Predictors of Serum Chlorinated Pesticide Concentrations among Prepubertal Russian Boys

Background: Few studies have evaluated predictors of childhood exposure to organochlorine pesticides (OCPs), a class of lipophilic persistent chemicals. Objectives: Our goal was to identify predictors of serum OCP concentrations—hexachlorobenzene (HCB), β-hexachlorocyclohexane (β-HCH), and p,p-dichlorodiphenyldichloroethylene (p,p´-DDE)—among boys in Chapaevsk, Russia. Methods: Between 2003 and 2005, 499 boys 8–9 years of age were recruited in a prospective cohort. The initial study visit included a physical examination; blood collection; health, lifestyle, and food-frequency questionnaires; and determination of residential distance from a local factory complex that produced HCB and β-HCH. Fasting serum samples were analyzed for OCPs at the U.S. Centers for Disease Control and Prevention. General linear regression models were used to identify predictors of the boys’ serum HCB, β-HCH, and p,p´-DDE concentrations. Results: Among 355 boys with OCP measurements, median serum HCB, β-HCH, and p,p´-DDE concentrations were 158, 167, and 284 ng/g lipid, respectively. Lower body mass index, longer breastfeeding duration, and local dairy consumption were associated with higher concentrations of OCPs. Boys who lived < 2 km from the factory complex had 64% (95% CI: 37, 96) and 57% (95% CI: 32, 87) higher mean HCB and β-HCH concentrations, respectively, than boys who lived ≥ 5 km away. Living > 3 years in Chapaevsk predicted higher β-HCH concentrations, and having parents who lacked a high school education predicted higher p,p´-DDE concentrations. Conclusions: Among this cohort of prepubertal Russian boys, predictors of serum OCPs included consumption of local dairy products, longer local residence, and residential proximity to the local factory complex. Citation: Lam T, Williams PL, Burns JS, Sergeyev O, Korrick SA, Lee MM, Birnbaum LS, Revich B, Altshul LM, Patterson DG Jr, Turner WE, Hauser R. 2013. Predictors of serum chlorinated pesticide concentrations among prepubertal Russian boys. Environ Health Perspect 121:1372–1377; http://dx.doi.org/10.1289/ehp.1306480


Terms of Use
This article was downloaded from Harvard University's DASH repository, and is made available under the terms and conditions applicable to Other Posted Material, as set forth at http:// nrs.harvard.edu/urn-3:HUL.InstRepos:dash.current.terms-of-use#LAA Introduction Persistent, lipid-soluble organochlorine pesticides (OCPs) such as hexachlorobenzene (HCB), β-hexachlorocyclohexane (β-HCH), dichlorodiphenyltrichlorethane (DDT), and its primary metabolite p,p´-dichlorodiphenyldichloroethylene (p,p´-DDE) are ubiquitous in the environment. Although these insecticides and fungicides were banned in the 1970s (United States) and in the 1980s (Russia) (Barber et al. 2005;Breivik et al. 1999;Jaga and Dharmani 2003), DDT is still used in some countries to control malaria and yellow fever (Jaga and Dharmani 2003), and HCB and β-HCH are generated as byproducts during the manufacture of other chlorinated chemicals (Courtney 1979;Jung et al. 1997). In addition to biomagnifying up the food chain, these compounds have long half-lives in both the environment and the body, ranging from years to decades, and persist long after use has ceased (Barber et al. 2005;Courtney 1979;Jaga and Dharmani 2003;Jung et al. 1997). Human exposure typically occurs through diet (e.g., fatty fish, dairy products, meats, poultry) and less commonly through inhalation or dermal absorption (Darnerud et al. 2006;Gasull et al. 2011). OCPs are stored in adipose tissue, concentrated in breast milk (Rogan et al. 1986), and can be passed from mother to child via transplacental transfer or breastfeeding, a primary route of exposure for children (Rogan et al. 1986;Sala et al. 2001). Compared with adults, young children have disproportionately elevated exposures to pesticides because they have higher ventilatory rates and food consumption relative to body weight (National Research Council 1993) and greater hand-to-mouth transfer of contaminated soil and dust. Moreover, the ability to metabolize, detoxify, and excrete pesticides may be reduced in young children (Landrigan et al. 2003).
