Using a Microsimulation of Energy Balance to Explore the Influence of Prenatal Sugar-sweetened Beverage Intake on Child Body Mass Index

Abstract

Objective:

Experiments with animals suggest that high sugar consumption during pregnancy may predispose offspring to obesity, but few human studies have examined this relationship. We explore the association between consumption of sugar-sweetened beverages (SSBs) during pregnancy and caloric intake through childhood.

Methods:

Using cohort data on child weight, height, and physical activity levels, we employed a lab-validated microsimulation model of energy balance to infer caloric intake of children through age 11. We then employ random effects models to explore the relationships between prenatal maternal consumption and inferred caloric intake during childhood.

Results:

An additional daily serving of SSBs during the second trimester of pregnancy was associated with an increase in child consumption of 13 kcal/day (95% CI: 1.2, 26.8). Age-stratified models adjusting for maternal and child covariates suggested that this association was strongest for children aged 2.5–5.5 years. We did not find that consumption of SSB during the first trimester had a consistently positive relationship to caloric intake.

Conclusions:

Our findings suggest SSB consumption during the second trimester of pregnancy is associated with child energy intake and may influence anthropometry in early childhood, which is consistent with and suggestive of the presence of biological causal pathways alongside likely simultaneous contributions of social and environmental influences.

INTRODUCTION:

The obesity epidemic continues to be a serious concern for the health and welfare of Americans. Childhood obesity is of particular concern because it not only is associated with an array of health challenges but also increases the likelihood of obesity in adulthood. According to the Centers for Disease Control and Prevention, 13.7 million children and adolescents have obesity, including 18.4 percent of children 6–11 years old and 13.9 percent of children 2–5 years old.(1) Childhood obesity is potentially influenced by a large number of factors, some of which are understudied; among these is the prenatal environment. A large body of literature finds that the intrauterine environment is not only directly influenced by maternal behaviors and physiology—including pregnancy weight gain, diabetes, over- and under-nutrition, and smoking status—but is also tied to obesity risk in children.(2–4) While evidence of the link between sugar-sweetened beverage (SSB) consumption in childhood and obesity is compelling and public health interventions to limit intake among children have been widespread (5–9), the implications of SSB consumption during the prenatal period for child obesity are not well known.

Several recent studies have examined the relationship between SSB consumption during pregnancy and child body weight. Phelan et al. (2011) found that among normal-weight mothers, SSB intake during pregnancy was the single strongest predictor of high child birth weight but not a predictor of weight at 6 months.(10) Jen et al. (2017) determined that an additional daily serving of sugar-containing beverages during pregnancy was associated with a 0.04 increase in BMI z-score at age 6.(11) Gillman et al. (2017) observed that an additional daily serving of SSB during the second trimester of pregnancy was associated with a 0.07 increase in BMI z-score at mid-childhood (median age 7.7 years).(12)

These findings suggest a potential causal impact of maternal consumption during pregnancy on child body weight that emerges only years after birth. However, the mechanisms that may drive this relationship are understudied.(3) One theory is that children whose mothers consume SSBs during pregnancy are more likely to be raised in obesogenic environments. Parents provide their children with food, serve as models of behavior, and shape their food choices, all of which influence BMI.(13–16) While many analyses attempt to account for environmental influence, accurately capturing child dietary behavior is challenging. Another set of theories concerns how biological pathways shape food preferences in children. Sensitivity to and preference for sweet taste have salient genetic components,(17, 18) and certain mothers may be predisposed to obesogenic diets and high SSB consumption. In turn, they may pass this disposition along to their children. However, the evidence on whether genetically-inherited sensitivity to and preference for sweetness is linked to obesity is equivocal.(19, 20) Flavor preferences in children are also shaped by maternal diet by way of amniotic fluid and breast milk(21)..

A separate biological theory contends that SSB consumption during pregnancy affects the metabolic programming of children and their body weight during childhood. Maternal overnutrition may lead to elevated levels of insulin production in the fetus and subsequent adverse metabolic changes in the offspring, including insulin resistance and overexpression of proteins related to lipid metabolism.(22–24) Much of the support for this theory comes from animal studies due to the infeasibility of such experiments with humans. In rats, obesogenic and high-fructose diets during gestation, for example, can induce maternal hyperglycemia and insulin resistance, and lead to high plasma glucose and leptin, hypertrophic adipocytes, and increased body weight in offspring in the perinatal period and at maturity.(25–28) While suggestive, the extremity of the diets induced in many of these studies and biological differences across species limit our ability to draw firm conclusions about the human context.

