THE ASSOCIATION BETWEEN PREGNANCY- RELATED WEIGHT AND OFFSPRING
BMI AT CHILD AGE 3 TO 6 YEARS: RESULTS FROM A MID-MICHIGAN COHORT
By
Taylor Seaton

A THESIS
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
Epidemiology- Master of Science
2017

ABSTRACT
THE ASSOCIATION BETWEEN PREGNANCY- RELATED WEIGHT AND OFFSPRING
BMI AT CHILD AGE 3 TO 6 YEARS: RESULTS FROM A MID-MICHIGAN COHORT
By
Taylor Seaton
To provide a broader perspective on the relationship between maternal and child obesity,
associations between several components of pregnancy-related weight and offspring body mass
index (BMI) were examined. The Archive for Child Health (ARCH) enrolled and interviewed
women at first prenatal visit in three clinics in mid-Michigan. A subset of mothers with
singleton children participated in an age 3-6 y follow-up visit (mean=4.8 y) where
anthropometric data was collected on both mother and child. Data collection included interview
at enrollment (pre pregnancy BMI), birth certificate review (gestational weight gain), and
measured height, weight, and waist circumference of mothers and children at follow-up. Multiple
linear regressions were used to examine the association between maternal weight and waist
circumference and offspring BMI. Maternal body composition at follow-up was also tested as a
mediator. Characteristics of the study sample (n=114) included: 54% White, 13% Black, 18%
Hispanic, and 15% other race/ethnicity; 82% had household incomes < $50,000/year; 84%
reported no history of smoking. After adjusting for race and smoking history, only pre pregnancy
BMI (β=0.84, 95% CI 0.1, 1.6) and maternal follow-up BMI (β =0.69, 95% CI 0.03, 1.3) were
significantly associated with childhood BMI, with no evidence of mediation as tested. Our
findings add to recent literature by revealing associations between a measure of maternal
postpartum body composition (i.e. follow-up BMI) and offspring BMI. These associations may
represent the influence of lifestyle and other environmental factors.

ACKNOWLEDGEMENTS
Completing this thesis would not have been possible without the help of several dedicated
people. I would like to thank my advisor and thesis committee chair, Dr. Jean Kerver, for her
guidance and support throughout this entire process, as well as taking the time to talk with me
every week. I would like to thank Dr. Zhehui Luo for her time in guiding me through the
statistical analysis and providing me the means to understand mediation analysis. To my
committee member Dr. Nigel Paneth, I am very grateful for his expertise and valuable insight.
Without him this thesis would not have been possible. I would also like to thank Dr. Shobha
Mehta for taking the time to be a part of this committee as it surely was not easy to call in from
out of state. Her experience and resources were immensely helpful when developing the
framework for this thesis. Lastly, I would like to thank my loved ones for their continuous
support. To my mom and dad, I would not be where I am today without their unwavering
encouragement. Thank you.

iii

TABLE OF CONTENTS

LIST OF TABLES ......................................................................................................................... vi
LIST OF FIGURES ..................................................................................................................... viii
KEY TO ABBREVIATIONS ........................................................................................................ ix
CHAPTER 1: INTRODUCTION ................................................................................................... 1
CHAPTER 2: BACKGROUND ..................................................................................................... 3
2.1 Childhood Obesity ................................................................................................................ 3
2.2 Maternal and Childhood Body Composition ........................................................................ 4
2.2.1 Pre Pregnancy BMI ........................................................................................................ 4
2.2.2 Gestational Weight Gain ................................................................................................ 7
2.2.3 Post Pregnancy Maternal Body Composition ................................................................ 9
2.2.4 Breastfeeding and Childhood Obesity ......................................................................... 11
2.3 Conceptual Framework ....................................................................................................... 12
2.4 Specific Aims ...................................................................................................................... 12
2.4.1 Aim 1: .......................................................................................................................... 13
2.4.2 Aim 2: .......................................................................................................................... 13
CHAPTER 3: METHODS ............................................................................................................ 14
3.1 Study Population ................................................................................................................. 14
3.2 Pre Pregnancy BMI ............................................................................................................. 14
3.3 Child BMI ........................................................................................................................... 15
3.4 Gestational Weight Gain ..................................................................................................... 15
3.5 Follow-up Body Composition............................................................................................. 16
3.6 Other Variables ................................................................................................................... 16
3.7 Analytic Plan ....................................................................................................................... 17
CHAPTER 4: RESULTS .............................................................................................................. 20
4.1 Demographic Variable Description .................................................................................... 20
4.2 Maternal Weight Description .............................................................................................. 21
4.3 Childhood BMI Description ............................................................................................... 21
4.4 Pre Pregnancy BMI Model ................................................................................................. 22
4.5 Follow-up Maternal Body Composition Models ................................................................ 22
4.6 Mediation Analysis ............................................................................................................. 23
CHAPTER 5: DISCUSSION ........................................................................................................ 26
5.1 Interpretations ..................................................................................................................... 26
5.2 Strengths ............................................................................................................................. 29
5.3 Limitations .......................................................................................................................... 29
iv

CHAPTER 6: CONCLUSION ..................................................................................................... 31
APPENDICES .............................................................................................................................. 33
APPENDIX A: Tables .............................................................................................................. 34
APPENDIX B: Figures ............................................................................................................. 62
REFERENCES ............................................................................................................................. 67

v

LIST OF TABLES

Table 1: Recommendations for Total Weight Gain During Pregnancy by Pre Pregnancy BMI . 35
Table 2: Explanation of Terminology Used for Mediation Analysis ........................................... 36
Table 3: Comparison of Mother and Child Characteristics Between the ARCH Child
Development Cohort and Analytic Sample ..................................................................... 37
Table 4: Maternal and Child Sociodemographic Characteristics by Pre Pregnancy BMI Category
......................................................................................................................................... 40
Table 5: Maternal and Child Sociodemographic and Perinatal Characteristics by Childhood BMI
Percentile Category ........................................................................................................ 42
Table 6: Association between Pre Pregnancy BMI and Childhood BMI Percentile ................... 44
Table 7: Association between Gestational Weight Gain and Childhood BMI Percentile ........... 45
Table 8: Association between Follow-up Maternal BMI and Childhood BMI Percentile .......... 46
Table 9: Association between Follow-up Waist Circumference and Childhood BMI Percentile 47
Table 10: Association between Postpartum Weight Difference and Childhood BMI Percentile 48
Table 11: Association between Post Delivery Weight Change and Childhood BMI Percentile 49
Table 12: Effects of Pre Pregnancy BMI on Childhood BMI Percentile: Influence of Follow-up
Maternal BMI ............................................................................................................... 50
Table 13: Effect of Pre Pregnancy BMI on Follow-up Maternal BMI ........................................ 51
Table 14: Summary of Total, Direct and Indirect Effect for Prepregnancy BMI on Childhood
Percentile BMI Percentile: Influenced by Follow-up Maternal BMI........................... 52
Table 15: Effects of Pre Pregnancy BMI on Childhood BMI Percentile: Influence of Maternal
Waist Circumference ................................................................................................. 53
Table 16: Effect of Pre Pregnancy BMI on Maternal Waist Circumference ............................... 54
Table 17: Summary of Total, Direct and Indirect Effect for Prepregnancy BMI on Childhood
BMI Percentile: Influenced by Maternal Waist Circumference ................................... 55
Table 18: Effects of Pre Pregnancy BMI on Childhood BMI Percentile: Influence of Postpartum
Weight Difference ....................................................................................................... 56
vi

Table 19: Effect of Pre Pregnancy BMI on Post Partum Weight Difference .............................. 57
Table 20: Summary of Total, Direct and Indirect Effect for Prepregnancy BMI on Childhood
BMI Percentile: Influenced by Post Partum Weight Difference .................................. 58
Table 21: Effects of Pre Pregnancy BMI on Childhood BMI Percentile: Influence of Post
Delivery Weight Change ............................................................................................. 59
Table 22: Effect of Pre Pregnancy BMI on Post Delivery Weight Change ................................. 60
Table 23: Summary of Total, Direct and Indirect Effect for Prepregnancy BMI on Childhood
BMI Percentile: Influenced by Post Delivery Weight Change .................................... 61

vii

LIST OF FIGURES

Figure 1: Inclusion and Exclusion Criteria Used for Analytic Sample ........................................ 63
Figure 2: Timing and Definitions of Maternal Body Composition Measurements ...................... 64
Figure 3: Conceptual Model of Pre Pregnancy BMI and Childhood BMI with Follow-up
Maternal Body Composition as a Mediator ................................................................... 65
Figure 4: Conceptual Model of Pre Pregnancy BMI and Childhood BMI with Follow-up
Maternal Body Composition as a Mediator and Adjusted for Confounders ................. 66

viii

KEY TO ABBREVIATIONS

aOR

Adjusted Odds Ratio

ARCH

Archive for Research on Child Health

BMI

Body Mass Index

CDC

Centers for Disease Control and Prevention

CI

Confidence Interval

GWG

Gestational Weight Gain

IOM

Institute of Medicine

OR

Odds Ratio

PDWC

Post Delivery Weight Change

PPWD

Postpartum Weight Difference

RR

Relative Risk

SES

Socioeconomic Status

WHO

World Health Organization

ix

CHAPTER 1: INTRODUCTION
According to the World Health Organization (WHO) “Childhood obesity is one of the
most serious public health challenges of the 21st century” (1). Obesity is a multifaceted issue that
is caused by several factors including genetics, environment, and behavior. Obese is a term to
describe a person who has excess body fat and is most often estimated by body mass index
(BMI). For children, a BMI in the 95th percentile or higher using age and sex specific categories
is considered obese (2). Approximately 17% of adolescents are considered obese in the United
States based on data from 2011-2014 (3). Obese children are more likely to have health problems
throughout childhood such as asthma, diabetes, and liver disease (4). These children are also
likely to remain obese into adulthood and have more severe health consequences compared to
adults who were not obese during childhood. These health consequences include coronary heart
disease, stroke, cancer, and a lower quality of life (5). It has become clear that an early start to
obesity prevention, perhaps as early as in utero is a crucial step in decreasing overall obesity
rates.
Pregnancy is a critical time period for child development. Several studies have shown
that high pre pregnancy body mass index (BMI) and excessive gestational weight gain (GWG)
are associated with high birth weight and later childhood obesity, which has been attributed to
fetal programming during pregnancy (6). Mothers who are overweight often have increased
levels of insulin, growth hormones, and leptin and are more likely to have higher insulin
resistance compared to women of a normal BMI (7). Because the fetus receives all its nutrients
through the placenta, it is extremely susceptible to any changes in the mother’s hormones.
Insulin is involved in the development of the central nervous system and is suggested to
influence the development of the infant’s metabolism (7). If a child is at higher risk of becoming
1

obese because the mother is overweight, this transgenerational cycle may explain some of the
increase in obesity rates over the past few decades.
Although many studies have examined the relationship between pre pregnancy BMI and
childhood BMI, post pregnancy factors have rarely been taken into account in those same
studies. When assessing factors related to child BMI, it is important to consider post pregnancy
factors to account for the child’s environment. Information is lacking about whether home and
social environment may contribute to the risk of childhood obesity as much as or more than
metabolic programming in utero. Literature on breastfeeding has been inconclusive but some
studies have found breastfeeding for more than 4 months to have a significant protective effect
against childhood obesity (8). Consideration of maternal post pregnancy body composition can
also provide insight into post birth behavior. A healthy post pregnancy body composition could
be indicative of a healthier lifestyle, which may indicate a healthier childhood environment. By
taking into account pre and post pregnancy factors, a more complete picture of the relationships
between maternal size and body composition at these time points and child BMI may be
elucidated.

