Descriptive statistics. A total of 196 participants were included in the final analyses. Reasons for exclusion were technical problems during recording (n = 4), too much noise in the EEG signal to produce a reliable ERP waveform (n = 3), and random responding on the task (n = 3). Compared with the retained participants, these 10 excluded participants had higher cord PCB-153 concentrations [t(199) = 2.57, p = 0.01] and tended to have higher current PCB-153 [t(200) = 1.83, p = 0.07] and Pb [t(201) = 1.84, p = 0.07] concentrations but did not differ on any of the other exposure or control variables (all p-values > 0.10). When the data from the children with too much noise in the EEG signal were included in the behavioral analyses and those with random responding were included, the results remained essentially unchanged. An additional three children were excluded from analyses involving stimulus-locked components because of an insufficient number of trials in their ERP averaged waveform, and 30 were excluded from analyses involving response-locked components for the same reason. Those 33 participants excluded from either ERP analysis had higher PCB-153 [t(189) = 3.16; p = 0.002] and Hg [t(191) = 2.10; p = 0.04] concentrations in cord blood and higher blood Hg concentrations at testing [t(191) = –2.42; p = 0.02].
Descriptive statistics of the final study sample are summarized in . Pb levels at time of testing are similar to those reported in other epidemiological studies of low-level postnatal Pb exposure at school age (e.g., Surkan et al. 2007
), although the levels measured at 11 years of age in this study are about half those seen at 5 years in a subsample of these participants (Boucher et al. 2009b
). Five children had blood Pb concentrations above the 10-μg/dL threshold used by U.S. and Canadian public health agencies for evaluating risk of Pb neurotoxicity and were therefore referred to the appropriate public health authorities. Boys had higher blood Pb concentrations at time of testing than did girls (3.4 vs. 2.0 μg/dL, p
< 0.001), which probably reflects their greater involvement in hunting-related activities. Cord PCB concentrations are about three times higher than in southern Québec (Ayotte et al. 2003
) and similar to those in the Dutch PCB study (Longnecker et al. 2003
). Cord Hg concentrations are about 20 times higher than in southern Québec (Muckle et al. 2001
) and similar to those reported in children from the Faroe Islands (Debes et al. 2006
Descriptive statistics for the study sample (n = 196).
Associations among the concentrations of contaminants measured at birth and at time of testing are presented in . Pb, PCB-153, and Hg are moderately associated with each other, presumably because these substances are found at relatively high concentrations in traditional Inuit food and families vary in the degree to which they consume traditional food.
Intercorrelations among contaminants (Pearson correlation coefficients, r).
Behavioral performance. Values for mean RT and rates of correct responses on the go and no-go trials are presented in . Repeated-measures analysis of variance (ANOVA) showed that, as expected, participants are more accurate on go than on no-go trials [F(1,195) = 655.4, p < 0.001]. Faster mean RT for go trials is associated with more accurate detection of go (r = –0.31, p < 0.001) and no-go (r = –0.19, p < 0.001) trials. Rates of correct responses for go trials are positively associated with rates of correct responses for no-go trials (r = 0.39, p < 0.001).
Results from the regression analyses testing for associations between contaminant concentrations and behavioral go/no-go performance are presented in . After control for confounding factors, higher Pb concentrations in cord blood are associated with fewer correct responses for both go and no-go trials. Higher blood Pb concentrations at time of testing are associated with an increased number of false alarms in no-go trials, suggesting difficulty in inhibiting a prepotent response. This association falls just short of statistical significance when participants with blood Pb concentrations ≥ 5 μg/dL are excluded (n = 175; r = –0.18, p = 0.02; standardized β = –0.14, p = 0.078). Current plasma PCB-153 concentrations are associated with slower responses to go trials. PCB-153 concentrations in cord plasma and Hg concentrations in cord and child blood samples were not associated with any of the behavioral measures after statistical control for confounding variables (all p-values > 0.10). illustrates mean performance per quartile of exposure on each outcome significantly associated with a contaminant. Postnatal Pb and PCB-153 effects seem linear, whereas the adverse effects of prenatal Pb are mainly observed in the highest-exposed children.
Associations between contaminants and behavioral performance in the go/no-go task (n = 196).
Figure 1 Mean ± SD behavioral performance in the go/no-go task according to quartiles of exposure: (A,B) cord blood Pb, (C) 11-year blood Pb, and (D) 11-year plasma PCB-153. Outcome measures are adjusted for the potential confounders listed in (more ...)
Testing for interactions revealed interaction effects for cord Pb with both cord PCB and cord Hg for three behavioral measures: incorrect no-go RT (standardized β: Pb × PCB = 0.14, p = 0.05; Pb × Hg = 0.12, p = 0.10 percent correct go (standardized β: Pb × PCB = –0.17, p = 0.04; Pb × Hg = –0.15, p = 0.07), and percent correct no-go trials (standardized β: Pb × PCB = –0.11, p = 0.12; Pb × Hg = –0.18, p = 0.01). Although not all these interactions reached conventional levels of statistical significance, the stratification analyses showed that the effects of cord Pb were seen primarily in the children with higher prenatal PCB and/or Hg exposures, indicating that the effects of prenatal Pb exposure were intensified by heavier PCB and Hg exposures.
