In Utero DDT and DDE Exposure and Obesity Status of 7-Year-Old Mexican-American Children in the CHAMACOS Cohort

Background: In utero exposure to endocrine disrupting compounds including dichlorodiphenyltrichloroethane (DDT) and dichlorodiphenyldichloroethylene (DDE) has been hypothesized to increase risk of obesity later in life. Objectives: The Center for the Health Assessment of Mothers and Children of Salinas (CHAMACOS) study is a longitudinal birth cohort of low-income Latinas living in a California agricultural community. We examined the relation of in utero DDT and DDE exposure to child obesity at 7 years of age. We also examined the trend with age (2, 3.5, 5, and 7 years) in the exposure–obesity relation. Methods: We included 270 children with o,p´-DDT, p,p´-DDT, and p,p´-DDE concentrations measured in maternal serum during pregnancy (nanograms per gram lipid) and complete 7-year follow-up data including weight (kilograms) and height (centimeters). Body mass index (BMI; kilograms per meter squared) was calculated and obesity was defined as ≥ 95th percentile on the sex-specific BMI-for-age Centers for Disease Control and Prevention 2000 growth charts. Results: At 7 years, 96 (35.6%) children were obese. A 10-fold increase in o,p´-DDT, p,p´-DDT, or p,p´-DDE, was nonsignificantly associated with increased odds (OR) of obesity [o,p´-DDT adjusted (adj-) OR = 1.17, 95% CI: 0.75, 1.82; p,p´-DDT adj-OR = 1.19, 95% CI: 0.81, 1.74; p,p´-DDE adj-OR = 1.22, 95% CI: 0.72, 2.06]. With increasing age at follow-up, we observed a significant trend toward a positive association between DDT and DDE exposure and odds of obesity. Conclusion: We did not find a significant positive relation between in utero DDT and DDE exposure and obesity status of 7-year-old children. However, given the observed trend with age, continued follow-up will be informative.

In utero exposure to endocrinedisrupting compounds has been hypothesized to increase risk of obesity in childhood and into adult hood (BaillieHamilton 2002;Diamanti Kandarakis et al. 2009;Heindel and vom Saal 2009). Increasing animal evidence supports a potential role of endocrinedisrupting com pounds either directly or indirectly in the pathogenesis of obesity (Grun and Blumberg 2006;Heindel and vom Saal 2009;Newbold et al. 2007Newbold et al. , 2009. Earlylife exposure might alter development of adipose tissue in terms of number, size, and distribution of adipo cytes formed, or it may affect the larger regu latory systems involved in weight homeostasis (Grun and Blumberg 2009).
The compound, dichlorodiphenyl trichloro ethane (DDT), and its primary metabolite, dichlorodiphenyldichloro ethylene (DDE), are persistent organic pollutants and known endo crine disruptors (Agency for Toxic Substances and Disease Registry 2002). A wide range of reproductive and developmental effects have been associated with DDT and DDE expo sure (Eskenazi et al. 2009), but recent studies have focused on possible obesogenic effects of these compounds. In vitro studies in two preadipocyte cell lines (3T3L1, 3T3F442) suggest that p,p´DDT has the ability to alter adipocyte differentiation, and that these effects correlate with changes induced in the expression of C/EBPα (CCAAT/enhancer binding protein α) and PPARγ (peroxisome proliferatoractivated receptor γ), the main transcription factors regulating the adipo genic process (MorenoAliaga and Matsumura 2002). The mechanisms leading to the stimu lation of these two nuclear factors, however, remain unclear. In a similar in vitro study, p,p´DDE had no effect on adipogenesis, but was shown to promote fatty acid uptake under nonstimulated conditions in mature adipo cytes (Howell and Mangum 2011). Thus, experimental evidence suggests that both compounds have the ability to promote some aspects of adipose dysfunction.
