£33.}.
. . . a... as. ,. ,
a ,. a ._ 3;... 1mm ,3
. .uiuﬁnrt; “as... 2%.. .. J
.. $3... 1..., , 3%... _
. ....m_m.. «a... 6w
“Evin .
"F.

1.12:
a! . r t
I 1:

 

i _
.31? a;
(‘I l 4

iv? .

‘
r.»

,ﬁwﬁ

a:

. .
t
.2

. 31:“.
..J..!..u,)n
«0.3.9
. w”... <3
.hlwﬂpa
sauce...
.2...

.Lasmwhwww .
.Psnéi fun
“.3 3 y. .I l

a :51 rt}. .
a 36.2.1311?
:1
.t 7:.) A.
. 323:3. T

 

A075... . ..
. Emiéﬁn: ..

.x. utU.

 

TRF’SIS

$.52‘W LIBRARY

Michigan State
University

 

 

 

 

This is to certify that the

dissertation entitled

DIETARY PATTERNS ASSOCIATED WITH RISK FACTORS
FOR CARDIOVASCULAR DISEASE IN U.S. ADULTS

presented by
Jean Marie Kerver
has been accepted towards fulfillment

of the requirements for

Ph.D. degree in Human Nutrition

 

1- MIA“ /

WT Maonprofess’J

Date WNW/l C; ”(L
r

MS U is an Afﬁrmatiw Action/Equal Opportunity Institution 0- 12771

 

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

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

6/01 c'JCIRC/Dateouopes-p. 1 5

DIETARY PATTERNS ASSOCIATED WITH RISK FACTORS FOR
CARDIOVASCULAR DISEASE IN US. ADULTS

By

Jean Marie Kerver

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

2003

ABSTRACT

DIETARY PATTERNS ASSOCIATED WITH RISK FACTORS FOR
CARDIOVASCULAR DISEASE IN US. ADULTS

By

Jean Marie Kerver

Dietary intake of speciﬁc nutrients, foods, and food groups are well-established
dietary risk factors for cardiovascular disease (CVD). As dietary patterns encompass the
dietary intake of individuals within the context of lifestyles, we tested the hypotheses that
complex dietary behaviors can be systematically classiﬁed into dietary patterns that are
associated with nutrient intakes and risk factors for CVD. This study utilized data from
the third National Health and Nutrition Examination Survey (NHANES III) and all
statistical analyses accounted for the survey design and sample weights and controlled for
confounding variables (age, gender, ethnicity, income, smoking status, alcohol intake,
vitamin/mineral supplement use, BMI, and physical activity level) in regression analyses.

Using the food-frequency questionnaire and collapsing the 64 food groups
available into 35, dietary patterns of healthy US. adults (220y; n=13,130) were identiﬁed
by factor analysis. Of six dietary patterns identiﬁed, two patterns emerged most
prominently and were characterized by high intakes of 1) processed meats, eggs, red
meats, and high-fat dairy products (“Western”); and 2) green, leafy vegetables, salad
dressings, tomatoes, other vegetables, cruciferous vegetables, and tea (“American-
healthy”). The Western pattern was associated (p<0.05) positively with sermn C-peptide,

serum insulin, and glycosylated hemoglobin and inversely with RBC folate

concentrations after adjusting for confounding variables. The American-healthy pattern
had no linear relationships with any of the biomarkers examined.

Using the 24-h dietary recall, meal and snack patterns of US. adults (220 y;
n=15,978) were described in relation to nutrient intakes. Daily eating frequency was
associated (p<0.05) positively with carbohydrate (% energy), folic acid, vitamin C,
calcium, magnesium, iron, potassium, and ﬁber intakes and inversely related to protein
(% energy), total fat (% energy), cholesterol, and sodium intakes. In additional analyses
(n=10,427), daily eating frequency was associated positively with serum C-peptide
concentrations (p<0.05). Daily eating frequency was associated positively with an
“American-healthy” dietary pattern but not with a “Western” dietary pattern. After
further controlling for the American-healthy dietary pattern, the relationship between
daily eating frequency and serum C-peptide concentrations was no longer statistically
signiﬁcant (p=0.09).

We identiﬁed common dietary patterns among free-living persons that
characterize high-risk groups at the US. population level. These dietary patterns are
remarkably similar to those reported in other subgroups and are correspondingly related
to health risks. Our ﬁndings conﬁrm the hypothesis that dietary pattern analysis is a
reliable method for assessing dietary intake that predicts CVD risk, and that meal and
snack patterns are predictive of nutrient intake and selected biomarkers. The concept of
the “total diet” inclusive of both food intakes and dietary behaviors should be considered

in investigations of dietary intake and health outcomes.

Dedicated to my Mother who instilled the value of an education in all of her children and

truly believed we could do whatever we set our minds to.

iv

ACKNOWLEDGEMENTS

I will be forever grateful for the vast support I have received in completing this
endeavor. I want to thank each of my Guidance Committee members for their brilliant
insights and their never-ending help: Drs. Lorraine Weatherspoon, Norman Hord, Wanda
Chenoweth, Leonard Bianchi, Howard Teitelbaum, and Won Song. A very special
gratitude is reserved for my major Advisor, Dr. Won Song who has taken me under her
wing and nurtured me along at every step of the way. I was able to get through each day
with the camaraderie and encouragement of the remarkable members of my lab group,
especially Andrea Padgitt, Yikyung Park, Dr. EunJu Yang, Emily Meier, and Saori
Obayashi. I am most indebted to my many dear ﬁ'iends and my enormous and generous
family, who gave me space when I needed it but more often rescued me from the tedium,

and without whom none of this matters.

TABLE OF CONTENTS

List of Tables ..................................................................................................... viii
List of Figures ..................................................................................................... ix
Working Deﬁnitions ............................................................................................ x
Chapter One: Introduction ................................................................................... 1

Chapter Two: Review of Literature
A. Cardiovascular Disease

1. Pathophysiology ............................................................................ 9
2. Prevalence .................................................................................... 10
3. Risk Factors ................................................................................. 11
4. Dietary Effects ............................................................................. 13
B. Dietary Patterns
1. Description .................................................................................. l4
2. Measurement ............................................................................... 14
3. Advantages .................................................................................. 17
4. Limitations ................................................................................... 19
5. Validity and Reliability ............................................................... 20
6. Studies of Dietary Patterns .......................................................... 21
C. Daily Eating Frequency
1. Description .................................................................................. 25
2. Studies of Daily Eating Frequency ............................................. 26
D. Summary ................................................................................................ 29
Chapter Three: Dietary Patterns Associated with Risk Factors for Cardiovascular Disease
in Healthy U.S. Adults
A. Abstract .................................................................................................. 31
B. Introduction ............................................................................................ 32
C. Subjects and Methods ............................................................................ 34
D. Results .................................................................................................... 39
E. Discussion .............................................................................................. 41

Chapter Four: Meal and Snack Patterns are Associated with Dietary Intake of Energy and
Nutrients in US. Adults

A. Abstract .................................................................................................. 56
B. Introduction ............................................................................................ 57
C. Subjects and Methods ............................................................................ 58
D. Results .................................................................................................... 61
E. Discussion .............................................................................................. 62

vi

Chapter Five: The Relation of Meal and Snack Frequency to Dietary Patterns and
Cardiovascular Disease Risk Factors in Healthy U. S. Adults

A. Abstract .................................................................................................. 71
B. Introduction ............................................................................................ 72
C. Subjects and Methods ............................................................................ 73
D. Results .................................................................................................... 77
E. Discussion .............................................................................................. 78
Chapter Six: Conclusion
A. Implications ........................................................................................... 84
B. Recommendations for Future Research ................................................ 86
Bibliography ....................................................................................................... 88

LIST OF TABLES

Chapter Three
Table 1. Food groupings used in the dietary pattern analysis ................................... 46

Table 2. Factor-loading matrix for major and minor dietary patterns in healthy
US. adults .................................................................................................................. 48

Table 3. Population percentages in sociodemographic and lifestyle categories
by quintile of dietary pattern scores in healthy US. adults ...................................... 49

Table 4. Age-adjusted mean biomarker values by quintile of Western dietary
pattern scores in healthy US. adults .......................................................................... 50

Table 5. Age-adjusted mean biomarker values by quintile of American-
healthy dietary pattern scores in healthy US. adults ................................................ 51

Table 6. Mulitvariate-adjusted mean biomarker values by quintile of
Western dietary pattern scores in healthy U. S. adults ............................................... 52

Table 7. Mulitvariate-adjusted mean biomarker values by quintile of
American-healthy dietary pattern scores in healthy US. adults ............................... 53

Chapter Four
Table 1. Sociodemographic and lifestyle characteristics by daily eating
frequency in US. adults ............................................................................................ 67

Table 2. Multivariate-adjusted mean nutrient intakes by daily eating
frequency in US. adults ............................................................................................ 68

Table 3. Sociodemographic and lifestyle characteristics by selected meal
patterns in US. adults ................................................................................................ 69

Table 4. Multivariate-adjusted mean nutrient intakes by selected meal

patterns in US. adults ................................................................................................ 70
Chapter Five

Table l. Mulitvariate—adjusted mean biomarker values by daily eating

frequency in healthy US. adults ................................................................................ 81

Table 2. Regression model for Western dietary pattern by eating frequency ........... 82

Table 3. Regression model for American-healthy dietary pattern by
eating frequency ......................................................................................................... 83

viii

LIST OF FIGURES

Figure 1. Hierarchical depiction of the two major (A, B) and four minor (C, D, E, F) dietary
patterns detected in healthy US. adults ................................................................................. 54

WORKING DEFINITIONS

Daily Eating Frequency. Number of eating occasions reported in a 24-hour time period.

Dietary Index. A summary measure of dietary intake designed to assess overall diet
quality, which is typically constructed on the basis of dietary recommendations. An
example is the Healthy Eating Index (HEI), which is a summary measure of the degree to

which an individual’s diet conforms to recommendations in the US. Dietary Guidelines for
Americans.

Dietary Intake. Foods, nutrients, and other food components ingested by an individual.

Dietary Patterns. Summary measure of foods, nutrients, and dietary behaviors within the
total diet.

Dietary Standards. Standards by which dietary intake is compared (e.g., Dietary
Reference Intakes).

Factor. A cluster of related variables that are a mathematically distinguishable component
of a larger group. of variables in factor analysis.

Factor Analysis. A method of analysis used to reduce a large number of variables to a
smaller number of variables, or factors; a factor is a set of variables, such as items on a
survey, that can be conceptually and statistically related or grouped together. Factor
analysis is done by ﬁnding patterns among the variations in the values of several variables;
a cluster of highly intercorrelated variables is a factor.

Factor Loadings. The correlations between each variable and each factor in a factor
analysis. They are analogous to regression coefﬁcients. The higher the loading, the closer
the association of the item with the group of items that make up the factor.

Factor Score. A measure of an individual’s association with a factor. The higher the factor
score of an individual, the closer the association of the individual with the factor.

Meal Frequency. The number of meals (i.e. breakfast/brunch, lunch, dinner) reported in a
24-hour time period.

Observational Study. A research design in which the investigator observes subjects but
does not impose experimental protocols.

Quintile. Division of the total cases in a study into ﬁve groups of equal size.

Total Diet. An overall pattern of foods eaten, rather than any one food or meal.

X

Chapter 1

INTRODUCTION

Cardiovascular disease (CVD) is the number one cause of death in the United
Sates (US) and is one of the most serious public health problems we face today. The risk
varies with age, gender, ethnicity, and socioeconomic status. Modiﬁable risk factors
include smoking, physical inactivity, hypertension, dyslipidemia, diabetes, and
overweight (Krauss et al. , 2000). The latter four risk factors can potentially be improved
by changes in dietary intake. Speciﬁc dietary components (e.g. excess saturated fat,
inadequate dietary ﬁber), however, explain only a portion of CVD risk.

Dietary intake consists of consumption of individual foods or food mixtures that
are in turn made up of a complex set of macronutrients, micronutrients, and other food
components. Nutrition research has largely focused on the study of speciﬁc nutrients and
other food components in the prevention of chronic disease. In many instances, however,
it is not accurate to presume that nutrient intake can represent food intake. A clear
example of this concept was evidenced in the attempt to prevent coronary heart disease
(CHD) with beta-carotene supplements (ATBC Study Group, 1994; Hennekens et al.,
1996; Omenn et al. , 1996). Based on animal studies and prospective human cohort
studies, beta-carotene was thought to play a role in the prevention of CHD through its
antioxidant properties. However, in randomized controlled primary prevention studies
beta-carotene supplements either had no effect or increased the incidence of coronary
events and cancer death (ATBC Study Group, 1994; Hennekens et a1. , 1996; Omenn et

al. , 1996). Partially in response to this occurrence, there has been a surge of interest in

the study of dietary patterns, which is an attempt to summarize the synergy of foods and
nutrients within the total diet in relation to human health and disease. Dietary patterns
are inclusive of the roles of bioactive compounds in speciﬁc foods or food groups
(phytochemicals, non-nutritive and nutritive compounds) and their interactions and are
consumed within the context of available food supplies and individual lifestyles.

Since dietary intake is such a complex exposure variable, it is now recognized that
we must develop new methods and reﬁne existing methods of assessing dietary intake
that focus on the total diet and not just individual dietary components (e. g. nutrients,
phytochemicals) or behaviors (e. g. adding salt at the table). The improvement of
methods for evaluating dietary intake will lead to a better understanding of the risk for
CVD that is attributable to diet and may ultimately lead to decreases in CVD incidence
through public education and other interventional efforts to promote optimal dietary
intakes.

While there are several methods of summarizing certain aspects of dietary intake
(e.g. nutrient adequacy ratios, number of servings of speciﬁc food groups), it is a
challenge to ﬁnd a method that accurately describes the complete dietary intake of an
individual in all of its complexity. Increasingly, multivariate data reduction techniques,
such as factor analysis and cluster analysis, are being used to translate food or food group
intake into identiﬁable dietary patterns. The underlying assumption of statistical data
reduction regarding food intake is that foods eaten together (i.e. within the total diet) can
be characterized as part of a dietary pattern that is more epidemiologically meaningful
than individual foods (Slattery et al., 1998). It is important to note, however, that data

reduction by any method can occur only after a complete understanding of the individual

components. Thus, dietary pattern analysis is an extension of the many years of research
involving speciﬁc dietary components. Additionally, while we now have the luxury of
studying dietary patterns because we have sufﬁcient information about the individual
components of diet, it will remain critically important to continue investigating singular
dietary components for many reasons, including the correct interpretation of dietary
pattern research.

By using multivariate data reduction techniques, dietary patterns have been
identiﬁed and associated with various sociodemographic characteristics and chronic
disease risks in age- gender- and race-speciﬁc populations. In a large case-control study
(n=1,993 cases and n=2,410 controls), Slattery, et al. (1998) identiﬁed dietary patterns in
both men and women that were associated with risk of colon cancer in a multiethnic
population. Study participants were from the Kaiser Permanente Medical Care Program
of Northern California and the majority were over 60 years old and thus represents an
age-speciﬁc population. In two large prospective studies including men (Health
Professionals F ollow-up Study) (Hu et al., 2000) and women (Nurses’ Health Study)
(Fung et al. , 2001b), dietary patterns were associated with increased risk for CHD.
Further analyses of the Health Professionals Follow-up Study revealed associations
between dietary patterns and plasma biomarkers of obesity and cardiovascular disease
risk in men (Fung et al., 2001a). Although these were important ﬁndings, the participants
in both the Health Professionals Follow-up Study and the Nurses’ Health Study were
mostly white, over 40 years of age, and of higher socioeconomic status, and thus the
ﬁndings must be reproduced in other populations before any general conclusions can be

drawn.

Interestingly, the two major dietary patterns identiﬁed in all three studies
mentioned above were qualitatively similar and labeled by the authors as l)“Westem,”
which was characterized by a high intake of red meat, processed meat, high-fat dairy
products, eggs, and reﬁned grains, and 2)“prudent,” which was characterized by a high
intake of vegetables, fruit, legumes, and ﬁsh. Broadly speaking, these studies have
identiﬁed similar major dietary patterns, with the Western pattern consistently associated
with increased disease risk, and the prudent pattern consistently associated with
decreased disease risk. The similarity of dietary patterns across these diverse study
populations lends credibility to the pursuit of identiﬁcation of dietary patterns related to
health and disease that can be used to guide public health recommendations. While the
study of dietary patterns appears to have merit based on current literature in speciﬁc
population groups, general dietary patterns need to be identiﬁed in nationally
representative populations in order to better understand the impact of diet on CVD risk at
the population level.

In addition, dietary patterns based on food intake still do not account for the full
range of dietary behaviors. Besides the composition of dietary intake, dietary behaviors
such as the daily number of meals or snacks may potentially also eﬁ‘ect risk for CVD.
Increased daily eating frequency has been shown to improve lipid proﬁles and markers
for glucose metabolism in men with diabetes and healthy men, which may mediate the
risk for CVD (Jenkins et a1. , 1989; McGrath & Gibney, 1994).

Early epidemiological observations revealed a higher incidence of obesity and
ischemic heart disease with fewer daily eating occasions in men (Fabry & Tepperrnan,

1970) (Fabry et al. , 1968). There have now been numerous observational and

experimental studies regarding daily eating frequency and CVD risk factors, but they
have provided inconsistent results. Large observational studies have consistently shown
more favorable lipid proﬁles with increasing daily eating frequency (Titan et al., 2001),
(Redondo et al. , 1997), (Edelstein et al., 1992). On the other hand, while short-term
clinical experiments tend to show that increased frequency of eating leads to favorable
blood lipid and glucose concentrations under controlled feeding conditions, longer
experiments do not (Gibney & Wolever, 1997). These ﬁndings may be explained in part
by the observation that increased meal frequency is associated with a decrease in human
cholesterol biosynthesis (Jones, 1997), but perhaps metabolic adaptations occur over
time.

In one study subjects were identiﬁed as men whose usual eating habits included
either three or six meals per day (McGrath & Gibney, 1994). The subjects were then
instructed to convert to the alternate eating pattern for 3 weeks duration and increasing
meal frequency was accompanied by a signiﬁcant reduction in total and LDL-cholesterol
concentrations. Although subjects were instructed not to change the nutritional
composition of their diet, signiﬁcant changes in nutrient intake (% energy from protein,
fat, saturated fat, and alcohol) occurred in the group who decreased their meal frequency.
This highlights the fact that changing eating frequency may precipitate changes in
nutrient composition and may also explain the changes in lipid concentrations seen in this
study. This implies that the discrepancy between epidemiological observations and
experimental trials may actually be because meal frequency is a surrogate measure that is
a strong covariate of some other variable (e.g. nutrient composition) that is affecting lipid

proﬁles and/or glucose metabolism.

In free-living populations, nutrient intakes may vary depending on meal or
snacking habits. In fact, increased eating frequency was associated with higher daily
intakes of energy, fat, carbohydrate, and protein in the Norfolk population of the
European prospective investigation into cancer (EPIC-Norfolk), however, more favorable
lipid concentrations were still associated with increasing meal frequency even after
adjusting for energy, macronutrient intake, and other confounding variables in
multivariate analyses (Titan et al., 2001). In the EPIC-Norfolk study, frequency of eating
was asSessed using the question, “How many times a day do you eat, including meals,
snacks, biscuits with coffee breaks, etc.?” and nutrient intake was assessed with a
quantitative 160 item food frequency questionnaire. The assumption was thus made that
people habitually eat the same number of meals or snacks each day, although there is no
published research to support this assumption. Despite this limitation, the EPIC-Norfolk
study provides important infonrration regarding eating frequency, macronutrient intake,
and serum lipid concentrations in the Norfolk population and should be veriﬁed in other
populations.

More research is needed to determine if eating frequency is related to nutrient
intake within a given day. In other words, within a 24-hr period, are diets that consist of
more frequent eating occasions related to more favorable nutrient intakes in free-living
populations? Moreover, since the nutritional composition of eating occasions is likely to
be as important as eating frequency, it is important to examine the nutrient intakes among
individuals with different daily eating frequencies in free-living populations.
Furthermore, the daily eating frequency may not only be related to nutrient intakes, but to

overall dietary patterns and both may be part of a larger pattern of health-related

behaviors. Therefore, this study describes the mean nutrient intakes and biomarker
concentrations, among groups with different daily eating frequencies and different dietary
pattern scores (i.e. factor scores) in an effort to determine associations between meal and
snack patterns, nutrient intake, dietary patterns, and risk factors for CVD.

To summarize, CVD imposes a huge burden of illness in the US. and both dietary ‘
patterns and daily eating frequency have been associated with risk factors for CVD in
speciﬁc epidemiological cohorts or small clinical trials. However, no similar work has
been done to explore the associations between dietary patterns or daily eating ﬁ'equency
and risk factors for CVD using nationally representative populations. Hence, this study
examined the associations between measures of dietary intake (i.e. dietary patterns and
daily eating frequency) and risk factors for CVD (i.e. sociodemographic and lifestyle
characteristics, physical measurements, and biomarkers) in the healthy U.S. adult
population. To test the hypotheses that in US. adults 1) individual dietary intake data
can be systematically classiﬁed into major dietary patterns based on food group intake
that are related to risk factors for CVD; 2) meal and snack patterns (including daily eating
frequency) are associated with nutrient intakes; and 3) meal and snack patterns are
associated with both risk factors for CVD and dietary patterns based on food group
intake, the speciﬁc objectives of this study were to use the most recent national dietary
and health data available (NHANES III) to:

1. a) Identify major dietary patterns with a factor analysis of the food frequency
questionnaire items; and b) determine the association between major dietary

patterns and CVD risk factors (Chapter Three).

