Received: 5 October 2017 Revised: 30 July 2018 Accepted: 15 September 2018

DOI: 10.1002/ajhb.23188

American Journal of Human Biology

ORIGINAL RESEARCH ARTICLE

Body mass index trajectories during the first year of life and their
determining factors

Junxiu Liu1,2 | Jihong Liu1 | Edward A. Frongillo Jr.3 | Nansi S. Boghossian1 | Bo Cai1 | Haiming
Zhou4 | Linda J. Hazlett1
and important and preventable prenatal and postnatal risk
ention programs.
1Department of Epidemiology and Biostatistics,
Arnold School of Public Health, University of South
Carolina, Columbia, South Carolina

2Friedman School of Nutrition Science and Policy, Tufts

University, Boston, Massachusetts

3Department of Health Promotion, Education, and

Behavior, Arnold School of Public Health, University
| INTRODUCTION
of South Carolina, Columbia, South Carolina

4Division
Obesity is a significant and global public health problem.
of Statistics - Northern Illinois
University, DeKalb, Illinois ecause obesity is very difficult to treat, prevention is criti- cal.
A new paradigm for prevention has emerged in recent years,
Correspondence Junxiu Liu, University of South
Carolina, 915 Greene Street, Columbia, SC 29208; and ocusing on the first 1000 days of life, from a woman's
Friedman School of Nutrition Science and Policy, regnancy until a child's second birthday (UNICEF, 2018). This
Tufts University, Boston, MA. Email:
junxiuliu99@gmail.com
aradigm evolved from the notion that environmental factors in
Objective: The purpose of this study was to examine the trajectories of bodytero and in early life have a profound influence on lifetime
x (BMI) in the first year of life and their determining factors. Methods: We ealth (Gillman, 2005; Oken & Gillman, 2003). Many putative
rom the Infant Feeding Practices Survey II restricted to children with 2 or sk factors in early life (up to 3 years of age) are associated
points of BMI data during follow-up visits within the first year of life (n = with an increased
ent class growth analysis was used to identify dis- tinct BMI trajectories. risk for obesity in childhood, including parental obesity, high
tinomial logistic regression, we examined the pre- natal and early life birth weight, “catch-up growth,” higher weight gain in the first
nts of the identified trajectories. Results: Three BMI trajectories were year, maternal smoking, TV watching, early adipos- ity rebound,
during the first year of life: “low- stable” (81.6%), “high-stable” (15.6%), and and short sleep duration (Ong, Ahmed, Emmett, Preece, &
8%) trajectories. Boys, pre- term infants, infants born to overweight Dunger, 2000; Reilly et al., 2005). An important limitation
Hispanic mothers, non-Hispanic Black mothers, and mothers who smoked apparent in many studies examining early life fac- tors for
gnancy were significantly more likely to have high-stable versus low-stable obesity is the lack of a developmental perspective. In most
s. Infants born to non- Hispanic Black mothers were more likely to have a studies, the obesity outcome is an assessment of body mass
us a low-stable tra- jectory. Household income ≥350% of the federal poverty index (BMI) at a single time point (Wen et al. 2014). Some
ull adherence to the guidelines of the American Academy of Pediatrics for studies assess weight at two time points, but ignore the dynamic
tfeeding exclusivity and duration reduced the likelihood of infants being in and possibly non-linear growth patterns of weight change across
versus the low-stable trajectory. Conclusion: Distinct BMI trajectories were age (Baird et al. 2005). It is unclear
early as infancy. The predic- tors of these trajectories offer information
Am J Hum Biol. 2019;31:e23188. wileyonlinelibrary.com/journal/ajhb © 2018 Wiley Periodicals, Inc. 1 of 11 https://doi.org/10.1002/ajhb.23188
whether some BMI trajectories develop in infancy, or whether
certain BMI trajectories are associated with an increased risk of
.1 | Study population
obesity later in life.
In recent years, latent growth modeling approaches Data came from the Infant Feeding Practice Survey (IFPS) II, a
have received more attention due to advances in statistical soft-ongitudinal survey of US mothers of healthy single- tons, who
ware and analytical packages. These methods are particu- larly were recruited in their third trimester and fol- lowed throughout
useful for identifying homogeneous subpopulations with similarhe first year of their baby's life. During the follow-up, a total of
growth patterns while allowing for non-linear trajectories (Jung0 mail questionnaires were sent to mothers at approximately 1,
& Wickrama, 2008; stbye, Malhotra, & Landerman, 2010)., 3, 4, 5, 6, 7, 9, 10.5, and
Several studies have characterized trajec- tories of BMI during 12 months after the infant's birth. The design of the IFPS II is
early, middle, and later childhood through modeling BMI (Ford, described in detail elsewhere (Ding et al., 2006). In brief, the
Martorell, Mehta, Ramirez- Zea, & Stein, 2016; Magee, Caputi, IFPS II was conducted by the Food and Drug Adminis- tration
& Iverson, 2013; Pryor et al., 2011; Slining, Herring, Popkin, and Centers for Disease Control and Prevention between 2005
Mayer-Davis, & Adair, 2013; van Rossem et al., 2014; Ventura, and 2007 through a consumer-opinion mail panel of about 500
Loken, & Birch, 2009), BMI z-score (Carter, Dubois, Tremblay, 000 households. To qualify for the IFPS II, mothers had to be at
Tal- jaard, & Jones, 2012; Giles et al., 2015; Haga et al., 2012; least 18 years old with a gestational duration of at least 35
Ziyab, Karmaus, Kurukulaaratchy, Zhang, & Arshad, 2014), weeks and newborns' birth weight of at least 2.25 kg. The
BMI percentage (Lane, Bluestone, & Burke, 2013), or a response rates for each postnatal ques- tionnaire varied from
dichotomous indicator for being overweight or obese (Li, Goran, 63% to 83%. The analytical sample for this study included those
Kaur, Nollen, & Ahluwalia, 2007; Mustillo et al., 2003). Among children with 2 or more time points of BMI data during the
these published studies, a few studies (Ford et al., 2016; Giles et measurement visits at birth, 3, 5, 7, and 12months of age
al., 2015; Slining et al., 2013) included at least two time points (further described in Figure 1).
for BMI data in the first year of life and these studies were
conducted in different infant popula- tions including Australia, 2.2 | Outcome—BMI trajectory during the first year of life
Philippines, and Guatemalan. Thus, evidence is still limited in
terms of identifying BMI trajectories in infancy. To further Through postnatal questionnaires, the mothers were asked to
guide prevention efforts in “the 1000 days” program, there is a provide their infant's weight and length as measured at their
need for more data about distinct BMI trajectories in the first most recent doctor's visit and their infant's ages at each of the
year of life. visits. BMI was calculated by dividing an infant's weight in
Infant growth does not happen in isolation, and kilograms by the square of length in meters. To identify
distinct BMI trajectories during infancy might contribute to biologically implausible values of BMI, we calculated sex- and
future risks of adverse health consequences (Barker, 1995; Liu age-specific BMI z-scores using SAS programs accord- ing to
et al., 2017). Our study (Liu et al., 2017) and others (Jones- the 2006 WHO Growth Charts (CDC, 2018). Further, we
Smith, Fernald, & Neufeld, 2007) indicated that high-stable and excluded children who had BMI z-scores <−5 or >5, considered
accelerated BMI trajectory groups had higher odds of obesity to be biologically implausible values (CDC, 2018).
later in life. BMI trajectories might capture the human body's Z scores are optimal for assessing children's growth sta-
dynamic physiological process. Therefore, understanding tus at a single occasion, but they are not good for assessing
potential determinants for BMI trajectories during infancy is changes over time because z-scores are calculated based on
important. The aim of this study was to identify distinct BMI external reference populations (Leroy, Ruel, Habicht, &
trajectories among children during their first year of life, and to Frongillo, 2015). The best scales for measuring weight changes
examine the relative influence of potential determining factors are BMI or BMI percentage (Cole, Faith, Pietro- belli, & Heo,
on the identified trajectories. 2005; Hall & Cole, 2006). Multiple studies (Magee et al., 2013;
Pryor et al., 2011; van Rossem et al., 2014; Ventura et al., 2009)
have utilized BMI to identify BMI trajectories. Details about
2 | METHODS how to use repeated measure- ments of BMI to identify distinct
 BMI trajectories during the first year of life are presented belowand behavioral factors. Our purpose was to identify the
determining factors that characterize sub-groups at risk for
accelerated BMI trajectories, so the determining factors were
2.3 | Exposures—Potential determining factors
categorized accordingly. For example, we categorized two
Based on prior literature (Giles et al., 2015; Pryor et al., 2011factors: meeting the professional guidelines for breastfeeding
we considered various determining factors, includinpractice and gestational weight gain. Similarly, because ges-
sociodemographic characteristics, maternal medical, antational weight gain guidelines vary by pre-gravida
behavioral factors during pregnancy, and infant demographi
2 of 11 LIU ET AL . American Journal of Human Biology
Household income was categorized as <185%, 185%-350%, and
350%.
Information on mother's age at childbirth (18-24,
5-34, and ≥ 35), smoking status during pregnancy (yes vs. no),
estational diabetes (yes vs. no), and maternal pre-gravid weight
nd height was self-reported in the prenatal question- naire.
Gestational duration reported in the neonatal question- naire was
ategorized as <37weeks or ≥37 weeks. Maternal BMI was
alculated by dividing maternal pre- gravid weight in kilograms
y the square of height in meters and was categorized as
nderweight (BMI < 18.5), normal weight (BMI: 18.5-24.99),
verweight (BMI: 25.0-29.9), or obese (BMI ≥ 30.0).
Gestational weight gain was self- reported by women when the
nfant was approximately
3 weeks old. Given that total gestational weight gain varies by
gestational age, we calculated the gestational weight gain
adequacy ratio according to the 2009 Institute of Medicine
(IOM) guidelines (Yaktine & Rasmussen, 2009) and gesta-
tional weight gain was further categorized as inadequate,
adequate, and excessive. The detailed calculation procedures by
maternal pre-gravid status are provided elsewhere (Bodnar,
Siega-Riz, Simhan, Himes, & Abrams, 2010; Liu et al., 2014).
Infant characteristics considered were sex, birth weight,
feeding practices, and age at solid food introduction. Both infant
sex and birth weight were reported by the mothers in the
BMI (categorical), maternal pre-gravida BMI was also cate-
neonatal questionnaire. Birth weight was categorized as low
gorized. Furthermore, categorical maternal pre-gravida BMI
birth weight (<2500 g), normal birth weight (2500-3999 g), and
also had the advantage of allowing non-linear relationships
high birth weight (≥4000 g). Infant feeding practices were
between the predictor and the BMI trajectories.
collected using multiple postnatal sur- vey questionnaires during
Data on maternal race/ethnicity (non-Hispanic white, the first year of life. Breastfeeding was expressed as adherence
non-Hispanic black, Hispanic, and non-Hispanic other), mar- to the 2005 American Academy of Pediatrics guidelines
ital status (married vs unmarried), education level (high school (Sharma, Dee, & Harden, 2014): never initiated breastfeeding,
or less, some college, and college graduate or higher), and initiated breastfeeding but did not exclusively breastfeed for ≥4
household income were collected in the demographic months, adhered to exclu- sivity for ≥4 months but breastfeeding
questionnaire. Household income was expressed as the per- duration <12 months, and adhered to both exclusive
centage of the federal poverty level (FPL), which is the ratio of breastfeeding for ≥4 months and breastfeeding ≥12 months (Lin,
annual family income to the appropriate poverty-threshold Tu, & Zhu, 2005; Schwalfenberg, Genuis, & Rodushkin, 2013).
values used by the US Census Bureau (Bureau, 2016). Age at solid food introduction was categorized as <4, 4 to <6,
 and ≥6 months.
LIU ET AL . 3 of 11 American Journal of Human Biology