Results of epidemiologic studies of pre natal DDT and DDE exposure and child growth are inconsistent and direct com parison is limited by variations in exposure and outcome assessment (Burns et al. 2012;CupulUicab et al. 2010;Gladen et al. 2000Gladen et al. , 2004Jusko et al. 2006;Karmaus et al. 2009;Mendez et al. 2011;Pan et al. 2010;Ribas Fito et al. 2006;Valvi et al. 2012;Verhulst et al. 2009). To date, three studies have uti lized both a direct measure of in utero expo sure in maternal serum during pregnancy or cord blood and an agestandardized measure of overweight based on body mass index (BMI) zscore ≥ 85th percentile (Mendez et al. 2011;Valvi et al. 2012;Verhulst et al. 2009). All three studies reported positive associations between prenatal DDE exposure and over weight status with followup periods rang ing from 14 months to 6.5 years, but only one study reported the association with DDT exposure. In a prospective birth cohort of 518 children in Spain, maternal serum DDE was associated with increased risk of overweight (BMI zscore ≥ 85th percentile) at 14 months of age (Mendez et al. 2011). DDT was mea sured but excluded from analysis due to the low detection frequency (99% nondetect). Additionally, higher cord blood DDE levels (DDT was not measured) were associated with increased BMI SD score at 3 years of age in a prospective birth cohort study of 138 children in Belgium (Verhulst et al. 2009). The associa tion was enhanced among children of smok ing mothers. Finally, in a prospective birth cohort of 344 children in Spain, higher cord blood p,p´-DDT and p,p´-DDE levels were nonmonotonically associated with increased BMI zscore and odds of overweight (≥ 85th percentile) at 6.5 years, but associations were modified by child sex (Valvi et al. 2012). The association with p,p´-DDT was limited to males, and the association with p,p´-DDE was stronger among females.
In the present study, we examined the relation of maternal serum concentrations of o,p´- DDT, p,p´-DDT, and p,p´-DDE (DDT and DDE) during pregnancy with obesity status of 7yearold children in the Center Background: In utero exposure to endocrine disrupting compounds including dichlorodiphenyltrichloroethane (DDT) and dichlorodiphenyldichloroethylene (DDE) has been hypothesized to increase risk of obesity later in life. oBjectives: The Center for the Health Assessment of Mothers and Children of Salinas (CHAMACOS) study is a longitudinal birth cohort of low-income Latinas living in a California agricultural community. We examined the relation of in utero DDT and DDE exposure to child obesity at 7 years of age. We also examined the trend with age (2, 3.5, 5, and 7 years) in the exposure-obesity relation. Methods: We included 270 children with o,p´-DDT, p,p´-DDT, and p,p´-DDE concentrations measured in maternal serum during pregnancy (nanograms per gram lipid) and complete 7-year follow-up data including weight (kilograms) and height (centimeters). Body mass index (BMI; kilograms per meter squared) was calculated and obesity was defined as ≥ 95th percentile on the sexspecific BMI-

Methods
Study participants. The CHAMACOS study is a longitudinal birth cohort study of the effects of pesticide and other environmental exposures on the health of pregnant women and their children living in the Salinas Valley, an agricultural region in California. Pregnant women were recruited between October 1999 and October 2000. Eligible women were ≥ 18 years of age, < 20 weeks gestation at enrollment, English or Spanishspeaking, qualified for governmentsponsored health care, and planned to deliver at the county hospital. The study was approved by the insti tutional review boards at participating insti tutions and written informed consent was obtained from all mothers.
Of 601 women who were initially enrolled, 527 were followed through deliv ery of a singleton live birth that survived the neonatal period, and 417 provided a mater nal serum sample during pregnancy for DDT and DDE analysis. Of these, complete followup interview and anthropometric measurements were available for children at 2 years (n = 307), 3.5 years (n = 271), 5 years (n = 265), and 7 years (n = 270) of age. Thus, the main study sample included the 270 chil dren with 7year data.
