Childhood obesity is a globally persistent disease with multifactorial causes and a wide range of complications that track into adulthood ¹. Increasingly, evidence suggests that exposures to environmental endocrine-disrupting compounds such as Bisphenol A (BPA) and phthalates may influence body weight and physiology, such as insulin resistance ^2-4. These multi-functional compounds are used in everyday products, leading to ubiquitous exposure ^5,6. BPA is used in the production of polycarbonate plastics and epoxy resins and is commonly found in food and beverage containers, medical equipment, toys, the linings of canned foods, thermal receipt paper, and dental sealants ⁷. Phthalates are used similarly as plasticizers in many consumer and industrialized products and as a stabilizing and solubilizing material ⁸.

Associations between exposures to these compounds and increased body mass index (BMI) and adiposity differ by sex, population, and age of exposure, but many studies suggest positive associations ^9-12. However, no longitudinal studies have investigated the role that in utero exposures to these compounds may have on growth trajectories. By assessing whether growth differs by levels of exposure, this could allow estimation of milestones, such as infancy peak or the adiposity rebound, which could be used to link childhood growth with adult diseases like obesity ¹³.

In this study, we add to our previously reported associations between in utero and childhood urinary concentrations of BPA and phthalates metabolites on childhood BMI and adiposity. We examine the BMI growth trajectories of children exposed to these compounds in utero to assess how different levels of exposures may influence growth. We applied an exploratory method of growth modeling to fit and assess the trajectories of BMI for males and females by tertiles of third trimester BPA and phthalates metabolites.

Among 249 children, 47% were male (Table 1). The average follow-up for children was 10.3 years. Among 2,162 total visits, 78% occurred at or before 5 years of age. Distributions of BPA and phthalate metabolites in maternal third trimester urine show detectable concentrations in most pregnancy samples, with no differences by sex (Supplementary Table 2).

We constructed BMI trajectories for models that showed a better fit with the inclusion of the BPA or phthalates metabolite (Supplementary Table 3). In boys, the likelihood ratio tests showed a better fit for models including the phthalates metabolites MiBP, MBzP, MEHP, and MEHHP. In girls, model fit was best with the inclusion of MECPP. We did not find a relationship between BPA exposure and BMI trajectories.

Trajectories for each tertile of metabolite exposure crossed at various times during the modeled ages, particularly in the first few years. Some BMI trajectories which appeared to be stable at the highest trajectory, such as the second tertile of exposure to MiBP in boys, did not remain so after the age of 5 years. Conversely, while exposure to the highest tertile of MEHP in infancy predicted the highest BMI trajectory in boys, after age 2 the highest trajectory was predicted by exposure to the second tertile of MEHP exposure. Confidence intervals widened as ages increased due to fewer observations. For modeled trajectories in boys, by age 14 the first and second tertiles of exposure had the highest predicted BMI trajectory (Figure 1). Exposure to the first tertile of MiBP and MBzP predicted the lowest BMI trajectory in infancy and early childhood, but crossed over to predict the highest BMI by age 14. With the phthalates metabolites MEHP and MEHHP, the second tertile of exposure was consistently predictive of the highest BMI trajectory from early childhood on. In girls, modeling showed that, by age 14, the third tertile of MECPP predicted the highest BMI trajectory for girls, though prior to age 9, the highest level of exposure predicted the lowest BMI trajectory (Figure 2).

In this prospective study modeling BMI trajectories, we observed that in utero exposure to phthalates metabolites differentially influenced the BMI trajectories of boys and girls. While tertile trajectories crossed frequently in early childhood, by age 14, we found that different tertiles of exposure predicted the highest attained BMI. This was observed with the highest tertile of exposure in girls, but the first or second tertile of exposure in boys. We did not detect any significant association with BPA.

Previous studies have found some positive associations between BPA and phthalates metabolites with child BMI; most longitudinal studies have outcomes in early childhood with few occurring in late childhood and early adolescence ^11,12,20-23. Results from these longitudinal studies have been mixed, with differences reported by metabolites, sex, and age. The mixed findings from prenatal exposures and child BMI outcomes may be due to the different ages that child outcomes are measured, which may be influenced by the timing of puberty or adiposity rebound. Therefore, we explored the use of growth trajectories as an alternate method of assessing the potential impact of early life exposures. The fluctuating trajectories indicated that the age outcomes are assessed at can influence the interpretation of the relationship between levels of exposures with child outcome. This is likely a contributing reason why some metabolites are positively associated with child BMI at one age but not another.

This is illustrated by contrasting the results we obtained here, with previous work and studies by others. Deierlein et al. (2016) observed higher BMI for girls aged 7-13 years when they were exposed at 6-8 to medium vs. low concentrations of MBzP while, in previous work, we found an negative relationship between prenatal MBzP exposure and child’s BMI at ages 8-14 years with no modification by sex ^10,12. However, we only observed a relationship between MBzP among boys in this study. Similarly, a pooled analysis of 3 prospective cohorts in the US showed sexually dimorphic associations with high molecular weight phthalates metabolites (MEHP, MEHHP, MEOHP, MECPP) negatively associated with BMI in girls, but no associations with boys ²⁴. In previous work, we found urinary concentrations of MEHP in boys were negatively associated with child adiposity but did not find a relationship among girls ¹². This was inconsistent with other studies which observed MEHP to be inversely associated with BMI in girls both cross-sectionally and longitudinally ^9,10. Using BMI trajectory modeling to re-assess our findings of MEHP exposure in utero, we found a relationship among boys, but not girls. These differences in results may potentially be due to the close trajectory paths and overlapping confidence intervals. However, with further refinement, growth trajectory modeling could be a useful tool to assess how exposures may alter trajectories of growth in children and may aid in interpreting how different levels of exposures may result in different associations.

