The PIN Study followed a cohort of central North Carolina women from early pregnancy through 12 months postpartum (PIN 2011
). In the third phase of PIN, women were recruited from the University of North Carolina prenatal care clinic, and they delivered infants at University of North Carolina hospitals between 2001 and 2005 (n
= 2009). The PIN Postpartum Study followed women through the first year postpartum (n
= 689). The PIN Babies Study began in January 2004 to follow the children of women participating in the later years of the PIN study through 3 years of age (n
= 585). By design, all children in the PIN Babies Study were singleton births free from major birth defects. This analysis was further limited to infants who were primarily breast-fed at least 3 months, to obtain a milk sample (n
= 304). Women participating in the PIN Babies Study were more likely to be white, have higher educational attainment, and be older than mothers in the larger PIN cohort (Daniels et al. 2010
; PIN 2011
). Self-administered questionnaires, telephone interviews, and home visits were used to collect pregnancy and postpartum health and lifestyle information throughout the PIN studies. All study protocols were approved by the institutional review boards at the University of North Carolina–Chapel Hill and the Centers for Disease Control and Prevention, and all mothers provided informed consent.
Details of the breast milk sample collection were published previously (Daniels et al. 2010
). Briefly, lactating women provided a morning milk sample at 3 months postpartum using the milk collection kit provided. Women stored the milk in their freezer until a research team picked up the sample during a home data collection visit later the same day. Samples were then stored at –80°C until analyzed.
The Organic Analytical Toxicology Branch of the National Center for Environmental Health at the Centers for Disease Control and Prevention (Atlanta, GA) analyzed milk samples for nine PBDE congeners using previously described methods (Sjödin et al. 2004
). The milk samples were added to diatomaceous earth packed in a solid-phase extraction cartridge (3 mL) and 13
C-labeled internal standards were added. Target analytes and lipids were extracted using an automated modular solid-phase extraction system (Cambridge Isotope Laboratories, Andover, MA), which dried the sample onto the diatomaceous earth with pressurized nitrogen and eluted analytes and lipids with dichloromethane. Lipid content was determined gravimetrically, and the final analytical determination of PBDEs was performed by gas chromatography/isotope-dilution high-resolution mass spectrometry. Both wet weight and lipid normalized concentrations (reported as nanograms per gram lipid) were produced. PBDE values below the limit of detection (LOD) were assigned a concentration of the LOD divided by the square root of 2. For quality control, two quality control and two blank samples were added to each batch of 16 study samples. Quality assurance practices in the laboratory were regularly monitored (Sjödin et al. 2004
Five PBDE congeners—BDEs 28, 47, 99, 100, and 153—were detected in > 91% of samples. We evaluated the level of these congeners, as well as their sum, in relation to developmental outcomes. The four PBDEs congeners infrequently detected [in < 70% of the PIN Babies sample; BDEs 66, 85, 154, and 183 (Daniels et al. 2010
)] were not evaluated in relation to developmental outcomes.
Social and emotional development assessment.
Mothers completed the Infant–Toddler Social and Emotional Assessment (ITSEA), of their child between 24 and 36 months of age (Carter and Briggs-Gowan 2006
). The ITSEA is a validated measure of social and emotional development that uses a parent-report questionnaire to assess a wide array of social-emotional and behavioral problems and competencies. Parents rate a broad range of 166 behaviors that are part of typical development, but can be problematic if they occur too frequently or infrequently, as well as behaviors that occur infrequently and represent deviations from the normal course of development. The parent scores each item as 0 = not true/rarely, 1 = somewhat true/sometimes, 2 = very true/often. Mothers were not aware of the levels of PBDEs in their breast milk at the time the ITSEA was completed.
The ITSEA produces scores for four primary domains of social and emotional behavioral development for young children; each domain contains several subscales: externalizing domain (activity/impulsivity, aggression/defiance, peer aggression), internalizing domain (depression/withdrawal, general anxiety, separation distress, inhibition to novelty), dysregulation (problems with sleeping, problems with eating, negative emotionality, sensory sensitivity), and social-emotional competence domain (positive behaviors that include compliance, attention regulation, imitation and pretend play skills, mastery motivation, empathy, prosocial peer relations). Previous research demonstrates good test–retest reliability, criterion validity, and a well-supported factor structure in the ITSEA (Carter and Briggs-Gowan 2006
; Carter et al. 2003
For each of the four main behavioral domains, subscales are averaged and converted to age- and sex-specific T-scores that have a normalized mean of 50 and a standard deviation of 10. Higher scores in the externalizing, internalizing, and dysregulation domains indicate more behavior problems, whereas higher scores in the social-emotional competence domain indicate better behavior. Normative T-scores are not available for each of the 17 subscales that constitute the ITSEA domain scores; only the raw score percentile ranks are available for each subscale. The ITSEA subscales conventionally use the 10th percentile to indicate clinical concern (using normative data). Compared with the normative population, distributions in our sample were, on average, shifted toward slightly better behavior; thus, we dichotomized each subscale at the 20th percentile to include scores indicating possible problems that may characterize children at risk, relative to the norm in this sample, as well as those with more extreme scores that would typically raise significant concern. Although subscales may be less reliable than domain scores because they are based on fewer items (5–13), and subscales are not age- and sex-standardized for our population, they provide additional detail for specific, more common, behavior clusters.
Additional maternal demographic, psychosocial, and lifestyle covariates were collected through PIN and PIN Postpartum maternal interviews (PIN 2011
). PIN Babies staff completed a modified version of the Home Observation for Measurement of the Environment (HOME) Inventory (Caldwell and Bradley 1984
) during the 36-month interviews to obtain information about parenting styles and influences of the home environment on the child. The modified HOME included only variables that could be accessed through interview. Study staff were not aware of a child’s ITSEA scores or PBDE exposures through breast milk at the time the HOME was administered.
Statistical analysis. We examined the shape of the relationship between the milk concentrations of each of the five PBDE congeners and their sum, using both wet weight and lipid-adjusted PBDE levels, and the continuous domain T-scores using a locally weighted regression smoother (LOESS). Patterns of association were similar using both methods; results using lipid adjusted levels are reported herein. On the basis of the visual inspection of plots of each smoothed association, we created three categories for breast milk PBDE levels: below the 50th percentile; from the 50th to 75th percentiles; and above the 75th percentile. We used separate multivariate regression models to examine the association between each of the five most detected PBDEs and their sum and the four ITSEA domains. We used logistic regression models to assess the relationship between these five PBDEs and their sum and the 17 dichotomous subscale scores.
We adjusted models for child sex and age; household income; maternal age, race, and education; parity; prenatal tobacco use; omega-3 fatty acid levels; and duration of breast-feeding. We carefully examined the appropriate form (continuous or nominal) for each covariate in relation to the outcomes and present the categorization used in . These covariates were chosen based on our a priori
expectation of their association with both PBDE milk concentrations and indicators of social and emotional problems and competence. We also considered including the HOME score in adjusted analyses, but did not because variability in the HOME was minimal within our population and generally explained by differences in income and educational attainment (Caldwell and Bradley 1984
). All analyses were conducted in SAS version 9.2 (SAS Institute Inc., Cary, NC).
Selected sample characteristics.