Secondary Analysis from the Seychelles Child Development Study: The Child Behavior Checklist

Environmental Research Section A 84, 12}19 (2000) doi:10.1006/enrs.2000.4085, available online at http://www.idealibrary.com on

Secondary Analysis from the Seychelles Child Development Study: The Child Behavior Checklist Gary J. Myers,* Philip W. Davidson,- Donna Palumbo,* Conrad Shamlaye,? Christopher Cox,A Elsa Cernichiari,# and Thomas W. Clarkson# Departments of *Neurology, -Pediatrics, ABiostatistics, and #Environmental Medicine, University of Rochester School of Medicine and Dentistry; and ?Ministry of Health, Republic of Seychelles Received December 16, 1999

INTRODUCTION Human exposure to methylmercury (MeHg), a known neurotoxin, is primarily from Ash consumption. As part of a large study examining the association between MeHg exposure and child development in a population with high Ash consumption we examined school-age behavior using the Achenbach Child Behavior Checklist (CBCL). The CBCL Total T score was a primary endpoint and was reported earlier to show no adverse association with prenatal or postnatal MeHg exposure. In this study we analyzed the T scores of the CBCL subscales to determine if more discrete aspects of measured behavior were associated with exposure. The Seychelles Child Development Study (SCDS) is a prospective, double blind, longitudinal evaluation of over 700 children. The index of prenatal exposure was maternal hair total mercury (T-Hg) in a segment growing during gestation. Postnatal exposure was T-Hg in the child’s hair taken at 66 months of age. The child’s primary caregiver completed the CBCL during the 66-month evaluation. No association between prenatal or postnatal MeHg exposure and the CBCL subscales was found. In Seychellois children exposed to MeHg from consumption of ocean Ash we found no association between either prenatal or postnatal MeHg exposure and behavior as measured by the CBCL subscales.  2000 Academic

Methylmercury (MeHg) is an environmental toxin to which the developing brain is particularly sensitive (WHO, 1990). Both prenatal and postnatal exposure could theoretically be associated with clinically detectable effects at low exposure levels. All 7sh contain MeHg and when consumed it is fully absorbed and readily crosses the placenta and the blood brain barrier. Fetal exposure can therefore result from maternal 7sh consumption during pregnancy. The lowest level of prenatal or postnatal MeHg exposure associated with clinically detectable effects is unknown. Some previous studies have reported prenatal effects of exposure at MeHg levels of 6}20 ppm in maternal hair (Marsh et al., 1987; Cox et al., 1989; McKeown-Eyssen et al., 1983; Kjellstrom et al., 1986, 1989; Myers et al., 1995a; Dahl et al., 1996; Grandjean et al., 1997, 1998) while others have found no effects at that exposure level (Marsh et al., 1995a; Myers et al., 1995b, 1997; Davidson et al., 1995, 1998; Axtell et al., 1998). Only a few studies have examined postnatal exposure at this concentration, and only one adverse association has been reported (Davidson et al., 1998; Grandjean et al., 1997, 1998, 1999). In the Faroe Islands the mercury concentration in the children’s hair at 7 years of age was inversely associated with performance on memory for visuospatial information (Grandjean et al., 1999). The primary source of human MeHg exposure is dietary. All 7sh contain some MeHg and levels can be very high if there is local pollution (Swedish Expert Group, 1971; Phelps et al., 1980). Frequent 7sh consumption even without local pollution can increase hair mercury levels above 10 ppm (Turner et al., 1980; Cernichiari et al., 1995a; Clarkson, 1995).

Press

Key Words: methylmercury; prenatal exposure; child development; Child Behavior Checklist; neurodevelopment; Ash; Seychelles. This paper was presented at Mercury as a Global Pollutant: 5th International Conference, Rio de Janeiro, Brazil, May 23}28, 1999.  To whom correspondence should be addressed at Division of Pediatric Neurology, Box 631, University of Rochester School of Medicine and Dentistry, 601 Elmwood Avenue, Rochester, NY 14642. 12 0013-9351/00 $35.00 Copyright  2000 by Academic Press All rights of reproduction in any form reserved.