A better understanding of the relative importance of biological and environmental pathways is important for policymakers, intervention leaders, and public health professionals. If SSB consumption during pregnancy affects mid-childhood BMI via biological programming mechanisms—a theory suggested by the limited human evidence—then obesity prevention efforts should not only target the myriad of social, environmental, and cultural influences on children, but maternal diet and behavior during pregnancy as well.

In the current study, we extend the work of Gillman et al. (2017), using data from the same cohort study to gain intuition about the causal linkages between maternal SSB intake and child anthropometric outcomes throughout early childhood. We used a lab-validated microsimulation model of energy balance to infer caloric intake from longitudinal, individual-level anthropometry and time use data of children in the Project Viva study, a pre-birth cohort of mother-child pairs. The resulting inferred caloric intake estimates thus represent residual energy net of growth and physical activity in familiar and easily interpretable units. We then explored differences in these inferred calorie intake trajectories by pre-natal maternal SSB exposure levels to provide additional insight into the potential causal pathways discussed above beyond what can be done with weight or BMI measures alone.

METHODS:

Input Data:

We used a microsimulation model of biological energy balance paired with time-specific, individual-level BMI and physical activity data from a cohort of children to infer unobserved caloric intake values throughout childhood. BMI and physical activity data came from Project Viva, a longitudinal study that tracked diet, physiology, and health-related behaviors among women and their children, beginning during pregnancy and extending through offspring childhood.¹ Between 1999 and 2002, Project Viva recruited pregnant women from obstetric offices of Atrius Harvard Vanguard Medical Associates in Eastern Massachusetts. These expectant mothers (n = 2,100) attended in-person study visits during the first and second trimesters of pregnancy as well as shortly after delivery, during their child’s infancy (median age = 6.6 months), early childhood (median age 3.4 years), and mid-childhood (median age 8.0 years).² They also completed mailed questionnaires on the diet and health behaviors of their children on an annual basis. Maternal data included daily SSB (including sugary soda and fruit drinks with added sugar), fruit juice, and diet soda consumption during pregnancy, as well as pre-pregnancy BMI, pregnancy weight change, education, smoking status, income, and race/ethnicity. Child data included daily physical activity, height, weight, and SSB consumption. Full details on cohort recruitment and participation are detailed elsewhere.(12, 29)

We transformed data on specific physical activities among children in the cohort into physical activity levels (PAL) for use in the microsimulation model. PAL is a measure of mean daily energy expenditure, expressed as a multiple of basal metabolic rate. To calculate PAL for each child at every observation, we multiplied physical activity ratios of each activity by the reported mean daily hours spent engaging in these activities at each age. The age-specific physical activity ratios of time use variables—which included sleep, TV watching, active play, walking, light/moderate exercise, and vigorous exercise—were derived from the literature.(30–32) Residual hours unaccounted for by cohort data (i.e., 24 minus the sum of all hours spent engaging in specific activities) were assigned a physical activity ratio value based on an age-dependent function.(33) Additional details on the calculation of PAL are available in the supplementary appendix. We used multiple imputation to obtain values when activity variables used to calculate PAL were missing. Specifically, we used SAS (Proc MI) to generate 50 imputation datasets, ran the microsimulation with each dataset independently, and then pooled regression coefficients during analysis in Stata.

Microsimulation model mechanics:

The principal component of the microsimulation is an existing physiology model.(34) A series of differential equations characterizes changes in fat and fat-free mass as a function of energy intake and energy expenditure while accounting for metabolic adaptations to changes in body weight. The original model was calibrated to adult data, but we relied on a modified version that extends to children and considers age-specific growth trajectories and metabolism, distinguishing, for example, healthy growth from excessive weight gain.(35)

As initially designed, the physiology model takes in a child’s height, weight, age, PAL, and caloric intake at each time point, and outputs a new weight value representing the result of that caloric intake consumed daily over a given time interval. We used an algorithm that adapts the physiology model to instead estimate unobserved values of daily caloric intake (i.e. inferred residual energy imbalance in caloric units for ease of interpretation) given a pair of sequential of weight observations. The algorithm iterates over the empirical observations stored for each participant in temporal order. At each observation, it initializes the physiology model with the participant’s height, weight, age, physical activity level and sex, and then calculates daily caloric intake required to reach the weight value at the subsequent observation. Additional details about the mechanics of the physiology model are available in the supplementary appendix.