2. a) Describe the number of meals and snacks eaten per day (based on 24-hr dietary
recall and self-reported deﬁnitions of meal and snack occasions); and
b) determine the association between the number of eating occasions and daily
macronutrient (% energy) and micronutrient intakes (Chapter Four).

3. a) Determine the association between daily eating frequency (using aggregate
data) and CVD risk factors; and b) determine the association between daily eating

frequency (using aggregate data) and the dietary patterns (using the factor score)

(Chapter Five).

Chapter 2

REVIEW OF LITERATURE

A. Cardiovascular Disease
1. Pathophysiology

The term “cardiovascular disease (CVD)” encompasses a broad range of diseases
that can include any part of the cardiovascular system. Coronary heart disease (CHD) is
the most common type of CVD and also the ﬁnal common pathway of most
cardiovascular diseases. The primary pathogenesis of the major cardiovascular diseases
is the obstruction of blood vessels by atherosclerosis or thrombosis, either singly or in
combination. Atherosclerosis is the principal pathologic process that underlies CHD and
other cardiovascular diseases such as peripheral artery disease and some kinds of stroke
(Cohn et al., 1997).

The process of atherosclerosis consists of both the accumulation of lipids within
the artery wall and a series of cellular and molecular inﬂammatory responses (Ross,
1999). The initiating event of atherosclerosis is the formation of lesions that obstruct the
ﬂow of blood somewhere in the cardiovascular system. These lesions get progressively
worse and are commonly classiﬁed as fatty streaks, ﬁbrous plaques, and complicated
lesions, in that order of severity. The fatty streaks contain atheromas, or hard masses of
fat, and foam cells. As the disease progresses, ﬁbrous plaques form, which are areas of
thickening of the arterial wall that contain the fatty streaks in the center that are
surrounded by smooth muscle cells, macrophages and collagen. Finally, the complicated

lesion is formed, which is made up of calciﬁed tissue containing various degrees of

necrosis, thrombosis and ulceration. When the blood vessel is approximately 70%
occluded from these lesions, the blood ﬂow is cut down sufﬁciently that the surrounding
tissue is often injured (Phibbs, 1997).
2. Prevalence

CVD is the leading cause of death in the United States (US) in both men and
women. In 1998 in the US, 949,619 people died from CVD, which is 40.6 percent of all
deaths or 1 of every 2.5 deaths. CVD was about 70 percent of “total mention mortality,”
which means that of the more than 2,000,000 deaths ﬁom all causes, CVD was listed as a
primary or contributing cause on about 1,400,000 death certiﬁcates (AHA, 2000). CHD
is the cause of approximately half of all CVD deaths, or about 500,000 deaths annually
(N CHS, 2001). Mortality rates from CHD have been declining in the US. for the past 30
years, but morbidity rates are less clear and the burden of illness is still enormous.

Women and both black and Mexican-American men appear to have a lower risk
for CHD than white men when the data is adjusted for other risk factors including
hypertension, age, and income, but mortality rates are declining faster in white men than
in women or blacks (Keil et al., 1995), (Ho et al., 1993), (Mitchell et al., 1991).
Mortality rates are increasing in poorer persons and decreasing in the wealthy. In the
1950’s, heart disease mortality rates were greater among persons with higher
socioeconomic status (SES), but now mortality rates are higher among persons with
lower SES (NCHS, 2001). The shift is presumably because higher SES groups more
quickly adopt practices to reduce the risk of chronic disease. This includes lifestyle
changes, but also reﬂects the greater access to health care and medication that people of

higher SES enjoy.

10

3. Risk Factors

Risk for CVD is multi-factorial and includes non-modiﬁable risk factors such as
genetic predisposition, male gender, and advanced age, but also modiﬁable risk factors
including smoking, physical inactivity, hypertension, dyslipidemia, type 2 diabetes, and
overweight (Krauss et al. , 2000). Especially notable is the recent increase in overweight
and obese individuals in the US, which is contributing substantially to the burden of
chronic health conditions. Results ﬁ'om the 1999-2000 NHANES estimate that 64% of
US. adults are either overweight (body mass index 25.0-29.9) or obese (body mass index
2 30.0) which represents an 8% higher prevalence than that reported from NHANES III,
1988-94 (Flegal et al. , 2002). Additionally, data from the 1999-2000 NHANES estimate
that 15 percent of adolescents aged 12-19 years are overweight, which represents a 4%
higher prevalence than estimates from NHANES III (Ogden et al. , 2002) and will likely
lead to continued increases in adult obesity.

Much attention has been focused on the relationship between diet and CVD as
mediated by the effect of diet on blood cholesterol and lipoprotein proﬁles. This is in
part because large epidemiological observations have shown that higher concentrations of
serum total cholesterol predict CVD mortality (Rosenman et al., 1976) and large clinical
prevention trials such as the Lipid Research Clinics Coronary Primary Prevention Trial
(LRC-CPPT) have shown that lowering serum total cholesterol and low density
lipoprotein cholesterol (LDL-C) will reduce risk for CVD mortality (LRC-CPPT, 1984a).
Regardless of the cause of elevated total serum cholesterol, the LRC-CPPT showed that a
1% decline in serum cholesterol correlates with a nearly 2% decrease in cardiovascular

events in a multicenter, prospective, randomized, double-blind study (LRC-CPPT,

11

1984b). According to the classic “diet—heart” hypothesis, excessive intakes of saturated
fats and dietary cholesterol increase the concentration of serum cholesterol, which then
leads to the development of coronary artery disease (Gordon, 1988). Furthermore, a large
body of research has shown that various other dietary practices can impact blood
cholesterol and lipoprotein proﬁles (Carleton et al. , 1991) (NHLBI, 2001).

Plasma homocysteine (Hcy) is also considered to be a biomarker for CVD risk
that is affected by speciﬁc nutrient consumption (i.e., vitamins B-6 and folate). Data
from approximately 80 clinical and epidemiological studies including more than 10,000
patients has indicated that elevated total plasma Hey is a strong risk factor for
athersclerotic vascular disease in the coronary, cerebral, and peripheral vessels and for
arterial and venous thromboembolism (Refsum et al. , 1998). Hyperhomocysteinemia is
caused by common genetic factors as well as deﬁciencies of folate and vitamin B6. Other
biomarkers for CVD include altered vascular and endothelial factors, Apolipoprotein A-I,
lipoprotein(a), ﬁbrinogen, insulin, C-peptide, glycosylated hemoglobin, and C-reactive
protein (F ung et al. , 2001a).

Creating an accurate risk factor proﬁle for an individual, however, is difﬁcult
because of the interactions between risk factors and the possible multiplicative effects of
certain risk factors in combination. For example, the effects of smoking on
atherosclerosis are thought to be precipitated by changes in vascular tone and increased
platelet aggregation, which may be aggravated by hyperhomocysteinemia. Additionally,
the beneﬁts of physical activity may be from its effects on maintaining a healthier body
weight, lowering total serum cholesterol, increasing high density lipoprotein cholesterol

(HDL-C), and improving muscle tone (Schroder et al. , 2002).

12

4. Dietary Effects

Speciﬁc dietary factors have been associated with risk of CVD morbidity and
mortality and biomarkers for CVD risk. In fact, there are many dietary factors that have
been shown to be associated with increased risk of CVD including: total energy,
macronutrient intake and distribution, speciﬁc fatty acid classes (e.g., saturated,
monounsaturated, polyunsaturated, n-3 fatty acids, and trans fatty acids), dietary
cholesterol, dietary ﬁber, antioxidant vitamins and minerals, pro-oxidants (e.g., iron), and
minerals affecting blood pressure (e.g., sodium, potassium, calcium, and magnesium).
Additionally, speciﬁc foods and food groups have been associated with altered CVD risk
including alcohol, fruits and vegetables, grains, nuts, ﬁsh, soy protein, red meat, and
dairy products (both whole fat and non fat dairy products have been implicated in
increased and decreased risk, respectively). Although the spectrum of research regarding
diet and CVD risk is vast indeed, current population recommendations address only those
issues that are known with relative certainty to decrease risk of CVD.

Recommendations from the third Adult Treatment Panel (ATPIII) of the National
Cholesterol Education Program (N CEP) include weight management, physical activity,
and restriction of calories from saturated fat (< 7% of calories), and dietary cholesterol (<
200 mg/day), and increasing intake of soluble ﬁber (10-25 g/day) (NHLBI, 2001). The
American Heart Association released new guidelines in the year 2000 stating that the
major emphasis for weight management should be on avoidance of excess total energy
intake and a regular pattern of physical activity. Fat intake of 330% of total energy is
recommended to assist in limiting consumption of total energy as well as saturated fat.

The guidelines continue to advocate a population-wide limitation of saturated fat to <10%

13

of energy and dietary cholesterol to <300 mg/d. Because of increased evidence for the
cardiovascular beneﬁts of ﬁsh (particularly fatty ﬁsh), consumption of at least 2 ﬁsh
servings per week is now recommended. Finally, recent studies support a major beneﬁt
on blood pressure of consuming vegetables, fruits, and low-fat dairy products, as well as
limiting salt intake (<6 grams per day) and alcohol (no more than 2 drinks per day for
men and 1 for women) and maintaining a healthy body weight (Krauss et al. , 2000).
B. Dietary Patterns
1. Description

As evidenced in the dietary recommendations above, the traditional focus of
nutrition science on the amount and distribution of nutrients is being replaced by a
combination focus on both nutrients/food components (i.e. fat, saturated fat, cholesterol,
ﬁber) and foods (i.e. ﬁsh, vegetables, fruit, low-fat dairy products). The possibility of
many undiscovered compounds in foods, and the enormity of interactions among food
components are now recognized. As a result, the study of patterns of nutrients, foods,
and food—group intakes has begun to emerge in nutrition research in an attempt to capture
a snapshot of the entire diet.
2. Measurement

One method of describing the overall diet is to assess the agreement of an
individual’s diet with pre-defrned dietary standards. This approach builds on previous
knowledge concerning the health or disease promoting effects of speciﬁc dietary
components, and is thus referred to as a priori. It involves the use of a graded score to
assess an individual’s diet against our best estimation of an ideal diet, which is usually

based on scientiﬁc evidence regarding diet-disease relationships and current dietary

l4

recommendations (Hu, 2002). Examples include a scoring system derived from studies
of food groups, individual food items, and nutrients in relation to common diseases in the
population (Nube et al. , 1987), and the healthy eating index (Kennedy et al. , 1995),
which is a summary measure describing how well an individual’s diet conforms to the
US. dietary guidelines. Necessarily, a priori approaches are limited by current scientiﬁc
knowledge of diet-disease relationships.

On the other hand, factor analysis and cluster analysis detect patterns from
existing data with no prior assumptions of health or disease relationships and are referred
to as a posteriori. Factor analysis is a statistical technique used both to reduce the
number of variables required to explain the relationships among a set of variables, and to
classify the variables into a smaller number of variables called factors (Lawley, 1963).
Factor analysis does not assess the relation between independent and dependent variables.
Instead, it is used to assess the patterns of relationship among many dependent variables,
with the goal of discovering something about the nature of the independent variables that
affect them, even though those independent variables were not measured directly
(Lawley, 1963). Therefore, conclusions reached by factor analysis are by deﬁnition more
hypothetical than when independent variables are directly observed. The inferred
independent variables are called factors.

Principal component analysis is the method of factor analysis that has been used
to deﬁne dietary patterns because the principal components of a correlation matrix are
mathematical functions of the observed variables. The central concept of principal
components analysis is summarization of a large set of variables that can be replaced by a

smaller set, which best represents the larger set (Kim, 1978). Results of a principal

15

components analysis include a factor loading matrix, which comprises the dietary
patterns (i.e. factors) for the entire sample, and a factor score for each individual, which
is derived by summing the individual intakes of the food items weighted by standardized
scoring coefﬁcients for each factor. Thus each subject has a factor score for each factor
that emerges from the data reduction.

Cluster analysis is another exploratory multivariate statistical technique used to
describe dietary patterns. While factor analysis aggregates food groups based on their
correlation with one another, cluster analysis aggregates individuals based on the
similarity of their diets. The purpose is to sort people into relatively uniform groups, (i.e.
clusters), so that members of each cluster have dietary intakes similar to one another, but
not similar to the dietary intakes of members of other clusters (Darlington, 1973).

The ability of cluster analysis to separate individuals into distinct groups is an
advantage over factor analysis in which an individual may score high on more than one
factor and thus results may be more difﬁcult to interpret. Another advantage of cluster
analysis is that it can be applied to any sort of similarity measures, such as ratings of the
similarity of likert scale responses, whereas factor analysis is usually only applied to a
correlation matrix. On the other hand, cluster analysis is not designed to recognize certain
unique properties of correlation matrices. For example, factor analysis is able to
recognize the directionality of correlations (-1 to 1) and changes the signs of the factor
loadings but does not change anything else in the factor analysis output, whereas negative
correlations completely change the output of a cluster analysis. Additionally, factor
analysis can recognize other properties of correlations that cluster analysis has no ability

to recognize, since the associations in cluster analysis are treated only as generic

16

similarity measures rather than as correlations. For instance, if variables A and B each
correlate 0.7 with variable C, and correlate 0.49 with each other, factor analysis can
recognize that A and B correlate zero when C is held constant because 0.72 = .49
(Darlington, 1973).

Both factor analysis and cluster analysis have distinctive features and the study
question at hand must guide the decision regarding which technique to use. Each method
has the ability to uncover associations and structure that are both sensible and useful,
even though not previously apparent. The results of both may contribute to the deﬁnition
of a formal classiﬁcation scheme of dietary patterns.

3. Advantages

Assessment of the total diet by any method has several advantages over single
nutrient analyses. For one thing, we have many examples of known interactions between
nutrients (e. g. enhanced calcium absorption in the presence of vitamin D) (Devine et al. ,
2002). Moreover, dietary intake is inherently complex with the possibility of unknown
components in foods that may interact with known nutrients or yet other unknown food
components. Similarly, single nutrient effects may be small, but their cumulative effect
may be beneﬁcial (or harmful) to human health. Dietary pattern analysis incorporates all
nutrient interactions in the diet and allows us to capture diet-disease or diet-biomarker
relationships without knowing the exact nutrient or food component involved (Jacobs &
Murtaugh, 2000).

Additionally, since many nutrients are highly correlated within foods, it is
difﬁcult to examine their effects separately. For example, other beneﬁcial compounds

found in high ﬁber foods such as antioxidants and other phytochemicals confound

l7

conclusions from early studies regarding dietary ﬁber intake. Likewise, nutrient intakes
may be confounded by entire dietary patterns. As previously discussed, beta-carotene
supplementation failed to prevent chronic disease in prospective trials that were
conducted aﬁer the observation that fruits and vegetables (i.e. beta-carotene containing
foods) were associated with lower incidence of disease. This has been partially attributed
to a combination of the other beneﬁcial components of diets high in fruits and vegetables,
including folate, ﬁber, magnesium, potassium, ﬂavonoids, and plant sterols (Bazzano et
al,2002)

The diet as a whole in free-living populations must be considered in assessing
both diet-disease relationships and the effects of dietary recommendations because the
addition or deletion of certain foods within a diet tends to displace other foods. There is a
move toward inclusive food-based recommendations in the prevention of chronic disease,
however, the addition of all foods thought to be beneﬁcial may lead to the consumption
of excess energy (Kris-Etherton et al. , 2002). While it is the goal of most dietary
recommendations to have healthy foods displace less healthy foods in the diet, it is
critical to examine the entire dietary pattern in order to determine the actual effects of
dietary changes. Additionally, certain foods (e.g. eggs) may contain both beneﬁcial (i.e.
unsaturated fats, essential amino acids, folate, other B vitamins) and detrimental nutrients
(i.e. saturated fat and cholesterol) and thus the individual nutrient effects may actually
cancel each other out.

Finally, it has been suggested that interventions based on total diet approaches are
easier to implement. It is the position of the American Dietetic Association that healthful

eating messages to the public should emphasize the total diet, or overall pattern of food

18

eaten, rather than any one food or meal (F reeland-Graves & Nitzke, 2002). Assessing
dietary patterns that reﬂect actual consumption may provide insights into possibilities for
dietary changes and also may help facilitate the ﬁndings into public health
recommendations (van Dam et al. , 2002). Two notable clinical trials have set precedents
not only in demonstrating that dietary patterns can have a marked beneﬁcial impact on
important risk factors for cardiovascular disease, but also in their ability to persuade
subjects to adopt and comply with new dietary habits (Appel et al. , 1997) (de Lorgeril et
al. , 1999).
4. Limitations

As with all nutrition research, dietary pattern analysis can only be as good as the
dietary assessment method it is based upon. Dietary recalls, food records, or food
frequency questionnaires (FFQ) are generally used to assess dietary intake. FF Qs require
subjects to recall their usual frequency of consumption of each food or food group from a
list of foods for a given period of time. Because the FF Q gathers less information
regarding speciﬁc foods, cooking methods, and portion size than 24-hour dietary recalls
or food records, the quantiﬁcation of nutrient intake is not considered as accurate.
Measurement errors inherent in this retrospective method of dietary assessment include
possible under or over-reporting of general food intake, or selective under or over-
reporting of certain foods, or both. Validation studies comparing FFQs with repeated
dietary recalls generally show correlations in the range of 0.4 to 0.7 for most foods and
nutrients (Subar et al. , 2001). In validation studies comparing FFQs with biomarkers of
nutrient intake, plasma triacylglycerol, vitamin C, and carotenoid concentrations also

correlate similarly with food frequency estimates of both macronutrients (Willett et al. ,

19

2001) and micronutrients (Block et al., 2001). F F Qs have been shown to correlate
moderately well with 24—h urinary nitrogen (r=0.21-0.29) and urinary potassium (r=0.32-
0.34), even though 7 day food diaries provided better estimates of nitrogen and potassium
intakes (McKeown et al. , 2001). However, the FFQ is easier to administer and less
costly to analyze, and because it is designed to assess usual intake, most large
epidemiological studies have utilized food frequency questionnaires. Accordingly, the
limitations of F F Qs also apply to dietary pattern analyses that are based on dietary
information collected from a FF Q.
5. Validity and Reliability

Dietary patterns resulting from a factor analysis of FFQ items have been validated
against dietary patterns derived from diet records (Hu et al. , 1999). Hu et al examined the
reproducibility and validity of dietary patterns deﬁned by factor analysis by comparing
results from a food-frequency questionnaire administered twice, 1 year apart, and two 1-
week diet records among a subset of subjects in the Health Professionals Follow-up
Study (Hu et al. , 1999). They identiﬁed two major eating patterns which they labeled
“Prudent” and “Western.” The Prudent pattern was characterized by a higher intake of
vegetables, fruits, legumes, whole grains, and ﬁsh, whereas the Western pattern was
characterized by a higher intake of processed meat, red meat, butter, high-fat dairy
products, eggs, and reﬁned grains. The correlations between the F FQ and the diet
records (0.45 to 0.74) for the two patterns and the correlations between factor scores and
plasma biomarkers, which were in the expected directions, indicated reasonable
reproducibility and validity of the major dietary patterns deﬁned by factor analysis.

Similar studies exist validating dietary patterns derived from cluster analysis by

20

comparing nutrient or biochemical proﬁles between the clusters (Tucker et al. , 1992;
Millen et al. , 2001).

The reproducibility of dietary patterns across populations, however, has not been
examined and doing so brings up the subjective nature of factor analysis. In conducting
factor analysis, the researcher must decide which variables to enter into the correlation
matrix. A bias in either the inclusion or exclusion of certain variables can cause
problems. The inclusion of unrelated variables can have the eﬁ‘ect of redeﬁning factors
because of shared extraneous variance, whereas exclusion of variables in order to
simplify the factorial structure can lead to erroneous conclusions (Kim, 1978). Although
dietary patterns would not be expected to be uniform across all populations, the use of
factor analysis has been criticized because of its subjective nature (Martinez et al., 1998).
However, similar ﬁndings across studies would support the use of factor analysis in
nutrition epidemiology. The present study attempted to replicate or refute existing
literature by deliberately using the same food groups that have previously been reported
(Hu et al. , 1999) and determining if similar dietary patterns exist in a nationally
representative sample. Further, we sought to ascertain if the relationship between dietary
patterns and risk factors for CVD is consistent across study populations.

5. Studies of Dietary Patterns

It is now recognized that observational studies regarding dietary patterns are
needed because the description of dietary patterns that exist in free-living populations
may allow us to discover new associations between diet and disease risk (van Dam et al. ,
2002). Although the study of dietary patterns is not novel, as Schwerin et al. (1982)

described dietary patterns from national nutrition survey data in 1982 using factor

21

analysis, the use of factor analysis gained greater credibility in nutritional epidemiology
after the study conducted by Slattery et al. (1998) associated dietary patterns with colon
cancer risk.

The associations between dietary patterns and the risk of CHD were examined in
two large ongoing cohort studies, the Nurses’ Health Study (NHS) and the Health
Professionals’ Follow-up Study (HPF U) (Hu et al., 2000; Fung et al., 2001b). In both
studies, after adjustment for confounding variables, risk for CHD was higher in persons
consuming the Western dietary pattern, while risk for CHD was lower in those
individuals consuming the prudent dietary pattern. In further analyses of a subset of men
from the HPFS, dietary patterns were signiﬁcantly associated with some biomarkers of
CHD and obesity (tissue plasminogen antigen, fasting insulin, C-peptide, leptin, C-
reactive protein, homocysteine, and plasma folate), but not with others (total plasma
cholesterol, HDL-cholesterol, LDL-cholesterol, triacylglycerol, apolipoprotien A-I,
lipoprotein(a), ﬁbrinogen, von Willebrand factor, factor VII antigen, and glycosylated
hemoglobin concentrations) (Fung et al., 2001a).