SEYCHELLES CHILD DEVELOPMENT STUDY: SECONDARY ANALYSIS

Subtle developmental changes at exposure levels below those that impair an organ system or affect overall psychological function have previously been reported following exposure to environmental toxins (Stollery, 1996; Feldman et al., 1980). When subtle deviations from normal are anticipated, functional tests may be more sensitive in detecting either the quantity or the quality of changes. However, the functional domain is very broad, varies with age, and is sensitive to situational variables. Therefore, detecting an effect depends greatly on utilizing a sensitive and speci7c assessment measure. Among the functional effects commonly examined are measures of behavior. Some authors have argued that behavior may be more sensitive and speci7c to neurotoxic exposures than more global functions such as cognition and language (Vorhees and Mollnow, 1987; Spyker, 1975). In children, the behavioral effects of low-level lead exposure have received particular attention (Needleman et al., 1979, 1996; Bellinger et al., 1994; Feldman et al., 1980, Sciarillo et al., 1992; Wasserman et al., 1998; Burns et al., 1999). The Child Behavior Checklist (CBCL) was chosen because associations using the parent or teacher report form to measure behavior have been reported with childhood lead exposure (Sciarillo et al., 1992; Bellinger et al., 1994; Needleman et al., 1996; Burns et al., 1999). Based on the parent and teacher CBCL given to 301 students at ages 7 and 11 years, Needleman and colleagues (1996) reported an increased risk for antisocial and delinquent behavior as lead exposure increased. Other studies in Australia and Baltimore using the parent CBCL and in Boston using the teacher CBCL have reported similar 7ndings (Burns et al., 1999; Sciarillo et al., 1992; Bellinger et al., 1994). The CBCL was part of a more extensive test battery given at 66 months of age to the children in the prospective longitudinal Seychelles Child Development Study (SCDS). The SCDS is examining the association between prenatal and postnatal exposure to MeHg from 7sh consumption and child development. The total CBCL T score was a primary endpoint at 66 months of age, and showed no association with either prenatal or postnatal MeHg exposure (Davidson et al., 1998). However, some authors have suggested that global measures of behavior may not be as sensitive as discrete measures in detecting deleterious effects (Stollery, 1996). Therefore, we examined the association between the subscales of the CBCL and MeHg exposure to see if a more detailed analysis would detect subtle effects.

13

METHOD

Subjects. A cohort of 779 children residing on the island of Mahe in the Republic of Seychelles was enrolled at 6 months of age. The children were born between February 1989 and February 1990 and constituted approximately 50% of all Seychellois births during that period. Thirty-nine children were excluded from the study. For 15 there was inadequate maternal hair to recapitulate mercury exposure and 24 met predetermined exclusion criteria (Marsh et al., 1995b). At 66 months of age 711 children were evaluated. Test instrument. The parent CBLC is a standardized report inventory that has been translated into over 50 languages ranging from Arabic to Zulu and used successfully in research studies throughout the world (Achenbach, 1991; Vignoe and Achenbach, 1998). It is acceptable to parents, easy to administer, brief, reliable, ef7cient to score, and has a high association with other behavioral indices such as the Connors Parent Questionnaire and the Quay}Peterson Revised Behavior Problem Checklist (Achenbach, 1991). The CBCL measures eight behavioral domains as reported by the caregiver and provides an overall behavioral index as well as 10 subscales. The eight domains, each having its own subscale, are Withdrawn, Somatic Complaints, Anxious/Depressed, Social Problems, Thought Problems, Attention Problems, Delinquent Behavior, and Aggressive Behavior (Achenbach, 1991). Two additional T scores are derived from combinations of these subscales. Internalization is the combination of Withdrawn, Somatic Complaints, and Anxious/Depressed and Externalization is Delinquent and Aggressive Behavior combined. Procedure. Children were evaluated at approximately 66 months of age (mean age at evaluation "5.6 years, SD"0.09, minimum"5.4 years, maximum 6.4 years). Evaluations occurred at a Child Development Center staffed by a team of specially trained Seychellois nurses. A comprehensive developmental test battery was used (Davidson et al., 1998). Primary caregivers (de7ned as the family member with whom the child spent at least 5 nights per week) completed the 1991 parent version of the CBCL during the evaluation. In 96% of the cohort the caregiver was the child’s mother. Exposure index. Total mercury was measured in maternal or child hair using cold vapor atomic absorption. Total mercury (T-Hg) in hair has been

14

MYERS ET AL.

shown to correlate well with both MeHg exposure (WHO, 1990) and brain levels of T-Hg (Cernichiari et al., 1995b) and has been the standard measure of exposure in nearly all studies. Prenatal exposure was measured in the longest available segment of maternal hair representing growth during pregnancy, assuming a growth rate of 1.1 cm per month. Postnatal exposure was determined by measuring the centimeter of hair closest to the child’s scalp from a sample taken at the 66-month evaluation. This postnatal age was chosen because the children were weaned and had consumed a 7sh diet for over 2 years, and hair could be taken during the evaluation. For 25 children, the hair taken during the 66-month evaluation was inadequate and hair obtained when the children were between 48 and 56 months of age was used. Statistical analysis. Associations between prenatal and postnatal MeHg exposure and the subscales from the CBCL were examined with adjustment for covariates. Full and reduced regression models were employed for each endpoint. Each model was examined with and without mercury by gender interaction for prenatal and postnatal exposure since differential effects related to gender have been reported (McKeown-Eyssen et al., 1983; Marsh et al., 1987; Myers et al., 1995a; Davidson et al., 1995, 1998). Each analysis was also examined with and without outliers. The analysis plan and covariates used in each model were the same as those used for the 66-month main study and have been described previously (Davidson et al., 1998). Lead, polychlorinated biphenyls (PCBs), and pesticides were not identi7ed as covariates since there is no known local pollution and measurements in Seychelles have been low (Cernichiari et al., 1995a; Davidson et al., 1998; Seychelles Public Health Laboratory Report, 1997). The standardized T scores for the subscales were used because Seychelles is a Westernized culture and there was no a priori reason to believe there were signi7cant differences. In addition these scores were considered the most meaningful in measuring behavior. Poisson (log-linear) regression analysis was used because the T-score distributions were skewed, and in particular showed 8oor effects. Goodness of 7t was measured by the Pearson chi-square statistic, and tests of signi7cance for the model and for individual predictors were based on likelihood ratio chi-square statistics. Residual analysis was based on the deviance residual (McCullagh and Nelder, 1989) using previously established, objective criteria (outliers were de7ned as standardized resid-