We inferred caloric intake for children from birth to a maximum age of 11. We implemented an ex-post interval threshold, dropping observations for which there were no subsequent observations within a three-year period. The purpose of this was to avoid including calorie estimates that relied on weight change based on temporally distal observations, which could produce inaccurate estimates.³

Statistical analysis:

We partially validated the daily calorie intake trajectories from the microsimulation model by comparing them to estimates from the U.S. Department of Agriculture’s (USDA) Dietary Guidelines, which provide age- and sex-specific energy requirements for children age 2 and older.(36) Some previous work has found child BMI to be predicted by maternal prenatal consumption of artificially sweetened beverages(37). We thus estimated associations between maternal beverage consumption, separately including measures of SSBs, fruit juice (discussed further below), and diet soda and inferred child caloric intake using a random effects model of the following form:

$$Y_{i,t}=B_0+ {\gamma }_t+\beta X_i+\delta C_{i,t}+{\epsilon }_{i,t}$$

Where Y_i,t is a measure of inferred daily caloric intake of child i at age t, γ is a vector of age terms, X is a continuous measure of prenatal maternal beverage consumption, C is a vector of maternal- and child-specific covariates, and ε is an error term. Our main exposure variables were daily maternal SSB consumption (servings/day) during the first and second trimester of pregnancy. SSB intake variables were continuous in our random effects models but are depicted as categorical in Figure 1 and Table 1. The other prenatal exposure variables examined were daily first and second trimester consumption of diet soda and fruit juice. We estimated separate models for the two trimesters to avoid autocorrelation. In each, we incrementally included covariates related to maternal and child demographics, behavior, and physiology. Our primary model included maternal covariates: pre-pregnancy BMI, pregnancy smoking status (binary, smoked at all while pregnant), education (binary, attained a college degree), socioeconomic status (binary, household income greater than $70,000/year), race/ethnicity, and child sex. In one alternative specification, we additionally included child physical activity level z-score and height z-score as covariates. In a final model, we also controlled for daily child SSB intake in the previous year, t-1. Physical activity level z-score was calculated based on the age-specific distributions from the cohort data and height z-score was calculated using age and sex-specific CDC child growth charts.(38) Maternal trimester 1 and 2 Alternate Healthy Eating Index score, a measure of dietary health, was excluded as a covariate because it did not appreciably alter results.

Figure 1: Simulated inferred caloric intake trajectories by age, stratified by second trimester maternal SSB intake group, and USDA estimated caloric needs per day.

Microsimulation model of caloric intake trajectories for children of mothers who consumed > 2.5 serv/day of SSB during the second trimester (red line) and for children of mothers who consumed 0 serv/day (blue line). Bars show 95 percent confidence intervals. The grey line represents USDA calorie guidelines (boy-girl mean) at each age for children with moderate physical activity levels. The upper and lower bounds of the grey polygon indicate age-specific USDA calorie guidelines for children with high and low physical activity levels, respectively.

Table 1:

Summary of Maternal Characteristics, Stratified by Second Trimester SSB Intake Level

	Maternal second trimester SSB consumption frequency
	0 serv/day	>0 to <=1 serv/day	>1 to <=2.5 serv/day	>2.5 serv/day	Total
	n = 198 (15.8%)	n = 845 (67.2%)	n = 131 (10.4%)	n = 83 (6.6%)	n = 1257
First trimester
SSB consumption, serv/day	0.19 (0.55)	0.50 (0.66)	1.16 (1.00)	1.60 (1.20)	0.59 (0.82)
Fruit juice consumption, serv/day	0.95 (0.85)	1.21 (0.90)	1.36 (1.05)	1.60 (1.28)	1.21 (0.95)
Diet soda consumption, serv/day	0.31 (0.57)	0.23 (0.50)	0.22 (0.37)	0.29 (0.67)	0.24 (0.51)
Second trimester
Fruit juice consumption, serv/day	0.93 (0.99)	1.24 (0.99)	1.46 (1.02)	1.94 (1.35)	1.26 (1.04)
Diet soda consumption, serv/day	0.23 (0.46)	0.13 (0.39)	0.11 (0.33)	0.16 (0.53)	0.14 (0.41)
Race / Ethnicity
White (%)	78.3	76.5	74.8	67.5	76.0
Black (%)	8.1	9.2	11.5	19.3	9.9
Hispanic (%)	3.0	5.8	8.4	2.4	5.4
Asian (%)	7.6	5.2	3.1	3.6	5.3
Other (%)	3.0	3.3	2.3	7.2	3.4
College education or more (%)	78.8	75.4	67.9	63.9	74.4
Family income < $70,000 (%)	29.3	33.6	40.5	44.6	34.4
Smoked during pregnancy (%)	6.6	9.9	13.7	16.9	10.3
Pre-pregnancy BMI, kg/m2	23.4 (4.28)	24.56 (5.13)	25.50 (5.57)	25.32 (5.97)	24.52 (5.14)