Procedure. Details of the study are pre sented elsewhere (Eskenazi et al. 2003). Briefly, after informed consent was obtained, women were interviewed in English or Spanish by trained, bilingual, bicultural interviewers twice during pregnancy [first: mean, 13 weeks (range, 4-29), second: mean, 26 weeks (range, 18-39) gestation], shortly after delivery, and when their children were 6 months and 1, 2, 3.5, 5, and 7 years of age. During each inter view, we collected information about family sociodemographic characteristics, household member work histories, maternal characteris tics and personal habits, pregnancy and medi cal histories, and childbased developmental milestones, diet, and behavioral information.
We made child anthropometric mea surements including weight (kilograms) and height (centimeters) at each followup visit. Beginning at 2 years of age, we measured barefoot standing height to the nearest 0.1 cm using a stadiometer and standing weight to the nearest 0.1 kg using a digital scale (Tanita MotherBaby scale, model 1582; Tanita Corp., Arlington Heights, IL). Starting at age 5 years, we also measured waist circumference to the nearest 0.1 cm by placing a measuring tape around the abdomen at the level of the iliac crest, parallel to the floor. All measure ments were made in triplicate and averaged for analysis.
Laboratory analyses. Maternal serum sam ples were collected by venipuncture during routine glucose tolerance testing at approxi mately 26 weeks gestation. Serum levels of o,p´-DDT, p,p´-DDT, and p,p´-DDE were measured by isotope dilution gas chroma tography-high resolution mass spectrometry methods (Barr et al. 2003), and reported on a wholeweight basis (picograms per gram). The sample mean (± SD) levels of detection for o,p´-DDT, p,p´-DDT, and p,p´-DDE were 1.2 ± 0.6, 1.5 ± 0.8, and 2.9 ± 1.4 pg/g serum, respectively. For nondetectable val ues, a serum level equal to onehalf the detec tion limit was assigned (Hornung and Reed 1990). Lipidadjusted values (nanograms per gram) were calculated by dividing o,p´-DDT, p,p´-DDT, and p,p´-DDE on a wholeweight basis by total serum lipid content, estimated by enzymatic determination of triglycerides and total cholesterol (Phillips et al. 1989).
Statistical analyses. Lipidadjusted levels of o,p´DDT, p,p´-DDT, and p,p´-DDE were log 10 transformed and analyzed as continu ous variables. We calculated BMI at 2, 3.5, 5, and 7 years as weight (kilograms) divided by height (meters) squared. We calculated age and sexspecific BMI zscores and per centiles for each child using 2000 Centers for Disease Control and Prevention growth charts (Kuczmarski et al. 2002). Obesity was defined as being at or above the 95th percentile of the sexspecific BMI for each child's age. Overweight was defined as being at or above the 85th percentile, but less than the 95th percentile of sexspecific BMI for age. Normal weight was defined as being less than the 85th percentile of sexspecific BMI for age. Waist circumference was dichotomized into at or below versus above the 90th percentile of sex specific waist circumference for age.
All statistical analyses were performed using Stata 11.2 (StataCorp, College Station, TX). We used linear regression to examine the relation of log 10 transformed maternal serum DDT and DDE concentrations with continuous outcomes (BMI zscore). We used logistic regression to examine the relation of log 10 transformed maternal serum DDT and DDE concentrations with categorical out comes including obesity (≥ 95th vs. < 95th percentile), overweight or obesity (≥ 85th vs. < 85th percentile), and waist circumfer ence (≥ 90th vs. < 90th percentile). We used polytomous logistic regression to examine the relation of log 10 transformed maternal serum DDT and DDE concentrations with ordered threecategory weight outcome [obese (≥ 95th percentile), overweight (< 95th and ≥ 85th percentile), normal weight (< 85th percen tile)]. For all regression models, standard errors were estimated using the robust Huber-White sandwich estimator. We performed regression diagnostics to ensure no unduly influential data points with standardized residuals > 3 or < -3, but there were none.