uals (!3 or '3 SD from the regression analysis). In case of lack of 7t, the Pearson statistic was used to estimate an extra-Poisson variation parameter. This parameter was used to compute F statistics for tests of individual predictors (McCullagh and Nelder, 1989). In this case the residual analysis was based on visual inspection of plots of the residuals and predicted values. RESULTS

Mercury exposure. The mean maternal hair MeHg level during pregnancy was 6.8 ppm (N" 711; SD"4.5; range, 0.5}26.7). The mean child hair MeHg level at 66 months was 6.5 ppm (N"708; SD"3.3; range, 0.9}25.8). The association between maternal and child hair MeHg concentrations was low (r"0.15; N"708; P(0.001). Test performance. Cognitive abilities in this population appeared normal as measured by the McCarthy Scales of Children’s Abilities General Cognitive Index (mean"94.0, SD 8.3) given when the children were 66 months ($6 months) of age. The mean CBCL total T score in Seychelles was 59.7 (SD 10.0) compared to a U.S. mean of 50.0 (SD 10.0), as reported previously (Davidson et al., 1998). The T scores for the eight subscales were all positively skewed and slightly higher than those seen in U.S. children (range 3.0}9.9 T-score points). In U.S. children the mean subscale scores range from 47 to 54, depending upon the subscale, and scores above 65 are considered clinically signi7cant. All subscales in Seychelles were within one standard deviation of U.S. values (Achenbach, 1991). The subscales for Delinquent and Aggressive Behavior and their combined Externalizing score showed the greatest deviation from U.S. norms. Test score means and standard deviations are shown in Tables 1 and 2 grouped by prenatal and postnatal exposure. The differences between means in the various groups are small and there are no consistent trends with increasing exposure. The Pearson r correlation coef7cient between Internalizing and Externalizing (r"0.64) was similar to that reported by Achenbach (r"0.59) (1991). The correlation between Aggression and Externalizing was high (r"0.91) as expected. The correlation between Attention and both Internalizing and Externalizing were lower (r"0.62). These 7ndings con7rm that use of the CBCL in Seychelles produced 7ndings similar to what would be expected in a normal U.S. population.

15

SEYCHELLES CHILD DEVELOPMENT STUDY: SECONDARY ANALYSIS

TABLE 1 CBCL Subscale Means and Standard Deviations by Prenatal Exposure Level Subgrouping by prenatal MeHg exposure level in ppm

Subscale Total T score Withdrawn Somatic complaints Anxious/Depressed Social problems Thought problems Attention problems Delinquent behavior Aggressive behavior Composite scores Internalizing Externalizing

Range: 43 Mean: 2.0 N: 159 60.4 (9.7) 60.0 (8.7) 57.7 (7.1) 57.0 (7.7) 57.6 (7.7) 57.7 (8.4) 56.8 (6.9) 61.2 (9.2) 60.3 (8.8)

'3}6 4.5 203 59.7 (10.5) 58.1 (8.1) 57.6 (7.3) 55.8 (6.8) 57.5 (8.2) 58.4 (8.2) 57.4 (7.0) 61.1 (8.8) 61.2 (9.0)

'6}9 7.4 156 59.3 (10.2) 58.5 (7.5) 57.6 (6.3) 56.1 (6.8) 57.3 (8.1) 58.4 (8.1) 57.7 (7.3) 59.5 (7.8) 60.0 (8.6)

'9}12 10.3 95 59.3 (9.8) 59.1 (8.8) 57.0 (6.3) 56.8 (7.1) 56.9 (7.5) 58.3 (8.7) 56.7 (7.0) 61.1 (8.5) 60.0 (8.5)

'12 15.3 95 59.7 (8.8) 58.8 (7.5) 56.7 (6.0) 55.6 (6.4) 57.5 (7.6) 57.8 (7.9) 56.7 (6.1) 61.3 (8.1) 60.3 (7.8)