Several other studies also observed dietary patterns that were similar to the two
major dietary patterns seen in men (HPFU) and women (N HS) (Slattery et al., 1998;
Osler et al., 2001; Tseng & DeVellis, 2001). In a study in Denmark, similar dietary
patterns were observed and the prudent dietary pattern was associated with lower risk of
CVD and total mortality, but the association with the western dietary pattern was not
signiﬁcant. Tseng and DeVillis (2001) examined major dietary patterns in US. whites
and found similar patterns. In a study conducted in Germany (Schulze et al. , 2001), the

dietary patterns were somewhat different from US. patterns. Other investigators have

22

examined dietary intake and other lifestyle practices and found dietary patterns that are
associated with high-risk health behaviors (Randall et al. , 1991); Tseng and DeVellis,
2001). An analysis of dietary patterns in relation to disease risk factors, however, has yet
to be conducted in a nationally representative sample.

Dietary indices have also been used to predict disease risk or mortality. In elderly
Greek subjects (Trichopoulou et al. , 1995), a composite score that reﬂects frequent
consumption of traditional Mediterranean foods (i.e. vegetables, fruits, legumes, alcohol,
dairy products, cereal, and a high ratio of monounsaturated to saturated fats) was
associated with reduced all-cause mortality. Likewise, a dietary index based on the
World Health Organization guidelines for the prevention of chronic diseases has been
shown to be signiﬁcantly associated with reduced all-cause mortality in a cohort of
elderly men in Finland, Italy, and the Netherlands (Huijbregts et al. , 1997). The
“recommended food score,” which is the sum of the number of foods recommended by
current dietary guidelines was associated with all-cause mortality in the Breast Cancer
Detection Demonstration Project (Kant et al., 2000). The Healthy Eating Index (HEI),
which is a measure of adherence to the Dietary Guidelines for Americans, was examined
in association with major chronic disease separately in men (HPF U) and women (NHS)
and showed only a weak association with major chronic disease for men but no
signiﬁcant association with chronic disease for women (McCullough et al. , 2000a;
McCullough et al. , 2000b). An “alternative healthy eating index” was developed by
McCullough et al. (2002) and shown to be strongly associated with CVD mortality in

both men and women.

23

The synergy of nutrients and non-nutrients within foods has also been
investigated in experimental regimens. Several studies have used an intervention in which
a speciﬁc dietary pattern was implemented and then the effects were observed. The
Dietary Approaches to Stop Hypertension (DASH) (Appel et al. , 1997) and the Lyon
Diet Heart studies (de Lorgeril et al. , 1999) are the most well-known of these types of
interventions. Other controlled feeding trials with short-term (8 week) changes in dietary
patterns have also shown favorable effects on measures of serum antioxidant capacity and
lipid peroxidation (Miller et al. , 1998), and serum homocysteine concentrations (Appel et
al. , 2000).

The DASH study was the ﬁrst study to illustrate clearly that a dietary pattern that
was high in fruits and vegetables (8-10 servings/day), with low-fat dairy products (2-3
servings/day), and low in total fat (27% energy) and saturated fat (7% energy)
signiﬁcantly improved multiple risk factors for CVD (i.e., systolic and diastolic blood
pressure, total cholesterol, LDL-cholesterol) when compared to a control diet with
nutrient content similar to the average diet in the US. It should be noted, however, that
the DASH diet also reduced HDL-cholesterol (Appel et al. , 1997). In the DASH-Sodium
trial, the DASH diet in combination with sodium restriction resulted in even greater
reductions in blood pressure (Sacks et al. , 2001)). The DASH studies demonstrated that
dietary interventions could effectively combine food and nutrient advice to achieve
changes in dietary patterns and risks for CVD.

The Lyon Diet Heart Study was a randomized controlled trial designed to test a
Mediterranean dietary pattern in the prevention of CHD recurrence in individuals who

had already suffered a myocardial infarction. Subjects in the experimental group were

24

instructed to consume a dietary pattern with more bread, root vegetables, green
vegetables, and ﬁsh, fruit 2. l/day, margarine high in alpha-linolenic acid (supplied by the
study) to replace butter and cream, rapeseed and olive oil, and less red meat (replaced
with poultry). The trial was actually terminated early because of the remarkable beneﬁts
in the experimental group including a 70% reduction in all-cause mortality. Although
total blood cholesterol concentrations were signiﬁcantly and independently associated
with recurrence, there were no differences in total blood cholesterol, triglycerides, HDL-
cholesterol, LDL-cholesterol, or lipoprotein(a) concentrations between the control and
experimental groups at the beginning or the end of the study (de Lorgeril et al., 1999).
This indicates that there are important risk factors other than serum lipids mediating the
relationship between diet and CHD.

The study of dietary patterns is emerging as an informative and powerful means
to augment our understanding of the role of diet in chronic disease in free-living
populations. Dietary patterns associated with disease risk have been identiﬁed in speciﬁc
population groups; however, nothing has been reported describing dietary patterns using
national data in relation to risk factors for CVD.

C. Daily Eating Frequency
1. Description

The deﬁnition of an eating occasion varies widely in the literature depending on
the purpose of the investigation. Often, eating occasions are classiﬁed as either meals or
snacks, but the deﬁnition of what constitutes a meal or snack is not uniform. The
majority of investigators have deﬁned a meal versus a snack based on the criteria of time

of consumption and/or nutrient composition of the eating occasion. Meals are generally

25

described as one of the main eating occasions of the day, with breakfast occurring in the
morning, lunch occurring mid-day and dinner occurring in the evening. Snacks refer to
other eating occasions, and are usually smaller and less structured than meals. Other
researchers use caloric cut-points to differentiate between meals and snacks. The
deﬁnition is important because it may signiﬁcantly inﬂuence the outcome and
interpretation of studies in which various deﬁnitions have been used. The relationship
between energy intake and eating frequency has been shown to be dependent on the
deﬁnition of what constituted a meal (Gatenby, 1997). Self-reporting of eating occasions
reveals that subjects have conﬂicting views about the difference between a meal and a
snack. For example, subjects may report a beverage and a cookie eaten at 12:30 pm.
either as a snack or a meal depending on whether the subject uses time of day or food-
type as a means of classiﬁcation. In the present study, the self-reported deﬁnitions of
eating occasions collected with the 24-h dietary recall in NHANES III were used and
consisted of: breakfast/brunch; lunch; dinner; and snack/beverage (National Center for
Health Statistics, 1996b).
2. Studies of Daily Eating Frequency

Epidemiological studies have consistently shown an inverse association between
daily eating frequency and blood lipid concentrations (Edelstein et al. , 1992; Redondo et
al. , 1997; Titan et al., 2001). Many of these studies adjusted for energy and
macronutrient intakes but none further assessed the diet quality or micronutrient intakes
associated with increasing meal frequency even though snacks have been shown to

contribute signiﬁcant energy and nutrients to overall dietary intakes in selected

26

populations (Cross et al. , 1994; Cross et al. , 1995; Haveman-Nies et al., 1998; Siega-Riz
et al., 1998).

Most experimental studies in healthy adults comparing serum lipid concentrations
while consuming different meal ﬁ'equencies show reduced blood concentrations of total
and LDL-cholesterol (Mann, 1997). Several of these studies, however, include
comparisons of 3 large meals per day to 9 or more smaller meals per day (Arnold et al. ,
1994; Arnold et al. , 1997), which may not be possible to translate into practical
recommendations. Furthermore, these studies are all of limited duration and it is not
known whether physiological adaptation occurs over time.

There is less consistency in the outcomes of studies comparing effects of eating
frequency on other measures of CVD risk. Some investigators (Arnold et al., 1993;
Murphy et al. , 1996) have found decreased HDL-cholesterol concentrations with
increased meal frequency, while others have not (Jenkins et al., 1989). Similarly, some
studies have found differences in triacylglycerol and free fatty acid concentrations
(Jenkins et al. , 1992), while others have not (Wolever, 1991). In a case-control study of
291 smokers and 828 control subjects, increased meal frequency was associated with a
reduced risk of developing symptomatic peripheral atherosclerosis (Powell et al., 1999).

Evidence from short-term studies suggests that increased daily eating ﬁ'equency
may be beneﬁcial for blood glucose control in healthy individuals and those with type 2
diabetes by reducing peak insulin secretion while maintaining blood glucose
concentrations. This ﬁnding may help deﬁne a mechanism for the association of
decreased serum lipid concentrations with increased meal frequency. Insulin stimulates

hydroxymethyglutaryl-CoA (HMGCoA) reductase, which is the rate-limiting enzyme in

27

hepatic cholesterol synthesis. Additionally, removal of cholesterol via reverse cholesterol
transport occurs only in the postprandial phase and therefore, may be facilitated when
meal frequency is increased (Mann, 1997).

Epidemiological studies have shown an inverse relationship between meal
frequency and body weight (Fabry et al., 1966; Kant et al., 1995). However, a review of
related literature concluded that this ﬁnding is likely due to changes in dietary intake as a
consequence of weight gain or dietary under-reporting of overweight persons (Bellisle et
al. , 1997). The conclusion is based in part on the outcome of studies using the rigorous
methods of whole-body calorimetry and doubly labelled water, which found no
difference in 24h energy expenditure between nibbling and gorging. It appears that any
effects of daily eating frequency on the regulation of body weight are likely to be because
those who eat more often tend to consume fewer calories, but this has not been
conﬁrmed.

Because there are methodological challenges in describing meal and snack
patterns of populations (Gatenby, 1997), we have very little information regarding the
meal and snack patterns of US. adults (Oltersdorf et al. , 1999). Results ﬁ'om one 24-
hour recall and two l-day diet records from the 1987-1988 Nationwide Food
Consumption Survey (NF CS) in the US. showed a mean daily eating frequency of 3.47
(SD 0.90) with 90% of the sample eating between 2 and 4 times per day (Longnecker et
al. , 1997). The NF CS data were not grouped by type of eating occasion (e.g., breakfast,
lunch, dinner, snack/beverage) or sociodemographic subgroups, nor were nutrient intake
data available by daily eating frequency. Furthermore, the NF CS had a response rate of

31% and the results are not necessarily generalizable to all US. adults. In a population-

28

based sample in Rancho Bernardo, CA of 2,034 white men and women aged 50-89 y,
19% of the sample had a daily eating frequency of 2 4 (Edelstein et al., 1992). Very little
additional information is available regarding frequency of meals in free-living U.S.
populations.

However, it is important to describe daily eating frequency and to determine if it
is related to macronutrient or micronutrient intakes in order to ascertain whether or not
daily eating frequency is a marker of some other variable (e. g. nutrient composition) that
is affecting lipid proﬁles and/or glucose metabolism. Accurate descriptions of meal and
snack patterns (including daily eating frequency) in free-living populations will assist in
designing appropriate clinical experiments and aid in understanding both the metabolic
and behavioral effects of different meal and snack patterns (Gibney & Wolever, 1997).
D. Summary

For decades, scientists have tried to show the cause and effect relation between
dietary intake and CVD risk at individual and population levels. Dietary intake research
has mainly focused on speciﬁc nutrients (e.g. saturated fat) or other food components
(e. g. dietary cholesterol) in relation to certain risk factors (e.g. serum cholesterol
concentration). Foods and nutrients are not eaten individually, however, but within the
context of the total diet and at various frequencies throughout the day. Dietary pattern
research evaluates individual dietary intake at the macro level, accounting for natural and
common clusters of foods consumed. Research ﬁndings within speciﬁc sub-population
groups support the hypothesis that dietary patterns can be identiﬁed from food intake data
collected with F FQs and that selected biomarkers are associated with intake of certain

food components, dietary patterns, and daily eating frequencies.

29

The study of dietary patterns as a method of evaluating dietary intake is emerging
as an informative and powerful tool to better understand the role of diet in risk for CVD.
Dietary patterns are easier to implement (if conﬁrmed beneﬁcial) than individual foods or
nutrients and the use of dietary patterns in disease prevention efforts may assist in
decreasing CVD incidence through promotion of optimal dietary intake. Dietary patterns
associated with disease risk have been identiﬁed in speciﬁc population groups; however,

nothing has been reported describing dietary patterns using national data in relation to

risk factors for CVD in the US.

30

Chapter 3

DIETARY PATTERNS ASSOCIATED WITH RISK FACTORS FOR
CARDIOVASCULAR DISEASE IN HEALTHY U.S. ADULTS

A. Abstract

Certain nutrients are well established as dietary risk factors for cardiovascular
disease (CVD), but dietary patterns may be a better predictor of CVD risk. This study
tested the hypothesis that the complex dietary behaviors of US. adults can be grouped
into major dietary patterns that are related to risk factors for CVD. Using the food-
frequency questionnaire from the third National Health and Nutrition Examination
Survey (NHANES III), dietary patterns of healthy US. adults (220y; n=13,l30) were
identiﬁed by factor analysis. Log-transforrned biomarker data were associated with
major dietary patterns while controlling for confounding variables in regression analyses.
All statistical analyses accounted for the survey design and sample weights. Of six
dietary patterns identiﬁed, two patterns emerged most prominently and were
characterized by high intakes of 1) processed meats, eggs, red meats, and high-fat dairy
products (“Western”); and 2) green, leafy vegetables, salad dressings, tomatoes, other
vegetables, cruciferous vegetables, and tea (“American-healthy”). The Western pattern
was associated (p<0.05) positively with serum C-peptide, serum insulin, and glycosylated
hemoglobin and inversely with RBC folate concentrations after adjusting for confounding
variables. The American-healthy pattern had no linear relations with any of the
biomarkers examined. The successful identiﬁcation of common dietary patterns among
free-living persons is promising in characterizing high-risk groups at the US. population

level. The dietary patterns identiﬁed here are similar to those reported in other non-

31

representative samples and are associated with biomarkers of CVD risk which conﬁrms
that dietary pattern analysis can be an important method for assessing dietary intakes
when predicting CVD risk.
B. Introduction

Cardiovascular disease (CVD) is a major public health problem in the United
States (U .S.) and dietary intake is thought to exert a great inﬂuence on the risk for CVD.
The known risk factors for CVD, however, including speciﬁc dietary components (e. g.
excess saturated fat) explain only a portion of morbidity and mortality from CVD. Since
dietary intake is such a complex exposure variable, it is now recognized that we must
develop and reﬁne methods of assessing dietary intake that focus on the total diet and not
just individual dietary components (e. g. nutrients). As a result, the study of patterns of
nutrients, foods, and food-group intakes has begun to emerge in nutrition research.

Assessment of the total diet takes into account all nutrient interactions and allows
us to capture diet-disease or diet-biomarker relationships without knowing the exact
nutrient or food component involved (Jacobs & Murtaugh, 2000). Furthermore, many
nutrients are highly correlated within foods, making them difficult to examine separately.
Likewise, nutrient intakes may be confounded by entire dietary patterns (Hu, 2002). For
example, beta-carotene supplementation failed to prevent chronic disease in prospective
trials that were conducted after the observation that fruits and vegetables (i.e. beta-
carotene containing foods) were associated with lower incidence of disease (ATBC Study
Group, 1994; Hennekens et al. , 1996; Omenn et al., 1996). These results have been

partially attributed to a combination of the other beneﬁcial components of diets high in

32

fruits and vegetables, including folate, ﬁber, magnesium, potassium, ﬂavonoids, and
plant sterols (Bazzano et al., 2002).

Two large prospective studies have recently reported that dietary patterns deﬁned
by factor analysis using dietary data collected with a food-frequency questionnaire (FFQ)
are associated with various sociodemographic characteristics, coronary heart disease
(CHD) risk (nonfatal myocardial infarction and fatal CHD), and biomarkers of CVD risk
(Hu et al., 2000; Fang et al., 2001a; Fung et al., 2001b). Findings from these studies,
however, are limited to men and women who participated in the Health Professionals
Follow-up Study and the Nurses’ Health Study. Subjects were mostly white, over 40
years of age, and of higher socioeconomic status, and thus we do not know if and how
dietary patterns differ in the overall US. adult population inclusive of all
sociodemographic strata. Furthermore, if dietary patterns of the US. adult population are
identiﬁable, we do not know if they are associated with CVD risk, as the risk for CVD
differs by age, gender, and race.

While dietary patterns have been identiﬁed and associated with CVD risk in
speciﬁc sub-population groups, they need to be identiﬁed in nationally representative
populations in order to better understand the impact of diet on CVD risk at the population
level. Therefore, the purposes of this study were to test the hypotheses that 1) dietary
intake data of the healthy US. population can be systematically classiﬁed into distinct

dietary patterns and 2) risk factors for CVD are associated with speciﬁc dietary patterns.

33

C. Subjects and methods
Dataset

Subjects in this study were participants in the third National Health and Nutrition
Examination Survey (NHANES III), 1988-94. The National Center for Health Statistics
(N CHS) conducted the survey to obtain nationally representative information on the
health and nutritional status of the US. population. The NHANES III sample represents
the total civilian, non-institutionalized population, two months of age or over, in the 50
states and the District of Columbia of the United States. In NHANES 111, 39,695 persons
were originally sampled over the six years. Of those, 33,994 (86% of sampled subjects)
were interviewed in their homes and provided information for the Household Adult
Questionnaire and the Dietary Food Frequency Questionnaire (ages 17 years and over).
All interviewed persons were invited to the Mobile Examination Center, where blood and
urine specimens were obtained, and a number of tests and measurements were performed
including body measurements and blood pressure testing (National Center for Health
Statistics, 1994).
Analytic sample

All adults aged 2 20 y were eligible for inclusion (n=18,125) in this study. From
this eligible sample, subjects excluded were: pregnant (n=288) and lactating (n=95)
women, subjects on drugs for hyperlipidemia or unspeciﬁed heart disease (n=1,251),
subjects who were told by a physician that they have diabetes (n=1,498), and subjects
who reported changing their diet in the past year for any reason (n=3,227). The ﬁnal
analytic sample in this study consisted of 13,130 individuals, aged 20 years and older,

who completed both the home questionnaire and the medical examination.

34

Dietary Assessment methods

Within the framework of the Food Frequency Questionnaire (F FQ), subjects were
asked how often over the past month they had eaten speciﬁed food items. The foods
were listed in groups, targeting those high in vitamins A and C and calcium, but
representing all major food groups consumed by Americans (see Table 1 for food
groups). Interviewers also probed for other food and beverage items, which were
included in the present analyses. It is important to note that portion sizes were not
deﬁned, and responses represent “number of times” as determined by the respondent.
Foods were reported as number of times consumed per day, per week, per month, or
never. All frequency of consumption variables were standardized as “times per month”
using the conversion factors 4.3 weeks/month and 30.4 days/month rounded to the
nearest whole number. Usual food intake to establish dietary patterns was based on the
FFQ. The food groups from the FFQ were collapsed to 35 predeﬁned food groups (Table
1) in order to closely approximate food groups that have previously been used in the
literature. This was done in order to limit the subjectivity in deﬁning food groups and
allow us to replicate or refute existing literature.
Laboratory methods

Laboratory data in NHANES III were available from whole blood and sera. Prior
to the phlebotomy (venipuncture), a questionnaire was administered to determine an
examinee's eligibility for all phlebotomy procedures. It included questions to determine
if it was safe to perform the venipuncture, to document and determine fasting
compliance, and to aid in analyzing the results of the laboratory tests performed.

Examinees were instructed to fast for 10-16 hours prior to the morning examination or for

35

six hours before the afternoon or evening examination (National Center for Health
Statistics, 1996c). Detailed specimen collection and processing instructions are discussed
in the Manual for Medical Technicians. The analytical methods used by each of the
participating laboratories are described in the Laboratory Procedures Used for NHANES
III (National Center for Health Statistics, 1996a).

Data for all adults were available for serum total cholesterol, HDL-cholesterol,
and triacylglycerol concentrations. LDL-cholesterol was calculated for sample persons
who reported fasting for _>_ 9 hours and who had triacylglycerol concentrations S 400
mg/dL (4.52 mmol/L) by using the equation developed by F riedewald et a1 (Friedewald et
al. , 1972). In the present study, we included triacylglycerol data only in subjects who
reported fasting for 2 9 hours, which is consistent with the guidelines recommended by
the third National Cholesterol Education Program Adult Treatment Panel for lipoprotein
analysis (2001). RBC folate, serum C-peptide, serum insulin, and serum C-reactive
protein concentrations, as well as glycosylated hemoglobin (%) and blood pressure were
available for all adults. Serum homocysteine data were available only in phase 2 of the
survey (1991-1994) and plasma ﬁbrinogen data were available only in subjects aged 40
years and older.

Data analyses

Statistical software. Data preparation was performed using SAS (Inc., 2000)
software (version 8.1). Because NHANES III was conducted in a stratiﬁed, multi-stage
probability design, traditional methods of statistical analysis based on the assumption of a
simple random sample are not applicable. As recommended by the NCHS, SUDAAN

(Shah BV, 2001) software (version 8.0) was used to estimate descriptive and inferential

36

statistics of interest and the associated variances. Sample weighting was used in
NHANES III to account for the unequal probability of selection, non-coverage, and non-
response bias. Older persons (>60 y), African-Americans, and Mexican Americans were
over-sampled to allow for more precise estimates of health and nutritional characteristics
for these speciﬁc population subgroups. Appropriate sample weights were applied in all
statistical analyses to produce estimates of means and percentiles that can be generalized
to the healthy adult US. population.

Statistical methods. Because the distribution of the dietary data was extremely
non-normal, data values were truncated at four standard deviations above the mean and
then log transformed. Factor analysis (principal component) was used to derive food
patterns based on the frequency of consumption of each of the 35 food groups collapsed
from the FF Q. The analysis was conducted using the FACTOR PROCEDURE in SAS.
In order to account for the complex survey design of NHANES III, a correlation matrix
was created from the weighted data on the 35 food groups after pooling the variance
within the sample strata using PROC GLM in SAS. Next, a data step was performed to
read the correlation matrix directly into the FACTOR PROCEDURE in SAS. The factors
were orthogonally transformed using varimax rotation to achieve a structure with
independent (nonoverlapping) factors.