Note: means and standard deviations in parentheses displayed for continuous variables. Summary includes only participants with values for each of the variables for whom we could also infer caloric intake.

To examine whether the association between SSB and other beverage consumption during pregnancy and inferred child caloric intake was dependent on age, we stratified the primary model a priori into age bins as follows: 0 to <2.5 years, 2.5 to <5.5 years, 5.5 to <8.5 years, and 8.5 to 11 years. To determine if associations were observed for all beverages with sugar, we conducted supplementary analyses with sugar-containing beverage consumption (the sum of SSB and fruit juice) as our primary exposure.

RESULTS:

Maternal and child summary statistics are shown in Table 1 (n = 1257). 83 percent consumed one serving or less per day during this period (with 16 percent consuming no SSBs). 7 percent of mothers consumed more than 2.5 servings of SSB per day during the second trimester. In addition, 76 percent of mothers were white, 34 percent had a household income up to $70,000, and 10 percent smoked during pregnancy. Maternal smoking, pre-pregnancy BMI, lower household income status, and black racial/ethnic classification were positively correlated with second trimester daily SSB intake, while college education was negatively correlated with intake. Daily consumption of SSB, diet soda, and fruit juice in trimester 1 were moderately correlated with their respective consumption levels in trimester 2 (r = 0.50, r = 0.61, r = 0.54, respectively) (Table 2). Among children, physical activity level and energy intake were positively correlated with age. Children ages 0 to 2.5 years had a mean (SD) physical activity level of 1.49 (0.73) and caloric intake of 693.19 (210.37) kcal/day; those ages 8.5 to 11 years had a mean (SD) level of 1.66 (0.07) and intake of 2010.62 (309.69) kcal/day (Table 3).

Table 2:

Pearson Correlation Matrix for Trimester 1 and 2 Maternal Beverage Intake Variables (n = 1257)

	SSB Tri 2	Diet Soda Tri 2	Fruit Juice Tri 2	SSB Tri 1	Diet Soda Tri 1	Fruit Juice Tri 1
SSB Tri 2	—
Diet Soda Tri 2	−0.01
Fruit Juice Tri 2	0.23	−0.08
SSB Tri 1	0.50	0.00	0.17
Diet Soda Tri 1	0.02	0.61	−0.09	0.00
Fruit Juice Tri 1	0.17	−0.06	0.54	0.17	−0.07	—

includes only participants represented in Table 1

Table 3:

Summary of Child Characteristics, Stratified by Age Group

	Child age
	0 to <2.5 years	2.5 to <5.5 years	5.5 to <8.5 years	8.5 – 11 years
	n = 1236	n = 781	n = 678	n = 544
Characteristic	Value
Age	0.79 (0.33)	3.82 (0.48)	7.03 (0.40)	9.58 (0.35)
Female (%)	51.38	53.52	52.94	52.38
Caloric intake (kcal/day)	693 (210)	1210 (198)	1632 (299)	2011 (310)
Physical activity level (multiple of basal metabolic rate)	1.49 (0.73)	1.57 (0.05)	1.62 (0.05)	1.66 (0.07)
Height (cm)	68.71 (6.74)	101.59 (5.12)	123.04 (5.71)	138.29 (6.72)
SSB consumption (serv/day)	0.18 (0.49)	0.26 (0.47)	0.37 (0.65)	0.37 (0.54)
Missing SSB data, (%)	35.11	5.10	10.32	19.30

Note: means and standard deviations in parentheses displayed for continuous variables. Summary includes only participants with values for each of the variables besides child SSB consumption for whom we could also infer caloric intake.