FIGURE 1 Flow-chart for deriving sample size

ajectories using serial mea- surements of BMI at birth, 3, 5, 7,
nd 12 months.
The conventional model is a random-effects model
Raudenbush & Bryk, 2002), which uses a single mean trajec-
ory to describe the BMI trajectories of all individuals and
llows intercept, slope, and quadratic coefficient for a qua-
ratic curve to be random and to vary across individuals. In
ontrast, growth mixture modeling allows for different groups of
2.4 | Statistical analyses ndividual trajectories to vary around different mean curves,
esulting in separate growth models for each latent class, with
Growth mixture modeling is a mixture of conventional modelsach class having its unique estimates of average growth factors.
commonly used to classify individuals into distinct groups based
Growth mixture modeling, often called a person-oriented
on the individual growth trajectory patterns, where the mixturepproach with a focus on describing the rela- tionships among
corresponds to latent trajectory classes. (Jung & Wickrama,ndividuals, ideally fits our goal of classify- ing individuals into
2008; Muthén, 2004; Muthén & Muthén, 2000) Our aim was toifferent groups based on their individual
capture unobserved subpopu- lations (ie, latent groups) of We first examined the functional shapes of trajectories
infants who shared similar BMI trajectories during their first by fitting both linear and quadratic models with different
year of life. Thus, growth mix- ture modeling (Jung & numbers of latent classes (Table 1). The results showed that
Wickrama, 2008) was chosen to identify group-based growth
quadratic models had much smaller Bayesian informatiointerpretability of the num- ber of classes, and the sample size in
criteria (BIC) values across all models. Therefore, furthea class.
analyses were based on the quadratic curve assumption. We nex We first examined the functional shapes of trajectories
began with the 2-class model and added an extra class until thby fitting both linear and quadratic models with different
best fitting model was found. Here, the best fitting model wanumbers of latent classes (Table 1). The results showed that
the one with the lowest BIC value and a signifi- cant P-value foquadratic models had much smaller Bayesian information
the Lo-Mendell-Rubin (LMR) likelihood ratio test. The LMR criteria (BIC) values across all models. Therefore, further
compared the current model against the model with one lesanalyses were based on the quadratic curve assumption. We next
class, and a significant P-value indi- cated that the currenbegan with the 2-class model and added an extra class until the
model was better (Lo, Mendell, & Rubin, 2001). Other criteribest fitting model was found. Here, the best fitting model was
we used to select a better model included high entropy value, ththe one with the lowest BIC value and a signifi- cant P-value for
interpretability of the num- ber of classes, and the sample size ithe Lo-Mendell-Rubin (LMR) likelihood ratio test. The LMR
a class. compared the current model against the model with one less
After identifying BMI trajectories, multinomial logisticlass, and a significant P-value indi- cated that the current
regression models were used to quantify the relative influ- encemodel was better (Lo, Mendell, & Rubin, 2001). Other criteria
of potential determining factors on the distinctiveness of thwe used to select a better model included high entropy value, the
identified BMI trajectories. Due to the small number of womeinterpretability of the num- ber of classes, and the sample size in
whose pre-gravid BMI was in the underweight category (n a class.
ABLE 1 Indices of goodness-of-fit for latent class growth analysis
68), we combined these women with those in the normal weigh
olution
category. All regression analyses were performed using SA
response patterns. M-plus version 5.1 was used for all growth
version 9.4. Significance was deter- mined at the P < .05 level.
mixture modeling analyses (Muthén & Muthén, 2010).

