Association of Maternal BMI and Rapid Infant Weight Gain With Childhood Body Size and Composition

This report used data from the Healthy Start Study, an ongoing, longitudinal, observational prebirth cohort study.¹³  From 2009 to 2014, 1410 pregnant individuals aged 16 years or older were enrolled before 24 weeks of gestation from obstetrics clinics at the University of Colorado Hospital in Denver, Colorado. Exclusion criteria included multiple gestation, history of serious chronic disease, or history of stillbirth or preterm birth before 25 weeks. In-person research visits occurred during pregnancy (median, 17 and 27 weeks), birth (median, 1 day old), infancy (median, 5 months old), and early childhood (median, 5 years old). Pregnant individuals provided written informed consent at enrollment. Children provided written assent starting at age 7 years. The study was approved by the Colorado Multiple Institutional Review Board and is registered as an observational study at www.clinicaltrials.gov (NCT02273297).

Participants were eligible for inclusion in the present analysis if all of the following measurements were available: weight at birth, infancy, and early childhood study visits; body composition in early childhood; infant primary feeding mode; and gestational age at birth (Supplemental Fig 4). Medical conditions were abstracted from electronic medical records and individually reviewed to exclude participants with conditions known to significantly impact growth and/or feeding. Participants with certain medical conditions were excluded, including 22q11.2 deletion (DiGeorge) syndrome, ichthyosis congenita, phenylketonuria, and others. We further excluded those with infancy visits before age 3 months or after 7 months. Differences between the entire Healthy Start cohort and the analytical sample are presented in Supplemental Table 5.

Prepregnancy weight was obtained from medical record abstraction (91%) or from self-report at the first pregnancy visit (9%). Maternal height was measured using a stadiometer at the first pregnancy research visit. The ppBMI was calculated as prepregnancy weight (kilograms) divided by height (meters) squared. Gestational weight gain was calculated using prepregnancy weight and the last clinically measured weight during pregnancy and compared with Institute of Medicine guidelines.

For offspring, recumbent length (birth and infancy visits) or standing height (early childhood visit) were measured with a stadiometer to the nearest 0.1 cm. Weight was measured on an electronic scale to the nearest 0.1 kg. We calculated birth weight-for-gestational age z scores and categorized participants as small for gestational age (SGA) if birth weight was <10th percentile based on reference data.¹⁴  Rapid weight gain (RWG) during infancy was defined as an increase in weight-for-age z score of ≥0.67 from the birth to infancy visit (average age, 4.8 months) because this cutoff has been previously correlated with increased risk for childhood obesity.^6,15  BMI z scores at the early childhood visit (average age, 4.7 years) were determined using the Centers for Disease Control and Prevention 2000 growth charts. Child body composition was measured using air displacement plethysmography (BOD POD, COSMED, Inc.) as previously described^13,16  to assess fat mass and fat-free mass. The %FM was calculated as fat mass divided by total mass.¹³  To examine whether ppBMI or RWG were associated with differences in early childhood fat free (lean) mass, we used FFMI, defined as fat-free mass in kilograms divided by height (meters) squared. We chose to use FFMI rather than total lean mass to avoid confounding from known correlations between fat-free mass and height.^17,18 

Child sex, race, and ethnicity were determined by maternal report at the birth visit. Mode of delivery was assessed by medical record abstraction and categorized as vaginal, vaginal with forceps or vacuum, or cesarean section. Infant feeding practices, including breastfeeding and introduction of complementary solid foods, were collected by maternal self-report at the infancy visit. Child caloric (energy) intake was measured in early childhood using two 24-hour dietary recalls (1 weekday and 1 weekend day) with parents acting as proxy reporters for their children. Dietary recalls were facilitated by trained personnel from the University of North Carolina using the Nutrition Data System for Research software.¹⁹  Although there are known limitations to 24-hour recall data,²⁰  previous work showed relative precision compared with other measures of energy intake.^21,22  Dietary data were processed to obtain daily caloric intake in childhood.

All analyses were performed in the Statistical Analysis Software, version 9.4 (SAS Institute). Population characteristics were assessed using frequencies and percentages for categorical variables and means and SDs for continuous variables. Characteristics were compared between children with and without rapid weight gain using χ² tests for categorical variables and Satterthwaite t tests for continuous variables.