Figure 1 shows means and standard deviations of caloric intake trajectories from the microsimulation model, clustered by age, for children of mothers who consumed ≥2.5 servings/day of SSB during the second trimester (red line) and for children of mothers who consumed 0 servings/day (blue line). The upper and lower bounds of the grey band represent daily calorie estimates from the USDA for physically active and sedentary children, respectively, while the middle dark grey line represents estimates for children with moderate physical activity. Because USDA estimates are distinct for each sex, we used the male-female mean, weighted according to the male-female ratio in the cohort data. Model caloric intake trajectories fall reasonably within the bounds of the USDA estimates for most of the sample, particularly for younger children. Modelled mean caloric intake among children of mothers who consumed ≥2.5 servings/day is modestly higher than intake among children of mothers who consumed 0 servings/day at all ages, particularly after age 2. Caloric intake trajectories of children of mothers with SSB consumption rates between 0 and 2.5 servings/day (not shown) fall within the USDA estimate range as well.

An additional maternal daily serving of SSB was associated with an increased energy intake among offspring of 7.84 kcal/day (95% CI: −4.42, 20.32) in the unadjusted trimester 1 model, and 13.82 kcal/day (95% CI: 1.2, 26.8) in the unadjusted trimester 2 model (Tables 4 and 5). The SSB coefficients were similar when daily intake of fruit juice and diet soda were introduced into the unadjusted models. In models including maternal and child covariates, the SSB coefficients remain positive for trimester 2 and are negative for trimester 1, although none of these were statistically significant.

Table 4:

Random Effects Model Estimates for Maternal Variables Predicting Child Caloric Intake, Trimester 1 Exposures

	SSB only, unadjusted		All beverage exposures, unadjusted		All beverage exposures, adjusted for maternal covariates		All beverage exposures, adjusted for maternal and child covariates		All beverage exposures, adjusted for maternal and child covariates, including child SSB
Variable	B	95 % CI	B	95 % CI	B	95 % CI	B	95 % CI	B	95 % CI
SSB	7.84	(−4.42, 20.32)	7.89	(−4.59, 20.36)	−2.54	(−10.18, 5.1)	−1.48	(−7.73, 4.76)	−5.15	(−11.69, 1.38)
Juice			−0.49	(−11.3, 10.34)	2.43	(−3.93, 8.8)	−0.15	(−5.34, 5.05)	−3.78	(−9.25, 1.69)
Diet Soda			15.49	(−4.8, 35.78)	8.46	(−3.41, 20.32)	4.79	(−4.93, 14.52)	6.46	(−3.62, 16.54)
Individuals	1,493		1,492		1,396		1,396		1,038
Observations	9.046		9,035		8,470		8,470		4,736
R2	.71		.71		.89		.92		.92

Note: Beverage exposures are measured in servings/day. Child caloric intake is measured in kcal/day. All models contain age terms. Maternal covariates include pregnancy BMI, smoking status, household income level, education, and race/ethnicity. Child covariates include sex, physical activity level, and height.

Table 5:

Random Effects Model Estimates for Maternal Variables Predicting Child Caloric Intake, Trimester 2 Exposures

	SSB only, unadjusted		All beverage exposures, unadjusted		All beverage exposures, adjusted for maternal covariates		All beverage exposures, adjusted for maternal and child covariates		All beverage exposures, adjusted for maternal and child covariates, including child SSB
Variable	B	95 % CI	B	95 % CI	B	95 % CI	B	95 % CI	B	95 % CI
SSB	13.82	(1.2, 26.8)	14.5	(1.52, 27.48)	6.27	(−1.53, 14.07)	5.7	(−0.72, 12.12)	1.66	(−5.2, 8.5)
Juice			−2.51	(−12.52, 7.49)	0.17	(−5.77, 6.11)	−0.26	(−5.18, 4.67)	−0.39	(−4.97, 5.74)
Diet Soda			14.28	(−11.93, 40.5)	7.75	(−7.43, 22.93)	5.90	(−6.6, 18.38)	0.54	(−12.27, 13.3)
Individuals	1,425		1,424		1,332		1,332		992
Observations	8,627		8,623		8,087		8,087		4,553
R2	.72		.72		.9		.92		.92

Note: Beverage exposures are measured in servings/day. Child caloric intake is measured in kcal/day. All models contain age terms. Maternal covariates include pregnancy BMI, smoking status, household income level, education, and race/ethnicity. Child covariates include sex, physical activity level, and height.