Model Model fit

3 | RESULTS

3.1 | BMI trajectories

The study sample consisted of half boys (49.4%) and half girl
(50.6%). The combined sample was used to estimate the overa
BMI trajectories during the first year of life to increas
statistical power. We compared the goodness-of-fit
Smallest
prop.

We first examined the functional shapes of trajectorie

by fitting both linear and quadratic models with differen
numbers of latent classes (Table 1). The results showed tha
quadratic models had much smaller Bayesian informatioLinear BIC 35 898 35 860 35 871 35 883
criteria (BIC) values across all models. Therefore, furthe Quadratic BIC 33 711 33 561 33 493 33 485
analyses were based on the quadratic curve assumption. We nex Entropy 0.718 0.733 0.552 0.598
began with the 2-class model and added an extra class until th LMR P-value 0.0000 0.0000 0.0346 0.3054
best fitting model was found. Here, the best fitting model wa We first examined the functional shapes of trajectories
the one with the lowest BIC value and a signifi- cant P-value foby fitting both linear and quadratic models with different
the Lo-Mendell-Rubin (LMR) likelihood ratio test. The LMR numbers of latent classes (Table 1). The results showed that
compared the current model against the model with one lesquadratic models had much smaller Bayesian information
class, and a significant P-value indi- cated that the currencriteria (BIC) values across all models. Therefore, further
model was better (Lo, Mendell, & Rubin, 2001). Other criterianalyses were based on the quadratic curve assumption. We next
we used to select a better model included high entropy value, thbegan with the 2-class model and added an extra class until the
 best fitting model was found. Here, the best fitting model wamodel was better (Lo, Mendell, & Rubin, 2001). Other criteria
the one with the lowest BIC value and a signifi- cant P-value fowe used to select a better model included high entropy value, the
the Lo-Mendell-Rubin (LMR) likelihood ratio test. The LMR interpretability of the num- ber of classes, and the sample size in
compared the current model against the model with one lesa class.
class, and a significant P-value indi- cated that the curren We first examined the functional shapes of trajectories
model was better (Lo, Mendell, & Rubin, 2001). Other criteriby fitting both linear and quadratic models with different
we used to select a better model included high entropy value, thnumbers of latent classes (Table 1). The results showed that
interpretability of the num- ber of classes, and the sample size iquadratic models had much smaller Bayesian information
a class. criteria (BIC) values across all models. Therefore, further
We first examined the functional shapes of trajectorieanalyses were based on the quadratic curve assumption. We next
by fitting both linear and quadratic models with differenbegan with the 2-class model and added an extra class until the
numbers of latent classes (Table 1). The results showed thabest fitting model was found. Here, the best fitting model was
quadratic models had much smaller Bayesian informatiothe one with the lowest BIC value and a signifi- cant P-value for
criteria (BIC) values across all models. Therefore, furthethe Lo-Mendell-Rubin (LMR) likelihood ratio test. The LMR
analyses were based on the quadratic curve assumption. We nexcompared the current model against the model with one less
began with the 2-class model and added an extra class until thclass, and a significant P-value indi- cated that the current
best fitting model was found. Here, the best fitting model wamodel was better (Lo, Mendell, & Rubin, 2001). Other criteria
the one with the lowest BIC value and a signifi- cant P-value fowe used to select a better model included high entropy value, the
the Lo-Mendell-Rubin (LMR) likelihood ratio test. The LMR interpretability of the num- ber of classes, and the sample size in
compared the current model against the model with one lesa class.
class, and a significant P-value indi- cated that the curren We first examined the functional shapes of trajectories
model was better (Lo, Mendell, & Rubin, 2001). Other criteriby fitting both linear and quadratic models with different
we used to select a better model included high entropy value, thnumbers of latent classes (Table 1). The results showed that
interpretability of the num- ber of classes, and the sample size iquadratic models had much smaller Bayesian information
a class. criteria (BIC) values across all models. Therefore, further
We first examined the functional shapes of trajectorieanalyses were based on the quadratic curve assumption. We next
by fitting both linear and quadratic models with differenbegan with the 2-class model and added an extra class until the
numbers of latent classes (Table 1). The results showed thabest fitting model was found. Here, the best fitting model was
quadratic models had much smaller Bayesian informatiothe one with the lowest BIC value and a signifi- cant P-value for
criteria (BIC) values across all models. Therefore, furthethe Lo-Mendell-Rubin (LMR) likelihood ratio test. The LMR
analyses were based on the quadratic curve assumption. We nexcompared the current model against the model with one less
began with the 2-class model and added an extra class until thclass, and a significant P-value indi- cated that the current
best fitting model was found. Here, the best fitting model wamodel was better (Lo, Mendell, & Rubin, 2001). Other criteria
the one with the lowest BIC value and a signifi- cant P-value fowe used to select a better model included high entropy value, the
the Lo-Mendell-Rubin (LMR) likelihood ratio test. The LMR interpretability of the num- ber of classes, and the sample size in
compared the current model against the model with one lesa class.
class, and a significant P-value indi- cated that the curren
4 of 11 LIU ET AL . American Journal of Human Biology
FIGURE 2 The identified body mass index (BMI) trajectory groups during the first year of life among 2322 infants from the infant feeding practices survey II
study, 2005-2007. Square lines represent observed values, and triangle lines represent expected values. Dashed line corresponds to high-stable trajectory; solid
line corresponds to low-stable trajectory; and dotted line corresponds to rising trajectory
indices among both linear and quadratic models with differ- entestational weight gain, marital status, education, parity,
numbers of classes (Table 1). Model 3 had relatively smallerestational dia- betes and infant's birth weight, maternal
BIC, with a significant Lo-Mendell-Rubin P-value, and aharacteristics asso- ciated with an increased risk of being in the
reasonable value of entropy. Hence, the three-class model withigh-stable versus low-stable trajectory were: overweight
quadratic terms was chosen as our final model used to compute mother (adjusted odds ratio [AOR] = 1.36, 95% CI: 1.02-1.81),
class memberships. His- panic mother (AOR = 1.92, 1.20-3.07), non-Hispanic black
The three BMI trajectories were identified in the firstmother (AOR=3.06, 1.83-5.14), multiparous mother (AOR =
year of life and labeled as “low-stable” (81.6% of infants),.37, 1.01-1.85), or mother who smoked during
“high-stable” (15.6%), and “rising” (2.8%) trajectories (Figure pregnancy (AOR = 1.48, 1.00-2.22) (Table 3). Additionally,
2). The low-stable trajectory could be visualized as a low rate of children's characteristics associated with high-stable trajec- tory
BMI increase from birth until 7 months, and then a relatively included preterm birth (AOR = 2.26, 1.42-3.61) and male
flat BMI throughout the remaining first year of life. Infants gender (AOR = 1.52, 1.20-1.93).
classified in the high-stable trajectory as com- pared with Birth to a non-Hispanic black mother (AOR = 3.47,
low-stable group had a much higher rate of increase in the first 7 1.24-9.72) was a risk factor for being in the rising versus
months of life and then a modest decrease between month 7 and low-stable trajectory. Full adherence to the guidelines of the
month 12. The rising trajec- tory had a similar rate of increase in American Academy of Pediatrics for both breastfeeding
the first 3-4 months as the low-stable group, but maintained the exclusivity and duration was associated with lower odds of the
same rate of increase throughout the first year of life. Due to its infant being in the rising (AOR = 0.17, 0.05-0.57) com- pared to
stable increase, this curve crossed with the low-stable trajectory the low-stable trajectory. Infants from households whose
at month 3 and with the high-stable trajectory at month 9. At the percentage of the FPL was >350% had lower odds of being in
end of the first year, the mean BMI for the rising trajec- tory was the rising BMI trajectory (AOR = 0.35, 0.15-0.85) versus in the
higher than that for the high-stable trajectory and low-stable low-stable category.
group.