We examined the effect of potential con founding variables identified a priori in the child obesity literature (Ebbeling et al. 2002). Potential confounders considered included family socioeconomic status (≤ poverty level, > poverty level), language spoken in the home (mostly Spanish, English/Spanish equally, mostly English), maternal educa tion level (≤ 6th grade, 7-12th grade, ≥ high school), maternal marital status (not mar ried vs. married/living as married), number of years mother had lived in the United States at the time of pregnancy (≤ 1, 2-5, 6-10, ≥ 11), maternal country of birth (United States, Mexico, other), maternal age (years), maternal prepregnancy BMI (categorical) (from reported weight and measured height at initial interview), maternal smoking dur ing pregnancy (no, yes), maternal soda con sumption during pregnancy (sodas/week), maternal BMI when the child was 7 years of age (measured weight and height), child sex, child birth weight (continuous), child birth order, whether child was breastfed [no, yes (duration in months)], child age at followup (months), and child health behaviors includ ing intake of diet and regular soda, sweet ened beverages, fast food, and sweet snacks, time spent watching television (hours/day), and time spent playing outside (hours/day) (maternal report). Covariates were kept in the model if they changed the coefficient for exposure (log 10 DDT/DDE) by > 10% or if they were independently associated with the outcome at p < 0.10. Final covariates included maternal prepregnancy BMI, birth weight, and agespecific time spent watching televi sion (continuous hours/day). We also con sidered possible interaction of child sex (male vs. female), breastfeeding status (continuous in months), and maternal prepregnancy BMI (continuous) with the exposure in all analyses by including a product term between expo sure and effect modifier. Interactions were considered significant if the p-value for the interaction term was < 0.2.
Using all children with any growth data at ages 2, 3.5, 5, or 7 years (n = 334, aver age number of observations = 3.3), we exam ined the trend with age at followup for BMI zscore and odds of obesity. We used a gen eralized estimating equation (GEE) model with the same set of covariates included in the 7year models and an additional interac tion term between exposure and exact age of child at time of evaluation. We considered the age interaction to be significant if the interaction term pvalue was < 0.20. We used the Stata lincom postestimation command to calculate the mean beta coefficient for BMI zscore (95% CI) or odds ratio (OR) for obe sity (95% CI), respectively, at each of the ages of interest. In sensitivity analyses, we repeated the final models excluding children who were low birth weight (n = 9) or preterm (n = 20). We also repeated the GEE models, limiting the analysis to children (n = 230) with com plete data at all four followup periods. Table 1 presents maternal and child charac teristics of the CHAMACOS birth cohort by obesity status at 7 years of age. Most mothers were Latina (98%), Mexicanborn (89.6%), had not completed high school (78.5%), and were living at or below the federal poverty line (70.0%). At the time of the pregnancy, moth ers were an average (± SD) of 26.1 ± 5.0 years old. Almost all mothers (96%) initiated breast feeding and the mean length of breastfeeding was 9.0 ± 8.3 months. Before pregnancy, 64.1% of mothers were overweight or obese (mean BMI = 27.6 ± 5.5 kg/m 2 ) and, by the 7year followup, this number had increased to 86.3% (mean BMI = 31.4 ± 6.2 kg/m 2 ).

Results
The 270 children were an average of 7.1 ± 0.2 years old at the 7year followup and 53.7% were female (Table 1). At birth, the children weighed an average of 3,462 ± 499 g, 9 (3.3%) were low birth weight (< 2,500 g), 20 (7.4%) were preterm (< 37 weeks), and about onethird were firstborn. The 7yearold children watched television an average of 2.0 ± 1.1 hr/day and played outside an average of 2.0 ± 1.4 hr/day. About half (53%) of children consumed less than one soda per week, but 40% consumed one or more sodas per week, and 7% consumed one or more sodas per day.