In determining the number of factors to retain, a step-wise process was utilized.
Eigenvalues >1.25 and the interpretability of the factors were used as the initial cut-point
to report dietary patterns, which resulted in six dietary patterns. Next, the Scree test was
considered which clearly identiﬁed two major dietary patterns, which were then used in

ﬁrrther analyses with CVD risk factors. The factors were labeled on the basis of

37

interpretation of the data as well as on prior literature (Yarnold et al. , 1995). A factor
score was created for each individual based on the monthly intake frequencies of the 35
food groups and the standardized scoring coefﬁcient of each food group for each factor.
Thus, each individual has a factor score for each factor that emerged from the data.
Percentage and standard error of means were calculated by the linearization
(Taylor series) variance estimation method for population parameters. Categorical
variable associations were assessed using a chi-square test. Ratio scale variables were
assessed using Wald F tests for determination of signiﬁcance between means of
biomarkers by quintile of dietary pattern scores, using the ﬁrst quintile as the reference
group. Linear regression analyses were conducted between biomarkers of CVD risk and
dietary pattern scores while controlling for confounding variables including: age, gender
(M/F), ethnicity (non-Hispanic White, non-Hispanic Black, Mexican-American),
smoking status (yes/no), alcohol intake (nondrinker—O drinks/day, light—>0 to V2
drink/day, moderate—V2 to <2 drinks/day, or heavy drinker—22 drinks/day),
vitamin/mineral supplement use (yes/no), BMI, physical activity (summation of the
frequency of multiple leisure-time activities multiplied by the respective estimated
oxygen consumption of each activity), and income (poverty income ratio calculated as
the ratio of family income to a Census Bureau-determined poverty threshold). Univariate
statistics of all dependent variables were assessed prior to regression analyses and log
transformed where extremely non-normal in order to more closely approximate a normal

distribution.

38

D. Results

Figure 1 depicts each of the six major and minor dietary patterns (i.e. factors) in a
pyramid shape with the foods that are highly correlated with the factor at the base of the
pyramid (factor loadings 2 0.60), the foods that are moderately correlated in the center of
the pyramid (factor loadings 0.40-0.59), and those that are correlated to a lesser extent
near the tip of the pyramid (factor loadings 0.20-0.39). Foods that were correlated to the
factor at a factor loading magnitude of <0.20 are not listed, while foods that are
negatively correlated with the factor (also at a magnitude of 2 0.20) are listed outside the
tip of the pyramid to represent their extreme infrequency of consumption in relation to
the factor. Table 2 reveals the complete factor-loading matrix for the 2 major and 4
minor dietary patterns. These six dietary patterns represent 37% of the explained
variance.

Further statistical analyses are only presented for the 2 major dietary patterns,
which were subjectively labeled as “Western,” and “American-healthy” and together
accounted for 20% of the explained variance. Subjects in the highest quintile of the
Western dietary pattern (Table 3) were more likely to be male, non-Hispanic Black or
Mexican-American, smokers, heavy drinkers, less educated, and less likely to take
vitamin/mineral supplements, while those in the highest quintile of the American-healthy
dietary pattern were more likely to be female, white, moderate drinkers, more educated,
and more likely to take vitamin/mineral supplements. Additionally, when compared to
subjects in the lowest dietary pattern quintile, those in the highest quintile of the Western

dietary pattern were younger, had a higher mean BMI, lower income, and reported less

39

physical activity, while those in the highest quintile of the American-healthy dietary
pattern were older, with higher income, and more physical activity.

In age-adjusted models, serum total cholesterol, sermn LDL-cholesterol, serum
triacylglycerol, serum homocysteine, glycosylated hemoglobin (%), and serum insulin
concentrations were higher, while serum HDL-cholesterol, serum C-peptide, and RBC
folate concentrations were signiﬁcantly lower in the higher quintiles (compared to the
ﬁrst quintile) of the Western dietary pattern (Table 4). Conversely, serum HDL-
cholesterol, and RBC folate concentrations were higher, while serum homocysteine and
serum insulin concentrations were lower in the higher quintiles (compared to the ﬁrst
quintile) of the American-healthy dietary pattern (Table 5).

Since the dietary pattern scores were related to sociodemographic and lifestyle
characteristics, further statistical analyses controlled for the effect of sex, ethnicity,
smoking status, alcohol intake, vitamin/mineral supplement use, BMI, physical activity,
and income. Table 6 shows signiﬁcant differences in serum LDL-cholesterol, serum
HDL-cholesterol, RBC folate, glycosylated hemoglobin (%), serum C-peptide and serum
insulin concentrations across quintiles of Western dietary pattern scores even after
controlling for confounding variables. To test for linear relationships, factor scores were
modeled as continuous variables and positive associations were found between the
Western dietary pattern and glycosylated hemoglobin (p<0.0001), serum C-peptide
(p<0.0001), and serum insulin concentrations (p<0.001). An inverse association was
seen between the Western dietary pattern and RBC folate concentrations (p=0.0001).
Table 7 shows signiﬁcant differences in serum total cholesterol and HDL-cholesterol

concentrations, and glycosylated hemoglobin (%) across quintiles of the American-

40

healthy dietary pattern, but none showed a signiﬁcant linear relationship (data not
shown). Additionally, systolic blood pressure, serum lipoprotein(a), plasma ﬁbrinogen,
and serum C-reactive protein were not found to be associated with the dietary patterns
(data not shown).
E. Discussion

The underlying assumption of statistical data reduction in regard to food intake is
that foods eaten together (i.e. within the total diet) can be characterized as part of a
dietary pattern that is more epidemiologically meaningful than its individual components
(Slattery et al. , 1998). The use of factor analysis to deﬁne dietary patterns, however, has
been criticized for its subjective nature with the concern that results are unable to be
replicated across populations or even within the same population (Martinez et al. , 1998).
Similar ﬁndings across studies would support the use of factor analysis in nutrition
epidemiology. Therefore, this research attempted to replicate dietary patterns that have
been reported in other epidemiological studies by using similar steps in the subjective
decision-making process while using data representative of the healthy U.S. adult
population. Remarkably similar results to those reported previously (Slattery et al., 1998;
Hu et al. , 2000; Fung et al., 2001b; Tseng & DeVellis, 2001) were found, given the fact
that the NHAN ES 111 sample is extremely diverse in sociodemographic and lifestyle
characteristics and previously reported data has been in age, gender, and race speciﬁc
populations. These results are in support of Tseng’s description of 2 fundamental U.S.
dietary patterns that she posits are a result of British culinary heritage (Western) and
nutrition science, industry, and government efforts to promote diets that prevent illness

(American-healthy) (Tseng, 1999).

41

The relation of the dietary patterns to sociodemographic and lifestyle
characteristics is in support of the theory that healthy food choices are a part of a larger
pattern of health-related characteristics and behaviors (Randall et al. , 1991). Our results
indicated associations between a Western dietary pattern and being nonwhite, male, less
educated, with lower income, smoking, decreased physical activity, and less
vitamin/mineral supplementation. Dietary patterns may either be independent from, or
interacting with other known risk factors for CVD, although many sociodemographic and
lifestyle characteristics were statistically controlled for in the multivariate analyses.
Additionally, the description of dietary patterns by using only food intake and not
additional dietary behaviors such as meal and snack patterns has been criticized (Tseng,
1999) and may indicate an area for further research.

It is not surprising that the Western dietary pattern is associated with lower RBC
folate concentrations, because of the lower consumption of folate containing foods in the
Western dietary pattern. It is somewhat surprising, although not inconsistent with prior
literature, that the Western dietary pattern is not associated with serum lipid
concentrations. An analysis of dietary intake in free-living persons using NHANES III
data revealed a positive relationship between dietary fat intake and serum total
cholesterol in men but not in women (Yang et al. , 2003). Results similar to those
reported in this study were found in the Health Professionals Follow-up Study in which a
Western dietary pattern characterized by high intakes of red meat, processed meat, high
fat dairy products, and reﬁned grains was associated with CHD mortality but was not
associated with serum lipid concentrations (Hu et al., 2000; Fung et al., 2001a). In

addition, the Lyon Diet Heart Study revealed a strong protective effect of a

42

Mediterranean dietary pattern in the prevention of CHD recurrence even though there
were no differences in total cholesterol, triacylglycerol, HDL-cholesterol, LDL-
cholesterol, or lipoprotein(a) concentrations between the control and experimental groups
at the beginning or the end of the study (de Lorgeril et al. , 1999). This indicates that
there are important risk factors other than serum lipids mediating the relationship
between diet and CHD. Perhaps the association shown here between markers of glucose
metabolism and the Western dietary pattern are indicative of the importance of the
relationship between dietary patterns and metabolic syndrome, which is in turn related to
CVD risk (Lakka et al. , 2002).

The American-healthy dietary pattern identiﬁed in this research is somewhat
similar to the prudent dietary pattern reported in both the Health Professionals Follow-up
Study and the Nurses’ Health Study, however, it does not include high intakes of
potentially beneﬁcial food groups such as ﬁsh, legumes, and low-fat dairy products. This
led us to label the pattern “American-healthy” implying an effort toward a healthy dietary
pattern, even if not exactly “prudent.” Because the dietary patterns described here are
based on actual food intake of free-living individuals and do not represent ideal dietary
patterns, strong associations with speciﬁc biomarkers were not expected. Additionally,
the two major dietary patterns described in this research represent only 20% of the
between-person variance in dietary patterns, thus further analyses on the minor dietary
patterns may reveal additional associations between dietary patterns and biomarkers of
CVD risk. Also described here are four minor dietary patterns and future research should
explore associations between minor dietary patterns and chronic disease risk in order to

more fully explore the diet-disease risk relationship at the population level.

43

Dietary pattern analysis is likely to be a more robust and longer-term measure of
dietary habits and more likely to be accurate than estimating intake of speciﬁc nutrient(s)
and their interactions. As with all nutrition research, however, dietary pattern analysis
can only be as good as the dietary assessment method it is based upon. Measurement
errors inherent in using F FQs for dietary assessment include possible under or over-
reporting of general food intake, or selective under or over-reporting of certain foods, or
both. Accordingly, the limitations of F FQs also apply to dietary pattern analyses that are
based on dietary information collected from a FFQ. However, since the FF Q is designed
to assess usual intake, and because of lower cost and relative ease of administration, most
large epidemiological studies have utilized FFQs.

As previously mentioned, subjectivity was limited by intentionally choosing
variables (food groups) to include in the factor analysis that have previously been used,
however, the exact re-creation of previously reported food groups was not possible. For
example, because of the food grouping system used in NHANES III (e.g. oatmeal and
other hot cereals), our best attempt to replicate food groupings in other studies led us to
include all hot cereals in our food group titled, “Whole grains.” The inclusion of
unrelated variables in a factor analysis can have the effect of redeﬁning factors because
of shared extraneous variance, whereas exclusion of variables in order to simplify the
factorial structure can lead to erroneous conclusions. Fine-tuning the food groups entered
into a factor analysis may improve associations between dietary patterns and markers for
disease risk.

A strength of this study is the statistical technique used to account for the complex

survey design of NHANES III by pooling the variance within the strata prior to creating a

44

correlation matrix based on weighted data that could then be used in a factor analysis
procedure using SAS software. Since the software programs generally available that will
account for the sample design of NHANES III in producing correct variance estimates

(e. g. SUDAAN) do not have the capability of conducting factor analyses this was
imperative in conducting an appropriate factor analysis. To our knowledge, the use of this
technique has not been reported elsewhere.

The use of multivariate data reduction techniques (e. g. factor analysis) to deﬁne
dietary patterns is emerging in nutritional epidemiology as an important method of
describing a snapshot of the entire diet that can then be associated with risk for chronic
disease. Dietary pattern analysis is an extension of the many years of research involving
speciﬁc dietary components and it will remain critically important to continue
investigating singular dietary components for many reasons, including the correct
interpretation of dietary pattern research. In the present study, no attempt has been made
to surmise cause and effect relationships both because of the cross-sectional design of
NHANES III and because certain dietary patterns may be part of a larger pattern of
healthy or unhealthy behaviors and thus may simply be a surrogate measure for other
variables. However, even after we statistically controlled for many potentially
confounding variables, dietary patterns alone were signiﬁcant predictors of biomarkers
for CVD risk. The dietary patterns identiﬁed here in this nationally representative
sample are associated with biomarkers of CVD risk. These data support the use of
dietary patterns in guiding public health recommendations for dietary prevention of

chronic disease.

45

Table 1. Food groupings used in the dietary pattern analysis

Food or food groups
Processed meats

Red meats

Organ meats

Fish and other seafood
Poultry

Eggs

Butter
Margarine
Low-fat dairy products

Food groups from the NHANES III FFQj

Bacon, sausage (chorizo) and luncheon meats such as hot dogs,
salami, and bologna

Beef, including hamburger, steaks, roast beef, and meatloaf; Pork
and ham, including roast pork, pork chops, and spare ribs

Liver and other organ meats such as heart, kidney, tongue, and tripe
(menudo)

Fish including ﬁllets, ﬁsh sticks, ﬁsh sandwiches, and tuna ﬁsh;
Shrimp, clams, oysters, crab, and lobster

Chicken, all types, including baked, fried, chicken nuggets, and
chicken salad; Turkey

Eggs including scrambled, fried, omelets, hard-boiled eggs, and egg
salad; Egg substitutes

Butter

Margarine

2%/low fat; 1%; Skim/nonfat; Buttermilk; Evaporated; and Other
milk; Chocolate milk and hot cocoa; Yogurt and frozen yogurt

High-fat dairy products Whole/regular milk; Half and half; Cheese, all types including

Liquor

Wine
Beer

Tea

Coffee
Fruits

Fruit juices

Cruciferous vegetables

American, Swiss, cheddar, and cottage cheese; Cheese dishes such
as macaroni and cheese, cheese nachos, cheese enchiladas, and
quesadillas; Ice cream, ice milk, and milkshakes

Hard liquor such as tequila, gin, vodka, scotch, rum, whiskey and
liqueurs, either alone or mixed

Wine, wine coolers, sangria, and champagne
Beer and lite beer; Non-alcoholic beer/wine
Regular tea with caffeine; Decaffeinated/herbal tea

Regular coffee with caffeine; Decaffeinated coffee and espresso
Citrus fruits including oranges, grapefruits, and tangerines; Melons
including cantaloupe, honeydew, and watermelon; Peaches
nectarines, apricots, guava, mango, and papaya; Other fruits such
as apples, bananas, pears, berries, cherries, grapes, plums, and
strawberries

Orange juice, grapefruit juice and tangerine juice; Other fruit juices
such as grape juice, apple juice, cranberry juice, and fruit nectars
Broccoli; Brussels sprouts and cauliﬂower; Cabbage, coleslaw, and
sauerkraut

Dark-yellow vegetables Carrots and vegetable mixtures containing carrots; Sweet potatoes,

Tomatoes

yams, and orange squash including acorn, butternut, hubbard, and
pumpkin

Tomatoes including fresh and stewed tomatoes, tomato juice, and
salsa

Green, leafy vegetables Spinach, greens, collards, and kale; Tossed salad

 

46

 

Table 1 (cont’d).

Food or food groups
Legumes

Other vegetables

Potatoes

Cold breakfast cereals

Whole grains

Reﬁned grains

Pizza

Snacks

Nuts
High-energy drinks

Low-energy drinks
Salad dressings
Soup

Sweets

Food groups from the NHANES III FFQ

Beans, lentils, chickpeas/garbanzos, including kidney, pinto,
refried, black, and baked beans; Soy products, including tofu, soy
milk, soy ice cream, soy cheese, soy hamburgers, soy hot dogs, soy
ﬂour, and textured vegetable protein

Hot red chili peppers; Peppers including green, red, and yellow
peppers; Any other vegetables such as green beans, corn, peas,
mushrooms, and zucchini

White potatoes, including baked, mashed, boiled, french-fries, and
potato salad

All-Bran, All-Bran Extra Fiber, 100% Bran, and Fiber One; Total,
Product 19, Most, and Just Right; All other cold cereals like corn
ﬂakes, Cheerios, Rice Krispies, and presweetened cereals
Cooked, hot cereals like oatmeal, cream of wheat, cream of rice,
and grits; Dark breads and rolls, including whole wheat, rye, and
pumpemickel

White bread, rolls, bagels, biscuits, English mufﬁns, and crackers;
Corn bread, com mufﬁns, and corn tortillas; Flour tortillas; Rice;
Spaghetti and pasta with tomato sauce

Pizza, calzone, and lasagna

Salted snacks such as potato chips, taco chips, com chips, and
salted pretzels and popcorn

Peanuts, peanut butter, other types of nuts, and seeds

Regular colas and sodas, not diet; Hi-C, Tang, Hawaiian Punch,
Kool-Aid, and other drinks with added vitamin C; Sports drinks,
fruit drinks (excluding juices but including tamarind drinks),
Popsicles

Diet colas, diet sodas, and diet drinks such as Crystal Light; Water,
including tap, mineral, spring, seltzer, soda

Oil and vinegar, mayonnaise and salad dressings such as Italian and
Thousand Island, including those added to salads and sandwiches
Stew or soup containing vegetables including minestrone, tomato,
and split pea

Cakes, cookies, brownies, pies, doughnuts, and pastries; Chocolate
candy and fudge

IFFQ=Food frequency questionnaire

47

Table 2. Factor-loading matrix for major and minor dietary patternsI in healthy US. adults2

 

 

 

 

 

 

Factor
Major Minor
1 2 3 4 5 6
Western American Califor— Breakfast South- Convenience

Food groups -healthy nian western

Processed meats 0.693 -0.02 0.02 005 -009 0.08
Red meats 0.64 0.07 0.07 -0.09 0.07 0.26
Organ meats 0.20 -0.21 0.41 -0.16 0.20 -0.27
Fish and other seafood 0.01 0.13 0.64 0.01 0.00 0.07
Poultry 0.09 0.1 l 0.55 0.08 0.00 0.16
Eggs 0.65 -0.07 0.12 0.07 0.07 -0.23
Butter 0.18 0.12 -0.07 0.02 -0.05 0.00
Margarine 0.23 0.19 -0.03 0.10 -0.06 -0.07
Low-fat dairy products -0.16 0.07 0.05 0.61 0.02 0.14
High-fat dairy products 0.42 0.10 -0.08 0.09 0.08 0.15
Liquor 0.06 0.04 0.02 0.00 -0.05 -0.03
Wine -0.15 0.13 0.13 0.07 -0.05 0.05
Beer 0.18 -0.08 -0.02 -0.09 0.13 0.14
Tea -0.10 0.41 0.04 -0.15 -0.17 0.05
Coffee 0.20 0.18 -0.20 -0.22 0.14 -0.38
Fruits 0.20 0.23 0.41 0.25 0.15 -0.02
Fruit juices 0.02 0.1 l 0.27 0.57 0.04 -0.05
Cruciferous vegetables -0.18 0.42 0.45 0.06 0.16 -0.22
Dark-yellow vegetables -0.13 0.37 0.36 0.21 0.25 -0.26
Tomatoes 0.03 0.51 0.04 0.10 0.38 -0.01
Green, leafy vegetables -0.07 0.67 0.30 0.1 l 0.06 -0.03
Legumes -0.02 0.05 0.02 0.06 0.70 -0.04
Other vegetables -0.01 0.47 0.17 0.05 0.43 0.02
Potatoes 0.35 0.34 -0.15 0.03 0.25 0.09
Cold breakfast cereals 0.07 0.02 -0.07 0.72 0.05 -0.03
Whole grains -0.10 0.28 0.21 0.34 -0.18 -0.37
Reﬁned grains 0.32 -0.02 0.04 -0.14 0.52 0.32
Pizza 0.07 0.05 0.06 0.08 0.01 0.62
Snacks 0.22 0.09 0.00 -0.1 l -0.05 0.49
Nuts -0.01 -0.01 0.09 0.14 0.17 0.00
High-energy drinks 0.29 -0.09 -0.06 -0. 10 0.09 0.37
Low-energy drinks -0.12 0.07 0.13 0.07 -0.07 0.14
Salad dressings 0.19 0.63 0.03 0.05 -0.09 0.08
Soup 0.01 0.07 0.34 0.06 0.44 -0.21
Sweets 0.22 0.07 -0.07 0.08 0.02 0.17
Variance explained (%) 11.06 8.85 5.20 4.68 3.86 3.63

 

lFactor retention criteria of scree plot (major) and eigenvalue cutoff of 1.25 (minor)
2Based on n=13,130 FFQ data; weighted N=128,068,510 (includes only cases with no
missing values on FFQ and who met other criteria as described in methods

3 Factor loadings represent magnitude and direction of association with factors (dietary
patterns) and can range from -1 .0 to +1.0

48

2:29:60

0:322: 5 32:23.: noun—2:88 combs voamgummn EoEEo>ow .825.“ 05 ‘3 “on 592 2.622. a 2:323.— mozem 2285
22302015. 823532: .«o 3.3 ESSA—5% $029 28... 8028 280:. .3 225:0” 02.35.? 72 32383 ”no?” .0: :o 389

 

 

 

 

 

 

 

 

 

 

 

a a a 2 8 2 2 2 2 an 22.x. 52K
cm 2 8 2 Na 2 a 2 8 mm 22.x. 58-2

285 2 2 2 on mm 88.9 mm a 2 2 2 22.x. 22V .228 .22.:
2 2 2 cm em 8 cm on 2 ﬂ oz

88.? mm S cm 2 2 885v 2 2 2 cm an we, a: .25 5:25
2 cm 8 2 3 am mm 8 2 2 28:
mm on em a 2 2 2 2 a 8 asses:
2 2 cm om E 2 cm cm 8 mm a»:

282. a 2 2 2 mm 885v 2 2 2 on 2.. 35682 295 3.282
2 cm 2 2 mm 2 2 2 a R exoemsoz

2 2d 2 2 2 cm 2 88.3 mm mm 2 2 2 exam was... maxoam
mm mm 8 2 2 2 2 cm 2 mm 2.2 n?
2 cm 2 cm mm mm on 2 2 a 2.2 3-22

88.? 2 2 2 mm x. 88.9 mm a 2 2 2 2.2 2: N .0585
a 8 cm 2 2 2 2 2.. mm 2 .22 A
a 2 2 om mm a S 2 2 2 E2

88.? 2 2 2 mm N... 88.? 8 cm 2 2 8 m2 2 v as... 82388
w 2 2 am 2.“ mm a 8 8 2 5.5.5:
a 2 2 mm mm mm 2 2 2 8 seem

Soodv a .m cm 2 8 88.9 2 8 cm 2 mm 223 2255
mm 2.. 2 2 cm 2 2 2 2 am sass".