Stimulus-locked ERPs. In repeated-measures ANOVAs with condition (go vs. no-go) as a within-subject variable, P3 amplitude at parietal electrode (Pz) was significantly larger for no-go than for go trials [10.1 vs. 5.0 μV; F(1,192) = 119.7, p < 0.001], reflecting the additional resources needed to inhibit a prepotent response. No condition effect was found for N2 [frontal electrode (Fz)] latency (go: 375.9 msec; no-go: 370.7 msec) and amplitude (go: –7.4 μV; no-go: –7.8 μV; p-values > 0.20). Slower N2 latency was associated with slower mean hit RT (go N2: r = 0.35, p < 0.001; no-go N2: r = 0.30, p < 0.001) and with lower percent correct go trials (go N2: r = –0.23, p = 0.001; no-go N2: r = –0.30, p < 0.001). Greater P3 amplitude was significantly associated with greater percent correct go trials (go P3: r = 0.21, p = 0.003; no-go P3: r = 0.29, p < 0.001), and P3 amplitude to go trials tended to be related to greater percent correct no-go trials (r = 0.14, p = 0.051).
Results from the regression analyses testing for associations between contaminants and the stimulus-locked ERP parameters are presented in . Higher current blood Pb concentrations are associated with reduced P3 amplitude to go and no-go trials. When participants with current blood Pb concentrations ≥ 5 μg/dL are excluded, this effect is still significant for no-go P3 amplitude (r = –0.24, p = 0.001; standardized β = –0.21, p = 0.007) but falls short of significance for go P3 amplitude (r = –0.16, p = 0.04; standardized β = –0.09, p = 0.22). The effect of current blood Pb concentrations on P3 amplitude is clearly observable when contrasting the ERP averages of the bottom and top quartiles of the distributions for current blood Pb concentration (adjusted mean ± SD no-go P3 amplitude, 12.8 ± 7.0 μV vs. 7.8 ± 8.1 μV, Cohen’s d = –0.66; see ). The association between cord PCB-153 and go N2 latency falls short of statistical significance after controlling for maternal age at delivery (p = 0.053), suggesting delayed N2 latency with increasing prenatal exposure. Analyses with Hg revealed an association between current blood Hg concentrations and reduced P3 amplitude in the go condition that fell short of statistical significance after controlling for confounders (p = 0.052). Cord blood Hg and current plasma PCB-153 concentrations levels were not significantly associated with any of the stimulus-locked ERPs.
Associations between contaminants and stimulus-locked ERP components recorded during the go/no-go task (n = 193).
Figure 2 Grand average for stimulus-locked go/no-go ERPs at midline electrodes comparing participants from the first quartile of blood Pb concentrations at time of testing (range, 0.4–1.3 μg/dL; n = 47; thin lines) and those from the fourth quartile (more ...)
Interaction terms revealed significant interactions between current Pb and current Hg on P3 amplitude elicited in the go condition (standardized β = 0.15, p = 0.04) and between current Pb and current PCB-153 on P3 amplitude elicited in both the go (standardized β = 0.17, p = 0.02) and the no-go (standardized β = 0.13, p = 0.08) conditions. In all cases, the Pb effects were stronger in children with lower coexposures to the other chemicals. There were also significant interactions between cord Hg and cord PCB-153 on no-go N2 latency (standardized β = 0.16, p = 0.05) and no-go P3 amplitude (standardized β = –0.20, p = 0.02). Stratification analyses revealed that higher cord Hg resulted in slower no-go N2 latency, and smaller no-go P3 amplitude in children with higher prenatal PCB exposure, but higher no-go N2 amplitude only in children with lower prenatal PCB exposure.
Response-locked ERPs. Repeated-measures ANOVAs with response type (correct go, incorrect no-go) as a within-subject variable revealed significant error-related effects on mean amplitude during the ERN [fronto-central electrode (FCz): 0.3 vs. –4.0 μV; F(1,166) = 81.6, p < 0.001], Pe [Cz: 1.7 vs. 8.2 μV; F(1,166) = 127.1, p < 0.001], and Pc [FCz: 4.4 vs. 6.7 μV; F(1,166) = 13.9, p < 0.001] latency intervals. Smaller ERN and Pe amplitudes tended to be associated with slower mean hit RT (ERN: r = 0.15, p = 0.052; Pe: r = –0.14, p = 0.07), and amplitudes of all three error monitoring components were significantly associated with higher percent correct go trials (ERN: r = –0.19, p = 0.02; Pe: r = 0.25, p = 0.001; Pc: r = 0.18, p = 0.02) and higher percent correct no-go trials (ERN: r = –0.23, p = 0.003; Pe: r = 0.25, p = 0.001; Pc: r = 0.16, p = 0.04).
Results from the regression analyses testing for associations between contaminants and response-locked ERP parameters are presented in . After statistical control for confounders, higher plasma PCB-153 concentrations at time of testing are associated with reduced amplitude of the components Pe, elicited by false alarms, and Pc, elicited by correct hits. These effects are illustrated in , which compares children from the bottom and top quartiles of current plasma PCB-153 concentrations (adjusted mean ± SD Pc amplitude, 5.5 ± 4.0 μV vs. 2.8 ± 4.8 μV, Cohen’s d = –0.61). Pb and Hg concentrations in cord and child blood samples were not associated with any of the response-locked ERP measures in standard regression analyses (all p-values > 0.10).
Associations between contaminants and response-locked ERP amplitudes recorded during the go/no-go task (n = 166).
Figure 3 Grand average for response-locked go/no-go ERPs at midline electrodes comparing participants from the first quartile of plasma PCB-153 concentrations at time of testing (range, 3.5–26.0 μg/g fat; n = 46; thin lines) to those from the fourth (more ...)
The effects of current plasma PCBs on response-locked ERPs did not interact with the other contaminants (all p-values > 0.15).