At the 7year followup, the mean (± SD) BMI zscore for the 270 children was 1.12 ± 1.0. In total, 96 (35.6%) children were classified as obese and an additional 48 (17.8%) were overweight. A total of 91 (33.7%) chil dren had a waist circumference ≥ 90th per centile for age, and of these, 82 (90.1%) were also obese. As presented in Table 1, obese chil dren were more likely to have an obese mother prepregnancy (p < 0.001) and at the 7year followup (p < 0.001). Obese children were also more likely to have higher birth weight (p = 0.05), and watch television > 2 hr/day (p = 0.01). There was no significant difference in obesity status of children by child dietary factors including soda consumption or mater nal sociodemographic indicators, including mother's country of birth, years lived in the United States, education, poverty, or mari tal status. The association with covariates was similar when different obesity measures (over weight, threecategory weight, BMI zscore, waist circumference) were considered (data not shown). Maternal serum levels of o,p´-DDT, p,p´-DDT, and p,p´-DDE were above the limit of detection for 100%, 96%, and 100% of the samples, respectively. The geometric mean [± geometric SD (GSD)] serum lev els were 1.66 (± 4.2) ng/g lipid o,p´-DDT, 20.45 (± 5.1) ng/g lipid p,p´-DDT, and 1,422 (± 3.3) ng/g lipid p,p´-DDE. Similar to what has been reported previously (Bradman et al. 2007), maternal levels of o,p´DDT, p,p´DDT, and p,p´DDE were significantly (p = 0.05) higher among mothers who were Mexicanborn, had lived in the United States ≤ 5 years, and were less educated (≤ 6th grade) (data not shown). Maternal levels were also significantly positively associated with longer duration of breastfeeding, but not with mater nal BMI (prepregnancy or at child's 7 years), child age, sex, birth order, or birth weight. Maternal levels, however, were significantly negatively associated with time the child spent watching television (p < 0.05) at 7 years.
As presented in Table 2, maternal serum concentrations in pregnancy of DDT and DDE were nonsignificantly positively related to odds of obesity and overweight when the child was 7 years of age, but not increased waist circumfer ence. After adjusting for maternal prepregnancy BMI, birth weight, and child television time, a 10fold increase in o,p´DDT or p,p´DDT was nonsignificantly associated with increased odds of obesity [o,p´DDT adjusted (adj) OR = 1.17; 95% CI: 0.75, 1.82; p,p´DDT adjOR = 1.19; 95% CI: 0.81, 1.74] and over weight (o,p´DDT adjOR = 1.32; 95% CI: 0.87, 2.00; p,p´DDT adjOR = 1.26; 95% CI: 0.87, 1.83). Results for DDE exposure were similar. A 10fold increase in p,p´DDE, was nonsignificantly associated with increased odds of obesity (adjOR = 1.22; 95% CI: 0.72, 2.06) and overweight (adjOR = 1.40, 95% CI: 0.84, 2.33). When we considered obesity status as an ordered threecategory variable (obese, over weight, normal weight), the results for DDT and DDE were consistent with estimates for the dichotomous outcomes (data not shown). We found no evidence of an association between DDT and DDE exposure and odds of waist circumference ≥ 90th percentile, after adjusting for maternal prepregnancy BMI, birth weight, and child television time (Table 2). Figure 1 presents the relationship of in utero DDT and DDE exposure with BMI zscore. After adjusting for maternal prepregnancy BMI, birth weight, and child television time, a 10fold increase in o,p´DDT or p,p´DDT was nonsignificantly positively associated with BMI zscore (o,p´DDT adjβ = 0.12; 95% CI: -0.07, 0.31; p,p´DDT adjβ = 0.10; 95% CI: -0.07, 0.27). In utero exposure to p,p´DDE was similarly nonsignificantly posi tively associated with BMI zscore (p,p´DDE adjβ = 0.12; 95% CI: -0.11, 0.35).
We found no evidence of effect modifi cation for any of the models (pinteraction > 0.20) by child sex, breastfeeding status, or maternal prepregnancy BMI (data not shown). We repeated the final models excluding 20 children who were preterm delivery or 9 who were low birth weight, and the results were not different (data not shown). The children included in the analysis did not differ signifi cantly from those who were excluded due to missing prenatal exposure or 7year anthropo metric data in terms of maternal characteristics (education, marital status, income), mater nal prepregnancy BMI or child birth weight, maternal serum DDT and DDE levels, or child obesity status (data not shown).