38.9 2 2 2 a R 88.3 mm a 8 2 2 2a: 2.80
8 2 2 2 mm 2 2 2 om am :8
a a a 2 2 2 2 2 E 8 22-2.

885v 2 2 2 cm 8 88¢ a 8 cm 2 cm 2 23m 2.8» .32..

a mo 5 8 mo 5 1 no so 8 No .5

E022. 220% >5_8;-§otoﬁ< 8028 52v 8883

 

$233 .m.D .328: E 823... Eaten .2806 we 0.25:6 an motowauo 2.03.:— vae oEQEwOEovoBOm E mowﬂeooeoa coca—33m .m 033.

49

$-33 a: mmz<=z .N 085 E baa 35302» 829 28.22283 3828225
.m2 no 38.823 350$qu Sm «no? «E a N v2.22 on? 28.33 28 baa vows—oar 43w:— 8v w 3232:3280 OH .528

can .23 .2: a N @822 2.3 28.3% no baa 5:250 Eaaovotm 05 .3 2.22330“ _o>o_ 88283—10. cm; H :38 no KEV u
52: v0 29¢- u :88 no 3%.? u :38 NO 52. u :88 _O mwvd - 34”.. u 08% E023 .0526 .3 own—um ”$029 Boom E028
50% mo cannon 83 @0232: ca 383 wanna 3:85:me =n 28.3283 .3 85> 02m =N .2235 3:22.250 $3 an 52>:

 

 

 

 

 

 

 

 

 

 

 

8.2 in V 2.2 S... V 3.2 02. V 3.: 0.3 V 3.2. 3;. V

385 an «.8 mom ”.2. v.3 80.: 22,532 £35
and 2.3 V 2.2 N3 $3 and V :3 23 V :3 23 V

886 one “no 23 .2 23 5.: 2223502225
83 an V 2.3 26 33 2a an.“ 26 22w 22

885V .2 man can 26 Ed 2.: "@0322 8.22825
:8 Sn V3.2 «2.. :5... 2m :2. Sn 3:. gm

885v m2. 2m 8m 8w 2; 2%.: 22:05:35. 82
:2: :a V 83 2; 225 8.» 25 2..» $3 an.”

252 2; 3a 2; R.” 2..» 83 22.22.85 20:
32 82 V 32 E2 32 2.. 3.2 2._ :2 :2

:23 n2 :2 82 2S 2 .2 3% $222.85 8
8mm 22 V 23 2a 2% Rm 83 NS 33 8....

885 2.». v3 3m 3m 2 .m 3 _ .m 2255 0-43
82 2.2 V 22 R2 :2 E... :2 82 22: c2

58.? R2 8.. ”2 32 a2 2.3 _ 22255 0252
8mm 2% V 2% 26 Sam :6 83 2.“ 2mm 22

0223 2a 2.“ 2% em.“ 26 25.: 22°85 E.
3.3 E V 2.3 QR £3 QR 23m ”.2 :3 EN

885 EN N8 08 SN v.3 22.2 ANEEV Em

.G 23: m 3 v0 8 No :0 a $22.55

 

N233 .m.D 3:8: 5 3.58 E023 .0526 E883 no 225:: .3 82.; 28—22283 $.88 noumsmvcbwax .v 035.

50

«5.32 .E mmZ<IZ .N own—E E baa 3.5322» mos—g how—.8503
Banana—3&2 .8 3.58.3 8:35am .8 “no? a: a N “.23“ 2.3 88.33 .8 baa cog—use 439: 2:. w 3235588
OH 838 can 93 WE a N tough on? $3.33. .8 baa noun—50 03303:..— 05 3 toga—330“ .32 02.20.3110. 3..— n :88

m0 856 u :38 #0 ”mud n :85 m0 ”wad. n :38 NO 6:- u :38 _O x _.m - de- u 088 Egg baﬁvmo owns»— anAOV 28m
8033 E06 .3 235:? 8% 3:30? no woman wsuma 358$ch =a rev—382a B 85> 3?. :N .9635 oocoumcoo $3 H 583:

 

 

 

 

 

 

 

 

 

 

 

9.3 34 V 3.: ”.8 V 3.2 33 V 8.2 3... V 3% 3% V

~23 2% we. 3“ «.8 v. V m 83 _ hA._\_o_EV £35
:3 :3 V 32 as V 83 .3 V 33 one V end and V

an _ No and N3 m3 mg as $3 _ K535 commend
2m.“ 26 V 3% W: V 33 :6 V 33 can V $2 :6 V

2a .o a.“ can 2a 2.“ an 25 _ 53% am: 835830
3:. Sn V38 Nam V3; an V89. a: :23. v2 V

£85 3m 3m “am an Sm 2%.: 53.8.5 828 oma
Sod a; V 2? S.» V 2? S.» V Ed «3 V $3: 25 V

885 m2 2.” Ma.» 3.” $5 32 E‘SoEEV 3:
$2 8.. V32 m: V32 :._ V22 2.. V32 e: V

:35 2.. VS VS «2 NS 83 g335:2
:2 m; V :3 m; V 3?. m: V Sam :3 V :3. 3.». V

$35 85 m2 .3 ”mm ”mm 33 n£22; 0-43
32 c2 V 32 :1 V :2 am; V :2 on. V 3.2 ”S V

$85 E in. NE a? 3... Sq: 3.3.50-8:
:2 :a V :3 2w V :3 8.“ V $2 in V :2 :w V

.83 a; can 2.“ an :6 33 _ QOEEV B
:2 EN V 2.8 E V 3.8 ”.2 V 3.8 5 V 2.8 68 V

353 EN .3 EN SN S.” SE ANEEV Em

a... 235V a no 5 no 8 10 c .8555

 

N333 .m.D 3:8: 5 8.88 E33 E06 >5_no:-§otoﬁ< mo 0.353 as mos—g gov—.8829 232: voamzﬂcaéw< .m 3an

51

3.32 .E mmZ<IZ .N 323 5 baa
353225 82? 5315803 35855-3. no cantata 3:85:me 3.: “no: 5.2 a N nouns.“ on? $833 58 3:0 tog—2:5

AEwE o9. w 8325:3500 GP 85.55 can Ea E: m N coma on? 38355 =5 .25 58:33 25303:“: 05 .3 “58532505

35. ooaohouoxlon E".— n :52: m0 ”:6 .I. 588 v0 Sod- u :52: m0 3nd. n :88 NO “52- n 508 _0 ”win - and- n 0.88
8853 30% .3 owns. Aug—0V 0.83 5.533 50% .«o 05:30.. 83 33303 no 583 9558 oogmmcwa :5 £85.55

3 85> can an BEBE 8:03:80 $3 a SonN 0885 53:8 Roam—E AoEooSo v: =zm 22.3 5388 £358 :3

=2m ”own A2590 3.: Evan—qua 505.5588? ”€836 >52. 583.6 25.53.: cog—5.6 Em: 583.655 8.85 .2325
x2590 5385 waxes—m x5§«8<-§2§2 £22m u_:5%_:-=o= .823 25355-55 32:50 556 x8 :8 59:53:55

 

 

 

 

 

 

 

 

 

 

 

$55 555 V 8.: 35 V 2.: 5.2 V 2.55 5.3 V :55 5.3. V
.8555 5. _ m 2:. 5.55 5.8 «.3 $55 535553 5.35
255 :5 V :55 25 V 55 3.5 V :55 35 V 355 35 V
5:55 85 :5 555 $5 555 $55 ”geasﬁaad
$55 5.5 V 3&5 25 V 82 :5 A25 :5 V 83. 5:5
:85 .55 555 25 25 25 5:5 555:»: 5223829
355 55m V355. ~55 :5; 555 3; an V3; 555
55555 MR am an 8:. N55 855 5355553555 0mm
:35 555 V 85.5 5; V :55 555 $55 555 V 355 R5
$555 8.5 85 2.5 555 :5 :55 235585 5:
SN: 2.. V 35.: :5 V :3 3.: SN: 52 V SN: 3..
5.555 5: 52 52 mm: 52 85 23555.5 8
£5 53 V 255 85 V $3 53 :53 52 V 3&5 555
5955 t .m :5 555 555 :5 $55 m.3522; 0.53
32 5.2 V GS 55.: V 82 8.: GS 52 V 32 mm:
5255 mm: mm: 52 mm: 2.: 5555 35255 025:
3:5 85 V 8% 555 V :55 :5 $55 25 V :55 555
2555 :5 :5 25 «N5 25 $55 EOEEV 8
25m 555 V 355 EN V :55 5.2 V 355 E V 3.8 .55
5855 5.55 SN 08 :55 5.55 855 £595 Em
“a 253V m 3 5 8 no .5 a 33:55

 

53:65 .m.D :23: 5 8.83. F533 .05»? E0533 mo 235:6 3 mos—g 35.8803 N508 .voamzncméamrazzsz 6 035,—.

52

3-32 .5 mmz<zz .N 05.5.... a. .28 35:58.25 52...; 58.55803 vogoumqgao. :o 550850.552. 8:85.95
.8 555.... 5... a N 55855.. 2.3 5.8.35.5 58 baa 3.5.05 5 42w... 85 w 5.50.53.53.58 O... 85.85 E5.— 55 55... a N 55255.. 9.3 5.83:5 co
bee .8233 3530.35. 2.. .3 3.5.3.50 .03. oocobﬁmlo... 5.5.. u :52: 50 ”who u :52: .50 .36 u .552: no .556- u 552.. NO
no. .7 u 550:. .0 .— _.m .. and- n 285 Ecuaa5§o€ .50 owcam ”$0.9 9.85 5355.. .0506 50 0:53:05 8% 58.303 .8 .5053 95.558
8:85.95 :55 ”58.55503 .3 85> cum :5 .563... 8:25....50 $5.5 5 55225 088:. 53.85 30.55.... ”50835.8 5. .35 22.3 5.5088
5.0558 E3 .35 mama ”35590 om: Soﬁe—5.35 30585588..» ”CS—5.6 .98.. 58.5.6 355.60... 55.56 3w: 58.5.6555 855:.
.3815 ”50550.0 5355 wﬁxoﬁm ”5.58.50.55-50.on ion—m 05.555.55.58: .823 0.5.55.2-.55 5.0.5.50 55.6 585 no.5 5.853.555

 

 

 

 

 

 

 

 

 

 

 

.555 5.55 V .555 5.55 V .555 5.55 V .555 5.55 V .5..5 5.55 V

5555.5 .55 5.55 5.55 . 5.55 5.55 555.5 .355qu 5.555.
.555 55.5 V .555 55.5 V .555 55.5 V .555 55.5 V .555 55.5

5555.5 .55 .55 .55 55.5 .55 .555 5355555555556
.555 5.5 V .555 5.5 V .5..5 :5 .555 5.5 .555 555

5555.5 .55 5.5 5.5 5.5 5.5 2.5 5.5555: 53255855
.555 555 :85 555 :55 555 .5.5 555 .555 555

5555.5 555 555 555 .55 555 555.5 ..§o§Vo.5.o5 om...
.555 55.5 V .555 55.5 V .555 55.5 .555 55.5 .555 55.5

555.5 55.5 55.5 5.5 55.5 55.5 .555 5.3558555:
.55.. 5... V.55.. 5... :55. 5... .55.. 5... .55.. 5...

5555.5 5... 55.. .5. 55.. 5... 555.5 5.5.1.5585 o...
.555 5.5 V .555 55.5 V .555 :5 ..5.5 5.5 .555 5.5

5.55.5 55.5 5. .5 5.5 55.5 55.5 555.5 53555.5 9......
.55.. 55.. :55. .5. :55. 55.. .55.. 55.. .55.. 55..

5555.5 55.. 55.. 55.. 55.. 55.. 555.5 35255 0-5.55.
.555 5.5 V .555 :5 V .25 55.5 .555 5.5 .555 :5

5555.5 5.5 5.5 55.5 55.5 5.5 555.5 .5385 P5
.555 5.55 V .555 5.55 V .555 5.55 .555 5.55 .555 5.55

5555.5 5.55 .55 5.55 5.55 5.55 555.5 .5...th .25.

5.5 5.55.: .5 50 5o 50 5o .5 5 55.58.25.

 

555—.65 .m.D 55.50: E 5285 E0558 .0306 55.50:-..35855 .50 5:25.56 .3 52...; 58.55.55... 5:52: .35525365553252 .5. 03.5...

53

Figure l. Heirarchical depiction (within each pyramid) of the factor loadings associated
with each food group in the two major (A, B) and four minor (C, D, E, F) dietary patterns
detected in healthy US. adults using factor retention criteria of a scree plot (major) and
eigenvalue cutoff of 1.25 (minor). Food groups shown outside the tip of the pyramid
represent negative factor loadings within the absolute value range of 0.20-0.39.

54

 
 
 
   
    
 

" Organ meats

Whole grains
Potatoes Factor loadings

Dark-yellow vegetable 0.20-0.39

 

   

     
     

High-energy drinks Factor loadings

 

 

 

 

 

 

 

 

 

   

   
      
      
   

 
     
  

  
  
     
    

 

 

Potatoes Reﬁned grains 0.20-0.39 Cruciferous vegetables Tea 0 40 0 59
/ High-fat dairy products :040-059 Tomatoes Other vegetables . '
Eggs Red meats 20.60 20.60
Processed meats Green, leafy vegetables Salad dressings
A. “Western” dietary pattern (factor 1) B. “American-healthy” dietary pattern (factor 2)
— -
' Coffee " Coffee
Whole Dark-
grains yellow
Fruit juices veg.
Factor loadiggs
Green, leafy veg. . Fruits
Factor loadings
Soup Whole grains 0.20-0.39

  

 

Dark-yellow vegetables

0.20-0.39
, Fruit juices 0.40-0.59
Fruits Organ meats
g. 0.40-0.59

 

 

 

 

 

Poultry Cruciferous ve Low-fat dairy products
20.60
/ Fish and other seafood \20.60 Cold breakfast cereals
C. “Californian” dietary pattern (factor 3) D. “Breakfast” dietary pattern (factor 4)

  
 
 
 
  
  
 

Coffee
Whole grains
- Organ meats
Dark-yellow veg.
Cruciferous veg.
Eggs
Factor loadings Soup

       
    
   

Dark-yellow
vegetables

Potatoes

Factor loadings

 

 

Red meats 0.20-0.39

  

Tomatoes 0.20-0.39

Soup Other vegetables
Snacks High-energy drinks 0.40—0.59
Reﬁned grains 0.40-0.59
/ Legumes \ 20.60 / Pizza \zoxso

E. “Southwestern” dietary pattern (factor 5) F. “Convenience” dietary pattern (factor 6)

 

 

55

Chapter 4

MEAL AND SNACK PATTERNS ARE ASSOCIATED WITH DIETARY

INTAKE OF ENERGY AND NUTRIENTS IN US. ADULTS

A. Abstract

This research tested the hypothesis that meal and snack patterns are associated
with nutrient intakes in US. adults. Using the 24-h dietary recall from the third National
Health and Nutrition Examination Survey (NHANES III), meal and snack patterns of
US. adults (220 y; n=15,978) were described in relation to nutrient intakes while
controlling for confounding variables in regression analyses. All statistical analyses
accounted for the survey design and sample weights. Daily eating frequency was
positively related to carbohydrate (% energy), folic acid, vitamin C, calcium, magnesium,
iron, potassium, and ﬁber intakes and inversely related to protein (% energy), total fat (%
energy), cholesterol, and sodium intakes. Meal patterns were further categorized into the
5 most commonly reported meal and snack combinations by population percentages
including: Breakfast (B), lunch (L), dinner (D) and 2 2 Snacks (S) (31.6%); B, L, D, and
l S (15.4%); B, D and Z 2 S (13.1%); B, L, D (8.3%); and L, D and _>. 2 S (7.6%). The
groups reporting B, L, D, and 2 l S had the highest intakes of all micronutrients
examined except cholesterol, vitamin B6, and sodium. Breakfast skippers had the lowest
intakes of all micronutrients examined except sodium. Findings from this cross-sectional
survey suggest that meal and snack patterns may be markers for nutrient intakes and
therefore nutrient intakes should be considered in investigations of meal patterns and

health outcomes.

56

B. Introduction

Meal and snack patterns including the frequency of daily eating occasions are
suspected to affect health outcomes including cardiovascular disease (CVD) risk and
glucose intolerance (F abry et al., 1968; Jenkins et al., 1989; Favero et al., 1998).
Epidemiological studies have consistently shown more favorable lipid proﬁles with
increasing number of meals (Edelstein etal., 1992; Redondo et aI. , 1997; Titan et al. ,
2001). Experimental dietary studies show variable responses in lipid proﬁles and
carbohydrate tolerance to changes in meal and snack patterns in different study
populations (Jenkins, 1997; Mann, 1997). Many of these studies adjusted for energy and
macronutrient intakes but none further assessed the diet quality or micronutrient intakes
associated with increasing meal frequency even though snacks have been shown to
contribute signiﬁcant energy and nutrients to overall dietary intakes in selected
populations (Cross et al. , 1994; Cross et al., 1995; Haveman-Nies et al., 1998; Siega-Riz
et al. , 1998).

Because there are methodological challenges in describing meal and snack
patterns of populations (Gatenby, 1997), we have very little information regarding the
meal and snack patterns of US. adults (Longnecker et al., 1997; Oltersdorf et al., 1999).
However, it is important to determine if meal and snack patterns are related to
macronutrient or micronutrient intakes in order to ascertain whether or not the patterns
are markers of some other variable (e. g. nutrient composition) that is affecting lipid
proﬁles and/or glucose metabolism. Accurate descriptions of meal and snack patterns in
free-living populations will assist in designing appropriate clinical experiments and aid in

understanding both the metabolic and behavioral effects of different meal and snack

57

patterns (Gibney & Wolever, 1997). Therefore, the purpose of this research was to
provide descriptive information on the meal and snack patterns of US. adults and test the
hypothesis that meal and snack patterns are associated with energy and nutrient intakes in
US. adults.
C. Subjects and methods
Dataset

Subjects in this study were participants in the third National Health and Nutrition
Examination Survey (NHANES 111), 1988-94. The National Center for Health Statistics
(N CHS) conducted the survey to obtain nationally representative information on the
health and nutritional status of the US. population. The NHANES 111 sample represents
the total civilian, noninstitutionalized population, two months of age or over, in the 50
states and the District of Columbia of the United States. In NHANES 111, 39,695 persons
were originally sampled over the six years. Of those, 33,994 (86% of sampled subjects)
were interviewed in their homes and provided information for the household adult
questionnaire. All interviewed persons were invited to the mobile examination center,
where the 24-hour dietary recall was administered (National Center for Health Statistics,
1994)
Analytic sample

All adults aged 2 20 y were eligible for inclusion in this study (n=18,125). From
this eligible sample, subjects excluded from the main analyses were those whose 24-hour
dietary recall data were not reliable and complete (as coded by NCHS) (n=2,147).

Therefore, the ﬁnal analytic sample consisted of 15,978 adults aged 2 20 y who

58

completed both the home questionnaire and the 24-hour dietary recall (a component of
the medical examination).
Dietary Assessment

The 24-hour dietary recall records (from NHANES III) and the United States
Department of Agriculture (USDA) Survey Nutrient Database were used to determine
meal pattern variables (number and self-reported meal or snack/beverage name) and daily
intake of energy and selected nutrients (i.e. macronutrient distribution; dietary
cholesterol; vitamin B6, folic acid; vitamin C; calcium; magnesium; iron; sodium;
potassium; and dietary ﬁber). Respondents reported all foods and beverages consumed
except plain drinking water for the previous 24-hour time period (midnight to midnight).
Eating occasions were self-reported from a list of possible options and included:
breakfast; brunch; lunch; dinner; snack/beverage and, Spanish language equivalents.
Breakfast and brunch were collapsed into one group hereafter referred to as breakfast and
the Spanish language groups were combined with the English language equivalent
groups. Subjects were divided into categories of daily eating frequency (1-2; 3; 4; 5; 2 6)
based on the distribution of data obtained and prior literature. Categories based on self-
reported meal and/or snack/beverage intakes were also identiﬁed based on the population
distribution and nutrient intakes were compared among groups.
Data Analyses

Statistical software. Data preparation was performed using SAS soﬁware (version
8.1). Because NHANES III was conducted in a stratiﬁed, multi-stage probability design,
traditional methods of statistical analysis based on the assumption of a simple random

sample are not applicable. As recommended by the NCHS, SUDAAN (Shah BV, 2001)

59

software (version 8.0) was used to estimate descriptive and inferential statistics of interest
and the associated variances. Sample weighting was used in NHANES III to account for
the unequal probability of selection, non-coverage, and non-response bias. Older persons
(<60 y), African-Americans, and Mexican Americans were over-sampled to allow for
more precise estimates of health and nutritional characteristics for these speciﬁc
population subgroups. Appropriate sample weights were applied in all statistical analyses
to produce estimates of means and percentiles that can be generalized to the healthy adult
US. population.

