes of food energy (Kcal),
fat (g), sugars (g), caffeine (mg), vitamin A (pg RAE), vitamin C (mg), folate (pg DFE),
calcium (mg), potassium (mg) and vitamin D (pg) from the total diet, from beverages,
and from non-beverages were assessed using the 24-hour recall dietary intake data.

Because of the risk for overweight, intakes of food energy (Kcal) and
macronutrients such as total fat (g), saturated fat (g), carbohydrate (g), total sugars (g)
and added sugars (g) are primarily of interest. It’s recommended that the diet contain
<30% calories from total fat, and <10% calories from saturated fat or added sugars.
Intakes of caffeine (mg) from soft drinks, coffee and tea might also be a concern for
teens. Teenagers, especially girls, are at risk for inadequate intake of some key nutrients.
Many of these vitamins and minerals can be obtained fi'om the diet by drinking juice or
eating cereal with milk for breakfast. Thus, additional nutritional risk factors that might
distinguish the ‘milk & juice’ (MSJ) group from other beverage pattern groups include
intakes of vitamin A (pg RAE), vitamin C (mg), folate (pg DFE), calcium (mg),
potassium (mg) and vitamin D (pg). The mean and median of the distribution of daily
nutrient intake were compared to Dietary Reference Intake (DRI) criteria, such as the

age- and gender-specific Estimated Average Requirement (EAR) or Adequate Intake (A1)

139

level (Institute of Medicine (IOM) of the National Academies of Science, 1997; IOM,
1998; IOM, 2000; IOM, 2001; IOM, 2002).

The primary source of food composition data for NHANES III was the U.S.
Department of Agriculture (USDA) Survey Nutrient Database; USDA provided two
nutrient files (first in 1993, second in 1995) for use in NHANES III (NCHS, 1997).
Because the Nutrition Coordinating Center (NCC) of the University of Minnesota had
developed the Dietary Data Collection system for NHANES III, survey foods were also
cross-referenced to the NCC nutrient composition files (NCHS, 1998). Although vitamin
D was not in the USDA Survey Nutrient Database, the second release of the NHANES III
24-hour recall dietary data included vitamin D intakes, because vitamin D was in the
NCC nutrient composition database (NCHS, 1998). The USDA Food and Nutrition
Database for Dietary Studies (FNDDS) (USDA, 2004) was used to access food
composition data for vitamin A (pg RAE), folate (pg DFE), caffeine and total sugars,
because intakes are either unavailable in the NHANES HI dataset, or if available, units
are inconsistent with DRI recommendations.

The added sugars content of food was obtained from the USDA Pyramid Food
Group Servings Database, Version 2.0 (Cook & Friday, 2004). Added sugars include
sugars and other sweeteners eaten separately or added to foods at the table as well as
sweeteners used as ingredients in prepared and processed foods such as breads, cakes and
other grain desserts, soft drinks, jams and jellies, candies and ice cream. Added sugars
and other sweeteners include white sugar, brown sugar, raw sugar, corn syrup, corn syrup
solids, high fructose corn syrup, malt syrup, maple syrup, pancake syrup, fructose

sweetener, liquid fi'uctose, honey, molasses, anhydrous dextrose, crystal dextrose, and

140

dextrin eaten separately or used as ingredients in prepared or processed food. Added
sugars do not include naturally occurring sugars such as lactose in milk and fructose in
fruit. In the USDA Pyramid Servings database, a serving of added sugars was defined as

the amount equivalent to 4.2 grams of sugar.

Statistical analysis

The Statistical Package for the Social Sciences, SPSS for Windows, Version 12.0 (SPSS
Inc. Chicago, IL, USA) was used to prepare the dataset for analysis, while all statistical
analyses were performed using SUDAAN version 8 (Research Triangle Institute,
Research Triangle Park, NC), to account for the effect of the complex sampling design
while conducting statistical tests.

For each beverage type (and for each subtype of beverages), the mean beverage
frequency was estimated fiom the frequency of consumption each youth reported in the
FF Q. The mean beverage share was estimated from the beverage’s share of total
consumption determined for each youth by expressing the beverage frequency as a
percentage of the summed frequency for all beverages. Two different methods, the
‘frequency’ method, and the ‘share’ method, were used to classify youth as frequent
beverage consumers. Criteria for frequent beverage consumption using the ‘frequency’
and ‘share’ methods, respectively, were fi'equency 21 time/day, and share 2.25% of total
beverage frequency. Therefore, in the ‘frequency’ method, the prevalence of fi'equent
beverage consumption was defined as the percentage of youth consuming the beverage at
least once per day, and in the ‘share’ method, prevalence of frequent beverage

consumption was the percentage of youth having shares that were least 25% of the total

141

beverage fi'equency. Beverage pattern prevalence estimates were the percentage of youth
classified by the combination of beverage types that were each reported frequently, using
the two different criteria of the ‘frequency’ and ‘share’ methods to define frequent
beverage consumption.

Mean beverage frequency and mean nutrient intakes of beverage pattern groups
were contrasted using the Bonferroni method to adjust statistical significance tests for the
number of comparisons. Three Bonferroni contrasts of interest were 'caloric beverage'
patterns vs. 'other beverage' patterns, 'milk' patterns vs. ’sugar-sweetened beverage'
patterns, and 'juice' patterns vs. 'non-juice' patterns. Mean and median nutrient intake of
boys/ girls age 12-13 and 14-16 years were estimated to distinguish gender-age groups
having different DRI levels. Mean frequency of beverage consumption at meal occasions
eaten at-home and away-from-home were estimated to determine the meal occasions’ and
meal locations’ share of the total beverage frequency.

The six-year MEC-examined sample weights were applied to the calculation of all
population parameters, such as the mean, median, or percentage of the population. The
sample weights account for unequal probability of selection, non-coverage and non-
response and adjust the data to the 1990 U.S. Census figures, therefore the estimates were
representative of adolescents in the 1990 U.S. population. The linearization (Taylor
series) method was used in SUDAAN to account for effects of the clustered sample

design in estimating the variance of population parameters (Shah et al., 1997).

142

RESULTS

Mean beverage fiequencies and the beverage’s percentage share of total beverage
frequency are presented in Table 2. The mean percentages of total beverage frequency
indicate that, on average, adolescents reported milk most frequently (39%), sugar-
sweetened beverages was second (30%), juice was consumed 20% of the time, with other
beverages contributing the remaining share (11%).

The distribution of beverage consumption frequency and the distribution of the
beverages’ percentage share of total beverage consumption are presented in Table 3.
Cumulative percentage distributions show, for example, milk was consumed <1 time/day
by 31% of youth, and conversely, 69% had milk 21 time/day. Two different methods
used to classify youth as frequent beverage consumers, the ‘frequency’ and ‘share’
methods, resulted in similar estimates of the prevalence of frequent beverage
consumption. Youth classified as frequent milk, sugar-sweetened beverage or juice
consumers using the ‘frequency’ method (i.e., fi'equency 21 time/day) were 69%, 49%,
and 35%, respectively, while 69%, 52% and 32% were classified as frequent consumers
of milk, sugar-sweetened beverage or juice, respectively, using the ‘share’ method (i.e.,
share 225% of total beverage frequency).

Beverage pattern prevalence estimates are shown in Table 4. Beverage patterns
were defined using the two different methods to classify youth as consumers of each
beverage combination. Based on beverage consumption data presented above, it was
expected that ‘milk’ patterns would be most prevalent, followed by ‘sugar-sweetened
beverages’ and ‘juice’ patterns, and beverage patterns using the ‘share’ method (share

225% of total) were ranked in that order. The ‘all caloric beverages’ (MSJ) pattern was

143

Table 2. Mean frequency of beverage consumption (# of times/day) and the
beverage's percentage share of total beverage consumption (% of total frequency),

using food frequency questionnaire (F FQ) data of adolescents aged 12-16 years1

 

 

 

Frequency (# times/d) Share Q/o of total)
Type of Beverage Mean 1 SE Mean :t SE
All beverages 4.2 i 0.1 100.0 i 0.0
Milk&milkdrinks 1.6101 38821.1
Fruit juice (100%) 0.9 :t 0.0 19.9 :l: 0.8
Sugar-sweetened beverages 1.2 :h 0.1 29.9 i 1.0
Fruit drinks 0.5 i 0.0 12.0 d: 0.5
Regular soft drinks 0.7 :h 0.0 17.9 i 0.7
Other beverages 0.5 :t 0.0 11.3 :1: 0.7
Diet soft drinks 0.2 :t 0.0 4.5 i 0.4
Coffee and tea 0.3 i 0.0 6.4 :l: 0.6
Beer, wine, liquor 0.0 :I: 0.0 0.5 :t 0.1

 

1 Source: NHANES 111, 1988-1994, adolescents aged 12-16 years, with complete,
reliable 24-hr recall dietary interview, and no missing or unknown responses to
beverage items on FFQ, excluding pregnant or lactating females (n=1,872)

Means and standard errors are sample-weighted and estimated by linearization
method of SUDAAN.

144

 

md 0.: fix m.wm odm «.2 03 8mm 9.8 900 .8: 9 N
00808 90:23:00.“.
88: 98880
880800 82:02.9

 

0.0 8 0.0 3 M: 0.: 3 Z S 9% a: v N
od 9.9 ed m.m in 0.9 m8 m.m 8.3 980 :0: m
0.0 2N a; A: 9m 0.: 06 we: WE 9:: :0: N
no 2 we :2 02 a: mom New 3m 3. 8 2
_.o m; m; m.» 88 2m m.: Qw wd 2003 :0: 9m
9.9 md vd m.~m va ow _ w. a QNN 9.2 0.003 :0: Wm
m.m 0: Wm ma: w.m_ 9.2 md 0.2 o6. #003 :2:
0.9 0.0 m9 9.9 82 W: mé wd N.m 8:08 :08 m-—
finw 98m ~68 ma: 2mm wdm w.m ms Nd 0:0N

9:0m200 90:05:08. 8 080002000 8 gm 0835209 98830 $083

.8 no 988089

 

8:84 a. 009. 0% 88.5 com 88.95 88.5 mom0:0>0m momflofim 02E. 88:3 052 88:83:00

 

8:: .08 880 020 com .8: “we: 550 $88 :8 $02 0 ea: 8823
00m80>0m 880 w>0m 00:2003m mow80>0m .8 00999. .8 9.82200
.8 009988 Lawn—m .8 009980m

 

099 08 8 90:25.8 022028 88: .8 0280 0880809
0.0990988 ems—2,08 08 8 90080330 8 8009 2&9 00mm 080002000 .8 o\ov 888808 0:: .099 08 8 00998983
088268 8:85:00 8 90.08: .8 *5 90:00:08 05 8 90.288800 8 8:09 87$ 00mm 080000800 .8 090v 8888me .m 0809.

145

AmanuE 0090809 8898009 8 8080.8 w89080x0 .099 :0 08099 088058 09 0080900: 85059:: .8 8898 0:
90:0 $69898 9.50990 99000: 89am 080990: .09099800 83 .8009 09-2 :08 08000090900 {009-309 .999 mm79<9979 H0080m

 

Nd 1w ~.m v.3 0.9.9 v.09 9.Nm w. 9 m W8 096.3 N
900508 .0980.
m8m: 9:0w0900

88:88 80:00: 9

 

06 Yo md m9 md 0.0 mg. v.9 9.9 $93 N
od 9N 0.0 :6 0N 8m 9.3 06 9mm $m9Vém
Nd ed 0.: 9.8 w.9 9 0.9 9 94mm méN msm XOmV-mN
5o No9 9.0 9.09.. odm ~89 fiwm 9.9m 0mm $mmvé9
9.N9 9.3 Wow odm odm QNM 9mm 9 odm 9.0 A.\..o9v.o.n
9.3 0.0m N60 m.o9 v.3 w.mm w.m ms NN 0:0N

9.8%0900 .0580. 8 08000090900 8 9099 08800909 983000

9980:9008 08:08:

90898 $9 008%

 

80909.9 0% 00.9. a. $98.5 9.8m 0098.5 0098.5 00w80>0m 00w0:0>0m9 008—. 898.89 099939 8988:0800

 

25: 080 8000 .20 000 .039 :8 H200 50-08 as: .989 a. 82 omens:
m0w80>0m 9050 w>0m 00:09003m.8w:m 00w80>0m 8 0099.9. 8 9:0w0900
8 009985 8 0099898

0.88 .m 2.5

146

Table 4. Percentage of adolescents aged 12-16 years classified by beverage pattern,
i.e., the combination of beverages frequently consumed in the FFQ, based on
criteria for frequent consumption using the 'frequency' method (frequency 21 time/
day) and 'share' method (share 225% of total frequency)l

 

 

 

_ Comb. 'Frequency' Method 'Share' Method
Beverage PatternZ CodeJ Rank Percent :1: SE Rank Percent :1: SE
All youth (%) 100.0 at 0.0 100.0 :1: 0.0
All caloric beverages MSJ 3 14.2 :1: 1.3 7 4.8 i 0.8
Milk & sugar-swt bevg MSj 2 20.4 i: 1.6 1 24.7 :1: 1.6
Milk &juice Ms] 4 12.8 i 1.4 4 14.8 d: 1.7
Milk Msj 1 21.4 :1: 1.5 2 24.2 i 1.7
Sugar-swt bevg & juice m8] 7 4.3 d: 0.8 5 6.5 :t 0.8
Sugar-swt bevg mSj 6 10.4 :t 1.1 3 16.1 :1: 1.8
Other beverages & juice msJ 8 3.8 i 0.8 6 5.7 i 0.9
Other beverages msj 5 12.8 d: 1.1 8 3.2 3c 0.6

 

' Source: NHANES III, 1988-1994, adolescents aged 12-16 years, with complete,
reliable 24-hr recall dietary interview, and no missing or unknown responses to
beverage items on FF Q, excluding pregnant or lactating females (n=1,872)
Percentages and standard errors are sample-weighted and estimated by
linearization method of SUDAAN.

2 Beverage patterns are combinations of milk (milk & milk drinks), sugar-swt bevg
(fruit drinks & regular soft drinks) and juice (100% fruit juice). Other beverages
include diet soft drinks, coffee/ tea, beer/ wine/ liquor & water.

3 Combination code refers to the pattern of beverages consumed or not consumed at
the specified level. Capital letters M, S and J refer to milk, sugar-swt bevg and
juice, respectively, consumed at the specified level, and lowercase letters m, s and;
represent consumption of less than the specified amount.

147

ranked high using the ‘frequency’ method (frequency 21 time/d), but this pattern was less
prevalent using ‘share’ method (share 225% of total) that requires roughly equal shares

for each of the three beverage types. The ‘other beverages’ (msj) pattern also ranked high
using consumption the ‘frequency’ method, because “under-reporters” would be

classified to the ‘other beverages’ (msj) pattern if they reported drinking all three

beverage types <1 time/d. Using the ‘share’ method, the ‘other beverages’ (msj) group
would have to be drinking ‘other beverages’ or water most fi'equently, when neither milk,
juice nor sugared-sweetened beverages was consumed with a share 225% of total
beverage frequency. Defining beverage patterns using the ‘share’ method (share 225%
of total) was preferred because each youth was classified by his/her own beverage
choices (i.e., the beverages he/she drinks most frequently), and this method was not
biased by under- or over-reporting. The ‘share’ method was used to classify youth by
beverage patterns when examining beverage frequencies and nutrient intakes of beverage
pattern groups.

Table 5 shows mean beverage fi'equencies of youth classified to beverage pattern
groups using the ‘share’ method. Mean beverage fi'equencies tended to be higher in the
FF Q than in the 24-hour recall, except consumption of regular soft drinks in the FFQ was
less than in the 24-hour recall. As expected, the ‘other beverages’ groups drank more
diet sofi drinks and more coffee/tea than the ‘caloric beverages’ groups, in both the FFQ
and 24-hr recall. Mean water (oz) was also high in the ‘other beverages’ (msj) group,

because youth would be classified by the ‘other beverages’ (msj) pattern if they did not
drink any beverage other than water. The ‘juice’ groups drank juice more frequently than

the ‘non-juice’ groups, the ‘milk’ groups drank milk more fiequently than ‘sugar-

148

 

 

Md H Nd _d H md dd H md a _d H nd dd H md _. _d H md M _d H dd M _d H md dd H dd EdN
_d H Nd dd H md m Md H MA m Md H wd dd H Nd dd H md M dd H ed M Md H 5d dd H md 0.:
83.6 33m
NdH dd NdH N; m MdH _.N m NdH a; MdH MA ~dH m._ M MdH o; M _.dH 0; ddH m; EdN
Nd H Nd Md H ed Zm Md H 5N m Nd H N; Z dd H md dd H md M Md H c; M _d H 0— do H N; 0mm
when $6-3m

 

dd H dd Nd H ed Md H Nd _. Md H nd dd H Nd p. Md H hd Md H Nd —d H dd dd H md EdN

O MdedO NdeN dded a MdeM dded H MdeN dded Mdeé dded 0mm

033.

 

NdHcd Mded _dH md MdHod 2 _dH mg 2 MdHNM M MdH M._ M MdH MA MdeA EdN

NdHcd MdHnd ddHcd _.oHdd 2 MdeN 2 _.de.NM Mde._ M MdH m; MdHcé 0%
M52

md H fin Nd H d.m _d H fim Nd H Nd Md H N.m Md H Wm Md H m.m Nd H fin Md H N.m EdN

hd H dd dd H dd Nd H md md H dd _d H Wm Nd H nd Md H dd md H dd Md H Nd 0mm
. mowfigofz

 

=88 EdN was 0.: 2d 5 @3525“ do do honed—wed ammuotfin Amm H 525 :82

 

an n 5 E: u 5 ARM n 5 an: n 5 d S u 5 a: u 5 a? u 5 as n 5 an? n 5 258m a.

 

.25 Ba .58 SE .32 22 .32 22 52.2 2538
w>om 033. on. whom 023. a. w>om M32 023 w>om Em w>om =<
.550 w>om .650 Emémwnm Eméwmsm a. Md: -wsm u% #52 QEOQO =<

«98.5 Eotam owfio>om

 

 

 

. hose:
com «recto no 6093 0mm 05 E @083 .A dog.“ :33 do o\omNN v.8

 

980% o - so am 0
M N M own 3208233 mo waaohw w: o baosgfl wowwhokun mo EOMWNMMHHOE .oafim.
. . EGO ..

Cg ~HNUOH EIVN mu 0 . a
5 9E 05 5 5335an Mvwsga omega; .3 Bugs
. oaozcoa .
0

mass as: .m 2.5.

149

 

 

 

 

 

 

 

ded dem dem mHMd ded mHmd dem dMHdd _HNd N353?
Md H Md dd H dd dd H dd dd H dd dd H dd dd H dd dd H dd dd H dd dd H dd EdN
_.d H _.d dd H dd m dd H dd m dd H dd dd H dd dd H dd dd H dd dd H dd dd H dd 0mm
3:32
0 Nd H dd 0 do H md Md H md dd H Md dd H md Md H md dd H Nd Md H Md dd H md EdN
O md H EM 0 _.d H md Z _.d H dd dd H _d Z dd H md dd H Md dd H Nd dd H Md dd H md 95
motoubou
O Nd H md 0 do H md Md H Md dd H dd dd H Md dd H _d dd H Md dd H dd dd H Md EdN
O md H d.— O Nd H md Z _d H md dd H Md Z dd H Nd dd H Md dd H Md dd H dd dd H Nd 0mm
83.6 65
O Nd H d; O Md H md Md H dd Md H do _d H dd _d H md dd H md —d H Md dd H dd EdN
O dd H d.N O Nd H dd 2 _.d H Nd _d H md 2 do H md dd H Nd dd H md dd H _d dd H md 93
w>on 550
Ndth Nded m MdHnM m NdeA Mded Mded MdH m4 MdeA ddH M." HadN
Nd H md dd H md Zm _d H m; m Nd H dd Z dd H md dd H md M Md H d; M dd H dd dd H 5d 0mm
mg dew

 

=32 EdN dam 0mm 05 E Axedhosfi do #3 Mugged owfigod Amm H 525 :82

 

Gm u 5 E: u 5 chm n 5 as: u 5 d S u 5 a: u 5 and u 5 as n 5
7:: :8 .55 $8 .32 :2 $2 $2
m>om 033 a. w>om 83—. do w>om M54 033. w>om Em w>om
sac mam 550 Emémsm Emémsm a. 53 -msm a. ea: E28 3.

 

macaw Egan owfio>om

at} n 5 83% a.

a 2-2
3.

09¢. w>om

5.58 .m 2.3.

150

$95.52 3.8th bod—.785 Z .m> 58:64.4 mega .335 m 98 c.9558 manta .w>od Sada; m .m> Ammgmmz
magma 2:89 2 Awning” 3.83am .owfigod 559V 0 .m> :3,sz 2:038 .owegon ate—8d M 0.83 3.0.888 Eoboddom
.Amddvfi 3:on Eaten owEo>od 850 2 6.380 E moo—SE hon—mm: 55 339% 8033 $8269 8 Home Z ._. .m .2 .O .M
2:68“ dommoam 2d 55 $3 do 533538 8 Home“ _. 28 m .8 $032 38532 28 Seconded owfigod 38 do admNN 25cm
23 383:8 50380de .Aoomsm and odddz 08% 93 A356 «8 Him»: a. 3:5 “3.8 mowfigod decouooBmémwsm A356
58 a. M35 5:: 8 Mode.— _. 28 m .2 £88— .fifimu dean—mace bacon—dud mowfiofifi mo 893388 06 mm 5033 ow80>on SE. a
.Z<<QDm do @0508 cedauwaofi— .3 33833 mm Hobo 235% was dofimmoBé—dnam fl >05:de owfiofin 502
33.7.3 8380M wfiufiofl .8 “5&8.” wands—88 .93 co mon omflgon 8 8309.8 850515 8 mammmmfi 0:
ca.» .3223: .6806 :32 EdN 0332 68588 5:5 £89» dTNM down 358203 .dddTwde .E mmZ<ZZ Hmadam _

6.28 .m 2.3

151

sweetened beverages’ groups, and the ‘sugar-sweetened beverages’ groups drank regular
soft drinks more frequently than the ‘milk’ groups, in both the FFQ and 24-hr recall. The
‘sugar-sweetened beverages’ groups drank more fruit drinks than the ‘milk’ groups in the
FFQ, but not in the 24-hr recall. The 24-hour recall intakes of fruit drinks were higher
for ‘juice’ groups than ‘non-juice’ groups. Some misclassification to ‘juice’ groups
might have occurred if youth mistakenly reported fruit drink frequency in response to the
100% fruit juice question in the F FQ.

Table 6 presents the share of total beverage frequency consumed at meal
occasions eaten at-home and away-from-home. Milk and juice were most frequently
consumed at the breakfast meal occasion, while fruit drinks, regular soft drinks, and other
beverages were most frequently consumed at snack/ beverage breaks. Milk and juice
were consumed at home ~75% of the time, while for regular soft drinks, the share
consumed at home was 52% of the total. The fast food restaurants’ share of consumption
was greater for regular soft drinks than for any other beverage.

The 24-hr recall nutrient intakes of beverage pattern groups are presented in
Table 7. It was hypothesized that ‘sugar-sweetened beverage’ groups might have high
energy intakes from both beverages and non-beverages, if regular sofi drinks were
frequently consumed with high-fat snack foods and fast foods. Comparisons of ‘milk’
and ‘sugar-sweetened beverage’ patterns showed that fat intake from beverages was
greater for ‘milk’ groups, and added sugar intake from beverages was greater for the
‘sugar-sweetened beverages’ groups. Percentage of calories from added sugars was
greater for ‘sugar—sweetened beverages’ groups than for ‘milk’ groups, but there were no

differences in energy or fat intakes from the total diet. Percentage calories from fat and

152

 

N.m Nod H mod Nd mod H Ed was mod H nod ~.N 8d H mod hd ddd H 8d .650
ON 5d H mod w.m 8d H bod m; 3d H Nod dd ddd H 5d vd ddd H 5d 398;.
Q: mod H 3d N.m Nod H 2d ad mod H 2d dd mod H odd 9N 3d H vod oEom m.o:ooEom
w...» 5d H bod 5m Nod H dfid vd ood H ddd Nd ddd H dod md dod H 3d «toaommobgfimom
d.m 8d H vod fid Nod H 2d 6; 5d H Nod wd odd H dod m; 3d H Nod 30 oxagoom “mam

dd mod H Qd md Nod H 2d 0: vod H VNd v.3 mod H 2d md~ mod H ©Nd 30an
0; 5d H Nod fim Sd H mod Nd dod H ddd fig Sd H 5d md ddd H 8d 3.53
93 mod H Rd 0.; mod H Rd v.3 bod H hwd m: odd H mad 5R. vod H 0N; ofiom

ode mod H 5“.— ddg mod H R; ode mod H 9m; ode mdd H wNA ode mod H No; moon—m 322 =<

 

ddv mod H cod odv 3d H mwd _.Nv vod H hmd dfiN Vdd H cmd N.w_ mod H dmd 3&on @630me
m.wN mod H Nvd w.wN mod H 3d ddN mod H mmd 9: mod H 2d QB Nod H oNd Hoamzmhmafifl
NAN mod H Rd 0.3 Nod H de QNN vod H Rd as; mod H mNd WE mod H dmd £254
dd mod H :d Nd 5d H nod md— Nod H 2d c.Nv vod H de mfiv mod H Rd noggmwgoum

ode mod H to; ode mod H E..— ddo_ mod H om; ode mod H wN.— dds mod H No; E88300 322 =<

 

.x. mm H as: .x. mm a as: .x. mm a an: .x. mm a. as: .x. mm 282 8.3 no: a.
6% u 5 at; u 5 $8 u 5 93 n5 53 n 5 8.2.80 no:
.8338 .050 .355 «cm 5.505 355 .35 8.3 gem x8. 935 5:2 5. be:

 

.5536on owfiofin .88 of 9 mach—moo— 308 and $8688 308 do 5:35:00 ommuaoopom d5
53.» 07$ comm E08328 owwuo>on macaw :88 BAN 05 E Quiche—EH do .3 husband owfio>o£ 532 d 28d.

153

.83: but,» too: cum «8 \oobOo .356 $8 86 3205 mowfigop 650 N
.ggm do 62:08 coantmoé 3 838:8 mm 8.5 Egg,“ can BEwBBéEEMm mm 55:”.on 09533 502
. 335:5 Naoaozcodv 3508 .3533? 05 28 Sad :32 BEEN 05 mama: 35:62 203 Evan—mace ommb>om
AanJHE 8380.“ $688— .8 :8:on mags—88 .93 no 38: owfiokun 8 830%»: 5505.5. 8 9:32: o: 28
.33385 E806 :88 EéN 035:8 .32an0 53, who» 07$ comm $58203 €09-33 .E mugs/ZZZ ”850m _

5.58 a 2.5

154

saturated fat, and caffeine intakes from the total diet and from beverages were lower for
‘juice’ groups than for ‘non-juice’ groups.

Energy, fat, carbohydrate and sugars intakes from the total diet and from
beverages were higher in ‘caloric beverages’ groups than in ‘other beverages’ groups, but
caffeine intakes from the total diet and from beverages were higher in ‘other beverages’
groups than in ‘caloric beverages’ groups (Table 7). Calcium and vitamin D intakes
from the total diet and from beverages were lower in ‘other beverages’ groups than in
‘caloric beverages’ groups. Calcium, potassium and vitamin D intakes from the total diet
and from beverages, and vitamin A intake from beverages were higher in ‘milk’ groups
than in ‘sugar—sweetened beverage’ groups. Vitamin C intakes from the total diet and
from beverages, and folate intake from beverages were higher in ‘juice’ groups than in
‘non-juice’ groups. As shown in Table 8, teen-age girls, in particular, appeared to be at
risk for inadequate intakes of calcium, potassium and vitamin D, and older adolescents

appeared to be at risk for inadequate vitamin A intake.

DISCUSSION

Beverage patterns developed in this study provided a simple classification scheme to
identify segments of the adolescent population having nutritional risks associated with
the types of beverages usually consumed. One other researcher identified similar
beverage pattern groups (Bowman et al., 2002), but the current study used F FQ rather
than 24-hour recall data to classify youth by the types of beverages most frequently
consumed. Others have used 24-hour recall data to identify segments of the adult

population eating different types of breakfast foods to examine the association of

155

Table 7. Mean daily nutrient intake from total diet, beverages and non-beverages among

groups of adolescents aged 12-16 years classified by beverage pattern‘

,2

 

 

 

 

 

 

 

 

 

 

Nutrient Beverage Pattern Group2
Amount from: All Milk & Sugar- Other Other
Total Diet Caloric Bevg Swt Bevg. Bevg & Juice Bevg
Beverages MSJ MSj msJ msj
Non-beverages (n = 109) (n = 435) (n = 103) (n = 55)
Mean (mean i SE) daily nutrient intake
Food Energy (Kcal)
Total Diet 2543 i 164 K 2470 i 103 K 1902 :t 119 2099 i 189
Beverages 494 i 37 K 542 i 27 K 410 i 39 369 i 58
Non-beverages 2049 d: 135 K 1928 i 90 K 1492 :t 107 1730 d: 156
Carbohydrate (g)
Total Diet 348 d: 31 K 321 i 12 K 257 i 15 269 d: 22
Beverages 97 :t 7 K 106 i: 5 K 87 :1: 8 74 :h 10
Non-beverages 250 i 27 K 215 :t 9 K 170 :1: 11 195 i 17
Vitamin A (mg RAE)
Total Diet 650 :t 72 629 3: 48 402 i 47 623 :1: 239
Beverages 153 :t 27 158 i 13 89 d: 17 123 i 65
Non-beverages 498 i 65 472 i 47 313 i 35 500 i 237
Vitamin C (mg)
Total Diet 134.8 a: 17.4 102.5 i 7.9 110.4 :1: 21.3 61.3 :I: 13.4
Beverages 82.4 at 14.9 40.2 :t 6.0 70.9 i 19.0 19.6 :t 9.5
Non-beverages 52.2 :1: 7.1 K 62.1 i 6.9 K 39.4 i 4.0 41.6 2+: 7.1
Folate (mg DFE)
Total Diet 772 :t 99 601 d: 24 432 :1: 45 666 :1: 164
Beverages 64.2 i 16.5 35.7 :t 4.1 46.5 i 11.3 41.6 :1: 17.8
Non-beverages 708 :1: 97 565 d: 23 386 :t 39 624 :1: 164
Calcium (mg)
Total Diet 1067 i 91 K 1049 i 44 K 648 at 43 722 d: 85
Beverages 415 :t 62 K 435 i: 29 K 251 i 39 239 i 58
Non-beverages 651 :1: 47 K 614 i 34 K 397 d: 39 482 :1: 64
Potassium (mg)
Total Diet 2553 i 219 2579 :t 116 2230 :1: 180 2102 d: 244
Beverages 789 i 121 690 i 48 658 :t 126 483 at 98
Non-beverages 1764 i121 1889 i 100 1573 i112 1619 i 196
Vitamin D (mg)
Total Diet 6.2 :1: 0.9 K 6.0 :1: 0.3 K 4.2 i: 0.7 2.8 i 0.4
Beverages 3.2 i 0.5 K 3.3 d: 0.2 K 1.7 i 0.3 1.5 :1: 0.4
Non-beverages 3.0 d: 0.6 2.7 :t 0.2 2.5 i 0.5 1.3 i 0.2

 

156

 

rs among

Omar

Bevg

m s]
n = 551

*1
H 1*
Z;

‘l\

l 3
H-
I‘m

\q Ci: '
\\
‘1‘
(I

A;
;;3 =95

Table 7, cont'd.

 

 

 

 

 

 

 

 

 

 

Nutrient Beverage Pattern Group2
Amount from: Milk & Sugar-Swt Sugar-Swt
Total Diet Juice Milk Bevg & Juice Bevg
Beverages MsJ Msj mSJ mSj
Non-beverages (n = 248) (n = 41 l) (n = 184) (n = 327)
Mean (mean t SE) daily nutrient intake
Food Energy (Kcal)
Total Diet 2384 at 113 2321 :t 77 2488 :t 170 2237 :t 96
Beverages 523 :t 29 485 i 25 505 :1: 47 462 :t 30
Non-beverages 1860 i 99 1837 i 67 1984 :1: 150 1775 :t 90
Carbohydrate (g)
Total Diet 328 i 16 301 i 11 357 i 29 298 i 12
Beverages 103 :1: 5 84 i 5 111 3:11 106 i 8
Non-beverages 225 at 14 218 :t 10 246 i 24 192 i 9
Vitamin A (mg RAE)
Total Diet 684 :t 74 646 :1: 38 561 :t 72 472 :1: 82
Beverages 188 i 28 M 212 $16 M 81 $14 60 at 8
Non-beverages 496 i 69 435 i 32 480 i 72 412 :t 80
Vitamin C (mg)
Total Diet 159.4 i 9.9 J 87.6 i 6.1 159.1 1 24.9 J 84.5 :1: 9.3
Beverages 103.1 :1: 9.3 39.2 i 5.7 101.1 2‘: 21.6 J 36.0 :1: 5.4
Non-beverages 55.9 i 5.1 48.1 d: 3.1 57.8 i 6.8 48.2 :1: 5.5
Folate (mg DFE)
Total Diet 720 i 81 627 d: 29 693 i 89 468 :1: 37
Beverages 73.7 i 7.5 J 43.3 :t 2.9 65.8 d: 18.5 J 22.1 :t 3.3
Non-beverages 646 :t 80 584 i: 29 627 i 80 446 :1: 35
Calcium (mg)
Total Diet 1065 i 68 M 1149 :1: 48 M 766 i: 58 752 :t 58
Beverages 486 :t 57 M 574 i 32 M 249 :h 43 197 i 21
Non-beverages 579 i 41 575 i 30 516 i 38 555 :1: 47
Potassium (mg)
Total Diet 2899 :1: 142 M 2615 i 74 M 2535 d: 242 2133 i155
Beverages 981 :1: 84 M 880 i 41 M 653 :1: 114 386 :1: 6O
Non-beverages 1918 i 96 1735 i 58 1883 i 168 1747 d: 125
Vitamin D (mg)
Total Diet 6.7 at 0.8 M 7.2 i 0.4 M 4.3 t 0.6 3.4 :t 0.3
Beverages 3.5 :1: 0.5 M 4.5 d: 0.3 M 1.6 d: 0.4 1.3 i 0.2
Non-beverages 3.1 i 0.3 2.7 :t 0.2 2.6 :h 0.3 2.1 at 0.2

 

157

 

Table 7, cont'd.