2

CHAPTER 2: BACKGROUND
2.1 Childhood Obesity
Obesity is most often defined by a categorization of body mass index (BMI--weight in
kilograms divided by height in meters squared). Because normal body composition of children
differs by age and sex, BMI is ranked in relation to a reference population to take into account
these variables. In 2000 the CDC revised the standard growth charts based on pooled data from
five cross sectional nationally representative health examination surveys: NHES II (1963-1965),
NHES III (1966-70), NHANES I (1971-74), NHANES II (1976-80), NHANES III (1988-94). To
calculate the percentile charts for child BMI the national data was smoothed with a variety of
parametric and nonparametric procedures (9). The cut-off points for the BMI percentiles were
defined in 1997 and were reviewed again in 2006 by an expert committee that represented 15
different health care organizations (10). The cut-off points capture different risk levels and
roughly correspond to the BMI categorization for adults. Overweight is defined as being at or
above the 85th percentile and below the 95th percentile (2). Obesity is defined as at or above the
95th percentile for children of the same age and sex (2). Extreme obesity is defined as a BMI at
or above 120% of the age and sex-specific 95th percentile on the CDC BMI-for-age growth
charts (3). In the United States in 2011-2014, 17% of youth aged 2 to 19 were classified as obese
and 5.8% of youth were classified within the extreme obesity category (3).
Results from the National Health and Nutrition Examination Survey (NHANES) from
1988-1994 and 2013-2014 indicate higher odds of obesity for children between the ages of 6 to
11 compared to the ages of 2 to 5 (OR 2.29) (3). The odds were also higher for non-Hispanic
Black children and Hispanic children compared to White children (OR 1.34 and 1.48

3

respectively) (3). Although the prevalence of childhood obesity has remained relatively stable
since 2007, the prevalence of extreme obesity is significantly increasing (3).
Childhood obesity has been associated with an increased risk of sleep apnea,
hypertension, abnormal glucose intolerance, and type 2 diabetes (4). Subsequently, children who
are obese are also at an increased risk for depression and impaired social and emotional
functioning (11). Obese children are more likely to be obese in adulthood and have more severe
health consequences compared to obese adults who were not obese as children (12). One study
looked at children’s obesity status from 1 year of age to 17 years and compared that to the
subject’s obesity status as an adult. The odds of being obese as an adult if the subject was obese
as a child were significant starting at age 3 and greatly increased as the child aged (13). At age
three the odds of being obese as an adult was 1.3, at the age of 9 the odds increased to 10.3, and
at the age of 17 the odds of being obese as an adult increased to 20.3. Because there is such a
strong association between childhood obesity and adult obesity, examining risk factors
associated with childhood obesity is critical.
2.2 Maternal and Childhood Body Composition
2.2.1 Pre Pregnancy BMI
For adults, obesity is defined as BMI at or greater than 30 kg/m2. Approximately 32% of
women between the ages of 20 and 39 are classified as obese and 6% are classified as morbidly
obese in the United States (5). Accordingly, in recent years, more women are entering pregnancy
either overweight or obese. The incidence of women being overweight or obese at the start of
pregnancy increased from 25% to 35% and the incidence of obesity at delivery increased from
29% and 39% between the years 1991 and 2001 (6). Therefore, it is important to understand if a
mother’s pre pregnancy BMI can negatively impact her child.
4

Research in animal models has suggested that there is a relationship between a mother’s
pre pregnancy weight and her offspring’s body composition. A study by Shankar et al studied the
impact of feeding pre pregnancy rats excess calories for three weeks before mating. All of the
dams were placed on the same diet for the length of pregnancy. The pups were distributed to
normal weight and obese mothers to ensure the nutritional intervention was solely from pre
pregnancy. Results from this study showed that offspring from obese dams had a higher body fat
percentage compared to the offspring from the lean dams (14).
The relationship between maternal body composition and her offspring’s body
composition has been studied in humans through prospective studies. In humans it has been
found, obesity increases risk for adverse pregnancy outcomes. Prospective studies with human
participants have shown that women who enter pregnancy at a higher BMI are more likely to
have comorbid conditions, such as diabetes, which can lead to adverse pregnancy outcomes and
fetal malformations (7). In addition, maternal obesity is associated with an increased risk of
preeclampsia, gestational diabetes, hypertension, and a need for operative delivery (7).
Not only does a high pre pregnancy BMI put the mother at higher risk during pregnancy,
but studies have shown that it also negatively impacts the offspring. Maternal obesity is related
to an increased risk of macrosomia, fetal distress, jaundice, stillbirth, and congenital
malformations (15). Furthermore, some research has shown that a high pre pregnancy BMI can
also have a long lasting negative impact on the child. A study by Fuemmeler et al examined the
association between pre pregnancy obesity and childhood weight growth trajectories in 704
mother-infant dyads. Measurements of the child were taken throughout the first 24 months of life
and the median number of measurements per child was 14. The results showed that children born
to women with a pre pregnancy BMI over 40 compared to women with a BMI between 18 and
5

24.9 were 8% larger during the first 24 months and had a delayed time for maximum growth; all
of which are risk factors for obesity and chronic disease later in life (16). Another study by Li et
al analyzed the relationship between pre pregnancy BMI and the BMI of children between 2 and
14 years of age (8). After adjusting for several confounders, it was found that compared to
children of mothers with a pre pregnancy BMI less than 25, children of mothers with a pre
pregnancy BMI between 25 and 29.9 were 2.48 times more likely to have a BMI in the 95th
percentile or higher (95% CI: 1.7, 3.4) (8). Children of mothers who had a pre pregnancy BMI
over 30 were 3.68 times more likely to have a BMI in the 95th percentile or higher (95% CI: 2.4,
5.5) (8). In a longitudinal study by Li et al, pre pregnancy BMI was found to be significantly
associated with offspring BMI Z-score from birth to 5 years of age (β=0.025 ¹ 0.005) (17).
Although the previous studies all focused on child BMI, several studies have also
examined the relationship between pre pregnancy BMI and other anthropometric measures for
the child. In a study conducted by Kaar et al, 313 mother child pairs were recruited to examine
maternal pre pregnancy BMI and offspring outcomes. The average age of follow up for the child
was 10.4 years. In this sample, children of overweight/obese mothers were significantly higher in
all anthropometric measurements that were assessed (18). In this study, they measured BMI
(95% CI: 1.57, 3.51), waist circumference (95% CI: 3.82, 8.98), visceral adipose tissue (95% CI:
36.14, 81.52), and subcutaneous adipose tissue (95% CI: 3.59, 9.53) (18). In another study by
Adriette et al, the main outcome of interest was waist to height ratio for the child at 2 to 6 years
of age (19). Their findings showed that maternal pre pregnancy BMI was positively associated
with the offspring’s waist to height ratio (β=0.025; 95% CI: 0.010, 0.039) (19).
Despite these studies finding statistically significant results, they all had several
limitations. One of the major limitations was the lack of data on environmental factors after
6

pregnancy such as breastfeeding (16-18). Additionally, some studies collected the weight
measurements through self-report (7, 16-18). Other studies sampled specific populations, such as
women at high risk for gestational diabetes, so results were not widely generalizable (18).
2.2.2 Gestational Weight Gain
In addition to examining the relationship between pre pregnancy BMI and child BMI,
several studies have examined the relationship between gestational weight gain (GWG) and child
BMI. GWG is closely associated with pre pregnancy BMI. It is important to examine both pre
pregnancy BMI and GWG because the intrauterine exposure to excessive gestational weight gain
may contribute to metabolic programming for the child (20). Because of the growing obesity
epidemic and the concern of gaining too much weight during pregnancy, the Institute of
Medicine (IOM) set guidelines based on pre pregnancy BMI. Recommendations for weight gain
are between 12.5 and 18 kg, 11.5 to 16 kg, 7 to 11.5 kg, and 5 and 9 kg, for women who are
classified as underweight, normal weight, overweight, or obese, respectively (21). Additionally,
the IOM has recommendations for the rate of GWG for the 2nd and 3rd trimester. Women who are
underweight should gain approximately 0.51 kg per week, normal weight women should gain
0.42 kg per week, overweight women should gain 0.28 kg per week, and obese women should
gain 0.22 kg per week (21). Women who exceed recommendations are at risk of complications at
birth and higher post pregnancy weight retention.
The study previously mentioned by Fuemmeler et al also examined GWG and found that
compared to children whose mothers gained adequate weight during pregnancy, children born to
mothers who exceeded the recommendation were 5% heavier at follow-up compared to mothers
who met the recommendation (95% CI: 0.3%, 8.6%) (16). In a study by Sridhar et al, women
who were members of the Kaiser Permanente Northern California group practice and completed
7

a follow-up survey between the years 2007 and 2010 were included in the analysis. Of the 4,145
mothers included, 66.2% of women exceeded the IOM GWG recommendation. Children of
mothers who exceeded the recommendation were larger at birth (3475 g vs 3344 g), more likely
to be macrosomic (15.0% vs 8.3%), and more likely to be overweight or obese at 2 to 5 years of
age (20.4% vs 14.5%) (20). After adjusting for maternal age, education, prepregnancy BMI and
race the results were attenuated but children of women who exceeded IOM recommendations
were still 1.51 times more likely to be overweight or obese at 2 to 5 years of age (95% CI: 1.231.87) (20).
As with examining pre pregnancy BMI, several studies also looked at different measures
of child anthropometrics and GWG. The previously mentioned study by Kaar et al also looked at
GWG. Similar to the results found for high pre pregnancy BMI, excessive GWG also was
significantly associated with child BMI (β=0.34; 95% CI: 0.25, 0.44), waist circumference
(β=0.83; 95% CI 0.58, 1.08), visceral adipose tissue (β = 0.72; 95% CI: 0.39, 1.06), and
subcutaneous adipose tissue (β=7.26; 95% CI: 4.90, 9.62) (18). Another study by Oken et al
examined 1044 mother child pairs 3 years after delivery. A child whose mother gained excess
weight during pregnancy had had an increased odds of being obese compared to a child whose
mother met the IOM recommendations (OR: 1.30; 95% CI: 1.0, 1.62) (22). This remained
significant after adjusting for both sociodemographic and biological variables. This study also
measured skinfold thickness in the child and found mothers with a greater GWG had children
with more adiposity at age 3, however this finding was not statistically significant (β=0.18; 95%
CI: -0.06, 0.42) (22).
Not all studies found a statistically significant relationship between GWG and child BMI.
In a study by Olson et al, data was extracted from the Basset Mothers Health Project which was
8

an observational cohort study of 622 health women that were followed from early pregnancy
until two years postpartum. Medical records of the offspring of these women were located and
audited for anthropometric information and 321 offspring had measured heights and weights
between the ages of 3.5 to 4.5. At age 4, there was no relationship found between GWG and
child BMI (RR=1.28; 95% CI 0.83, 1.96) (23). Von Kries et al also examined GWG and its
association with obesity in the offspring in using data from the German Health Interview and
Examination Survey for Children and Adolescents. This study looked at children 3 to 17 and did
not find a statistically significant association (RR=1.10; 95% CI: 0.99,1.22) (24). Additionally,
in a prospective study by Diesel et al. they used GWG z-scores and looked at the relative risk for
childhood obesity when the z-score was 1.5. After adjustment, children at the age of ten whose
mother’s GWG z-score was 1.5 had a relative risk of 1.50 compared to children whose mother’s
GWG z-scores were 0 (0.94, 2.40) (25).
Research examining GWG and child obesity has been inconclusive. Studies on GWG
have limitations similar to the pre pregnancy BMI literature, the studies on GWG also have
limitations. This includes limited data on post pregnancy environmental factors such as
breastfeeding (15, 17, 20, 22). Also the calculation of GWG can lead to errors due to not having
a true pre pregnancy weight or a true weight at delivery (15,17, 20, 22). Also some studies had a
specific population such as an unhealthy population (17) or a population of high socioeconomic
status (SES) (22).
2.2.3 Post Pregnancy Maternal Body Composition
The studies previously mentioned did not examine the maternal body composition post
pregnancy. Post pregnancy maternal body composition is also an important component when
looking at how maternal body composition can influence childhood obesity. Post pregnancy
9

