. From 2003–2005, we recruited 8- to 9-year-old boys in Chapaevsk, Russia, to participate in the Russian Children’s Study, as previously described (Burns et al. 2012
). Enrollment exclusion criteria included being institutionalized or having severe cerebral palsy. A total of 499 boys were thus identified, and they have been followed annually since recruitment. For the present analysis, 10 boys with severe chronic illnesses that could affect growth were excluded. Of the remaining 489, two pubertal boys were also excluded from the analysis because of implausibly low IGF-1 concentrations (< 50 ng/mL) despite normal height and body mass index (BMI) z
-scores. Of the remaining 487 boys, 394 met inclusion criteria for this analysis, specifically, by availability of a baseline BLL (at 8–9 years) and follow-up serum IGF-1 concentration at both the 2-year (at 10–11 years) and the 4-year (at 12–13 years) follow-up visits. None of the participants had IGF-1 measured at baseline or BLL measured at follow-up. The Russian Children’s Study was approved by the human studies institutional review boards (IRBs) of the Chapaevsk Medical Association, Harvard School of Public Health, Brigham and Women’s Hospital, and University of Massachusetts Medical School. The parent or guardian of each Russian Children’s Study participant signed an informed consent form, and each boy signed an assent form. The present analysis was a secondary data analysis that was exempt from requirement for IRB review of already collected, deidentified data under federal and Children’s Hospital Boston policies.
Study assessment protocol
. At study entry, boys underwent a physical examination and blood collection. We used a validated Russian Institute of Nutrition semiquantitative food-frequency questionnaire to estimate dietary intakes during the previous year (Martinchik et al. 1998
; Rockett et al. 1997
). Mothers or guardians also completed nurse-administered health and lifestyle questionnaires that included information on birth, family and child medical histories, occupational and residential history, and measures of socioeconomic status (SES) such as household income and parental education. We obtained birth weight and gestational age from medical records.
Physical examination. A single study nurse who did not have knowledge of the boys’ BLLs performed standardized anthropometric examinations, and a single investigator (O.S.) performed pubertal assessments at study entry and at annual follow-up visits. We measured height to the nearest 0.1 cm with a stadiometer. We measured weight to the nearest 100 g with a metric scale. We calculated age-adjusted percentiles for BMI (kilograms per meter squared) using the World Health Organization (WHO) standards (WHO 2011). For this analysis, pubertal status was based on testicular volume measured by Prader beads (orchidometer).
Blood lead levels. Venous blood samples (3.0 mL) were collected in trace metal-free Vacutainer tubes (Becton-Dickinson, Franklin Lakes, NJ, USA), after cleansing the venipuncture site with alcohol. Whole-blood samples were diluted with a matrix modifier solution and analyzed using Zeeman background corrected, flameless graphite furnace, atomic absorption spectrometry (ESA Laboratories, Chelmsford, MA, USA). BLLs below the limit of detection (1 μg/dL) were imputed as 0.5 μg/dL for 9 (2.2%) of 394 boys.
Serum IGF-1 concentrations. Serum IGF-1 concentrations were measured by a chemiluminescent immunometric assay using Siemens Immulite 2000 (Siemens AG, Munich, Germany). The assay is highly specific for IGF-1 with undetectable cross-reactivity with insulin, pro-insulin, luteinizing hormone (LH), thyroid-stimulating hormone, or insulin-like growth factor-2. The detection limit was 20 ng/mL; no IGF-1 values were below the limit of detection. The intra-assay coefficient of variation (CV) was < 3.9%, and the inter-assay CV was < 8.1% for the Immulite 2000 kit.
Statistical analysis. We used repeated measures analysis to estimate the association between BLL measured at 8–9 years of age and serum IGF-1 concentration at both the 2-year (at 10–11 years) and 4-year (at 12–13 years) follow-up visits. The distribution of BLLs was right-skewed with outliers. We considered several different ways of evaluating BLLs, including dichotomized as high (≥ 5 μg/dL) versus low (< 5 μg/dL) based on the new CDC threshold, as a continuous measure (log-transformed), and categorized as 0–2, 3–4, 5–9, or ≥ 10 μg/dL.
We fit linear regression models using a generalized estimating equation (GEE) approach to account for the repeated measures and slight skewness of IGF-1 concentrations. We first fit GEE linear regression models to evaluate unadjusted associations of high versus low BLL (≥ 5 or < 5 μg/dL) with serum IGF-1 concentrations. Next, we created a full multivariable model that included dichotomized BLL, birth weight (continuous), gestational age at birth (continuous), breastfeeding duration (< 12, 12–24, or > 24 weeks), maximum parental education (secondary education or less, junior college/technical training, or university graduate), monthly household income (< US$175, $175–250, or > $250), nutritional intake (total caloric intake and percent calories from protein, fat, and carbohydrate as continuous variables), baseline and follow-up age (continuous), and baseline and follow-up BMI [underweight (BMI < 10th percentile), overweight (BMI > 85th percentile), or normal weight]. We decided a priori
not to consider height as a covariate due to its strong correlation with IGF-1 during puberty (Silbergeld et al. 1986
). We then reduced this model by excluding covariates that did not predict the outcome with p
≤ 0.10, or did not change the estimated association between BLL and IGF-1 by > 10% when removed from the model, resulting in a final model that included baseline parental education, birth weight, nutritional intake, and baseline and follow-up age and BMI.
Because pubertal status is influenced by lead exposure (Selevan et al. 2003
; Williams et al. 2010
) and may be considered in the causal pathway between BLL and serum IGF-1 concentration, we fit the final reduced model both with and without adjustment for pubertal status [categorized based on testicular volume (TV) as prepubertal, ≤ 3 mL TV; early pubertal, > 3–6 mL TV; or mid-pubertal, > 6–15 mL TV]. No boys in this analysis had a TV > 15 mL. We also evaluated whether the association between BLL and IGF-1 concentration differed by pubertal stage by including an interaction term between pubertal status (prepubertal, early pubertal, or mid-pubertal) and BLL (≥ 5 or < 5 μg/dL).
Sensitivity analyses included the following modifications to the final reduced model: a) inclusion of maternal and paternal heights as potential confounding variables for the subset of boys who had these values available (n = 337), b) inclusion of 44 additional boys with only one follow-up IGF-1 concentration, c) BLL categorized as ≤ 2, 3–4, 5–9, or ≥ 10 μg/dL, and d) BLL modeled as a natural log-transformed continuous variable. We conducted all analyses using SAS version 9.3 (SAS Institute Inc., Cary, NC, USA), and we considered two-sided p-values ≤ 0.05 statistically significant.