4 | DISCUSSION
3.2 | Predictors of BMI trajectories

Maternal race/ethnicity, marital status, education, household Infancy is a critical period of child development due to the
income, smoking status during pregnancy, adherence to occurrence of rapid weight gain and altered adiposity, which
breastfeeding guidelines, and child's sex were significantly might shift a child's growth trajectory toward a more obese
related to BMI trajectories in bivariate analyses (Table 2). In phenotype in childhood and further into adulthood. Using
brief, children in the low-stable BMI trajectory were more likely growth mixture models, we identified three distinct patterns of
to be born to non-Hispanic white mothers, living in households BMI trajectory through the first year of life. Eighty-one percent
with mothers having a college education or more, households of children in our sample had a low-stable BMI tra- jectory in
with >350% of the FPL, or mothers who did not smoke during infancy, while over 15% had a high-stable BMI trajectory, and
pregnancy, had a longer duration of breastfeeding, or were girls. about 3% were in the rising BMI trajectory. Studies on the BMI
Those in high-stable BMI tra- jectories were more likely to be trajectories during the first 1000 days, especially the first year of
children born to non- Hispanic black or Hispanic mothers, life, are extremely limited. Thus, the novelty of this study is to
unmarried mothers, mothers with some college education or add to literature to show that distinct BMI trajectories start in
education at high school level or below, mothers who smoked the first year of life, espe- cially during the first 6 months of life.
during preg- nancy, or were boys. Finally, those in the Further, our findings also provide information about the
rising-BMI- trajectory were more likely to be non-Hispanic characteristics of children in different BMI trajectories, which
blacks and non-Hispanic others, living in poor households might be used in design- ing future programs to reduce the risk
(<185% of the FPL), born to mothers who smoked, or who of some high-risk chil- dren being in high-risk groups.
never initi- ated breastfeeding. A number of studies have attempted to identify distinct
After adjusting for maternal age at delivery, BMI trajectories among children using latent class analysis.
 These studies were primarily conducted in developed couninfancy by including at least two data points in the first year of
tries, namely the Netherlands (van Rossem et al., 2014), Japalife (Ford et al., 2016; Giles et al., 2015; Slining et al., 2013).
(Haga et al., 2012), the UK (Ziyab et al., 2014), Australia (GileThe number of trajectories identified ranged from three to seven,
et al., 2015; Magee et al., 2013), Canada (Carter et al., 2012but in most cases four trajectories were found. To our
Pryor et al., 2011; Tu, Masse, Lear, Gotay, & Richardson, 2015knowledge, the current study is the first one to identify latent
and the United States (Lane et al., 2013; Li et al., 2007; MustillBMI trajectories during infancy among American children.
et al., 2003; Ventura et al., 2009). Age ranges covered in thesCompared to other studies (Ford et al., 2016; Giles et al., 2015;
studies varied, and a few studies examined the BMI trajectory iSlining et al., 2013), our study had a larger sample
LIU ET AL . 5 of 11 American Journal of Human Biology
6 of 11 LIU ET AL . American Journal of Human Biology
TABLE 2 Basic maternal and infant characteristics according to the identified BMI trajectories: The infant feeding practices survey II study,
2005-2007, United States
Total sample (n, %)
BMI trajectory group (n, %)
Characteristics
Low-stable High-stable Rising
(n = 2322) (n = 1895) (n = 363) (n = 64) P value Maternal characteristics
Age (year)
18-24 452 (19.5) 353 (18.7) 85 (23.4) 14 (21.9)
25-34 1473 (63.5) 1206 (63.7) 226 (62.3) 41 (64.1) 0.187
≥35 395 (17.0) 334 (17.6) 52 (14.3) 9 (14.1)
Race/ethnicity
Non-Hispanic white 1947 (83.9) 1630 (86.1) 269 (74.1) 48 (75.0)
Non-Hispanic black 87 (3.8) 54 (2.9) 28 (7.7) 5 (7.8)
Hispanic 122 (5.3) 92 (4.9) 27 (7.4) 3 (4.7) <0.001
Others 164 (7.1) 117 (6.2) 39 (10.7) 8 (12.5)
Married, % 1791 (77.2) 1474 (77.9) 266 (73.3) 51 (79.7) 0.046
Education, %
High school or less 404 (18.5) 315 (17.6) 76 (23.2) 13 (22.4)
Some college 845 (38.8) 688 (38.4) 133 (40.6) 24 (41.4) 0.038
College graduate or higher 930 (42.7) 790 (44.1) 119 (36.3) 21 (36.2)
Income (% of poverty threshold)
<185% 892 (38.5) 715 (37.8) 143 (39.4) 34 (53.1)
185%-350% 848 (36.6) 688 (36.3) 138 (38.0) 22 (34.4) 0.047
>350% 580 (25.0) 490 (25.9) 82 (22.6) 8 (12.5)
Parity, %
Primiparous 695 (29.9) 583 (30.8) 97 (26.7) 15 (23.4) 0.157
Multiparous 1627 (70.1) 1312 (69.2) 266 (73.3) 49 (76.6)
Maternal smoking, %
Yes 193 (8.4) 145 (7.7) 41 (11.3) 7 (11.1) 0.048
No 2119 (91.7) 1742 (92.3) 321 (88.7) 56 (88.9)
Prepregnancy BMI status, %
BMI < 25 kg/m2 1046 (45.6) 866 (46.3) 154 (43.3) 26 (40.6) 0.256
BMI: 25.0-29.9 kg/m2 634 (27.7) 499 (26.7) 114 (32.0) 21 (32.8)
BMI ≥ 30 kg/m2 612 (26.7) 507 (27.1) 88 (24.7) 17 (26.6)
Gestational weight gain, %
Inadequate 404 (17.4) 318 (16.8) 72 (19.8) 14 (21.9)
Adequate 805 (34.7) 671 (35.5) 115 (31.7) 19 (29.7) 0.390
Excessive 1111 (47.9) 904 (47.8) 176 (48.5) 31 (48.4)
Gestational diabetes, % 149 (6.5) 117 (6.2) 28 (7.8) 4 (6.3) 0.744
Adherence to breastfeeding guideline, %
Never initiated breastfeeding 313 (13.5) 246 (13.0) 53 (14.6) 14 (21.9)
Initiated breastfeeding, did not exclusively breastfeed for
≥4 months
1556 (67.0) 1260 (66.5) 252 (69.4) 44 (68.8) 0.044
Adherence to exclusivity for ≥4 months, breastfeeding
duration<12 months
102 (4.4) 83 (4.4) 17 (4.7) 2 (3.1)
Adherence to both exclusivity for ≥4 months and breastfeeding
for ≥12 months
351 (15.1) 306 (16.2) 41 (11.3) 4 (6.3)
Offspring characteristics
Infant gender, %
Boy 1147 (49.4) 902 (47.6) 214 (59.0) 31 (48.4) <0.001
Girl 1175 (50.6) 993 (52.4) 149 (41.1) 33 (51.6)
Infant's birth weight, %