 

Nutrient Beverage Pattern Group2
Amount from: All Milk & Sugar- Other Other
Total Diet Caloric Bevg Swt Bevg. Bevg & Juice Bevg
Beverages MSJ MSj msJ msj
Non-beverages (n = 109) (n = 435) (n = 103) (n = 55)

 

Mean (mean i: SE) daily nutrient intake

 

 

 

 

 

 

Caffeine (mg)
Total Diet 41.3 i 7.7 68.2 i 6.8 57.4 :t 12.2 0 138.9 i 19.6 0
Beverages 36.4 :1: 7.5 66.2 i 6.8 55.1 :1: 12.3 0 136.8 :t 20.2 0
Non-beverages 4.9 at 2.0 1.9 :1: 0.2 2.3 i 0.5 2.1 i 1.0
Total Sugars (g)
Total Diet 187.6 i 20.1 K 170.1 i 8.1 K 134.0 3: 8.6 124.7 :1: 15.9
Beverages 95.1 :1: 6.8 K 103.2 :1: 5.5 K 82.2 i 7.7 70.7 i 10.7
Non-beverages 92.6 :1: 16.3 66.9 d: 4.2 51.8 i: 4.8 53.9 :t 13.8
Added Sugars (g)
Total Diet 138.0 :1: 15.7 K 130.2 i 7.8 K 92.9 a: 8.3 99.6 :1: 15.4
Beverages 63.1 i 6.1 78.5 i 5.7 57.0 i 6.9 58.9 d: 9.2
Non-beverages 74.9 i 14.8 51.7 :h 4.1 36.0 d: 3.8 40.7 i: 13.8
Total Fat (g)
Total Diet 93.5 i 5.6 K 96.2 :1: 4.9 K 72.3 :t 6.1 83.5 d: 9.0
Beverages 7.6 d: 1.2 K 7.5 i 0.5 K 4.8 d: 1.0 3.3 :t 1.0
Non-beverages 85.8 i 5.1 88.7 :t 4.9 67.5 :t 6.0 80.2 i 8.8
Saturated Fat (g)
Total Diet 32.7 i 2.4 K 34.1 i 1.5 K 23.8 d: 2.2 26.7 :t 2.1
Beverages 4.7 :h 0.8 K 4.6 i 0.3 K 2.9 i 0.6 2.0 i 0.6
Non-beverages 27.9 d: 2.3 K 29.5 d: 1.5 K 20.9 i 2.1 24.7 :t 2.0
Percent of calories:
Added sugars 20.6 :1: 1.2 21.5 :t 0.9 20.6 i 2.2 17.4 d: 2.4
Total Fat 32.8 :1: 1.5 34.3 :t 0.5 32.8 i 1.0 35.7 i 1.2
Saturated Fat 11.5 :t 0.8 12.4 :1: 0.3 10.9 i 0.5 11.7 :1: 0.6

 

I Source: NHANES 111, 1988-1994, adolescents aged 12-16 years, with complete,
reliable 24-hr recall dietary interview, and no missing or unknown responses to
beverage items on FFQ, excluding pregnant or lactating females (n=1,872).

2 The beverage pattern is the combination of beverages frequently consumed. Capital

letters M, S and J refer to milk (milk & milk drinks), sugar-sweetened beverages (fruit
drinks), sugar-sweetened beverages (fruit drinks & regular soft drinks) and juice (100%
fruit juice), respectively, consumed with a share 325% of total beverage frequency,

and lowercase letters m, s and j refer to consumption of less than the specified amount.

158

 

Table 7, cont'd.

Nutrient
Amount from:
Total Diet
Beverages
Non-bever_ages

Beverage Pattern Group2

 

Milk &
Juice
M s J

(n = 248)

Milk
M s j

(n=411)

Sugar-Swt

Bevg & Juice

mSJ
(n=184)

Sugar-Swt
Bevg
m S j

(n = 327)

 

Mean (mean 3: SE) daily nutrient intake

 

 

 

 

 

 

Caffeine (mg)
Total Diet 43.4 i 6.3 52.4 :1: 4.2 N 44.6 i 7.4 76.5 :t 6.7 N
Beverages 40.2 i 6.3 49.1 :1: 4.3 N 43.2 :1: 7.3 74.1 i 6.7 N
Non-beverages 3.3 :1: 0.7 M 3.3 d: 0.5 M 1.3 :h 0.3 2.4 :1: 0.4
Total Sugars (g)
Total Diet 175.4 :1: 8.4 152.1 :1: 7.5 175.2 i 13.8 166.1 d: 8.6
Beverages 99.9 d: 4.7 81.2 i 4.7 108.6 :1: 10.7 104.2 :t 7.7
Non-beverages 75.5 :1: 6.1 70.8 d: 5.5 66.6 :1: 6.4 61.9 d: 4.3
Added Sugars (g)
Total Diet 113.3 :E 7.2 108.4 i 6.7 132.7 d: 14.6 133.5 i 6.6
Beverages 58.7 d: 4.3 51.9 d: 4.5 81.4 :t 11.0 S 84.5 :t 5.8 S
Non-beverages 54.6 d: 5.4 56.5 :1: 4.7 51.3 d: 6.6 49.0 :t 3.9
Total Fat (g)
Total Diet 87.1 i 5.3 91.8 :1: 3.6 85.8 i 5.5 86.3 :1: 5.1
Beverages 7.9 i 1.0 10.5 i 0.8 M 5.0 :1: 1.3 3.1 :1: 0.4
Non-beverages 79.2 :I: 5.0 81.3 i 3.2 80.8 :t 5.5 83.2 :t 5.0
Saturated Fat (g)
Total Diet 30.2 :1: 1.9 33.4 :1: 1.7 28.1 :1: 1.8 29.5 :1: 1.8
Beverages 4.7 :t 0.6 6.5 :1: 0.5 M 3.0 :1: 0.8 1.9 :1: 0.3
Non-beverages 25.4 i 1.7 26.9 :1: 1.4 25.1 i 1.8 27.6 :1: 1.8
Percent of calories:
Added sugars 19.3 i 0.8 18.9 i 0.9 23.5 i 2.1 24.9 :1: 1.2 S
Total Fat 31.4 i: 0.7 34.5 :1: 0.5 N 31.2 d: 1.1 33.8 i 1.0 N
Saturated Fat 11.0 :1: 0.4 12.4 d: 0.3 N 10.4 :t 0.5 11.5 :t 0.4 N

 

Means and standard errors are sample-weighted and estimated by linearization

method of SUDAAN.

K, O, M, S, J, N refer to beverage pattern groups with higher intakes in contrast to other
beverage pattern groups (p<0.05). Bonferroni contrasts include K ('caloric beverage'
patterns MSJ/MSj) vs. 0 ('other beverage' patterns msJ/msj), M ('milk' patterns
MSJ/Msj) vs. S (sugar-sweetened beverage' patterns mSJ/mSj) and J ('juice' patterns

patterns MSJ/mSJ vs. N ('non-juice' patterns Msj/mSj).

159

Table 8. Dietary Reference Intakes (DRI) and daily nutrient intake in gender-age
groups of adolescents aged 12-16 years‘

Boys 12-13 yr Boys 14-16 yr Girls 12-13 yr Girls 14-16 yr

 

Nutrient (n = 109) (n = 435) (n = 248) (n = 411)
Food Energy (Kcal)

Mean :t SE ** 2583 d: 111 2832 i 90 1996 i 68 1927 i 66

Median i SE 2344 :t 79 2665 :1: 73 1914 i 82 1798 :1: 79
Carbohydrate (g) ‘

EAR 100 100 100 100

MeaniSE ** 339i15 378 113 269i9 256i8

MedianiSE 306::11 351 i 10 255 ill 238::7
Vitamin A (mg RAE)

EAR 445 630 420 485

Mean :1: SE H 723 i 62 625 :1: 28 630 d: 78 475 d: 47

Median :t SE 565 d: 43 522 i 31 401 at 32 339 i 21
Vitamin C (mg)

EAR 39 63 39 56

MeaniSE ** 108:1:6 131::9 104::7 91.1:1:7.1

MedianiSE 82i9 833:11 85::9 553:6
Folate (mg DFE)

EAR 250 330 250 330

Mean i SE ** 682 :1: 33 772 i 51 535 :t 32 452 i 23

Median d: SE 564 d: 34 596 d: 34 440 :t 23 371 i 27
Calcium (mg)

A1 1300 1300 1300 1300

MeaniSE ** 1142 :t56 1144i51 866 $39 773 $34

Mediani SE 1023 i 71 1017 :1: 56 767 3c 48 653 :h 32
Potassium (mg)

A1 4500 4700 4500 4700

Mean :t SE ** 2677 :1: 112 3038 :1: 143 2250 3: 72 2097 :t 84

Mediani SE 2387 :1:-122 2838 i150 2132 i102 1868 i114
Vitamin D (mg)

A1 5.0 5.0 5.0 5.0

Mean i SE ** 6.5 d: 0.4 6.7 :t 0.4 5.3 :1: 0.4 4.4 i 0.3

Median t SE 5.5 i 0.4 5.4 i 0.5 4.0 i 0.4 3.4 i 0.3

 

160

Table 8, cont'd.

1 Source: NHANES IH, 1988-1994, adolescents aged 12-16 years, with complete,
reliable 24-hr recall dietary interview, and no missing or unknown responses to
beverage items on FFQ, excluding pregnant or lactating females (n=1,872).
Dietary Reference Intakes (DRI) include Estimated Average Requirements (EAR)
and Adequate Intake (AI) levels.

Means, medians and standard errors are sample-weighted and estimated by
linearization method of SUDAAN.

** Higher nutrient intakes of boys than girls by Bonferroni contrast (p < 0.01)

H Higher nutrient intakes of youth 12-13 yr than 14-16 yr by Bonferroni contrast

(p < 0.01)

161

breakfast patterns with energy/nutrient intakes and the prevalence of overweight (Siega-
Riz et al., 2000; Cho et al., 2003).

Estimates of the percentage share of total beverage frequency were used in the
current study in the same way that others have used 24-hour recall data to estimate the
percentage share of total beverage calories (French et al., 2003). The share of soft drinks
obtained at table-service and fast-food restaurants by children and adolescents age 6-17
years was 14.4% in 1977-78 and 22.1% in 1994-98 (French et al., 2003). These data
were not comparable to the 14.8% estimate for 1988-94, because only the older subgroup
was examined in the current study. The CSFII 1994-98 study indicated that children and
adolescents obtained only 4.1% and 3.0% of soft drinks from vending machines and
school cafeterias, respectively (French et al., 2003), while 9.0% of soft drinks were
consumed at school by adolescents age 12-16 years in NHANES 1988-94. The estimate
of school consumption was averaged over the entire year. The share of regular soft
drinks consumed at school would be higher if analyses were limited to youth who were
attending school and not on summer vacation.

Percentage shares of beverages consumed at eating occasions were calculated
from relative frequency data reported for adolescents age 13-18 years in NFCS 1977-78
(Guenther et al., 1986). The percentage share of soft drinks consumed at snacks was 31%
(Guenther et al., 1986), compared to 48% for adolescents in the current study. These and
other data show a trend toward increased consumption of soft drinks at the snacking
occasion (Nielsen et al., 2002; Nielsen & Popkin, 2004). The shares of milk and juice
consumed at breakfast in 1977-78 were 38% and 76% (Guenther et al., 1986), while

comparable estimates for 1988-94 from the current study were 47% and 43%,

162

respectively. Drinking juice was highly specific to the breakfast occasion (Guenther et
al., 2006). Among adolescent girls age 12-19 years, between 36-44% of non-milk
drinkers did not eat breakfast, compared to 11-12% of milk drinkers (Bowman et al.,
2002). Thus, consumption of both milk and juice is likely a good indicator that the youth
ate breakfast. That milk and juice were found in this analysis to be associated with
breakfast, and sugared-sweetened beverages were associated with the snacking occasion,
suggests that ‘milk’ and ‘juice’ groups would have more nutrient-dense diets than ‘sugar-
sweetened beverages’ groups.

Percent calories from added sugars were high for ‘sugar-sweetened beverages’
groups examined in the current study, as expected from other reports (Bowman et al.,
2002; Striegel-Moore et al., 2006), and the expected nutritional benefits of drinking
‘milk’ and ‘juice’ patterns were observed in the current study. The methods used here to
develop beverage patterns were unique, but the results were consistent with observations
from various studies that used the CSFH dataset (Hamack et al., 1999; Ballew et al.,
2000; Bowman et al., 2002; Frary et al., 2004). In adolescents age 13-18 years, those
drinking 213 oz/d soda drank less milk and juice, and had higher vitamin A, calcium,
vitamin C and folate intakes than non-soda consumers (Hamack et al., 1999). An
association between the amount of soft drinks consumed and calcium adequacy was
reported first by Guenther (1986). Others also examined calcium and folate intakes of
adolescents and found them to be inadequate in high sugar-sweetened beverage
consumption categories (Frary et al., 2004). In adolescent girls, not drinking milk was
associated with inadequate vitamin A, calcium and folate intakes (Bowman et al., 2002).

By speculation, folate intake of milk consumers was attributed to RTE cereal consumed

163

at breakfast (Bowman et al., 2002). In the current study, folate intakes of the ‘milk and
juice’ group were attributed to juice.

In a study of children and adolescents age 2-17 years, while controlling for
intakes of all other beverages in logistic regression models for all age strata, the
likelihood that vitamin A and calcium intakes were adequate increased with amounts of
milk consumed; the likelihood that vitamin C intake was adequate increased with juice;
and the likelihood that folate intake was adequate increased with intakes of both milk and
juice (Ballew et al., 2000). Even after Ballew et a1. (2000) controlled for milk and all
other beverages, soft drinks were inversely associated with decreased vitamin A
adequacy in all age strata, but an inverse association between soft drinks and calcium
adequacy was observed only for children <12 years old. Among children/adolescents age
9-18 years in another study, controlling for servings of milk and dairy foods, other food
group servings, in addition to gram amounts of all other beverages, showed a very weak
positive association between soft drinks and calcium (mg) intake (Storey et al., 2004).

Only one study examined the association of calcium with all types of beverages,
including diet soft drinks and coffee/tea as well as caloric beverages (Striegel-Moore et
al., 2006). In a 10-year followup study of adolescent girls, increased intake of milk and
diet soft drinks was associated with increased calcium; regular soft drinks, fruit drinks
and coffee/tea decreased calcium; and fruit juice had no association with calcium intake,
after controlling for all other beverages (Striegel-Moore et al., 2006). In the current
study, the ‘other beverages’ pattern included coffee/tea in addition to ‘diet’ soft drinks,

and adolescent females in the ‘other beverages’ group were likely at risk for inadequate

164

 

calcium and vitamin D intakes. The current study was the first to report potassium and
vitamin D intakes of beverage pattern groups.

Beverage patterns may be firnctional indicators of the risk for overweight when
they are associated with lifestyle factors. Drinking milk and juice was associated with
the breakfast meal, and other research has suggested that breakfast consumption might be
associated with other components of a healthy lifestyle such as high levels of physical
activity (Cavadini et al., 2000a; Keski-Rahkonen et al., 2003; Godin et al., 2005). Sugar-
sweetened beverages were associated with the snack occasion and meals eaten at fast
food restaurants. Youth fi'equently eat snacks and drink soft drinks while watching TV
(Coon et al., 2001; Giammattei et al., 2003; Matheson et al., 2004; Phillips et al., 2004),
and as a sedentary behavior TV watching would preclude physical activity (Nelson et al.,
2005). Besides drinking diet soft drinks, the ‘other beverages’ group was drinking
coffee, tea and alcoholic beverages. Older female adolescents in this group might also be
skipping breakfast and smoking cigarettes to control their weight (Malinauskas et al.,
2006). Health-compromising behaviors tend to cluster in high-risk youth (Keski-
Rahkonen et al., 2003). Thus, the ‘other beverages’ group might be at risk for becoming
dependent on caffeine, tobacco, alcohol or other drugs in addition to having risks for

nutrient inadequacy.

Strengths and Limitations
A major strength of this study was that youth were classified into beverage pattern
groups. The association of beverage patterns with energy intakes was investigated by

examining mean nutrient intakes of beverage pattern groups, because the mean estimate

165

for a group would not be biased by intraindividual variation in 24-hour recall intakes
(Guenther et al., 1997). In the current study, each youth was classified to a beverage
pattern group based on his/her usual beverage consumption pattern as reported in the

F FQ. Beverage patterns were defined as the combinations of beverages most frequently
consumed so that independent effects of individual beverage types could be determined
by contrasting nutrient intakes of different beverage pattern groups.

There were several limitations that pertain to the current analyses. An under-
reporting bias is a concern, because overweight individuals tend to under-report energy
more than those with normal BMI (Bandini et al., 1990; Johnson et al., 1996; Johnson,
2000). Under-reporting occurs not only in the 24-hour recall, but also in the FFQ (Subar
et al., 2003). Misclassification to beverage pattern groups due to under-reporting in
response to the FFQ was addressed by using the ‘share’ method (share 225% of total
beverage frequency). For example, the percentage share was used to reclassify “under-
reporters” assigned to the ‘other beverages’ (msj) group by the ‘frequency’ method who
reported drinking all three beverage types <1 time/d. Youth remaining in the ‘other
beverages’ group who were overweight and trying to lose weight by drinking ‘diet’ soft
drinks might be under-reporting intakes in the 24-hour recall dietary interview.

There were no energy differences between ‘milk’ and ‘sugar-sweetened
beverages’ groups possibly because of under-reporting of energy intakes.
Misclassification due to an under-estimate of sugared-sweetened beverages’ percentage
share is also a concern. Overall, regular soft drinks were reported less often in the F FQ
than in the 24-hour recall, while milk and juice frequencies were higher in the FFQ than

in the 24-hour recall. The ‘sugar-sweetened beverages’ category did not include sugar-

.166

sweetened tea or coffee, thus the percentage of calories from added sugars in the ‘other
beverages’ group was relatively high. Youth classified to ‘juice’ groups might have
mistakenly reported fruit drink frequency in response to the 100% fruit juice question,
because the fruit juice question was sequenced before the question about fruit drinks in
the FFQ. The ‘sugar-sweetened beverages’ groups drank more fruit drinks than the
‘milk’ groups in the FF Q, but not in the 24-hr recall, and the 24-hour recall intakes of
fruit drinks were higher for ‘juice’ groups than ‘non-juice’ groups. Nevertheless, the
‘juice’ groups drank juice more frequently than the ‘non-juice’ groups, the ‘milk’ groups
drank milk more frequently than ‘sugar-sweetened beverages’ groups, and expected

nutritional benefits of consuming ‘milk’ and ‘juice patterns were observed.

Implications

Because beverage choice was associated with the meal context, the beverage pattern
could be used to indicate how often youth consume different types of meals (e. g.,
breakfast, snacks, meals eaten at fast food restaurants). Higher added sugar and lower
calcium intakes in ‘sugar-sweetened beverages’ groups than in ‘milk’ groups suggest that
the beverage pattern could potentially be a risk factor for overweight. Consumption of
‘other beverages’ patterns might put youth at risk for inadequate calcium, potassium or

vitamin D intakes.

167

Chapter 5
RISK FACTORS FOR OVERWEIGHT ASSOCIATED WITH PATTERNS OF
BEVERAGE CONSUMPTION AMONG ADOLESCENTS IN THE U.S.
ABSTRACT
Objective To examine the associations of beverage patterns (i.e., the types of beverages
most fiequently consumed) with sociodemographic and lifestyle characteristics that are
risk factors for overweight among adolescents in the U.S.
Design Cross-sectional analysis of data from the Third National Health and Nutrition
Examination Survey (NHANES HI, 1988-94). Food frequency questionnaire (FFQ) data
were used to classify youth by beverage patterns. The prevalence of overweight/risk for
overweight and distributions of sociodemographic and lifestyle factors were examined
within the beverage pattern groups.
Subjects A U.S. population-based sample of adolescents age 12-16 years (n=1,872).
Statistical analysis Sample-weighted percentage distributions and standard errors were
estimated using SUDAAN software. The Chi-square test for association was used to
compare the groups’ observed and expected distributions (p<0.05).
Results Sociodemographic and lifestyle factors associated with the prevalence of
overweight/risk for overweight included race-ethnicity, SES, parental overweight/obesity,
sports team participation, TV watching, weight loss behavior, smoking, and breakfast
consumption. The prevalence of overweight/risk for overweight was lowest (17%) in the
‘milk & juice’ group, and highest (58%) in a pattern of consuming ‘other beverages’ such
as diet soft drinks, coffee and tea. A greater percentage of the ‘other beverages’ group

than the ‘milk & juice’ group were smoking cigarettes (19% vs. 2%) and trying to lose

168

weight (60% vs. 15%). A greater percentage of the ‘milk & juice’ group than the ‘sugar-
sweetened beverages’ group ate breakfast every day (57% vs. 21 %), exercised daily (35%
vs. 28%), played on 23 sports teams (37% vs. 16%), and watched TV $1 hour/day (39%
vs. 20%). TV was watched 24 hours/day by 34% of the ‘sugar-sweetened beverages’
group vs. 10% of the ‘milk & juice’ group. Beverage patterns were also associated with
income and other indicators of SES (p<0.01).

Implications Beverage patterns might be markers of sociodemographic and lifestyle
factors associated with the prevalence of overweight/risk for overweight. This suggests
that complex interrelationships between sociodemographic/lifestyle factors, beverage

patterns and adolescent obesity exist.

INTRODUCTION
Because the etiology of obesity in youth is complex, not one single associative factor,
such as soft drink consumption, has been strongly linked to the high prevalence of
overweight in youth. Rather than a single beverage, patterns of beverages consumed
together can recognize the interrelationships between a variety of food and beverage
choices. A broad view of the sociodemographic and lifestyle context in which beverage
patterns are consumed is needed to consider how dietary habits are associated with
multiple risk factors for childhood overweight. Insight into the complex interrelationship
of beverage patterns with sociodemographic and lifestyle factors might provide new
strategies to decrease energy intake and overweight in adolescents.

Over the past 20 years, the prevalence of obesity in African American and

Hispanic youth has been increasing at a faster rate than in non-Hispanic whites (Strauss

169

& Pollack, 2001). Among adolescents age 12-19 years, the prevalence of overweight is
currently 19% among non-Hispanic blacks and 21% in Mexican-Americans compared to
15% in non-Hispanic whites (Hedley et al., 2004). In NHANES III, 1988-1994, an
inverse relationship between family income and obesity was observed in adolescents of
white but not other race-ethnic groups (Troiano & Flegal, 1998; Alaimo et al., 2001;
Miech et al., 2006). An increased prevalence of overweight among low-income
adolescents may be due in part to the quality of their diets, if only less expensive, energy-
dense foods and beverages are consumed (Alaimo et al., 2001; Drewnowski & Specter,
2004; Drewnowski & Darrnon, 2005). Additionally, low-income adolescents living in
urban environments might watch TV, if they have few opportunities to be physically
active. Compounding such sedentary behavior is the fact that soft drinks are frequently
consumed while watching TV (Coon et al., 2001; Giammattei et al., 2003; Utter et al.,
2003; Matheson et al., 2004; Phillips et al., 2004). Sugar-sweetened beverage
consumption has been associated with poverty among adolescents age 12-17 years in
NHANES, 1999-2002 (Miech et al., 2006). Still, it is unclear whether disparities in
overweight prevalence are driven by income and education of parents and/or familial
food and physical activity preferences and practices (Gordon-Larsen et al., 2003).
Beverages typically consumed at breakfast are one example of a familial food
pattern that could relate to obesity. For example, parental obesity was a risk factor for
overweight in adolescents age 12-16 years, but among those having one or two obese
parents, eating breakfast some days or every day was associated with normal body weight
(Fiore et al., 2006). Breakfast skipping has been associated with poverty, and could

potentially be a risk factor for overweight among low-income adolescents (Miech et al.,

170

2006). If youth consumed both milk and orange juice frequently at breakfast, then those
youth who regularly eat breakfast might be distinguished from breakfast skippers by
identifying those who do or do not consume milk and juice in combination. In
developing strategies to prevent overweight in youth, it might be usefirl to identify
beverage patterns that could function as markers of sociodemographic and lifestyle
factors associated with the risk for overweight.

The aim of this study was to examine the associations of beverage patterns (i.e.,
the types of beverages most fiequently consumed) with risk factors for overweight in 12-
16-year-old adolescents in the U.S. Youth were classified by beverage pattern, and
population distributions of sociodemographic and lifestyle factors were examined within
these groups. Youth were classified by consumption of milk, 100% fruit juice and/or
sugar-sweetened beverages (e. g., regular soft drinks and fruit drinks) to examine the
association of beverage patterns with risk factors such as race-ethnicity, household
income, parental overweight/obesity, physical activity, TV watching and breakfast
consumption. A pattern of drinking milk and juice was expected to be associated with
breakfast consumption and other healthful behaviors, while a pattern of drinking sugar-
sweetened beverages was expected to be associated with low SES, TV watching and

other environmental risk factors for overweight.

METHODS
Dataset
The Third National Health and Nutrition Examination Survey (NHANES HI, 1988-1994)

data was used for analyses, rather than the more recent 1999-2002 NHANES data,

171

because the dietary food frequency questionnaire used to assess the youth’s typical
beverage pattern is available only in NHANES IH. The National Center for Health
Statistics (N CHS) of the Centers for Disease Control and Prevention (CDC) conducts
NHANES to obtain nationally representative data on the nutritional and health status of
the civilian, non-institutionalized U.S. population. Participants were selected using a
complex, stratified, multistage, probability cluster sample design (DHHS, 1994).

Interviews were conducted in the household and at a mobile examination center (MEC).

Subjects

Adolescents age 12-16 years who were interviewed and examined in NHANES IH
comprised the analytic sample for this study. The NHANES 111 sample included 2,216
adolescents aged 12-16 years selected for examination, and, of those, 2,079 (93.8% of
2,216) were examined in the MEC. Upon examination, 19 respondents (0.9% of 2,216)
found to be ineligible (16 teenage girls were pregnant, and three were lactating) were
excluded from the current study. Because 24-hour recall data were used in a preliminary
analysis necessary to conduct the current study, the analytical sample was limited to
adolescents age 12-16 years who completed a 24-hour recall dietary interview as well as
all beverage items in the FFQ. There were 1,997 youth (90.1% of 2,216) who had a
complete, reliable 24-hour recall dietary interview, and of these, 97 (4.4% of 2,216) were
missing beverage frequency data and another 28 (1.3% of 2,216) were missing data on
water intake. After excluding those with missing data, the remaining analytic sample of
1,872 youth (84.5% of 2,216) included 871 boys (46.5% of 1,872) and 1,001 girls (53.5%

of 1,872) aged 12-16 years.

172

Definition of beverage patterns

Youth were classified by the combination of beverage types most frequently consumed,
as reported in the FFQ. Four main categories of the 12 FFQ beverage items listed in
Table l were (1) milk, including milk on cereal as well as milk drinks, (2) 100% film
juice, (3) sugar-sweetened beverages (e.g., fi'uit drinks and regular soft drinks) and (4)
other beverages (e.g., diet soft drinks, coffee/tea and beer/wine/liquor).

Beverage pattern classifications were based on ‘usual’ intake in the FFQ of a
combination of up to three beverage types: milk, sugar-sweetened beverages (e. g., regular
soft drinks and mu drinks) and/or 100% fruit juice. To classify youth into beverage
pattern groups, the FF Q beverage frequencies were used to identify the combination of
beverages that characterized a youth by his/her ‘typical’ beverage pattern. The
combination of beverages having shares that were at least 25% of the total were identified
by expressing each beverage’s frequency as a percentage of the summed frequency for all
beverages. Notations for the eight beverage patterns were combinations of the capital
letters M, S and J to refer to milk, sugar-sweetened beverages and juice, respectively,
having shares 2 25% total frequency, and lowercase letters m, s and j represented
consumption of less than the specified amount. The beverage patterns were named
according to the eight beverage combinations: ‘all caloric beverages’ (MSJ), ‘milk &
sugar-sweetened beverages’ (MSj), ‘milk & juice’ (MSJ), ‘milk’ (Msj), ‘sugar- sweetened
beverages & juice’ (mSJ), ‘sugar-sweetened beverages’ (mSj), ‘other beverages & juice’

(msJ) and ‘other beverages’ (msj).

173

Table 1. Beverage categories and beverage subcategories corresponding to beverage

items in the NHANES HI food frequency questionnaire (F FQ)

 

Beverage categories &

beverage subcategories

FFQ beverage items

 

1. Milk and milk drinks,

chocolate & regular

2. 100% fi'uit juice,

citrus & non—citrus

3. Sugar-sweetened bevg.
33. Fruit drinks
3b. Regular soft drinks
4. Other beverages
4a. Diet soft drinks

4b. Coffee/ tea

4c. Alcoholic

beverages

1a. How often did you have chocolate milk and hot cocoa?

1b. How often did you have milk to drink or on cereal?

Do not count small amounts of milk added to coffee or tea.

2a. How often did you have orange juice, grapefruit juice
and tangerine juice?

2b. Other fruit juices such as grape juice, apple juice,
cranberry juice, and fruit nectars?

3a. How often did you have Hi-C, Tang, Hawaiian Punch,
Koolaid, and other drinks with added vitamin C?

3b. Regular colas and sodas, not diet?

4a. Diet colas, diet sodas, and diet drinks such as Crystal
Light?

4b (1). Regular coffee with caffeine?

4b (2). Regular tea with caffeine?

4c (1). Beer and lite beer?

4c (2). Wine, wine coolers, Sangria, and champagne?

4c (3). Hard liquor such as tequila, gin, vodka, scotch,

rum, whiskey and liqueurs, either alone or mixed?

 

174

BMI and criteria for overweight in youth

Weight and height were measured at the MEC to assess Body Mass Index (BMI=kg/m2),
and youth were classified by weight status based on their BMI. The SAS program
available from NCHS was run to calculate the youth’s BMI percentile using the gender-
specific BMI-for-age CDC Growth Charts reference data (NCHS; Kuczmarski et al.,
2000). The BMI percentile was used to classify youth by weight status (i.e.,
underweight: BMI <5th percentile, normal body weight: BMI 25th percentile and BMI

<85th percentile, at risk for overweight: BMI 285th percentile and <95th percentile, and

overweight: BMI 295th percentile).

Potential risk factors for overweight

Sociodemographic factors examined were gender, age, race-ethnicity, region,
urbanization, parental overweight/obesity, household income and other indicators of
social economic status (SES). Lifestyle factors included smoking, alcohol use, weight
loss behavior, physical activity, TV watching, frequency of breakfast consumption, and
type of milk (e.g., whole, 2%, 1%, skim) usually consumed.

Categories for. race-ethnicity included non-Hispanic white, non-Hispanic black,
Hispanic and other race-ethnic groups. Four broad regions of the U.S. included the
Northeast, Midwest, South and West. The urban/rural classification was a metropolitan
area having a population 21 million vs. all other areas. The poverty income ratio was the
ratio of household income to the poverty level for household size in an urban or rural
area. Other indicators of SES included the family size and characteristics of the

household head or family reference person (FRP), such as their gender, age group,

175

 

marital status, education and employment status. BMI was calculated from reported
weight and height of the youth’s mother and father to determine whether both or one of
two parents were overweight (BMI 225 kg/mz) or obese (BMI 230 kg/mz).

The survey included lifestyle questions, such as those asking youth whether they
currently smoked cigarettes or drank alcohol in the past year. Youth were also asked
whether they were currently trying to lose weight and how often they ate breakfast (e.g.,
every day, some days, rarely, never, weekends only). The FFQ included a question
asking whether whole, 2%, 1% or skim/nonfat milk was usually consumed. Physical
activity was assessed using the following questionnaire items asking youth about their
exercise habits and participation in team sports:

1. How many times per week do you play or exercise enough to make you sweat

or breathe hard?

2. In the past year, how many sports teams or organized exercise programs have

you been involved in (not including physical education or gym classes)?
Youth were asked twice, once in the household and again at the MEC, about time spent
watching TV the previous day, and the average hours/day in two days was used in the

current study.

Statistical analysis
The Statistical Package for the Social Sciences, SPSS for Windows, Version 12.0 (SPSS
Inc. Chicago, IL, USA) was used to prepare the dataset for analysis, while all statistical

analyses were performed using SUDAAN version 8 (Research Triangle Institute,

176

Research Triangle Park, NC), to account for the effect of the complex sampling design
while conducting statistical tests.

The prevalence of overweight risk and overweight (BMI 285th percentile) was
estimated for youth classified by categories of each sociodemographic/lifestyle factor,
and the percentage distribution of each sociodemographic/lifestyle factor was estimated
for all youth and for each beverage pattern group. The percentage of each beverage
pattern group classified by weight status (e.g., underweight, normal, at risk for
overweight, overweight) was estimated as well. The Chi-square test for the association
of categorical variables was performed to detennine whether the observed and expected
distributions were different (p<0.05).

The six-year MEC-examined sample weights were applied to the calculation of all
population parameters, such as the percentage of youth who were overweight. The
sample weights account for unequal probability of selection, non-coverage and non-
response and adjust the data to the 1990 U.S. Census figures, therefore the estimates were
representative of adolescents in the 1990 U.S. population. The linearization (Taylor
series) method was used in SUDAAN to account for effects of the clustered sample

design in estimating the variance of population parameters (Shah et al., 1997).

RESULTS

Risk factors for overweight

Table 2 shows the distribution of individual/household sociodemographic and lifestyle
characteristics among all adolescents age 12-16 years, and the association of these

characteristics with the prevalence of overweight risk and overweight. Sociodemographic

177

Table 2. Distribution (% of adolescents aged 12-16 years in categories) of

sociodemographic and lifestyle characteristics, and the prevalence of overweight!
risk for overweight among adolescents classified by these characteristics2

 

 

 

 

 

 

 

 

Distribution of Prevalence of Overwt/

Characteristic Sample Characteristic Risk for Overwt
Category Size (n) Percent t SE Percent :1: SE

All adolescents age 12-16 years 1,872 100.0 :1: 0.0 27.3 i 2.1

** Weight status

<5 pctl 45 2.4 :t 0.5 0.0 :t 0.0
5-84 pctl 1,236 70.4 :1: 2.0 0.0 :1: 0.0
85-94 pctl 298 16.5 :1: 1.4 100.0 :1: 0.0
295 pctl 258 10.7 :1: 1.4 100.0 :1: 0.0
Missing 35

Gender
Male 871 50.5 :1: 2.1 26.5 :1: 2.6
Female 1,001 49.5 :t 2.1 28.0 :1: 2.4
Missing 0

Age (yr)
12-13 773 39.4 :1: 1.6 28.2 i 1.8
14 369 22.4 :L- 1.9 32.0 :1: 3.8
15-16 730 38.2 :1: 1.7 23.6 :1: 3.3
Missing 0

** Race-ethnicity

Non-Hisp. White 492 67.1 :1: 2.4 26.9 :t 2.9
Non-Hisp. Black 671 14.9 :1: 1.4 29.8 i 1.7
Hispanic 670 12.6 d: 1.7 35.1 d: 4.8
Other 39 5.4 i 1.0 6.0 :1: 3.2
Missing 0

Region of the U.S.
Northeast 212 19.5 :1: 2.1 24.1 i 2.7
Midwest 339 21.8 :1: 2.0 26.7 :t 4.7
South 834 34.7 i 3.2 27.0 :1: 3.9
West 487 23.9 i 3.8 30.8 i 4.1
Missinfig 0

Rural/urban code
Metro areas 873 46.6 :h 5.3 25.9 d: 2.6
All other areas 999 53.4 d: 5.3 28.5 :t 2.7
Missing 0

 

Table 2, cont'd.