maternal body composition could help explain some of the environmental exposures that
influence the child’s body composition. Mothers who are at a healthy weight after pregnancy or
return close to their pre pregnancy weight may partake in certain healthy lifestyle behaviors and
these behaviors could translate to how the child behaves as well. However, very few studies have
assessed the relationship with post pregnancy maternal body composition and childhood obesity.
In the study by Whitaker et al, parental obesity status was calculated using a
mathematical model that incorporated the parents BMI status at two separate time points.
Offspring obesity was measured at 5 different time points: 1-2 years, 3-5 years, 6-9 years, 10-14
years, and 15-17 years. Compared to children of mothers who were not obese, children of obese
mothers were more likely to also be obese at all five time points (OR 3.6; 95 % CI: 2.1-5.9, OR
3.6; 95% CI: 2.2-5.7, OR 3.3; 95% CI: 2.2-5.1, OR 3.1; 95% CI: 2.0-4.6, OR 2.8; 95% CI: 1.94.1) (13). In a secondary analysis by Robinson et al, weight was measured in children at 4 to 5
years of age and post delivery maternal weight change (PDWC) was defined as the difference in
maternal weight 3 to 36 months after delivery and weight at delivery. This study found that a 5
kg increase in PDWC was associated with a 12% increase in the odds of the child being
overweight (26). Further examination of this data showed that the relationship between PDWC
and child weight was stronger when the PDWC was measured within the first year of delivery.
These two studies also have several limitations. Neither collected data on breastfeeding
(13, 26). Also the time frame of when data was collected was variable across the studies (13, 26)
Lastly, the way post pregnancy body composition was measured was not ideal. In the study by
Whitaker et al the maternal body composition was not directly measured, but predicted (13). To
predict maternal BMI the authors abstracted several weight and height measurements from the
medical record and linear interpolation was used to estimate what the BMI of the mother would
10

be in each time period. The article does not state when the abstracted measurements were from
relative to the delivery of the child so it is impossible to know if pregnancy weight was
considered. In the study by Robinson et al to examine the post pregnancy environment they used
the difference between weight at delivery and post-partum weight which could be misclassified
as retained GWG (26). Furthermore, other methods of measuring post pregnancy body
composition, such as body fat percentage and waist circumference, to my knowledge, is
completely missing from the literature when examining the relationship between post pregnancy
body composition and childhood obesity.
2.2.4 Breastfeeding and Childhood Obesity
The majority of the previously mentioned studies did not collect data on breastfeeding.
Per the American Academy of Pediatrics, infants should be breastfed for the first 6 months
exclusively and be supplemented for at least a year (27). Many studies have identified
breastfeeding as a protective factor for childhood obesity, and others have found breastfeeding is
not a protective factor. A meta-analysis by Yan et al examined 25 studies that looked at
childhood obesity and feeding patterns. Pooling all 25 studies, children who had ever been
breastfed where 23% less likely to be overweight compared to children who had never been
breastfed (aOR= 0.78; 95% CI: 0.74,0.81) (28). Additionally, 17 of the studies measured the
length of time the infant was breastfed and a dose response was found. Infants who breastfed for
less than 3 months compared to never breast fed were 10% less likely to be obese. Infants who
breastfed 7 or more months were 21% less likely to be obese (28). This meta-analysis does have
several limitations. First, it is unclear what the follow up time for calculating childhood obesity
was for each study. Second, the results only report the adjusted findings. Some of the studies

11

may have over adjusted. Lastly, the meta-analysis did not distinguish between infants who were
exclusively breastfed and those who were partially breastfed.
2.3 Conceptual Framework
As obesity in children and adults becomes increasingly more severe, it is important to
study various risk factors associated with both. Because previous research has shown children
who are obese are more likely to be obese as adults and have more severe consequences,
examining childhood obesity is critical if we want to reduce the prevalence of obesity for all ages
(12).
Although there have been numerous experiments using animal models and several
prospective cohort studies examining the effect of pre pregnancy maternal weight composition
on childhood obesity, less research has been done on the association between post pregnancy
body composition and child BMI. Where the pregnancy related weight could give insight into the
biological context of a child’s BMI, post pregnancy body composition could help explain the
environmental context. When researching the potential driving force for the obesity epidemic it
is important to examine all time points. Post pregnancy body composition can give insight on
environmental factors such as diet and activity level. By examining pre pregnancy BMI, GWG,
and post pregnancy body composition, a fuller picture of how the maternal body composition
affects the child can be accomplished.
2.4 Specific Aims
This study assessed components of maternal obesity and its association with childhood
obesity. The aims for this study were as follows:

12

2.4.1 Aim 1:
To describe the relationship between components of maternal body composition and the
outcome of childhood obesity in offspring at ages 3 to 6 years with the separate
components of maternal composition being defined as:
1. Pre Pregnancy BMI
2. GWG
3. Follow-Up Maternal BMI
4. Postpartum Weight Difference (PPWD)
5. Post Delivery Weight Change (PDWC)
6. Waist Circumference (WC)
2.4.2 Aim 2:
To assess mediation by maternal body composition measurements (PPWD, PDWC, WC)
at 3 to 6 year post pregnancy follow-up in the relationship between pre pregnancy BMI
and childhood BMI percentile.

13

CHAPTER 3: METHODS
3.1 Study Population
The study population consisted of a subset of women who participated in the Archive for
Research on Child Health (ARCH). Results are from an ongoing prospective pregnancy cohort
initiated in 3 mid-Michigan prenatal care clinics that began recruitment in 2008. The goal of
ARCH is to archive urine and blood samples from pregnancy and placenta samples for the
purpose of etiological research on adverse maternal and child health outcomes. Both mothers and
offspring are followed throughout childhood and are asked to complete annual questionnaires.
These data are intended to be used for future research questions regarding pregnancy and
childhood development. Women were recruited from participating clinics, which included a
university faculty obstetric clinic, hospital residency clinic, and a county health clinic. Women
were approached prior to 14 weeks’ gestation, and were eligible if they were 18 or older and
could complete the survey in English. The Institutional Review Board for Michigan State
University approved the protocols and procedures of the ARCH study.
A sub study of ARCH, ARCH Child Development, was initiated in 2014 and collected
anthropometric data on the child and woman three to seven years post pregnancy. Of the 132
women and child pairs enrolled in ARCH Child Development, 130 pairs had available data on
height and weight. For this analysis only singleton pregnancies were included and women who
were not pregnant at the time of the follow up measurement, which resulted in an analytic
sample of 114 mother-child pairs.
3.2 Pre Pregnancy BMI
At time of enrollment, women were asked to report their pre pregnancy weight and
height. Additionally, height and weight of the woman was extracted from the birth certificate.
14

Both sources of height and weight were examined for this analysis and are termed ‘intake pre
pregnancy BMI’ and ‘birth certificate pre pregnancy BMI’. A correlation test was done to
determine if the intake pre pregnancy BMI differed from the birth certificate BMI. Pre pregnancy
BMI from the intake questionnaire was used when analyzing pre pregnancy BMI due to the
sample being larger compared to the birth certificate BMI sample. BMI was calculated by taking
weight in kilograms and dividing it by the height in meters squared.
3.3 Child BMI
Children ranged from 36 months to 83 months at the time of anthropometric data
collection. Height and weight were both measured twice and the average of these two
measurements was used to calculate BMI. CDC’s ‘Children’s BMI tool for schools’ was used to
convert each child’s BMI into the appropriate percentile based on the child’s age and gender
(29). Children who were below the 85th percentile were classified as normal weight, children
between the 85th and 94.99th percentile were classified as overweight, and children in the 95th
percentile or higher were classified as obese.
3.4 Gestational Weight Gain
Gestational weight gain (GWG) was reported in kilograms and calculated by subtracting
the women’s pre pregnancy weight from the reported weight at delivery. The source for both
weights came from the birth certificate. GWG examined as both a continuous variable and a
categorical variable. GWG was categorized as inadequate, adequate, or excessive based on the
IOM guidelines which considers the woman’s pre pregnancy BMI (21). Table 1 shows the
categorization for gestational weight gain.

15

3.5 Follow-up Body Composition
Follow-up anthropometric measurements were taken at 3 to 6 years postpartum. Weight
and height were measured without shoes on, and both measurements were taken twice then
averaged to calculate BMI for the mother. Postpartum weight difference (PPWD) was calculated
by taking the difference between the measured maternal weight at follow up and the self-reported
pre pregnancy weight. Post delivery weight change (PDWC) was calculated by taking the
difference between the measured maternal weight at follow up and the delivery weight recorded
on the Birth Certificate. Figure 2 represents the various maternal body composition
measurements. Waist circumference was measured twice in centimeters and the average was
taken.
3.6 Other Variables
The following variables were extracted from the birth certificate: child’s race, child’s sex,
maternal smoking history, maternal age, parity, mode of delivery, gestational age, and birth
weight. Child’s race was categorized as either Black Non-Hispanic, or Other. Maternal smoking
history was coded as a ‘yes’ if the mother reported ever smoking before pregnancy. Mode of
delivery was categorized as either Vaginal or Cesarean.
Two variables to account for socio-economic status, maternal education and household
income, were extracted from the intake questionnaire. For analysis, maternal education was
categorized into either “High school diploma or less” and “Some college or more.” For analysis,
income was categorized into either “Less than $50,000” and “$50,000 or more.”
Covariates that were extracted from the post-pregnancy follow-up included parity and
breastfeeding. Both variables were extracted from the one month follow up, and the yearly
follow-up questionnaires. Mothers were asked if they were breastfeeding, and how long they
16

breastfed for. Breastfeeding was then categorized as either: Breastfed for less than 1 month,
breastfed for more than 1 month but less than 6 months, breastfed for more than 6 months but
less than 12, and breastfed for 12 or more months. No information was collected on whether
breastfeeding practices were exclusive or partial at any time point. Mothers were also asked if
they had a change in health and reported if they were currently pregnant or had recently
delivered a new baby.
3.7 Analytic Plan
The Pearson’s correlation coefficient was calculated to examine the correlation between
the intake pre pregnancy BMI and the birth certificate BMI to determine which source for pre
pregnancy to use. The Pearson’s correlation coefficient was also calculated to examine the
correlation between all exposure variables and childhood BMI percentile. Descriptive Statistics
(means, standard deviations, and proportions) were calculated for all variables for interest.
Three types of models were created using linear regression. The first model was to
examine the fetal environment effect on Child BMI where pre pregnancy BMI was the main
exposure of interest. The second model was to examine the infant environment where the main
exposure of interest was examining the various follow-up maternal body composition
measurements separately: Follow-up BMI, waist circumference, PPWD, and PDWC. This
second set of models does not allow for causal interpretation but lays groundwork for the third
model. The third model was to examine the fetal and infant environment together using the
various follow-up maternal body composition measurements as a mediator.
To look at the follow-up maternal body composition measurements as a mediator, a
simple mediation model was utilized (30). The following three equations were used:

17

1. 𝑌 = 𝑖1 + 𝑐𝑋 + 𝑒1
2. 𝑌 = 𝑖2 + 𝑐 ′ 𝑋 + 𝑏𝑀 + 𝑒2
3. 𝑀 = 𝑖3 + 𝑎𝑋 + 𝑒3
Where i1, i2, i3 are intercepts, Y is childhood BMI percentile, X is pre pregnancy BMI, and M is
the follow-up maternal body composition measurement.
There are several important assumptions that must be met to test for mediation. First, the
residuals from equation 2 and 3 are independent and M and the residual in equation 2 are
independent. Second, there is no interaction between X and M in equation 3. Third, the specified
model includes no misspecification of the causal order (ie., Y M X instead of X M Y).
Fourth, there is no reciprocal causation between the mediator and the dependent variable. Lastly,
there is no misspecification due to imperfect measurement (30).
The direct effect was estimated by the effect pre pregnancy BMI had on child BMI
percentile while holding the follow-up maternal body composition measurements constant. The
indirect effect was the effect pre pregnancy BMI had on the follow-up maternal body
composition measurements that then affected child BMI percentile. Also reported in the
mediation analysis is the total effect which was derived from estimating the effect pre pregnancy
BMI had on the follow-up maternal body composition measurements without taking into account
the follow-up body composition measurement as a mediator (i.e. results from the first model
mentioned above). Table 2 shows a basic derivation of the indirect, direct, and total effect for
unadjusted models.
Beta coefficients, a coefficient of determination, and 95% confidence intervals were
calculated for all models. The 95% confidence intervals for the indirect effect were calculated by

18

the bootstrapping method where 10,000 samples were taken. Covariates were included in the
model if: 1) forward and backward selection found the variable as significant, 2) if there was a
biological rationale for including the covariate as a confounder based on previous literature. An
alpha level of 0.05 was used to determine statistical significance. All statistical procedures were
performed in SAS 9.4 and the mediation analysis was performed by the SAS macro PROCESS
(30).

19

CHAPTER 4: RESULTS
Pre pregnancy BMI was measured in two different ways for this study. Women were
asked to report their weight and height at the intake questionnaire. Pre pregnancy weight was
also recorded on the birth certificate. A correlation test was performed to examine how these two
variables compare. The results from this test showed the two variables were highly correlated (r=
0.85). Pre pregnancy BMI for the following analyses was extracted from the intake questionnaire
because there were fewer missing observations for the intake questionnaire compared to the birth
certificate (N=131 vs N=124). After excluding women who gave birth to twins and women who
were currently pregnant at the time of follow up there was a total of 114 mother-child pairs
included in the analysis. Table 3 compares the demographics and weight component variables
from the entire ARCH childhood development sample to the analytic sample.
4.1 Demographic Variable Description
For the analytic sample at the time of delivery, maternal age ranged from 18 to 39 with
the mean age of 26.7 years. Fifty four percent of the mothers had given birth before the index
pregnancy. Approximately 61% of the mothers had more than a high school education, and 60%
reported a household income of less than $23,000 a year. From the birth certificate, 15% of
mothers had reported a history of smoking. In this population, 53% of children were White, NonHispanic, 12% were Black, Non-Hispanic, 18% were Hispanic, and 17% were classified as other
race/ethnicity. The mean gestational age was 39.2 weeks, the mean birth weight was 3,400.9 g,
and 73% of the children were delivered vaginally. In the follow-up questionnaires, 44% of
mothers reported breastfeeding for less than 1 month, 31% reported breastfeeding between 1 and
6 months, 7% reported breastfeeding between 7 and 12 months, and 18% reported breastfeeding

20

for more than 1 year. The mean age at follow up was 58.2 months with a range of 36 months to
82 months.
4.2 Maternal Weight Description
The mean pre pregnancy BMI was 27.9; 23% of women were classified as overweight
and 33% were classified as obese. Table 4 shows the characteristics among women of a normal
BMI and women who were classified as overweight or obese. Women who were classified as
overweight (N=64) were more likely to have delivered a girl (RR 1.5, 95% CI:1.0-2.2). The
mean gestational weight gain was 15.0 kg with 58% of women exceeding the IOM
recommendation of gestational weight gain (N=66). Women who were classified as overweight
or obese were more likely to gain excessive weight during pregnancy (RR 2.1, 95% CI: 1.5,2.8).
The mean BMI for follow-up for the mother was 30.3, which was significantly higher than the
pre pregnancy BMI (mean difference= -2.28, 95% CI -3.10, -1.47). When looking at the weight
difference between delivery and at the follow up visit, the mean difference was -4.8 kg. The most
weight lost was 69 kg, and the most weight gained was 61 kg. When looking at the difference
between pre pregnancy weight and the follow up weight, the mean difference was 5.9 kg with
one woman weighing 31.8 kg more than her pre pregnancy weight and one woman weighing
51.7 kg less than her pre pregnancy weight.
4.3 Childhood BMI Description
The mean percentile, compared to the national percentiles, for this study population was
the 55th percentile. Twenty four percent of children were classified as being overweight or obese
(N=27). Table 5 shows the sociodemographic and perinatal characteristics of children who were
classified as normal compared to those classified as overweight or obese. None of the

21

characteristics examined were significantly different between children classified as a normal
weight compared to children classified as overweight or obese.
4.4 Pre Pregnancy BMI Model
The relationship between pre pregnancy BMI and child BMI percentile was statistically
significant before adjusting for confounding (β=0.86; 95% CI: 0.1, 1.6). For the pre pregnancy
BMI model there were 10 additional covariates that could be examined. By using both forward
and backwards selection procedures only 2 covariates were significant at Îą= 0.2. Both
race/ethnicity and smoking history remained significant when using Îą =0.05. Once race and
smoking history were controlled, pre pregnancy BMI remained significant (β=0.84; 95% CI: 0.1,
1.6). Because GWG has been found to be an important confounder in previous research, a third
model controlled for GWG, race, and smoking history. After adjusting for GWG, race, and
smoking history, 37% of the variation in childhood BMI percentile was explained. The
coefficient for pre pregnancy BMI increased and remained significant after controlling for the
significant covariates (β=0.99; 95% CI: 0.3, 1.7). Table 6 shows a summary of the results for this
model. A model to examine gestational weight gain was also constructed. GWG in the
unadjusted model was statistically insignificant and remained insignificant after controlling for
race and smoking status. Table 7 shows the results for the GWG model.
4.5 Follow-up Maternal Body Composition Models
Follow-up maternal BMI was not significantly associated with child BMI percentile
(β=0.63; 95% CI: -0.04, 1.3). However, after adjusting for race, follow-up BMI became
statistically significantly associated with childhood BMI percentile (β=0.69; 95% CI: 0.03, 1.3).
After adjusting for race, 26% of the variation in childhood BMI was explained. Table 8 shows
the results of this relationship.
22

The relationship between waist circumference and child BMI percentile was significant
in the unadjusted model (β=0.26; 95% CI: 0.01, 0.5). After performing selection procedures,
none of the nine possible covariates were significant at Îą =0.2. To compare to other models
race/ethnicity was controlled for in model 2 where waist circumference no longer remained
significant; Table 9 shows the results of both models.
When examining the relationship between PPWD, no statistically significant relationship
was found in the unadjusted model (β=0.19: 95% CI: -0.3, 0.7). Through selection procedures
race/ethnicity was the only covariate that was found to be significant. After adjusting for
race/ethnicity, PPWD remained insignificant (β=0.20: 95% CI: -0.3,0.7). This adjusted model,
shown in Table 10, explains 17% of the variation in childhood BMI percentile.
The relationship between PDWC and childhood BMI percentile was also found to be
statistically insignificant in the unadjusted model (β=0.17; 95% CI: -0.2, 0.5). Using both
backward and forward selection procedure, race/ethnicity was the only covariate that was
significant. After adjustment PDWC remained statistically insignificant (β=0.17; 95% CI: 0.2,0.5). After adjusting for race/ethnicity the model explained 20% of the variation in childhood
BMI percentile; Table 11 shows these results.
4.6 Mediation Analysis
After examining the relationship between each maternal body composition variable and
childhood BMI percentile, mediation models were explored. Covariates for the adjusted
mediation model were chosen by forward and backwards selection procedure by looking at the
following three models where follow-up maternal body composition is defined as one of the
following: follow-up maternal BMI, waist circumference, PPWD, or PDWC. The following
models are written in the format of dependent variable= independent variable.
23

1.

Follow-up maternal body composition= Pre pregnancy BMI

2.

Childhood BMI percentile= Pre pregnancy BMI + Significant covariates from equation
1

3.

Childhood BMI percentile= Pre pregnancy BMI + follow-up maternal body
composition + Significant covariates from equation 1+ Significant covariates
from equation 2

From this process GWG, race, and smoking history were found to be significant at alpha=0.20
and remained significant at alpha=0.05. Figure 3 provides a visual representation of the
mediation model.
All four of the follow-up maternal body composition measurements were examined.
From the previously mentioned analysis of pre pregnancy BMI and childhood BMI shown in
Table 6, the total effect of pre pregnancy BMI on childhood BMI was significant in the
unadjusted model and remained significant after adjusting for GWG, race/ethnicity, and smoking
history (β=0.98; 95% CI 0.28, 1.69).
Looking at the direct effect of pre pregnancy BMI when controlling for follow-up
maternal BMI, pre pregnancy BMI was no longer statistically significant (β=0.98; 95% CI:-0.43,
2.40). Looking at the indirect effect, follow-up maternal BMI was statistically insignificant,
indicating that there is no evidence of this variable being a mediator (β=0.10; 95% CI: -1.13,
1.29). Table 12-14 summarizes the findings for the mediation model with follow-up maternal
BMI. Similar results were found for maternal waist circumference. After controlling for waist
circumference, pre pregnancy BMI no longer remained significant (β=0.73; 95% CI: -0.42,
1.88). The indirect effect for waist circumference was also insignificant, indicating no evidence

24

of mediation (β=0.24; 95% CI: -0.68, 0.91). Table 15-17 shows the summary of the analysis
using waist circumference as a mediator.
After controlling for postpartum weight difference, pre pregnancy BMI remained
statistically significant and also remained statistically significant after further controlling for the
other covariates (β=1.00; 95% CI: 0.29, 1.72). The indirect effect of PPWD was statistically
insignificant indicating no evidence of mediation (β=-0.01; 95% CI: -0.21, 0.06). Table 18-20
shows the results for the analyses of PPWD as a mediator. Post delivery weight difference had
similar results to PPWD. After controlling for PDWC, pre pregnancy BMI remained statistically
significant (β=0.99; 95% CI: 0.25, 1.73). The indirect effect of PDWC was statistically
insignificant suggesting that this variable does not act as a mediator between pre pregnancy BMI
and childhood BMI percentile (β=0.005; 95% CI: -0.24, 0.31). Table 21-23 shows the results
from the analysis of PDWC as a mediator.