For each outcome, 3 general linear regression models were fit to characterize associations between ppBMI, RWG, and their interaction, and the outcomes of BMI-for-age z score, %FM, and FFMI. All hypothesis tests, including tests of interaction, were conducted using the Wald test and an α level of 0.05. Model assumptions, including linearity, were assessed by plotting jackknifed studentized residuals by predictors. Model 1 had ppBMI as the major predictor of interest. Model 2 had RWG as the major predictor of interest. Model 3 had ppBMI, RWG, and the interaction between ppBMI and RWG as the major predictors of interest.

Additional covariates included infant sex, SGA (yes or no), and the 2-, 3-, or 4-way interactions between the major predictors, sex, and an indicator variable for SGA. Child’s age at the early childhood visit was included in models of %FM and FFMI, but not in models of BMI z score, which already accounts for age. If sex significantly modified the association between the major predictors and outcome, the model was stratified by sex. Interactions and the precision covariates, age at introduction of complementary foods, and mode of delivery were removed if nonsignificant. β estimates, 95% confidence intervals (95% CI), and figures of model predicted means were reported for each final and best-fitting model.

Two additional analyses were conducted using each best-fitting model (models 1–3) to determine if either breastfeeding or childhood caloric intake explained part of the association. Best-fitting models were adjusted for either exclusive breastfeeding (yes or no) at the time of the infancy visit or average childhood caloric intake per day. Hypothesis tests were then reassessed to determine if model inference remained the same. A directed acyclic graph is shown in Supplemental Fig 5.

Poisson regression was conducted to assess the relative risk of overweight or obesity among children who were exposed to rapid weight gain. The outcome was overweight or obesity (yes or no). The predictors were RWG, sex, and SGA. Relative risk and 95% CIs were reported.

A total of 414 subjects were included in this analysis. Of those, 77 (18.6%) experienced RWG (increase in weight-for-age z score of ≥0.67 from birth visit to infancy visit) and 337 did not (Table 1). Infants who experienced RWG were on average smaller at birth (2.8 kg vs 3.2 kg), born 1 week earlier (38.7 vs 39.7 weeks) and were more likely to be SGA (33.8% vs 13.1%). RWG and normal weight gain groups did not differ in any other characteristic including maternal ppBMI, excessive gestational weight gain, race, ethnicity, infant sex, delivery mode, or primary feeding mode. There was no association between ppBMI and RWG in this cohort (P = .79), similar to findings of others.^23–25 

Maternal ppBMI and RWG were both positively associated with offspring BMI z score in childhood (Table 2). There was a 3-way interaction between ppBMI, RWG, and offspring sex (P = .0025), so results were stratified by sex. Among females, RWG significantly modified the association between ppBMI and early childhood BMI z scores (Fig 1A). For example, females exposed to a maternal ppBMI of 30 followed by RWG had 1.25 (95% CI, 0.84 to 1.65) higher average BMI z scores in early childhood than those exposed to a ppBMI of 30 without RWG, and 1.50 (95% CI, 1.10 to 1.90) higher average BMI z scores compared with those exposed to maternal ppBMI of 23 and no RWG (Fig 1B). Among males, RWG did not significantly modify the association between ppBMI and early childhood BMI z scores (P = .73; Fig 1A). Additional adjustment for breastfeeding status or caloric intake in childhood did not substantially alter the findings (Supplemental Table 6). Children who experienced RWG were 1.84 (95% CI, 1.04 to 3.26) times more likely to develop overweight or obesity, when compared with children who did not experience RWG (P = .037).

We next tested the effect of maternal ppBMI and RWG on childhood %FM as measured by air displacement plethysmography. Maternal ppBMI and RWG in infancy were both positively associated with childhood %FM (Table 3). RWG appeared to modify the association between ppBMI and childhood %FM, although the interaction P value was just above the cut point for statistical significance (P = .056). For example, children exposed to a maternal ppBMI of 30 followed by RWG had 4.01% (95% CI, 2.00% to 6.02%) higher average %FM in early childhood than those exposed to a ppBMI of 30 without RWG, and 4.92% (95% CI, 2.93% to 6.91%) higher average %FM than children whose mothers had a ppBMI of 23 and normal weight gain in infancy (Fig 2). There was no interaction by child sex. Additional adjustment for breastfeeding status in infancy or caloric intake in childhood did not substantially alter the findings (Supplemental Table 7).