Statistical methods. Percentage and standard error of means were calculated by

 

the linearization (Taylor series) variance estimation method for population parameters.
Categorical variable associations were assessed using a chi-square test. Linear regression
analyses were conducted between nutrients and both daily eating frequency and meal
pattern groups, while controlling for confounding variables including: age, gender (M/F),
ethnicity (non-Hispanic White, non-Hispanic Black, Mexican-American), smoking status
(yes/no), alcohol intake (nondrinker—O drinks/day, light—>0 to V2 drink/day,
moderate—V2 to <2 drinks/day, or heavy drinker—22 drinks/day), vitamin/mineral
supplement use (yes/no), BMI, physical activity (summation of the frequency of multiple
leisure-time activities multiplied by the respective estimated oxygen consumption of each
activity), income (poverty income ratio calculated as the ratio of family income to a
Census Bureau-determined poverty threshold), and energy intakes. Ratio scale variables
were assessed using Wald F tests for determination of signiﬁcance between means of
nutrient intakes by daily eating frequency or meal pattern groups. The fewest eating

occasions per day (1-2) was used as the reference group in the daily eating frequency

60

analyses, while breakfast, lunch, and dinner only (B, L, D only) was used as the reference
group in the meal and or snack/beverage pattern analyses.
D. Results

On average, subjects reported a daily eating frequency of 4.90 (SE 0.04) with a
range of 1-18 (median=4.18; mode=4). Daily eating frequency was categorized into 5
groups (1 -2; 3; 4; 5; 6+) based on the distribution of the data and prior literature. More
frequent eaters were more likely to be middle-aged (40-59 y), white, smokers, heavy
drinkers, vitamin/mineral supplement users, with higher income and education levels than
less frequent eaters (Table 1). Because daily eating frequency was associated with many
sociodemographic and lifestyle characteristics, further statistical analyses controlled for
the effect of age, sex, ethnicity, smoking status, alcohol intake, vitamin/mineral
supplement use, BMI, physical activity, income, and energy intakes. Aﬁer controlling
for these confounding variables, more frequent eaters had higher intakes of carbohydrate
(% energy), folic acid, vitamin C, calcium, magnesium, iron, potassium, and dietary ﬁber
and lower intakes of dietary fat (% energy), protein (% energy), cholesterol, and sodium
than less frequent eaters (Table 2).

When subjects were categorized in terms of self-reported speciﬁc meal (breakfast,
lunch, dinner) and snack/beverage consumption, the prevalence of reported meal skipping
was highest for lunch (26.1%) and lowest for dinner (10.4%), while 17.7% of the
population reported skipping breakfast. The majority of subjects reported consuming at
least two snacks (62.3%), while 25.2% of the population reported consuming one snack

and 12.5% reported consuming no snacks. Twenty-three different meal/snack

61

combinations were reported, but ﬁve of these described 75.9% of the population and thus
were the only ﬁve considered in further analyses regarding nutrient intake.

The most common meal pattern was reported by 31.6% of the population and
consisted of breakfast, lunch, dinner and at least 2 snacks (B, L, D + 2 2 S). Subjects
who reported consuming this meal pattern were more likely to be female, middle-aged
(40-59 y), white, nonsmokers, moderate drinkers, vitamin/mineral supplement users with
higher education and income levels and moderate activity levels. Also of note is that
within the non-Hispanic Black and Mexican-American subpopulation groups, over 40%
reported consuming a meal pattern type other than the ﬁve most commonly reported meal
patterns (Table 3). Subgroup analyses revealed a wide distribution of meal pattern types
within each race/ethnicity category, rather than a few prominent meal patterns in each
(data not shown).

After controlling for confounding variables in the multivariate analyses, those
reporting no snacks consumed the least amount of energy and carbohydrate (% energy)
and the highest amount of protein (% energy) and total fat (% energy). Those consuming
B, L, D + 2 2 S had the highest energy and carbohydrate (% energy) and lowest total fat
(% energy) intakes. Breakfast skippers had the lowest intakes of all micronutrients
examined except sodium. The groups reporting breakfast, lunch, dinner and at least one
snack had the highest intakes of all micronutrients examined except cholesterol, vitamin
B6 and sodium (Table 4).

E. Discussion
The results of this study provide descriptive information regarding meal and

snack patterns of US. adults. We report 30% of US. adults aged 20 y and older

62

consumed _>. 6 daily meals and snacks based on self-reported designation of eating
occasions from one 24-h dietary recall. By comparison, results from one 24-hour recall
and two l-d diet records from the 1987—1988 Nationwide Food Consumption Survey
(NFCS) in the US. showed a mean daily eating frequency of 3.47 (SD 0.90) with 90% of
the sample eating between 2 and 4 times per day (Longnecker et al. , 1997). The higher
number of meals and snacks in the NHANES III data reported here is likely due in part to
the fact that the classiﬁcation scheme in the NFCS for eating occasions was not based on
self-report but rather the researcher considered everything eaten within a one-hour time
period to be part of the same eating occasion. The NF CS data was not broken down by
type of eating occasion (e. g., breakfast, lunch, dinner, snack/beverage) or
sociodemographic subgroups, nor were nutrient data available by daily eating frequency.
Furthermore, the NFCS had a response rate of 31% and the results are not necessarily
generalizable to all US. adults.

In a population-based sample in Rancho Bernardo, CA of 2,034 white men and
women aged 50-89 y, 19% of the sample had a daily eating frequency of 2 4 (Edelstein et
al. , 1992). Similarly, in the Norfolk cohort of the European prospective investigation
into cancer (EPIC-Norfolk), involving over 14,000 men and women aged 45-75 y;
approximately 10% of the sample had a daily eating frequency of 2 6 (Titan et al. , 2001).
In both of these studies, daily eating frequency was assessed by a single question asking
participants to estimate the number of daily meals and snacks they consume each day.
Again, the discrepancy between the NHANES 111 results reported here may be due in part
to differences in methods for deriving daily eating frequency and also due in part to the

differences in geographic location and age of study populations. In our study, 22.8% of

63

participants aged 2 60 y had a daily eating frequency 2 6, which is similar to results
reported by Edelstein et al. (1992).

The deﬁnition of an eating occasion varies widely in the literature depending on
the purpose of the investigation. Often, eating occasions are classiﬁed as either meals or
snacks, but the deﬁnition of what constitutes a meal or snack is not uniform. Before the
effects of different food consumption patterns on health outcomes can be ﬁilly explored,
there is a need to develop appropriate methods for identifying and assessing eating
patterns within and between populations (Oltersdorf et al. , 1999). In the present study,
the self-reported deﬁnitions of eating occasions collected with the 24-h dietary recall in
NHANES III were used, however, perception of meals versus snacks may be different in
different populations, which could explain differences between studies.

This study also answers important questions regarding the relationship of
differing meal and snack patterns to macronutrient and rrricronutrient intakes. Because
one-day dietary recall information is insufﬁcient to represent usual dietary intake, we
were careful to never compare individual intake data to other individual characteristics.
We examined whether daily eating frequency is related to nutrient intake within diets, not
within individuals. While others have reported macronutrient intakes by daily eating
frequency, we are not aware of other studies that report micronutrient intakes by daily
eating frequency.

In the study by Edelstein et al. (1992), total fat (% energy), dietary cholesterol and
dietary ﬁber were higher with increasing daily eating frequency. In the EPIC-Norfolk
study, with increased daily eating frequency higher fat (% energy) and carbohydrate (%

energy) and lower protein (% energy) intakes were seen in men, while no difference in fat

64

(% energy) but higher carbohydrate (% energy) and lower protein (% energy) intakes
were seen in women. Similarly, our study shows lower protein (% energy) and higher
carbohydrate (% energy) intakes with increasing daily eating frequency, however, our
results indicate lower fat intake (% energy).

The issue of diet composition by daily eating frequency is very important in
studying the effects of daily eating frequency on health outcomes. Most experimental
studies in healthy adults comparing serum lipid concentrations while consuming different
meal frequencies show reduced concentrations of total and LDL-cholesterol (Mann,
1997). In one study in which subjects were identiﬁed as men whose usual eating habits
included either six meals per day or three meals per day, and were then instructed to
convert to the alternate eating pattern for 3 weeks duration, increasing meal frequency
was accompanied by a signiﬁcant reduction in total and LDL-cholesterol. Interestingly,
however, changes in nutrient intake also occurred in the group who decreased their meal
frequency. Speciﬁcally, even though subjects were instructed to change their eating
frequency but not to change the composition of their diet, nutrient analysis of food
records indicated an increase in protein, fat, saturated fat and alcohol (all as % energy),
which may explain the changes in lipid concentrations (McGrath & Gibney, 1994).

Finally, this study is an important ﬁrst step in describing meal patterns consumed
by US. adults and understanding the nutritional importance of speciﬁc meal patterns. In
particular, breakfast skippers (L, D + Z 2 S) had the lowest intakes of all nutrients except
sodium indicating that the breakfast meal provides signiﬁcant daily nutrients. This was
true regardless of daily eating frequency as evidenced by the fact that all ﬁve of the most

commonly reported meal patterns included a daily eating frequency of at least three, but

65

the breakfast skippers actually reported four or more eating occasions, whereas one of the
meal patterns was comprised of only three daily eating occasions (B, L, D). This
indicates that meal patterns may be better described by speciﬁc meal type rather than
simple daily eating frequency. In conclusion, ﬁndings from this cross-sectional survey
suggest that meal and snack patterns may be markers for macronutrient and micronutrient
intakes and therefore diet quality and/or complete nutrient intakes should be considered

in investigations of meal patterns and health outcomes.

66

Table 1. Sociodemographic and lifestyle characteristics by daily eating frequency in

 

 

 

 

US. adults"2
Daily eating frequency
All 1-2 3 4 5 2 6 P3
Population (%) 4.2 16.5 25.0 24.3 30.0
Gender
Male 47.4 4.8 16.1 24.7 23.5 30.9 0.0958
Female 52.6 3.7 16.9 25.4 24.8 29.2
Age Group
20-39 y 46.3 5.8 16.4 24.7 23.3 29.8 <0.0001
40-59 y 31.3 2.7 13.7 21.6 26.5 35.5
60+ y 22.5 3.2 20.5 30.6 22.9 22.8
Ethnicity
White 82.7 3.0 14.2 24.0 25.1 33.7 <0.0001
Black 11.8 10.8 24.9 27.5 19.0 17.9
Mex-Amer 5.5 8.7 26.9 28.6 19.7 16.1
Education Level
<12 yrs 24.7 6.9 22.5 29.4 21.1 20.2 <0.0001
12 yrs 33.8 4.3 16.2 25.1 24.3 30.1
>12 yrs 41.5 2.6 13.1 22.4 26.1 35.8
Income Level
2 1.85 PIR“ 30.0 7.4 21.7 28.9 19.6 22.4 <0.0001
l.86-3.5 PIR 34.0 3.6 14.3 24.5 24.9 32.7
>3.5 PIR 36.1 2.0 13.3 22.1 27.2 35.4
Smoking status
Smoker 28.3 5.2 15.2 22.0 23 .7 33 .9 0.0001
Nonsmoker 71.7 3.9 17.0 26.3 24.4 28.5
Alcohol intake
Nondrinker 44.5 4.9 18.8 28.2 23.2 25.0 <0.0001
Light drinker 37.0 3.3 15.9 23.7 24.7 32.4
Moderate 9.3 4.8 12.8 19.9 25.9 36.5
Heavy drinker 9.3 3.8 11.9 20.5 25.5 38.4
Vit/MinSupplement
Yes 42.3 2.7 14.4 24.0 26.4 32.5 <0.0001
No 57.7 5.4 18.0 25.8 22.6 28.2
Activity Level5
<33rd percentile 33.5 3.6 17.3 24.8 23.8 30.5 0.0973
33-66th percentile 33.1 4.5 14.5 23.9 25.0 32.1
>66th percentile 33.4 3.2 15.3 24.2 25.8 31.5

 

In=15,978; N=17l,457,892 2Population percentage; rows may not add up to 100% because of
rounding 3 Chi square test of independence 4PIR=Poverty Income Ratio; represents a poverty
index set by the federal government sEstimated oxygen consumption measured in metabolic

equivalents

67

355 3.55 3088 ~ 80503 5 3%: 333.9
E. .8 82.3. nggﬁuzm 3&2"... mm a =32N 8.55 3.58 Ea vase 63:8 .8353 aim awn x885
om: Eon—2&3 Ens—SEES? ”too—=36 h>aoa coo—cit 8839: £8.56 Em: coo—5.6.85 06—85 6:02“ R2590
maﬁa was—25. “Agotogéaoio—Z Jon—m omgnmmméoc .823 ﬁgaﬂméocv .3258 55.6 x8 com voumﬁw<_

 

 

 

 

 

885 ma 3.: no «a: no «a: 3 3.2 2. «n: @363 535
885v an «3cm mm «3% «a “Sam 2 3an an “EN 358.538
:86 R 3%... 4m «Ron 2 38m 2. «86m 2: «SR @5858
:85 3 3.2 3 3.2 to «02 no «02 3. an: 3.58:
885v m .32 m «N: m «8... m 3% m .38 356236.“:
385 S in» 2 «8a 2 £3 2 .13 mm a”: 358328
~33 mm «2: mm “2.2: n... 3.2: mm «was 3 «So $505635
:53 4 «8m m «as a “8m 6 «can a 4me @2528 35
at; 85 33 85 «a: mod 33 85 ”$3 85 «a»... 3.58 5:55
885 m .28 5 ﬁg a 3% a «:m : «Sm 353.8835
885v .8 3.2 no «3.». no «2% 8 «3m .2 «SM 3925 $8288
885v 2. «2m to ﬁne. no a”? .2 an? 3 «as. 565 $3522.30
.885 S «a: 3 «a: 3 an: 3 3.2 3 3.2 3.25 $525
885v mm «3.3 mm «22 mm «9% mm «22 8 $3; assigns

3m 3m 92 of 3. A5 8:230;
.C 235: cm m a. m N;

5:395 wage EEO

 

«3.36m .m.D E kocozvot wage >23 .3 moo—SE Betas: NcmoE .toamswcuouawazxsz .N 2%...

68

Table 132. 3Sociodemographic and lifestyle characteristics by selected meal patterns in US.
adults ’ '

 

B,L,D B,L,D B,L+ B,L,D L,D+
A11+228 +1S 22S only 228 Other 10‘

 

 

Population (%) 31.6 15.4 13.1 8.3 7.6 24.1
Gender
Male 47.4 29.4 13.7 14.3 7.8 8.5 26.3 <0.0001
Female 52.6 33.6 16.9 12.0 8.6 6.8 22.2
Age Group
20-39 y 46.3 28.5 13.0 12.5 6.9 9.9 29.3 <0.0001
40-59 y 31.3 37.2 14.2 14.0 6.9 8.0 19.7
60+ y 22.5 30.2 21.8 13.0 13.0 2.3 19.8
Ethnicity
White 82.7 35.7 16.0 13.4 7.9 8.0 19.0 <0.0001
Black 11.8 16.1 12.1 13.8 8.1 6.5 43.3
Mex-Amer 5.5 15.3 13.4 8.0 10.7 6.9 45.6
Education Level
<12 yrs 24.7 19.9 16.3 13.6 10.3 5.8 34.2 <0.0001
12 yrs 33.8 30.4 15.5 13.4 8.2 8.3 24.1
>12 yrs 41.5 39.5 14.8 12.6 7.0 8.1 18.1
Income Level
2 1.85 PIR5 30-0 20.3 15.3 12.4 9.3 7.1 35.8 <0.0001
1.86-3.5 PIR 34.0 33.8 14.9 14.2 7.7 8.2 21.3
>3.5 PIR 36.1 40.3 15.7 12.4 7.5 8.2 16.0
Smoking status
Smoker 28.3 27.8 10.3 15.8 6.3 10.8 28.9 <0.0001
Nonsmoker 71.7 33.1 17.4 12.0 9.0 6.3 22.2
Alcohol intake
Nondrinker 44.5 29.9 17.2 11.6 10.1 6.0 24.6 <0.0001
Light drinker 37.0 31.8 14.5 13.7 6.9 9.2 24.0
Moderate 93 37.6 10.5 15.9 7.3 8.2 20.6

Heavydrinker 9-3 33.4 11.7 15.2 6.1 8.2 25.3
Vit/MinSupplement

Yes 42-3 36.6 15.8 14.2 8.1 6.4 18.9 <0.0001
N0 57-7 27.9 15.1 12.3 8.4 8.5 27.9
Activity Level6

<33rdpercentile 33.5 31.2 14.7 12.8 9.1 8.4 23.9 0.0003
33-66thpercentile 33.1 34.6 14.5 12.1 6.9 8.6 23.2

>66thpercentile 33.4 33.3 15.8 15.3 7.9 6.4 21.4

In=15,978; Weighted N=17l,457,892 rPopulation percentage; rows may not add up to 100%
because of rounding 3B=breakfast; L=lunch; D=dinner; S=snackfbeverage 4Chi square test of
independence 5P1R=Poverty Income Ratio; represents a poverty index set by the federal
government 6Estimated oxygen consumption measured in metabolic equivalents

 

69

8.55 3.55 E086 _ 80508 5 team: mosatg =a 8m 88:33 9.8m 35.5%: me new: 9.8» baa Odd“ owEo>o§o~=mum

magnum— Eoﬁzua 33:35qu gﬁmmmdiuz mmcmﬁncn mm 8 .823 8.85 43.55 28 6:82: 5338 .833;

”gm ”own ”353$ 83580153 REEF—{£883 58—56 .93: £356 886.608 coo—5.6 Em: coo—£6.85 8.85 .200?
”35340 258.6. was—08m Agotogéuoioz Jon—m 25925-5: .82? o_§%_m-=o=v 5:253 536 xom tom Bu:€<~

 

 

 

 

_8c.cv :3 n no. one 0 4.: one 0 6.2 :3 n 3: 3 n 3: 23 one page
82:. 3o 6 3% 0:. 8 88 oz. 8 £8 9?. 0 2 3. EN n R: 35 season
885v moo n can 3:. n 3% tom 4 on? a? 0 $3 28 6 Son 3.5 eaoom
885v mg n n: 23 n 52 2. n 2 o}. n o.: 63 n w: 325 no:
585v a; 0 2m mod n a: Re 0 8... N3. 6 Sn Sn 4 am as; 538:on
885v mom on w; m: n so 92 n a.» mom 0 so mi 4 mg 3.5 826.8
38... mi. n. 93 a... n. N: o: n :_ 2:. n E N; n o: 95 o 5585
885v 63 n. ma 2: n E m :3 0 an 8.6 8 an of. 8 N2 825 28 6:8
885v :3 n m: 85 8 2d 85 n 8.. 35 0 SN 85 8 :2 3.5 8 £885
885v 8.” n on 2: 8 am So 0 am a.» n oom 31. n o8 REV 368220
28... as n 2% mod 8 mm 23 n 3” one 4 4.3. 2 n 4.2 $.28 § 5 58
585v 2 n 3:. as n at. 86 n we. ”2 n 2:. and n 9% $85 § 852380
38.? Rd n 4.2 3° n 4.2 o; n n: a; n S: :5 n «.2 $85 é 582a
885v 6.: n 32 Nov 8 88 :4 4 3R 6.8 n :2 EN 8 Sea .93ch broom

3 3 2: 4.2 o. _ M Q; ooasaoa
E263: mNNEJ booadd m~N+a.m mini; STA—41m

 

$3.3m .m.D E mEonmm :38 @8023. .3 moo—SE “combs: NSEE .38336533222 .v 635.

70

Chapter 5

THE RELATION OF MEAL AND SNACK FREQUENCY TO DIETARY
PATTERNS AND CARDIOVASCULAR DISEASE RISK FACTORS IN

HEALTHY U.S. ADULTS

A. Abstract

Dietary patterns based on food groupings have been associated with risk factors
for cardiovascular disease (CVD), but may not represent the entire relationship between
dietary intake and CVD risk. Daily eating frequency has also been associated with risk
factors for CVD in speciﬁc populations but may be a marker for nutrient intake, dietary
patterns, or both. Hence, this study examined the associations between daily eating
frequency and both risk factors for CVD and dietary patterns in healthy US. adults.
Subjects included were from the third National Health and Nutrition Examination Survey
(N HANES III) (_>.20y; n=10,427). Subjects excluded were: pregnant (n=288) and
lactating (n=95) women, individuals on drugs for hyperlipidemia or unspeciﬁed heart
disease (n=1,251), individuals told by a physician that they have diabetes (n=1,498),
those who reported changing their diet in the past year for any reason (n=3,227), those
with implausibly low (n=1,512) or high (n=6) reported energy intakes, and those without
complete and reliable dietary data as reported by the NCHS (n=2,147). The survey design
and sample weights were accounted for in all statistical analyses. Self-reported eating
occasions were categorized by daily frequency (1-2; 3; 4; 5; and 2 6). After controlling
for confounding variables in multivariate analyses, daily eating frequency was positively

associated with serum C-peptide concentrations (p<0.05). Daily eating frequency was

71

positively associated with an “American-healthy” dietary pattern but not related to a
“Western” dietary pattern. After further controlling for the American-healthy dietary
pattern, the relationship between daily eating frequency and serum C-peptide
concentrations was no longer statistically signiﬁcant (p=0.087l). In this cross-sectional
national probability sample, daily eating frequency was associated with serum C-peptide
concentrations but may only be a marker for a healthy dietary pattern.