Birth weight < 4000 g 2043 (88.0) 1662 (87.7) 324 (89.3) 57 (89.1) 0.678
Birth weight ≥ 4000 g 279 (12.0) 233 (12.3) 39 (10.7) 7 (10.9)
(Continues)
size (n = 2230) and included a national sample. We should be cautious in making a comparison between the
trajectories found by other studies such as Giles et al. (2015) and our study because Giles et al. (2015) examined the
BMI z-score trajectories and covered the time period from birth up to 3.5 years of age. But, similar to Giles et al.
(2015), we also found that the trajectories were generally characterized by the postnatal growth in the first 5 months
of life. Because we restricted data to a group of relatively healthy infants at births, we did not see many differences
at birth.
It is important for clinicians and policy-makers to be aware of the importance of growth trajectories in the first 6
months of life. A few studies that included one measure of size at birth or during the first year of life (Haga et al.,
2012; Pryor et al., 2011), also confirmed that the trajectories usually begin in infancy and continue into mid- or late
childhood. In our sepa- rate study (Liu et al., 2017), our group found that compared with those in the low-stable
trajectory, the children in the high- stable trajectory had a 1.8 times higher risk of being obese at 6 years of age,
while those in rising trajectory were not differ- ent. This might be due to our small sample size.
Because BMI trajectories begin in infancy and predict childhood obesity, it is important to characterize which
infants are more likely to be in one trajectory versus the other. Our study also provided useful information on this.
We found the most important factors associated with a high- stable trajectory in our sample, in descending order,
were non-Hispanic black ethnicity, preterm delivery, Hispanic ethnicity, male sex, smoking during pregnancy,
multiparity, and maternal overweight status. The important factors in descending order for having a rising BMI
trajectory were failure to meet breastfeeding guidelines, non-Hispanic black ethnicity, and a household income of
<350% of the FPL.
Our finding of infants of non-Hispanic black or Hispanic mothers having higher odds of being in the high-stable and
rising BMI trajectories, as compared with those born to non- Hispanic white mothers, is in line with prior studies
(Martinson, McLanahan, & Brooks-Gunn, 2015). Our study identified higher household income as a protective
factor against being in the rising BMI trajectory, which is somewhat consistent with prior studies reporting maternal
or parental education as determining factors for an offspring's BMI trajec- tory (Magee et al., 2013). Similar to a
prior study, (Giles
TABLE 2 (Continued)
Total sample (n, %)
BMI trajectory group (n, %)
Characteristics
Low-stable High-stable Rising
(n = 2322) (n = 1895) (n = 363) (n = 64) P value Age at solid food introduction
<4 months 1040 (46.6) 824 (45.4) 186 (53.0) 30 (46.7)
4 to <6 months 1019 (45.7) 846 (46.6) 146 (41.6) 27 (42.2) 0.060
≥6 months 173 (7.8) 147 (8.1) 19 (5.4) 7 (10.9)
Abbreviations: BMI, body mass index. P value determined using chi-square test for categorical variables or analysis of variance F test.
LIU ET AL . 7 of 11 American Journal of Human Biology
et al., 2015) multiparity was linked with higher odds of chil- dren being in the high-stable or “accelerating” BMI
trajectory. A limited number of studies have examined the influence of maternal prenatal factors on offspring's BMI
trajectories (Jedrychowski et al., 2015; Metayer et al., 2014; Ramakrishnan et al., 2016). Our finding that maternal
smoking during preg- nancy was associated with an increased risk of a child being in the high-stable BMI trajectory,
compared with non-smoking mothers, is largely consistent with previous findings (Carter et al., 2012; Haga et al.,
 2012; Li et al., 2007; Magee et al., 2013; Ziyab et al., 2014). Additionally, mothers who were overweight before
pregnancy had increased odds of their chil- dren being in the high-stable BMI trajectory as compared with mothers
who were under/normal weight. Giles et al. (2015) reported that maternal obesity in early pregnancy is the most
important factor differentiating BMI z-score trajectories. Fur- thermore, our data did not find meeting IOM
guidelines for gestational weight gain as a significant factor contributing to the divergence of the BMI trajectories
during infancy, while Li et al. (2007) reported that total gestational weight gain larger than 20.43 kg was a risk factor
associated with “early onset overweight trajectory” from age 2 years until 12 years.
In early life, breastfeeding is known as a protective factor for childhood obesity (Horta, Loret de Mola, & Victora,
2015), but its influence on BMI trajectories is not well-studied. We found that adherence to breastfeeding guidelines
was associated with reduced odds (AOR = 0.17) of being in the rising BMI trajectory compared to the low-stable
trajectory. This finding was consistent with prior studies (Carling, Demment, Kjolhede, & Olson, 2015; Li et al.,
2007; Oddy et al., 2014). Children's age at the introduction of solid food has been related to obesity in later life in
some studies (Pearce, Taylor, & Langley-Evans, 2013), yet our study did not observe a signifi- cant association of
age at solid food introduction in relation to children's BMI trajectory in their first year of life. Another study using
the same database as ours also did not find a positive link between the timing of solid food introduction and obesity
at the age of 6 years (Barrera, Perrine, Li, & Scanlon, 2016).
Due to the richness of the data source, our study was able to characterize the children who were in different BMI
trajecto- ries based on sociodemographic factors, prenatal, and postnatal factors. Yet the underlying biological or
behavioral etiology that would explain the different trajectories of BMI during infancy cannot be examined in this
study. It is possible that the
8 of 11 LIU ET AL . American Journal of Human Biology