We also performed stratified analysis of the adjusted model (without child SSB as a covariate) to examine whether the association between prenatal beverage consumption and child intake trajectories was specific to certain age strata (Tables 6 and 7). Results suggest this association was restricted to children age 2.5 to 5.5 years old. Adjusted for covariates, an additional maternal daily serving of SSB during the second trimester was associated with an additional 10.45 kcal/day child intake (95% CI: 3.688, 17.22).

Table 6:

Random Effects Model Estimates of Maternal Variables Predicting Child Caloric Intake, Stratified by Age Group, Trimester 1 Exposures

	Child age
	0 – 2.5 years		2.5 – 5.5 years		5.5 – 8.5 years		8.5 – 11 years
Variable	B	95 % CI	B	95 % CI	B	95 % CI	B	95 % CI
SSB	−1.85	(−7.73, 4.04)	−0.79	(−7.278, 5.70)	−12.77	(−30.63, 5.10)	−7.70	(−26.64, 11.23)
Juice	2.29	(−2.60, 7.18)	0.68	(−4.77, 6.14)	−9.27	(−24.46, 5.92)	−7.85	(−23.29, 7.59)
Diet Soda	4.21	(−4.58, 13.01)	5.61	(−4.58, 15.80)	1.34	(−29.38, 32.06)	30.12	(−0.78, 61.01)
Individuals	1,372		868		742		591
Observations	3,508		2,053		1,822		1,087
R2	.81		.85		.56		.70

Note: All models contain age terms and covariates for maternal pregnancy BMI and smoking status, socioeconomic status, education and race, as well as child sex.

Table 7:

Random Effects Model Estimates of Maternal Variables Predicting Child Caloric Intake, Stratified by Age Group, Trimester 2 Exposures

	Child age
	0 – 2.5 years		2.5 – 5.5 years		5.5 – 8.5 years		8.5 – 11 years
	B	95 % CI	B	95 % CI	B	95 % CI	B	95 % CI
SSB	3.71	(−2.30, 9.72)	10.45	(3.69, 17.22)	2.00	(−16.61, 20.62)	1.02	(−17.50, 19.53)
Juice	−1.63	(−6.18, 2.92)	3.78	(−1.50, 9.07)	−1.28	(−15.21, 12.66)	2.74	(−12.41, 17.89)
Diet Soda	11.10	(−0.54, 22.74)	7.77	(−4.90, 20.44)	−16.33	(−56.63, 23.98)	5.18	(−31.40, 41.75)
Individuals	1310		819		712		566
Observations	3,341		1,950		1,746		1,050
R2	.81		.85		.54		.69

Note: All models contain age terms and covariates for maternal pregnancy BMI and smoking status, socioeconomic status, education and race, as well as child sex.

Additional analyses:

We conducted sensitivity analyses around our characterization of the main exposure variables. Specifically, we repeated the analyses described above using maternal sugar-containing beverage consumption during trimesters 1 and 2 as the exposure variables (i.e. both consumption of sugar-sweetened and naturally sweetened fruit juices). Results from these supplementary models are presented in a supplemental appendix. They are similar to our main findings, suggesting that consumption of beverages specifically with added sugar was a stronger determining factor. In addition, we explored variations of our age-stratified models that do include child SSB consumption. Results are presented in our supplemental appendix and are also generally similar to those shown here.

DISCUSSION:

In the current study, we combined data from a longitudinal, pre-birth cohort of mother-child pairs and a microsimulation model of child energy dynamics to generate caloric intake trajectories of children from birth to adolescence. We subsequently examined the relationship between maternal prenatal beverage and these trajectories and found that consumption of an additional serving of SSB during the second—but not first—trimester of pregnancy was modestly associated with additional caloric intake among children. Models stratified by child age indicated that the association was strongest between ages 2.5 and 5.5 years.

Though the influence of child diet and SSB consumption on obesity onset has been well documented (7, 9), the current study provides suggestive evidence to support research that finds maternal prenatal consumption of sugar-sweetened beverages during the second trimester of pregnancy may be an important predictor of offspring weight during childhood as well. Child obesity intervention and policy efforts may therefore be well advised to consider targeting not just child consumption of added sugars, but maternal consumption during the pregnancy.