B. Introduction

Dietary patterns based on food groupings have been associated with
cardiovascular disease (CVD) and CVD risk factors in speciﬁc sub-population groups
(Hu et a1. , 2000; Fung et al. , 2001a; Fung et al., 2001b) and in nationally representative
populations (Chapter 3). Dimensions of dietary patterns other than food intake
frequencies and their intercorrelations (e. g., daily eating frequency), however, may
provide additional information regarding the relationship between dietary intake and
CVD risk (Tseng, 1999). Daily eating frequency has also been associated with risk
factors for CVD in speciﬁc populations (Edelstein et al., 1992; Redondo et al., 1997;
Titan et al. , 2001) but may be a marker for nutrient intake, dietary patterns, or both.

The issue of diet composition by daily eating frequency is very important in
studying the effects of daily eating ﬁequency on health outcomes. In free-living
populations, nutrient intakes may vary depending on meal or snacking habits (e.g., eating
frequency, breakfast skipping). In fact, increased daily eating frequency was associated
with higher intakes of energy, fat, carbohydrate, and protein in the Norfolk population of
the European prospective investigation into cancer (EPIC-Norfolk), however, more

favorable lipid concentrations were still associated with increasing meal frequency after

72

adjusting for macronutrient intake and other confounding variables in multivariate
analyses. Many of these aforementioned studies adjusted for energy and macronutrient
intakes but none further assessed the diet quality or micronutrient intakes associated with
increasing meal frequency even though snacks have been shown to contribute signiﬁcant
energy and nutrients to overall dietary intakes in selected populations (Cross et al. , 1994;
Cross et a1. , 1995; Haveman-Nies et al. , 1998; Siega—Riz et al. , 1998).

In US. adults, more frequent eaters have been shown to be more likely to be
white, older, smokers, heavy drinkers, vitamin/mineral supplement users, with lower
BMI, higher education levels, and higher incomes than less frequent eaters. Aﬁer
controlling for confounding variables, daily eating frequency has also been shown to be
positively related to carbohydrate (% energy), folic acid, vitamin C, calcium, magnesium,
iron, potassium, and ﬁber intakes and inversely related to protein (% energy), total fat (%
energy), cholesterol, and sodium intakes (Chapter 4). Most observational and
experimental studies of daily eating frequency and health outcomes control for the effects
of macronutrient composition, but not the effects of dietary patterns, diet quality and/or
micronutrient composition. This study tested the hypothesis that daily eating frequency is
associated with biomarkers for CVD and speciﬁc dietary patterns.

C. Subjects and methods
Dataset

Subjects in this study were participants in the third National Health and Nutrition
Examination Survey (N HANES III), 1988-94. The National Center for Health Statistics
(NCHS) conducted the survey to obtain nationally representative information on the

health and nutritional status of the US. population. The NHANES 111 sample represents

73

the total civilian, noninstitutionalized population, two months of age or over, in the 50
states and the District of Columbia of the United States. In NHANES 111, 39,695 persons
were selected over the six years. Of those, 33,994 (86%) were interviewed in their homes
and provided information for the Household Adult Questionnaire and the Dietary Food
Frequency Questionnaire (ages 17 years and over). All interviewed persons were invited
to the Mobile Examination Center, where blood and urine specimens were obtained, and
a number of tests and measurements were performed including body measurements and
blood pressure testing (U .S. DHHS, 1996).
Analytic sample

Individuals included in this study were adults 2 20 yrs (n=10,427). Subjects
excluded were: pregnant (n=288) and lactating (n=95) women, individuals on drugs for
hyperlipidemia or unspeciﬁed heart disease (n=1,251), individuals told by a physician
that they have diabetes (n=1,498), those who reported changing their diet in the past year
for any reason (n=3,227), those with implausibly low (n=1,512) or high (n=6) reported
energy intakes, and those without complete and reliable dietary data as reported by the
NCHS (n=2,147). Cut-off values for implausibly low or high energy intakes were
determined based on the distribution of the data. Reported energy intakes of less than 15
kcals per kg body weight or higher than 10,000 kcals per day were considered
implausibly low and high intakes, respectively.
Dietary Assessment

Assessment of daily eating frequency has been described in detail elsewhere
(Chapter 4). Brieﬂy, the 24-hour dietary recall records (from NHANES III) were used to

determine daily eating frequency. Respondents reported all foods and beverages

74

consumed except plain drinking water for the previous 24-hour time period (midnight to
midnight). Subjects were divided into groups based on total daily eating occasions (1 -2;
3; 4; 5; .>. 6). The plausibility of energy intakes was determined based upon the
distribution of the data and prior literature. Subjects who consumed less than 15 kcals
per kg of body weight were thought to have implausibly low energy intakes, while those

reporting more than 10,000 kcals were considered to have implausibly high energy

intakes.
Dietary Pattern Assessment

Methods for derivation of major dietary patterns have been reported in detail
elsewhere (Chapter 3). Brieﬂy, dietary patterns of healthy US. adults (220y; n=13,130)
were described by factor analysis of the food-frequency questionnaire (F F Q) from the
third National Health and Nutrition Examination Survey (NHANES III) while accounting
for the survey design and sample weights. Major dietary patterns included high intakes
of: 1) processed meats, eggs, red meats, and high-fat dairy products (“Western”); and 2)
green, leafy vegetables, salad dressings, tomatoes, other vegetables, cruciferous
vegetables, and tea (“American-healthy”).

Factor analysis (principal component) was used to derive food patterns based on
35 predeﬁned food groups collapsed from the 62 food groups available from the F F Q.
The analysis was conducted using the FACTOR PROCEDURE in SAS. In order to
account for the complex survey design of NHANES III, a correlation matrix was created
from the weighted data on the 35 food groups after pooling the variance within the
sample strata using PROC GLM in SAS. Next, a data step was performed to read the

correlation matrix directly into the FACTOR PROCEDURE in SAS. The factors were

75

orthogonally transformed using varimax rotation to achieve a structure with independent
(nonoverlapping) factors. Each individual has a factor score for both the Western and the
American-healthy dietary patterns.
Blood Collection

Laboratory testing in NHANES 111 included determination of serum total
cholesterol, HDL-C, and triacylglycerol concentrations in all adults. LDL-C was
calculated for sample persons who reported fasting for 2 9 hours and who had
triacylglycerol concentrations 5 400 mg/dL (4.52 mmol/L) by using the equation
developed by Friedewald et al. In these analyses, triacylglycerol measurements were
included only in subjects who reported fasting for 2 9 hours. This is consistent with the
guidelines recommended by the third National Cholesterol Education Program Adult
Treatment Panel for lipoprotein analysis. Serum homocysteine, serum folate, red blood
cell folate, serum C-peptide, serum insulin, and serum C-reactive protein concentrations,
as well as glycosylated hemoglobin and blood pressure were measured in all adults.
Serum homocysteine was determined only in phase 2 of the survey (1991-1994).

Lipoprotein(a) and plasma ﬁbrinogen were measured only in subjects aged 40 years and

older.
Data Analyses

Statistical software. The survey was conducted in a stratiﬁed, multi-stage
probability design, thus traditional methods of statistical analysis based on the
assumption of a simple random sample are not applicable. As recommended by the
NCHS, SUDAAN (Shah, 1995) soﬁware (version 8.0) was used to estimate statistics of

interest and the associated variance. Appropriate sample weights were applied in all

76

analyses to produce estimates of means and percentiles that can be generalized to the US.
population. Data preparation was performed using SAS software (version 8.01).
Descriptive and inferential statistical analyses were performed using SUDAAN software
(version 8.0).

Statistical methods. Percentage and standard error of means were calculated by
the linearization (Taylor series) variance estimation method for population parameters.
Ratio scale variables were assessed using Wald F tests for determination of signiﬁcance
between means of biomarkers by daily eating frequency.

Linear regression analyses were conducted between biomarkers of CVD risk and
daily eating ﬁequency while controlling for confounding variables. Possible confounding
variables included based on the current literature were: age, gender (M/F), ethnicity (non-
Hispanic White, non-Hispanic Black, Mexican-American), smoking status (yes/no),
alcohol intake (nondrinker—O drinks/day, light—>0 to 1/2 drink/day, moderate—V2 to <2
drinks/day, or heavy drinker—22 drinks/day), vitamin/mineral supplement use (yes/no),
BMI, physical activity (summation of the frequency of multiple leisure-time activities
multiplied by the respective estimated oxygen consumption of each activity), income
poverty income ratio calculated as the ratio of family income to a Census Bureau-
determined poverty threshold), and energy intake. Linear regression analyses were also
conducted between factor scores of the Western and American-healthy dietary patterns
and daily eating frequency while controlling for age and energy intakes.

D. Results
After controlling for confounding variables in multivariate analyses, daily eating

frequency was positively associated with serum C-peptide concentrations (Table 1), but

77

not other biomarkers of CVD risk. Additionally, systolic blood pressure, serum
lipoprotein(a), plasma ﬁbrinogen, serum C-reactive protein, serum triacylglycerol, and
serum homocysteine were not found to be associated with daily eating frequency (data
not shown). Daily eating frequency was not related to a “Western” dietary pattern after
adjusting for age and energy intake (Table 2), however, daily eating frequency was
positively associated with an “American-healthy” dietary pattern (p<0.0001) (Table 3).
After further controlling for the American-healthy dietary pattern, the relationship
between daily eating frequency and serum C-peptide concentrations was no longer
statistically signiﬁcant (p=0.087l).
E. Discussion

Evidence from short-term studies suggests that increased daily eating frequency
may be beneﬁcial for blood glucose control in healthy individuals and those with type 2
diabetes by reducing insulin secretion while maintaining blood glucose concentrations
(Jenkins et al. , 1989; Jenkins et al., 1992). This ﬁnding may help deﬁne a mechanism for
the association of decreased serum lipid concentrations seen with increased meal
frequency in epidemiologic observations (Edelstein et al. , 1992; Redondo et al. , 1997;
Titan et a1. , 2001). Insulin stimulates hydroxymethyglutaryl-CoA (HMGCoA) reductase,
which is the rate-limiting enzyme in hepatic cholesterol synthesis. Additionally, removal
of cholesterol via reverse cholesterol transport occurs only in the postprandial phase and
therefore, may be facilitated when meal frequency is increased (Mann, 1997). However,
after controlling for confounding variables, our results indicate a signiﬁcant relationship
between daily eating frequency and serum C-peptide concentrations only. None of the

other biomarkers examined were signiﬁcantly related to daily eating frequency.

78

Furthermore, after we controlled for the effects of a healthy dietary pattern, the
relationship between daily eating frequency and serum C-peptide concentrations was no
longer statistically signiﬁcant. This implies that daily eating frequency may be a marker
of nutrient intakes (Chapter 4) and entire dietary patterns.

Epidemiological studies have shown an inverse relationship between meal
frequency and body weight. However, a review of related literature concluded that this
ﬁnding is likely due to changes in dietary intake as a consequence of weight gain or
dietary under-reporting of overweight persons (Bellisle et al., 1997). The conclusion is
based in part on the outcome of studies using the rigorous methods of whole-body
calorimetry and doubly labelled water, which found no difference in 24h energy
expenditure between nibbling and gorging. It appears that any effects of daily eating
frequency on the regulation of body weight are likely to be because those who eat more
oﬁen tend to consume fewer calories, but this has not been examined in depth.

Since these issues could also confound a relationship between daily eating
frequency and biomarkers for CVD risk, we addressed the issue of possible confounding
effects of post hoc dietary changes by excluding all individuals who reported changing
their diet in the past year for any reason. Additionally, we addressed the issue of
confounding by dietary-underreporting by excluding all individuals who reported
consuming implausibly low calorie intakes for their body weight. Further, we controlled
for characteristics that are associated with dietary under-reporting in our multivariate
analyses (i.e. gender and BMI) (Briefel et a1. , 1997). The present analyses only utilized

group data in any comparisons between the one day dietary recall data and individual

79

characteristics (i.e. sociodemographics, lifestyle characteristics, biomarkers, dietary
pattern scores).

In this cross-sectional national probability sample, daily eating frequency was
associated with serum C—peptide concentrations, which is not inconsistent with
experimental studies of eating frequency and markers of glucose metabolism (Jenkins,
1997). However, daily eating frequency was also associated with a dietary pattern
characterized by high intakes of green, leafy vegetables, salad dressings, tomatoes, other
vegetables, cruciferous vegetables, and tea and when analyses were further controlled for
this dietary pattern the relationship between daily eating frequency and serum C-peptide
concentrations was attenuated. Thus, daily eating frequency may be a marker for a

healthy dietary pattern.

80

02:? 0:382: vogoMmgmo. :o 608:8..0: 00:85:20 3.: «.00...n

:39: co... m 823::0050 0... 880m :0: :5 E: a N :82: 0:3 3003:... :0 b5 5:250 20303:... 05 .3 33:200.. 80: @0220?
:o :08: 352 00:85:20. :0 cow—3:85 .3 85> 020. :n .0225 00:02.28 $3 a 502.. 2:85 5:500 .0293: ”380050 2 95
20:3 E088 0.25:. :3 Em mow: “3500.0 00: 808039.... 5058\5583 ”GS—:5 .900: 503.6 2803:: may—:5. 3w: 5:58:05
0:55 3:020 .3585 0380 mica... ”A50E0E<-:02x02 £005 25:23-3: .233 2:805:05 .9258 552V .80 :8 Essa}...

 

 

 

 

 

 

 

 

 

 

3.2. 2% V 3:. 3:. V 3.3. 3:. V 8..“ .3. V 3% 3.. V

3%... 3.. 2.. 3:. 3:. as. «8.0 .3225: 5.3...
3.... 3.... .Vlmmn... a... V an... 3.... VINE: on... is... 2.... V

3...... ”I. 8... cm... mm... .2. .3... .goaaVuuzaoaé
£3 3.“ V 32 :0 V 8mm 2: V 8..“ 3m V 22 an V

3.3... 2.0 2.“ 2w 2.0 :6 «3.0 .30: 33.2826
33. a; V3... 23 V32. Em V2.9. .R V22. 8m V

22... an .2. ”an a... .2 R3 ..§o=aVo.~.o: um:
33 .3. V :3 :5. V 33 02 V :3 8.... V 22 a: V

83... on. a. .m 03 on... o. .m 33 .goeaV 9.5..
Gm. 3... V30. 2.. V2.2 3.. V82 8.. :3... .2 V

is... 2.. 2.. .2 8.. an. $0.0 goes: 0-4::
23 2.0 V 2.2 :2 V 3.3 80 V :2 a... V :3 a... V

30.... ”3 02 2: 8w 2: «8.0 £255 8
:3 :8 V 3.2 .2 V 8.8 .02 V 3...... 02 V 8.2 SN V

$3.... v.2 v.2 0.3 ”.8 EN 83 can: .2:

a 2...: a. 0 N n v m N.. a .2382:

 

«9:60 .m.D 2:00: 5 5:250: 3300 DE: A: 002.; 0:382: ~52: .c0um:_.:0-0§..0>:_:2 ._ 030...

81

Table 2. Regression model for Western dietary pattern by eating frequency

 

 

Category Variable 81d: SE P2
Eating frequency 1-2 0
3 0.01 0.06 0.8189
4 -0.02 0.05 0.7617
5 -0.03 0.06 0.641 1
2 6 -0.02 0.06 0.7633
Age (yrs) <-0.01 <0.01 0.2849
Calories (kcal) <00] <00] < 0.0001

 

lRegression coefficient

2P for comparison between group with B=0 and other groups within category

82

Table 3. Regression model for American-healthy dietary pattern by eating
ﬁequency

 

 

Category Variable B] :l: SE P2
Eating frequency 1-2 0
3 0.07 0.09 0.4065
4 0.27 0.08 0.0009
5 0.32 0.08 0.0001
2 6 0.43 0.08 <0.0001
Age (yrs) 0.01 <0.01 <0.0001
Calories (kcal) <00] <00] 0.0246

 

lRegression coefficient

2P for comparison between group with B=0 and other groups within category

83

Chapter 6

CONCLUSION

A. Implications

Major and minor dietary patterns of US. adults were identiﬁed using multivariate
data reduction techniques. After statistically controlling for many potentially
confounding variables, dietary patterns alone were signiﬁcant predictors of biomarkers
for CVD risk. The dietary patterns identiﬁed here are similar to those reported in other
nonrepresentative samples and are associated with biomarkers of CVD risk which
supports the use of dietary patterns in guiding public health recommendations for dietary
prevention of chronic disease.

These results are also in support of Tseng’s description of 2 fundamental U.S.
dietary patterns that she posits are a result of British culinary heritage (Western) and
nutrition science, industry, and government efforts to promote diets that prevent illness
(American-healthy) (Tseng & DeVellis, 2001). The concept of two ﬁmdarnental U.S.
dietary patterns is also in agreement with other research conducted in the Food and
Nutrition Database Research Center by Yang (unpublished data) on dietary intake of
Korean Americans. Korean Americans were shown to have dietary patterns classiﬁed as
“westernized” or “American-healthy,” both of which showed acculturation from the
traditional Korean diet, but in different qualitative directions. The extension of this
theory is‘that nutrition intervention efforts do make a difference. The challenge to
researchers is to identify core aspects of dietary patterns as targets and then practitioners

can develop menus around those targets.

84

While the Western dietary pattern was associated with biomarkers for CVD, the
American-healthy dietary pattern was not, which is not surprising because the dietary
patterns described here were based on actual intake and not ideal intake patterns. The
American-healthy diet described here is less than ideal in part because it does not contain
high amounts of foods thought to be beneﬁcial in the prevention of chronic disease (e. g.,
ﬁsh, legumes, dairy products). We labeled this diet “American-healthy,” because we
think it is a good representation of the foods that Americans perceived as healthy during
the time period of data collection (1988-1994). The perception of a healthy diet versus an
actual “ideal” diet may be related to the fact that 77% of Americans believe that there are
“good” and “bad” foods (Freeland-Graves & Nitzke, 2002), which may in turn stem from
“reductionism and the narrowing nutrition perspective” (Messina et al., 2001). This
implies that there is a disconnect between the messages nutrition educators are trying to
disperse and the messages that consumers are perceiving.

The results presented here also provide descriptive information regarding meal
and snack patterns (including daily eating frequency) of US. adults. After controlling for
confounding variables in multivariate analyses, daily eating frequency was signiﬁcantly
related to macronutrient and micronutrient intakes in healthy US. adults. Daily eating
frequency was positively associated with serum C-peptide concentrations and also with
an “American-healthy” dietary pattern but not related to a “Western” dietary pattern.
After further controlling for the American-healthy dietary pattern, the relationship
between daily eating frequency and serum C-peptide concentrations was no longer
statistically signiﬁcant, indicating that daily eating frequency may only be a marker for a

healthy dietary pattern. Diligence is required to capture accurate diet-disease

85

relationships without assuming a causal relationship between a dietary factor and a diet-
associated biomarker of risk.
B. Recommendations for future research

Further research should explore the minor dietary patterns identiﬁed here. Even
though the dietary patterns that emerge may be similar enough in all adult subgroups to
use one factor analysis, the relationship between the factor scores and the biomarkers of
CVD risk may be different. Exploration of dietary patterns in relation to biomarkers for
CVD risk in sub-population groups, including adolescents and children may provide new
insights into the relationships between dietary patterns and chronic disease risk.

Using current nutrition knowledge, an in-depth examination of each pattern in
terms of what it does or does not contain that may be harmful or beneﬁcial and how the
harmful and beneﬁcial components displace each other in diets of ﬁ'ee-living pe0ple is
warranted. The importance of not only the presence of harmful foods, but also the
absence of beneﬁcial foods in the Western dietary pattern is conﬁrmed by the low RBC
folate concentrations. Additional research on dietary patterns may explore the possibility
that animal food is displacing plant food in US. diets. Furthermore, any description of
an ideal dietary pattern using the inclusionary food approach must consider if all foods
will ﬁt into a diet without excessive energy content (Kris-Etherton et al. , 2002).

Additionally, we need to explore whether or not the dietary pattern itself is a
marker for an entire healthy lifestyle. Even though we statistically controlled for many
confounding variables, we still acknowledge the possibility in observational research that
other unknown confounders exist. As such, we may ﬁnd a psychological determinant

that predicts the healthy lifestyle behaviors, but ultimately, we need to establish a

86

deﬁnition of healthy dietary patterns so that we can educate people about the composition
and beneﬁts. Yet other research may seek a way to convert consumers of the traditional
dietary pattern to a healthier dietary pattern.

Even though it appears that eating frequency may be related to biomarkers for
CVD risk simply because it is a surrogate marker for nutrient intake and/or dietary
patterns, other factors such as who you eat with and where you eat should be explored in
order to fully characterize dietary patterns. Future research should include further
analysis of speciﬁc foods eaten at the same meal together (e. g. iron fortiﬁed cold

breakfast cereal with orange juice) and the effects on health consequences.

87

BIBLIOGRAPHY

AHA (2000) 2000 Heart and stroke statistical update. Dallas: American Heart
Association.

Appel LJ, Miller ER, 3rd, Jee SH, Stolzenberg-Solomon R, Lin PH, Erlinger T, Nadeau
MR & Selhub J (2000) Effect of dietary patterns on serum homocysteine: results
of a randomized, controlled feeding study. Circulation 102, 852-857.

Appel LJ, Moore TJ, Obarzanek E, Vollmer WM, Svetkey LP, Sacks FM, Bray GA,
Vogt TM, Cutler JA, Windhauser MM, Lin PH & Karanja N (1997) A clinical
trial of the effects of dietary patterns on blood pressure. DASH Collaborative
Research Group [see comments]. N Engl J Med 336, 1117-1124.