HCB, β-HCH, and p,p´-DDE have been associated with adverse health effects in animal studies, including neurodevelopmental toxicity (Courtney 1979;Ecobichon et al. 1990), cancer [Courtney 1979; International Agency for Research on Cancer (IARC) 1987(IARC) , 1991, decreased male reproductive performance (e.g., mating index), and testicular abnormalities (Courtney 1979;Gralla et al. 1977;Gray et al. 2001;Van Velsen et al. 1986). Among humans, there is evidence of neurodevelopmental toxicity following chronic exposure (reviewed by Ecobichon et al. 1990). Understanding the predictors of OCP exposure can improve child health and inform public health recommendations and policies to reduce childhood exposure. To date, some studies have evaluated predictors of childhood exposure to OCPs (Barr et al. 2006;Den Hond et al. 2009;Gallo et al. 2011;Karmaus et al. 2001;Lu et al. 2010;Petrik et al. 2006).
We assessed predictors of serum HCB, β-HCH, and p,p´-DDE concentrations among children residing in Chapaevsk, Russia, a city of approximately 72,000 people located 950 km southeast of Moscow that has a history of environmental organochlorine compound contamination. In the 1930s, a complex of factories began producing chemical agents for military use (e.g., mustard gas, lewisite). From 1967 to 1987, production focused on organo chlorine chemicals including HCB, HCH, and its derivatives (α-HCH, β-HCH, and γ-HCH) (Akhmedkanov et al. 2002). However, DDT was never produced . By 2003, production of all chemicals ceased. Environmental contamination by HCB and β-HCH may have resulted from improper disposal, storage of hazardous waste from factories, or environmental release of organochlorine by-products of the manufacturing process ).

Study population.
The Russian Children's Study is an ongoing prospective cohort study of 499 prepubertal boys in Chapaevsk, Russia. Briefly, 623 boys 8-9 years of age were identified from the town-wide health insurance system. Of these, 572 were eligible and 516 agreed to participate (recruitment rate of 90%), although 17 boys were subsequently excluded because they were orphans (precluding collection of residential history and other information). The first 144 9-year-old boys enrolled did not have OCPs measured, leaving 355 boys with serum OCP measurements. The study was approved by the Human Studies Institutional Review Boards of the Chapaevsk Medical Association, Harvard School of Public Health, University of Massachusetts Medical School, and Brigham and Women's Hospital. The parent/guardian of each child gave informed consent, and the children signed assent forms before participation.
At study entry, eligible boys had a physical examination and a fasting blood sample collected. A study nurse administered health and lifestyle questionnaires to the parent/guardian that included questions concerning the child's birth history, medical history, and physical activity; parental occupational and educational background; family medical history and residential history; and household income. The child's current residence was identified using an electronic map.
Physical examination. At the initial visit, height and weight measurements were obtained according to a written protocol. Height in stocking feet was measured to the nearest 0.1 cm using a stadiometer. Weight was measured to the nearest 100 g with a metric scale. Body mass index (BMI) was calculated as kilograms divided by height in meters squared.
Blood sample analysis. Fasting blood samples collected at study entry were centrifuged, and the serum was aliquoted and stored at -35°C until shipment on dry ice to the U.S. Centers for Disease Control and Prevention for analysis. Five OCPs (HCB, β-HCH, p,p´-DDE, aldrin, and mirex) were measured in the 355 boys. The samples, including method blank and quality control samples, were spiked with 13 C 12 -labeled pesticides, extracted by a C18 solid-phase extraction (SPE) followed by a multicolumn automated cleanup and enrichment procedure using either large-volume (Turner et al. 1997) or small-volume SPE (Sjödin et al. 2004). High-resolution mass spectrometry in selective ion monitoring was used to analyze the samples (Barr et al. 2003). Because aldrin and mirex were below the limit of detection (LOD) in almost all (> 95%) participants (median aldrin LOD, 0.40 ng/g lipid; median mirex LOD, 4.5 ng/g lipid), we limited our analysis to HCB, β-HCH, and p,p´-DDE, all of which were above the LOD in all samples (median HCB LOD, 4.1 ng/g lipid; median β-HCH LOD, 4.5 ng/g lipid; median p,p´-DDE LOD, 5.6 ng/g lipid). The analytical coefficients of variation for individual OCPs in quality assurance/quality control samples ranged between 10% and 15% over the course of the study. All OCP concentrations were expressed on a per-lipid basis. Total serum lipid content was determined from enzymatic measurements of total cholesterol and triglycerides (Phillips et al. 1989).