TABLE 3 Results from adjusted multinomial logistic regression of the association between prenatal and early life factors with high-stable and rising BMI
trajectories in comparison to low-stable trajectory among 2320 children the Infant Feeding Practices Survey II study, 2005-2007, United States

Variable High-stable Rising

AOR (95% CI) AOR (95% CI)

Prepregnancy BMI status

BMI < 25 kg/m2 1 (reference) 1 (reference)

BMI: 25.0-29.9 kg/m2 1.36 (1.02, 1.81) 1.30 (0.70, 2.42)

BMI ≥30 kg/m2 0.99 (0.72, 1.35) 0.91 (0.47, 1.79)

Gestational weight gain

Inadequate 1.20 (0.85, 1.71) 1.53 (0.74, 3.18)

Adequate 1 (reference) 1 (reference)

Excessive 1.11 (0.84, 1.47) 1.22 (0.65, 2.29)

Adherence to breastfeeding guideline

Never initiated breastfeeding 1 (reference) 1 (reference)

≥4 months and breastfeeding for ≥12
Initiated breastfeeding, did not exclusively
months
breastfeed for ≥4 months
0.84 (0.51, 1.39) 0.17 (0.05, 0.57)
0.94 (0.66, 1.34) 0.58 (0.30, 1.13)

Adherence to exclusivity for ≥4 months,

Age (year), %
breastfeeding duration<12 months
1.23 (0.65, 2.35) 0.32 (0.07, 1.52)

Adherence to both exclusivity for

18-24 1.14 (0.82, 1.58) 1.05 (0.51, 2.14)

 25-34 1 (reference) 1 (reference)

35-43 0.81 (0.57, 1.14) 0.86 (0.40, 1.84)

Race/ethnicity, %

Non-Hispanic white 1 (reference) 1 (reference)

Non-Hispanic black 3.06 (1.83, 5.14) 3.47 (1.24, 9.72)

Hispanic 1.92 (1.20, 3.07) 1.07 (0.32, 3.59)

Others 2.08 (1.34, 3.22) 2.16 (0.87, 5.34)

Married or cohabiting, %

Yes 1.24 (0.87, 1.75) 1.80 (0.80, 4.03)

No 1 (reference) 1 (reference)

Education, %

High school or less 1.49 (1.02, 2.17) 1.03 (0.46, 2.32)

Some college 1.18 (0.88, 1.50) 1.00 (0.52, 1.92)

College graduate or higher 1 (reference) 1 (reference)

Income (% of poverty threshold), %

<185% 1 (reference)

185%-350% 1.22 (0.91, 1.62) 0.71 (0.39, 1.28)

>350% 1.26 (0.88, 1.81) 0.35 (0.15, 0.85)

Gestational age

<37 weeks 2.26 (1.42, 3.61) 0.44 (0.06, 3.25)

≥37 weeks 1 (reference) 1 (reference)

Parity, %

Primiparous 1 (reference) 1 (reference)

Multiparous 1.37 (1.01, 1.85) 1.08 (0.55, 2.12)

Maternal smoking

Yes 1.48 (1.00, 2.22) 1.12 (0.44, 2.82)

No 1 (reference) 1 (reference)

Gestational diabetes

Yes 1.31 (0.82, 2.07) 1.01 (0.35, 2.92)

No 1 (reference) 1 (reference)

SexBoy 1.52 (1.20, 1.93) 1.04 (0.62, 1.75)

(Continues)
 his prospective longitudinal study suggests that distinct BMI
ajectories are evident during an infant's first year of life,
specially during the first 6 months of life. This study provides
dditional evidence to support the importance of the ongoing
first 1000 days” program spearheaded by
UNICEF (2018). We were able to characterize children into
different BMI trajectories. For example, a high-stable trajec-
tory is likely put those children at increased risk for obesity at
age 6 (Liu et al., 2017). While our findings require repli- cation
in other studies, clinicians can begin to screen for some
differentiation of BMI trajectories during the first year of a
sub-groups of infants who are at higher risk for accel- erated
child's life stems from genetics, maternal conditions, behavioral growth in infancy. Efforts to prevent these children from
and socio-demographic factors, and their interactions. The transitioning to accelerated BMI trajectories in infancy will
complex determinants of early child growth need to be further ultimately help these children reduce their risk for obe- sity in
investigated using well-designed cohort studies with a large childhood, and help to reduce their risk for various
sample size, serial BMI measures, and objective measures of obesity-related chronic diseases.
infant and maternal factors such as measured weight, length/
height, total gestational weight gain, and pre-pregnancy BMI.
The IFPS II was a valuable and unique data source for CONFLICT OF INTEREST The authors declare that there is
this study because of its longitudinal prospective design and no conflict of interest.
having approximately 10 measurement visits during infancy.
Studies including >2 body size measurements to examine BMI
trajectories in infancy are limited. Additionally, IFPS II is a rich
AUTHOR CONTRIBUTIONS Junxiu Liu and Jihong Liu conceived
database which allowed us to characterize the chil- dren in
and designed the study. Junxiu Liu drafted the manuscript and
different BMI trajectories. However, our data have several
did all data analysis. All authors provided critical revision of the
potential limitations. First, the sample is not nationally
manuscript.
representative of the US population, although the sample was
well distributed throughout the US. Second, data were self-
reported by the child's mother. No medical records were ORCID
examined to verify weight, height, health status, or other med-
Junxiu Liu https://orcid.org/0000-0003-3303-0217
ical characteristics for children. Third, longitudinal attrition
across postnatal questionnaires in IFPS II may introduce
potential bias. A simulation study has shown that the magni- REFERENCES
tude of bias depends on different levels of loss to follow-up and
Baird, J., Fisher, D., Lucas, P., Kleijnen, J., Roberts, H., & Law, C. (2005). Being big
missing mechanism (Kristman, Manno, & Cote, 2004). The or growing fast: Systematic review of size and growth in infancy and later obesity.
authors reported no significant bias for 5%-60% of sub- jects BMJ, 331(7522), 929. https://doi.org/10.1136/bmj.38586. 411273.E0 Barker, D. J.
who were lost to follow-up if the missing mechanism was (1995). Fetal origins of coronary heart disease. BMJ, 311, 171–174. Barrera, C. M.,
Perrine, C. G., Li, R., & Scanlon, K. S. (2016). Age at introduc- tion to solid foods
completely at random or missing at random. However, if the and child obesity at 6 years. Childhood Obesity, 12, 188–192. Bodnar, L. M.,
missing pattern is not random, a small proportion of loss to Siega-Riz, A. M., Simhan, H. N., Himes, K. P., & Abrams, B. (2010). Severe obesity,
follow-up would introduce bias. The response rates in our study gestational weight gain, and adverse birth outcomes. The American Journal of
Clinical Nutrition, 91, 1642–1648. Bureau. 2016. US cencus. Retreived from
ranged from 64.5% to 87.6%, which are relatively high
https://www.census.gov/hhes/www/
follow-up rates. Additionally, the M-plus software handles
poverty/index.html. Accessed on May 1, 2016. Carling, S. J., Demment, M.
missing data as completely missing at random. M., Kjolhede, C. L., & Olson, C. M. (2015). Breast- feeding duration and weight gain
trajectory in infancy. Pediatrics, 135, 111–119.