In Figures 2 and 3, we present the asso ciations of DDT and DDE for these children at younger ages. We observed a significant trend with age at followup towards a positive association between DDT and DDE expo sure and BMI zscore (p-interaction = 0.123 for o,p´DDT; p-interaction = 0.087 for p,p´DDT; p-interaction = 0.196 for p,p´DDE). For odds of obesity (see Figure 3), we observed a significant trend with age at followup towards a positive association between o,p´DDT (p-interaction = 0.192) and p,p´DDE (p-interaction = 0.185) but not p,p´DDT (p-interaction = 0.215) expo sure. We maintained interaction terms in all models for consistency. When we limited the sample to children who had complete data at all four followup ages, the results were com parable (data not shown).

Discussion
This longitudinal birth cohort study of a pre dominantly MexicanAmerican population residing in a California agricultural com munity provides some evidence that in utero DDT and DDE exposure may alter risk for obesity with age. We found that current data do not support a statistically significant posi tive association between in utero DDT and DDE exposure and obesity status of 7year old children. However, we observed a signifi cant trend with age (2, 3.5, 5, 7 years) toward a positive association between maternal serum concentrations of DDT and DDE and odds Table 2. Results of logistic regression models for associations of in utero DDT and DDE exposure with childhood obesity, overweight or obesity, and waist circumference at 7 years, CHAMACOS, 2007-2008.
The prevalence of obesity at 7 years of age in this study is very high (36%). This preva lence is higher than the 18% prevalence reported in NHANES for all U.S. children, 6-11 years old, and also higher than the 22% prevalence reported for MexicanAmerican children of the same age range (Ogden et al. 2012). Further, more than half of the children in this study were overweight or obese, and 1% had a BMI of ≥ 30 (the adult definition of obesity). In fact, the prevalence of overweight/ obesity (53.3%) in this study is twice the preva lence (26.7%) reported in the Spanish prospec tive birth cohort study with a similar length of followup (6.5 years) (Valvi et al. 2012).
The results are consistent with the positive associations reported in the three prospec tive birth cohort studies that used a direct measure of prenatal DDT and DDE exposure (maternal serum at pregnancy or cord blood) and a standardized measure of overweight (BMI zscore ≥ 85th percentile) at 14 months of age (Mendez et al. 2011), at 3 years of age (Verhulst et al. 2009), and at 6.5 years of age (Valvi et al. 2012). Valvi et al. (2012) reported a nonmonotonic increase in the risk for overweight at 6.5 years that was modified by child sex; the risk with cord blood DDT exposure was limited to males. In contrast to Valvi et al. (2012), we found no differ ence in associations by child sex with either DDT or DDE exposure. Consistent with other previous studies (Mendez et al. 2011;Valvi et al. 2012;Verhulst et al. 2009), we found no change in associations after exclud ing low birth weight and preterm children. We also found no evidence of a relation of DDT or DDE exposure with waist circumfer ence, but these children are still young and waist circumference may not be as sensitive a measure of overweight status in prepuber tal populations. To our knowledge, no other studies have examined the relation of DDT and DDE exposure on waist circumference.
In this study, we found similar associations of o,p´DDT and p,p´DDT on 7yearold child obesity status. Two other studies have examined prenatal o,p´DDT and p,p´DDT separately but neither used an age and sex standardized measure of BMI (Gladen et al. 2004;Jusko et al. 2006). Jusko et al. (2006) found no association between maternal serum o,p´DDT, p,p´DDT or p,p´-DDE and child weight zscore or height zscore at 5 years of age, but did not examine BMI zscore. In a subsample of 304 males from the Collaborative Perinatal Project, no association was reported between prenatal o,p´DDT, p,p´DDT or p,p´DDE exposure and BMI with followup to 10-20 years of age (Gladen et al. 2004). Given the very wide age range of followup, interpretation is limited without an age standardized measure of BMI (zscore). We found similar associations for p,p´DDT or p,p´DDE exposure on 7yearold child obesity status measures. Given the high correlation in this study between maternal serum levels of DDT and DDE (r = 0.8 to 0.9), it is difficult to separate out the individual associations of each compound.
