Results

Recruitment and participation.
During the study period, 417 pregnancies were identified in the targeted communities. Of them, we excluded 47 pregnant women (11.3%) who had already been enrolled in the study during a previous pregnancy and 3 women (0.7%) due to miscarriage, and we were unable to contact 9 women (2.2%). Of the 358 eligible women asked to participate, 110 (30.7%) refused. This refusal rate is comparable with that of other prospective studies with several interviews in populations of low socioeconomic status. Of the 248 women willing to participate, we were unable to review the medical charts of 43 infants for the following reasons: 10 (4.0%) moved to another village, 14 (5.6%) were adopted in another village, 11 (4.4%) because of miscarriage or perinatal mortality, and 8 (3.2%) because the mother withdrew from the study. Finally, we excluded 6 (2.4%) participants because no biological samples were available for exposure analysis. A total of 199 participants were included in the final analyses.

Population characteristics.
Mothers included in the analysis were mostly from Puvirnituq (45.4%) and Inukjuaq (39.3%). The mean age at delivery was 25.2 years, and most of them smoked during pregnancy (91.4% reported smoking at least 1 cigarette/ day; mean, 10.6 cigarettes/day). Only 2.6% of the infants were not exposed to secondhand smoke during their first year of life. The mean parity was 2.1. There were more males than females (57.6%), and the mean birth weight and length were 3,454 g and 50.3 cm, respectively. Breast-feeding was very common, and only 12.2% were not breast-fed (most of them because they were adopted).

Incidence of infections.
Incidence proportions and rates for selected infections are shown in Table 2. Otitis media was the most frequent infection diagnosed, with a mean of 2.8 episodes per infant-year, followed by URTIs, with 2.4 episodes per infant-year. During the first year of life, almost all infants had at least one episode of otitis (96.0%), and 17.1% had five episodes or more. LRTIs required hospitalization in 31.4% of cases. More than half of the infants (56.8%) were hospitalized at least once during their first year of life.

Contaminant burden in plasma.
Table 1 shows the concentration of contaminants in maternal and infant plasma. The geometric mean concentration of the sum of the 14 PCB congeners (∑PCBs) in maternal plasma was 308 μg/kg (range, 60–1,951 μg/kg). The concentration of the ∑PCBs was highly correlated with that of PCB-153 in maternal plasma (r = 0.99). The correlation between cord plasma and maternal plasma was also very high, both for the ∑PCBs and for PCB-153 (r = 0.95 and 0.94, respectively). The geometric mean concentration for DDE in maternal plasma was 294 μg/kg (range, 54–2,269 μg/kg). The correlation between cord and maternal plasma samples for DDE was also very strong (r = 0.94). Mean concentrations of PCBs and DDE were lower in infant plasma compared to those in maternal plasma.

Prenatal exposure to PCB-153 and infections.
The association between prenatal exposure to PCB-153 and incidence of infections is shown in Table 3. In preliminary analyses we found that the associations between OCs and incidence rates were somewhat stronger during the first 6 months of life. Although this study was designed for a 12-month follow-up, we also present the results for the first 6 months of life. Regarding infections during the first 6 months of life and prenatal exposure to PCBs, we observed statistically significant associations only for LRTIs (3rd quartile; RR = 1.54 and 1.68 for the unadjusted and adjusted models, respectively). Although not statistically significant, almost all other RRs were above the unity. When the four types of infections were combined, the relative rates ranged from 1.19 to 1.20 in the unadjusted model and from 1.17 to 1.27 in the adjusted model. The trend was statistically significant in the adjusted model (p = 0.04).
Compared to the first 6 months of life, the effect size was lower when the first 12 months of life were considered, and only GI infections still pointed toward a positive association. The association was significant for the 3rd quartile in the adjusted model only (RR = 1.59). Globally, rate ratios were similar in the unadjusted and adjusted models.

Prenatal exposure to DDE and infections.
The association between incidence of infections and prenatal exposure to DDE (Table 4) was similar to that observed for exposure to PCB-153. For the first 6 months of life, we detected significant associations with otitis (RR = 1.63, 3rd quartile) and LRTIs (RR = 1.52, 2nd quartile) in the unadjusted model, and with URTIs (RR = 1.56, 2nd quartile) and otitis (RR = 1.83, 3rd quartile) in the adjusted model. The trend was significant for otitis in the unadjusted model (p = 0.04) and borderline significant in adjusted model (p = 0.07). When the four types of infections were combined, we observed significant associations for the 2nd quartile (RR = 1.49) in the unadjusted model, and for the 2nd (RR = 1.38) and 3rd (RR = 1.33) quartiles in the adjusted model. As observed for PCB exposure, almost all RRs were above the unity.
When considering the first 12 months of life, we observed significant associations for GI infections (RR = 1.49, 2nd quartile) in the unadjusted model, and for URTIs (RR = 1.34, 2nd quartile) and GI infections (RR = 1.59, 2nd quartile) in the adjusted model. For all infections combined, the association reached statistical significance only for the 2nd quartile in the unadjusted model (RR = 1.17).

Postnatal exposure to OCs and infections.
We used OC concentrations in infant plasma to evaluate the effect of postnatal exposure on incidence of infections (sampling done at a median age of 7.0 months). We observed no association between postnatal exposure and the incidence of infections (data not shown). The only significant association was for PCBs (12-month follow-up, 2nd quartile, RR = 1.19) in the unadjusted model, but the statistical significance was lost when adjustment for confounding was done.

Effects of exposure to OCs on hospitalization rate.
We found no significant association between prenatal or postnatal exposure and incidence rate of hospitalization for LRTIs (data not shown). However, statistical power was poor because of the limited number of admissions.