 

 

 

 

 

 

 

Distribution of Prevalence of Overwt/
Characteristic Sample Characteristic Risk for Overwt
Category Size (n) Percent d: SE Percent :t SE
** Parent’s weight status
2 overwt/obese 759 33.6 i: 1.9 39.7 i 3.3
1 overwt/obese 621 40.9 :t 2.2 22.8 i 2.9
Both normal 386 25.4 i 1.7 17.6 :1: 2.8
Missing 106
Family ref. person (FRP) gender
Male 1,149 69.4 :1: 1.9 26.4 :1: 2.4
Female 707 30.6 i 1.9 28.2 i 3.2
Missi_ng l6
FRP age (W)
15-29 72 2.9 i 0.9 33.7 i 9.1
30-39 834 45.6 i 2.2 28.4 :t 2.8
40-49 695 41.2 d: 2.3 22.7 d: 2.4
350 250 10.3 i 1.2 33.9 :1: 7.3
Missflg 21
FRP marital status
Married 1,162 69.0 :1: 1.9 26.1 :t 2.4
Not married 706 31.0 :1: 1.9 29.8 :t 3.2
Missing 4
** FRP education
<9th grade 406 11.0 :1: 1.4 32.9 i 5.3
9-11th grades 341 12.8 d: 1.3 35.3 :1: 4.8
HS grad 618 35.5 :1: 2.0 29.5 :t 3.8
Some college 502 40.7 i 2.4 21.3 :1: 3.5
Missing 5
FRP employed
Yes 1,352 83.2 :t 1.5 27.1 i 2.5
No 424 16.8 i 1.5 29.8 d: 4.1
Missing 96
** Poverty Income Ratio (PIR)
0-130% 808 29.1 :1: 2.2 34.9 :1: 2.9
131-185% 239 14.4 i 1.6 28.6 :1: 5.5
186-350% 473 35.0 :h 2.6 26.5 :h 3.8
>350% 197 21.5 3: 2.7 15.3 i 4.2
MissinL 155

 

179

Table 2, cont'd.

 

 

 

 

 

 

 

Distribution of Prevalence of Overwt/
Characteristic Sample Characteristic Risk for Overwt
Category Size (n) Percent :1: SE Percent i SE
Family size (# persons)
1-3 360 20.7 d: 1.7 32.2 i 3.9
4-5 926 57.8 i 2.4 27.6 :1: 2.8
26 586 21.6 d: 2.5 21.5 :1: 3.1
Missing 0
** Trying to lose weight
Yes 351 18.5 :1: 1.6 57.7 d: 4.7
No 1,457 81.5 i 1.6 20.2 a: 1.8
Missing 64
Exercise (times/wk)
$2 428 20.7 :t 1.5 30.4 i: 3.6
3-4 421 21.9 i 1.8 30.1 d: 3.6
5-6 528 27.6 :1: 2.2 21.7 :1: 2.6
27 434 29.9 :1: 2.4 27.1 :t 3.9
MissinL 61
** # sports teams past yr
None 775 38.2 :1: 1.9 37.0 :1: 3.3
1-2 735 40.8 :1: 1.9 23.6 :1: 2.8
23 301 21.0 :1: 2.1 15.3 :1: 2.8
Missing 61
** TV watched (hr/d)
$1 394 26.7 i 1.8 24.8 :1: 3.6
2-3 891 49.0 :1: 2.1 22.5 :1: 2.5
24 587 24.3 :t 1.4 39.7 i 3.4
Missingfi 0
** Smoke cigarettes
Yes 81 6.8 :t 1.0 40.3 i 7.3
No 1,730 93.2 i 1.0 26.0 :1: 2.3
Missifl 61
Alcohol past yr
Yes 191 11.9 :1:1.1 25.5 a: 5.4
No 1,608 88.1 i 1.1 27.3 i 2.4
MissinL 73

 

Table 2, cont'd.

 

 

Distribution of Prevalence of Overwt]

Characteristic Sample Characteristic Risk for Overwt

Category Size (11) Percent t SE Percent :1: SE
** How often eat breakfast

Every day 681 41.5 :t 2.2 22.9 d: 2.9
Some days 617 33.0 d: 1.8 24.4 i 3.0
Rarely 246 13.9 :1: 1.3 35.7 :1: 5.4
Never 46 2.5 :t 0.8 26.7 d: 10.4
Weekends only 221 9.1 i 0.9 42.6 :1: 6.5
Missing 61

Type of milk used
Whole/regular 879 36.8 :1: 2.2 26.0 i 2.6
2% 568 43.4 i 2.0 30.1 i 2.7
1% 63 6.0 i 1.2 17.2 :1: 7.7
Skim/nonfat 155 11.5 :1: 1.6 24.9 :1: 6.4
Never drink milk 63 2.4 i 0.6 42.2 :t 12.2
Missing 144

 

' Source: NHANES 111, 1988-1994, adolescents aged 12-16 years, with complete,
reliable 24-hr recall dietary interview, and no missing or unknown responses to
beverage items on F FQ, excluding pregnant or lactating females (n=1,872)
Percentages and standard errors are sample-weighted and estimated by linearization
method of SUDAAN.

2 Overweight/risk for overweight defined by BMI 285th percentile of BMI-for-age

** p<0.01, * p<0.05, by Chi-square test for association between characteristic and
prevalence of overweight/risk for overweight

181

groups more likely to have a BMI 285th percentile included Hispanics and adolescents of
low SES. Education level of the household head and income were both inversely
associated with the prevalence of overweight risk and overweight. Adolescents whose
parents’ were both overweight/obese were more likely to be at risk/overweight
themselves.

Lifestyle factors potentially associated with lower BMI included physical activity
and breakfast consumption. Sports team participation was inversely associated, and TV
watching, positively to BMI. Exercise did not have a linear relationship with overweight;
22% of those exercising 5-6 times/week were at risk/overweight, and the prevalence was
higher among those exercising both <5 and >6 times/week. A positive association
between overweight and weight loss behavior was highly significant. The prevalence of
overweight risk and overweight was 60% among those trying to lose weight. Smoking
was associated with overweight risk and overweight. Those who ate breakfast every day
or some days were less likely to be at risk or overweight than those who ate breakfast

rarely or on weekends only.

Risk factors associated with beverage patterns

Profiles describing characteristics of beverage pattern groups were based on the
distributions of sociodemographic and lifestyle factors shown in Table 3. The
percentage of a beverage pattern group having a characteristic was highlighted in this
table using arrows pointing upwards or downwards when the distribution within that

group was disproportionate relative to distributions in other beverage pattern groups.

182

 

dd H dd dd H N.N~ e v.“ H 5N fin H Rd m.N H Wm fin H md w; H w.m w.m H md .550
as H wdd EN. H N.d_ e d.N H dd— hd H d.NN e d.N H wd~ fim H dd; xi; H vd 9m H odd 6 653$:
Wm H d.: w.m H m.m_ d.N H fl: d.N H déN 6 mg H 5.: m.N H Q2 a; H Q: dd H N.mN e xoflm %wmcoz
Nd H Wm“. e md H d.mm Wm H ddd 6 m6 H 5.? w.m H d.NN. F dd H 53 6 Wm H d.Nh e d.N. H Ndv 83>? 35:62

 

bEEfioéomm H...

 

 

 

 

 

 

3 H «.3 e 3 H N2 2. S. H Sq e 3 H 5.? mm H QR M3 H 3m 3 H 33 3 H as. e 2.:
E H 5.8 2 H MEN 3 H SN 3 H 3m e 3 H SN e mm H on mm H 5.: 3 H 3 E
3 H 3m 5.5 H g 3. H EN 3 H 5.8 3 H 5% e 3 H can e mm H w? e 2 H 93. 2-2
dd 8< ..
E H H8 2. mm H is e 3 H man 3 H 3m 2. mm H ”.5. 3 H 3m 3 H 33 3 H me. 238
E H 8.8 an H 2m 9m H New 3 H 2:. 3 H 3m 5. H 8.8 mm H N5 9 3 H 5% e 222
.5680 .3."
3 H 0.8 e E H H2 e 3 H 2: 3 H 5.2 e 2 H 2 3 H «.8 a 2 H a: e S H 8.2 e =8 mom
2: H ”.2 e 3 H we e 2 H 8.: e 3: H 52 mm H 3: e 3 H 5.: H mm H 8.2 3 H 4.: a :8 88w
«.8 H ”.3 a 5% H 35 3 H :5 5.5 H n: we H m8 3 H «.8 e E. H 2K 3 H 0% =8 33
5.5 H 5.5 3. H 0.5 I H mm 3. H 3 2 H mm 2 H 3 3 H 5. mm H on :8 mv

 

mfifim Ems? H...

 

26:28.20 mimoEoEamaoEmuoBOm do 203.8 5 3:88.86 do Em H 2:60.613 ommucoemd

 

 

Gm u 5 S: n 5 <8 u 5 as: n 5 d S n 5 38 u 5 a? u 5 as n 5 .6888
.25 25 $5 25 .32 22 .52 a: 058385
w>om 632. a. w>om w>om 622. a. w>om 0:32 623. .w>om Em w>om etc—mo
850 850 Emémam Emémsm a. ea: .83 a 52 5.

 

~95on 583m omfio>om

 

mama» dTNd own 358302.. do 3:on Egan omfiokfi GEE» mozmtfiofifio 2.98m: 98 ofiqfiwofiodeog do couzflbma .m 655,—.

183

 

Nd" H ddm fie H m.—N m6 H fidm Wm H ham Wm H Yam :H H m.dm d.m H fidm >6 H ddv e 0:“th

 

 

«.2 H Woe fie H m.wh e mé H add v6 H m.dh Wm H ddh :q H mdd d.m H add 5d H d.~m 232
HEWHE
m6 H 02 d6 H fimN #6 H déN dd H d.N.N 9m H fivm fin H N.mN dd H d.wN dd H 9mm Echo: 50m

m.d_ H 06¢ 6 9w H Ndm Nd H 1mm NH H m.dv in H fidm d.m H 0.2» mé H m.wv 6 d6 H ddm 83703393 _
5.3 H dém Wm H fimm Wm H ddv e v.0 H d._m 5v H wdm e w.m H N.mm d.N H m.m~ 5d H mu? 9 omonobzcgo N

 

mash
55$ H.553 H...

 

d.: H 5N0 dd H Q? 56 H d.md F m.w H w.mv m6 H fidm mg. H d.mv dd H v.3 w.w H flow 6 93.8 .550 =<

 

 

184

 

d.: H mfim dd H 9mm 56 H d.mm Wm H New 6 m6 H d.dm 2. H dfim 6 90 H ofiv w.w H ddm 88a 882
38

gngdfiam

Nw H «2mm md H 5mm 6 Wm H 5.: ms H fimm e fim H 03 Wm H mdm SH H flaw w.w H fivm «3.3
cd H m.mm e d.w H mém d6 H flow 6 ms H mdm 0v H ENm do H Ydm w.m H 5mm wd H d.m~ 550m
06 H 02 m6 H 3: d6 H d.NN v6 H 5.3 d.m H 5.3 Om H QB Wm H m.w~ e We H ENN $0322
Wm H d6 Nd H .3. En H 5mm 9 m6 H m.N_ dd H m.d~ 5.5 H m.wm e Nd H .3: m6 H Nmm 9 $85.82
.m.D

2: 5 sewed t.

 

ozmtodofimco mEmotdoEQEmoEmueoom do dowofio :_ 3:83.03 dkom H «520$ mascoeoa

 

 

an n 5 E: u 5 :2 u 5 2% u 5 d s. u 5 3% u 5 a? n 5 as u 5 935
.28 7:: .88 3,5 .32 22 .52 22 2358ng
w>om 023. a. w>om w>om 022. a. w>om v52 3:: .m>om Em w>om 2.8—«U
550 550 Emémsm Emémsm H 52 -awsm a. 5:2 .2

 

$380 Eotam Hwy—96m

 

6.58 .m 03¢

 

 

 

 

 

 

 

 

 

 

 

 

”@580 2.8th owfigom

3 a 3 3 a 3 3 n." 2: 3 a E: 3 a Q: 3 3 3.3 3 a 3. 3 a non e oz
3 a 3m 5 3 a 3.8 e 3 « 3w 3 a n; 3 a 3m 3 a 3m 3 a 3m 3 a 9% 8>
@32an Eu
3 a 33 a 3 a 33 3 a 3m 3 a 33 e 3 a 33 3 « Em e 3 « 3m 3 a 33 a n3.58 28m
3.: a 3m 3 a 3m 3 a 3.9. e 3 a 33 3 a 3m 3 a 33 3 « 3m 3 a 33 2:» mm
3 a 2: 3 a 35 3 a 3_ 3 a 3.2 M: a 3 S a 2 3 a 3_ e 3 a 5.: 825 52$
3 a 3_ 3 a 3 3 a a: 3 a 3: 3 a 5.: S a 3 3 a 5.: 3 a 02 $3 3v
Gaga—duo nan—m 3..
3 « gm 3 a 33 3 a 3m 3 a 33 3 « 33 3 a 33 3 a 3m 3 a 5.3 e 8535 82
3 a 35 G a 3.: e 3 a 35 3 a w? 3 a 9:. 3 a n: 3 a 35 3 a 3m Bra:
33%
355 mac
3 a 33 e 3 a 3 .2 a 3 3 « 3.: e 3 a 3.: 3 a 3 3 a 3: 3 a 3 on
3 a 2m 3 a 33 3 a 33 E a 3m 3 a 33 3 a 3.3 3 a 3m 3 um 3m 38
n: a 23 3 a 33 3 a N3 3 a 33 3 um 33 3 a 33 3 a 33 3 a 3.3 e 38
S a 3 3 a no 3 a 3 3 a 3 2 a 3 E a 3 3 a 3 e no a 3 3-2
A5 awn «an ...
25:23.20 EbmoEoEafigoquOm n.o bouxofio 5 3:88.08 .0 Gm H 2.020% mdmgeoa
3 u 5 E: u 5 3n n 5 :3: u 5 2 3 u 5 :3 u 5 63 n 5 as u 5 338
we 25 .88 as .32 2.2 $2 32 0558829
m>om 032. a. 96m w>om 022. Q w>om 5:2 032. .w>um Em w>om 3.5—«U
550 550 Eméwam Eméwsm a 5:2 .555 a. 55 3

 

3.58 .m 255

185

 

 

 

 

 

 

 

mg: H 2:. fix H fimn in H Wow _.w H v.2. Qm H wéw Qm H Qmw Wm H new msm H m.mw 6 oz
3: H 5% 9 fix H QR e in H :2 fie H QVN e Qm H ~23 Qm H Q3 Wm H 92 Qm H m.: 8%
Haws?
82 8 march.
Qv H 9: Qo H QNN Nd H «.2 N6 H :N 6 5m H 33 e fie H m.w~ e Qm H 5.3 5v H 03 0N
0.2 H v.3 ms H wdm Qm H mém m6 H mac Qm H #8 Wm H me 3V H Qmm >6 H v.9 WV
2: H Q2” 9 E. H mdN fin H 03 QN H 5.2 ~.N H ".3 Wm H 02 Qv H m.mN 6 m.\. H m.—N m;
438qu

5 8: 55$ .1.
Qm H 3 9 Wm H OR 6 Qm H de Qc H Yon Wm H w.m~ 6 N6 H :h .._‘ QM H QE 4. mg. H Qom °\oommA
2: H N? 6 m8 H Qmm a v6 H Qom Qm H me Qv H QwN N6 H mém Wm H 33 9 Q5 H mam Samméw—
Qw H w.mm fiw H NB Wm H v.2 Wm H 92 9 EN H QC 9 m.m H 3: Nam H QQ Wm H m6 $373”—
.md H QR Wm H QNm _.m H wém #6 H Ném e Qm H m.w~ wé H 9mm Wm H Dom ms H 03 6 $025
9an 23

 

26:98.65 mgggofawmmEmuofiom E as 5 3:33.03 21mm H 285% $9583

0885 €26.“ 3.,

 

3 n 5
.3 E
w>om

550

S: u 5
a m E
033. a. w>om

850

A3 u 5
.5 a
w>om

Emfiawnm

A3: n5 2:35
H m a .3 2
SE a. mam Va:
Emfiwmsm

33 u 5

a m 2
023.

a. V52

3 n 5
.3 m 2
.w>om Em
.356 a. 55

as n 5
a m E
w>om etc—a0

=<

bowouau

 

 

3985 Satan owfigum

0853820

 

.380 .m 033

186

 

Qw H MAw YN H Woo QN H wa Qo H v.3 e QM H v.8 NA H Qwa e Nd H vdo Wm H Ema oz

 

 

 

 

 

 

187

 

QM H Qw~ e in H QM QN H fl: 6 Qo H Q_ Q_ H Qo N; H QN Nd H QB Wm H MS I 3.x.
838an 838m *

2. H v.wM e QM H M.MM e wé H QMM 6 Q0 H QMM e QN H QE 9 MN H Q3 4. HM H flew M.m H M.©N VN
Q5 H NdM M.© H Qom fim H Q? _.n H QMM mM H Qmm e Qm H MAM QM H QMm e 5.5 H Qwv M-N
in H QNN NM H _.MM 6 v6 H vdm a 5.5 H NMN QM H QwN M.m H QMM e QM H v.8 a MS H NMN W
35

3:895 >.H .2.
5.5 H fifi 4, Q5 H QNN QN H Q3 QM H me e QN H Q2 4, v.5 H QhM e M.M H v.2 QM H QMN 6 MN
M6 H QNN 4, MS H QaM M.m H Name If M6 H QwM QM H de Md H 3% e QM H v.3 ".w H N.NM NA
QM H _.mo 6 MN H NMM Qv H QMM 56 H QNM Qv H Q3 9 QM H Q3 a “M H NdM md H de 282
b» «man

338 35% 5 3..
Q0 H Q3 Qc H v.3 Nd H M.wN NS H QMM QM H QNM QM H vdM e QM H Qua v6 H QMN hN
v.5 H VAN M.m H flow Nd H NMN fim H wdm M.M H NVN Wm H ~QNM v6 H QMM M6 H wém 9m
Md H finM e QM H QNM e Qv H M.NN MN H v.5 A, QM H QMN QM H Qom N.N H Q3 Qa H QMN 7M
m6 H Q~N Q5 H me e QM H Mdm e wd H «MN 6 QM H Q3 QN H Q: A. QM H Mdfi Em H NAN NW
c.2885

DmMOhOKm a. *

 

ozmtmfiflmco oEmmEoEafiQoEmcoBOm ho bodmfio E £58203 ho Em H E853 mmflcmema

 

 

Gm u 5 E: u 5 ARM u 5 A3: u 5 a S u 5 a: n 5 a? n 5 as u 5 $330
.28 7.5 $5 $8 .32 :2 $2 a: 055630
m>om 023. a. w>om w>om 033 Q w>om x32 33.. .w>om gm w>om 3.830
550 550 Emémam Emémsm a. é: 55% a. “a: =<

 

Nazca 583m owfigom

 

.380 .m 05¢

 

Md H Qw m; H QN QM H QC Qv H Md C.C H CC CC H QC C.C H CC CC H C.C VEE #58 852
CC H C.: 6 CS H Q3 6 QN H QC QN H C.C M.M H v.2 wé H QE NM H Q2 Nd H Qv A, «flaccgfim

 

Qv H QM SH H M.M v; H QC QM H QM QN H 2. 6 WM H C.C_ e QN H QM QM H vé o\o_
QC H #3 M.M H «.3 QM H M.MM QC H M.MM QM H QDV e M.M H fiwv 9 N6 H M.Mv S H va cam
QC H QQ M.M H M.MN QM H QCm e QC H Q? 6 Qv H NMM QM H v.3 QM H QMM : H va 83383—95,

Com:

use we 25 :

 

QM H wé Qm H QNN e M.M H Q: Cd H Q3 .1 H QC CA H 3‘ M._ H QB M.M H M.w baa 388—83

 

 

a5 H 3 No H 3 3 H 3 e 3. H 2 2 H 3 Ho H 3 .3 H E 3. H 2 532
es H EN 9 3 H EN 9 2 H as e 3 H n: _.m H w: 3 H E 3 H .2: 3 H 3 boss
3 H as E H 3m 3 H in e .3 H 93 e 3 H 3; 3 H 5m 3. H cam 3. H 3:. e 2% oeom
m: H 93 e to H 5.8 3 H 2N a S H new M? H as“ 3 H H? e on H 34 E H 5.3 E. 88m
awnings
«no Etc 303 I

S H Eh _.m H :a e 3 H 3w E H 3% mm H 0% E H a? e 2 H mam Hm H Sm oz
3 H 3N e 3 H 3 2 H as E H as 3 H n: E H 3 ad H N: E H ms HS

 

tn “man 8834 H

 

2888820 oEmeoEaEmeoceoom 8 bodmfio 5 3:88.08 8 Amw H 8883 $980th

 

 

an n 5 as u 5 :2 u 5 as u 5 C :V n 5 a3 u 5 a? n 5 as u 5 835
.28 :8 .88 Be .32 22 $2 32 055825
w>om 008—. Q 96m w>om 032. Q m>om Va: 032. .w>om Em w>om otoBU
550 E8 Emémam .5955 a. 55 .i5 H vs: =<

 

 

N539:0 88th owa8>om

 

.958 .m 22¢

188

80000 030050 C00 05:80.880 8003000 80000000000 00C 0000 000000.30 3 .mQCva * JCCVQ .i
600: 20:08.0: 05 £000: 003 3MB 008a 008000.30 >288 05. H
8:080 02.08000 05 :05 000— 00058 800.0800 M 000 m .8 8030—
00000032 000 $000000C 0w80>0p :33 C0 $3M 08:0 0 wig: 0000 5030009000 .008“ 0:0 m0w80>0n “0080000309830 .058 C0
00x88 00 0300 M 05 m .2 80:3 3800 .93 05 8 6080000 b80000b 0008 m0w80>00 C0 020080800 05 mm 80300 088030 05. H
.Z<<QDm m0 C0508 000008008— 3 “00008300 000 8080 800080 000 00280303800 0.8 0080800000 .05 8002
Gum. W5 00380.0 @580”: 00 8838 862008 .95 :0 m80: 098260 00 008090.08 85058: .5 9:38 00

C00 .3030008 506 C0000 Eém 030200 6003800 :23 800% C72 00mm 080002200 €92.33 ,5 mg 008cm _

.958 .m 05¢

189

‘All caloric beverages’ (MSJ)

The ‘all caloric beverage’ (MSJ) pattern was associated with overweight; 16% of
youth classified by this beverage pattern had a BMI 2.95th percentile of BMI-for-age.
Those consuming a combination of all three caloric beverages (milk, sugar-sweetened
beverages and juice) were more likely to be older, non-Hispanic black or Hispanic, male
adolescents living in rural areas or in the Northeast region of the U.S. The ‘all caloric
beverage’ (MSJ) pattern was associated with low SES; household heads were more likely
young, unmarried or female, less likely to have a college education, less likely to be
employed in the past 2 weeks or more likely to have income $130% poverty. For many
youth in this group, both parents were overweight or obese. Although this beverage
pattern was not associated with exercise or TV watching, it was associated with sports
team participation; many youth in this group played 23 sports (28%). The ‘all caloric
beverage’ (MSJ) group tended to consume breakfast only some days. The group was less
likely to use skim/nonfat milk.

‘Milk & sugar-sweetened beverages’ (MS j)

The ‘milk & sugar-sweetened beverages’ (MSj) pattern was moderately
associated with overweight; although the majority of youth classified by this beverage
pattern had normal BMI (70%), 12% had a BMI 295th percentile of BMI-for-age.
Although it was less likely that both parents’ were overweight or obese, one was. Those
consuming both milk & sugar-sweetened beverages were more likely to be younger,
male, Midwestern, non-Hispanic white adolescents of moderately hi gh-income
households (186-350% poverty). Because a disproportionate number of household heads

were age 15-29 years and not graduated from High School, it seems that some youth were

190

independent fiom their parents and living with other young adults. The youth in the
‘milk & sugar-sweetened beverages’ (MSj) group tended to watch TV 2-3 hours/day, but
otherwise physical activity and breakfast consumption levels were normally distributed.

‘Milk & juice’ (MsJ)

The ‘milk & juice’ (MSJ) pattern was associated with a low prevalence of
overweight risk and overweight (17%); more than 80% had normal body weight. An
even number of boys and girls were classified by this beverage pattern. Those
consuming both milk and juice tended to be younger, non-Hispanic white adolescents
living in the Northeast or in urban areas of the U.S. SES was high; 57% of household
heads were college-educated and 27% were wealthy (>350% poverty). The majority of
youth in this group ate breakfast every day (57%). The group usually drank 1-2% milk.
Other healthy characteristics of the ‘milk & juice’ (MSJ) group included a very high level
of physical activity; many exercised 27 times/week (35%). Most adolescents in the ‘milk
& juice’ (MSJ) group participated in team sports (81%) and many played 23 sports
(3 7%). Few watched TV 24 hours/day (10%) and a high proportion watched TV _<_1
hour/day (39%). In addition, these youth were less likely to smoke cigarettes (2%) or
drink alcohol (4%).

‘Milk’ (MSj)

Some youth in the ‘milk’ (Msj) group were at risk for overweight (19%), but the
group tended to have an average distribution of many risk factors for overweight. Few
characteristics distinguished this beverage pattern from others. Gender groups were close
to having equal numbers. Like the ‘milk & sugar-sweetened beverages’ (MSj) and ‘milk

& juice’ (MSJ) groups, the youth in this group tended to be younger and non-Hispanic

191

white. The ‘milk’ (Msj) pattern represented both low- and hi gh-income groups. There
was a tendency for some youth to have two overweight/obese parents (37%). Like the
‘milk & juice’ (MsJ) group, the ‘milk’ group also tended to use 1-2% milk, but team
sports participation was low (54%) and many watched TV 2-3 hours/day (54%).

‘Sugar-sweetened beverages & juice’ (mSJ)

Like the ‘milk & sugar-sweetened beverages’ (MSj) pattern, the ‘sugar-sweetened
beverages & juice’ (MSj) pattern was moderately associated with overweight; 12% had a
BMI 2 95th percentile of BMI-for-age, and many were trying to lose weight. This
beverage pattern represented non-Hispanic black or Hispanic, 14-year-old, females living
in urban areas or in the West. SES was moderate; half of the household heads had a
college education (50%). Nevertheless, for 52% of the group, income was low ($185%
poverty). Some adolescents might have been living in multi-generational family units;
household heads age 50 or older were over-represented in this group and family size
tended to be large. Youth in this group tended to exercise either a lot (25 times/week) or
very little (.<.2 times/week), perhaps because this beverage pattern was associated with
both sports team participation and TV watching. Large proportions of youth in this group
played 23 sports (29%) or watched TV 24 hours/day (33%). The group tended to
consume breakfast only some days, and many youth in the group drank whole milk.

‘Sugar-sweetened beverages’ (mSj)

The prevalence of overweight risk/risk was 30% in the ‘Sugar-sweetened
beverages’ (mSj) group. This beverage pattern represented older non-Hispanic white
adolescents living in the Northeastern, Southern or rural areas of the U.S. SES indicators

were normally distributed within this group. Many youth in the ‘sugar-sweetened

192

beverages’ (mSj) group (27%) exercised infrequently ($2 times/week), although 46% of
the group played 1-2 sports per year. Most youth watched TV for 22 hours/day (80%).
There was a very low rate of regular breakfast consumption in this group (21%). A high
percentage of youth in the group drank whole milk (50%).

‘Other beverages & juice’ (msJ)

The ‘other beverages & juice’ (msJ) pattern was associated with overweight/risk
and weight loss behavior; 32% had BMI 285‘h percentile and 28% were trying to lose
weight. The group consisted mostly of 15-16-year-old girls representing Hispanics and
other race-ethnic groups living in the Western region of the U.S. SES was high;
household heads were mostly male (79%), married (77%), employed (91%) and 28%
were wealthy (>350% poverty). A low rate of alcohol use (6%) distinguished this group
from the ‘other beverages’ (msj) pattern discussed below. Most youth watched TV either
:1 hour/day (3 7%) or 24 hours/day (33%). There was a low rate of regular breakfast
consumption in this group (26%); many ate breakfast rarely (23%) or on weekends only
(23%). Many youth in the group drank skim milk (18%).

‘Other beverages’ (msj)

The ‘other beverages’ (msj) pattern was highly associated with overweight/risk
for overweight and weight loss behavior. Of all the beverage pattern groups, the ‘other
beverages’ (msj) group had the highest prevalence of overweight/risk for overweight.
Those with BMI 285th percentile were 5 8%, and 60% were trying to lose weight. The
majority had parents that were overweight or obese (65%). Most were non-Hispanic
white girls age 15-16 years. SES was moderate; household heads tended to be age 250

years, most were employed (93%) and incomes were high (186-350% poverty), although

193

few were college-educated (29%). Many lived in the South (55%), so the ‘other
beverages’ consumed might have been sweet tea. There were high rates of smoking
(19%) and alcohol use 924%). Coffee and other caffeineated beverages were might have
been consumed as well. Those who drank diet soft drinks might have been also
exercising 3-4 times/week in their efforts to lose weight. Few were involved with team
sports (65%) and TV was watched 24 hours/day (3 8%). Although the majority ate
breakfast every day (52%), another large proportion ate breakfast rarely (28%). Many

youth in the group drank skim milk (18%).

Summary

Sociodemographic and lifestyle factors associated with the prevalence of overweight/risk
for overweight included race-ethnicity, SES, parental overweight/obesity, sports team
participation, TV watching, weight loss behavior, smoking and breakfast consumption.
Two caloric/sugar-sweetened beverage patterns associated with a high prevalence of
overweight, the ‘caloric beverages’ (MSJ) pattern consumed more often by
black/Hispanic boys and the ‘sugar-sweetened beverages & juice’ (mSJ) pattern
consumed more often by black/Hispanic girls, were also associated with low income and
other indicators of low SES. Two different sugar-sweetened beverage patterns consumed
more often by non-Hispanic whites, the ‘milk & sugar-sweetened beverages’ (MSj) and
‘sugar-sweetened beverages’ (mSj) patterns, were associated with parental
overweight/obesity and TV watching. The ‘milk & juice’ pattern was positively
associated with SES, breakfast consumption, exercise and sports team participation; and

inversely associated with TV watching and overweight/risk for overweight. The ‘other

194

beverages’ group had the highest prevalence of overweight/risk for overweight, and this
pattern was positively associated with weight loss behavior, smoking, TV watching and

inversely with sports team participation.

DISCUSSION

Beverage patterns defined in this study were based on a simple classification scheme
(Bowman, 2002) and used to identify segments of the adolescent population that had
distinct sociodemographic and lifestyle characteristics. Because these sociodemographic
and lifestyle factors were associated with the prevalence of overweight/risk for
overweight, understanding their association with beverage patterns gave insight into the
complex inter-relationship between these risk factors, beverage consumption and the
prevalence of overweight/risk for overweight. For example, the discriminating
characteristic of the ‘milk & juice’ group was regular breakfast consumption, and the
‘milk & juice’ group had the lowest prevalence of overweight. Other sociodemographic
and lifestyle factors associated with the ‘milk & juice’ pattern included high SES,
frequent exercise, sports team participation and not watching TV. In contrast, the
‘sugared-sweetened beverage’ patterns were associated with low SES and TV watching,
and overweight was more prevalent in the ‘sugared-sweetened beverage’ groups than in
the ‘milk & juice’ groups. Perhaps SES was the common denominator behind breakfast
habits, TV watching and physical activity levels (Miech et al., 2006). Nevertheless,
beverage patterns functioned effectively as a marker of these risk factors for overweight,

because the beverage patterns were associated with these sociodemographic/lifestyle

195

factors, and both the beverage patterns and sociodemographic/lifestyle factors were
associated with the prevalence of overweight/risk for overweight.

Beverage patterns were defined a priori to classify youth into mutually exclusive
groups according to the combination of beverages he/she most frequently consumed.
One other researcher classified youth by consumption of milk and/or soda (Bowman et
al., 2002), but the current study used F FQ rather than 24-hour recall data, and also
controlled for juice consumption while contrasting beverage pattern groups to determine
effects of the various types of beverages. Cluster analysis could have been used to
empirically derive beverage patterns and classify youth into mutually exclusive groups
(Kant, 2004), but each individual would not have necessarily consumed all beverages that
characterized his/her cluster. A cluster analysis of water, beverages and selected foods
consumed by adults showed that higher-educated, older adults (>60 years) were more
likely to be grouped with a “water/fruit & vegetable/low fat milk” cluster, while young
adults (18-35 years) were more likely to be grouped with a “fast food/snack/soft drink”
cluster (Popkin et al., 2005).

Youth were classified into the various beverage pattern groups according to their
beverage preferences. Others have found that beverage choices differed by age, gender,
race-ethnicity or region (Guenther, 1986; Hamack et al., 1999; Bowman, 2002; Forshee
& Storey, 2003). For example, age was associated with decreased intakes of milk and
increased intakes of soft drinks, both regular and diet, especially in girls (F orshee &
Storey, 2003). Boys tended to drink greater quantities of milk (Guenther, 1986) and soft
drinks (Guenther, 1986; Harnack et al., 1999) than girls. African Americans drank less

milk (Guenther, 1986; Forshee & Storey, 2003), coffee/tea (Guenther, 1986) and soft

196

drinks (Hamack et al., 1999; Bowman, 2002; Forshee & Storey, 2003) but more fruit
drinks (Guenther, 1986; Forshee & Storey, 2003) than whites. Hispanic girls drank more
citrus juice than non-Hispanic girls (Forshee & Storey, 2003). Adolescent girls living in
the Midwest tended to drink sofi drinks, while those living in the Northeast were more
likely to drink milk (Bowman, 2002). Adolescents in the South tended to drink
sweetened iced tea (Guenther et al., 1986). The age, gender, race-ethnic and regional
distributions of beverage patterns defined by the present study reflected the different
beverage choices of population subgroups that others have found (Guenther, 1986;
Hamack et al., 1999; Bowman, 2002; Forshee & Storey, 2003).