As expected based on prior reports,¹⁸  BMI z score and total fat-free mass were positively correlated in this data set (both sexes: ρ = 0.47; females: ρ = 0.50; males: ρ = 0.43; all P < .0001) such that children with higher BMI z scores had more fat-free mass. Adjusting for height through FFMI (defined as fat-free mass [kg] divided by height [m²]) can make this association less pronounced.¹⁷  Although maternal ppBMI was positively associated with offspring FFMI, RWG was not (Table 4). There was a 3-way interaction between ppBMI, RWG, and offspring sex (P = .0073), so results were stratified by sex. RWG significantly modified the association between ppBMI and FFMI among females but not males (Fig 3A). For example, females exposed to a maternal ppBMI of 30 followed by RWG had a 0.93 kg/m² (95% CI, 0.42 to 1.43) higher average FFMI in early childhood than females exposed to a ppBMI of 30 without RWG, and 1.15 kg/m² (95% CI, 0.65 to 1.65) higher average FFMI compared with those exposed to maternal ppBMI of 23 and no RWG (Fig 3B). Additional adjustment for breastfeeding status in infancy or caloric intake in childhood did not substantially alter the findings (Supplemental Table 8).

In this study, we show that maternal ppBMI and RWG in infants interact to produce higher early childhood BMI z scores in a sexually dimorphic manner. Females whose mothers had a ppBMI of 23 and normal infancy weight gain had an average BMI at the 51st percentile, whereas females whose mothers had a ppBMI of 30 followed by RWG in infancy had an average BMI at the 94th percentile in early childhood, nearly at the cutoff for classification of obesity. Our results also demonstrate independent significant effects of ppBMI and of RWG on childhood %FM, as well as potentially an interactive effect, highlighting the importance of body composition end points to more fully characterize long-term effects of early life exposures. Development of childhood obesity is multifactorial and involves both biological and socioecological inputs, including contributions from pre- and postnatal environments. Although many studies have investigated the individual effects of pre- and postnatal factors, few have examined their interactive effects.

Several other groups have shown separate correlations of both maternal obesity¹¹  and RWG^26,27  with later risk of increased adiposity. Our results show that each of these exposures is associated with %FM in childhood and suggest possible combined effects of RWG and maternal ppBMI on offspring %FM. A similar interaction between maternal obesity during pregnancy and RWG was previously described.²³  RWG was defined from birth to the second birthday; triceps skin fold measurements were used to estimate fat mass annually from age 2 to 6 years. The authors concluded that maternal obesity influenced that rate of change in %FM from 2 to 6 years of age among children with RWG but did not specifically assess differences in adiposity at the study’s conclusion. In both their study and ours, children were studied during the physiologic BMI nadir and adiposity rebound,²⁸  which may make detection of %FM differences more challenging.

Our findings suggest that maternal ppBMI and RWG are each risk factors for childhood adiposity, regardless of sex, and may have interactive effects. The age range of our early childhood body composition measurements at 4 to 6 years corresponds to the physiologic nadir of adiposity and before the BMI and adiposity rebounds.²⁹  Studies at older ages may clarify whether there is truly an interactive effect on %FM. For FFMI, we found a significant synergistic effect of maternal ppBMI and RWG in females only. Because fat-free (lean) mass is a major component of and correlates with BMI in childhood,¹⁸  the combined effects of maternal ppBMI and RWG on FFMI may reflect their joint effects on childhood BMI.