57.7 (10.5) 59.9 (10.2)

55.9 (10.3) 60.5 (10.4)

56.4 (9.6) 59.1 (9.8)

57.1 (9.9) 59.6 (9.8)

56.7 (7.9) 60.1 (9.5)

Regression analyses. For all eight subscales the Pearson chi-square showed an adequate 7t, but for two subscales the model was not signi7cant (Somatic Complaints and Anxious/Depressed). A prenatal MeHg by Gender interaction was present for Delinquent Behavior ( P"0.035) and Externalizing ( P" 0.051). No interaction between postnatal MeHg exposure and gender was present in any model, and removal of outliers did not change the results of any analysis. To further examine the MeHg by Gender interaction with Delinquent Behavior separate slopes were calculated for each gender and are shown in Fig. 1. Despite the signi7cant interaction, the

association between prenatal MeHg and Delinquent Behavior was not signi7cant for either slope ( P"0.15 male and 0.12 female). Although there is evidence that the two slopes are signi7cantly different from each other, neither slope is signi7cant individually. For Externalizing and Internalizing (the two composite scores) both models were signi7cant but the Pearson chi-square did not indicate an adequate 7t. Therefore the extra-Poisson variation adjustment described earlier was required for both endpoints. No gender interaction was present and there was no association with either prenatal or postnatal MeHg exposure.

TABLE 2 CBCL Subscale Means and Standard Deviations by Postnatal Exposure Level Subgrouping by postnatal MeHg exposure level in ppm

Subscale Total T score Withdrawn Somatic complaints Anxious/Depressed Social problems Thought problems Attention problems Delinquent behavior Aggressive behavior Composite Internalizing Externalizing

Range: 43 Mean: 2.2 N: 73 57.6 (9.4) 59.8 (8.2) 57.1 (7.3) 54.7 (6.5) 56.3 (7.3) 55.8 (7.6) 55.4 (6.5) 58.5 (8.1) 58.3 (8.5)

'3}6 4.6 299 60.1 (10.0) 58.7 (8.1) 57.5 (6.8) 56.6 (7.4) 57.6 (8.5) 58.2 (8.3) 57.5 (7.1) 60.7 (8.6) 61.0 (8.8)

'6}9 7.4 213 60.2 (10.7) 59.2 (8.5) 57.6 (6.7) 56.4 (7.0) 57.4 (7.8) 58.9 (8.5) 57.5 (7.0) 61.7 (8.6) 60.9 (8.5)

'9}12 10.2 76 59.3 (9.3) 58.6 (7.4) 57.5 (6.6) 56.0 (6.7) 57.3 (7.5) 58.6 (8.4) 56.6 (7.4) 60.5 (8.6) 60.1 (9.0)

'12 14.9 47 59.5 (7.6) 56.9 (6.9) 56.6 (6.0) 55.4 (5.3) 57.8 (6.3) 57.6 (7.3) 56.5 (5.0) 61.3 (8.2) 59.3 (6.7)

56.0 (9.5) 57.3 (9.8)

56.7 (10.2) 60.4 (9.9)

57.1 (10.0) 60.2 (10.8)

56.9 (8.9) 59.7 (9.5)

55.2 (9.0) 59.8 (7.8)

16

MYERS ET AL.

maternal age increased the CBCL subscales improved and 7ve reached signi7cance. Children with lower HOME scores had higher CBCL subscales, suggesting increased levels of maladaptive behaviors. Gender, socioeconomic status, and caregiver intelligence in8uenced test scores in the expected direction. Higher CBCL scores were associated with males and lower socioeconomic status (SES) while lower scores were associated with higher caregiver intelligence. DISCUSSION

FIG. 1. Individual slopes for males ( P"0.15) and females ( P"0.12) in the analysis of MeHg by Gender relationship for Delinquent Behavior. Only one outlier (large dark circle) was present, in the female analysis, and the regression line for females is plotted with and without its inclusion.

Table 3 shows the parameter estimates for mercury effects from the reduced models without outliers. For Delinquency the model with both gender interactions is reported. The parameter estimates for these slopes, shown in Table 3, indicated higher scores with increasing MeHg in males and lower scores in females. The absolute magnitudes of these changes are nearly identical. Multiple covariates included in the analyses showed signi7cant associations with the subscales and nearly all were in the expected direction. As