As indicated by preliminary analyses of beverage frequency (Chapter 4, Table 5),
the ‘other beverages’ groups included many adolescents who were drinking diet soft
drinks in addition to those who drank coffee and tea. A high percentage of youth in this
group reported trying to lose weight. Therefore, the association of the high prevalence of
overweight with the ‘other beverages’ pattern should be interpreted with caution.
Because cross-sectional data lack temporality, reverse causality should be considered
when interpreting these observations. Rather than causing overweight, drinking “diet”
soft drinks, exercising, smoking cigarettes and breakfast skipping were possibly related to
the weight loss efforts of the ‘other beverages’ group. Drinking “diet” beverages or
bottled water instead of sugared-sweetened beverages decreased BMI among those in the
upper BMI tertile at baseline in a 25-week intervention study of non-dieting youth age
13-18 years (Ebbeling et al., 2006). Dieting in adolescents was associated with unhealthy

weight loss practices rather than exercise in a prospective study (French et al., 1995).

197

Among school-aged children in Ireland, dieters reported drinking more coffee, higher
levels of tobacco use, and lower exercise levels than non-dieters (Gabhainn et al., 2002).

Reverse causality, i.e., that overweight adolescents were exercising to lose
weight, provides an alternative explanation for the high prevalence of overweight/risk
found among infrequent exercisers in the present study compared to those exercising 5-6
times per week. However, as observed in a previous investigation using NHANES 111
data (Crespo et al., 2001), in the current study, overall, no association was found between
exercise and the prevalence of overweight/risk, because the prevalence was also high
among those exercising 7 or more times per week. Others omitted a discussion of the
exercise variable and used only the measurement of team sports participation to examine
the relationship between physical activity and BMI in NHANES HI (Storey et al., 2003;
Forshee et al., 2004). A positive association between exercise and the prevalence of
normal BMI was found in CSFII (Boumtje et al., 2005). The high prevalence of
overweight observed in the current study among daily exercisers compared to those
exercising 5-6 times per week might be due to increased lean body mass, not fat. BMI
does not accurately assess body composition, e.g., overweight youth with a high
proportion of lean body mass are not overfat (Forbes, 1987).

The association between smoking and overweight might also be due to reverse
causality. College women have reported they smoke cigarettes to control body weight
(Malinauskas et al., 2006). That smoking might be used as a weight loss strategy even
among teenagers is a disturbing trend. Youth who reported they were trying to lose
weight were more likely to initiate smoking in two prospective studies (French et al.,

1994; Austin et al., 2001). Smoking was associated with trying to lose weight (Lowry et

198

al., 2002) and also with other unhealthy weight loss practices in the 1999 Youth Risk
Behavior Survey (YRBS) (Delnevo et al., 2003). Daily smokers were 2—4 times more
likely to fast, use pills and purge to control weight than nonsmokers (Delnevo et al.,
2003), and smoking was associated with vomiting or laxative use in another study
(Krowchuk et al., 1998).

College women reported that they used breakfast skipping as a weight loss
strategy even more often than smoking (Malinauskas et al., 2006), and breakfast skipping
was also associated with unhealthy weight loss practices, as breakfast skippers were more
likely to report fasting to lose weight in the YRBS (Zullig et al., 2006). The inverse
association between overweight prevalence and breakfast consumption observed in the
current study could then be due to overweight youth skipping breakfast to lose weight,
although it’s possible that breakfast skipping might lead to weight gain over time.
Breakfast skippers in a longitudinal study tended to gain more weight (measured as BMI)
than regular breakfast eaters, but only among normal-weight subjects; BMI decreased in
overweight adolescents who skipped breakfast (Berkey etal., 2003). Potential
confounding provides another explanation for the association of breakfast skipping with
overweight. The ‘milk and juice’ pattern was associated not only with regular breakfast
consumption, but also with high SES and other healthy behaviors like exercise and not
watching TV, and these other factors could account for the association between breakfast
consumption and the low prevalence of overweight. Drinking milk and juice was
associated with breakfast consumption, and breakfast consumption might be associated
with other components of a healthy lifestyle such as high levels of physical activity

(Cavadini et al., 2000b; Keski-Rahkonen et al., 2003; Godin et al., 2005).

199

Cross-sectional analyses cannot be used to make causal inferences, but the model
in Figure 1 presumes that factors like breakfast consumption and TV watching are
downstream from beverage patterns in the causal pathway leading from beverage patterns
to overweight (causal pathway A). Numerous studies have shown that regular soft drinks
and snack foods are consumed while watching TV (Coon et al., 2001; Giammattei et al.,
2003; Utter et al., 2003; Matheson et al., 2004; Phillips et al., 2004). Preliminary
analyses of the frequency with which adolescents consumed beverages at the various
meal occasions (shown in Table 6 of Chapter 4) demonstrated that sugared-sweetened
beverages were most frequently consumed at snack/ beverage breaks, while milk and
juice were most frequently consumed at the breakfast meal occasion. The model shown
in Figure 1 suggests that TV watching and the frequency of breakfast consumption
determine the meal situation and beverage patterns that accompany the snack or breakfast
occasion. Because of their proximity, beverage patterns might have a stronger
relationship to the mediating and outcome variables (energy intake and adolescent
obesity) than those downstream. However, as shown in Figure l, the effects of TV
watching would be larger than beverage patterns (causal pathway A), because TV
watching would be associated with physical activity levels (causal pathway B) as well
(Robinson et al., 1993; Nelson et al., 2005).

Recent research showing the effects of SES on adolescent overweight suggested
that poverty was associated with breakfast skipping, physical inactivity, and sugar-
sweetened beverage consumption (Miech et al., 2006). Therefore, SES might be shown
downstream from breakfast, TV watching and exercise habits in the model of Figure 1.

The current analyses of the association between beverage patterns and overweight were

200

binge .53 2:03.:— 0923...... he neufioemma 2: new .25:— _a=E3=eU A PEER

 

 

 

 

/——\

\

//

 

 

 

 

 

 

 

 

 

 

Eat—Eonxm 53304
3.55— . u... .agmham
< A V
533m
3.85 3.55— owauogm—

 

 

 

 

5.52.2. a... 2%?
.332 .mafieiw .
Em 325.3.—

32— : he mam .9885 E
saga—95m £39m
Eaaem 525$ .03.
g?

 

 

n=1—89$ >h.

 

 

 

 

 

‘l

 

 

 

 

 

new-83 :62
5.3390 302

 

 

 

 

fl man“: “mafia—“Ohm w

 

201

not adjusted for covariates. Adjusting for SES, TV watching and other
sociodemographic/lifestyle factors depicted downstream in the model might lead to
‘over-control’ in examining the association between sugar-sweetened beverages and
overweight.

There were several limitations that pertain to the current analyses. Some youth
might have been misclassified to a ‘juice’ group, if they mistakenly reported fruit drink
frequency in response to the 100% mu juice question, because the fruit juice question
was sequenced before the question about fruit drinks in the F F Q. In preliminary analyses
of beverage frequency (Table 5 of Chapter 4), 24-hour recall intakes of fruit drinks were
higher for ‘juice’ groups than ‘non-juice’ groups. Youth of low SES among the ‘sugared-
sweetened beverages & juice’ or ‘other beverages & juice’ groups, for example, might
have been more likely to confuse fiuit drinks with fi'uit juice than those with high SES in
the ‘milk & juice’ group. Small variances indicated that consumers of the ‘milk & juice’
pattern were a homogeneous group.

Large variances indicated that the ‘other beverages’ groups were comprised of
subgroups. Because the FFQ did not allow the respondent to indicate whether sugar was
added to coffee or tea, the ‘other beverages’ group included both sugar-sweetened
coffee/tea consumers and diet soft drink consumers, and risk factors might have varied
widely between these subgroups. On the other hand, because there were four separate
patterns that included sugar-sweetened beverages and shared similar risk factors, it might
be necessary to collapse them to see a stronger association between sugar-sweetened

beverage consumption and obesity in youth.

202

The major strength of this study was that youth were classified into beverage
pattern groups according to their ‘usual’ beverage intakes reported using the FFQ.
Assessing each youth’s ‘typical’ beverage pattern using ‘usual’ intakes in the FF Q should
strengthen the relationship of beverage consumption with sociodemographic and lifestyle
factors. Because of intra-individual variation in dietary intake, youth might have been
misclassified to beverage patterns if a single 24-hour recall were used instead. In
addition, misclassification to beverage pattern groups due to under-reporting in response
to the FF Q was addressed by identifying most frequently consumed beverage types based
on their percentage share of total beverage frequency. Rather than using the beverage
frequency to identify youth who reported consumption 21 time/day, for example, youth
were classified by the combination of beverages consumed with shares 225% of total
beverage frequency. If fi'equency-based criteria were used, youth who reported drinking
all caloric beverage types <1 time/d would have been classified to the ‘other beverages’
(msj) pattern, and these “under-reporters” were removed from this group by using
consumption criteria based on the percentage share. Avoiding potential bias due to
under-reporting was important to the design of this study; overweight youth might be
more likely to under-report beverage consumption (Bandini et al., 1990; Johnson et al.,

1996; Johnson, 2000).

Implications
Beverage patterns might be markers of sociodemographic and lifestyle factors that are
risk factors for overweight. This suggests that complex interrelationships between

sociodemographic/lifestyle factors, beverage patterns and adolescent obesity exist.

203

Chapter 6
PATTERNS OF BEVERAGE CONSUMPTION ASSOCIATED WITH
ADOLESCENT OBESITY IN THE U.S.

ABSTRACT

Objective To examine associations of beverage patterns (i.e., the types of beverages
most frequently consumed) with BMI percentile, percentage body fat and overweight/risk
for overweight among adolescents in the U.S.

Design Cross-sectional analysis of data from the Third National Health and Nutrition
Examination Survey (NHANES III, 1988-94). Food frequency questionnaire (FF Q) data
were used to classify youth into beverage pattern groups. BMI percentile and
overweight/risk for overweight were determined from measured BMI using CDC Growth
Charts, and percentage body fat from two skinfold measurements.

Subjects A U.S. population-based sample of adolescents age 12-16 years (n=1,872).
Statistical analysis SUDAAN software was used to run linear and logistic regression
analyses. The crude and adjusted associations of each beverage pattern group with body
weight and with body fat, least-square (LS) mean BMI percentile and percentage body
fat, as well as the odds ratios for overweight/risk for overweight were examined.
Covariates examined were gender, age, race-ethnicity, region, urbanization, parental
overweight/obesity, SES indicators, smoking, alcohol use, weight loss behavior, exercise,
team sports participation and TV watching.

Results The LS means of both BMI percentile and percentage body fat were lower in
the ‘milk & juice’ group than in the ‘other beverages’ group, and LS mean BMI

percentile was higher in the ‘sugar-sweetened beverages’ than in the ‘milk & juice’

204

group, in the unadjusted model that included only the beverage pattern, amount of water
and type of milk usually consumed. The crude odds ratio for overweight/risk for
overweight was low in the ‘milk & juice’ group (OR=O.52, 95% CI = 0.28-0.96), and
high in the ‘other beverages’ group (OR=2.74, 95% CI = 1.12-6.68), compared to the
‘milk’ group. The adjusted associations of ‘milk & juice’ with BMI percentile,
percentage body fat and overweight/risk for overweight were significant in most models,
except those adjusted for SES, physical activity or TV watching. The adjusted
association of the ‘sugar-sweetened beverages’ pattern with BMI percentile was
significant while controlling for physical activity, but it was not significant while
controlling for SES or TV watching. The associations of the ‘other beverages’ pattern
with obesity outcomes were not significant with adjustments for youth who were trying
to lose weight.

Implications This study suggests that several sociodemographic/lifestyle factors
examined are highly correlated with beverage patterns, and together these
sociodemographic/lifestyle factors and beverage patterns are associated with adolescent
obesity. The interrelationships between sociodemographic/lifestyle factors and beverage
patterns must be carefully considered to elucidate the potential role each specific

etiologic factor has in increasing the prevalence of adolescent overweight.

INTRODUCTION
The prevalence of overweight among adolescents has been steadily increasing over the
past three decades in the U.S. Among adolescents age 12-19 years, the current

prevalence rate of 17 .4% is more than three times higher than the rate of 5.0% observed

205

30 years ago (Ogden et al., 2002; Hedley et al., 2004; Ogden et al., 2006). At the same
time, intakes of soft drinks and other sugar-sweetened beverages have increased, for
example, soft drink intakes of adolescents age 14-17 years doubled (from 7 to 14 oz/day)
for girls, and more than tripled (from 7 to 22 oz/day) for boys between 1977-78 and
1994-98 (French et al., 2003), and other researchers have observed similar trends
(Cavadini et al., 2000a; Nielsen & Popkin, 2004). Some researchers have suggested that
consumption of sugar-sweetened beverages is related to the high prevalence of
overweight in youth (Hamack et al., 2000; Bray et al., 2004).

Consumption of milk has declined, for example, 77% of girls age 12—19 years
drank milk at least once per day in 1977-78 compared to 53% in 1994-96, while soft
drink consumption increased from 41% to 56% of girls at the same time (Bowman et al.,
2002). Other researchers have corroborated these trends in milk and soft drink intakes
among children and adolescents of various age and gender groups (Morton & Guthrie,
1998; Moshfegh et al., 2000; Enns et al., 2002; Emis et al., 2003). An inverse
relationship between calcium and adiposity has been hypothesized (Zemel et al., 2000),
which suggests that low fat milk intakes of youth might be inversely correlated with body
fat. Age was associated with decreased intakes of milk and increased intakes of soft
drinks, especially in girls (Lytle et al., 2000; Bowman et al., 2002; Forshee & Storey,
2003). If intakes of milk are low and soft drink intakes are high in adolescents compared
to children of younger age groups, then the low calcium intakes combined with excess
calories from intakes of sugar-sweetened beverages might be associated with the
prevalence of overweight in adolescents.

Several prospective studies (Ludwig et al., 2001; Berkey et al., 2004) and

206

intervention studies (James et al., 2004; Ebbeling et al., 2006) have shown an association
between soft drinks and weight gain in youth, although cross-sectional studies examining
sugar-sweetened beverage consumption in relation to Body Mass Index (BMI) or
overweight status have had inconsistent results (F orshee & Storey, 2003; Forshee et al.,
2004; Boumtje et al., 2005). Those studies that used the Continuous Survey of Food
Intake by Individuals (CSFII) data to examine the association of beverage intakes with
BMI or overweight in adolescents (Forshee & Storey, 2003; Boumtje et al., 2005) were
limited in that self-reported information rather than measured height and weight were
used to compute BMI (Kuczmarski et al., 2001; Brener et al., 2003). In addition,
analyses of children or adolescents using National Health and Nutrition Examination
Survey (NHANES) data that examined BMI in relation to beverage intakes (F orshee &
Storey, 2003; Forshee et al., 2004) were limited, because BMI was not adjusted for
gender and age using the CDC Growth Charts reference data (Kuczmarski et al., 2000;
Bachman et al., 2006). In the current study, BMI percentile was calculated fiom
measured height and weight using the gender-specific BMI-for-age percentile reference
data provided by the CDC Growth Charts, because criteria based on the BMI-for-age
percentiles were recommended to assess overweight/risk for overweight in children and
adolescents (Kuczmarski et al., 2000). In addition, percentage body fat was determined
from summed triceps and subscapular skinfold measurements (Slaughter et al., 198 8),
because BMI does not accurately assess body composition, e. g., overweight youth with a
high proportion of lean body mass are not overfat (Forbes, 1987 ).

Cross-sectional studies using the CSFII or NHANES data (Forshee & Storey,

2003; Forshee et al., 2004; Boumtje et al., 2005) might have been limited by inaccurate

207

assessment of beverage consumption due to intra-individual variation in intake (Beaton et
al., 1979; Sempos et al., 1985) or under-reporting of intake (Bandini et al., 1990; Johnson
et al., 1996; Johnson, 2000; Subar et al., 2003). In the current study, beverage intakes
were assessed using a food frequency questionnaire (FFQ) rather than the 24-hour recall
to address intra-individual variation. Instead of focusing on individual beverages, the
current study examined beverage intake by defining patterns that reflected multiple
beverages consumed in combination. Youth were classified into mutually exclusive
groups according to the combinations of beverage types most frequently reported as
‘usual’ intake. A method for classifying youth by his/her ‘typical’ beverage pattern was
developed to address under-reporting of intakes.

The aim of this study was to examine the associations of beverage patterns (i.e.,
the types of beverages most frequently consumed) with obesity (i.e., overweight and
adiposity) in l2-l6—year-old adolescents in the U.S. Indicators of overweight and
adiposity examined separately included the BMI percentile, percentage body fat as
assessed by skinfold measurements, and overweight/risk for overweight (i.e., BMI 285th
percentile of BMI-for-age). Youth were classified into beverage pattern groups
according to the combination of beverage types reported most frequently in the F FQ. In
addition to beverage pattern groups, other beverage variables considered in analyses were
fluid ounces of water and type of milk (e.g., whole, 2%, 1%, skim) usually consumed.
Potential covariates examined were gender, age, race-ethnicity, region, urbanization,
parental overweight/obesity, SES indicators, smoking, alcohol use, weight loss behavior,

exercise, team sports participation and TV watching. Covariates were examined

208

separately one at a time to investigate whether each one might potentially confound the

relationship between beverage patterns and overweight/ adiposity.

METHODS
Dataset
The Third National Health and Nutrition Examination Survey (NHANES 111, 1988-1994)

dataset was used for analyses, rather than the more recent 1999-2002 NHANES dataset,

 

because the dietary food frequency questionnaire used to assess the youth’s usual
beverage pattern is available only in NHANES III. The National Center for Health
Statistics (NCHS) of the Centers for Disease Control and Prevention (CDC) conducts
NHANES to obtain nationally representative data on the nutritional and health status of
the civilian, non-institutionalized U.S. population. Participants were selected using a
complex, stratified, multistage, probability cluster sample design (N CHS, 1994). The
survey included interviews conducted in the household and at a mobile examination
center (MEC). The examination included an anthropometric component and sexual
maturation staging. A food frequency questionnaire and a 24-hour recall dietary

interview were also administered at the MEC.

Subjects

Adolescents age 12-16 years who were interviewed and examined in NHANES III
comprised the analytic sample for this study. The NHANES IH sample included 2,216
adolescents aged 12-16 years selected for examination, and, of those, 2,079 (93.8% of

2,216) were examined at the MEC. Upon examination, 19 respondents (0.9% of 2,216)

209

found to be ineligible (l6 teenage girls were pregnant, and three were lactating) were
excluded from the current study. Because 24-hour recall data were used in a preliminary
analysis necessary to conduct the current study, the analytical sample was limited to
adolescents age 12-16 years who completed a 24-hour recall dietary interview as well as
all beverage items in the FF Q. There were 1,997 youth (90.1% of 2,216) who had a
complete, reliable 24-hour recall dietary interview, and of these, 97 (4.4% of 2,216) were
missing beverage frequency data and another 28 (1.3% of 2,216) were missing data on
water intake. Alter excluding those with missing data, the remaining analytic sample of
1,872 youth (84.5% of 2,216) included 871 boys (46.5% of 1,872) and 1,001 girls (53.5%

of 1,872) aged 12-16 years.

Definition of beverage patterns

Youth were classified by the combination of beverage types most frequently consumed,
as reported in the FFQ. Four main categories of the 12 FFQ beverage items listed in
Table l were (1) milk, including milk on cereal as well as milk drinks, (2) 100% fruit
juice, (3) sugar-sweetened beverages (e.g., fruit drinks and regular soft drinks) and (4)
other beverages (e. g., diet soft drinks, coffee/tea and beer/wine/liquor). An additional
item in the FF Q asked what type of milk was usually consumed: whole, 2%, 1%, or
skim/nonfat milk. The dietary interview included a question to respondents about how
many glasses/cups or fluid ounces of plain water they usually drank in a 24-hour period.
If the response was in terms of glasses or cups, participants specified the fluid ounces per

glass or cup, and numbers of glasses/cups were converted to fluid ounces. The beverage

210

Table l. Beverage categories and beverage subcategories corresponding to beverage

items in the NHANES 111 food frequency questionnaire (F FQ)

 

Beverage categories &

beverage subcategories

FFQ beverage items

 

1. Milk and milk drinks,

chocolate & regular

2. 100% mu juice,

citrus & non-citrus

3. Sugar-sweetened bevg.
3a. Fruit drinks
3b. Regular soft drinks
4. Other beverages
4a. Diet soft drinks

4b. Coffee/ tea

4c. Alcoholic

beverages

1a. How often did you have chocolate milk and hot cocoa?

lb. How often did you have milk to drink or on cereal?

Do not count small amounts of milk added to coffee or tea.

2a. How often did you have orange juice, grapefruit juice
and tangerine juice?

2b. Other fruit juices such as grape juice, apple juice,
cranberry juice, and fruit nectars?

3a. How often did you have Hi-C, Tang, Hawaiian Punch,
Koolaid, and other drinks with added vitamin C?

3b. Regular colas and sodas, not diet?

4a. Diet colas, diet sodas, and diet drinks such as Crystal
Light?

4b (1). Regular coffee with caffeine?

4b (2). Regular tea with caffeine?

4c (1). Beer and lite beer?

4c (2). Wine, wine coolers, Sangria, and champagne?

4c (3). Hard liquor such as tequila, gin, vodka, scotch,

rum, whiskey and liqueurs, either alone or mixed?

 

211

pattern, type of milk and fluid ounces of water usually consumed were all included in the
current analyses.

Beverage pattern classifications were based on ‘usual’ intake in the F FQ of a
combination of up to three beverage types: milk, sugar-sweetened beverages and/or 100%
fruit juice. To classify youth into beverage pattern groups, the F FQ beverage frequencies
were used to identify the combination of beverages that characterized a youth by his/her
‘typical’ beverage pattern. Each beverage’s frequency was expressed as a percentage of
the summed frequency for all beverages to identify the combination of beverages each
having a share that was at least 25% of the total beverage frequency. Notations for the
eight beverage patterns were combinations of capital letters M, S and J to refer to milk,
sugar-sweetened beverages and juice, respectively, having shares 225% total beverage
frequency, and lowercase letters m, s and j represented consumption of less than the
specified amount. Beverage patterns were named according to the eight beverage
combinations: ‘all caloric beverages’ (MSJ), ‘milk & sugar-sweetened beverages’ (MSj),
‘milk & juice’ (MSJ), ‘milk’ (Msj), ‘sugar-sweetened beverages & juice’ (mSJ), ‘sugar-

sweetened beverages’ (mSj), ‘other beverages & juice’ (msJ) and ‘other beverages’ (msj).

Assessment of overweight and adiposity

Weight and height measured at the MEC (DHHS, 1998) were used to assess Body Mass
Index (BMI=kg/m2), and skinfold thickness measures (DHHS, 1998) were used to assess
percentage body fat. The SAS program available from NCHS was run to calculate the
youth’s BMI percentile using the gender-specific BMI-for-age CDC Growth Charts

reference data (N CHS; Kuczmarski et al., 2000). The BMI percentile and percentage

212

body fat were dependent variables in two separate linear regression analyses. The BMI
percentile was used to classify youth by weight status (i.e., underweight: BMI <5th
percentile, normal body weight: BMI 25th percentile and BMI <85th percentile, at risk
for overweight: BMI 285th percentile and <95th percentile, and overweight: BMI 295th
percentile). The presence of overweight/risk for overweight (i.e., BMI 285th percentile)
was the dependent variable for logistic regression analyses. Underweight youth were
excluded from logistic regression analyses.

Percentage body fat was determined from summed triceps and subscapular
skinfold thickness (mm) measurements using prediction equations for adolescents
developed by Slaughter et a1. (1988). There were 35 youth whose triceps or subscapular
skinfold thickness was too large for the calipers. Rather than excluding these obese
adolescents from analyses, percentage body fat was calculated using the group’s
maximum skinfold thickness (mm) measurements in place of the missing values.
Slaughter et a1. (1988) included separate skinfold equations for prepubescent, pubescent,
and postpubescent white and black male adolescents and all female adolescents whose
summed tricep and subscapular skinfolds were 35 mm or less, and male and female
adolescents whose summed tricep and subscapular skinfolds were greater than 35 mm.
Adolescent boys were classified by race and sexual maturation (SMA) stage to use the
Slaughter equations. White and black male adolescents were classified by the composite
pubic hair and genitalia rating as follows: prepubescent (SMA stages 1.0 - 2.0), pubescent
(SMA stages 2.5-3.5) and postpubescent (SMA stages 4.0-5.0) (Boileau et al., 1984;

Slaughter et al., 1984).

213

Assessment of sexual maturation stage

The physician’s examination included an assessment of SMA according to the 5-stages of
developing pubic hair (male and female), male genitalia, and female breasts (Tanner,
1962; DHHS, 1991). A composite score was calculated by averaging the pubic hair
rating with the genitalia rating for boys and with the breast rating for girls (Tanner,

1962). The composite score was used to classify both males and females by pubertal
status: prepubescent (SMA stages 1.0 - 2.0), pubescent (SMA stages 2.5-3.5) and
postpubescent (SMA stages 4.0-5.0) (Boileau et al., 1984; Slaughter et al., 1984).
Pubertal status of those missing both SMA ratings was determined by age (years). The
regression models included pubertal status in addition to the beverage variables and

potential covariates described next.

Covariates in regression analyses

Potential covariates considered in regression analyses included sociodemographic factors
such as gender, age, race-ethnicity, region, urbanization, parental overweight/obesity,
household income and other indicators of social economic status (SES). Lifestyle factors
examined were smoking, alcohol use, weight loss behavior, physical activity and TV
watching. Categories for race-ethnicity included non-Hispanic white, non-Hispanic black,
Hispanic and other race-ethnic groups. Four broad regions of the U.S. included the
Northeast, Midwest, South and West. The urban/rural classification was a metropolitan
area having a population 21 million vs. all other areas. The poverty income ratio was the
ratio of household income to the poverty level for household size in an urban or rural

area. Other indicators of SES included the family size and characteristics of the

214

household head or family reference person (F RP), such as their gender, age group,
marital status, education and employment status. BMI was calculated from reported
weight and height of the youth’s mother and father to determine whether both or one of
two parents were overweight (BMI _>_25 kg/mz) or obese (BMI 230 kg/mz).

The survey included lifestyle questions, such as those asking youth whether they
currently smoked cigarettes or drank alcohol in the past year. Youth were also asked
whether they were currently trying to lose weight. Physical activity was assessed using
the following questionnaire items asking youth about their exercise habits and
participation in team sports:

1. How many times per week do you play or exercise enough to make you sweat

or breathe hard?

2. In the past year, how many sports teams or organized exercise programs have

you been involved in (not including physical education or gym classes)?
Youth were asked twice, once in the household and again at the MEC, about time spent
watching TV the previous day, and the average hours/day in two days was used in the

current study.

Statistical analysis

The Statistical Package for the Social Sciences, SPSS for Windows, Version 12.0 (SPSS
Inc. Chicago, IL, USA) was used to prepare the dataset for analysis. All statistical
analyses were performed using SUDAAN version 8 (Research Triangle Institute,
Research Triangle Park, NC), to account for the effect of the clustered sample design in

estimating the variance of population parameters while conducting statistical tests (Shah

215

et al., 1997). The six-year MEC-examined sample weights were applied to the
calculation of all statistics. The sample weights adjust the data to the 1990 U.S. Census
figures, therefore the data are representative of adolescents in the 1990 U.S. population.
The association of obesity outcomes with beverage pattern groups was examined
after controlling for covariates using regression analysis. Multiple regression analysis
was used to implement a multiple analysis of variance (MANOVA) with categorical
variables (i.e., dummy variables) in a general linear model. The p-value of the Wald F
test for the treatment effect was interpreted to test for a significant association between
any one of the beverage pattern groups and obesity. Because the p-value to test the
significance of an effect of a particular beverage pattern group indicates only whether the
beta coefficient was different fiom zero, Bonferroni contrasts of least square (LS) means
were performed to indicate whether the ‘milk & juice’ beverage pattern group was
significantly different from each of the other seven beverage pattern groups (i.e., LS
mean BMI percentile and LS mean body fat (%) were adjusted for effects of covariates).
The 95% confidence intervals for odds ratios were calculated and interpreted to indicate
whether a beverage pattern group’s risks were significantly different from 1.0. Odds
ratios for overweight/risk for overweight were adjusted for effects of covariates as well.
Each covariate was added individually to the model that included only the
beverage pattern, amount of water and type of milk usually consumed (Model 1).
Covariates were examined separately one at a time to investigate whether each one might
potentially confound the relationship between beverage patterns and overweight/
adiposity. If the effect of beverage patterns was no longer significant with the inclusion

of the potential confounder in the model then it was concluded that this covariate was

216

indeed confounding the relationship between beverage patterns and overweight/adiposity.
Covariates not found to be confounders were added to Model 1 together to form an
adjusted model (Model 2), and the remaining covariates were each added individually

one at a time to the adjusted model (Model 2).

RESULTS

Table 2 presents mean values of overweight and adiposity measurements in beverage
pattern groups. Bonferroni contrasts indicated that BMI percentile, percentage body fat
and overweight prevalence were highest in the ‘other beverage’ (msj) group compared to

the ‘milk’ (Msj) group.

Percentile of BMI-for-age
The associations of beverage patterns with the BMI percentile are shown in Table 3. The
R-square for the unadjusted model (Model 1) indicated that the beverage pattern, water
and milk type explained 3% of the variance in the percentile of BMI-for-age. The
association between BMI percentile and beverage patterns (Wald F p<0.05) was such that
the least square (LS) mean BMI percentile of the ‘sugar-sweetened beverage’ (mSj) and
‘other beverage’ (msj) pattern groups were higher than the ‘milk & juice’ (MSJ) group
(p<0.05).

Adding parent’s weight status individually to Model 1 increased R-square to
~10% of the variance explained, and the overall beverage pattern effect was significant at
the p=0.01 level. The overall beverage pattern effect was no longer significant (Wald F

p=0.l 1) when indicators of social economic status (SES), such as income and education

217

 

 

 

 

 

2% u 5 :s n 5 8% n 5 E: u 5 as. u 5 as u 5 a? u 5 as u 5 as; u 5 2.58on
%a a Sm as a 3m an « EN 3 « EN 2. a MEN 2 a 4.: .3 a EN 3 a SN 3 a SN 52% Em
3 u 5 :s n 5 8% n 5 as u 5 as. u 5 as n 5 as. u 5 as n 5 as; u 5 2:828
3 a 3m 5 a as 2 a es 3 a as Z a Na 3 a we 3. a m: S a as E a ms 52% Em
Aoacoobm 59%. :26 E30 .8.“ xmtbzcgo was $5582” 5mmN Ems SE36 .«0 Amm “m “3803 coco—«$5
an n 5 as n 5 as u 5 as n 5 as. n 5 as u 5 a? u 5 as u 5 as; u 5 2:888
*3. a we» 2. an ode A: a mg 3 a New 3 a 3m 3 a 3m 2 a mam 3. « ado S a <8 Em
an n 5 :s u 5 as n 5 :s u 5 as. u 5 as u 5 as. u 5 as u 5 as; u 5 aces
:25 a mad :5 a 33 85 a a8 25 a a3 :5 a and 8d a s5 85 a 25 £5 a £3 85 a £5 Em
an n 5 as u 5 82 u 5 as u 5 as. n 5 an n 5 a? u 5 as u 5 as; n 5
a5 a ham mo a 98 3 a a: no a EN 3 a 2 no a 5.8 no a EN we an em. 3 « EN €35 Em
an n 5 :s u 5 82 u 5 as u 5 as. n 5 as u 5 am... u 5 as u 5 as; n 5
9m .1. Se 3 a 58 3 a new 2 a 2m 3 a" man 3 a 5;. 3 a 2m g a 5.8 we a 3m 35 Ema?
«$63.85 $83 8 tom: EoEoSmmoE Emma? been Amm a :aoEv :82
an n 5 as u 5 as. u 5 as u 5 c s. u 5 a3 u 5 a? u 5 as u 5 at: u 5 “880585
.28 7.5 “we as .32 2E .52 SE 32.2 3:83.
w>om 033 a. w>om 623. a. m>om #34 023 w>om Em w>om =< Emma? Atom
550 MEm see 53% Emma a 5E .95 a 5E 2.5.6 =<

 

Nazca Egan owfio>om

 

.Eotma owfigon .3 woman—o

memo» 07$ own 358203 waste 362:? ES “@6595 $83 8 new: EoEoSmmoE we.“ been can Emma? been :82 .N 033.

218

95mm Q36 2:8. 3 anew Eaten ow8u>on mo memtmaEOo so.“ 8.0 V a .1. .36 v a
E3098 BEBE 05 55 $3 mossy: 382%: .328 m .E whats .
88832 98 Samoa—comm omfio>on 130:0 XmNN 0.8% a warren :08 533033“ do?“ can mommmgon coguooBmmemm i=8 mo
moss": 8 some." _. 98 m .2 @832 3:50 .0“; of 5 v8.88» banana as: mowmbgn mo cousin—moo 2.: s 2.5th owfigun 2:. N
.Z<<n5m «o @0508 metamtmomz .3 33:53 08 88.8 2353 emu cflawBBéEEmm 8a momsmsobm 28 $802
Amthnmv 8:35.“ wages .5 “gamma mafia—88 .Omm :0 mac: owfigon 8 830%": 2322:: me wmmmsE o:

as .5st base :32 54m 0352 .3238 a? as“; sé Rs 38828.. «arse .E mmz<mz H858 _

 

 

 

219

an n 5 aa u 5 as u 5 :s u 5 as. u 5 as n 5 as. u 5 as n 5 :2: u 5
:3 a as S a EN me a 3m 3 « :2 S a is as a. as 3 a EN 3 a SN 3 a as $5 3 38m
am u 5 aa n 5 as u 5 :s n 5 as. u 5 as u 5 a; u 5 as n 5 :8; u 5 A55 2&5”
:2 a 2“. 3 a ms. 2 a as E a EN 2 a SN 2 q. ”.3 E a MEN 2 a. EN 3 a QR 858%
am u 5 aa u 5 as n 5 :s u 5 as. u 5 as u 5 a; u 5 as u 5 as; u 5 €85 285%.
:5 a as I a as no a as ad a as Z a as so a ~.: 3 a ms 2 a as so a. as 53835
3 u 5 aa u 5 as m u 5 as u 5 as. u 5 as u 5 as. u 5 as n 5 as; u 5 55 285%
.3 a q: 3 a as so a as 3 a as .3 a 3: ed a as S a n: 2 a n: as a. as 885

bass?“ $83 8 wow: Ensues—mace “mm been Amm a .825 :82
am n 5 as n 5 as u 5 as u 5 2 :. n 5 as u 5 as. u 5 as n 5 as; u 5 assesses
.35 28 $8 as a: :2 $2 SE ass sigma.
955 622. a. whom 022. a. w>om 6:32 023. m>om Em w>om =< Emma? boom
550 “5m 850 Essa 33% a. 5E -mam a. xE E26 3.

 

«95.5 Sufism owmmgom

5.58 .N 2am.