The results of this study are biologi cally plausible. In experimental studies, both DDT and DDE have been associated with adipose dysfunction (Howell and Mangum 2011;MorenoAliaga and Matsumura 2002). Given that DDT, an estrogen agonist, is metabolized to DDE, an androgen antago nist, there may be more than one mechanism to consider. Early developmental exposure to DDT and DDE could affect weight by affect ing normal weight homeostasis either directly on adipose cells through differentiation and proliferation or indirectly via disruption of the endocrine feedback loop (Cooke and Naaz 2004;DiamantiKandarakis et al. 2009).
This study has several strengths. The CHAMACOS study is a longitudinal birth cohort with a relatively long followup period for which considerable information was col lected about potential confounders. The study population is relatively homogenous (Mexican American) for factors such as diet, breastfeeding, country of origin, and socio economic status, which can reduce uncon trolled confounding. We were able to measure o,p´DDT, p,p´DDT, and p,p´DDE expo sure in maternal serum collected during the pregnancy. Exposure levels were high rela tive to other Mexican Americans (Centers for Disease Control and Prevention 2004), likely due to the mothers' recent immigration from Mexico, but there was a wide range of expo sure. Finally, we used a standardized measure of overweight based on BMI zscore which facilitates comparison across studies.
This study has some limitations. First, the positive associations we observed between in utero DDT and DDE exposure and obe sity status at 7 years were not statistically sig nificant; we cannot eliminate chance as an alternative explanation. It will be important to follow up the CHAMACOS cohort through puberty to examine the longitudinal trends in obesity with DDT and DDE exposure and to be able to consider the impact of such Figure 2. Trend in estimated association for BMI z-score at 2, 3.5, 5, and 7 years with in utero log DDT and DDE exposure, CHAMACOS 2007-2008. Age-specific association was derived using lincom after GEE model for exposure and interaction with exact age at measurement. Measures were based on participants with growth data at any age (n = 334, average no. of observations = 3.3). All models were adjusted for child's exact age in months, maternal prepregnancy BMI, birth weight, and child age-specific television time. Age (years) 7 3.5 2 5 7 3.5 Figure 3. Trend in estimated OR for obesity at 2, 3.5, 5, and 7 years with in utero log DDT and DDE exposure, CHAMACOS 2007-2008. Age-specific associations were derived using lincom after GEE model for exposure and interaction with exact age at measurement. Measures were based on participants with growth data at any age (n = 334, average no. of observations = 3.3). All models were adjusted for child's exact age in months, prepregnancy BMI, birth weight, and child age-specific television time. Age (years) 7 3.5 2 5 7 3.5 volume 121 | number 5 | May 2013 • Environmental Health Perspectives exposure on the adrenal hormonemediated increase in weight and the sex steroid-induced pubertal growth spurt. Second, we were not able to consider the potential confounding effects of other chemical exposures such as polychlorinated biphenyls (PCBs) or hexa chlorobenzene that have previously been asso ciated with child obesity (Smink et al. 2008;Valvi et al. 2012). However, in CHAMACOS, concentrations of DDT and DDE were only weakly associated with PCB congeners and hexachloro benzene (r = 0.09-0.19) (Chevrier et al. 2008). Finally, compared with the 417 mothers who had maternal serum DDT and DDE measurements, only 270 children had complete anthropometric data at 7 years. However, in utero DDT and DDE expo sure levels of those with and without 7year anthropometric data were not significantly different. In addition, predictors of maternal levels of o,p´DDT, p,p´DDT, and p,p´DDE were similar to those reported previously in the larger group (Bradman et al. 2007).
In summary, we examined the asso ciation of in utero exposure to o,p´DDT, p,p´DDT, and p,p´DDE with body weight in the CHAMACOS longitudinal birth cohort through 7 years of age. The present data do not support a statistically significant association between in utero DDT and DDE exposure and obesity status of 7yearold children. However, the need for further research in aging children is supported by the statistically significant trend with age (2, 3.5, 5, 7 years) toward a positive association between maternal serum concen trations of DDT and DDE and the odds of childhood obesity. Continued followup of the CHAMACOS cohort will be informative.