Distance information. The straight-line distance from each participant's residence to the primary factory complex was calculated using ArcView GIS 3.0 (ESRI, Redlands, CA, USA) by the Moscow Ecological Analytical Center.
Food frequency questionnaire (FFQ). Study nurses administered a validated Russian Institute of Nutrition FFQ that was modified to assess local food consumption (Hauser et al. 2005). The FFQ included questions on the usual frequency of consumption and portion sizes (classified based on photographs) of > 70 food items consumed during the previous year (Martinchik et al. 1998). The participating boy and his parent/guardian completed the FFQ together.
Statistical analysis. We used Spearman correlations to assess relationships among OCPs. We used linear regression models to identify predictors of the serum concentration of each OCP. To improve normality and reduce the influence of outliers, serum OCP concentrations were log 10 -transformed. Potential predictors were identified using a priori knowledge from the literature. All covariates (except food consumption) with p ≤ 0.20 in bivariate analyses were included in an initial multivariable linear regression model for each OCP that was subsequently reduced to include predictors with p ≤ 0.10 only. In addition, breastfeeding duration was retained in all models because it is a known route of exposure for children, and there was a significant linear trend (p ≤ 0.05) in the concentration of all three OCPs over categories of breastfeeding duration. We categorized breastfeeding duration as none, 1-13 weeks, and > 13 weeks to limit the potential influence of a small number of observations with a long duration of breastfeeding. To account for nonlinearity, BMI was grouped according to World Health Organization (WHO) child growth standards: underweight (> 1 SD below the mean), normal, and overweight/ obese (> 1 SD above the mean) (de Onis et al. 2007). We modeled other continuous predictors as simple continuous variables and as categorical variables, and examined point estimates and compared the goodness-of-fit of each model based on adjusted coefficient of determination (R 2 ) values to determine the most appropriate coding.
Individual food items were categorized into groups (eggs, dairy products, poultry, nonpoultry meats, fish, and fruits and vege tables), and the usual frequency and portion size (grams/year) of each category was determined. In addition, we estimated consumption of local foods in each food category, and included each local food category [in tertiles (low, medium, or high) or dichotomized as any vs. none when < 50% reported eating local foods] in multivariable models while simultaneously adjusting for total consumption of each food category (in tertiles). Local foods reflect possible local contamination, whereas total foods are adjusted in models to reflect background levels. Because of the specific interest in dietary sources of OCP concentrations, each food group was considered separately as well as simultaneously, adjusting for other factors.
For ease of interpretation, regression coefficients are presented as the estimated percent change in the serum concentration of each OCP with a 1-unit increase in exposure volume 121 | number 11-12 | November-December 2013 • Environmental Health Perspectives (for continuous variables) or relative to the reference category (for categorical variables) (10 β , where β is the regression coefficient for a given predictor), holding all other variables constant. We also estimated adjusted mean concentrations of each OCP according to residential distance from the factory using least-square means, adjusted for all other predictors in the final model for each OCP. Additionally, we performed sensitivity analyses using log 10 -transformed whole-weight serum OCP concentrations adjusted for total lipids instead of direct lipid-adjusted OCPs in the final model. A p-value of < 0.05 was considered statistically significant. Tests for trend were performed by modeling categorical variables as an ordinal variable using integer values (0, 1, 2). All data analyses were performed using SAS version 9.2 (SAS Institute Inc., Cary, NC, USA).