Table 3. Building the linear regression model of the effects of beverage pattern on

BMI percentile in adolescents age 12-16 years1

 

 

Model BMI Percentile P-value

Beverage Least Square Beta SE T-test P-value Model

Pattern2 Mean :t SE Coeff. Beta H0: B=0 Wald F3 Wald F3 R-Square
MODEL 1: beverage pattern, water & milk type 2.50 0.028 0.030

MSJ 62.8 :t 4.4 2.98 4.87 0.543

MSj 59.6 d: 2.6 -0.15 3.91 0.970

MsJ 54.9 d: 2.6 -4.86 3.44 0.164

Msj 59.8 d: 2.9 0.00 0.00 .

mSJ 57.5 :1: 4.0 -2.26 5.48 0.681

mSj 64.9 i 2.2 ** 5.14 3.31 0.127

msJ 60.4 d: 3.9 0.61 5.03 0.904

msj 74.1 i 5.5 ** 14.33 5.73 0.016
MODEL 1 + A (gender, age, ethnicity & sexual maturation) 2.42 0.033 0.085
MODEL 1 + B (region & rural/urban area) 2.50 0.028 0.034
MODEL 1 + C (parent's weight status) 3.03 0.010 0.098
MODEL 1 + D (smoking, alcohol & drug use) 2.80 0.016 0.044
MODEL 1 + E (SES’) 1.79 0.111 0.062
MODEL 1 + F (trying to lose weight) 2.14 0.056 0.169
MODEL 1 + G (physical activity) 2.13 0.058 0.039
MODEL 1 + H (TV watching) 1.93 0.085 0.050
MODEL 2: model 1 + A, B, C & D 3.35 0.005 0.155

MSJ
MSj
MSJ
Msj
mSJ
mSj
msJ

msj

61.4 :t 5.5
61.4 :t 2.5
54.4 :1: 2.4
60.8 :t 2.5
57.4 i 3.5
63.1 i 2.1
60.4 :1: 3.8
72.3 :1: 5.4

*

*1!

*4!

0.56
0.58
-6.39
0.00
-3.35
2.30
-0.38
11.55

220

5.63
3.55
3.13
0.00
4.16
2.83
4.83
5.31

0.921
0.872
0.046

0.424
0.422
0.938
0.035

Table 3, cont'd.

 

Model BMI Percentile P-value
Beverage Least Square Beta SE T-test P-value Model
Pattemz Mean :1: SE Coeff. Beta H0: B=0 Wald F3 Wald F3 R-Square
MODEL 2 + E (SES‘) 2.00 0.075 0.176
MSJ 61.8 i 5.6 1.44 5.78 0.804
MSj 60.3 :t 2.6 -0.03 3.87 0.993
MsJ 55.2 i 2.1 -5.15 3.11 0.104
Msj 60.4 :t 2.7 0.00 0.00 .
mSJ 57.5 :t 3.7 -2.86 4.20 0.499
mSj 62.3 :1: 2.3 1.91 3.19 0.553
msJ 61.5 i 4.3 1.18 5.10 0.818
msj 69.8 i 5.4 "‘ 9.42 5.51 0.094
MODEL 1 + F (trying to lose weight) 2.16 0.055 0.260
MSJ 62.7 d: 4.7 1.24 4.87 0.800
MSj 61.8 i 2.2 * 0.32 3.23 0.921
MSJ 55.2 d: 2.4 -6.28 3.28 0.061
Msj 61.5 i 2.3 0.00 0.00 .
mSJ 56.0 i 3.6 -5.47 4.25 0.205
mSj 62.5 i 1.8 * 1.03 2.64 0.697
msJ 60.7 :t 3.7 -0.80 4.49 0.859
msj 63.3 :1: 5.7 1.85 5.81 0.752
MODEL 2 + G (physical activity) 2.45 0.031 0.167
MSJ 61.3 :1: 5.5 1.23 5.60 0.827
MSj 61.5 i 2.5 1.49 3.56 0.679
MsJ 54.9 :t 2.5 -5.08 3.23 0.122
Msj 60.0 i 2.5 0.00 0.00 .
mSJ 58.3 i 3.6 -l.70 4.19 0.686
mSj 63.3 i 2.1 * 3.23 2.90 0.270
msJ 60.7 i 3.7 0.65 4.78 0.893
msj 71.5 i 5.4 * 11.48 5.42 0.039

221

Table 3, cont'd.

 

 

Model BMI Percentile P-value
Beverage Least Square Beta SE T-test P-value Model
Pattern2 Mean t SE Coeff. Beta Ho: B=0 Wald F3 Wald F3 R—Square
MODEL 2 + H (TV watching) 2.08 0.064 0.166
MSJ 61.7 d: 5.6 0.63 5.74 0.914
MSj 60.9 i: 2.5 -0.19 3.52 0.957
MsJ 55.5 d: 2.4 -5.63 3.15 0.080
Msj 61.1 i 2.6 0.00 0.00 .
mSJ 57.3 :h 3.5 -3.80 4.12 0.361
mSj 62.5 :t 2.1 1.41 3.01 0.641
msJ 60.0 :1: 4.0 -1.08 4.98 0.829
msj 71.2 i 5.5 * 10.08 5.43 0.070
OVERALL MODEL: model 2 + E, F, G & H 1.09 0.385 0.284
MSJ 63.7 i 5.1 2.37 5.38 0.661
MSj 60.5 :1: 2.3 -0.83 3.57 0.816
MsJ 56.2 i 2.4 -5.07 3.27 0.128
Msj 61.3 :1: 2.5 0.00 0.00 .
mSJ 56.4 :t 3.6 -4.87 4.18 0.249
mSj 61.6 :L- 1.9 0.25 2.88 0.931
msJ 61.7 :1: 4.2 0.40 5.02 0.936
msj 59.6 :1: 6.1 -1.70 6.24 0.786

 

1 Source: NHANES III, 1988-1994, adolescents aged 12-16 years, with complete,
reliable 24-hr recall dietary interview, and no missing or unknown responses to
beverage items on F F Q, excluding pregnant or lactating females (n=1,872).

2 The beverage pattern is the combination of beverages most frequently reported in the
FFQ. Capital letters M, S and J refer to intakes of milk, sugar-sweetened beverages
and juice, respectively, each having a share 225% of total beverage frequency, and
lowercase letters m, s and j represent intakes less than the specified amount

*** p< 0.01, **p<0.05, *p<0.10 for contrast of beverage pattern to 'milk & juice' (MSJ)

3 Wald F and p-value for the overeall beverage pattern effect

4 Characteristics pertaining to social economic status (SES) include household size, income,

and gender, age, marital status, education and employment of the household head

222

of the household head, were added to Model 1. Weight loss behavior increased the R-
square to ~17%, compared to ~4-5% of the variance explained with physical activity or
TV watching each added individually to Model 1.

To examine the effects of adding SES, weight loss behavior, physical activity and
TV watching individually and incrementally to an adjusted model, Model 2 was formed
from Model 1 plus parents’ weight status and other model components. Model 2
explained >15% of the variance in the BMI percentile, and the overall beverage pattern
effect was significant at the p=0.005 level.

When SES was added to Model 2, the association between the BMI percentile and
beverage patterns became significant at the p<0.1 level only (Wald F p=0.075), and the
effects of the ‘sugar-sweetened beverage’ (mSj) pattern disappeared. On the other hand,
effects of the ‘other beverage’ (msj) pattern disappeared, while the BMI percentile of
‘sugar-sweetened beverage’ (mSj) pattern was still higher than the ‘milk & juice’ (MSJ)
pattern (p<0.1) when weight loss behavior was added to Model 2. Adding weight loss
behavior increased the R-square by more than 10% fiom 15.5% for Model 2 to 26% of
the variance explained. The additional variance explained was less than 1% when
physical activity and TV watching were each added individually to Model 2. Physical
activity and TV watching were similar to SES in the way that these covariates changed
the association between beverage patterns and the BMI percentile. No association
between beverage patterns and BMI percentile was observed in the overall model formed
by adding SES, weight loss behavior, physical activity and TV watching together to

Model 2.

223

.1--

Percentage Body Fat

Linear regression models of beverage pattern effects on percentage body fat are shown in
Table 4. In these analyses, beverage pattern effects were mainly due to an association of
percentage body fat with the ‘other beverages’ patterns (msJ and msj). Effects of ‘sugar-
sweetened beverage’ patterns were not significant in either the unadjusted or adjusted
models.

In the unadjusted model (Model 1), beverage variables explained 3.7% of the
variance in percentage body fat, and in the adjusted model (Model 2), the R-square
increased to 21.4% of the variance explained. Adding weight loss behavior to Model 2
increased the R-square by more than 15% to almost 37% variance explained. With this
adjustment, the percentage body fat in the ‘other beverages’ groups (msJ and msj) was
lower than in Model 2, and no longer significantly different from the ‘milk & juice’
(MSJ) pattern.

SES, physical activity and TV watching were similar in the way that these
covariates made adjustments to Model 2. The percentage body fat of the ‘milk & juice’
(MSJ) group was slightly higher than in Model 2 with adjustments for SES, physical
activity or TV watching. The Wald F was still significant with physical activity added to
Model 2, but there was no effect of beverage patterns in the overall model when SES,

weight loss behavior and TV watching were added as well.

224

 

Table 4. Building the linear regression model of the effects of beverage pattern on

percentage body fat in adolescents age 12-16 years1

 

 

Model Body Fat (%) P-value

Beverage Least Square Beta SE T-test P-value Model

Pattern2 Mean 2L. SE Coeff. Beta H0: B=0 Wald F3 Wald F3 R-Square
MODEL 1: beverage pattern, water & milk type 4.57 0.001 0.037

MSJ 22.5 i 1.9 0.03 2.19 0.989

MSj 21.7 :t 1.0 -0.74 1.59 0.644

MsJ 20.5 i: 0.9 -1.95 1.38 0.163

Msj 22.4 :1: 1.3 0.00 0.00 .

mSJ 22.0 :h 1.6 -0.43 2.13 0.842

mSj 23.0 :1: 1.0 0.54 1.46 0.713

msJ 26.8 :1: 1.2 *** 4.34 1.91 0.027

msj 31.9 :1: 2.9 *** 9.43 3.27 0.006
MODEL 1 + A (gender, age, ethnicity & sexual maturation) 3.19 0.007 0.176
MODEL 1 + B (region & nrral/urban area) 3.96 0.002 0.041
MODEL 1 + C (parent's weight status) 4.27 0.001 0.063
MODEL 1 + D (smoking, alcohol & drug use) 5.30 0.000 0.054
MODEL 1 + E (SES’) 4.46 0.001 0.078
MODEL 1 + F (trying to lose weight) 2.55 0.025 0.267
MODEL 1 + G (physical activity) 2.70 0.019 0.039
MODEL 1 + H (TV watching) 4.49 0.001 0.045
MODEL 2: model 1 + A, B, C & D 2.85 0.014 0.214

MSJ
MSj
MsJ
Msj
mSJ
mSj
msJ

msj

229ill
227¢L0
206i08
224iL2
2L4iL0
226i09
245iL1
305:11

*1:

**

0.53
0.33
-1.79
0.00
-0.96
0.16
2.15
8.11

225

2.24
1.50
1.31

0.00 .

1.43
1.31
1.66
3.43

0.815
0.826
0.178

0.504
0.906
0.202
0.022

Table 4, cont'd.

 

Model Body Fat (%) P-value
Beverage Least Square Beta SE T-test P-value Model
Pattern2 Mean a SE Coeff. Beta Ho: B=0 Wald F3 Wald F3 R-Square
MODEL 2 + E (SE33) 2.88 0.013 0.286
MSJ 23.1 d: 2.1 1.75 2.07 0.401
MSj 22.4 :1: 0.9 1.03 1.23 0.406
MSJ 21.2 d: 0.7 -0.14 0.94 0.881
Msj 21.4 i 0.7 0.00 0.00 .
mSJ 21.5 :1: 1.0 0.14 1.14 0.901
mSj 22.1 at 0.9 0.77 1.10 0.488
msJ 25.8 d: 1.3 ** 4.47 1.39 0.002
msj 29.5 i 2.8 ** 8.19 2.90 0.007
MODEL 1 + F (trying to lose weight) 2.27 0.044 0.370
MSJ 23.6 :1: 1.8 0.97 1.78 0.588
MSj 23.0 :1: 0.9 0.36 1.13 0.752
MsJ 21.0 i 0.9 -1.65 1.03 0.117
Msj 22.7 :1: 0.8 0.00 0.00 .
mSJ 20.9 :t 1.0 -1.76 1.20 0.148
mSj 22.1 :1: 0.9 -0.59 1.11 0.601
msJ 24.3 :1: 1.4 1.63 1.71 0.346
msj 26.7 :t 2.3 4.04 2.58 0.124
MODEL 2 + G (physical activity) 1.78 0.113 0.246
MSJ 22.9 :1: 2.1 1.03 2.17 0.636
MSj 22.7 i 0.9 0.85 1.41 0.551
MsJ 21.3 :t 0.9 -0.53 1.31 0.689
Msj 21.8 :1: 1.1 0.00 0.00 .
mSJ 22.1 :1: 1.0 0.21 1.40 0.882
mSj 22.6 :1: 0.9 0.72 1.29 0.579
msJ 24.6 a: 1.1 2.78 1.58 0.085
msj 29.8 :t 2.9 ** 7.95 3.21 0.017

226

Table 4, cont'd.

 

Model Body Fat (%) P-value
Beverage Least Square Beta SE T-test P-value Model
Patternz Mean :h SE Coeff. Beta H0: B=0 Wald F3 Wald F3 R-Square
MODEL 2 + H (TV watching) 2.13 0.058 0.231
MSJ 23.1 :I: 2.2 0.53 2.31 0.819
MSj 22.5 i 1.0 -0.10 1.43 0.946
MsJ 21.2 :1: 0.8 -1.41 1.26 0.267
Msj 22.6 i 1.2 0.00 0.00 .
mSJ 21.4 :1: 1.0 -1.22 1.43 0.397
mSj 22.3 :t 0.9 -0.30 1.30 0.818
msJ 24.3 :t 1.2 1.74 1.73 0.320
msj 29.9 :t 3.0 ** 7.33 3.33 0.032
OVERALL MODEL: model 2 + E, F, G & H 1.78 0.114 0.416
MSJ 24.1 i 1.9 2.46 1.86 0.193
MSj 22.4 d: 0.8 0.73 1.11 0.513
MsJ 22.1 :1: 0.8 0.48 1.01 0.635
Msj 21.6 :t 0.6 0.00 0.00 .
mSJ 21.3 i 0.9 -0.35 1.05 0.738
mSj 21.7 i 0.8 0.04 0.93 0.965
msJ 25.6 i 1.3 3.97 1.45 0.009
msj 25.1 :1: 1.9 3.52 2.07 0.095

 

1 Source: NHANES III, 1988-1994, adolescents aged 12-16 years, with complete,
reliable 24-hr recall dietary interview, and no missing or unknown responses to
beverage items on FF Q, excluding pregnant or lactating females (n=1,872).

2 The beverage pattern is the combination of beverages most frequently reported in the

FFQ. Capital letters M, S and J refer to intakes of milk, sugar-sweetened beverages
and juice, respectively, each having a share 225% of total beverage frequency, and
lowercase letters m, s and j represent intakes less than the specified amount

*** p< 0.01, **p<0.05, *p<0.10 for contrast of beverage pattern to 'milk & juice' (MSJ)

3 Wald F and p-value for the overeall beverage pattern effect

4 Characteristics pertaining to social economic status (SES) include household size, income,

and gender, age, marital status, education and employment of the household head

227

Overweight/Risk for Overweight

Table 5 shows the odds ratios for overweight/risk for overweight (285th percentile) in
each beverage pattern group. Compared to the ‘milk’ (Msj) group (reference category),
the odds ratio for overweight/risk for overweight was low in the ‘milk & juice’ (MSJ)
group (OR=O.52, 95% CI = 0.28-0.96), while the ‘other beverage’ (msj) group was more
likely to be at risk for overweight or overweight (OR=2.74, 95% CI = 1.12-6.68).
Although the linear regression analyses of BMI percentile showed effects of ‘sugar-
sweetened’ beverage patterns, none were observed in these logistic regression models.

When SES was added to Model 2, the odds ratio for the ‘other beverage’ pattern
(OR=2.47, 95% CI = 1.03-5.87), but not the ‘milk & juice’ pattern (OR=O.52, 95% CI =
0.25-1.11), was significant, while the Wald F for the overall beverage pattern effect
became significant at the p<0.01 level. The opposite was true of weight loss behavior.
When weight loss behavior was added to Model 2, the odds ratio for the ‘milk & juice’
pattern (OR=0.47, 95% CI = 0.23-0.97), but not the ‘other beverage’ pattern (OR=1.62,
95% CI = 0.65-4.02), was significant, even though the Wald F for the overall beverage
pattern effect was no longer significant (p>0.2).

Physical activity and TV watching were similar to SES in their effects. When
either of these covariates were added to Model 2, the odds ratio for the ‘other beverage’
pattern, but not the ‘milk & juice’ pattern, was significant. The Wald F was no longer
significant with TV watching added to Model 2. There were no further associations of
beverage patterns with overweight/risk in the overall model formed by adding SES,

weight loss behavior, physical activity, and TV watching together to Model 2.

228

Table 5. Building the logistic regression model of the effects of beverage pattern on
the likelihood of overweight/risk for overweight in adolescents age 12-16 yearsl’2

 

 

Model component Odds Lower Upper P-value

Beverage pattern3 Ratio4 Limit4 Limit" Wald F5 Wald F5
MODEL 1: beverage pattern, water & milk type 2.85 0.0142

MSJ 0.95 0.39 2.30

MSj 0.94 g 0.50 1.78

MsJ " , 0.52 ‘ 0.28 . _ 0.96

Msj 1.00 1.00 1.00

mSJ 0.89 0.43 1.81

mSj 1.14 0.67 1.92

msJ 0.97 0.42 2.20

msj 2.74 1.12   6.68
MODEL 1 + A (gender, age, ethnicity & sexual maturation) 2.33 0.0393
MODEL 1 + B (region & rural/urban area) 2.63 0.0216
MODEL 1 + C (parent's weight status) 3.12 0.0084
MODEL 1 + D (smoking, alcohol & drug use) 3.24 0.0066
MODEL 1 + E (SE85) 3.13 0.0083
MODEL 1 + F (trying to lose weight) 1.80 0.1082
MODEL 1 + G (physical activity) 1.88 0.0929
MODEL 1 + H (TV watching) 2.13 0.0581
MODEL 2: model 1 + A, B, C & D 2.55 0.0256

MSJ 0.88 0.31 2.52

MSj _ 0.99 _ 0.47 2.06

MsJ w 0.50 ' 0.26 .0499.

Msj 1.00 1.00 1.00

mSJ 0.87 0.44 1.73

mSj 1.13 0.66 1.93

msJ 1.02 0.45 2.32

msj ‘ f 2.91 , , 1.21 f 16.96

229

 

Table 5, cont'd.

 

Model component Odds Lower Upper P-value
Beverage pattern3 Ratio Limit Limit Wald F4 Wald F4
MODEL 2 + E (SE86) 3.65 0.0030
MSJ 1.19 0.42 3.34
MSj 1.00 0.46 2.18
MsJ 0.52 0.25 1.11
Msj 1.00 1.00 1.00
mSJ 0.87 0.46 1.65
mSj 1.18 0.64 2.20
ms] g 1.28 0.51 3 .22
msj 2.47 1 1.03 5.87
MODEL 1 + F (trying to lose weight) 1.42 0.2193
MSJ 0.93 0.35 2.51
MSj 0.97 . 0.46 g 2.01
W 40.47 0.2.3   . 0.97
Msj 1.00 1.00 1.00
mSJ 0.77 0.38 1.55
mSj 1.02 0.59 1.75
ms] 0.98 0.40 2.44
msj 1.62 0.65 4.02
MODEL 2 + G (physical activity) 1.96 0.0799
MSJ 0.97 0.33 2.89
MSj 1.09 0.51 2.30
MS] 0.64 0.33 1.25
Msj 1.00 1.00 1.00
mSJ 1.15 0.58 2.28
mSj 1.29 0.75 2.22
msJ 1.15 0.49 2.72
msj 2.80 f , 1.12 6.96

230

 

Table 5, cont'd.

 

Model component Odds Lower Upper P-value
Beverage pattern3 Ratio Limit Limit Wald F4 Wald F4
MODEL 2 + H (TV watching) 1.64 0.1465
MSJ 0.88 0.30 2.60
MSj 0.92 0.44 1.92
MSJ 0.54 0.28 1.06
Msj 1.00 1.00 1.00
mSJ 0.84 0.44 1.60
mSj 1.04 0.60 1.82
msJ 20.95 _ g 0.39 2.34
msj _ 2.56 1.04 6.34
OVERALL MODEL: model 2 + E, F, G & H 1.35 0.2471
MSJ 1.46 0.53 3.99
MSj 0.94 0.42 2.12
MSJ 0.58 0.24 1.38
Msj 1.00 1 .00 1.00
mSJ 0.94 0.48 1.84
mSj 1.09 0.58 2.04
ms] 1.32 0.43 4.05
msj 1.09 0.3 8 3 .07

 

1 Source: NHANES 111, 1988-1994, adolescents aged 12-16 years, with complete,
reliable 24-hr recall dietary interview, and no missing or unknown responses to
beverage items on F F O, excluding pregnant or lactating females (n=1,872).

2 The beverage pattern is the combination of beverages most frequently reported in the
FFQ. Capital letters M, S and J refer to intakes of milk, sugar-sweetened beverages
and juice, respectively, each having a share 225% of total beverage frequency, and
lowercase letters m, s and j represent intakes less than the specified amount

3 Overweight/risk for overweight defined by BMI 285th percentile of BMI-for-age

4 Shaded odds ratio (95% confidence interval) is significantly different from 1.0 (p<0.05)

5 Wald F and p-value for the overall beverage pattern effect

6 Characteristics pertaining to social economic status (SES) include household size, income,

and gender, age, marital status, education and employment of the household head

231

DISCUSSION

This study provides much needed data regarding interrelationships between
sociodemographic/lifestyle factors, beverage patterns and adolescent obesity in the U.S.
That low least square (LS) mean BMI percentile, percentage body fat, and the odds ratio
for overweight/risk for overweight in the ‘milk & juice’ group were adjusted upwards
when covariates were added to the models indicated that the ‘milk & juice’ pattern was
associated with high levels of SES and physical activity, and with low levels of TV
watching. Adjustments for the effects of covariates that reduced the high LS mean BMI
percentile of the ‘sugar-sweetened beverage’ group indicated that the ‘sugar-sweetened
beverage’ pattern was associated with low levels of SES and physical activity, and with
high levels of TV watching. These results were consistent with other analyses of soft
drinks and BMI controlled for income, sports team participation and watching TV
(Forshee & Storey, 2003; Forshee et al., 2004).

An increased prevalence of overweight among low-income adolescents may be
due in part to the quality of their diets, when primarily less expensive, energy-dense
foods and beverages are consumed (Alaimo et al., 2001; Drewnowski & Specter, 2004;
Drewnowski & Darmon, 2005). Additionally, low-income adolescents living in urban
environments might watch TV, if they have few opportunities to be physically active
(Elkins et al., 2004). Compounding such sedentary behavior is the fact that soft drinks
are frequently consumed while watching TV (Coon et al., 2001; Giammattei etal., 2003;
Utter et al., 2003; Matheson et al., 2004; Phillips et al., 2004). Sugar—sweetened beverage
consumption and physical inactivity have been associated with poverty among

adolescents age 12—17 years in NHANES, 1999-2002 (Miech et al., 2006). Still, it is

232

 

unclear whether disparities across poverty groups in overweight prevalence are driven by
the social environment, built environment, income and education of parents and/or
familial food and physical activity preferences and practices (Link & Phelan, 1995; Diez
Roux, 2005; Sallis & Glanz, 2006; Gordon-Larsen et al., 2003).

Research showing the effects of SES on adolescent overweight suggested that
poverty was associated not only with sugar-sweetened beverage consmnption and
physical inactivity, but with breakfast skipping as well (Miech et al., 2006). In analyses
of NHANES 111 data, eating breakfast some days or every day was associated with
normal body weight only among adolescents having one or two obese parents (Fiore et
al., 2006). Parental overweight/obesity was a potential covariate considered in the
current analyses as well.

The current analyses of beverage patterns and adolescent obesity were not
adjusted for the effects of breakfast consumption habits. There were several reasons for
excluding this potential covariate from analyses. First, the association between breakfast
skipping and overweight has not been firmly established (Berkey et al., 2003;
Rampersaud et al., 2005), and other NHANES HI analyses of the association of
beverages and overweight have also omitted breakfast habits from their models (Forshee
& Storey, 2003; Forshee et al., 2004). Second, preliminary research (Chapter 4, Table 6)
and other analyses of beverages consumed by adolescents (Guenther et al., 1986;
Bowman et al., 2002) suggested that the ‘milk & juice’ pattern would be a good indicator
of breakfast consumption, and breakfast habits might then ‘over-control’ the potential

association of the ‘milk & juice’ pattern with overweight.

233

Preliminary analyses of the frequency with which adolescents consumed
beverages at the various meal occasions demonstrated that milk and juice were most
frequently consumed at the breakfast meal occasion (Chapter 4, Table 6). In 1977-78, the
shares of milk and juice consumed at breakfast were 38% and 76%, respectively
(Guenther et al., 1986). Drinking juice was highly specific to the breakfast occasion
(Guenther et al., 1986). Among adolescent girls age 12-19 years, between 36-44% of
non-milk drinkers did not eat breakfast, compared to 11-12% of milk drinkers (Bowman
et al., 2002). Thus, consumption of both milk and juice is likely a good indicator that the
youth ate breakfast.

The model shown in Figure 1 depicts how the beverage pattern assignment can
be a marker for meal situation variables, such as the meal occasion (e.g., breakfast, lunch,
dinner, snack) and meal location (e.g., home or places away from home such as school or
a fast food restaurant) that hypothetically determine which beverages are frequently
consumed by youth. Because milk and 100% fruit juice were frequently consumed at
breakfast (Chapter 4, Table 6), breakfast-eating habits might determine whether youth
were classified as milk and 100% fi'uit juice consumers.

Beverage patterns were defined a priori to classify youth into mutually exclusive
groups according to the combination of beverages he/ she most frequently consumed.
Only one other researcher has classified youth by consumption of milk and/or soft drinks
(Bowman et al., 2002), but the current study differed by using FFQ rather than 24-hour
recall data, and also controlled for juice consumption while contrasting beverage pattern
groups to determine effects of the various types of beverages. Cluster analysis could

have been’used to empirically derive beverage patterns and classify youth into mutually

234

.362... .53 manor-E omega: he meufloemma 2: me .25.: 12.33.30 A 2:»:—

 

 

 

 

r—fl

\

//

 

9.35595—

 

553.
33s.:

 

 

 

 

 

w

«

 

3.5—.m— . u ,

 

4‘

mam—.833 PH

 

 

 

 

$13.: Rheum

 

 

 

 

332—2. «.8— Emma?
.38? Emu—cam
Em REP—am
use: Eco mam .0385 a:
5.2223— .539“
beifim .8..an .63
33%

 

Eaten
mus—95m

 

 

 

 

 

 

 

.5339— 132
5:330 .32

 

 

 

 

 

fl 85.... .8388:

 

235

exclusive groups (Kant, 2004), but each individual would not have necessarily consumed
all beverages that characterized his/her cluster. A cluster analysis of water, beverages
and selected foods consumed by adults showed that higher—educated, older adults (>60
years) were more likely to be grouped with a “water/fi'uit & vegetable/low fat milk”
cluster, while young adults (18-35 years) were more likely to be grouped with a “fast
food/snack/soft drink” cluster (Popkin et al., 2005).

Preliminary analyses showed that sugar-sweetened beverages were frequently
consumed at fast food restaurants and at snacks (Chapter 4, Table 6). Other researchers
have shown that soft drinks are frequently consumed while snacking and watching TV
(Coon et al., 2001; Giammattei et al., 2003; Matheson et al., 2004; Phillips et al., 2004),
and with meals eaten at fast food restaurants (French et al., 2001b; Bowman et al., 2004).
Thus, fast food restaurant use, TV watching and snack-eating habits might determine
whether youth are classified as sugar-sweetened beverage consumers. Youth who are
characterized by beverage patterns that include sugar-sweetened beverages might be
more overweight than other youth because of the energy intakes contributed by these
beverages (Hamack et al., 1999; Troiano et al., 2000; Bowman, 2002), or because of the
energy intakes contributed by snack foods and fast foods typically consumed with sugar-
sweetened beverages (Hampl et al., 2003; Utter et al., 2003; Bowman et al., 2004;
Phillips et al., 2004).

The model shown in Figure 1 depicts energy intake as the mediator between
beverage patterns and obesity. Investigators who conducted a prospective study of
preschool children excluded energy intake from analyses, because it was the mediating

variable in the pathway between the exposure (beverage consumption) and the outcome

236

(changes in weight or BMI) (Newby et al., 2004). For this reason, energy intake was
excluded from the current analyses. Instead, the mean energy intake of beverage pattern
groups was examined in a separate study (Chapter 4, Table 7). In the current study,
youth classified by beverage patterns associated with fast food restaurant use or snacking
behavior might be more overweight than youth classified by beverage patterns associated
with breakfast consumption, due to potential energy intake differences between the
beverage pattern groups.

Inverse associations of the ‘milk & juice’ pattern with BMI percentile, percentage
body fat, and overweight/risk for overweight were observed in the current study. Weight
loss behavior was included in the current study to examine whether the correlation
between overweight and ‘milk & juice’ was due to overweight youth skipping breakfast
to lose weight. College women have reported that they skipped breakfast as a weight loss
strategy (Malinauskas et al., 2006), even though breakfast skipping might lead to weight
gain over time (Berkey et al., 2003). Breakfast skippers in a longitudinal study tended to
gain more weight (measured as BMI) than regular breakfast eaters, but only among
normal-weight subjects; BMI decreased in overweight adolescents who skipped breakfast
(Berkey et al., 2003).

Potential confounding provides another explanation for the inverse association of
the ‘milk & juice’ pattern with overweight. Other research suggested that breakfast
consumption might be associated with components of a healthy lifestyle such as high
levels of physical activity (Cavadini et al., 2000b; Keski-Rahkonen et al., 2003; Godin et
al., 2005). In preliminary analyses, the ‘milk and juice’ pattern was associated not only

with regular breakfast consumption, but also with high SES and other healthy behaviors

237

like exercise and not watching TV (Chapter 5, Table 2), and these other factors could
account for the association between the ‘milk & juice’ pattern and the low prevalence of
overweight. In the current analyses, effects of the ‘milk & juice’ pattern were no longer
significant when indicators of SES, physical activity and TV watching were added to the
models.

In this study, sports team participation was associated with obesity outcomes, but
exercise was not (data not shown), similar to another study using NHANES HI data
(Crespo et al., 2001). Others have omitted the exercise variable and used only the
measurement of team sports participation to examine the relationship between physical
activity and BMI (Forshee et al., 2004). Exercise has been positively associated with the
prevalence of normal BMI in analyses of CSFII data (Boumtje et al., 2005).

In cross—sectional analyses, the associations of obesity outcomes with lifestyle
factors like exercise, smoking and dietary patterns should be interpreted with caution.
Because cross-sectional data lack temporality, reverse causality must be considered when
interpreting these observations, i.e., overweight youth might be exercising to lose weight.
The association of smoking with overweight might also be due to reverse causality.
Youth who reported they were trying to lose weight were more likely to smoke (French et
al., 1994; Austin et al., 2001; Lowry et al., 2002; Delnevo et al., 2003).

The association of the ‘other beverages’ pattern with obesity outcomes was
another example of reverse causality. As indicated by preliminary analyses of beverage
frequency, the ‘other beverages’ groups drank ‘diet’ soft drinks, coffee and tea more
frequently than the ‘caloric beverages’ groups (Chapter 4, Table 5), and a high

percentage of youth in the ‘other beverages’ group reported trying to lose weight

238

 

 

 

(Chapter 5, Table 2). Rather than causing overweight, drinking ‘diet’ sofi drinks was
possibly related to the weight loss efforts of the ‘other beverages’ group. In the current
study, the association of the ‘other beverages’ group with obesity outcomes was no
longer statistically significant when analyses controlled for youth who were trying to lose
weight. Drinking “diet” beverages or bottled water instead of sugar-sweetened beverages
decreased BMI among those in the upper BMI tertile at baseline in a 25-week

intervention study of non-dieting youth age 13-18 years (Ebbeling et al., 2006).

Strengths and Limitations

The anthropometric and dietary assessment methods were the major strengths of this
study. The youth’s BMI percentile was calculated from measured height and weight
using the gender-specific BMI-for—age percentile reference data provided by the CDC
Growth Charts. In addition, a measure of percentage body fat was used to assess body
composition. Youth were classified into beverage pattern groups according to their
‘usual’ beverage intakes reported using the FFQ. Because of intra-individual variation in
dietary intake, youth might have been misclassified to beverage patterns if a single 24-
hour recall were used instead of the FFQ. Therefore, an assessment of ‘typical’ beverage
patterns (i.e., ‘usual’ intake) should strengthen the relationship of beverage consumption
with obesity outcomes. In addition, misclassification to beverage pattern groups due to
under-reporting in response to the F FQ was addressed by identifying most frequently
consumed beverage types based on their percentage share of total beverage frequency.
Rather than using the beverage frequency to identify youth who reported consumption 21

time/day, for example, youth were classified by the combination of beverages consumed

239

with shares 225% of total beverage frequency. If frequency-based criteria were used,
youth who reported drinking all caloric beverage types <1 time/d would have been
classified to the ‘other beverages’ (msj) pattern, and these “under-reporters” were
removed from this group by using consumption criteria based on the percentage share.
Avoiding potential bias due to under-reporting was important to the design of this study,
because overweight youth might be more likely to under-report beverage consumption
(Bandini et al., 1990; Johnson et al., 1996; Johnson, 2000).

There were several limitations that pertain to the current analyses. Some youth
might have been misclassified to a ‘juice’ group, if they mistakenly reported fruit drink
frequency in response to the 100% fruit juice question, because the fruit juice question
was sequenced before the question about fruit drinks in the FFQ. Beverage items in the
FFQ did not distinguish sugar-sweetened coffee/tea from coffee/tea consumed without
added sugar. Therefore, the ‘other beverages’ group might have included both sugar-
sweetened coffee/tea consumers and diet soft drink consumers. Effects of sugar-
sweetened beverages on overweight might have been diluted by misclassification. In
addition, there were four separate patterns that included sugar-sweetened beverages
compared to the ‘milk & juice’ pattern, and it might be necessary to collapse these four
patterns to identify a potential association between sugar-sweetened beverages intake and

obesity in youth.
Implications

This study suggests that several sociodemographic/1ifestyle factors examined are highly

correlated with beverage patterns, and together these sociodemographic/lifestyle factors

240

and beverage patterns are associated with adolescent obesity. The interrelationships
between sociodemographic/1ifestyle factors and beverage patterns must be carefully
considered to elucidate the potential role each specific etiologic factor has in increasing

the prevalence of adolescent overweight.