Arnold L, Ball M & Mann J (1994) Metabolic effects of alterations in meal frequency in
hypercholesterolaemic individuals. Atherosclerosis 108, 167-174.

Arnold L, Mann JI & Ball MJ (1997) Metabolic effects of alterations in meal frequency
in type 2 diabetes. Diabetes Care 20, 1651-1654.

Arnold LM, Ball MJ, Duncan AW & Mann J (1993) Effect of isoenergetic intake of three

or nine meals on plasma lipoproteins and glucose metabolism. Am J Clin Nutr 57,
446-45 1 .

ATBC Study Group T (1994) The effect of vitamin E and beta carotene on the incidence
of lung cancer and other cancers in male smokers. N Engl J Med 330, 1029-1035.

Bazzano LA, He J, Ogden LG, Loria CM, Vupputuri S, Myers L & Whelton PK (2002)
Fruit and vegetable intake and risk of cardiovascular disease in US adults: the ﬁrst
National Health and Nutrition Examination Survey Epidemiologic Follow-up
Study. Am J Clin Nutr 76, 93-99.

Bellisle F, McDevitt R & Prentice AM (1997) Meal frequency and energy balance. Br J
Nutr 77 Suppl 1, 857-70.

Block G, Norkus E, Hudes M, Mandel S & Helzlsouer K (2001) Which plasma
antioxidants are most related to fruit and vegetable consumption? Am J Epidemiol
154,1113-1118.

Briefel RR, Sempos CT, McDowell MA, Chien S & Alaimo K (1997) Dietary methods

research in the third National Health and Nutrition Examination Survey:
underreporting of energy intake. Am J Clin Nutr 65, 12038-12098.

88

Carleton RA, Dwyer J, Finberg L, Flora J, Goodman DS, Grundy SM, Havas S, Hunter
GT, Kritchevsky D, Lauer RM & et al. (1991) Report of the Expert Panel on
Population Strategies for Blood Cholesterol Reduction. A statement from the
National Cholesterol Education Program, National Heart, Lung, and Blood
Institute, National Institutes of Health. Circulation 83, 2154-2232.

Cohn JN, Bristow MR, Chien KR, Colucci WS, Frazier OH, Leinwand LA, Lorell BH,
Moss AJ, Sonnenblick EH, Walsh RA, Mockrin SC & Reinlib L (1997) Report of
the National Heart, Lung, and Blood Institute Special Emphasis Panel on Heart
Failure Research. Circulation 95, 766-770.

Cross AT, B'abicz D & Cushman LF (1994) Snacking patterns among 1,800 adults and
children. J Am Diet Assoc 94, 1398-1403.

Cross AT, Babicz D & Cushman LF (1995) Snacking habits of senior Americans. J Nutr
Elder 14, 27-3 8.

Darlington RB, Weinberg, S.L., Walberg, H.J. (1973) Canonical variate analysis and
related techniques. Review of Educational Research 43, 433-454.

de Lorgeril M, Salen P, Martin JL, Monjaud I, Delaye J & Mamelle N (1999)
Mediterranean diet, traditional risk factors, and the rate of cardiovascular
complications after myocardial infarction: ﬁnal report of the Lyon Diet Heart
Study. Circulation 99, 779-785.

Devine A, Wilson SG, Dick IM & Prince RL (2002) Effects of vitamin D metabolites on

intestinal calcium absorption and bone turnover in elderly women. Am J Clin Nutr
75, 283-288.

Edelstein SL, Barrett-Connor EL, Wingard DL & Cohn BA (1992) Increased meal
frequency associated with decreased cholesterol concentrations; Rancho
Bernardo, CA, 1984-1987. Am J Clin Nutr 55, 664-669.

F abry P, Fodor J, Hejl Z, Geizerova H & Balcarova O (1968) Meal frequency and
ischaemic heart-disease. Lancet 2, 190-191.

Fabry P, Hejda S, Cemy K, Osancova K & Pechar J (1966) Effect of meal frequency in

schoolchildren. Changes in weight-height proportion and skinfold thickness. Am J
Clin Nutr 18, 358-361.

F abry P & Teppennan J (1970) Meal frequency-~a possible factor in human pathology.
Am J Clin Nutr 23, 1059-1068.

F avero A, Franceschi S, La Vecchia C, Negri E, Conti E & Montella M (1998) Meal
frequency and coffee intake in colon cancer. Nutr Cancer 30, 182-185.

89

Flegal KM, Carroll MD, Ogden CL & Johnson CL (2002) Prevalence and trends in
obesity among US adults, 1999-2000. Jama 288, 1723-1727.

Freeland-Graves J & Nitzke S (2002) Position of the American Dietetic Association: total

diet approach to communicating food and nutrition information. J Am Diet Assoc
102, 100-108.

Friedewald WT, Levy RI & Fredrickson DS (1972) Estimation of the concentration of
low-density lipoprotein cholesterol in plasma, without use of the preparative
ultracentrifuge. Clin Chem 18, 499-502.

Fung TT, Rimm EB, Spiegelrnan D, Rifai N, Toﬂer GH, Willett WC & Hu FB (2001a)
Association between dietary patterns and plasma biomarkers of obesity and
cardiovascular disease risk. Am J Clin Nutr 73, 61-67.

Fung TT, Willett WC, Stampfer MJ, Manson JE & Hu FB (2001b) Dietary patterns and
the risk of coronary heart disease in women. Arch Intern Med 161, 1857-1862.

Gatenby SJ (1997) Eating frequency: methodological and dietary aspects. Br J Nutr 77
Suppl 1, S7-20.

Gibney MJ & Wolever TM (1997) Periodicity of eating and human health: present
perspective and future directions. Br J Nutr 77 Suppl 1, S3-5.

Gordon T (1988) The diet-heart idea. Outline of a history. Am J Epidemiol 127, 220-225.

Haveman-Nies A, de Groot LP & van Staveren WA (1998) Snack patterns of older
Europeans. J Am Diet Assoc 98, 1297-1302.

Hennekens CH, Buring JE, Manson JE, Stampfer M, Rosner B, Cook NR, Belanger C,
LaMotte F, Gaziano JM, Ridker PM, Willett W & Peto R (1996) Lack of effect of
long-term supplementation with beta carotene on the incidence of malignant
neoplasms and cardiovascular disease. N Engl J Med 334, 1145-1149.

Ho KK, Pinsky JL, Kannel WB & Levy D (1993) The epidemiology of heart failure: the
Framingham Study. J Am Coll Cardi0122, 6A-13A.

Hu PB (2002) Dietary pattern analysis: a new direction in nutritional epidemiology. Curr
Opin Lipidol 13, 3-9.

Hu FB, Rimm E, Smith-Warner SA, Feskanich D, Stampfer MJ, Ascherio A, Sampson L
& Willett WC (1999) Reproducibility and validity of dietary patterns assessed

with a food- frequency questionnaire [see comments]. Am J Clin Nutr 69, 243-
249.

90

 

Hu FB, Rimm EB, Stampfer MJ, Ascherio A, Spiegelman D & Willett WC (2000)

Prospective study of major dietary patterns and risk of coronary heart disease in
men. Am J Clin Nutr 72, 912-921.

Huijbregts P, Feskens E, Rasanen L, Fidanza F, Nissinen A, Menotti A & Kromhout D
(1997) Dietary pattern and 20 year mortality in elderly men in Finland, Italy, and
The Netherlands: longitudinal cohort study. ij 315, 13-17.

SAS Inc. SI (2000) The SAS system for Windows, version 8.1. Cary, NC: SAS Institute
Inc.
Jacobs DR, Jr. & Murtaugh MA (2000) It's more than an apple a day: an appropriately

processed, plant-centered dietary pattern may be good for your health. Am J Clin
Nutr 72, 899-900.

Jenkins DJ (1997) Carbohydrate tolerance and food frequency. Br J Nutr 77 Suppl 1,
871-81.

Jenkins DJ, Ocana A, Jenkins AL, Wolever TM, Vuksan V, Katzrnan L, Hollands M,
Greenberg G, Corey P, Patten R & et al. (1992) Metabolic advantages of
spreading the nutrient load: effects of increased meal frequency in non-insulin-
dependent diabetes. Am J Clin Nutr 55, 461-467.

Jenkins DJ, Wolever TM, Vuksan V, Brighenti F, Cunnane SC, Rao AV, Jenkins AL,
Buckley G, Patten R, Singer W & et al. (1989) Nibbling versus gorging:
metabolic advantages of increased meal frequency. N Engl J Med 321, 929-934.

Jones PJ (1997) Regulation of cholesterol biosynthesis by diet in humans. Am J Clin Nutr
66, 438-446.

Kant AK, Schatzkin A, Graubard BI & Ballard-Barbash R (1995) Frequency of eating
occasions and weight change in the NHANES I Epidemiologic Follow-up Study.
Int J Obes Relat Metab Disord 19, 468-474.

Kant AK, Schatzkin A, Graubard BI & Schairer C (2000) A prospective study of diet
quality and mortality in women. Jama 283, 2109-2115.

Keil JE, Sutherland SE, Hames CG, Lackland DT, Gazes PC, Knapp RG & Tyroler HA
(1995) Coronary disease mortality and risk factors in black and white men.
Results from the combined Charleston, SC, and Evans County, Georgia, heart
studies. Arch Intern Med 155, 1521-1527.

Kennedy ET, Ohls J, Carlson S & Fleming K (1995) The Healthy Eating Index: design
and applications. J Am Diet Assoc 95, 1103-1108.

Kim J -O, Mueller, Charles W. (1978) Factor Analysis: Statistical Methods and Practical
Issues, First ed. Iowa City: Sage Publications, Inc.

91

Krauss RM, Eckel RH, Howard B, Appel LJ, Daniels SR, Deckelbaum RJ, Erdman JW,
Jr., Kris-Etherton P, Goldberg IJ, Kotchen TA, Lichtenstein AH, Mitch WE,
Mullis R, Robinson K, Wylie-Rosett J, St Jeor S, Suttie J, Tribble DL & Bazzarre
TL (2000) AHA Dietary Guidelines: revision 2000: A statement for healthcare
professionals from the Nutrition Committee of the American Heart Association.
Circulation 102, 2284-2299.

Kris-Etherton PM, Etherton TD, Carlson J & Gardner C (2002) Recent discoveries in
inclusive food-based approaches and dietary patterns for reduction in risk for
cardiovascular disease. Curr Opin Lipidol 13, 397-407.

Lakka HM, Laaksonen DE, Lakka TA, Niskanen LK, Kumpusalo E, Tuomilehto J &
Salonen JT (2002) The Metabolic Syndrome and Total and Cardiovascular
Disease Mortality in Middle-aged Men. Jama 288, 2709-2716.

Lawley DNaM, A. E. (1963) Factor Analysis as a Statistical Method. London:
Butterworth & Co. Ltd.

Longnecker MP, Harper JM & Kim S (1997) Eating frequency in the Nationwide Food
Consumption Survey (U .S.A.), 1987-1988. Appetite 29, 55-59.

LRC-CPPT (1984a) The Lipid Research Clinics Coronary Primary Prevention Trial
results. 1. Reduction in incidence of coronary heart disease. Jama 251, 351-364.

LRC-CPPT (1984b) The Lipid Research Clinics Coronary Primary Prevention Trial
results. 11. The relationship of reduction in incidence of coronary heart disease to
cholesterol lowering. Jama 251, 365-374.

Mann J (1997) Meal frequency and plasma lipids and lipoproteins. Br J Nutr 77 Suppl 1,
$83-90.

Martinez ME, Marshall JR & Sechrest L (1998) Invited commentary: Factor analysis and
the search for objectivity. Am J Epidemiol 148, 17-19.

McCullough ML, Feskanich D, Rimm EB, Giovannucci EL, Ascherio A, Variyam JN,
Spiegehnan D, Stampfer MJ & Willett WC (2000a) Adherence to the Dietary

Guidelines for Americans and risk of major chronic disease in men. Am J Clin
Nutr 72, 1223-1231.

McCullough ML, Feskanich D, Stampfer MJ, Rosner BA, Hu F B, Hunter DJ, Variyam
JN, Colditz GA & Willett WC (2000b) Adherence to the Dietary Guidelines for
Americans and risk of major chronic disease in women. Am J Clin Nutr 72, 1214-
1222.

92

 

McGrath SA & Gibney MJ (1994) The effects of altered frequency of eating on plasma
lipids in free-living healthy males on normal self-selected diets. Eur J Clin Nutr
48, 402-407.

McKeown NM, Day NE, Welch AA, Runswick SA, Luben RN, Mulligan AA,
McTaggart A & Bingham SA (2001) Use of biological markers to validate self-
reported dietary intake in a random sample of the European Prospective
Investigation into Cancer United Kingdom Norfolk cohort. Am J Clin Nutr 74,
188-196.

Messina M, Lampe JW, Birt DF, Appel LJ, Pivonka E, Berry B & Jacobs DR, Jr. (2001)
Reductionism and the narrowing nutrition perspective: time for reevaluation and
emphasis on food synergy. J Am Diet Assoc 101, 1416-1419.

Millen BE, Quatromoni PA, Copenhafer DL, Demissie S, O'Horo CE & D'Agostino RB
(2001) Validation of a dietary pattern approach for evaluating nutritional risk: the
Framingham Nutrition Studies. J Am Diet Assoc 101, 187-194.

Miller ER, 3rd, Appel LJ & Risby TH (1998) Effect of dietary patterns on measures of
lipid peroxidation: results from a randomized clinical trial. Circulation 98, 2390-
2395.

Mitchell BD, Hazuda HP, Haffner SM, Patterson JK & Stern MP (1991) Myocardial
infarction in Mexican-Americans and non-Hispanic whites. The San Antonio
Heart Study. Circulation 83, 45-51.

Murphy MC, Chapman C, Lovegrove JA, Isherwood SG, Morgan LM, Wright JW &
Williams CM (1996) Meal frequency; does it determine postprandial lipaemia?
Eur J Clin Nutr 50, 491-497.

National Center for Health Statistics N (1994) Plan and operation of the third National
Health and Nutrition Examination Survey, [988-94. Washington, DC: US
Government Printing Office.

National Center for Health Statistics N (1996a) Third National Health and Nutrition
Examination Survey, 1 988-94, NHANES 111 Reference mauals and Reports (CD-
ROM. Hyattsville, MD: US. Department of Health and Human Services
(DHHS), Centers for Disease Control and Prevention (CDC).

National Center for Health Statistics N (1996b) Third National Health and Nutrition
Examination Survey, 1 988-1994, NHANES III Examination Data File (CD-ROM).
Hyattsville, MD: US. Department of Health and Human Services (DHHS),
Centers for Disease Control and Prevention (CDC).

National Center for Health Statistics N (1996c) Third National Health and Nutrition
Examination Survey, [988-1994, NHANES III Laboratory Data File (CD-ROM.

93

Hyattsville, MD: US. Department of Health and Human Services (DHHS),
Centers for Disease Control and Prevention (CDC).

NHLBI (2001) Executive Summary of The Third Report of The National Cholesterol
Education Program (NCEP) Expert Panel on Detection, Evaluation, And
Treatment of High Blood Cholesterol In Adults (Adult Treatment Panel III). Jama
285, 2486-2497.

Nube M, Kok FJ, Vandenbroucke JP, van der Heide-Wessel C & van der Heide RM

(1987) Scoring of prudent dietary habits and its relation to 25-year survival. J Am
Diet Assoc 87, 171-175.

Ogden CL, Flegal KM, Carroll MD & Johnson CL (2002) Prevalence and trends in

overweight among US children and adolescents, 1999-2000. Jama 288, 1728-
1732.

Oltersdorf U, Schlettwein-gsell D & Winkler G (1999) Assessing eating pattems-an
emerging research topic in nutritional sciences: introduction to the symposium.
Appetite 32, 1-7.

Omenn GS, Goodman GE, Thornquist MD, Balmes J, Cullen MR, Glass A, Keogh JP,
Meyskens FL, Valanis B, Williams JH, Barnhart S & Hammar S (1996) Effects of
a combination of beta carotene and vitamin A on lung cancer and cardiovascular
disease. N Engl J Med 334, 1 150-1 155.

Osler M, Heitrnann BL, Gerdes LU, Jorgensen LM & Schroll M (2001) Dietary patterns

and mortality in Danish men and women: a prospective observational study. Br J
Nutr 85, 219-225.

Phibbs B (1997) The human heart: a basic guide to heart disease, First ed. MV-NY:
Lippincott-Raven Publishers.

Powell JT, Franks PJ & Poulter NR (1999) Does nibbling or grazing protect the
peripheral arteries from atherosclerosis? J Cardiovasc Risk 6, 19-22.

Randall B, Marshall JR, Graham S & Brasure J (1991) High-risk health behaviors
associated with various dietary patterns. Nutr Cancer 16, 135-151.

Redondo MR, Ortega RM, Zamora MJ, Quintas ME, Lopez-Sobaler AM, Andres P &
Gaspar MJ (1997) Inﬂuence of the number of meals taken per day on
cardiovascular risk factors and the energy and nutrient intakes of a group of
elderly people. Int J Vitam Nutr Res 67, 176-182.

Refsum H, Ueland PM, Nygard O & Vollset SE (1998) Homocysteine and cardiovascular
disease. Annu Rev Med 49, 31-62.

94

Rosenman RH, Brand RJ, Sholtz RI & Friedman M (1976) Multivariate prediction of
coronary heart disease during 8.5 year follow-up in the Western Collaborative
Group Study. Am J Cardiol 37, 903-910.

Ross R (1999) Atherosclerosis is an inﬂammatory disease. Am Heart J 138, S419-420.

Sacks FM, Svetkey LP, Vollmer WM, Appel LJ, Bray GA, Harsha D, Obarzanek E,
Conlin PR, Miller ER, 3rd, Sirnons-Morton DG, Karanja N & Lin PH (2001)
Effects on blood pressure of reduced dietary sodium and the Dietary Approaches
to Stop Hypertension (DASH) diet. DASH-Sodium Collaborative Research
Group. N Engl J Med 344, 3-10.

Schroder H, Marrugat J, Elosua R & Covas MI (2002) Tobacco and alcohol
consumption: impact on other cardiovascular and cancer risk factors in a southern
European Mediterranean population. Br J Nutr 88, 273-281.

Schulze MB, Hoffrnann K, Kroke A & Boeing H (2001) Dietary patterns and their
association with food and nutrient intake in the European Prospective
Investigation into Cancer and Nutrition (EPIC)-Potsdam study. Br J Nutr 85, 363-
373.

Shah BV BB, Bieler GS (2001) SUDAAN user's manual, release 8. 0. Research Triangle
Park, NC: Research Triangle Institute.

Siega-Riz AM, Carson T & Popkin B (1998) Three squares or mostly snacks--what do

teens really eat? A sociodemographic study of meal patterns. J Adolesc Health 22,
29-36.

Slattery ML, Boucher KM, Caan BJ, Potter JD & Ma KN (1998) Eating patterns and risk
of colon cancer [see comments]. Am J Epidemiol 148, 4-16.

Subar AF, Thompson FE, Kipnis V, Midthune D, Hurwitz P, McNutt S, McIntosh A &
Rosenfeld S (2001) Comparative validation of the Block, Willett, and National
Cancer Institute food frequency questionnaires : the Eating at America's Table
Study. Am J Epidemiol 154, 1089-1099.

Titan SM, Bingham S, Welch A, Luben R, Oakes S, Day N & Khaw KT (2001)
Frequency of eating and concentrations of serum cholesterol in the Norfolk
population of the European prospective investigation into cancer (EPIC-Norfolk):
cross sectional study. ij 323, 1286-1288.

Trichopoulou A, Kouris-Blazos A, Wahlqvist ML, Gnardellis C, Lagiou P,

Polychronopoulos E, Vassilakou T, Lipworth L & Trichopoulos D (1995) Diet
and overall survival in elderly people. ij 311, 1457-1460.

95

Tseng M (1999) Validation of dietary patterns assessed with a food-frequency
questionnaire [letter; comment]. Am J Clin Nutr 70, 422.

Tseng M & DeVellis RF (2001) Fundamental dietary patterns and their correlates among
US whites. J Am Diet Assoc 101, 929-932.

Tucker KL, Dallal GE & Rush D (1992) Dietary patterns of elderly Boston-area residents
deﬁned by cluster analysis. J Am Diet Assoc 92, 1487-1491.

van Dam RM, Rimm EB, Willett WC, Stampfer MJ & Hu PB (2002) Dietary patterns
and risk for type 2 diabetes mellitus in US. men. Ann Intern Med 136, 201-209.

Willett W, Stampfer M, Chu NF, Spiegelman D, Holmes M & Rimm E (2001)
Assessment of questionnaire validity for measuring total fat intake using plasma
lipid levels as criteria. Am J Epidemiol 154, 1 107-1 1 12.

Wolever TM (1991) Diet and blood lipid levels: effect of "nibbling". Cmaj 144, 729.

Yang EJ, Chung HK, Kim WY, Kerver JM & Song WC (2003) Carbohydrate Intake Is
Associated with Diet Quality and Risk Factors for Cardiovascular Disease in US.
Adults: NHANES 111. J Am Coll Nutr 22, 71-79.

Yarnold PR, Soltysik RC, McCormick WC, Burns R, Lin EH, Bush T & Martin GJ

(1995) Application of multivariable optimal discriminant analysis in general
internal medicine. J Gen Intern Med 10, 601-606.

96

    

   

IIIIIIIIIIIIIIIIIIIIIIIIIIIIIII

will.willlimitljljll

  

     

31930 690