25

CHAPTER 5: DISCUSSION
5.1 Interpretations
In this study of 114 mother and child pairs, the relationship between maternal body
composition and childhood BMI was examined. By examining maternal body composition
before pregnancy, at delivery, and 3 to 6 years after delivery, a fuller picture can be created to
help explain how maternal body composition relates to childhood BMI. Looking at all three time
points gives us the opportunity to begin to account for both biological and environmental factors
that could impact the child.
In the analysis, prepregnancy BMI was significantly associated with childhood BMI
percentile and this relationship remained after controlling for GWG, race/ethnicity, and smoking
history. This finding supports previous research. Follow-up maternal BMI was also significantly
associated with childhood BMI percentile after controlling for race, and maternal waist
circumference was statistically significantly associated in the unadjusted model. Both variables
had a weaker relationship with childhood BMI percentile, however, as compared to pre
pregnancy BMI. After adjustment, one unit increase in pre pregnancy BMI was associated with
an increase in childhood BMI percentile of 0.99. A one unit increase in follow-up maternal BMI
was associated with an increase in childhood BMI percentile of 0.69 and waist circumference
was associated with an increase in childhood BMI percentile of 0.24. Postpartum weight
difference and post delivery weight change both had little effect on childhood BMI percentile,
even after controlling for race/ethnicity.
The next step for this study was to examine if follow-up maternal body composition
measurements served as a mediator between prepregnancy BMI and childhood BMI as there is
no literature to our knowledge examining this relationship. In the first analysis using follow-up
26

maternal BMI as a mediator, pre pregnancy BMI no longer is statistically significant after
adjusting for follow-up maternal BMI. This suggests that there is no evidence that pre pregnancy
BMI influenced child BMI percentile independent of follow-up maternal BMI. However,
because pre pregnancy BMI and follow-up maternal BMI are highly correlated this result should
be interpreted with caution (r=0.86). Due to the highly-correlated nature of these two variables
the chance of a type II error is increased. The indirect effect for this model is also statistically
insignificant; thus, there is no evidence in these data that follow-up maternal BMI acts as a
mediator between prepregnancy BMI and childhood BMI. Similar results were found when
analyzing maternal waist circumference as a mediator. After controlling for waist circumference,
pre pregnancy BMI no longer remained significantly associated with childhood BMI percentile
however waist circumference and pre pregnancy BMI were also highly correlated and thus
skewing the results (r=0.77). The indirect effect for this model was also insignificant indicating
there is no evidence maternal waist circumference serves as a mediator.
The analysis of postpartum weight difference resulted in prepregnancy BMI remaining
statistically significant after adjusting for PPWD indicating that prepregnancy BMI is associated
with childhood BMI percentile independent of PPWD. The same results were found when
looking at post delivery weight change.
These results indicate that pre pregnancy BMI is the strongest predictor of childhood
BMI despite examining maternal body composition at multiple time points throughout her life
and using different techniques (i.e. mediation) to attempt to examine causality. It is likely that a
woman’s lifestyle and home environmental factors, such as socioeconomic status, are similar
before, during, and after pregnancy. The factors that lead to a woman’s body composition before
pregnancy most likely remain after pregnancy and helps explain why after pre pregnancy BMI,
27

follow-up maternal BMI and waist circumference impact child BMI the greatest. Working under
the assumption that a woman’s lifestyle habits that contributed to her pre pregnancy BMI are
similar to her post pregnancy lifestyle habits, post pregnancy BMI and waist circumference do
not necessarily provide a better insight into the home environment post pregnancy compared to
her pre pregnancy BMI.
Examining change in maternal weight provides a slightly different picture because it
allows us to see if the environmental factors that led to the mother’s BMI allows her to lose the
weight gained from pregnancy. When investigating post delivery weight change there was a
significant difference between women who had a pre pregnancy BMI less than 25, a BMI
between 25 and 29.9, and over 30. Women who were classified as obese before pregnancy lost
less weight between the follow-up measurement and delivery. For women who did not have
subsequent pregnancies, women classified as having a normal or overweight BMI before
pregnancy lost, on average, 9.5 kg and 10.7 kg respectively between follow-up and delivery.
Women classified as being obese before pregnancy on average lost 1.3 kg between follow-up
and delivery. The in-home environment of the mother likely explains this finding. If unhealthy
behavior contributed to a women’s obesity status before pregnancy, the unhealthy behavior
likely remains after pregnancy and therefore would make it difficult for the mother to lose the
pregnancy weight. This potential unhealthy behavior at home could translate to a poor diet or
physical inactivity in the child. However, in our results examining weight change affected the
child BMI very little. Pre pregnancy BMI remains to be the best predictor of child BMI but it is
difficult to differentiate how pre pregnancy BMI is affecting the child in-utero versus how the
behaviors that led to the pre pregnancy BMI is affecting the child’s BMI do to the home
environment.

28

5.2 Strengths
There were several strengths of this study. Unlike previous literature, this study measured
maternal body composition after pregnancy. To our knowledge, this is the first study that
assessed the association between measured maternal body composition after pregnancy and
childhood BMI. Additionally, many of the previously mentioned studies examining the
relationship between pre pregnancy BMI and childhood BMI did not account for post pregnancy
factors such as breastfeeding. This study did collect information on breastfeeding as well as the
parity of the mother at the time of the follow up visit; however, both variables did not
significantly influence the final models. As a second strength, we had access to several
measurements to attempt to capture post pregnancy maternal weight composition.
5.3 Limitations
There are limitations in this study. First off, using maternal body composition as a
variable to help explain the in-home environment after pregnancy is not ideal. Theoretically, a
desirable maternal body composition after pregnancy could indicate healthy eating habits at
home as well as other healthy behaviors; however, the home environment after pregnancy could
be measured in a more accurate way such as examining the child’s diet and physical activity.
Also, the data collected on smoking history is limited. This is a known important covariate when
examining pre pregnancy BMI and childhood BMI, however in our data we only know if the
mother had a history of smoking. Only 40% of mothers with a history of smoking reported the
date they had quit smoking. For the 60% of mothers who did not report a quit date, it is unknown
if they had smoked through pregnancy and never quit smoking or if this is missing data. Also,
there was no follow-up questions on smoking so it is not known if any of the mothers smoked
after pregnancy. Next, the breastfeeding data did not differentiate between mothers who solely

29

breastfed from mothers who supplemented with bottle feeding. More complete breastfeeding
data could result in the breastfeeding variable significantly impacting the analysis. Furthermore,
the variable waist circumference was only measured post pregnancy; a better variable to examine
how central adiposity affects child BMI percentile would have been the difference between post
pregnancy waist circumference and pre pregnancy waist circumference. Also as a limitation,
most mothers in this population have a household income under $50,000 a year and are White,
non-Hispanic. Lastly, this study had a small sample size. The results of this study may not be
generalizable to the general population.

30

CHAPTER 6: CONCLUSION
In this study, we aimed to look at maternal body composition at various points
throughout her life and how that can affect her child’s BMI. By examining body composition
before, at, and after delivery through different methods we can start to explore how a mother can
impact a child’s weight status. The results from this study showed that pre pregnancy BMI is the
strongest predictor of childhood BMI. When using follow-up maternal body composition as a
mediator, no mediation was observed. All three time periods for maternal body composition
could potentially affect a child’s risk but there is no evidence that follow-up maternal body
composition can better explain the child’s BMI compared to pre pregnancy BMI. Pre pregnancy
BMI can affect the in-utero environment and cause the child to be at an increased risk for obesity
at a biological level, but pre pregnancy BMI can also be reflective of a mother’s lifestyle and
habits. Unhealthy habits before pregnancy are likely to remain after pregnancy.
To tease out the complexity of the mechanism in which pre pregnancy BMI affects child
BMI future studies need to be performed. It is difficult to understand if the effect of fetal
programming caused by a high pre pregnancy BMI is the driving force for child obesity, or if it
is the home environment that is contributing to the obesity status of the child, or some
combination of both. If possible, it would be worth studying women who either lost a significant
amount of weight or gained a significant amount of weight between pregnancies and how this
impacted their offspring. By doing this type of experiment, a better understanding of how fetal
programming may affect child BMI can be attained.
Despite not fully knowing the complexity of factors that lead to a specific pre pregnancy
BMI, there remains a strong relationship between a high pre pregnancy BMI and childhood
obesity. This study supports the hypothesis that overweight mothers provide an environment that
31

is a risk factor for offspring obesity. Interventions should be developed to help women lose
weight if they are planning to become pregnant and to help women stay healthy throughout
pregnancy. Further studies should also look at other in-home environmental factors such as diet
and physical activity at a young age and how these measurements influence the relationship
between pre pregnancy BMI and childhood BMI.

32

APPENDICES

33

APPENDIX A:
Tables

34

Table 1: Recommendations for Total Weight Gain During Pregnancy by Pre Pregnancy BMI a
Pre Pregnancy BMI
Underweight (<18.5 kg/m2)
Normal Weight (18.5-24.9 kg/m2)
Overweight (25.0-29.9 kg/m2)
Obese (≥30.0 kg/m2)
a

Weight Gain Range (kg)
12.5
18.0
11.5
16.0
7.0
11.5
5.0
9.0

Recommendations are from the Institute of Medicine (IOM) (21)

35

Table 2: Explanation of Terminology Used for Mediation Analysis
Effects
Total Effect = (a*b)+c’
Direct Effect = c’
Indirect Effect = (a*b)
a is the path between the exposure and mediator
b is the path between the mediator and outcome
c’ is the path between exposure and outcome when controlling for the mediator

36

Table 3: Comparison of Mother and Child Characteristics Between the ARCH Child
Development Cohort and Analytic Sample
Table 3 (cont’d)
ARCH Child
Development
N=132
N (%)

Analytic Sample
N=114

P value

72 (55)
18 (14)
22 (16)
20 (15)

60 (53)
14 (12)
21 (18)
19 (17)

0.8047

66 (50)
66 (50)

54 (47)
60 (53)

0.9314

27 (20)
47 (36)
47 (36)
11 (8)
39.3 (2.9) a
3411.7 (545.4) a

22 (19)
40 (35)
41 (36)
11(10)
39.2 (3.0) a
3400.9 (541.60) a

92 (74)
32 (26)

83 (73)
31 (27)

0.7715

57 (43)
38 (29)
14 (11)
23 (17)
16.0 (2.0) a

50 (44)
35 (31)
8 (7)
21 (18)
16 (2.0) a

0.5908

96 (75)
18 (14)
14 (11)

88 (76)
16 (14)
10 (10)

0.8978

N (%)

Child Characteristics
Child Race/ Ethnicity
White
Black
Hispanic
Other
Child Gender
Male
Female
Child Age at measurement
3 Years
4 Years
5 Years
6 years
Gestational Age (Weeks)
Birth Weight (Grams)
Mode of Delivery
Vaginal
Cesarean
Breastfeeding
Less than 1 Month
1 Month to 6 Months
7 to 11 Months
≥ 12 Months
Child BMI
Child BMI percentile b
Less than 85th percentile
85th to 94.99th
95th and above

37

0.8217
0.8313

0.7715

Table 3 (cont’d)
ARCH Child
Development
N=132
N (%)

Analytic Sample
N=114

P value

16 (13)

12 (11)

33 (25)

30 (26)

47 (36)
34 (26)

40 (36)
30 (27)

74 (58)
30 (23)
13 (10)
11 (9)

67 (60)
23 (21)
11 (10)
10 (9)

0.9487

105 (85)
18 (15)

96 (84)
18 (16)

0.8134

57 (46)
67 (54)

52 (46)
62 (54)

0.9341

41 (33)
83 (67)
26.8 (5.2) a
28.4 (7.8) a

37 (32)
77 (68)
26.7 (5.0) a
28.0 (7.7) a

0.8920

53 (41)
33 (25)
45 (34)
14.8 (7.1) a

50 (44)
26 (23)
38 (33)
15.0 (7.3) a

0.7939

31 (23)
31 (23)
70 (54)
30.7 (8.7) a

22 (19)
26 (23)
66 (58)
30.3 (8.6) a

0.6061

N (%)

Maternal Characteristics
Maternal Education
Less than High School
High School Graduate or
Equivalent
Some College
College Graduate or More
Household Income
Under $25,000
$25,000 to $49,000
$50,000 to $74,999
$75,000 or Above
Maternal History of Smoking c
No
Yes
Parity During Index Pregnancy
0
≥1
Parity at Follow-Up
1
≥2
Maternal Age at Delivery
Pre Pregnancy BMI
Pre Pregnancy BMI
Distribution
Normal
Overweight
Obese
Gestational Weight Gain (kg)
Gestational Weight Gain
Distribution
Inadequate
Adequate
Excessive
Follow-up Maternal BMI