Although most studies have likewise found that early RWG promotes a higher BMI later in life,^4,26,30  the effect of RWG may depend on additional exposures or characteristics. In our study, female children were particularly affected by the combined effects of ppBMI and RWG, possibly a reflection of earlier adiposity rebound in females²⁸  or hormonal differences during the minipuberty of infancy.³¹  Others have looked for sex differences in the response to early RWG, although results varied from ours. One study noted a stronger effect in males³²  although many, including a meta-analysis,⁶  found no evidence of sex dimorphism.²⁶  The scarcity of previous publications reporting on the interaction of maternal ppBMI and RWG makes us unable to make definitive conclusions about sex differences. We recommend that pediatric providers monitor infants of both sexes for RWG. Additional studies adequately powered to detect sex differences and examination of time points into adolescence and adulthood may provide a more complete picture of sexual dimorphism related to RWG.

Previous reports have also demonstrated anthropometric outcome differences related to RWG based on race and/or ethnicity,³³  whether the child was primarily breast or formula fed,^34,35  and cigarette smoke exposure.^12,23  The influence of socioeconomic status may differ between study populations.^1,33  Our results appear to be independent of primary infant feeding mode, but examinations of race and ethnicity or socioeconomic status may be incorporated in future analyses. More work is needed in larger cohorts to clarify the contributions of each factor.

Our study focused on the combined effects of elevated maternal ppBMI and RWG, regardless of the underlying cause of RWG. Others have investigated various contributors to RWG, including duration of breastfeeding,³⁶  formula feeding,^35,37  timing of introduction of complementary foods,^38,39  infant food responsiveness and eating speed,⁴⁰  and feeding on a schedule.³⁷  In this cohort, breastfeeding rates and timing of complementary introduction did not differ between infants with and without RWG, and adjustment for breastfeeding status at 6 months of age did not significantly change the findings. Similarly, adjustment for caloric intake in childhood did not significantly alter the magnitude of associations between ppBMI, RWG, and offspring outcomes. In the current study, we did not examine specific aspects of childhood dietary intake such as consumption of added sugar⁴¹  or physical activity levels.⁴²  We did not consider possible interventions for prevention of RWG. Recent studies showed lower risk of RWG with promotion of responsive feeding practices.^43–45  Ongoing studies are testing whether increased paternal engagement prevents RWG⁴⁶  and the impact of sleep and the intestinal microbiome.⁴⁷ 

Although the present report has many strengths, there are some limitations. Our relatively small analytic sample size benefited from location in a high resource area with a high national prevalence of overweight and obesity.⁴⁸  Our results may not be generalizable to a more global setting. Others have remarked that prevention of RWG may only be appropriate in communities with a high percentage of childhood overweight/obesity and a low percentage of underweight children.⁶  It is thus possible that RWG may benefit children in low-resource settings, particularly those born small, with previous evidence that RWG improves human capital.⁴⁹  Additional populations that may benefit from accelerated postnatal growth are infants born prematurely,⁵⁰  who were excluded from this analysis, and infants born SGA²⁷  because modest catch-up growth in early infancy may be a normal physiologic process for SGA infants.^51,52  Specific studies focused on these unique populations are warranted to define ideal postnatal growth patterns.

Our inclusion of body composition by air displacement plethysmography more clearly connects RWG to increased adiposity than use of BMI alone, although 2-compartment body composition models do not allow for characterization of fat mass distribution. We and others have shown that RWG may increase visceral⁵³  and hepatic⁷  fat accumulation in childhood and visceral fat in mid-adulthood.⁵⁴  Higher visceral fat has been previously linked to higher adult mortality⁵⁵  and may be an important outcome for future work. Incorporation of body composition in the newborn period may more precisely identify infant risk factors.^13,56  However, because body composition is not routinely available in the clinical setting, our use of weight-for-age z scores offers better practical guidance for pediatric providers. Finally, pursuing follow-up into adulthood will shed additional clarity on the long-term ramifications of RWG during infancy.

There is synergistic interaction between RWG in infancy and maternal prepregnancy weight status in relation to childhood BMI and body composition. Identifying modifiable factors contributing to RWG will be crucial to develop effective interventions. Pediatric providers should monitor infants for RWG during early life, especially in the context of maternal obesity, to mitigate future risk of overweight and obesity.

%FM

percent fat mass

CI

confidence interval

FFMI

fat-free mass index

ppBMI

prepregnancy BMI

RWG

rapid infant weight gain

SGA

small for gestational age