No adverse effects on child behavior of either prenatal or postnatal MeHg exposure from consuming ocean 7sh were identi7ed in this analysis. We chose the CBCL because it is a standardized questionnaire that collects data on a variety of problem behaviors, has been widely used in cross-cultural research, and has extensive data supporting its reliability and validity. It has also been used in research on neurodevelopmental syndromes such as Attention De7cit Disorder where there are behavioral features of brain dysfunction, often in the presence of normal cognition. In addition, it is reported to be sensitive and speci7c enough to be of value in evaluating the effects of low-level lead exposure. However, no speci7c behavioral abnormalities have been described following MeHg exposures at these levels and if MeHg does affect behaviors at these exposure levels, the CBCL might not measure the right domains or the adverse effects might have been obscured by covariates that were not measured. These data suggest that the CBCL is a valid measure of behavior in this population. The subscale scores were normally distributed and had the expected associations with a number of covariates (socioeconomic status, caregiver intelligence, and SES) known to be associated with child development. The correlation between the subscales was similar to those reported by Achenbach (1991) on a U.S. population. Although the CBCL has not been used previously in the Seychellois population, it has been used extensively around the world (Vignoe and Achenbach, 1998). The subscale T-scores were consistently higher than those reported for U.S. norms, but all the subscale mean scores were within one standard deviation of those reported by Achenbach (1991) for a normal population. The generally higher subscale T scores may represent a cultural artifact resulting from variations in interpretation by the Seychellois of either the child’s behavior or the questions. Parents appeared to answer the CBCL carefully.

TABLE 3 Mercury and Covariate Effects Present in Multiple Regression Analyses

Parameter estimate Maternal MeHg Child MeHg Gender x Maternal MeHg Separate slopes Gender x Child MeHg Gender (ref male) Birth Weight Child’s Medical History Maternal Age HOME (ref '35) Low (0}31) Medium ('31}35) SES (ref professional) Unskilled (("19) Skilled (20}29) Semi-professional (30}39) Care-Giver IQ (ref '29) Lower Third ('"16) Middle Third (17}28) Hearing (ref '35 dB) 0}25 dB 26}35 dB

Withdrawn

Somatic complaints

Anxious/ Depressed

Social problems

0.0003(0.001) !0.001(0.001) !0.0008(0.001) !0.001(0.001) !0.004(0.002) !0.0003(0.002)!0.0008(0.002) !0.0003(0.002)

Thought problems 0.000(0.001) 0.001(0.002)

Attention problems

Delinquent behavior**

Aggressive behavior

ExternalizingA

InternalizingA

!0.002(0.001) !0.0015(0.0015) 0.0004(0.0012) 0.001(0.002) 0.0013(0.002) !0.0024(0.0016)

!0.045(0.06) !0.064(0.08) * M 0.0024(0.0016) F!0.0025(0.0016)

!0.006(0.01) !0.03(0.01)? !0.02(0.01)* 0.03(0.01)* !0.001(0.01) 1.53(0.52) !0.002(0.01) 0.003(0.01) 0.004(0.01) !0.001(0.01) 0.02(0.01) !0.29(0.55) !0.02(0.03) 0.01(0.03) !0.03(0.03) 0.002(0.03) 0.006(0.03) !1.74(1.47) !0.004(0.001)? !0.002(0.001)* !0.003(0.001) ? !0.001(0.001) !0.003(0.001)? !0.09(0.05) * * 0.03(0.01) !0.02(0.01) 0.02(0.01) 0.03(0.01) 0.03(0.01) 2.23(0.68) 0.02(0.01) 0.007(0.01) 0.01(0.01) 0.03(0.01) 0.04(0.01) 1.41(0.67) * 0.03(0.02) 0.008(0.02) 0.02(0.02) 0.05(0.02) 0.03(0.02) 2.54(0.94) 0.02(0.02) 0.02(0.02) 0.02(0.02) 0.06(0.02) 0.03(0.02) 2.40(0.88) 0.002(0.02) 0.005(0.02) 0.01(0.02) 0.05(0.02) 0.04(0.02) 2.38(0.93) * * 0.03(0.01) 0.001(0.01) 0.001(0.01) 0.04(0.01) 0.0006(0.01) 0.20(0.71) 0.02(0.01) !0.004(0.01) 0.01(0.01) 0.007(0.01) !0.03(0.01) 0.17(0.64) !0.03(0.04) !0.01(0.05)

0.02(0.04) 0.05(0.05)

!0.002(0.04) 0.02(0.05)

0.04(0.04) 0.03(0.05)

!0.02(0.04) 0.004(0.05)

!0.16(1.98) !0.66(2.26)

0.018(0.026) 0.001(0.01) 0.0049(0.013) !0.026(0.011) 0.0092(0.011) 0.01(0.01) 0.0065(0.014) 0.0037(0.011) 0.013(0.03) !0.06(0.03)* !0.044(0.037) !0.033(0.030) !0.0014(0.001) !0.001(0.001) !0.0019(0.0012) !0.0061(0.0010)* ? ? * 0.044(0.014) 0.04(0.01) 0.042(0.018) 0.040(0.014) 0.04(0.013) 0.04(0.01) 0.047(0.017) 0.030(0.014) 0.028(0.019) 0.024(0.018) 0.021(0.019)