5 | CONCLUSION
LIU ET AL . 9 of 11 American Journal of Human Biology
TABLE 3 (Continued)

Variable High-stable Rising

AOR (95% CI) AOR (95% CI)

Girl 1 (reference) 1 (reference)

Birth weight

Birth weight < 4000, % 1.04 (0.71, 1.52) 1.12 (0.49, 2.57)

Birth weight ≥ 4000, % 1 (reference) 1 (reference)

Age at solid food introduction

<4 months 1.15 (0.88-1.50) 0.69 (0.39-1.23)

4 to <6 months 1 (reference) 1 (reference)

≥6 months 0.75 (0.44-1.29) 1.81 (0.75-4.37)

AOR, adjusted odds ratio; CI, confidence interval.

Carter, M. A., Dubois, L., Tremblay, M. S., Taljaard, M., & Jones, B. L. (2012).51–760. Lane, S. P., Bluestone, C., & Burke, C. T. (2013). Trajectories of BMI from
Trajectories of childhood weight gain: The relative importance of local envi- ronmentarly childhood through early adolescence: SES and psychosocial predictors. Brit- ish
versus individual social and early life factors. PLoS One, 7, e47065. CDC. 2018. Aournal of Health Psychology, 18, 66–82. Leroy, J. L., Ruel, M., Habicht, J. P., &
SAS program for the 2006 WHO growth charts. Retrieved fromrongillo, E. A. (2015). Using height-for-age differences (HAD) instead of
http://www.cdc.gov/nccdphp/dnpao/growthcharts/resources/sas.htm. Accessed on eight-for-age z-scores (HAZ) for the meaningful measurement of population-level
April 23, 2018. Cole, T. J., Faith, M. S., Pietrobelli, A., & Heo, M. (2005). What is theatch-up in linear growth in children less than 5 years of age. BMC Pediatrics, 15,
best mea- sure of adiposity change in growing children: BMI, BMI%, BMI z-score or 45. Li, C., Goran, M. I., Kaur, H., Nollen, N., & Ahluwalia, J. S. (2007). Develop-
BMI centile? European Journal of Clinical Nutrition, 59, 419–425. Ding, Y. S., Yan, ental trajectories of overweight during childhood: Role of early life factors. Obesity
X. J., Jain, R. B., Lopp, E., Tavakoli, A., Polzin, G. M., ... Watson, C. H. (2006). Silver Spring), 15, 760–771. Lin, D., Tu, Y., & Zhu, L. (2005). Concentrations and
Determination of 14 polycyclic aromatic hydrocar- bons in mainstream smoke from ealth risk of polycyclic aromatic hydrocarbons in tea. Food and Chemical
U.S. brand and non-U.S. brand cigarettes. Environmental Science & Technology, 40, oxicology, 43, 41–48. Liu, J., Gallagher, A. E., Carta, C. M., Torres, M. E., Moran,
1133–1138. Ford, N. D., Martorell, R., Mehta, N. K., Ramirez-Zea, M., & Stein, A. D. , & Wilcox, S. (2014). Racial differences in gestational weight gain and
(2016). Life-course body mass index trajectories are predicted by childhood egnancy-related hypertension. Annals of Epidemiology, 24, 441–447. Liu, J. X.,
socioeconomic status but not exposure to improved nutrition during the first 1000 iu, J. H., Frongillo, E. A., Boghossian, N. S., Cai, B., & Hazlett, L. J. (2017). Body
days after conception in Guatemalan adults. The Journal of Nutrition, 146, ass index trajectories during infancy and pediat- ric obesity at 6 years. Annals of
2368–2374. Giles, L. C., Whitrow, M. J., Davies, M. J., Davies, C. E., Rumbold, A. pidemiology, 27, 708.e1–715.e1. Lo, Y., Mendell, N. R., & Rubin, D. B. (2001).
R., & Moore, V. M. (2015). Growth trajectories in early childhood, their relation- shipesting the number of compo-
with antenatal and postnatal factors, and development of obesity by age 9 years: nents in a normal mixture. Biometrika, 88, 767–778. Magee, C. A.,
Results from an Australian birth cohort study. International Journal of Obesity, 39, aputi, P., & Iverson, D. C. (2013). Identification of distinct body mass index
1049–1056. Gillman, M. W. (2005). Developmental origins of health and disease. Theajectories in Australian children. Pediatric Obesity, 8, 189–198.
New Martinson, M. L., McLanahan, S., & Brooks-Gunn, J. (2015). Variation in child body
England Journal of Medicine, 353, 1848–1850. Haga, C., Kondo, N., mass index patterns by race/ethnicity and maternal nativity status in the United States
Suzuki, K., Sato, M., Ando, D., Yokomichi, H., ... Yamagata, Z. (2012). and England. Maternal and Child Health Journal, 19, 373–380. Metayer, C., Milne,
Developmental trajectories of body mass index among Japanese children and impact E., Dockerty, J. D., Clavel, J., Pombo-de-Oliveira, M. S., Wesseling, C., ...
of maternal factors during pregnancy. PLoS One, 7, e51896. Hall, D. M. B., & Cole, Infante-Rivard, C. (2014). Maternal supplementation with folic acid and other
T. J. (2006). What use is the BMI? Archives of Disease vitamins and risk of leukemia in offspring: A childhood leukemia international
consortium study. Epidemiology, 25, 811–822. Mustillo, S., Worthman, C., Erkanli,
in Childhood, 91, 283–286. Horta, B. L., Loret de Mola, C., & Victora,
A., Keeler, G., Angold, A., & Costello, E. J. (2003). Obesity and psychiatric disorder:
C. G. (2015). Long-term conse- quences of breastfeeding on cholesterol, obesity,
Developmental tra- jectories. Pediatrics, 111, 851–859. Muthén, B. (2004). Latent
systolic blood pressure and type 2 diabetes: A systematic review and meta-analysis.
variable analysis. In The Sage handbook of quantita- tive methodology for the social
Acta Paediatrica, 104, 30–37. Jedrychowski, W. A., Perera, F. P., Maugeri, U.,
sciences (pp. 345–368). Thousand Oaks, CA: Sage Publications. Muthén, B., &
Majewska, R., Mroz, E., Flak, E., ... Jacek, R. (2015). Long term effects of prenatal