The cohort data we relied on has been previously examined in a similar context (12, 45), but we made several important extensions. Specifically, we incorporated information from the literature on age- and activity-specific physical activity ratio values to capture heterogeneity in energy expenditure of children and used a microsimulation model to evaluate child trajectories in terms of energy intake rather than BMI. Finally, we analyzed outcomes longitudinally between age 0 to 11 rather than at a single time point. Like the earlier work that utilized this cohort dataset, we found that maternal SSBs were most strongly tied to child energy balance, and that the second but not first trimester of pregnancy was critical in explaining this effect.

A key question is to what extent maternal consumption of SSBs during pregnancy itself drives child obesity and high BMI. On the one hand, SSB consumption during pregnancy may simply be a proxy for the food environment and consumption habits of children that drive BMI change. Alternatively, biological pathways, such as a nutritional programming mechanism, could account for some of the observed relationship. During the second trimester, the fetus begins to ingest amniotic fluid in addition to receiving glucose through direct blood transfer, and chronically high glucose levels driven by maternal consumption may be matched by chronic high levels of fetal insulin production.(23, 39) Hyperinsulinemea can trigger increased lipogenesis as well as altered expression of genes and proteins related to metabolic functioning in the offspring.(22) In particular, the high fructose content of many SSBs may be a major driver of these metabolic perturbations.(27, 40, 41) However, without extensive data on child consumption, dietary habits, or longitudinal hormone assays of mother and child, it is difficult to draw definitive conclusions about the plausibility of a biological pathway. One study found the association between maternal intake of SSB and child adiposity at age 6 was independent of child insulin concentration at this age, although insulin at earlier years was not examined.(11)

Our estimates of the additional daily child caloric intake associated with second trimester maternal SSB consumption are modest as a proportion of total calories consumed. Estimates are also small when compared to estimates of youth energy intake from SSBs alone, which, while declining recent years (42), are around 150 kcal/day. (43) Nonetheless, when considered over an extended timespan, relatively small daily discrepancies in energy intake may produce significant differences in body weight and rates of obesity.

A major strength of this study is our use of longitudinal cohort data that begin prior to birth and extend more than a decade, with detailed information on prenatal health, diet, environmental characteristics, and offspring anthropometry throughout childhood. In addition, we quantify differences in outcomes in terms of calories rather than BMI, which may be useful when designing interventions that directly target consumption.

One limitation of our study is that we relied heavily on parental self-report data of child diet, which may be subject to measurement error. Because child intake data was based exclusively on questionnaires completed by mothers, it is possible that actual intake is underreported, particularly among older children, who may have SSB consumption habits that are not well observed by their parents.(44) Another limitation is that even though the microsimulation was calibrated to children aged 5 years and older, we used it from birth onwards. This may bias calorie trajectories for children under 5 in our sample due to age-related differences in metabolism (35), although model calorie trajectories correspond closely with USDA estimates for children in this age range. Subsequent efforts that extend lab-validated metabolic dynamics to birth might help to better capture exposure effects in toddlers and infants.

Future work should more explicitly characterize biological and environmental pathways that influence child consumption and obesity patterns to better differentiate them. Research may incorporate more extensive and high quality data on child consumption, social and familial environments in early childhood (e.g. parental diet and time use), and the metabolic markers—including insulin concentration and sensitivity—from the prenatal period through childhood.

Study importance questions:

What is already known about this subject?

• Experiments with animals suggest that high sugar consumption during pregnancy may predispose offspring to obesity.

• There is limited evidence from human studies indicating the presence of such a relationship.

What are the new findings in this manuscript?

• an additional daily serving of SSB during the second trimester of pregnancy was associated with an increase in child consumption of 13 kcal/day (95% CI: 1.2, 26.8).

• Age-stratified models adjusting for maternal and child covariates suggested that this association was strongest for children aged 2.5–5.5 years.

• We did not find that consumption of SSB during the first trimester had a consistently positive relationship to caloric intake.

How might results change the direction of research or the focus of clinical practice?

• Findings are suggestive of the presence of biological causal mechanisms linking maternal SSB consumption and later childhood physiological outcomes, but do not preclude the simultaneous contribution of social and environmental influences.

• This might motivate additional research to confirm the presence and illuminate the operation of such mechanisms.

• Findings are supportive of interventions to reduce maternal SSB consumptions.