0.04(0.02) 0.03(0.02) 0.007(0.02)

0.042(0.024) 0.038(0.023) 0.013(0.024)

0.027(0.020) 0.022(0.018) 0.015(0.019)

0.0046(0.014) 0.0089(0.013)

0.005(0.01) 0.01(0.01)

0.0074(0.018) 0.0076(0.016)

0.023(0.015) 0.022(0.013)

0.011(0.04) !0.003(0.045)

!0.0001(0.04) !0.017(0.051) 0.02(0.05) !0.013(0.058)

0.014(0.041) 0.037(0.047)

Note. Parameter estimates and standard errors for reduced models without interaction with outliers removed. All reduced models were signi7cant ( P("0.05 using two-sided test). ref, reference value. ?"P(0.01; *"P'0.0140.05. **"Reduced model without outliers with separate slopes for gender by MeHg interaction. A Log linear models with scale adjustment using the Pearson statistic to estimate an extra-Poisson variation parameter and with outliers included.

SEYCHELLES CHILD DEVELOPMENT STUDY: SECONDARY ANALYSIS

Regression coef7cients (slope"test point/ppm MeHg) for CBCL endpoints

17

18

MYERS ET AL.

Anecdotally, our observations are that Seychellois parents hold their children to a high standard of behavior. Nearly all of the children live in extended families with parents and relatives all participating in shaping the child’s behavior. The children are generally well behaved, quiet, and respectful and tend to be compliant with parents and other authority 7gures. In over 3000 evaluations spanning 5.5 years the children have generally been quiet, cooperative, and respectful. The 7nding of a gender interaction with Delinquent Behavior was intriguing. Differential gender effects of MeHg have been previously reported with males showing greater sensitivity to adverse effects (Marsh et al., 1987; McKeown-Eyssen et al., 1983). Gender interactions were also found previously in the SCDS. At the 66-month evaluations from the main study there was a signi7cant gender interaction with postnatal MeHg exposure present on the Bender Gestalt Visual Motor Test. Males had a decrease in errors or improved scores that reached signi7cance while females had a slight increase in errors that was not signi7cant (Davidson et al., 1995, 1998). With the Delinquency subscale the gender interaction indicated a signi7cant difference between males and females, but when separate slopes were estimated neither indicated a signi7cant difference from zero with respect to the exposure index. The slopes of the associations indicated that males had an increase in delinquent behaviors with increasing prenatal exposure while females had fewer delinquent behaviors. This 7nding could be interpreted as both a negative effect of exposure on males and a positive effect on females and be an early indication of signi7cant behavioral effects. It could also be a spurious 7nding. The association in females of better behavior with increasing exposure is of particular interest. In the main cohort SCDS 66-month evaluations MeHg exposure in these low ranges was associated with improved performance on several developmental endpoints (Davidson et al., 1998). The authors postulated that the MeHg concentration was associated with 7sh consumption and that the nutritional bene7ts of 7sh may have accounted for the improvement. The SCDS also found during the 29-month evaluations that activity level as measured on the Infant Behavior Record of the Bayley Scales of Infant Development decreased in males as prenatal Me Hg exposure increased (Davidson et al., 1995). However, it was unclear if this was a bene7cial or adverse effect. The complete parent CBCL has not previously been used to study MeHg exposure. However, 16 of

the 113 questions from the test were incorporated into the maternal questionnaire administered in the Faroe Islands study (Grandjean et al., 1997). That study also used The Nonverbal Analogue Pro7le of Mood States, but no association was reported with either prenatal or postnatal exposure. The Faroe Islands study did report adverse associations between MeHg exposure and a number of other endpoints. The Seychelles and Faroe Islands studies have similar goals, to study exposure to dietary MeHg, but differ in many ways. The source and measure of exposure, secondary exposures, test batteries, and analysis methods all differed and one or more of these may account for the differing 7ndings. In summary, no adverse associations between the behaviors measured by the CBCL in Seychelles and either prenatal or postnatal MeHg exposure at the levels achieved by consuming a high 7sh diet were found in this study. These data are consistent with our earlier reports. Through 66 months of age in the main cohort we have not identi7ed adverse associations between either prenatal or postnatal MeHg exposure and primary or secondary endpoints (Myers et al., 1995b, 1997; Davidson et al., 1995, 1998; Axtell et al., 1998). REFERENCES Achenbach, T. M. (1991). ‘‘Manual for the Child Behavior Checklist/4-18 and 1991 Pro7le.’’ University of Vermont Department of Psychiatry, Burlington, VT. Axtell, C. D., Myers, G. J., Davidson, P. W., Choi, A. L., Cernichiari, E., Sloane-Reeves, J., Shamlaye, C., Cox, C., and Clarkson, T. W. (1998). Semiparametric modeling of age at achieving developmental milestones after prenatal exposure to methylmercury in the Seychelles child development study. Environ. Health Perspect. 106(9), 559}564. Bellinger, D., Hu, H., Titlebaum, L., and Needleman, H. L. (1994). Attentional correlates of dentin and bone lead levels in adolescents. Arch. Environ. Health 49, 98}105. Burns, J. M., Baghurst, P. A., Sawyer, M. G., McMichael, A. J., and Tong, S. (1999). Lifetime low-level exposure to environmental lead and children’s emotional and behavioral development at ages 11}13 years. Am. J. Epidemiol. 149, 740}749. Cernichiari, E., Toribara, T. Y., Liang, L., Marsh, D. O., Berlin, M. W., Myers, G. J., Cox, C., Shamlaye, C. F., Choisy, O., Davidson, P., and Clarkson, T. W. (1995a). The biological monitoring of mercury in the Seychelles study. Neurotoxicology 16, 613}628. Cernichiari, E., Brewer, R., Myers, G. J., Marsh, D. O., Lapham, L. W., Cox, C., Shamlaye, C. F., Berlin, M., Davidson, P. W., and Clarkson, T. W. (1995b). Monitoring methylmercury during pregnancy: Maternal hair predicts fetal brain exposure. Neurotoxicology 16, 705}710.