241

Chapter 7

CONCLUSION

The concluding chapter provides a summary of key findings, and an assessment
of the overall strengths and limitations of this research. The chapter concludes with

implications for firture research suggested by findings from this dissertation.

A. Summary of Findings

The aims of this research were to examine the associations of beverages with the
meal context; to examine the associations of beverage patterns (i.e., types of beverages
most frequently consumed) with dietary intake and sociodemographic/1ifestyle risk
factors for overweight; and to examine associations of beverage patterns with BMI
percentile, percentage body fat and overweight/risk for overweight, adjusted for potential
covariates of interest. Youth were classified by their ‘typical’ beverage pattern using the
‘share’ method to identify the combination of beverage types (i.e., milk, juice and/or
sugar-sweetened fruit drinks/soft drinks) that were each reported frequently (i.e.,
beverage frequency ‘share’ 225% of the total beverage frequency) in a food fi'equency
questionnaire. -

A key, and somewhat unexpected, finding of this study was that the ‘milk &
juice’ pattern, not the ‘milk’ pattern, was inversely associated with the BMI percentile
and percentage body fat in most adjusted regression models. In addition, the odds ratio
for overweight/risk for overweight was low in the ‘milk & juice’ group, compared to the
‘milk’ group, but these effects disappeared when the models were adjusted for SES,

physical activity (i.e., exercise and sports team participation) or TV watching.

242

Another key finding pertaining to ‘milk & juice’ was that this pattern was
associated with breakfast consumption. Milk and juice were consumed frequently at the
breakfast meal, and a high percentage of the ‘milk & juice’ group reported eating
breakfast daily. The ‘milk & juice’ pattern was also positively associated with exercise
and sports team participation, and inversely with TV watching. In addition, the ‘milk &
juice’ pattern was positively associated with income, suggesting that income might also
be associated with breakfast consumption and other lifestyle factors correlated with the
‘milk & juice’ pattern.

Sugar-sweetened beverages were consumed more frequently at the snack occasion
than at meals. In addition, soft drinks were consumed more frequently than other
beverages at fast food restaurants. Energy intakes did not differ between beverage
pattern groups, but ‘sugar-sweetened beverages’ groups had higher added sugars and
lower calcium intakes than ‘milk’ groups. The ‘sugar-sweetened beverages’ pattern was
positively associated with BMI percentile after adjusting for physical activity, but not
after adjusting for TV watching, suggesting that watching TV was strongly associated
with intakes of soft drinks for overweight youth in this study. That sugar-sweetened
beverage patterns were associated inversely with income, and positively with watching
TV, suggests that income might be inversely associated with watching TV as well.

The ‘other beverages (diet soft drinks, coffee and tea)’ pattern was positively
associated with both BMI percentile and percentage body fat. In addition, the odds ratio
for overweight/risk for overweight was high in the ‘other beverages’ group, compared to
the ‘milk’ group. However, a high percentage of youth in the ‘other beverages’ group

reported that they were trying to lose weight. Together, these findings indicate that

243

overweight youth were drinking diet sodas to lose weight. Cigarette smoking was also
prevalent among those youth in the ‘other beverages’ group. In addition, energy and
calcium intakes were lower, and caffeine intakes were higher in ‘other beverages’ groups
than in ‘caloric beverages’ groups. However, mean intakes of the ‘other beverages’
group should be interpreted with caution, because the sample size of this group, in

particular, was relatively small (n=55).

B. Strengths and Limitations

Strengths of this study include the methods used to assess overweight and
adiposity in adolescents. Cross-sectional studies using the Continuous Survey of Food
Intake by Individuals (CSFII) data to examine the association of beverage intakes with
BMI or overweight in adolescents (Forshee & Storey, 2003; Boumtje et al., 2005) were
limited in that self-reported information rather than measured height and weight were
used to compute BMI (Kuczmarski et al., 2001; Brener et al., 2003). In addition,
analyses of children or adolescents using National Health and Nutrition Examination
Survey (NHANES) data that examined BMI in relation to beverage intakes (F orshee &
Storey, 2003; F orshee et al., 2004) were limited, because BMI was not adjusted for
gender and age using the CDC Growth Charts reference data (Kuczmarski et al., 2000;
Bachman et al., 2006). In the current study, BMI percentile was calculated fi'om
measured height and weight using the gender-specific BMI-for-age percentile reference
data provided by the CDC Growth Charts, because criteria based on the BMI-for-age
percentiles were recommended to assess overweight/risk for overweight in children and

adolescents (Kuczmarski et al., 2000). In addition, percentage body fat was determined

244

from summed triceps and subscapular skinfold measurements (Slaughter et al., 1988),
because BMI does not accurately assess body composition, e.g., overweight youth with a
high proportion of lean body mass are not overfat (Forbes, 1987).

Cross-sectional studies using 24-hour recall data available fi'om CSFII or
NHANES (Forshee and Storey, 2003; Forshee et al., 2004; Boumtje etal., 2005) might
have been limited by inaccurate assessment of beverage consumption due to intra-
individual variation in intake (Beaton et al., 1979; Sempos et al., 1985). When intake of
an individual is assessed, intra—individual variation can be addressed by averaging
repeated measurements obtained from multiple 24-hour recall interviews or by asking the
individual to average his/her intake over multiple days in response to a FFQ.

When the mean nutrient intake of a group is estimated using a random sample of
the population like that found in the national surveys, intra-individual variation is
addressed through the use of a random process, according to the stochastic model
(Parzen, 1962). In the stochastic model, because of the random process, theoretically,
each sample person in a group represents a repeated measurement obtained from the
average individual characterized by that group (Parzen, 1962). The mean nutrient intake
of a group is not biased by intra-individual variation, because it represents the mean of
repeated measurements obtained from the average individual in the group.

Although the mean intake of the group is an unbiased estimate of the average
individual’s true mean intake (Guenther et al., 1997), intra-individual variation affects the
precision of the mean estimate, and two groups are not significantly different if the
within-group (or within-person) variation is larger than the between-group (or between-

person) variation (Sempos et al., 1985). Increasing the sample size increases the

245

precision of the estimate. Therefore, the variance in the estimate of the mean intake of a
group is reduced when the sample size is large, as in the national surveys like NHANES.
In the current study, youth were classified into mutually exclusive groups

according to the combinations of beverage types most frequently reported as ‘usual’
intake using a F FQ. The FFQ was comprehensive (i.e., all beverage items were
included), but it had not been quantified or ‘validated’ to estimate nutrient intakes
(Briefel, et al., 1992; Briefel, 1994). Youth were classified by beverage patterns to allow
for the use of 24-hour recall data to estimate mean nutrient intake of beverage pattern
groups, because the mean estimate for a group would not be biased by intra-individual
variation (Guenther et al., 1997).

Beverage patterns were defined as combinations of beverage types so that
independent effects of individual beverage types on energy intake or obesity outcomes
could be determined through comparisons of various beverage pattern groups. In
addition, using dummy variables to represent the beverage pattern groups instead of
continuous measures of beverage intakes in the regression models addressed the potential
for multicollinearity interfering with the investigation of obesity outcomes.

Youth were classified into beverage pattern groups according to their ‘usual’
beverage intakes reported using the FFQ. Because each individual was classified by
his/her own beverage intakes, due to intra-individual variation in dietary intake, youth
might have been classified incorrectly into beverage pattern groups if a single 24-hour
recall were used instead. Therefore, assessing each youth’s ‘typical’ beverage pattern
using ‘usual’ intakes in the FFQ should strengthen the relationship of beverage

consumption with sociodemographic/ lifestyle factors or obesity outcomes.

246

Under-reporting of intake was still a concern, however, as the Observing Protein
and Energy Nutrition (OPEN) study has shown that it occurs when either the 24-hour
recall or FFQ methods are used to assess intake (Subar et al., 2003). Avoiding potential
bias due to under-reporting was important to the design of this study, because overweight
youth might be more likely to under-report intake (Bandini et al., 1990; Johnson et al.,
1996; Johnson, 2000). This study was limited by the potential bias due to under-
reporting in estimating mean energy intakes of beverage pattern groups that were
associated with overweight. Low energy intakes in the ‘other beverages’ group might be
due to overweight youth trying to lose weight or under-reporting of intakes in the 24-hour
recall dietary interview. In addition, if ‘sugar—sweetened beverage’ groups tended to be
more overweight than ‘milk’ groups, then under-reporting among the ‘sugar—sweetened
beverage’ groups could explain why there were no energy intake differences between
these groups.

Misclassification to beverage pattern groups due to under-reporting in response to
the F FQ was addressed by identifying most fi'equently consumed beverage types based
on their percentage share of total beverage frequency. Rather than using the beverage
frequency to identify youth who reported consumption 21 time/day, for example, youth
were classified by the combination of beverages consumed with shares 225% of total
beverage frequency. If fi'equency-based criteria were used, youth who reported drinking
all caloric beverage types <1 time/d would have been classified to the ‘other beverages’
pattern, and these “under-reporters” were removed from this group by using consumption

criteria based on the percentage share.

247

However, misclassification due to under-reporting may still be a concern if sugar-
sweetened beverages were under-reported more often than milk or juice, for example. In
the current study, overall, regular soft drinks were reported less often in the FF Q than in
the 24-hour recall, while milk and juice frequencies were higher in the FFQ than in the
24-hour recall. If the percentage share of ‘sugar—sweetened beverages’ were under-
reported in the F FQ, then youth might have been incorrectly classified to beverage
pattern groups. Effects of drinking sugar-sweetened beverages on overweight might have
been diluted, if youth were not classified correctly to ‘sugar—sweetened beverages’
groups.

Some sugar-sweetened beverage consumers might have been misclassified to a
‘juice’ group, if they mistakenly reported fi'uit drink frequency in response to the 100%
fruit juice question, because the fruit juice question was sequenced before the question
about fi'uit drinks in the FFQ. The ‘sugar—sweetened beverage’ category was limited to
sofi drinks and fruit drinks that contained added sugar. Sugar-sweetened coffee and tea
were not included in the ‘sugar—sweetened beverage’ category, because the F FQ did not
distinguish these beverage items fiom coffee and tea consumed without added sugar.

The FFQ instrument available in NHANES III was also limited because it had not
been quantified or ‘validated’ to estimate energy intakes (Briefel, et al., 1992; Briefel,
1994). Although it was tested in the National Center for Health Statistics Cognitive
Laboratory and pretested, the NHANES III F FQ was not ‘validated’ for nutrient intake
estimates against another dietary method, because it was not intended for quantifying
nutrient intakes (Briefel, personal communication). Further, the validity of the FFQ

nutrient intake estimates, and the ability of respondents to appropriately respond to a FFQ

248

 

in order to estimate quantities of foods consumed (e. g., “How often did you eat a '/2 cup
serving of broccoli over the past year?”) has been called into question (Kristal et al.,
2005). In addition, results of the OPEN study indicated that the low attenuation factor
(0.04-0.16) observed with a single administration of a FFQ is not improved by repeat
administration, but the attenuation factor for a single 24-hour recall (O.10-O.20) increases
to 0.35-0.50 with four repeats (Schatzkin et al., 2003). These results from the OPEN
study (Schatzkin et al., 2003) suggest that the 24-hour recall nutrient intake estimates
would be more accurate than those obtained from a FFQ, if multiple days of 24-hour
recall intake were available, unless the test-retest reliability of the FFQ can be improved
(perhaps by reordering or clarifying the wording of the questionnaire items, for example).
Results from a food-based validation study that examined frequency of food and
beverage consumption from repeated FFQ and repeated 7-day records indicated intra-
individual week-to-week variation in intake was lower for beverages than meat or
vegetables (Salvini et al., 1989). Therefore, beverages may require fewer repeated
measures than does food to reliably assess frequency of consumption (Salvini et al.,
1989) using the 24-hour recall method. Results of this study also suggested that response
to the FFQ tended to over-represent “socially desirable” foods (Salvini et al., 1989).

Because the F FQ instrument in NHANES III was not intended for quantifying
nutrient intakes (Briefel, et al., 1992; Briefel, 1994), reference amounts customarily
consumed (RACC) per eating occasion (Food and Drug Administration, 2002) were not
given with instructions for the respondent to adjust his/her frequency response for intake
amounts that were more or less than the RACC. Even if the caloric intake from

beverages were estimated using the frequency response and the RACC, the analysis plan

249

in the current study was to classify youth by the percentage of total beverage frequency
rather than by the percentage of total calories in order to account for intakes of non-
caloric beverages containing caffeine like coffee, tea, and diet soft drinks that might be
associated with overweight. Classifying youth by beverage frequency or weighting the
frequency by the RACC would have the same result, because the RACC, the serving size
used for labeling purposes, is the same, 240 mL (8 fl oz), for carbonated beverages,
noncarbonated beverages, milk and juice (Food and Drug Administration, 2002). If
actual serving sizes that are larger than the RACC were more likely for carbonated soft
drinks than for milk and juice, then the ‘sugar-sweetened beverage’ group would be more
likely to have higher energy intakes fiom beverages than ‘milk and juice’ group.
Because of intra-individual variation in the actual amounts consumed per eating occasion,
it may be difficult for an individual to weight his/her frequency response by the number
of recommended portions he actually consumed. Therefore, in the current study, energy
intake from beverages was estimated using the 24-hour recall rather than the FFQ data.
The mean intake of the beverage pattern group accounts for intra-individual variation,
because it represents the mean of repeated measurements obtained from the average
individual in the group.

The FFQ was limited in that it did not include a question about water intake.
Water intake was also not assessed in the 24-hour recall dietary interview. Instead, youth
were asked a separate question to estimate the amount (fluid ounces) of water they
usually drink per day. Because water intake was not reported in terms of frequency
(times/day), water intake could not be added to the total beverage frequency to adjust the

beverage’s percentage share for the amount of water consumed. If the ‘milk and juice’

250

group tended to drink more water, then presumably they would need to obtain less water
for hydration from beverages and the total beverage frequency would be lower than the
‘sugar-sweetened beverage’ group. However, the FFQ had not been validated to estimate
the amount of water obtained from beverages, and the question about water intake had
not been validated either. Further, there was a concern that overweight youth might be
biased toward under-reporting intake in the FFQ and over-reporting water intake at the
same time. Therefore, youth were classified into beverage pattern groups according to
the beverage’s percentage share without adjusting for the total beverage frequency or
amount of water consumed. Instead, the mean water intake and total beverage frequency
of beverage pattern groups were examined separately, and the measure of association
between beverage patterns and overweight/risk for overweight was adjusted for the
amount of water consumed. Although there may have been a reporting bias, the
estimates of the groups’ mean water and beverage intakes were not biased by intra-
individual variation.

Because the F FQ did not allow the respondent to indicate whether sugar was
added to coffee or tea, the ‘other beverages’ group included both sugar-sweetened
coffee/tea consumers and diet soft drink consumers, and risk factors might have varied
widely between these subgroups. In addition, the ‘juice’ groups included both orange
juice and apple juice consumers. Because the association between 100% fruit juice
consumption and overweight/risk for overweight might depend on the type of juice
usually consumed, it might be important to indicate whether it was citrus juice (e.g.,
orange juice and grapefruit juice) or non-citrus juice (e. g., grape juice, apple juice,

cranberry juice and fruit nectars). On the other hand, because there were four separate

251

patterns that included sugar-sweetened beverages and shared similar risk factors, it might
be necessary to collapse them to identify a potential association between sugar-sweetened
beverage consumption and obesity in youth.

The most important limitation of this study is that cross-sectional data lack
temporality, and cannot be used to make causal inferences. Interpretation of results of
this study was difficult when an unexpected finding could be explained by reverse
causality. For example, when low calorie “diet” soft drinks were positively associated
with body weight, there were two possible explanations to consider, i.e., “diet” soft
drinks promote weight gain (which is contradictory to what was expected), or overweight
youth drink “diet’ beverages because they are trying to lose weight. Although these
cross-sectional data cannot be used to establish which explanation for the positive
association between “diet’ soft drinks and body weight was correct, a theoretical model
was proposed and the statistical analysis adjusted the measures of association for the

body weight of youth who were trying to lose weight.

C. Implications

The frequency with which beverages were consumed in various meal settings was
examined in this study, and youth were classified into beverage pattern groups according
to the types of beverages he/she consumed most frequently. Because beverage choice
was associated with the meal context, the beverage pattern could be used to indicate how
often youth consume different types of meals (e. g., breakfast, snacks, meals eaten at fast
food restaurants). Because milk and juice were both consumed more frequently with the

breakfast meal than any other type of meal occasion, when an adolescent was classified

252

by the ‘milk & juice’ pattern (by determining that he/she most frequently consumed milk
and juice from ‘usual’ intakes reported in the FFQ) it was a good indication that he/she
usually ate breakfast. A high percentage of the ‘milk & juice’ group (57%) reported that
they ate breakfast every day.

The ‘milk & juice’ pattern and breakfast eating were both inversely associated
with overweight/risk for overweight. These results provide additional support for health
professionals to recommend that parents encourage youth to eat breakfast daily to reduce
the risk for overweight in adolescence (Ritchie et al., 2005). The inverse associations of
the ‘milk & juice’ pattern with indicators of adolescent obesity remained after controlling
for youth who were trying to lose weight, which suggests that it’s unlikely these results
were due to reverse causality, i.e. overweight youth skipping breakfast to lose weight.
Analyses of various sociodemographic/lifestyle risk factors for overweight that were
associated with the ‘milk & juice’ pattem suggested that eating breakfast regularly was
associated with other components of a healthy lifestyle like drinking reduced fat (2%) or
low fat (1%) milk, exercising daily, and watching TV less than one hour per day. Thus,
the ‘milk & juice’ pattern was effectively a marker for breakfast consumption and other
components of a healthy lifestyle, and these lifestyle characteristics together with the
‘milk & juice’ pattern were inversely associated with adolescent obesity. The results of
regression analyses that were adjusted for income suggested that youth in low-income
households might be especially at risk for overweight if they skip breakfast and have low
levels of physical activity.

Analyses of the meal context in which sugar-sweetened beverages were consumed

indicated that fruit drinks and soft drinks were more frequently consumed at the

253

snack/beverage occasion than with meals like breakfast, lunch or dinner, and soft drinks
frequently accompanied meals eaten at fast food restaurants. Because youth were
classified by beverage patterns according to the types of beverages he/she consumed most
frequently, that sugar-sweetened beverages were associated with the snack occasion and
meals eaten at fast food restaurants suggests that classification into the ‘sugar—sweetened
beverage’ group was determined by how often youth eat snack foods and fast foods.
Other researchers have shown that soft drinks are often consumed with pizza, chips, and
fast food while viewing television (Coon et al., 2001; Matheson et al., 2004), and in the
current study, the ‘sugar-sweetened beverages’ pattern was associated with TV watching.
These results then suggest that snacks consumed while watching TV were likely to
include soft drinks, consumed either alone as a snack/beverage or with other snack foods.
A pattern of consuming ‘sugar-sweetened beverages’ could potentially be a risk
factor for overweight, because of the higher added sugar and lower calcium intakes in
‘sugar-sweetened beverages’ groups than in ‘milk’ groups, even though energy intakes
did not differ. Because the ‘sugar-sweetened beverages’ pattern was associated with TV
watching, regression analyses indicated that the ‘sugar-sweetened beverages’ pattern and
TV watching were both positively associated with the BMI percentile. Results of
regression analyses adjusted for income suggested that youth in low-income households
might be especially at risk for overweight if they drink sugar-sweetened beverages and
watch TV for 4 hours/day or more. These results provide additional support for health
professionals to recommend that parents limit television-viewing time to reduce the risk

for overweight in adolescence (Ritchie et al., 2005).

254

 

Because beverage choice was associated with the meal context, theses findings
support that parents should establish regular meal times and encourage youth to limit the
number of snacks consumed between meals. Family meals should be eaten while seated
at a table rather than in front of a TV. Rather than serving foods and beverages typically
consumed while watching TV, like fast foods, pizza and soft drinks (Coon et al., 2001;
Matheson et al., 2004), the meal preparer should plan to serve foods and beverages that
help youth meet the recommendations of the MyPyramid Plan (USDA, 2005). Because
the typical adolescent should be consuming 3 cups of milk per day (USDA, 2005), it is
important that milk is consumed at meal times. Because beverage choice might be
limited in some settings, for example, only soft drinks might be available in fast food
restaurants, a suggestion to intervene outside the home environment would be for policy-
makers to develop and implement a marketing plan to offer alternative beverages with
meals consumed away-from-home.

Although the ‘other beverages (diet soft drinks, coffee and tea)’ pattern was
associated with adolescent obesity, results of this study suggested that this association
was likely due to reverse causality, i.e., overweight youth were drinking ‘diet’ soda to
lose weight. An intervention study has shown that replacing regular soft drinks with low-
calorie soft drinks could potentially prevent weight gain in youth (Ebbeling et al., 2006).
Health practitioners should note, however, that youth who diet to lose weight might be
potentially at risk for developing unhealthy weight loss practices, like breakfast skipping,
smoking cigarettes, using substances containing caffeine, fasting or purging (French et
al., 1995; Krowchuk et al., 1998; French et al., 2001; Austin et al., 2001; Lowry et al.,

2002; Delnevo et al., 2003, Malinauskas et al., 2006). Results of the current study

255

suggested that consumption of ‘other beverages’ patterns might put youth at risk for

inadequate calcium, potassium or vitamin D intakes.

D. Recommendations for Future Research

Analyses of beverage frequency reported in the F F Q and 24-hour recall suggested
that youth should be educated to distinguish 100% fi'uit juice from fruit drinks that
contain added sugars. It may also be necessary to distinguish citrus juice (e. g., orange
juice) from non-citrus juice (e. g., apple juice, grape juice) when educating youth about
the nutritional benefits of drinking 100% fruit juice. It may be important to know that
orange juice is consumed more frequently than apple juice at the breakfast occasion, for
example, to recommend that youth drink juice with breakfast. Therefore, future research
recommendations are to separate the different types of juice in examining beverages
consumed at the various meal occasions.

Future research should also examine the amounts of beverages consumed at the
various meal occasions. In the current study, energy intakes from beverages in the total
diet were examined. However, it may be useful to know that greater amounts of
beverages are consumed at the snack occasion than at breakfast, for example. It’s
possible that drinking ‘milk & juice’ was inversely associated with overweight, because
youth in the ‘milk & juice’ group drank the recommended V2 cup serving (USDA, 2005)
of orange juice at breakfast. If large portions of apple juice are fiequently consumed as
snack/beverages throughout the day, consumption of excessive amounts of apple juice
might be associated with a risk for overweight (Dennison et al., 1997; Dennison et al.,

1999). Likewise, if large portions of sugar-sweetened beverages were consumed per

256

occasion, increased energy intakes from beverages in the ‘sugar—sweetened beverages’
group could explain why this pattern was positively associated with the BMI percentile.
Researchers should note the potential for bias in under-reporting intake in the 24-hour
recall could still lead to null results.

The type of milk (e.g., whole, 2%, 1%, skim) usually consumed by beverage
pattern groups was examined, and regression analyses adjusted for the type of milk when
examining the associations of beverage patterns with obesity outcomes. If analyses were
not adjusted for the type of milk, those beverage pattern groups that tended to consume
whole milk might be associated with obesity. If these unadjusted analyses were
replicated with the latest NHANES data, perhaps there would be significant differences
in the risk for overweight between ‘milk’ and ‘sugar—sweetened beverages’ patterns,
because the milk fat content is lower today than it was in NHANES IH, 1988-94. In
addition, replication of analyses of the association of beverages with away-fiom-home
meal locations (e.g., school, fast food restaurants) might show that the school’s share of
the sugar-sweetened beverage consumption would be greater in NHANES, 1999-2004
than in NHANES III, 1988-94. These changes in the beverage market may affect the
percentage of youth who are distributed among beverage pattern groups. Therefore, it is
important to replicate this research using the most recent survey data.

The most recent NHANES data might have improved the assessment of physical
activity by obtaining a complete inventory of activities, instead of including only the two
items to assess exercise-induced sweating and sports team participation in youth. The
assessment of ‘screen’ time might also be improved by including the use of video games

and the Computer in addition to television-viewing time. The dietary assessment

257

methodology has also been improved in the most recent NHANES. If the second day of
intake that was collected from respondents were available in the public-use dataset, then
the ‘usual’ intake of the group could be assessed using the method developed at Iowa
State University to adjust the intake distribution for intra-individual variation (Carriquiry,
1999; Caniquiry, 2003). It is necessary to estimate the ‘usual’ intake of the beverage
pattern groups to assess the prevalence of inadequate nutrient intake using the Estimated
Average Requirement and DRI-based methods (Suitor & Gleason, 2002). The NHANES,
2003-O4, dataset will also include responses to a Food Propensity Questionnaire that was
developed for use with two 24-hour recalls to improve the estimation of the ‘usual’ intake
of food (Freedman et al., 2004; Subar et al., 2004; Tooze et al., 2005). Because the
second day of intake was available for only 10% of the sample, it was not possible to
estimate ‘usual’ intake in the current study. Therefore, it will be important to replicate
this research using a dataset that has two days of 24-hour recall days for all respondents.
In the current study, the ‘milk & juice’ pattern was positively associated with
physical activity and inversely associated with TV watching, while the ‘sugar-sweetened
beverages’ pattern was positively associated with TV watching. In addition, the ‘milk &
juice’ pattern was effectively a marker for breakfast consumption, and other researchers
have shown that both breakfast skipping and TV watching are inversely associated with
income (Miech et al., 2006). Therefore, further examination of TV watching and income
is warranted in reanalyses of these data. The analyses of the current study included a test
for an interaction between gender and beverage patterns, and because the interaction term
was not significant, analyses were not stratified by gender. Future reanalyses of these

data shOuld test for an interaction and possibly stratify by levels of TV watching or by

258

income. Because income and the household head’s SES were significant in this study, it
might be useful to separate the youth’s race-ethnicity from other individual
characteristics like age and gender when potential covariates are reexamined. Because
there were four separate patterns that included sugar-sweetened beverages and shared
similar risk factors, it might be necessary to collapse them to identify a potential
association between sugar-sweetened beverage consumption and obesity in youth.

It was not clear from this study whether the low prevalence of overweight/risk for
overweight in those who ate breakfast every day was due to beverage patterns (e.g., milk
and juice consumption), other dietary patterns (e.g., food patterns or meal patterns),
nutrient intakes (e.g., calcium intakes), sociodemographic/lifestyle factors (e.g., income
and education of parents, exercise habits, sports team participation, TV watching), or
more fundamental causes (e. g., neighborhood effects associated with SES). Therefore,
firrther research in this area is recommended to clarify the inter-relationships between
beverage patterns and other risk factors for overweight.

That ‘milk & juice’ was associated with breakfast and ‘sugar-sweetened
beverages’ was associated with watching TV suggests beverage patterns, meal patterns,
diet patterns and physical activity patterns are all inter-related, and the context in which
beverages are consumed require further investigation. Because cross-sectional data
cannot be used to make causal inferences, a prospective study is recommended for further
research in this area. The FFQ is limited and could not be used to investigate meal
patterns, while the potential for bias due to under-reporting of intake in the 24-hour recall
remains. Therefore, continued improvement in dietary assessment methodology is

essential to progress in this area of study.

259

BIBLIOGRAPHY

Acheson KJ, Zahorska-Markiewicz B, Pittet P, Ananthararnan K, J équier E. Caffeine and
coffee: their influence on metabolic rate and substrate utilization in normal weight and
obese individuals. Am J Clin Nutr. 1980; 33: 989-97.

Ainsworth BE, Haskell WL, Leon AS, Jacobs DR Jr, Montoye HJ, Sallis JF,
Paffenbarger RS Jr. Compendium of physical activities: classification of energy costs of
human physical activities. Med Sci Sports Exerc. 25(1): 71-80. 1993.

Alaimo K, Olson CM, Frongillo EA, Jr. Low family income and food insufficiency in
relation to overweight in US children: is there a paradox? Arch Pediatr Adolesc Med.
2001;155:1161-7.

Alexy U, Sichert-Hellert W, Kersting M, Manz F, Schoch G. Fruit juice consumption and
the prevalence of obesity and short stature in German preschool children: results of the
DONALD Study. Dortmund Nutritional and Anthropometrical Longitudinally Designed.
J Pediatr Gastroenterol Nutr. 1999; 29: 343-9.

Almiron-Roi g E, Chen Y, Dreqnoski A. Liquid calories and the failure of satiety; how
good is the evidence. Obes Rev. 2003; 4: 201-212.

American Academy of Pediatrics. Policy Statement. Calcium requirements of infants,
children, and adolescents. Pediatrics. 1999; 104: 1152-7.

American Academy of Pediatrics. Committee on Nutrition. The use and misuse of fruit
juice in pediatrics. Pediatrics. 2001; 107: 1210-3.

American Diabetes Association (ADA). Type 2 diabetes in children and adolescents.
Pediatrics. 2000; 105: 671-680.

Anderson PM, Butcher KF, Levine PB. Maternal employment and overweight children. J
Health Econ. 2003; 22: 477-504.

Anonymous. Pulp Friction. Tufts University Diet & Nutrition Letter 1995; 13: 1-2.

Anonymous. Can energy drinks really boost my energy? Mayo Clin Womens
Healthsource. 2006; 10: 10.

Ashida H, Furuyashiki T, Nagayasu H, Bessho H, Sakakibara H, Hashimoto T,
Kanazawa K. Anti-obesity actions of green tea: possible involvements in modulation of

the glucose uptake system and suppression of the adipogenesis-related transcription
factors. Bio/actors. 2004; 22: 135-40.

260

 

Astrup A, Tuobro S, Cannon S, Hein P, Breum L, Madsen J. Caffeine: a double-blind,
placebo-controlled study of its thermogenic, metabolic, and cardiovascular effects in
healthy volunteers. Am J Clin Nutr. 1990; 51: 759-67.

Astrup A, Breum L, Tuobro S, Hein P, Quaade F. The effect and safety of an
ephedrine/caffeine compound compared to ephedrine, caffeine and placebo in obese
subjects on an energy restricted diet. A double blind trial. Int J Obes. 1992; 16: 269-77.

Astrup A, Buemann B, Flint A, Raben A. Low-fat diets and energy balance: how does the
evidence stand in 2002? Proc Nutr Soc. 2002; 61: 299-309.

Austin SB, Gortrnaker SL. Dieting and smoking initiation in early adolescent girls and
boys: a prospective study. Am J Public Health. 2001; 91: 446-50.

Avogaro A, Watanabe RM, Dall'Arche A, De Kreutzenberg SV, Tiengo A, Pacini G.
Acute alcohol consumption improves insulin action without affecting insulin secretion in
type 2 diabetic subjects. Diabetes Care 2004; 27: 1369-74.

Bachman CM, Baranowski T, Nicklas TA. Is there an association between sweetened
beverages and adiposity? Nutr Rev. 2006; 64: 153-74.

Ballew C, Kuester S, Gillespie C. Beverage choices affect adequacy of children's nutrient
intakes. Arch Pediatr Adolesc Med. 2000; 154: 1148-52.

Bandini LG, Schoeller DA, Cyr HN, Dietz WH. Validity of reported energy intake in
obese and nonobese adolescents. Am J Clin Nutr. 1990; 52: 421-5.

Baran D, Sorensen A, Grimes J, Lew R, Karellas A, Johnson B, Roche J. Dietary
modification with dairy products for preventing vertebral bone loss in premenopausal
women: a three-year prospective study. J Clin Endocrinol Metab. 1990; 70: 264-70.

Barba G, Troiano E, Russo P, Venezia A, Siani A. Inverse association between body
mass and frequency of milk consumption in children. Br J Nutr. 2005; 93: 15-19.

Barr SI, McCarron DA, Heaney RP, Dawson-Hughes B, Berga SL, Stern, J S, Oparil S.
Effects of increased consumption of fluid milk on energy, nutrient intake, body weight,

and cardiovascular disease risk factors in healthy older adults. J Am Diet Assoc. 2000;
100: 810-817.

Barr SI. Increased dairy product or calcium intake: Is body weight or composition
affected in humans? J Nutr. 2003; 133: 2458-2488.

Basiotis PP, Welsh SO, Cronin FJ, Kelsay J L, Mertz W. Number of days of food intake

records needed to estimate individual and group nutrient intakes with defined confidence.
JNutr. 1987; 117: 1638-41.

261

 

Beaton GH, Milner J, Corey P, McGuire V, Cousins M, Stewart E, de Ramos M, Hewitt
D, Grambsch PV, Kassirn N, Little JA. Sources of variance in 240hour dietary recall

data: implications for nutrition study design and interpretation. Am J Clin Nutr. 1979; 32:
2456-59.

Berkey CS, Rockett HR, Gillrnan MW, Field AE, Colditz GA. Longitudinal study of
skipping breakfast and weight change in adolescents. Int J Obes Relat Metab Disord.
2003; 27: 1258-66.

Berkey CS, Rockett HR, Field AE, Gillrnan MW, Colditz GA. Sugar-added beverages
and adolescent weight gain. Obes Res. 2004; 12: 778-788.

Berkey CS, Rockett HRH, Willett WC, Colditz GA. Milk, dairy fat, dietary calcium, and
weight gain: A longitudinal study of adolescents. Arch Pediatr Adolesc Med. 2005; 159:
543-550.

Beulens J W, van Beers RM, Stolk RP, Schaafsma G, Hendriks HF. The effect of
moderate alcohol consumption on fat distribution and adipocytokines. Obes Res. 2006;
14: 60-6.

Bialostosky K, Briefel RR, Pennington J. Nutrition monitoring in the United States. In
Handbook of Nutrition, ed. CD Berdanier, pp. 407-33. Washington, DC: CRC Press;
2001.

Blackburn GL, Kanders BS, Lavin PT, Keller SD, Whatley J. The effect of aspartame as
part of a multidisciplinary weight-control program on short- and long-term control of
body weight. Am J Clin Nutr. 1997; 65: 409-18.

Block G, Dresser CM, Hartman AM, Carroll MD. Nutrient sources in the American diet:
quantitative data from the NHANES II survey. 1. Vitamins and minerals. Am J
Epidemiol. 1985; 122: 13-26.

Block G, Dresser CM, Hartman AM, Carroll MD. Nutrient sources in the American diet:
quantitative data from the NHANES 11 survey. 11. Macronutrients and fats. Am J
Epidemiol. 1985; 122: 27-40.

Bobak M, Skodova Z, Marmot M. Beer and obesity: a cross-sectional study. Eur J Clin
Nutr. 2003; 57: 1250-3.

Boileau RA, Lohman TG, Slaughter MH, Ball TE, Going SB, Hendrix MK. Hydration of
the fat-free body in children during maturation. Human Biology. 1984; 56: 651-6.

Bonjour JP, Carrie AL, Ferrari S, Clavlen H, Slosmarr D, Theintz G, Rizzoli R. Calciurn-

enriched foods and bone mass in prepubertal girls: a randomized, double-blind, placebo-
controlled trial. J Clin Invest. 1997; 99: 1287—1294.