Results
Study population. Characteristics of the 355 boys are shown in Table 1. At study entry, 84% of boys were 8 years old, 17% were overweight/obese, and 25% were underweight (de Onis et al. 2007). Among boys with and without serum OCP measurements (n = 355 vs. 144), there were no significant differences in height, weight, BMI z-scores, birth, or family characteristics, but there was a significant difference in household income (44% vs. 26% of families in the highest income category, respectively). Also, more boys with measured serum OCP concentrations had fathers who were employed at the factory complex compared with those without OCP measurements (14% vs. 9%).
Predictors of serum HCB, β-HCH, p,p´-DDE. BMI and residential distance from the factory complex were the strongest predictors, explaining 18% and 23% of the variability in the serum concentrations of HCB and β-HCH, respectively. All other covariates combined explained an additional 3% (HCB) and 13% (β-HCH) of the variability. For p,p´-DDE, BMI and breastfeeding duration combined explained 23% of the variability, all other model covariates explained an additional 7%.
Factors that predicted higher concentrations of only one of the OCPs were having parents who had a high school education or less (57% higher serum p,p´-DDE relative to boys whose parents had university/ postgraduate training; 95% CI: 17, 111%), living in Chapaevsk for ≥ 3 years (higher β-HCH concentrations compared with those living < 3 years in Chapaevsk, trend p = 0.003) and having a father employed at the factory complex (higher β-HCH concentrations, p = 0.10).
The percent of variation (R 2 ) explained by the predictors in each final OCP model ranged from 0.21 (for the model of HCB) to 0.36 (for β-HCH). Sensitivity analyses of predictors for whole weight of log 10transformed serum OCPs adjusted for total lipids were consistent with predictions of lipid-adjusted serum OCPs (see Supplemental Material, Table S1).

Discussion
In the present study, we measured serum OCP concentrations and identified several demographic, lifestyle, and environmental predictors among boys living in Chapaevsk, Russia, a town contaminated by previous industrial activity. These results complement a publication describing predictors of dioxins and polychlorinated biphenyls (PCBs) among these boys (Burns et al. 2009). Despite the young age of our cohort (8-9 years), concentrations of OCPs were similar to or higher than concentrations reported for somewhat older pediatric populations (range, 8-19 years) in the United States and Europe (Barr et al. 2006;Den Hond et al. 2011;Patterson et al. 2009;Petrik et al. 2006).
Consistent with other studies of persistent organic pollutants in this cohort and other populations (Burns et al. 2009;Den Hond et al. 2009;Gallo et al. 2011;Humblet et al. 2010;Karmaus et al. 2001), lower BMI predicted higher serum OCP concentrations. This finding may be attributable to a smaller volume of distribution in boys with lower BMI, resulting in higher serum concentrations (Wolff et al. 2005).
Breastfeeding is a known route of early life exposure to lipophilic persistent compounds (Rogan et al. 1986). Consistent with other studies (Barr et al. 2006; Den Hond  Gallo et al. 2011;Karmaus et al. 2001), longer breastfeeding duration (> 13 weeks) predicted higher OCP concentrations. Although breastfeeding in our cohort ended years before OCP measurement, childhood concentrations of lipophilic compounds track closely with breastfeeding exposure in infancy (Patandin et al. 1999). Residential distance from the primary factory may provide insight on chemical-specific pathways of exposure in this area. Specifically, living < 2 km from the complex was associated with higher serum HCB and β-HCH concentrations, which is consistent with these compounds having been manufactured at the factory as a source of exposure. However, serum p,p´-DDE levels were highest for boys living 2 to < 5 km from the factory and only moderately and nonsignificantly elevated among boys living within 2 km. DDT, the parent compound for p,p´-DDE, was not manufactured at the complex, and other exposure sources likely contributed to the boys' p,p´-DDE levels.