SEYCHELLES CHILD DEVELOPMENT STUDY: SECONDARY ANALYSIS Clarkson, T. W. (1995). Environmental contaminants in the food chain. Am. J. Clin. Nutr. 1(Suppl.), 682S}686S. Cox, C., Clarkson, T. W., Marsh, D. O., Amin-Zaki, L., Tikriti, S., and Myers, G. J. (1989). Dose}response analysis of infants prenatally exposed to methylmercury. An application of a single compartment model to single-strand hair analysis. Environ. Res. 31, 640}649. Dahl, R., White, R. F., Weihe, P., Sorensen, N., Letz, R., Hudnell, H. K., Otto, D. A., and Grandjean, P. (1996). Feasibility and validity of three computer-assisted neurobehavioral tests in 7-year-old children. Neurotoxicol. Teratol. 18, 413}419. Davidson, P. W., Myers, G. J., Cox, C., Axtell, C., Shamlaye, C. F., Sloane-Reeves, J., Cernichiari, E., Needham, L., Choi, A., Wang, Y., Berlin, M., and Clarkson, T. W. (1998). Effects of prenatal and postnatal methylmercury exposure from 7sh consumption at 66 months of age: The Seychelles Child Development Study. JAMA 280(8), 701}707. Davidson, P. W., Myers, G. J., Cox, C., Shamlaye, C. F., Marsh, D. O., Tanner, M. A., Berlin, M., Sloane-Reeves, J., Cernichiari, E., Choisy, O., Choi, A., and Clarkson, T. W. (1995). Longitudinal neurodevelopmental study of Seychellois children following in utero exposure to methylmercury from maternal 7sh ingestion: Outcomes at 19 and 29 months. Neurotoxicology 16, 677}799. Feldman, R. G., Ricks, N. L., and Baker, E. L. (1980). Neuropsychological effects of industrial toxins: A review. Am. J. Ind. Med. 1, 211}227. Grandjean, P., Budtz-Jorgensen, E., White, R. F., Jorgensen, P. J., Weihe, P., Debes, F., and Keiding, N. (1999). Methylmercury exposure biomarkers as indicators of neurotoxicity in children aged 7 years. Am. J. Epidemiol. 150, 301}305. Grandjean, P., Weihe, P., White, R. F., and Debes, F. (1998). Cognitive performance of children prenatally exposed to ‘‘safe’’ levels of methylmercury. Environ. Res. 77, 165}172. Grandjean, P., Weihe, P., White, R. F., Debes, F., Araki, S., Yokoyama, K., Murata, K., Sorensen, N., Dahl, R., and Jorgensen, P. J. (1997). Cognitive de7cit in 7-year-old children with prenatal exposure to methylmercury. Neurotoxicol. Teratol. 19, 417}428. Kjellstrom, T., Kennedy, P., Wallis, S., and Mantell, C. (1986). ‘‘Physical and Mental Development of Children with Prenatal Exposure to Mercury from Fish. Stage 1. Preliminary Tests at Age 4.’’ National Swedish Environmental Board Report No. 3080, Solna, Sweden. Kjellstrom, T., Kennedy, P., Wallis, S., Stewart, A., Friberg, L., Lind, B., Witherspoon, P., and Mantell, C. (1989). ‘‘Physical and Mental Development of Children with Prenatal Exposure to Mercury from Fish. Stage 2. Interviews and Psychological Tests at Age 6.’’ National Swedish Environmental board Report 3642, Solna, Sweden. Marsh, D. O., Clarkson, T. W., Cox, C., Myers, G. J., Amin-Zaki, L., and Al-Tikriti, S. (1987). Fetal methylmercury poisoning. Arch. Neurol. 44, 1017}1022. Marsh, D. O., Turner, M. D., Smith, J. C., Allen, P., and Richdale, N. (1995a). Fetal methylmercury study in a Peruvian 7sheating population. Neurotoxicology 16, 717}726. Marsh, D. O., Clarkson, T. W., Myers, G. J., Davidson, P. W., Cox, C., Cernichiari, E., Tanner, M. A., Lednar, W., Shamlaye, C. F., Choisy, O., Hoareau, C., and Berlin, M. (1995b). The Seychelles study of fetal methylmercury exposure and child development: Introduction. Neurotoxicology 16, 577}582.