The observed sex differences are thought to be due to the endocrine-disrupting mechanisms of phthalates metabolites. Phthalates are anti-androgenic and are able to influence energy homeostasis through increasing the number of adipocytes, as well as increasing lipid accumulation; their action is not through direct interaction with the androgen receptors, but through stimulation of the peroxisome proliferator-activated receptors family of nuclear receptors (PPARs) ^8,25. PPARα in particular has been found to have sex-differential effects, with exposure to MEHP significantly increasing body weight and fat pads of male, but not female mice ²⁶.

Limitations of this study include the relatively small sample size and fewer number of observations of BMI measures after the age of 5, as follow-up was not systematic after that point in time. Children had 1-3 measurements after that age and few (n=21) had only 1 measurement after the age of 5. This modeling is exploratory in examining whether growth trajectories could be constructed using available data. Restricting growth modeling to observations before the age of 5 would have altered reported results and further work will need to ensure greater data support for more precise modeling. Our interpretations should therefore be viewed with caution until further studies are conducted with more observations. We also chose to use fractional age polynomials rather than splines to model growth trajectories; spline models are more flexible but require pre-defined decisions on the number and placement of age knots as well as requiring continuous data support, whereas we wanted to explore how growth curves would form based on the data, and much of our data on BMI was collected before the age of 5.

Another limitation was the use of a single spot urine in pregnancy for analyses of metabolites. Concentrations of metabolites are known to have short metabolic half-lives, with <24 hours for phthalates metabolites and <5-6 hours for BPA, and exhibit temporal variability; a single spot measure may therefore not be capable of accurately assessing individuals ^27,28. However, individuals tend to exhibit consistent behaviors that would result in more consistent exposures, which has been shown by other studies ²⁹. Our geometric mean urinary concentrations of metabolites were comparable or lower compared with those reported by others in the pregnancy period. In a Californian population, Harley et al. (2013) reported a GM of 1.2 ng/mL for BPA averaged across 2 pregnancy time points, which was higher than our 0.7 ng/mL and 0.9 ng/mL for boys and girls, respectively. Similarly, higher values were reported by Vafeiadi et al. (2016) among a Greek population during the first trimester of pregnancy (1.2 ng/mL). For phthalates metabolites, higher GM values were reported by Buckley et al. (2016) among 3 pooled cohorts from the US (μg/L; MEP: 154.1; MiBP: 6.45; MCPP: 2.08; MBzP: 11.6; MECPP: 35.7; MEHHP: 21.4; MEHP: 4.92; MEOHP: 18.0) and by Bellavia et al. (2017) among women in their first trimester of pregnancy in Boston (μg/L; MEP: 137.8; MEP: 0.9; MCPP: 2.1). While prenatal exposure to BPA was positively associated with BMI z-score in boys, but negatively in girls, at 4 years of age in the Rhea cohort ²¹,we did not find any differences in BMI trajectories by levels of BPA exposure, which may be due to the high proportion of values measured <LOQ. It is also possible that trimester-specific associations may reveal vulnerable windows for in utero exposures, but there are too few studies with trimester-specific exposures to draw any conclusions. Additionally, we also chose to correct for specific gravity when controlling for urinary dilution, rather than creatinine, as creatinine is influenced by factors such as age, sex, and muscle mass ³⁰.

Exposures to BPA and phthalates metabolites are influenced by a variety of factors such as diet, physical activity, and use of personal care products. We did not include dietary intakes, consumer product use, or physical activity as potential confounders, which will impact the amount of residual confounding. We also relied on maternal BMI collected at 1 month postpartum rather than prepregnancy BMI due to limited availability of self-reported heights and weights of the mothers at recruitment into the original birth cohorts.

However, strengths of our study include the extensive follow-up within this prospective study design, allowing the construction of growth trajectories. Unlike cross-sectional studies, where there is possibility of reverse causality from higher urinary metabolites in heavier individuals due to increased dietary intake or volume of consumer product use, longitudinal studies provide a temporal sequence of events and we were able to measure exposures in utero, during a critical time of development. Additionally, the construction of BMI trajectories based on exposure in pregnancy is a novel method of assessing the potential impacts that these metabolites have on child growth on a continuous scale, rather than exploring these associations at various ages.

We found different levels of exposure to phthalates metabolites were associated with child growth by sex. In girls, the highest level of exposure to MECPP was associated with the highest predicted BMI by age 14, while lower levels of exposure to MiBP, MBzP, MEHP, and MEHHP were associated with the highest predicted BMI in boys. Our findings suggest that timing of outcome measurement may influence associations observed. Further follow-up in this study will provide us with anthropometric measures to capture trajectories through adolescence.Other studies are needed and should consider more evenly distributed measurements and confounding variables for growth modeling. Trajectory modeling is a potential method to better understand how exposures to endocrine-disrupting compounds may influence growth.