Muthén, L. K. (2000). Integrating person-centered and variable-centered analyses:
and postnatal air- borne PAH exposures on ventilatory lung function of non-asthmatic
Growth mixture modeling with latent trajectory classes. Alcoholism: Clinical and
preado- lescent children. Prospective birth cohort study in Krakow. Science of the
Experimental Research, 24, 882–891. Muthén, L. K., & Muthén, B. O. (2010). Mplus
Total Environment, 502, 502–509. Jones-Smith, J. C., Fernald, L. C., & Neufeld, L.
user's guide: Statistical analysis
M. (2007). Birth size and accel- erated growth during infancy are associated with
increased odds of childhood overweight in Mexican children. Journal of the American with latent variables: User'ss guide. Los Angeles, CA: Muthén & Muthén. Oddy, W.
Dietetic Associa- tion, 107, 2061–2069. Jung, T., & Wickrama, K. A. S. (2008). An H., Mori, T. A., Huang, R. C., Marsh, J. A., Pennell, C. E., Chivers, P. T., ... Beilin, L.
introduction to latent class growth analysis and growth mixture modeling. Social and J. (2014). Early infant feeding and adiposity risk: From infancy to adulthood. Annals
Personality Psychology Compass, 2, 302–317. Kristman, V., Manno, M., & Cote, P. of Nutrition & Metabolism, 64, 262–270. Oken, E., & Gillman, M. W. (2003). Fetal
(2004). Loss to follow-up in cohort studies: origins of obesity. Obesity Research,
How much is too much? European Journal of Epidemiology, 19, 11, 496–506. Ong, K. K., Ahmed, M. L., Emmett, P. M., Preece, M. A., &
 Dunger, D. B. (2000). Association between postnatal catch-up growth and obesity in J., & Rodushkin, I. (2013). The benefits and risks of consuming brewed tea: Beware
child- hood: Prospective cohort study. BMJ, 320, 967–971. stbye, T., Malhotra, R., of toxic element contamination. Journal of Toxicology, 2013, 370460. Sharma, A. J.,
& Landerman, L. R. (2010). Body mass trajectories through adulthood: Results fromDee, D. L., & Harden, S. M. (2014). Adherence to breastfeeding guidelines and
the National Longitudinal Survey of Youth 1979 Cohort (1981–2006). International maternal weight 6 years after delivery. Pediatrics, 134(Suppl 1), S42–S49. Slining, M.
Journal of Epidemiology, 40, 240–250. Pearce, J., Taylor, M. A., & Langley-Evans, M., Herring, A. H., Popkin, B. M., Mayer-Davis, E. J., & Adair, L. S. (2013). Infant
C. (2013). Timing of the intro- duction of complementary feeding and risk of BMI trajectories are associated with young adult body composition. Journal of
childhood obesity: A system- atic review. International Journal of Obesity, 37, Developmental Origins of Health and Disease, 4, 56–68. Tu, A. W., Masse, L. C.,
1295–1306. Pryor, L. E., Tremblay, R. E., Boivin, M., Touchette, E., Dubois, L., Lear, S. A., Gotay, C. C., & Richardson, C. G. (2015). Body mass index trajectories
Genolini, C., ... Cote, S. M. (2011). Developmental trajectories of body mass index ifrom ages 1 to 20: Results from two nationally representative Canadian longitudinal
early childhood and their risk factors: An 8-year longitudinal study. Archives of cohorts. Obesity (Silver Spring), 23, 1703–1711. UNICEF. 2018. Why 1,000 days?
Pediatrics & Adolescent Medicine, 165, 906–912. Ramakrishnan, U., Retrieved from https://thousanddays.org/
Gonzalez-Casanova, I., Schnaas, L., DiGirolamo, A., Quezada, A. D., Pallo, B. C., .. the-issue/why-1000-days/. Accessed on April 23, 2018. van Rossem, L.,
Martorell, R. (2016). Prenatal supplementa- tion with DHA improves attention at 5 yWijga, A. H., Brunekreef, B., de Jongste, J. C., Kerkhof, M., Postma, D. S., ... Smit,
of age: A randomized controlled trial. The American Journal of Clinical Nutrition, H. A. (2014). Overweight in infancy: Which pre- and perinatal factors determine
104, 1075–1082. Raudenbush, S. W., & Bryk, A. S. (2002). Hierarchical linear overweight persistence or reduction? A birth cohort followed for 11 years. Annals of
models: Applica- Nutrition & Metabolism, 65, 211–219. Ventura, A. K., Loken, E., & Birch, L. L.
tions and data analysis methods. Thousand Oaks, California: Sage. Reilly,(2009). Developmental trajectories of girls' BMI across childhood and adolescence.
J. J., Armstrong, J., Dorosty, A. R., Emmett, P. M., Ness, A., Rogers, I., ... Avon Obesity (Silver Spring), 17, 2067–2074.
Longitudinal Study of Parents and Children Study Team. (2005). Early life risk facto
for obesity in childhood: Cohort study. BMJ, 330, 1357. Schwalfenberg, G., Genuis,
10 of 11 LIU ET AL . American Journal of Human Biology
Wen, L. M., Baur, L. A., Rissel, C., Xu, H., & Simpson, J. M. (2014). Correlates of ge: Prenatal determinants and health consequences. Journal of Epidemiology and
body mass index and overweight and obesity of children aged 2 years: Findings from ommunity Health, 68, 934–941.
the healthy beginnings trial. Obesity, 22(7), 1723–1730. Yaktine, A. L., & Rasmussen, How to cite this article: Liu JX, Liu JH, Frongillo Jr.
K. M. (2009). Weight gain during pregnancy. In
EA, et al. Body mass index trajectories during the first
Reexamining the Guidelines. Washington: National Academies Press.
Ziyab, A. H., Karmaus, W., Kurukulaaratchy, R. J., Zhang, H., & Arshad, S. H.
year of life and their determining factors. Am J Hum Biol.
(2014). Developmental trajectories of body mass index from infancy to 18 years of 2019;31:e23188. https://doi.org/10.1002/ ajhb.23188
LIU ET AL . 11 of 11 American Journal of Human Biology