38

0.8914

0.8335
0.5627

0.8139

0.6298

Table 3 (cont’d)

Follow-up Maternal BMI
Distribution
Normal
Overweight
Obese
Follow-up Waist
Circumference (cm)
Postpartum Weight Difference
(kg)
Post Delivery Weight Change
(kg)

ARCH Child
Development
N=132
N (%)

Analytic Sample
N=114

P value

37 (28)
39 (30)
54 (42)
101.3 (24.6) a

32 (29)
35 (31)
45 (40)
100.2 (24.2) a

0.9232

6.3 (11.8) a

6.3 (12) a

0.9994

-6.7 (17.7) a

-7.5 (17.1) a

0.6002

N (%)

0.6308

Note: Observations missing for education (n=2), income (n=4), follow-up BMI (n=2), follow-up waist
circumference (n=12), postpartum weight difference (n=3), post delivery weight change (n=10)
a
Presented as Mean (Standard Deviation)
b
Percentiles were defined by the Centers for Disease Control and Prevention and based on the United States’
population
c
Smoking History is defined as having ever smoked

39

Table 4: Maternal and Child Sociodemographic Characteristics by Pre Pregnancy BMI Category
Table 4 (cont’d)
Normal BMI a
N=50

Overweight/Obese
BMI b
N=64

N

%

N

%

29
6
8
7

58
12
16
14

31
8
13
12

48
13
20
19

P value

Child Characteristics
Child Race/ Ethnicity
White, Non-Hispanic
Black, Non-Hispanic
Hispanic
Other
Child Gender
Male
Female
Child Age at Measurement
3 Years
4 Years
5 Years
6 Years
Gestational Age (Weeks)
Birth Weight (Grams)
Mode of Delivery
Vaginal
Cesarean
Breastfeeding
≤ 1 month
1 to 6 months
7 to 11 months
≥ 12 months

0.7634

0.0445
29
21

58
42

25
39

39
61

8
18
18
6
50
50

16
36
36
12
39.4 (2.7) c
3322.3 (537.6) c

14
22
23
5
64
64

22
34
36
8
39.1 (3.2) c
3462.2 (540.9) c

37
13

74
26

46
18

72
28

0.7932

0.6988
0.1721
0.8002
0.2789

25
12
2
11

50
24
4
22

25
23
6
10

39
36
9
16

Maternal Characteristics
Maternal Education
Less than High School
High School Graduate or
Equivalent
Some College
College Graduate or More
Household Income
Under $25,000
$25,000 to $49,000
$50,000 to $74,999
$75,000 or above

0.1092
7

14

5

8

11

22

19

31

14
18

28
36

26
12

42
19
0.2338

25
14
4
5

52
29
8
10

40

42
9
7
5

67
14
11
8

Table 4 (cont’d)
Normal BMI a
N=50
N
Maternal History of Smoking d
No
Yes
Parity at Intake Pregnancy
0
≥1
Parity at Follow-up
1
≥2
Maternal Age (Years)
Gestational Weight Gain (kg)
Gestational Weight Gain
Categorized
Inadequate
Adequate
Excessive

%

Overweight/Obese
BMI b
N=64
N

P value

%
0.1249

45
5

90
10

51
13

79
21
0.0491

28
22

56
44

24
40

37
63

21
29
50
50

42
48
25.9 (4.6) c
14.54 (6.4) c

16
48
64
64

25
75
27.4 (5.3) c
15.29 (7.9) c

0.2830

0.1120
0.8385
<0.0001

17
19
14

34
38
28

5
7
52

Note: Observations missing for education (n=2), income (n=3),
a
Normal BMI defined as a having a Body Mass Index of less than 25
b
Overweight/ Obese BMI defined as having a Body Mass Index of 25 or higher
c
Presented as Mean (Standard Deviation)
d
Smoking history is defined as having ever smoked

41

8
11
81

Table 5: Maternal and Child Sociodemographic and Perinatal Characteristics by Childhood BMI
Percentile Category
Table 5 (cont’d)
Normal BMI
N=88
N
Child Characteristics
Child Race/ Ethnicity
White, Non-Hispanic
Black, Non-Hispanic
Hispanic
Other
Child Gender
Male
Female
Child Age at Measurement
3 years
4 Years
5 Years
6 Years
Gestational Age (Weeks)
Birth Weight (Grams)
Mode of Delivery
Vaginal
Cesarean
Breastfeeding
≤ 1 month
1 to 6 months
7 to 11 months
≥12 months
Maternal Characteristics
Maternal Education
Less than High School
High School Graduate
or Equivalent
Some College
College Graduate or
more
Household Income
Under $25,000
$25,000 to $49,000
$50,000 to $74,999
$75,000 or above

a

%

Overweight/ Obese
BMI b
N=26
N
%

P value

0.3957
46
11
14
17

52
13
16
19

14
3
7
2

54
12
27
8
0.8877

42
46

48
52

12
14

46
54

17
30
34
7
88
88

19
38
39
8
39.2 (3.0) c
3376.0 (555.9) c

5
10
7
4
26
26

19
38
27
15
39.4 (3.0) c
3484.9 (490.7) c

65
23

74
26

18
8

69
31

0.5676

0.7190
0.3701
0.6409

0.4937
20
18
36
4

23
20
41
16

6
5
7
7

24
20
28
28
0.5079

9

10

3

12

21

24

9

36

34

39

6

24

23

26

7

28
0.4176

51
18
10
6

60
21
12
7
42

16
5
1
4

62
19
4
15

Table 5 (cont’d)
Normal BMI
N=88
N
Maternal History of
Smoking d
No
Yes
Parity at Intake Pregnancy
0
≥1
Parity at Follow-up
1
≥2
Maternal Age (Years)
Gestational Weight Gain
(kg)
Gestational Weight Gain
Categorized
Inadequate
Adequate
Excessive

a

%

Overweight/ Obese
BMI b
N=26
N
%

P value

0.9913
14
74

16
84

4
22

16
84
0.0837

44
44

50
50

8
18

31
69

69
19
88

78
22
26.3 (5.0) c

20
6
26

77
23
28.2 (5.0) c

0.9073

88

14.72 (6.47) c

26

15.77 (9.59) c

0.5198

0.0832

0.6762
18
21
49

20
24
56

4
5
17

15
19
65

Note: Descriptive Results shown for the analytic sample (N=114)
Observations missing for education (n=2), income (n=3),
Percentiles were defined by the Centers for Disease Control and Prevention and based on the United States’
population
a
Normal BMI defined as having a Body Mass Index less than the 85th percentile
b
Overweight/ Obese defined as having a Body Mass Index at the 85 th percentile or higher
c
Presented as mean (standard deviation)
d
Smoking history is defined as having ever smoked

43

Table 6: Association between Pre Pregnancy BMI and Childhood BMI Percentile a

Variables
Pre Pregnancy
BMIe
Race: Black f
Smoking
History: Yes g
GWG (kg) h, i

N

Model 1b
R2=0.05
β
95 %
CI

Model 2c
R2=0.13
95 % CI

β

Model 3d
R2=0.14
β
95 % CI

114

0.86*

0.1,1.6

0.84*

0.1, 1.6

0.99*

0.3, 1.7

16

-

-

-22.21

-38.72, 5.70

-21.76*

-38.2, -5.3

18

-

-

-16.66

-38.72, -5.70

-17.1*

-32.2, -2.0

114

-

-

-

-

0.41

-0.3, 1.1

Results from Linear Regression using the analytic sample (N=114)
Percentiles were defined by the Centers for Disease Control and Prevention and based on the United
States’ population
Abbreviation: GWG = gestational weight gain
a
Pearson Coefficient for pre pregnancy BMI and Child BMI percentile r=0.22
b
Model 1: Unadjusted
c
Model 2: Adjusted for race and smoking history
d
Model 3: Adjusted for race, smoking history, and gestational weight gain
e
Beta Coefficient interpreted as a 1 unit increase in pre pregnancy BMI
f
Referent Group non-Black N=98
g
Referent Group No Smoking History N=96
h
Beta Coefficient interpreted as a 1 kg increase in weight gained during pregnancy
i
Pearson Coefficient for pre pregnancy BMI and GWG is r=0.01
* Indicates p<0.05

44

Table 7: Association between Gestational Weight Gain and Childhood BMI Percentile a
Model 1b
R2=0.008
β
95 % CI

Model 2c
R2=0.08
β
95 % CI

Model 3d
R2=0.14
β
95 % CI

Variables

N

GWG (kg) h, i
Race: Black f
Smoking
History: Yes g

114
16

0.36
-

-0.4, 1.1
-

0.42
-20.9*

-0.3, 1.2
-37.9, -3.9

0.41
-21.76*

-0.3, 1.1
-38.2, -5.3

18

-

-

-14.13

-29.5, 1.2

-17.1*

-32.2, -2.0

0.99*

0.3, 1.7

Pre Pregnancy
BMIe

114

Results from Linear Regression using the analytic sample (N=114)
Percentiles were defined by the Centers for Disease Control and Prevention and based on the United
States’ population
Abbreviation: GWG = gestational weight gain
a
Pearson Coefficient for pre pregnancy BMI and Child BMI percentile r=0.36
b
Model 1: Unadjusted
c
Model 2: Adjusted for race and smoking history
d
Model 3: Adjusted for race, smoking history, and pre pregnancy BMI
e
Beta Coefficient interpreted as a 1 unit increase in pre pregnancy BMI
f
Referent Group non-Black N=98
g
Referent Group No Smoking History N=96
h
Beta Coefficient interpreted as a 1 kg increase in weight gained during pregnancy
i
Pearson Coefficient for pre pregnancy BMI and GWG is r=0.01
* Indicates p<0.05

45

Table 8: Association between Follow-up Maternal BMI and Childhood BMI Percentile a

Variables

N

Follow-up Maternal
BMI d
Race: Black e

112
13

Model 1b
R2=0.03
β
95 % CI

β

Model 2c
R2=0.07
95 % CI

0.63

-0.04, 1.3

0.69*

0.03, 1.3

-

-

-17.7

-35.1, -0.4

Results from Linear Regression using the analytic sample – 2 for missing data on Follow-up BMI (N=112)
Percentiles were defined by the Centers for Disease Control and Prevention and based on the United States’
population
a Pearson Coefficient for follow-up BMI and Child BMI percentile r=0.19
b Model 1: Unadjusted
c Model 2: Adjusted for race
d Beta Coefficient interpreted as a 1 unit increase in follow-up BMI
e Referent Group non-Black N=98
* Indicates p<0.05

46

Table 9: Association between Follow-up Waist Circumference and Childhood BMI Percentile a

Variables
Waist Circumference
(cm) d
Race: Black e

N

Model 1b
R2=0.04
β
95 % CI

β

Model 2c
R2=0.06
95 % CI

103

0.26*

0.01, 0.5

0.24

-0.02, 0.5

12

-

-

-16.0

-34.4, 2.4

Results from Linear Regression using the analytic sample – 11 for missing waist circumference data (N=103)
Percentiles were defined by the Centers for Disease Control and Prevention and based on the United States’
population
a Pearson Coefficient for Waist Circumference and Child BMI percentile r=0.20
b Model 1: Unadjusted
c Model 2: Adjusted for race
d Beta Coefficient interpreted as a 1 cm increase in waist circumference
e Referent Group non-Black N=91
* Indicates p<0.05