19

McCullagh, P., and Nelder, J. A. (1989). ‘‘Generalized Linear Models,’’ 2nd ed. Chapman and Hall, London. McKeown-Eyssen, G. E., Ruedy, J., and Neims, A. (1983). Methyl mercury exposure in northern Quebec II. Neurologic 7ndings in children. Am. J. Epidemiol. 118, 470}479. Myers, G. J., Marsh, D. O., Cox, C., Davidson, P. W., Shamlaye, C. F, Tanner, M. A., Choi, A., Cernichiari, E., Choisy, O., and Clarkson, T. W. (1995a). A pilot neurodevelopmental study of Seychellois children following in utero exposure to methylmercury from a maternal 7sh diet. Neurotoxicology 16, 629}638. Myers, G. J., Marsh, D. O., Davidson, P. W., Cox, C., Shamlaye, C. F., Tanner, M. A., Choi, A., Cernichiari, E., Choisy, O., and Clarkson, T. W. (1995b). Main neurodevelopmental study of Seychellois children following in utero exposure to methylmercury from a maternal 7sh diet: Outcome at six months. Neurotoxicology 16, 653}664. Myers, G. J., Davidson, P. W., Shamlaye, C. F., Axtell, C. D., Cernichiari, E., Choisy, O., Choi, A., Cox, C., and Clarkson, T. W. (1997). Effects of prenatal methylmercury exposure from a high 7sh diet on developmental milestones in the Seychelles Child Development Study. Neurotoxicology 18, 819}830. Needleman, H. L., Gunnoe, C., Leviton, A., Reed, R., Peresie, H., Maher, C., and Barrett, P. (1979). De7cits in psychologic and classroom performance of children with elevated dentine lead levels. N. Engl. J. Med. 300, 689}695. Needleman, H. L., Riess, J. A., Tobin, M. J., Biesecker, G. E., and Greenhouse, J. B. (1996). Bone lead levels and delinquent behavior. JAMA 275, 363}369. Phelps, R. W., Clarkson, T. W., Kershaw, T. G., and Wheatley, B. (1980). Interrelationships of blood and hair mercury concentrations in a North American population exposed to methylmercury. Arch. Environ. Health 35(3), 161}168. Sciarillo, W. G., Alexander, G., and Farrell, K. P. (1992). Lead exposure and child behavior. Am. J. Public Health 82, 1356}1360. ‘‘Seychelles Public Health Laboratory Report’’ (1997). Spyker, J. M. (1975). Behavioral teratology and toxicology. In ‘‘Behavioral Toxicology’’ (B. Weiss and V. G. Laties, Eds.), pp. 311}349. Plenum Press, New York. Stollery, B. T. (1996). Cognitive neurotoxicology: A luxury or necessity. Neurotoxicol. Teratol. 18, 359}364. Swedish Expert Group. (1971). Methylmercury in 7sh. Nord. Hyg. Tid. Suppl. 4. Turner, M. D., Marsh, D. O., Smith, J. C., Clarkson, T. W., Inglis, J. B., Rubino, E. C., Chiriboga, J., and Chiriboga, C. C. (1980). Methylmercury in populations eating large quantities of marine 7sh. Arch. Environ. Health 35, 367}378. Vignoe, D., and Achenbach, T. M. (1998). ‘‘Bibliography of Published Studies Using the Child Behavior Checklist & Related Materials.’’ Univ. of Vermont Department of Psychiatry. Vorhees, C. V., and Mollnow, E. (1987). Behavioral teratogenesis: Long-term in8uences on behavior from early exposure to environmental agents. In ‘‘Handbook of Infant Development.’’ (J. D. Osofsky, Ed.), pp. 913}971. Wiley, New York. Wasserman, G. A., Staghezza-Jaramillo, B., Shrout, P., Popovac, D., and Graziano, J. (1998). The effect of lead exposure on behavior problems in preschool children. Am. J. Public Health 88, 481}486. WHO. (1990). ‘‘Environmental Health Criteria 101, Methylmercury.’’ World Health Organization, Geneva, Switzerland.