262

 

Borchardt RT, Huber J A. Cathechol O-methyltransferase. Structure-activity relationships
for inhibition by flavonoids. J Med Chem. 1975; 18: 120-2.

Boumtje PI, Huang CL, Lee J -Y, Lin B-H. Dietary habits, demographics, and the
development of overweight and obesity among children in the United States. Food
Policy. 2005; 30: 115-128.

Bowen J, Noakes M, Clifton PM. A high dairy protein, high-calcium diet minimizes bone
turnover in overweight adults during weight loss. J Nutr. 2004; 134: 568-573.

Bowman SA. Beverage choices of young females: Changes and impact on nutrient
intakes. J Am Diet Assoc. 2002; 102: 1234-9.

Bowman SA, Gortrnaker SL, Ebbeling CB, Pereira MA, Ludwig DS. Effects of fast-food
consumption on energy intake and diet quality among children in a national household
survey. Pediatrics. 2004; 113 (1) Pt 1: 112-8.

Bracco D, Ferrarra JM, Arnaud MJ, J équier E, Schutz Y. Effects of caffeine on energy

metabolism, heart rate, and methylxanthine metabolism in lean and obese women. Am J
Physiol. 1995; 269(4) Pt 1: E671-8.

Bray GA, Nielsen SJ, Popkin BM. Consumption of high-fi'uctose corn syrup in
beverages may play a role in the epidemic of obesity. Am J Clin Nutr. 2004; 79: 537-43.

Bray GA, Champagne CM. Beyond energy balance: there is more to obesity than
kilocalories. J Am Diet Assoc. 2005; 105 (Suppl 1): $17-$23.

Brener ND, McManus T, Galuska DA, Lowry R, Wechsler H. Reliability and validity of
self-reported height and weight among high school students. J Adolesc Health. 2003; 32:
281-7.

Briefel RR, Flegal KM, Winn DM, Loria CM, Johnson CL, Sempos CT. Assessing the
nation's diet: limitations of the food frequency questionnaire. J Am Diet Assoc. 1992; 92:
959-62.

Briefel RR. Assessment of the US diet in national nutrition surveys: national
collaborative efforts and NHANES. Am J Clin Nutr. 1994; 59 Suppl: 164S-167S.

Briefel RR, Sempos C, McDowell M, Chien S, Alaimo K. Dietary methods research in
NHANES III: underreporting of energy intake. Am J Clin Nutr. 1997; 65 (Suppl): 1203-
9S.

Briefel RR. Nutrition monitoring in the United States. In Present Knowledge in
Nutrition, ed. BA Bowman, RM Russell, pp. 617-35. Washington, DC: ILSI Press, 8th
ed.; 2001.

263

 

Briefel RR, Johnson CL. Secular trends in dietary intake in the United States. Annu Rev
Nutr. 2004; 24: 401-31.

Buemann B, Toubro S, Astrup A. The effect of wine or beer versus a carbonated soft
drink, served at a meal, on ad libitum energy intake. Int J Obes Relat Metab Disord.
2002; 26: 1367-72.

Buggs C, Rosenfield RL. Polycystic ovary syndrome in adolescence. Endocrinol Metab
Clin North Am. 2005; 34: 677-705.

Burke V, Milligan RA, Thompson C, Taggart AC, Dunbar DL, Spencer MJ, Medland A,
Gracey MP, Vandongen R, Beilin J. A controlled trial of health promotion programs in

1 1-year-olds using physical activity "enrichment" for higher risk children. J Pediatr.
1998; 132: 840-8.

Cadogan J, Eastell R, Jones N, Barker ME. Milk intake and bone mineral acquisition in
adolescent girls: randomised, controlled intervention trial. BMJ. 1997; 315: 1255—1260.

Carriquiry AL. Assessing the prevalence of nutrient inadequacy. Public Health Nutr.
1999; 2: 23-33.

Carriquiry AL. Estimation of usual intake distributions of nutrients and foods. J Nutr.
2003; 133: 6018-6088.

Carruth BR, Skinner JD. The role of dietary calcium and other nutrients in moderating
body fat in preschool children. Int J Obes. 2001; 25: 559-566.

Cavadini C, Siega-Riz AM, Popkin BM. US adolescent food intake trends from 1965 to
1996. Arch Dis Child. 2000a; 83: 18-24.

Cavadini C, Decarli B, Grin J, Narring F, Michaud PA. Food habits and sport activity
during adolescence: differences between athletic and non-athletic teenagers in
Switzerland. Eur J Clin Nutr. 2000b; 54 Suppl 1: 816-20.

Centers for Disease Control and Prevention (CDC). Health, United States, 2003 with
Chartbook on Trends in the Health of Americans. Atlanta, Georgia: U.S. Department of

Health and Human Services, Centers for Disease Control and Prevention; 2003a.
[Online]. Available: http://www.cdc.gov/nchs/data/hus/husO3.pdf

Centers for Disease Control and Prevention (CDC). Physical activity levels among
children aged 9-13 years - United States, 2002. MM WR 2003b; 52: 757-788.

Centers for Disease Control and Prevention (CDC). Youth Risk Behavior Surveillance —
United States, 2003. MM” 2004; 53 (No. SS-2): 1-96.

264

Chan GM, Hoffinan K, McMurry M. Effects of dairy products on bone and body
composition in pubertal girls. J Pediatr. 1995; 126: 551—556.

Chanmugam P, Guthrie JF, Cecilio S, Morton JF, Basiotis P, Anand R. Did fat intake in
the United States really decline between 1989-1991 and 1994-1996? J Am Diet Assoc.
2003; 103: 867-872.

Child Trends. 2002. Watching Television. [Online]. Available:
hgp://www.childtrendsdatabankorg/pdf/S5_PDF.pdf

Cho S, Dietrich M, Brown CJP, Clark CA, Block G. The effect of breakfast type on total
daily energy intake and body mass index: results from the Third National Health and
Nutrition Examination Survey (NHANES HI). J Amer Coll Nutr. 2003; 22: 296-302.

Clark N. Milk: destroying the myths. Physicians and Sports Med. 1990; 18: 133-5.

Cleghom DB, O’Loughlin PD, Schroeder BJ, Nordin BEC. An open, crossover trial of
calcium-fortified milk in prevention of early postmenopausal bone loss. Med J Aust.
2001; 175: 242—245.

Cole TJ, Bellizzi MC, Flegal KM, Dietz WH. Establishing a standard definition for child
overweight and obesity worldwide: international survey. BM]. 2000; 320: 1240-3.

Conway JM, Ingwersen LA, Vinyard BT, Moshfegh AJ. Effectivenes of the USDA’s
five-step multiple-pass method to assess food intake in obese and nonobese women. Am
J Clin Nutr. 2003; 77: 1171-8.

Conway JM, Ingwersen LA, Moshfegh AJ. Accuracy of dietary recall using the USDA
five-step multiple-pass method in men: an observational validation study. J Am Diet
Assoc. 2004; 104: 595-603.

Cook AJ, Friday J E. Pyramid Servings Database for USDA Survey Food Codes Version
2. 0. [Online]. Beltsville, MD: USDA, Agricultural Research Service, Community
Nutrition Research Group; 2004. Available at http://www.barc.udeng/bhnrc/cnrg

Cook S, Weitzman M, Auinger P, Nguyen M, Dietz WH. Prevalence of a metabolic
syndrome phenotype in adolescents: Findings from the Third National Health and
Nutrition Examination Survey, 1988-1994. Arch Pediatr Adolesc Med. 2003; 157: 821-
827.

Coon KA, Goldberg J, Rogers BL, Tucker KL. Relationships between use of television
during meals and children’s food consumption patterns. Pediatrics. 2001; 107: E7.

Coon KA, Tucker KL. Television and children’s consumption patterns: A review of the
literature. Minerva Pediatr. 2002; 54: 423-36.

265

Cordain L, Bryan ED, Melby CL, Smith MJ. Influence of moderate daily wine
consumption on body weight regulation and metabolism in healthy free-living males. J
Am Coll Nutr. 1997; 16: 134-9.

Cordain L, Melby CL, Hamamoto AE, O’Neill DS, Comier MA, Barakat HA, Israel RG,
Hill JO. Influence of moderate chronic wine consumption on insulin sensitivity and other
correlates of syndrome X in moderately obese women. Metabolism. 2000; 49: 1473-8.

Crepinsek MK, Burstein NR. Maternal Employment and Children ’s Nutrition. Volume
11, Other Nutrition-Related Outcomes. E—FAN-O4-006-l. Washington, DC: Economic
Research Service, USDA; 2004.

Crespo CJ, Smit E, Troiano RP, Bartlett SJ, Macera CA, Andersen RE. Television
watching, energy intake, and obesity in US children: results item the third National
Health and Nutrition Examination Survey, 1988-1994. Arch Pediatr Adolesc Med. 2001;
1 5 5: 3 60-5.

Dallongeville J, Marécaux N, Ducimetiere P, F errieres J, Arveiler D, Bingham A,
Ruidavets JB, Simon C, Amouyel P. Influence of alcohol consumption and various

beverages on waist girth and waist-to-hip ratio in a sample of French men and women. Int
J Obes Relat Metab Disord. 1998; 22: 1178-83.

Daniels SR, Arnett DK, Eckel RH, Gidding SS, Hayrnan LL, Kumanyika S, Robinson
TN, Scott BJ, St Jeor S, Williams CL. Overweight in children and adolescents:
pathophysiology, consequences, prevention, and treatment. Circulation. 2005; 111: 1999-
2012.

Davies KM, Heaney RP, Recker RR, Lappe JM, Barger-Lux MJ, Rafferty K, Hinders S.
Calcium intake and body weight. J Clin Endocrinol Metab. 2000; 85: 4635-4638.

Dawson-Hughes B, Harris S, Krall EA, Dallal GE. Effect of calcium and vitamin D
supplementation on bone density in men and women 65 years of age or older. N Engl J
Med. 1997; 337: 670—676.

DeCastro JM. The effects of the spontaneous ingestion of particular foods or beverages,
on the meal pattern and overall nutrient intake of humans. Phys Behav. 1993; 53: 1133-
1144.

Deckelbaum RJ, Williams CL. Childhood obesity: The health issue. Obes Res. 2001; 9
(S4): 2398-2438.

Delnevo CD, Hrywna M, Abatemarco DJ, Lewis MJ. Relationships between cigarette
smoking and weight control in young women. Fam Community Health. 2003; 26: 140-6.

Dennison BA. Fruit juice consumption by infants and children: a review. J Am Coll Nutr.
1996; 15(5) Suppl: 4S-118.

266

 

Dennison BA, Rockwell HL, Baker SL. Excess fruit juice consumption by preschool-
aged children is associated with short stature and obesity. Pediatrics. 1997; 99: 15-22.

Dennison BA, Rockwell HL, Nichols MJ, Jenkins P. Children's growth parameters vary
by type of fruit juice consumed. J Am Coll Nutr. 1999; 18: 346-52.

Devine A, Prince RL, Bell R. Nutritional effect of calcium supplementation by skim milk
powder or calcium tablets on total nutrient intake in postmenopausal women. Am J Clin
Nutr. 1996; 64: 731—737.

U.S. Department of Health and Human Services (DHHS). National Center for Health
Statistics. NHANES III Physician Examiner 's Training Manual. Rockville, MD: Westat,
Inc.; 1991.

U.S. Department of Health and Human Services (DHHS). National Center for Health
Statistics. Vital and Health Statistics. Plan and operation of the Third National Health
and Nutrition Examination Survey, I988-94. Hyattsville, MD: Public Health Service;
1994. DHHS Publication No. (PHS) 94-1308, Series 1, No. 32.

U.S. Department of Health and Human Services (DHHS). National Center for Health
Statistics. NHANES III Dietary Interviewer ’5 Training Manual, Revised. Rockville, MD:
Westat, Inc.; 1996a.

U.S. Department of Health and Human Services (DHHS). Centers for Disease Control
and Prevention (CDC). Physical Activity and Health: A Report of the Surgeon General.
Atlanta, GA: USDHHS, CDC, NCCDPHP; 1996b.

U.S. Department of Health and Human Services (DHHS). National Center for Health
Statistics. NHANES 111 Body Measurements (Antropometry) Reference Manual.
Rockville, MD: Westat, Inc.; 1998.

Dhurandhar NV. Fruit juice consumption. Pediatrics. 1997; 100: 728-30.

Dibba B, Prentice A, Ceesay M, Stirling DM, Cole TJ, Poskitt EME. Effect of calcium
supplementation on bone mineral accretion in Gambian children accustomed to a low-
calcium diet. Am J Clin Nutr. 2000; 71: 544—549.

Dicker D, Belnic Y, Goldsmith F, Green M, Nitzan-Kaluski D. On the relationship
between dietary calcium intake, body mass index and waist size. Presented at the 13th
European Congress on Obesity. Int J Obes. 2004; Supplement 1(28): SS9.

Diepvens K, Kovacs EM, Nijs IM, Vogels N, Westerterp-Plantenga MS. Effect of green

tea on resting energy expenditure and substrate oxidation during weight loss in
overweight females. Br J Nutr. 2005; 94: 1026-34.

267

 

Dietz WH. Critical periods in childhood for the development of obesity. Am J Clin Nutr.
1994; 99: 955-999.

Dietz WH. Childhood weight affects adult morbidity and mortality. J Nutr. 1998; 128
(28); 4118-4148.

Diez Roux AV. Persistent social patterning of cardiovascular risk: rethinking the familiar.
Circulation. 2005; 111: 3020-1.

DiMeglio DP, Mattes RD. Liquid versus solid carbohydrate: effects on food intake and
body weight. IntJ Obes. 2000: 24: 794-800.

Doucette RE, Dwyer J T. Is fruit juice a "no-no" in children's diets? Nutr Rev. 2000; 58:
1 80—3.

Drewnowski A, Specter 8E. Poverty and obesity: the role of energy density and energy
costs. Am J Clin Nutr. 2004; 79: 6-16.

Drewnowski A, Darmon N. The economics of obesity: dietary energy density and energy
cost. Am J Clin Nutr. 2005; 82(1) Suppl: 2658-2738.

Dulloo AG, Seydoux J, Girardier L. Potentiation of the thermogenic antiobesity effects
of ephedrine by dietary methylxanthines: adenosine antoagonism or phosphodiesterase
inhibition? Metabolism. 1992; 41: 1233-41.

Dulloo AG, Duret C, Rohrer D, Girardier L, mensi N, F athi M, Chjantre P,
Vandermander J. Efficacy of a green tea extract rich in catechin polyphenols and caffeine

in increasing 24-h energy expenditure and fat oxidation in humans. Am J Clin Nutr. 1999;
70: 1040-5.

Dulloo AG, Seydoux J, Girardier L, Chantre P, Vandermander J. Green tea and

thermogenesis: interactions between catechin-polyphenols, caffeine and sympathetic
activity. Int J Obes Relat Metab Disord. 2000; 24: 252-8.

Duncan GE, Li 8M, Zhou XH. Prevalence and trends of a metabolic syndrome phenotype
among us. Adolescents, 1999-2000. Diabetes Care. 2004;27:2438-43.

Dwyer J, Ellwood K, Moshfegh AH, Johnson CL. Integration of the Continuing Survey
of Food Intakes by Individuals and the National Health and Nutrition Examination
Survey. J Am Diet Assoc. 2001; 101: 1142-3.

Dwyer J, Picciano MF, Raiten DJ. Future directions for the integrated CSFH-NHANES:
What We Eat in America-NHANES. J Nutr. 2003a; 133: 5768-818.

Dwyer J, Picciano MF, Raiten DJ. Collection of food and dietary supplement intake data:
What We Eat in America-NHANES. J Nutr. 2003b; 133: 590s-600s.

268

 

Ebbeling CB, Feldman HA, Osgania SK, Chomitz VR, Ellenbogen SJ, Ludwig DS.
Effects of decreasing sugar-sweetened beverage consumption on body weight in
adolescents: a randomized controlled pilot study. Pediatrics. 2006; 117: 673-80.

Elders PJ M, Lips P, Netelenbos J C, van Ginkel FC, Khoe E, van der Vijgh WJ, van der
Stelt PF. Long-terrn effect of calcium supplementation on bone loss in perimenopausal
women. J Bone Miner Res. 1994; 9: 963—970.

Elkins WL, Cohen DA, Koralewicz LM, Taylor 8N. After school activities, overweight,
and obesity among inner city youth. J Adolesc. 2004; 27: 181-9.

Elliott SS, Keim NL, Stern J 8, Teff K, Havel PJ. Fructose, weight gain, and the insulin
resistance syndrome. Am J Clin Nutr. 2002;76:911-22

Enns CW, Mickle SJ, Goldman JD. Trends in food and nutrient intakes by children in
the United States. Fam Econ Nutr Rev. 2002; 14: 56-68.

Enns CW, Mickle SJ, Goldman JD. Trends in food and nutrient intakes by adolescents in
the United States. Fam Econ Nutr Rev. 2003; 15: 15-27.

Fagot-Campagna A, Saadine JB, Flegal KM, Beckles GL. Diabetes, impaired fasting
glucose, and elevated HbAlc in U.S. adolescents: The Third National Health and
Nutrition Examination Survey. Diabetes Care. 2001; 24: 834-837.

F ellers PJ, Nikdel 8, Lee HS. Nutrient content and nutrition labeling of several processed
Florida citrus juice products. J Am Diet Assoc. 1990; 90: 1079-84.

Finkelstein EA, Ruhm CJ, Kosa KM. Economic causes and consequences of obesity.
Annu Rev Public Health. 2005; 26: 239-57.

F iore H, Travis S, Whalen A, Auinger P, Ryan S. Potentially protective factors associated
with healthful body mass index in adolescents with obese and nonobese parents: a
secondary data analysis of the third national health and nutrition examination survey,
1988-1994. J Am Diet Assoc. 2006; 106: 55-64.

Flegal KM. Defining obesity in children and adolescents: epidemiologic approaches. Crit
Rev Food Sci Nutr. 1993; 33: 307-12.

Flegal KM, Troiano RP. Changes in the distribution of body mass index of adults and
children in the US population. Int J Obes Relat Metab Disord. 2000; 24: 807-818.

Food and Drug Administration (FDA). 2002. Code of federal regulations: 21 CFR

101.12 a Reference amounts customarily consumed per eating occasion. [Online]
Available at http://www.cfsan.fda.gov/~lrd/CF101-12.HTML

269

 

Food Marketing Institute. Trends in the United States: Consumer Attitudes & the
Supermarket, 1996. Washington, DC: Food Marketing Institute; 1996.

Food Marketing Institute. Trends in the United States: Consumer Attitudes & the
Supermarket, 2003. Washington, DC: Food Marketing Institute; 2003.

Forbes GB. Lean body mass-body fat interrelationships in humans. Nutr Rev. 1987; 45:
225-31.

Forshee RA, Storey ML. Total beverage consumption and beverage choices among
children and adolescents. Int J Food Science Nutr. 2003; 54: 297-307.

F orshee RA, Anderson PA, Storey ML. The role of beverage consjmption, physical
activity, sedentary behavior, and demographics on body mass index of adolescents. Int J
Food Sci Nutr. 2004; 44: 463-78.

Fox MK, Pac S, Devaney B, J ankowski L. Feeding Infants and Toddlers Study: What
foods are infants and toddlers eating? J Am Diet Assoc. 2004: 104 (Suppl 1): 822-830.

Frary CD, Johnson RK, Wang MQ. Children and adolescents' choices of foods and
beverages high in added sugars are associated with intakes of key nutrients and food
groups. J Adolesc Health. 2004; 34: 56—63.

Freedman D8, Dietz WH, Srinivasan SR, Berenson G8. The relation of overweight to

cardiovascular risk factors among children and adolescents: The Bogalusa Heart Study.
Pediatrics. 1999; 103: 1175-1182.

Freedman LS, Midthune D, Carroll RJ, Krebs-Smith S, Subar AF, Troiano RP, Dodd K,
Schatzkin A, Ferrari P, Kipnis V. Adjustments to improve the estimation of usual dietary
intake distributions in the population. J Nutr. 2004; 134: 1836-43.

Freeman R, Sheiham A. Understanding decision-making processes for sugar
consumption in adolescence. Community Dent Oral Epidemiol. 1997; 25: 228-32.

French 8A, Perry CL, Leon GR, Fulkerson J A. Weight concerns, dieting behavior, and
smoking initiation among adolescents: a prospective study. Am J Public Health. 1994;
84: 1 8 1 8-20.

French SA, Perry CL, Leon GR, Fulkerson JA. Changes in psychological variables and
health behaviors by dieting status over a three-year period in a cohort of adolescent
females. J Adolesc Health. 1995; 16: 438-47.

French SA, Story M, Jeffery RW. Environmental influences on eating and physical
activity. Ann Rev Public Health. 2001a; 22: 309-335.

270

French SA, Story M, Neumark-Sztainer D, Fulkerson J A, Harman P. Fast food restaurant
use among adolescents: associations with nutrient intake, food choices and behavioral
and psychosocial variables. Int J Obes Relat Metab Disord. 2001b; 25: 1823-33.

French SA, Lin BH, Guthrie JF. National trends in soft drink consumption among
children and adolescents age 6 to 17 years: Prevalence, amounts, and sources, 1977/78 to
1994/1998. J Am Diet Assoc. 2003; 103: 1326-1331.

Gabhainn SN, Nolan G, Kelleher C, Friel S. Dieting patterns and related lifestyles of
school-aged children in the Republic of Ireland. Public Health Nutr. 2002;5: 457-62.

Giammattei J, Blix G, Marshak HH, Wollitzer AO, Pettit DJ. Television watching and
soft drink consumption associations with obesity in 11- to 13-year-old school children.
Arch Pediatr Adol Med. 2003; 157: 882-6.

Gleason P, Suitor C. 2001. Children ’s Diets in the Mid-1990 ’s: Dietary Intake and its
Relationship with School Meal Participation. Alexandria, VA: U.S. Department of
Agriculture. Report No. CN-01-CD1.

Godin G, Anderson D, Lambert LD, Deshamais R. Identifying factors associated with
regular physical activity in leisure time among Canadian adolescents. Am J Health
Promot. 2005; 20: 20-7.

Gold DR, Darnokosh AI, Dockery DW, Berkey C8. Body-mass index as a predictor of
incident asthma in a prospective cohort of children. Pediatr Pulmonol. 2003; 36: 514-21.

Goodman E. The role of socioeconomic status gradients in explaining differences in US
adolescents health. Am J Public Health. 1999; 89: 1522-1528.

Goodman E, Slap GB, Huang B. The public health impact of socioeconomic status on
adolescent depression and obesity. Am J Public Health. 2003; 93: 1844-1850.

Goran MI, Ball GD, Cruz ML. Obesity and risk of type 2 diabetes and cardiovascular
disease in children and adolescents. J Clin Endocrinol Metab. 2003; 88: 1417-1427.

Gordon-Larsen P, Adair LS, Popkin BM. The relationship of ethnicity, socioeconomic
factors, and overweight in US adolescents. Obes Res. 2003; 11: 121-129.

Greenberg J A, Axen KV, Schnoll R, Boozer CN. Coffee, tea and diabetes: the role of
weight loss and caffeine. Int J Obes. 2005; 29: 1121-9.

Gruner C, Alig F, Muntwyler J. Validity of self-reported exercise-induced sweating as a
measure of physical activity among patients with coronary artery disease. Swiss Med
Wkly. 2002; 132: 629-32.

271

Guenther PM. Beverages in the diets of American teenagers. J Am Diet Assoc. 1986; 86:
493-9.

Guenther PM, DeMaio TJ, Ingwersen LA, Berline M. The multiple-pass approach for
the 24-hour recall in the Continuing Survey of Food Intakes by Individuals (CSFII) 1994-
1996. Presented at the International Conference on Dietary Assessment Methods,
January 1995; Boston, Mass.

Guenther PM, Kott PS, Carriquiry AL. Development of an approach for estimating usual
nutrient intake distributions at the population level. J Nutr. 1997; 127: 1106-12.

Gunther CW, Legowski PA, Lyle RM, McCabe GP, Eagan MS, Peacock M, Teegarden
D. Dairy products do not lead to alterations in body weight or fat mass in young women
in a l-y intervention. Am J Clin Nutr. 2005; 81: 751-756.

Guo SS, Wu W, Chumlea WC, Roche AF. Predicting overweight and obesity in
adulthood from body mass index values in childhood and adolescence. Am J Clin Nutr.
2002; 76: 653-658.

Guthrie JF, Morton J F. Food sources of added sweeteners in the diets of Americans. J Am
Diet Assoc. 2000; 100: 43-48, 51.

Halkjaer J, Sorensen TI, Tjonneland A, Togo P, Holst C, Heitmarm BL. Food and
drinking patterns as predictors of 6-year BMI-adjusted changes in waist circumference.
BrJ Nutr. 2004; 92: 735-48.

Harnpl J S, Heaton CLB, Taylor CA. Snacking patterns influence energy and nutrient
intakes but not body mass index. J Hum Nutr Diet. 2003; 16: 3-11.

Han LK, Takaku T, Li J, Kimura Y, Okuda H. Anti-obesity action of oolong tea. Int J
Obes Relat Metab Disord. 1999; 23: 98-105.

Hansen K, Remington PL. The effect of alcohol consumption on body weight among
Wisconsin adults. Wis Med J. 1992; 91: 348-50.

Hamack L, Stang J, Story M. Soft drink consumption among US children and
adolescents. Nutritional consequences. J Am Diet Assoc. 1999; 99: 436-41.

Hamack LJ, Jeffery RW, Boutelle KN. Temporal trends in energy intake in the United
States: an ecologic perspective. Am J Clin Nutr. 2000; 71: 1478-84.

Harvey-Berino J, Gold BC, Lauber R, Starinski A. The irnpact of calcium and dairy
product consumption on weight loss. Obes Res. 2005; 13: 1720-1726.

Heaney RP, Davies KM, Barger-Lux MJ. Calcium and weight: clinical studies. J Am Coll
Nutr. 2002; 21: 1528-1558.

272

Hedley AA, Ogden CL, Johnson CL, Carroll MD, Curtin LR, F legal KM. Prevalence of
overweight and obesity among US children, adolescents, and adults, 1999-2002. JAMA.
2004; 291: 2847-50.

Hellerstedt WL, Jeffery RW, Murray DM. The association between alcohol intake and
adiposity in the general population. Am J Epidemiol. 1990; 132: 594-611.

Huang TTK, McCrory MA. Dairy intake, obesity, and metabolic health in children and
adolescents: knowledge and gaps. Nutr Rev. 2005; 63: 71-80.

Institute of Medicine of the National Academies of Science (IOM). Dietary Reference
Intakes for Calcium, Phosphorus, Magnesium, Vitamin D, and F louride. Washington,
DC: National Academies Press; 1997. [Online]. Available: http://print.nap.edu

Institute of Medicine of the National Academies of Science (IOM). Dietary Reference
Intakes for Thiamin, Riboflavin, Niacin, Vitamin B6, Folate, Vitamin B12, Pantothenic
Acid, Biotin, and Choline. Washington, DC: National Academies Press; 1998. [Online].
Available: http://print.nag).edu

Institute of Medicine of the National Academies of Science (IOM). Dietary Reference
Intakes for Vitamin C, Vitamin E, Selenium, and Carotenoids. Washington, DC: National
Academies Press; 2000. [Online]. Available: http://print.nap.edu

Institute of Medicine of the National Academies of Science (IOM). Dietary Reference
Intakes for Vitamin A, Vitamin K, Arsenic, Boron, Chromium, Copper, Iodine, Iron,
Manganese, Molybdenum, Nickel, Silicon, Vanadium, and Zinc. Washington, DC:
National Academies Press; 2001. [Online]. Available: http://print.nap.edu

 

Institute of Medicine of the National Academies of Science (IOM). Dietary Reference
Intakes for Energy, Carbohydrates, Fiber, Fat, Protein and Amino Acids
(Macronutrients). Washington, DC: National Academies Press; 2002. [Online].
Available: http://print.n_ap.edu

J acobsen BK, Thelle D8. The Tromso Heart Study: the relationship between food habits
and the body mass index. J Chronic Dis. 1987; 40: 795-800.

Jacqmain M, Doucet E, Despres JP, Bouchard C, Tremblay A. Calcium intake, body
composition, and lipoprotein-lipid concentrations in adults. Am J Clin Nutr. 2003; 77 :
1448-1452.

J ahns L, Siega-Riz AM, Popkin BM. The increasing prevalence of snacking among US
children from 1977 to 1996. J Pediatr. 2001; 138: 493-498.

273

 

James J, Thomas P, Cavan D, Kerr D. Preventing childhood obesity by reducing
consumption of carbonated drinks: cluster randomized controlled trial. BMJ. 2004; 328:
1237.

J anz KF, Nielsen DH, Cassady 8L, Cook J 8, Wu Y-T, Hansen JR. Cross-validation of
the Slaughter skinfold equations for children and adolescents. Med Sci Sport Exerc.
1993; 25: 1070-1076.

Jensen LB, Kollerup G, Quaade F, Sorensen OH. Bone mineral changes in obese women
during a moderate weight loss with and without calcium supplementation. J Bone Miner
Res. 2001; 16: 141-147.

Johnson RK, Driscoll P, Goran MI. Comparison of multiple-pass 24-hour recall
estimates of energy intake with total energy expenditure determined by the doubly
labeled water method in young children J Am Diet Assoc. 1996; 96: 1140-44.

Johnson RK. What are people really eating and why does it matter? Nutr Today. 2000;
35: 40-46.

Johnson RK, Frary C. Choose beverages and foods to moderate your intake of sugars:
the 2000 Dietary Guidelines for Americans —- what’s all the firss about? J Nutr. 2001;
131, 27668-27718.

Johnston CC, Miller JZ, Slemenda C8, Reister TK, Hui 8, Christian JC, Peacock M.
Calcium supplementation and increases in bone mineral density in children. N Engl J
Med. 1992; 327: 82—87.

Johnston CS. Uncle Sam's diet sensation: MyPyramid--an overview and commentary.
MedGenMed 2005; 7: 78.

Kabmova K, Braunerova R, Aldhoon B, Hlavaty P, Wagenkrrecht M, Kunesova M,
Parizkova J, Hainer B. Association of changes in macronutrient and calcium intakes with

body weight change in obese subjects. Presented at the 13th European Congress on
Obesity. Int J Obes. 2004; Supplement 1(28): 8138.

Kalkwarf HJ, Specker BL, Dianchi DC, Ranz J, Ho M. The effect of calcium
supplementation on bone density during lactation and after weaning. N Engl J Med. 1997;
337 : 523—529.

Kant AK. Dietary patterns and health outcomes. J Am Diet Assoc. 2004; 104: 615-35.

Kennedy E, Riddick H, Kramer-LeBlanc C, Davis C. Fruit juice consumption. Pediatrics.
1997;100:728; author reply 729-30.

274

Keski-Rahkonen A, Kaprio J, Rissanen A, Virkkunen M, Rose RJ. Breakfast skipping
and health-compromising behaviors in adolescents and adults. Eur J Clin Nutr. 2003; 57:
842-53.

Kimm SY, Glynn NW, Kriska AM, Barton BA, Kronsberg SS, Daniels SR, Crawford
PB, Sabry 21, Liu K. Decline in physical activity in black girls and white girls during
adolescence. N Engl J Med. 2002; 347: 709-15.

Kloeblen-Tarver AS. Fruit juice consumption not related to grth among preschool-
aged children enrolled in the WIC program. J Am Diet Assoc. 2001; 101: 996.

Kohl H, Blair 8, Paffenbarger R, Macera C, Kronenfeld J. A mail survey of physical
activity habits as related to measured physical fitness. Am J Epidemiol. 1988; 127: 1228-
39.

Koplan JP, Liverrnan CT, Kraak VI, eds. 2005. Preventing Childhood Obesity. Health in
the Balance. Committee on Prevention of Obesity in Children and Youth, Food and
Nutrition Board, Board on Health Promotion and Disease Prevention, Institute of
Medicine of the National Academies. Washington, DC: National Academies Press.
http://www.nap.edu

Kovacs EM, Lejeune MP, Nijs I, Westerterp-Plantenga M8. Effects of green tea on
weight maintenance after body-weight loss. Br J Nutr. 2004; 91 : 431-7.

Krebs-Smith SM, Subar AF, Guenther P, Tooze JA, Midthune D, Kipnis V, Dodd K.
Food propensity questionnaires (FPQs) in the National Health and Nutrition Examination
Survey (NHANES): A step in the evolution of usual dietary intake estimation. FASEB J.
2004; 18 (5): A480.

Kristal AR, Peters U, Potter JD. Is it time to abandon the food frequency questionnaire?
Cancer Epidemiol Biomarkers Prev. 2005; 14: 2826-8.

Krowchuk DP, Kreiter SR, Woods CR, Sinal 8H, DuRant RH. Problem dieting behaviors
among young adolescents. Arch Pediatr Adolesc Med. 1998; 152: 884-8.

Kubik MY, Lytle LA, Story M. Schoolwide food practices are associated with body mass
index in middle school students. Arch Pediatr Adolesc Med. 2005;159:1 l 1 1-1 1 l4.

Kuczmarski RJ, Ogden CL, Grummer—Stawn LM, Flegal Km, Guo SS, Wei R, Mei Z,
Curtin LR, Roche AF, Johnson CL. CDC grth charts: United States. Adv Data. 2000; 8
(314); 1-27.

Kuczmarski MF, Kuczmarski RJ, Naj jar M. Effects of age on validity of self-reported

height, weight, and body mass index: findings fi'om the Third National Health and
Nutrition Examination Survey, 1988-1994. J Am Diet Assoc. 2001; 101: 28-34.

275

Lappe JM, Rafferty KA, Davies M, Lypaczewski G. Girls on a high-calcium diet gain

weight at the same rate as girls on a normal diet: a pilot study. J Am Diet Assoc. 2004;
104: 1361-1367.

Lau EMC, Woo J, Lam V, Hong A. Milk supplementation of the diet by postmenopausal

Chinese women on a low calcium intake retards bone loss. J Bone Miner Res. 2001; 16:
1704-1709.

Lawlor DA, Chaturvedi N. Treatment and prevention of obesity--are there critical periods
for intervention? Int J Epidemiol. 2006; 35: 3-9.

Lee WTK, Leung 88F, Wang 8H, Xu YC, Zeng WP, Lau J, Oppenheimer SJ, Cheng
J CY. Double-blind, controlled calcium supplementation and bone mineral accretion in
children accustomed to a low-calcium diet. Am J Clin Nutr. 1994; 60: 744—750.

Lee WTK, Leung 88F, Leung DMY, Tsand HSY, Lau J, Cheng J CY. A randomized,
double-blind controlled calcium supplementation trial, and bone and height acquisition in
children. Br J Nutr. 1995; 74: 125—139.