47

Table 10: Association between Postpartum Weight Difference and Childhood BMI Percentile a
Model 1b
R2=0.0005
Variables
PPWD (kg) d
Race: Black e

Model 2c
R2=0.03

N

β

95 % CI

β

95 % CI

112
13

0.19
-

-0.3, 0.7
-

0.20
-15.96

-0.3, 0.7
-33.5,1.6

Results from Linear Regression using the analytic sample – 2 for missing data for follow-up weight (N=112)
Percentiles were defined by the Centers for Disease Control and Prevention and based on the United States’ population
Abbreviation: PPWD=Post partum weight difference (defined as the difference between pre pregnancy weight and
follow-up weight)
a Pearson Coefficient for PPWD and Child BMI percentile r=0.05
b Model 1: Unadjusted
c Model 2: Adjusted for race
d Beta Coefficient interpreted as a 1 kg increase in the difference between (follow-up – pre pregnancy weight)
e Referent Group non-Black N=98
* Indicates p<0.05

48

Table 11: Association between Post Delivery Weight Change and Childhood BMI Percentile a

Variables
PDWC d
Race: Black e

N

β

112
13

0.17
-

Model 1b
R2=0.009
95 % CI
-0.2,0.5
-

β
0.17
-16.0

Model 2c
R2=0.04
95 % CI
-0.2, 0.5
-33.6, 1.5

Results from Linear Regression using the analytic sample – 2 for missing data for follow up weight (N=112)
Percentiles were defined by the Centers for Disease Control and Prevention and based on the United States’ population
Abbreviation: PDWC = Post Delivery Weight Change (defined as the difference between weight at delivery and weight
at follow up)
a Pearson Coefficient for PDWC and Child BMI percentile r=0.02
b Model 1: Unadjusted
c Model 2: Adjusted for race
d
Beta Coefficient interpreted as a 1 kg increase in the difference between (follow-up– delivery weight)
e Referent Group non-Black N=98
* Indicates p<0.05

49

Table 12: Effects of Pre Pregnancy BMI on Childhood BMI Percentile: Influence of Follow-up
Maternal BMI a

Total Effect Model
d

Pre Pregnancy BMI
GWG e
Race: Black f
Smoking History:
Yes g
Controlled Direct
Effect Model
Pre Pregnancy BMI d
Follow-up BMI h
GWG e
Race: Black f
Smoking History:
Yes g

N
112
112
13
18

Model 1b
R2 =0.05
β
95% CI
0.84* 0.12, 1.56
R2 = 0.05

N
112
112
112
13
18

β
0.74
0.10
-

Model 2c
R2=0.13
β
95% CI
0.98*
0.28, 1.69
0.44
-0.33, 1.20
20.34*
3.20, 37.48
-17.21*
-32.43, -1.99
R2=0.13

95% CI
-0.70, 2.18
-1.20, 1.41
-

β
0.98
-0.01
0.44
20.33*
-17.21

95% CI
-0.43, 2.40
-1.29, 1.28
-0.33, 1.20
3.03, 37.64
-32.57, -1.85

Results from Linear Regression using the analytic sample (N=114)
Percentiles were defined by the Centers for Disease Control and Prevention and based on the
United States’ population
Abbreviation: GWG = Gestational Weight Gain
a Pearson Coefficient for pre pregnancy BMI and Follow-up BMI r=0.86
b Model 1: Unadjusted
c Model 2: Adjusted for GWG, race, and Smoking history
d Beta Coefficient interpreted as a 1 unit increase in pre pregnancy BMI
e
Beta Coefficient interpreted as a 1 kg increase in GWG
f Referent Group non-Black N=99
g Referent Group no smoking history N= 94
h Beta Coefficient interpreted as a 1 unit increase in Follow-up BMI
* Indicates p<0.05

50

Table 13: Effect of Pre Pregnancy BMI on Follow-up Maternal BMI a

Mediation Model
Pre Pregnancy BMI
GWG e
Race: Blackf
Smoking History:
Yesg

d

N
112
112
13
18

Model 1b
R2 =0.75
β
95% CI
0.95*
0.85, 1.05
-

Model 2c
R2=0.75
β
95% CI
0.96*
0.85, 1.06
0.03
-0.08, 0.15
-1.25
-3.83, 1.33
-17.21*
-32.43, -1.99

Results from Linear Regression using the analytic sample (N=114)
Percentiles were defined by the Centers for Disease Control and Prevention and based on the United States’
population
Abbreviation: GWG = Gestational Weight Gain
a
Pearson Coefficient for pre pregnancy BMI and follow-up BMI is r=0.86
b Model 1: Unadjusted
c
Model 2: Adjusted for GWG, race and smoking history
d Beta coefficient interpreted as a 1 BMI unit increase
e Beta coefficient interpreted as a 1 kg increase in GWG
f Referent Group non-Black N=99
g Referent group is no smoking history N=94
* Indicates p<0.05

51

Table 14: Summary of Total, Direct and Indirect Effect for Prepregnancy BMI on Childhood
Percentile BMI Percentile: Influenced by Follow-up Maternal BMI

Total Effect c
Controlled Direct Effect d
Indirect Effect e

Model 1a
β
95% CI
0.84*
0.12, 1.56
0.74
-0.70, 2.18
0.10
-1.13, 1.29

Model 2b
β
95% CI
0.99*
0.28, 1.69
0.92
-0.49, 2.34
0.07
-1.12, 1.21

c

The coefficient reported is for pre pregnancy BMI from the total effect model in Table 11

d

The coefficient reported is for pre pregnancy BMI from the Controlled direct effect in Table 11

e

The coefficient reported is the product of pre pregnancy BMI from the controlled direct effect in Table 11 and
follow-up BMI from Table 12

52

Table 15: Effects of Pre Pregnancy BMI on Childhood BMI Percentile: Influence of Maternal
Waist Circumference a

Total Effect Model
Pre Pregnancy BMI
GWG e
Race: Black f
Smoking History: Yes g

N
103
103
12
17

Controlled Direct Effect
Model
Pre Pregnancy BMI d
Waist Circumference h
GWG e
Race: Black f
Smoking History: Yes g

N
103
103
103
12
17

d

Model 1b
R2 =0.04
β
95% CI
0.79*
0.04, 1.53
2
R = 0.05

Model 2c
R2=0.13
β
95% CI
0.97*
0.24, 1.70
0.44
-0.34, 1.22
21.74*
3.53, 39.94
-19.17*
3.29, 35.05
2
R =0.14

β
0.48
0.13
-

β
0.73
0.10
0.45
21.37*
-19.00*

95% CI
-0.70, 1.66
-0.26, 0.52
-

95% CI
-0.42, 1.88
-0.28, 0.48
-0.33, 1.24
3.05, 39.69
-34.95, -3.05

Results from Linear Regression using the analytic sample (N=114)
Percentiles were defined by the Centers for Disease Control and Prevention and based on the United States’ population
Abbreviation: GWG = Gestational Weight Gain
a Pearson Coefficient for pre pregnancy BMI and Waist Circumference is r=0.77
b Model 1: Unadjusted
c Model 2: Adjusted for GWG, race, and Smoking history
d Beta Coefficient interpreted as a 1 unit increase in pre pregnancy BMI
e Beta Coefficient interpreted as a 1 kg increase in GWG
f Referent Group non-Black N=91
g Referent Group no smoking history N= 86
h Beta Coefficient interpreted as a 1 cm increase in waist circumference
* Indicates p<0.05

53

Table 16: Effect of Pre Pregnancy BMI on Maternal Waist Circumference a

Mediation Model
d

Pre Pregnancy BMI
GWGe
Race: Blackf
Smoking History: Yesg

N
103
103
12
17

Model 1b
R2 =0.59
β
95% CI
2.33*
1.95, 2.71
-

β
2.35*
-0.15
3.56
1.69*

Model 2c
R2=0.60
95% CI
1.97, 2.74
-0.57, 0.27
-6.14, 13.25
-6.77, 10.14

Results from Linear Regression using the analytic sample (N=114)
Percentiles were defined by the Centers for Disease Control and Prevention and based on the United States’ population
Abbreviation: GWG = Gestational Weight Gain
a Pearson Coefficient for pre pregnancy BMI and waist circumference is r=0.77
b Model 1: Unadjusted
c Model 2: Adjusted for GWG, race and smoking history
d Beta coefficient interpreted as a 1 BMI unit increase
e Beta coefficient interpreted as a 1 kg increase in GWG
f Referent Group non-Black N=91
g Referent group is no smoking history N=86
* Indicates p<0.05

54

Table 17: Summary of Total, Direct and Indirect Effect for Prepregnancy BMI on Childhood
BMI Percentile: Influenced by Maternal Waist Circumference

Total Effect c
Controlled Direct Effect d
Indirect Effect e

Model 1a
β
95% CI
0.79*
0.04, 1.53
0.48
-0.70, 1.66
0.31
-0.56, 1.08

Model 2b
β
95% CI
0.97*
0.24, 1.70
0.73
-0.42, 1.88
0.24
-0.67, 0.94

c

The coefficient reported is for pre pregnancy BMI from the total effect model in Table 14

d

The coefficient reported is for pre pregnancy BMI from the Controlled direct effect in Table 14

e

The coefficient reported is the product of pre pregnancy BMI from the controlled direct effect in Table 14 and
waist circumference from Table 15

55

Table 18: Effects of Pre Pregnancy BMI on Childhood BMI Percentile: Influence of Postpartum
Weight Difference a

Total Effect Model
Pre Pregnancy BMI d
GWG e
Race: Black f
Smoking History: Yes g

N
112
112
13
18

Controlled Direct Effect
Model
Pre Pregnancy BMI d
PPWD h
GWG e
Race: Blackf
Smoking History: Yesg

N
112
112
112
13
18

Model 1 b
R2 =0.05
β
95% CI
0.84*
0.12, 1.56
2
R = 0.05

Model 2 c
R2=0.13
β
95% CI
0.99*
0.29, 1.70
0.46
-0.30, 1.22
20.78*
3.62, 37.93
-17.72*
-32.94, -2.50
2
R =0.13

β
0.85*
0.11
-

β
1.00*
0.08
0.45
20.81*
-17.57*

95% CI
0.13, 1.58
-0.36, 0.59
-

95% CI
0.29, 1.72
-0.38, 0.54
-0.31, 1.22
3.58, 38.04
-32.88, -2.27

Results from Linear Regression using the analytic sample (N=114)
Percentiles were defined by the Centers for Disease Control and Prevention and based on the United States’ population
Abbreviation: GWG = Gestational Weight Gain
PPWD = post partum weight difference( defined as follow-up weight – pre pregnancy weight)
a Pearson Coefficient for pre pregnancy BMI and PPWD r=-0.10
b Model 1: Unadjusted
c Model 2: Adjusted for GWG, race, and Smoking history
d Beta Coefficient interpreted as a 1 unit increase in pre pregnancy BMI
e Beta Coefficient interpreted as a 1 kg increase in GWG
f Referent Group non-Black N=99
g Referent Group no smoking history N= 94
h Beta Coefficient interpreted as a 1 kg increase in the change between follow-up and pre pregnancy weight
* Indicates p<0.05

56

Table 19: Effect of Pre Pregnancy BMI on Post Partum Weight Difference a

Mediation Model
d

Pre Pregnancy BMI
GWGe
Race: Blackf
Smoking History: Yesg

N
112
112
13
18

Model 1b
R2 =0.01
β
95% CI
-0.16
-0.45, 0.13
-

Model 2c
R2=0.12
β
95% CI
0.99*
0.29, 1.70
0.46
-0.30, 1.22
20.78*
3.62, 37.93
17.72*
2.50, 32.94