Duration of residence in Chapaevsk and father's prior employment at the factory were positive predictors of β-HCH. We hypothesized that longer residence in Chapaevsk would be associated with higher exposure to and bioaccumulation of both β-HCH and HCB, because both were produced locally. Therefore, it is unclear why duration of residence was a significant predictor of β-HCH but not HCB. Similarly, fathers' prior occupation at the factory was not a significant predictor of HCB, despite the same potential for exposure from residues on the fathers' work clothing, boots, tools, or skin (Lu et al. 2000). Mother's employment at the factory did not predict any of the OCPs, but only 5% of mothers reported previous employment at the factory, limiting power to detect a statistically significant association.
Non-occupational exposure to HCB, β-HCH, and p,p´-DDE is primarily dietary (Darnerud et al. 2006;Gasull et al. 2011); therefore, we expected consumption of local foods high in fat, such as dairy and fish, to predict higher serum concentrations. However, local dairy consumption was the only significant predictor of all three OCPs. Studies of OCP exposure and local diet among children have been limited to two assessments concerned about local environmental contamination (Den Hond et al. 2009;Gallo et al. 2011), that did not find an association with local dairy. Dairy consumption has been associated with serum OCP concentrations in several adult populations, although none differentiated whether dairy foods were from local sources (Arrebola et al. 2009(Arrebola et al. , 2012Lee et al. 2007). Although previous studies among children reported associations between serum OCPs and consumption of local fish (Gallo et al. 2011) and consumption of fatty meats and vegetables (Den Hond et al. 2009), these foods were not significant predictors in our cohort.
We previously reported that consumption of most local foods predicted higher serum dioxins and PCBs in the same study cohort (Burns et al. 2009), consistent with findings reported for other study populations concerned about environmental contamination (Choi et al. 2006;Gallo et al. 2011;Schecter et al. 2003). It is unclear why local food consumption, apart from dairy, was associated with dioxin-like compounds but not serum OCP concentrations in our study cohort.
One limitation of our diet analyses is the inability to assess consumption of specific types of fish or fat content. We had only father's reported employment history with no independent verification, and therefore can only speculate on the association observed with β-HCH. Although we attempted to evaluate many potential determinants of these exposures (e.g., parental education), there were probably other predictors that we could not assess.
A major strength of this study is the large sample size of young boys with serum HCB, β-HCH, and p,p´-DDE measurements. For all three OCPs, all serum concentrations were above the limit of detection with wide ranges of concentrations. Detailed dietary information, including local food consumption, as well as calculated residential distance from the factory complex, was also available. In this context, this study contributes to understanding of determinants of serum OCP levels among children, and in particular, highlights the potential importance of local risk factors for exposure.

Conclusion
Our findings suggest that contamination from the local factory may be an important source of HCB and β-HCH exposure for boys in Chapaevsk. Residential distance from the primary factory was a significant predictor of serum HCB and β-HCH, both of which were manufactured at the complex. Father's past employment at the factory and longer residence in Chapaevsk also predicted higher serum β-HCH, and local dairy consumption predicted higher serum concentrations of all three OCPs; these results add further support to local environmental contamination, at least partly from the factory, as a source of exposure. Consistent with other studies, longer breastfeeding duration and lower BMI predicted higher serum OCPs.
Our findings provide insight on determinants of OCP exposure, which may lead to local monitoring and continuation of remediation measures (e.g., soil removal) to reduce childhood and community exposure. Although our findings suggest that local dairy consumption and longer breastfeeding duration are primary determinants of OCP exposure, it would be premature to recommend reduced intake of local dairy foods or reduced breastfeeding without fully understanding the exposure pathway and the risk-benefit trade-offs from such a recommendation. It is important to keep in mind that these local food products were central to the children's diet in this region, and that breastfeeding has well-established benefits. Recommendations to prevent childhood exposure include environmental cleanup of contaminated areas, regulatory enforcement of safe practices for industrial waste disposal and emissions control, and preferential consumption, when available, of foods produced in noncontaminated areas. Figure 1. Adjusted mean OCP serum concentrations of Russian prepubertal boys in relation to residential distance from factory complex. Adjusted means (95% CIs) calculated using least-square means adjusted for all other covariates in Table 4. Adjusted mean OCP serum concentrations (ng/g lipid) HCB p,p'-DDE β-HCH < 2 km 2 to < 5 km ≥ 5 km