Lelovics Z,Tarnavolgyi G. Relation between calcium intake and obesity. Presented at the
13th European Congress on Obesity. Int J Obes. 2004; Supplement 1(28): 8169.

Levin BE. The obesity epidemic: Metabolic imprinting on genetically susceptible neural
circuits. Obes Res. 2000; 8: 342-347.

Levin S, Lowry R, Brown DR, Dietz WH. Physical activity and body mass index among
US adolescents: youth risk behavior survey, 1999. Arch Pediatr Adolesc Med. 2003; 157:
816-20.

Levine AA. Excessive fi'uit juice consumption: how can something that causes failure to
thrive be associated with obesity? J Pediatr Gastroenterol Nutr. 1997; 25: 554-5.

Lifshitz F, Ament ME, Kleinman RE, Klish W, Lebenthal E, Udall JN. Role of juice
carbohydrate malabsorption in chronic nonspecific diarrhea in children. J Pediatrics.
1992; 89: 825-9.

Lin BH, Guthrie J, Frazao E. Quality of children’s diets at and away from home; 1994-
96. Food Review. 1999; 2-10.

Lin YC, Lyle RM, McCabe LD, McCabe GP, Weaver CM, Teegarden D. Dairy calcium
is related to changes in body composition during a two-year exercise intervention in

young women. J Am Coll Nutr. 2000; 19: 754-760.

Link BG, Phelan J. Social conditions as fundamental causes of disease. J Health Soc
Behav. 1995; (extra issue):80-94.

276

Lloyd T, Andon MB, Rollings N, Martel JK, Landis JR, Demers LM, Eggli DF,
Kieselhorst K, Kulin HE. Calcium supplementation and bone mineral density in
adolescent girls. JAMA. 1993; 270: 841—844.

Lobstein T, Baur L, Uauy R, IASO International obesity Task Force. Obesity in children
and young people: A crisis in public health. Obes Res. 2004; (Suppl 1): 4-85.

Lohman TG. Applicability of body composition techniques and constants for children
and adults. Exerc Sport Sci Rev. 1986; 14: 325-57.

Loos RJF, Rankinen T, Leon AS, Skinner J S, Wihnore JH, Rao DC, Bouchard C.
Calcium intake is associated with adiposity in black and white men and white women of
the HERITAGE Family Study. J Nutr. 2004; 134: 1772-1778.

Lopez-Garcia E, van Dam RM, Rajpathak S, Willett WC, Manson J E, Hu FB. Changes in
caffeine intake and long-term weight change in men and women. Am J Clin Nutr. 2006;
83: 674-80.

Lovejoy J C, Champagne CM, Smith SR, de Jonge L, Xie H. Ethnic differences in dietary
intakes, physical activity, and energy expenditure in middle-aged, premenopausal
women: the Healthy Transitions Study. Am J Clin Nutr. 2001; 74: 90-95.

Lowry R, Galuska DA, Fulton J E, Wechsler H, Kann L. Weight management goals and
practices among U.S. high school students: associations with physical activity, diet, and
smoking. J Adolesc Health. 2002; 31: 133-44.

Ludwig D8, Peterson KE, Gortrnaker SL. Relation between consumption of sugar-
sweetened drinks and childhood obesity: a prospective, observational analysis. Lancet.
2001; 357: 505-8.

Lukasiewicz E, Mennen LI, Bertrais S, Amault N, Preziosi P, Galan P, Hercberg 8.
Alcohol intake in relation to body mass index and waist-to-hip ratio: the importance of
type of alcoholic beverage. Public Health Nutr. 2005; 8: 315-20.

Lyle J, Hoffman H. 1972. Children’s use of television and other media. In: Rubenstein
GA, Murray JP, eds. Television and Social Behavior. Vol. 4. Washington, DC: US
Government Printing Office. Pp. 129-256.

Lytle LA, Seifert S, Greenstein J, McGovern P. How do children's eating patterns and
food choices change over time? Results from a cohort study. Am J Health Promot. 2000;
14: 222-8.

Madans JH, Sondik EJ, Johnson CL. Future Directions for What We Eat in America-
NHANES: The Integrated CSFII-NHANES. Foreword. J Nutr 2003a; 133: 575s.

277

 

Malinauskas BM, Raedeke TD, Aeby VG, Smith J L, Dallas MB. Dieting practices,
weight perceptions, and body composition: a comparison of normal weight, overweight,
and obese college females. Nutr J. 2006; 5: ll.

Marques-Vidal P, Montaye M, Haas B, Bingham A, Evans A, Juhan-Vague I, Ferrieres J,
Luc G, Amouyel P, Arveiler D, Yamell J, Ruidevets JB, Scarabin PY, Ducirnetiere P.
Relationships between alcoholic beverages and cardiovascular risk factor levels in
middle-aged men, the PRIME Study. Prospective Epidemiological Study of Myocardial
Infarction Study. Atherosclerosis. 2001; 157: 431-40.

Matheson DM, Killen JD, Wang Y, Varady A, Robinson TN. Children’s food
consumption during television viewing. Am J Clin Nutr. 2004; 79: 1088-94.

Mattes RD. Dietary compensation by humans for supplemental energy provided as
ethanol or carbohydrate in fluids. Physiol Behav. 1996; 59: 179-87.

Merrilees MJ, Smart EJ, Gilchrist NL, Frampton C, Turner JG, Hooke E, March RL,
Maguire P. Effects of dairy food supplements on bone mineral density in teenage girls.
Eur J Nutr. 2000; 39: 256—262.

Miech RA, Kumanyika SK, Stettler N, Link BG, Phelan J C, Chang VW. Trends in the
association of poverty with overweight among US adolescents, 1971-2004. JAMA. 2006;
295: 2385-2393.

Miller GD, Jarvis JK, McBean LD. The importance of meeting calcium needs with foods.
J Am Coll Nutr. 2001; 20(2) Supplzl688-1858.

Mirrniran P, Esmaillzadeh A, Azizi F. Dairy consumption and body mass index: an
inverse relationship. Int J Obes. 2005; 29: 115-121.

Mokdad AH, Ford ES, Bowman BA, Dietz WH, Vinicor F, Bales V8, Marks J 8.
Prevalence of obesity, diabetes, and obesity-related health risk factors, 2001. J Am Med
Assoc. 2003; 289: 76-79.

Morton JF, Guthrie JF. Changes in children’s total fat intakes and their food group
sources of fat, 1989-91 versus 1994-95: Implications for diet quality. Family Econ Nutr
Review. 1998: 44-57. ‘

Moshfegh AJ, Mickle 8M, Borrud LG, Goldman JD, LaComb RP. Dietary intakes of
children in America: Beverage choices. FASEB J. 2000; 14 (4); A533.

Mrdj enovic G, Levitsky DA. Nutritional and energetic consequences of sweetened drink
consumption in 6- to 13-year-old children. J Pediatr. 2003; 142: 604-610.

278

Murase T, Nagasawa A, Suzuki J, Hase T, Tokimitsu I. Beneficial effects of tea catechins
on diet-induced obesity: stimulation of lipid catabolism in the liver. Int J Obes Relat
Metab Disord. 2002; 26: 1459-64.

Murase T, Hararrrizu S, Shirnotoyodome A, Tokimitsu 1. Reduction of diet-induced

obesity by a combination of tea-catechin intake and regular swimming. Int J Obes. 2006;
30: 561-8.

Murphy SP. Collection and analysis of intake data from the integrated survey. J Nutr.
2003;133:5855-98.

Must A, Strauss R8. Risks and consequences of childhood and adolescent obesity. Int J
Obes Relat Metab Disord. 1999; 23 (28); 82-811.

Must A, Anderson SE. Effects of obesity on morbidity in children and adolescents. Nutr
Clin Care. 2003; 6: 4-12.

Nagao T, Komine Y, Soga S, Meguro S, Hase T, Tanaka Y, Tokimitsu I. Ingestion of a
tea rich in catechins leads to a reduction in body fat and malondialdehyde-modified LDL
in men. Am J Clin Nutr. 2005; 81: 122-9.

National Center for Health Statistics (N CH8). A SAS Program for the CDC Growth
Charts. Available: http://www.cdc.gov/nccdphp/dnpa/growthcharts/sas.htrn

National Center for Health Statistics (N CH8). National Health and Nutrition
Examination Survey, III 1988-94. CD-ROM Series 11, No. 1. SETS Version 1.22a,
Revised October 1997. Hyattsville, Nfl): National Center for Health Statistics; 1997.

National Center for Health Statistics (NCHS). National Health and Nutrition
Examination Survey, III 1988-94. CD-ROM Series 11, No. 2A. ASCII Version 1.223,
Issued April 1998. Hyattsville, MD: National Center for Health Statistics; 1998.

National Heart, Lung & Blood Institute (NHLBI) Obesity Task Force. Clnical guidelines
on the identification, evaluation, and treatment of overweight and obesity in adults- the
evidence report. Obes Res. 1998; 6 (Suppl 2): 518-2098.

National Heart, Lung, and Blood Institute (NHLBI). Third Report of the Expert Panel on
Detection, Evaluation, and Treatment of High Blood Cholesterol in Adults (Adult
Treatment Panel III). Publication No. 02-5215. Rockville, MD: NIH; 2002.

National Research Council (NRC). Working Families and Growing Kids: Caring for
Children and Adolescents. Washington, DC: National Academies Press; 2003.

Nelson MC, Gordon-Larsen P, Adair LS, Popkin BM. Adolescent physical activity and

sedentary behavior: patterning and long-term maintenance. Am J Prev Med. 2005; 28:
259-66.

279

Newby PK, Peterson KE, Berkey CS, Leppert J, Willett WC, Colditz GA. Beverage
consumption is not associated with changes in weight and body mass index among low-
income preschool children in North Dakota. J Am Diet Assoc. 2004; 104: 1086-94.

Nicklas TA, Yang SJ, Baranowski T, Zakeri I, Berenson G. Eating patterns and obesity in
children. The Bogalusa Heart Study. Am J Prev Med. 2003; 25: 9-16.

Nielsen SJ, Siega-Riz AM, Popkin BM. Trends in energy intake in U.S. between 1977
and 1996: similar shifts seen across age groups. Obes Res. 2002; 10: 370-8.

Nielsen SJ, Popkin BM. Patterns and trends in food portion sizes, 1977-1998. JAMA.
2003; 289: 450-3.

Nielsen SJ, Popkin BM. Changes in beverage intake between 1977 and 2001. Am J Prev
Med. 2004; 27 : 205-10.

Nielsen Media Research. 2000 Report on Television: The First 50 Years. New York:
AC Nielsen Company; 2000.

Novotny R, Daida YG, Acharya S, Grove JS, Vogt TM. Dairy intake is associated with
lower body fat and soda intake with greater weight in adolescent girls. J Nutr. 2004; 134:
1 905-1 909.

Nowson CA, Green RM, Hopper J L, Sherwin AJ, Young D, Kaymakci B, Guest CS,
Smid M, Larkins RG, Wark JD. A co-twin study of the effect of calcium supplementation
on bone density during adolescence. Osteoporos Int. 1997; 7: 219—225.

Ochner CN, Lowe MR. Opposing effects of calcium and caloric intake on weight regain
after diet. Presented at the 13th European Congress on Obesity. Int J Obes. 2004;
Supplement 1(28): 8143.

Ogden CL, Flegal KM, Carroll MD, Johnson CL, Prevalence and trends in overweight
among US children and adolescents, 1999-2000. J Am Med Assoc. 2002; 288: 1728-
1732.

Ogden CL, Carroll MD, Curtin LR, McDowell MA, Tabak CJ, Flegal KM. Prevalence of
overweight and obesity in the United States, 1999-2004. JAMA. 2006; 295: 1549-55.

Oken E, Gillrnan MW. Fetal origins of obesity. Obes Res. 2003; 11: 496-506.
Ortega RM, Requejo AM, Lopez-Sobaler AM, Andrés P, Quintas ME, Navia B,

Izquierdo M, Rivas T. The importance of breakfast in meeting daily recommended
calcium intake in a group of schoolchildren. J Am Coll Nutr. 1998; 17: 19-24.

280

Paeratakul S, Ferdinand DP, Champagne CM, Ryan DH, Bray GA. Fast-food
consumption among US adults and children: dietary and nutrient intake profile. J Am Diet
Assoc. 2003; 103: 1332-8.

Parikh SJ, Yanovski J A. Calcium intake and adiposity. Am J Clin Nutr. 2003; 77 : 281-
287.

Parzen E. Stochastic processes. San Francisco: Holden-Day, 1962.

Pereira MA, Jacobs DR Jr, Van Horn L, Slattery ML, Kartashov AI, Ludwig DS. Dairy
consumption, obesity, and the insulin resistance syndrome in young adults. JAMA. 2002;
287: 2081-2089.

Perez-Jaraiz MD, Revilla M, Alvarez de los Heros J I, Villa LF, Rico H. Prophylaxis of
osteoporosis with calcium, estrogens and/or eelcatonin: comparative longitudinal study of
bone mass. Maturitas. 1996; 23: 327-332.

Phillips SM, Bandini LG, Cry H, Colclough-Douglas S, Naumova E, Must A. Dairy food
consumption and body weight and fatness studied longitudinally over the adolescent
period. Int J Obes. 2003; 27: 1 106-1113.

Phillips SM, Bandini LG, Naumova EN, Cyr H, Colclough S, Dietz WH, Must A.
Energy-dense snack food intake in adolescence: longitudinal relationship to weight and
fatness. Obes Res. 2004; 12: 461-72.

Pinhas-Hamiel O, Dolan LM, Daniels SR, Standiford D, Khoury PR, Zeitler P. Increased
incidence of non-insulin—dependent diabetes mellitus among adolescents. J Pediatr.
1996; 128: 608-615.

Popkin BM, Barclay DV, Nielsen SJ. Water and food consumption patterns of U.S. adults
from 1999 to 2001. Obes Res. 2005; 13: 2146-52.

Popkin BM, Armstrong LE, Bray GM, Caballero B, Frei B, Willett WC. A new
proposed guidance system for beverage consumption in the United States. Am J Clin
Nutr. 2006; 83: 529-42.

Poppitt SD, Eckhardt JW, McGonagle J, Murgatroyd PR, Prentice AM. Short-term
effects of alcohol consumption on appetite and energy intake. Physiol Behav. 1996; 60:
1063-70.

Prentice A, J arjou LM, Cole TJ, Stirling DM, Dibba B, F airweather-Tait 8. Calcium
requirements of lactating Gambian mothers: effects of a calcium supplement on breast-

milk calcium concentration, maternal bone mineral content, and urinary calcium
excretion. Am J Clin Nutr. 1995; 62: 58—67.

281

 

Putnam RD. Tuning in tuning out: The strange disappearance of social capital in
America. Political Science. 1995; 28: 664-683.

Raben A, Basalara TH, Moller AC, Astrup A. Sucrose compared with artificial
sweeteners: different effects on ad libitum food intake and body weight after 10 wk of
supplementation in overweight subjects. Am J Clin Nutr. 2002; 76: 721-729.

Rampersaud GC, Bailey LB, Kauwell GP. National survey beverage consumption data
for children and adolescents indicate the need to encourage a shift toward more nutritive
beverages. J Am Diet Assoc. 2003; 103: 97-100.

Rampersaud GC, Pereira MA, Girard BL, Adams J, Metzl JD. Breakfast habits,
nutritional status, body weight, and academic performance in children and adolescents. J
Am Diet Assoc. 2005; 105: 743-60.

Recker RR, Hinders 8, Davies KM, Heaney RP, Stegrnan MR. Lappe JM, Kimmel DB.
Correcting calcium nutritional deficiency prevents spine fractures in elderly women. J
Bone Miner Res. 1996; 11: 1961—1966.

Reid IR, Home A, Mason B, Ames R, Bava U, Gamble GD. Effects of calcium
supplementation on body weight and blood pressure in normal older women: a
randomized controlled trial. J Clin Endocrinol Metab. 2005; 90: 3824-3829.

Ricci TA, Chowdhury HA, Heymsfield SB, Stahl T, Pierson RN Jr, Shapses SA. Calcium
supplementation suppresses bone turnover during weight reduction in postmenopausal
women. J Bone Miner Res. 1998; 13: 1045—1050.

Rideout VJ, Vandewater EA, Wartella EA. Zero to Six: Electronic Media in the Lives of
Infants, Toddlers and Preschoolers. Menlo Park, CA: Henry J. Kaiser Family
Foundation; 2003.

Riggs BL, O’Fallon WM, Muhs J, O’Connor MK, Kumar R, Melton LJ 111. Long-term
effects of calcium supplementation on serum parathyroid level, bone turnover, and bone
loss in elderly women. J Bone Miner Res. 1998; 13: 168—174.

Ritchie LD, Welk G, Styne D, Gerstein DE, Crawford PB. Family environment and
pediatric overweight: what is a parent to do? J Am Diet Assoc. 2005; 105(5) Suppl 1:
870-9 ‘

Roberts D, Foehr U, Rideout V, Brodie M. Kids and Media at the New Millenium.
Menlo Park, CA: Henry J. Kaiser Family Foundation; 1999.

Robinson TN, Hammer LD, Killen JD, Kraemer HC, Wilson DM, Hayward C, Taylor

CB. Does television viewing increase obesity and reduce physical activity? Cross-
sectional and longitudinal analyses among adolescent girls. Pediatrics. 1993; 91: 273-80.

282

Robinson TN. Reducing children’s television viewing to prevent obesity: a randomized
controlled trial. JAMA. 1999; 282: 1561-7.

Roche AF, Siervogel RM, Chumlea WC, Webb P. Grading body fatness fi'om limited
anthropometric data. Am J Clin Nutr. 1981; 34: 2831-8.

Rolls BJ, Laster LJ, Surnmerfelt A. Hunger and food intake following consumption of
low-calorie foods. Appetite. 1989; 13: 115-27.

Rolls BJ. The supersizing of America. Portion size and the obesity epidemic. Nutrition
Today. 2003: 38: 42-53.

Sakurai Y, Umeda T, Shinchi K, Honjo S, Wakabayashi K, Todoroki I, Nishikawa H,
Ogawa S, Katsurada M. Relation of total and beverage-specific alcohol intake to body

mass index and waist-to-hip ratio: a study of self-defense officials in Japan. Eur J
Epidemiol. 1997; 13: 893-8.

Sallis J F, Glanz K. The role of built environments in physical activity, eating, and obesity
in childhood. Future Child. 2006; 16: 89-108.

Salvini 8, Hunter DJ, Sampson L, et a1. Food-based validation of a dietary questionnaire:
the effects of week-to-week variation in food consumption. Int J Epidemiol. 1989; 18:
85 8-67.

Sanchez-Villegas A, Bes-Rastrollo M, Martinez-Gonzalez MA, Serra-Majem L.
Adherence to a Mediterranean dietary pattern and weight gain in a follow-up study: the
SUN cohort. Int J Obes. 2006; 30: 350-8.

Schatzkin A, Kipnis V, Carroll RJ, et al. A comparison of a food frequency questionnaire
with a 24-hour recall for use in an epidemiological cohort study: results fi'om the
biomarker-based Observing Protein and Energy Nutrition (OPEN) study. Int J Epidemiol.
2003; 32: 1054-62.

Schmidt M, Affenito SG, Striegel-Moore R, Khoury PR, Barton B, Crawford P,
Kronsberg S, Schreiber G, Obarzanek E, Daniels 8. Fast-food intake and diet quality in
black and white girls: the National Heart, Lung, and Blood Institute Growth and Health
Study. Arch Pediatr Adolesc Med. 2005; 159: 626-31.

Schrb'der H, Marrugat J, Vila J, Covas MI, Elosua R. Adherence to the traditional
Mediterranean diet is inversely associated with body mass index and obesity in a Spanish
population. J Nutr. 2004; 134: 3355-61.

Sempos CT, Johnson NE, Smith EL, Gilligan C. Effects of intraindividual and
interindividual variation in repeated dietary records. Am J Epidemiol. 1985; 121: 120-30.

283

 

Shah BV, Barnwell BG, Bieler GS. SUDAAN User 's Manual, Release 7.50. Research
Triangle Park, NC: Research Triangle Institute; 1997.

Shapses SA, Heshka S, Heymsfield SB. Effect of calcium supplementation on weight and
fat loss in women. J Clin Endocrinol Metab. 2004; 89: 632-637.

Shields DH, Corrales KM, Metallinos-Katsaras E. Gourmet coffee beverage consumption
among college women. J Am Diet Assoc. 2004; 104: 650-3.

Sibbald B. Sugar industry sour on WHO report. CMAJ. 2003; 168: 1585.

Siconolfi S, Lasater 8, Snow R, Carleton R. Self-reported physical activity compared
with maximal oxygen uptake. Am J Epidemiol. 1985; 122: 101-05.

Siega-Riz AM, Popkin BM, Carson T. Trends in breakfast consumption for children in
the United States from 1965-1991. Am J Clin Nutr. 1998; 67: 7488-7568.

Siega-Riz AM, Popkin BM, Carson T. Differences in food patterns at breakfast by
sociodemographic characteristics among a nationally representative sample of adults in
the United States. Prev Med. 2000; 30: 415-24.

Silvennan BL, Cho NH, Rizzo TA, Metgert BE. Long-term effects of the intrauterine

environment: The Northwestern University Diabetes in Pregnancy Center. Diabetes
Care. 1998; 21 (82): B142-B149.

Skinner JD, Carruth BR, Moran J, 3rd, Houck K, Coletta F. Fruit juice intake is not
related to children's growth. Pediatrics. 1999; 103: 58-64.

Skinner JD, Bounds W, Carruth BR, Ziegler P. Longitudinal calcium intake is negatively
related to children’s body fat indexes. J Am Diet Assoc. 2003; 103; 1626-1631.

Slaughter NIH, Lohman TG, Boileau RA, et al. Influence of maturation on relationship of
skinfolds to body density: a cross-sectional study. Hum Biol. 1984;56: 681-9.

Slaughter MH, Lohman TG, Boileau RA, Horswill CA, Stillman RJ, Van Loan MD,
Bemben DA. Skinfold equations for estimation of body fatness in children and youth.
Hum Biol. 1988; 60: 709-23.

Sloan EA. What, when, and where Americans eat. Food Technology. 2003; 57: 48-66.
Smiciklas-Wright H, Mitchell D, Mickle S, Goldman J, Cook A. Foods commonly eaten
in the United States, 1989-1992 and 1994-1996: Are portion sizes changing? J Am Diet
Assoc. 2003; 103: 41-47.

Smith MM, Lifshitz F. Excess fi'uit juice consumption as a contributing factor in
nonorganic failure to thrive. Pediatrics. 1994; 93: 438-43.

284

Steegrnann AT, Jr. Climate, racial category, and body proportions in the U.S. Am J Hum
Biol. 2005; 17: 393-402.

Stinson W. Fruit juice consumption. Pediatrics. 1997; 100: 728; author reply 729-30.

St-Onge MP. Dietary fats, teas, dairy, and nuts: potential functional foods for weight
control? Am J Clin Nutr. 2005; 81: 7-15.

Storey ML, Forshee RA, Weaver AR, Sansalone WR. Demographic and lifestyle factors

associated with body mass index among children and adolescents. Int J Food Sci Nutr.
2003; 54: 491-503.

Storey ML, Forshee RA, Anderson PA. Associations of adequate intake of calcium with
diet, beverage consumption, and demographic characteristics among children and
adolescents. J Am Coll Nutr. 2004; 23: 18-33.

Storm D, Eslin R, Porter ES, Musgrave K, Vereault D, Patton C, Kessenich C, Mohan S,
Chen T, Holick MF, Rosen CJ. Calcium supplementation prevents seasonal bone loss and

changes in biochemical markers of bone turnover in elderly New England women: a
randomized placebo-controlled trial. J Clin Endocrinol Metab. 1998; 83: 3817-3825.

Strauss RS, Pollack HA. Epidemic increase in childhood overweight, 1986-1998. J Am
Med Assoc. 2001; 286: 2845-2848.

Striegel-Moore RH, Thompson D, Affenito 8G, F ranko DL, Obarzanek E, Barton BA,
Schreiber GB, Daniels SR, Schmidt M, Crawford PB. Correlates of beverage intake in
adolescent girls: the National Heart, Lung, and Blood Institute Growth and Health Study.
J Pediatr. 2006; 148: 183-7.

Stroehla BC, Malcoe LH, Velie EM. Dietary sources of nutrients among rural Native
American and white children. J Am Diet Assoc. 2005;105:1908-16.

Sturrn R. The econonrics of physical activity: Societal trends and rationales for
interventions. Am J Prev Med. 2004: 27 (38): 126-135.

Sturm R. 2005a. Childhood obesity — What can we learn fi‘om existing data on societal
trends. Part 1. Preventing Chronic Disease [Online]. Available:
http://www.cdcgov/pcd/issues/2005/jan/O4 0038.htm

Sturrn R. 2005b. Childhood obesity — What can we learn from existing data on societal

trends. Part 2. Preventing Chronic Disease [Online]. Available:
http://www.cdcgov/pcd/issues/ZOO5/zmr/04_0039.htm

285

Subar AF, Krebs-Smith 8M, Cook A, Kahle LL. Dietary sources of nutrients among US
children, 1989—1991. Pediatrics. 1998; 102(4) Pt 1: 913-23.

Subar AF, Kipnis V, Troiano RP, Midthune D, Schoeller DA, Bingham S, Sharbaugh
CO, Trabulsi J, Runswick S, Ballard-Barbash R, Sunshine J, Schatzkin A. Using intake

biomarkers to evaluate the extent of dietary misreporting in a large sample of adults: the
OPEN study. Am J Epidemiol. 2003; 158: 1-13.

Subar AF, Dodd K, Thompson FE, Kipnis V. What, exactly, is the food propensity
questionnaire? FASEB J. 2004; 18 (5): A481.

Suitor CW, Gleason PM. Using Dietary Reference Intake-based methods to estimate the

prevalence of inadequate nutrient intake among school—aged children. J Am Diet Assoc.
2002; 102: 530-6.

Suter PM, Hasler E, Vetter W. Effects of alcohol on energy metabolism and body weight
regulation: is alcohol a risk factor for obesity? Nutr Rev. 1997; 55: 157-71.

Tanasescu M, Ferris AM, Himmelgreen DA, Rodriguez N, Perez-Escarnilla R.
Biobehavioral factors are associated with obesity in Puerto Rican children. J Nutr. 2000;
130: 1734-1742.

Tanner JM. Growth at Adolescence (2"d edition). Oxford, England: Blackwell Scientific
Publications; 1962.

Tavani A, Negri E, La Vecchia C. Determinants of body mass index: a study from
northern Italy. Int J Obes Relat Metab Disord. 1994; 18: 497-502.

Teegarden D. Calcium intake and reduction in weight of fat mass. J Nutr. 2003; 133:
2498-25 1 8.

Teufel NT, Dufour DL. Patterns of food use and nutrient intake of obese and non-obese
Hualapai Indian women of Arizona. J Am Diet Assoc. 1990; 90: 1229-35.

Thompson WG, Holdman NR, J anzow DJ, Slezak JM, Morris KL, Zemel MB. Effect of
energy-reduced diets high in dairy products and fiber on weight loss in obese adults.
Obes Res. 2005; 13: 1344-53.

Thomson PL. Sweetened drinks as a source of sugar intake in selected teenagers in south-
east England. Community Dent Oral Epidemiol. 1977; 5: 65-6.

Tippett KS, Enns CW, Moshfegh AJ. Food consumption surveys in the US Department
of Agriculture. Nutr Today. 1999; 34: 33-46.

286

Tooze J A, Dodd KW, Midthune D, Krebs-Snrith 8M, Subar AF, Kipnis V. Estimating
the distribution of usual intake of episodically consumed foods using two 24-hour recalls
and a food propensity questionnaire. FASEB J. 2005; 19: A481.

Tordoff MG, Alleva AM. Effect of drinking soda sweetened with aspartame or high-
fructose corn syrup on food intake and body weight. Am J Clin Nutr. 1990; 51: 963-9.

Troiano RP, F legal KM. Overweight children and adolescents: Description,
epidemiology, and demographics. Pediatrics. 1998: 101: 497-504.

Troiano RP, F legal KM. Overweight prevalence among youth in the United States: Why
so many different numbers? Int J Relat Metab Disord. 1999; 23: 822-827.

Troiano RP, Briefel RR, Carroll MD, Biolostosky K. Energy and fat intake of children
and adolescents in the United States: Data from the National Health and Nutrition
Examination Surveys. Am J Clin Nutr. 2000; 72: 13438-538.

Tsakalou Z, Yannakoulia M, Fotios A, Terzidou M, Kokkevi A, Sidossis L. Prevalence
of obesity/overweight and eating habits in Greek adolescents. Presented at the 13th
European Congress on Obesity. Int J Obes. 2004; Supplement 1(28): 8203.

U.S. Department of Agriculture (USDA). Center for Nutrition Policy and Promotion.
The Food Guide Pyramid. Washington, DC: US Government Printing Office; 1996.

U.S. Department of Agriculture (USDA). Pyramid Servings Intakes by U.S. Children
and Adults: 1994-96, 1998. Table Set No. 1. Agricultural Research Service, Community
Nutrition Research Group; 2000. [Online].

Available at http://www.barc.usda.gov/bhnrc/foodsurvey.

U.S. Department of Agriculture (USDA). Agriculture Research Service, ARS Food
Surveys Research Group (FSRG). USDA Food and Nutrient Database for Dietary
Studies, 1.0. Beltsville, MD: FSRG; 2004. [Online].

Available at http://www.barc.usda.gov/bhnrc/foodsurvey.

U.S. Department of Agriculture (USDA). Center for Nutrition Policy and Promotion.
MyPyramid Plan. 2005. [Online]. Available at http://www.mypyramid.g_o_\_r

 

U.S. Department of Labor. 2004. Women in the Labor Force: A Databook. Bureau of
Labor Statistics. Report 973. [Online]. Available: http://www.bls.gov/

Utter J, Neumark-Sztainer D, Jeffery R, Story M. Couch potatoes or french fries: are

sedentary behaviors associated with body mass index, physical activity, and dietary
behaviors among adolescents? J Am Diet Assoc. 2003; 103: 1298-305.

287

Vadstrup ES, Petersen L, Sorensen TI, Gronbaek M. Waist circumference in relation to
history of amount and type of alcohol: results from the Copenhagen City Heart Study. Int
J Obes Relat Metab Disord. 2003; 27: 23 8-46.

van Dam RM, Hu F B. Coffee consumption and risk of type 2 diabetes: a systematic
review. JAMA. 2005; 294: 97-104.

Venti CA, Tataranni A, Salbe AD. Lack of relationship between calcium intake and
body size in an obesity-prone population. J Am Diet Assoc. 2005; 105: 1401-1407.

Videon TM, Manning CK. Influences on adolescent eating patterns: the importance of
family meals. J Adolesc Health. 2003; 32: 365—73.

Wannamethee SG, Shaper AG. Alcohol, body weight, and weight gain in middle-aged
men. Am J Clin Nutr. 2003; 77: 1312-7.

Wannamethee SG, Field AE, Colditz GA, Rimm EB. Alcohol intake and 8-year weight
gain in women: a prospective study. Obes Res. 2004 ;12: 1386-96.

Wannamethee SG, Shaper AG, Whincup PH. Alcohol and adiposity: effects of quantity
and type of drink and time relation with meals. Int J Obes. 2005; 29: 1436-44.

Westerterp-Plantenga MS, Lejeune MP, Kovacs EM. Body weight loss and weight
maintenance in relation to habitual caffeine intake and green tea supplementation. Obes
Res. 2005;13:1 195-204.

Whitaker RC, Dietz WH. Role of the prenatal environment in the development of
obesity. J Pediatr. 1998: 132: 768-776.

Wilsgaard T, Jacobsen BK, Amesen E. Determining lifestyle correlates of body mass
index using multilevel analyses: the Tromso Study, 1979-2001. Am J Epidemiol. 2005;
162: 1179-88.

Wilson JF. Preschoolers’ mid-afternoon snack intake is not affected by lunchtime food
consumption. Appetite. 1999; 33: 3 19-327.

World Health Organization. 2003. Diet, nutrition and the prevention of chronic
diseases. Report of the joint WHO/FAO expert consultation. WHO Technical Report
Series, No. 916 (TRS 916). [Online]. Available:
http://www.who.int/dietphvsicalactivity/trs91 6/en/

Wosje KS, Kalkwarf HJ. Lactation, weaning, and calcium supplementation: effects on
body composition in postpartum women. Am J Clin Nutr. 2004; 80: 423-429.

WoteCki CE. Integrated NHANES: uses in national policy. J Nutr. 2003;133:5823-48.

288

Wu CH, Lu FH, Chang CS, Chang TC, Wang RH, Chang CJ. Relationship among
habitual tea consumption, percent body fat, and body fat distribution. Obes Res. 2003; 11:
1088-95.

Zemel MB, Shi H, Greer B, Dirienzo D, Zemel PC. Regulation of adiposity by dietary
calcium. FASEBJ. 2000; 14: 1132-1138.

Zemel MB. Calcium modulation of hypertension and obesity: mechanisms and
implications. J Am Coll Nutr. 2001; 20: 4288-4358.

Zemel MB. Regulation of adiposity and obesity risk by dietary calcium: mechanisms and
implications. J Am Coll Nutr. 2002; 21 : 1468-1518.

Zemel MB, Miller SL. Dietary calcium and dairy modulation of adiposity and obesity
risk. Nutr Rev. 2004; 62: 125-131.

Zemel MB, Teegarden DB, Van Loan MB, Schoeller DB, Matkovic VB, Lyle RB, Craig
BB. Role of dairy products in modulation weight and fat loss: A multi-center trial.
FASEB J. 2004a; 18: A845.

Zemel MB, Thompson W, Milstead A, Morris K, Campbell P. Calcium and dairy
acceleration of weight and fat loss during energy restriction in obese adults.
Obes Res. 2004b; 12: 582-590.

Zemel MB. Role of calcium and dairy products in energy partitioning and weight
management. Am J Clin Nutr. 2004c; 79(suppl): 9078-9128.

Zemel MB, Richards J, Mathis A, Gebhardt L, Silva E. Dairy (yogurt) augmentation of
total and central fat loss in obese subjects. Int J Obes. 2005a; 29: 391-7.

Zemel MB, Richard J, Milstead A, Campbell P. Effects of calcium and dairy on body
composition and weight loss in African-American adults. Obes Res. 2005b; 13: 1218—25.

Zizza C, Siega-Riz AM, Popkin BM. Significant increase in young adults' snacking
between 1977-1978 and 1994-1996 represents a cause for concern! Prev Med. 2001; 32:
303-10.

Zullig K, Ubbes VA, Pyle J, Valois RF. Self-reported weight perceptions, dieting

behavior, and breakfast eating among high school adolescents. J Sch Health. 2006; 76:
87 -92.

289

   

_ uliltingjijl‘lllljjutjtl