Results from Linear Regression using the analytic sample (N=114)
Percentiles were defined by the Centers for Disease Control and Prevention and based on the United States’
population
Abbreviations: GWG = Gestational Weight Gain,
PPWD = post partum weight difference (defined as the difference between follow-up and pre
pregnancy weight)
a Pearson Coefficient for pre pregnancy BMI and PPWD is r=-0.10
b Model 1: Unadjusted
c Model 2: Adjusted for GWG, race and smoking history
d Beta coefficient interpreted as a 1 BMI unit increase
e Beta coefficient interpreted as a 1 kg increase in GWG
f Referent Group non-Black N=99
g Referent group is no smoking history N=94
* Indicates p<0.05

57

Table 20: Summary of Total, Direct and Indirect Effect for Prepregnancy BMI on Childhood
BMI Percentile: Influenced by Post Partum Weight Difference

Total Effectc
Controlled Direct Effectd
Indirect Effecte

Model 1a
β
95% CI
0.84*
0.12, 1.56
0.85*
0.13, 1.58
-0.02
-0.24, 0.05

Model 2b
β
95% CI
0.99*
0.29, 1.70
1.00*
0.29, 1.72
-0.01
-0.23, 0.06

c

The coefficient reported is for pre pregnancy BMI from the total effect model in Table 17

d

The coefficient reported is for pre pregnancy BMI from the Controlled direct effect in Table 17

e

The coefficient reported is the product of pre pregnancy BMI from the controlled direct effect in Table 17 and
Post Partum Weight Difference from Table 18

58

Table 21: Effects of Pre Pregnancy BMI on Childhood BMI Percentile: Influence of Post
Delivery Weight Change a

Total Effect Model
Pre Pregnancy BMI d
GWG e
Race: Black f
Smoking History: Yes g

N
112
112
12
17

Controlled Direct Effect
Model
Pre Pregnancy BMI d
PDWC h
GWG e
Race: Blackf
Smoking History: Yesg

N
112
112
112
12
17

Model 1b
R2 =0.05
β
95% CI
0.84*
0.12, 1.56
2
R = 0.05

Model 2c
R2=0.13
β
95% CI
0.99*
0.29, 1.70
0.46
-0.30, 1.22
20.78*
3.62, 37.93
-17.72
-32.94, 2.50
2
R =0.13

β
0.84*
-0.01
-

β
0.99*
0.01
0.46
20.78*
-17.69*

95% CI
0.09, 1.59
-0.36, 0.33
-

95% CI
0.25, 1.73
-0.34, 0.35
-0.33, 1.26
3.54, 38.01
-33.03, -2.36

Results from Linear Regression using the analytic sample (N=114)
Percentiles were defined by the Centers for Disease Control and Prevention and based on the United States’ population
Abbreviation: GWG = Gestational Weight Gain
PDWC = post delivery weight change (defined as the difference between follow-up weight and weight at
delivery)
a Pearson Coefficient for pre pregnancy BMI and PDWC r=0.27
b Model 1: Unadjusted
c Model 2: Adjusted for GWG, race, and Smoking history
d Beta Coefficient interpreted as a 1 unit increase in pre pregnancy BMI
e Beta Coefficient interpreted as a 1 kg increase in GWG
f Referent Group non-Black N=99
g Referent Group no smoking history N= 94
h Beta Coefficient interpreted as a 1 unit increase in the difference between follow-up weight and weight at delivery
* Indicates p<0.05

59

Table 22: Effect of Pre Pregnancy BMI on Post Delivery Weight Change a

Mediation Model
d

Pre Pregnancy BMI
GWG e
Race: Blackf
Smoking History: Yesg

N
112
112
13
18

Model 1b
R2 =0.07
β
95% CI
0.59*
0.19, 0.99
-

β
0.62*
-0.63
0.54
3.20

Model 2c
R2=0.15
95% CI
0.23, 1.01
-1.05, -0.20
-8.97, 10.05
-5.24, 11.63

Results from Linear Regression using the analytic sample (N=114)
Percentiles were defined by the Centers for Disease Control and Prevention and based on the United States’ population
Abbreviation: GWG = Gestational Weight Gain
PDWC = post delivery weight change (defined as the difference between follow-up weight and weight
at delivery)
a Pearson Coefficient for pre pregnancy BMI and PDWC is r=0.27
b Model 1: Unadjusted
c Model 2: Adjusted for GWG, race and smoking history
d Beta coefficient interpreted as a 1 BMI unit increase
e Beta coefficient interpreted as a 1 kg increase in GWG
f Referent Group non-Black N=99
g Referent group is no smoking history N=94
* Indicates p<0.05

60

Table 23: Summary of Total, Direct and Indirect Effect for Prepregnancy BMI on Childhood
BMI Percentile: Influenced by Post Delivery Weight Change

Total Effectc
Controlled Direct Effectd
Indirect Effecte

Model 1a
β
95% CI
0.84*
0.12, 1.56
0.84
-0.9, 1.59
-0.01
-0.28, 0.23

Model 2b
β
95% CI
0.99*
0.29, 1.70
0.99
0.25, 1.73
0.01
-0.23, 0.06

c

The coefficient reported is for pre pregnancy BMI from the total effect model in Table 20

d

The coefficient reported is for pre pregnancy BMI from the Controlled direct effect in Table 20

e

The coefficient reported is the product of pre pregnancy BMI from the controlled direct effect in Table 20 and Post
Delivery Weight Change from Table 21

61

APPENDIX B:
Figures

62

Figure 1: Inclusion and Exclusion Criteria Used for Analytic Sample

63

Figure 2: Timing and Definitions of Maternal Body Composition Measurements

64

Figure 3: Conceptual Model of Pre Pregnancy BMI and Childhood BMI with Follow-up
Maternal Body Composition as a Mediator

65

Figure 4: Conceptual Model of Pre Pregnancy BMI and Childhood BMI with Follow-up
Maternal Body Composition as a Mediator and Adjusted for Confounders

66

REFERENCES

67

REFERENCES

1. World Health Organization. World Health Organization. [Online]. [cited 2016 12 16.
Available from: http://www.who.int/dietphysicalactivity/childhood/en/.
2. Center for Disease Control. Center for Disease Control. [Online].; 2016 [cited 2016.
Available from: https://www.cdc.gov/obesity/childhood/index.html.
3. Ogden C. Trends In obesity prevalence mong children and adolescents in the United States,
1988-1994 Through 2013-2014. JAMA. 2016; 215(21): p. 2292-2299.
4. Daniels S. Complications of Obesity in Children and Adolescence. International Journal of
Obesity. 2009; 33: p. S60-S65.
5. Center for Disease Control. Centers for Disease Control and Prevention. [Online].; 2013
[cited 2016. Available from: https://www.cdc.gov/nchs/data/databriefs/db131.htm.
6. Zhang S. Maternal Obesity and the Early Origins of Childhood Obesity: Weighing up the
benefits and costs or maternal weight loss in the periconceptional period of the offspring.
Experimental Diabetes Research. 2011.
7. Lessa N, Berard A. Update on Prepregnancy Maternal Obesity: Birth Defects and Childhood
Outcomes. J Pediatr Genet. 2015; 4: p. 71-83.
8. Li C. Additive Interactions of Maternal Prepregnancy BMI and Breast-feeding on
Childhood Overweight. Obestiy research. 2005; 13(2): p. 362-371.
9. Kuczmarski RJ, Ogden CL, Guo SS. 2000 CDC Growth Charts for the United States:
Methods and Development. National Center for Health Statistics. 2002; 11(246).
10. Barlow S. Expert Committee Recommendations Regarding the Prevention, Assessment and
treatment of Child and Adolescent Overweight and Obesity: Summary Report. Pediatrics.
2007; 120(4).
11. Morrison K. Association of Depression and Health Related Quality of Life with body
Composition in CHildren and youth with Obesity. Journal of Affective Disorders. 2015
February; 172: p. 18-23.
12. Must A. Morbidity and Mortality Associated with Elevated Body Weight in Childre and
Adolescents. The American Journal of Clinical Nutrition. 1996; 63(3).
13. Whitaker R. Predictuing preschooler obesity at birth: the role of maternal obesity in early
pregnancy. Pediatrics. 2004; 114(1).
14. Shankar K, Harrell A, Liu X. Maternal Obesity at Conception Programs Obesity in the
Ofsspring. American Journal of Physiology. 2008; 294(2).

68

15. Galliano D, Bellver J. Female Obesity:short and long term consequences on the offspring.
Gynecological Endocrinology. 2013; 29(7).
16. Fuemmeler B, Wang L, Edwin IS. Size Tempo and Velocity of Infant Growth: Analysis of
Newborn Epigenetic Study Cohort. Childhood Obesity. 2016; 12(3): p. 210-218.
17. Li A, Teo K, Morrison K. A Genetic Link Between Prepregnancy Body Mass Index,
Postpartum Weight Retention and Offspring Weight in Early Childhood. Obesity. 2017; 25:
p. 236-243.
18. Kaar J, Crume T, Brinton J. Maternal Obesity, Gestational Weight Gain and Offsrping
Adiposity: The Exploring Perinatal Outcomes among Children Study. J Pediatr. 2014; 165:
p. 509-15.
19. Adriette J, Oostvogels JM, Stronks K. Maternal Prepregnancy BMI, Offspring's Early
Postnatal Growth, and Metabolic Profile at Age 5-6 Years: the ABCD Study. JCEM. 2014;
99(10): p. 3845-3854.
20. Sridhar S, Darbinian J, Ehrlich S. Maternal gestational weight gain and offspring risk for
childhood overweight or obesity. Am J obstet Gynecol. 2014; 211.
21. Center for Disease Control. Weight Gain During Pregnancy. [Online].; 2016. Available
from: https://www.cdc.gov/reproductivehealth/maternalinfanthealth/pregnancy-weightgain.htm.
22. Oken E, Taveras E, Kleinman K. Gestational Weight gain and child adiposity at age 3 years.
Am J Obstet Gynecol. 2007; 196.
23. Olson C, Demment M, Carling S. Associaitons Between Mothers and Their Children's
Weights at 4 years of age. Childhood Obesity. 2010; 6(4).
24. Von Kries R, Ensenauer R, Beyerlein A. Gestational Weight Gain and overweight in
children: results from the cross sectional German KiGGS study. Internationaly Journal of
Pediatric Obesity. 2011; 6: p. 45-52.
25. Diesel JC, Eckhardt CL, Day NL. Gestational Weight Gain and the risk of offspring obesity
at 10 and 16 years: a prospective cohort study in low income women. International Journal
of Obsterics and Gynaecology. 2015 September; 122(10).
26. Robinson C, Cohen A, Rehkopf D. Pregnancy and post-delivery maternal weight changes
and overweight in preschool children. Preventive Medicine. 2014 March; 60: p. 77-82.
27. American Academy of Pediatrics. [Online].; 2017 [cited 2017. Available from:
https://www2.aap.org/breastfeeding/faqsBreastfeeding.html.
28. Yan J, Liu L, Zhu Y. The association between breastfeeding and childhood obesity: a metaanalysis. BMC Public Health. 2014 December; 14.

69

29. Center of Disease Control. Children's BMI Tool for Schools. [Online].; 2015. Available
from:
https://www.cdc.gov/healthyweight/assessing/bmi/childrens_BMI/tool_for_schools.html.
30. Hayes A. Introduction to Mediation, Moderation, and Conditional Process Analysis New
York: The Guildford Press; 2013.
31. Iessa N. Update on Prepregnancy Maternal Obesity: Birth Defects and Childhood
Outcomes. J Pediatr Genet. 2015; 4: p. 71-83.

70