Women were primarily young and parous, and relatively few had completed high school (Table 1). Most mothers were born in Mexico and spoke Spanish (89%), and 23% had been in the United States ≤ 1 year at enrollment. Nearly all (92%) lived with the child’s father, and two-thirds of families lived below the federal poverty threshold (U.S. Census Bureau 2000). Most women (82%) lived with one or more farmworkers and 43% worked in agriculture themselves during pregnancy. Almost no mothers consumed alcohol regularly during pregnancy, and few smoked or lived with a smoker. Half had symptoms of depression 1 year postpartum. The average maternal PPVT score was in the low normal range, averaging 86 ± 21.

One-third of children were first-born (Table 1). Almost all were initially breast-fed (96.6%), half for ≥ 6 months, and 29% still breast-fed at 12 months of age. Few children lived in home environments considered to be of low quality in terms of stimulation and interaction (HOME scores < 26) at 6 (8%) and 12 (< 1%) months of age.

The geometric mean (GM) of pregnancy DAPs was 114.9 nmol/L (Table 2). Child levels increased with age, with total DAPs averaging 45.5, 59.5, and 70.9 nmol/L at 6, 12, and 24 months of age, respectively. Maternal DAPs were uncorrelated with child DAPs (r = 0.04 with each age, p = 0.40 to 0.47); The 12- and 24-month DAPs were correlated (r = 0.20, p < 0.001) but neither was correlated with 6-month DAPs. DMs were similarly correlated, but there was no correlation of DE measurements across ages. However, at any given age, the DEs and DMs were significantly correlated (r = 0.28–0.50, p < 0.001).

As shown in Table 2, < 40% of MDA levels were above the LOD; the median detectable level was around 1 μg/L. Most (91%) mothers had detectable serum TCPy, with a median level around 4 μg/L.

At 6, 12, and 24 months of age, respectively, Bayley PDI scores (mean ± SD) were 96.4 ± 10.6, 106.0 ± 12.6, and 97.5 ± 10.6, and MDI scores were 95.7 ± 7.0, 100.6 ± 8.9, and 85.9 ± 11.8 (Table 3). The proportion of MDI scores < 85 increased dramatically at 24 months with 50% < 85 relative to 3–4% on earlier assessments. On the CBCL, more children scored in the clinical range on the DSM-oriented pervasive developmental disorder scale (14.4%) than the national reference sample (≤ 3%) (binomial test p < 0.0001), although similar to the expected proportion of children scored in the clinical range on the attention problems syndrome (2.0%) and ADHD (3.3%) scales (Table 3).

Table 4 presents the change [and 95% confidence intervals (CIs)] in PDI and MDI scores associated with a 10-fold increase in metabolites, controlling for covariates. Results were similar in unadjusted models. We did not observe any statistically significant associations between metabolite levels and PDI at any age. We observed an overall pattern of negative associations between pregnancy metabolites and MDI, and positive associations between concurrent child metabolites and MDI, which were not statistically significant until the 24-month assessment. Specifically, for every 10-fold increase in pregnancy DAPs, we found a 3.5 point decrease in the 24-month-olds’ MDI (95% CI, −6.6 to −0.5; p = 0.02); however, for every 10-fold increase in the DAPs at 24 months, we observed a 2.4 point increase in MDI (95% CI, 0.5 to 4.2; p = 0.01). We attribute the associations with total DAPs largely to the DM metabolites, which indicated a similar pattern of results.

Table 5 summarizes multiple regression results for CBCL outcomes. Because of small numbers (Table 3), we used less conservative “borderline” cut points for attention-related outcomes. We did not observe any significant associations for these two scales with any DAPs measurement. The only covariate we found to be significantly related to risk of attention problems and ADHD was maternal depression: Mothers who reported depressive symptomatology 12 months postpartum had nearly 3-fold odds for reporting that their 2-year-old had attention problems and ADHD [odds ratio (OR) = 3.1; 95% CI, 1.3 to 7.2; p = 0.01; and OR = 2.7; 95% CI, 1.2 to 5.9; p = 0.01, respectively].

We did, however, observe that children with higher prenatal and postnatal total DAPs were at significantly higher risk of PDD, with an approximately 2-fold increase in risk for each 10-fold increase in metabolites (prenatal DAPs OR = 2.3; 95% CI, 1.0 to 5.2, p = 0.05; 24-month DAPs OR = 1.7; 95% CI, 1.0 to 2.9, p = 0.04) We observed similar statistically significant associations for prenatal DM and 24-month DE exposures (Table 5). The only other covariate associated with report of PDD was maternal PPVT, which was negatively associated with risk of PDD (for each point increase in PPVT, OR = 0.98; 95% CI, 0.96 to 0.99; p = 0.001).

Controlling for birth weight, gestational age, abnormal neonatal reflexes, or additional chemical exposures did not substantially change our results. Results did not differ by sex of the child for any of the analyses. In addition, although we previously observed an association with maternal DDT and 24-month MDI (Eskenazi et al. 2006), we did not observe an interaction between DDT and prenatal DAPs. Overall, adjusting metabolites by creatinine yielded results similar to primary analyses, with some coefficients becoming slightly stronger and some weaker. The observed negative association of prenatal metabolites and 24-month MDI diminished somewhat (total DAPS: β = −3.0; 95% CI, −6.1 to 0.1; p = 0.06; DMs: β = −3.1; 95% CI, −5.9 to −0.4; p = 0.03). Creatinine-adjustment of child metabolites also reduced the positive associations of total DAPS (β = 1.7; 95% CI, −0.3 to 3.7; p = 0.09) and DMs (β = 1.4; 95% CI, −0.5 to 3.3; p = 0.14) on MDI at 24 months. CBCL results changed little with creatinine adjustment.

No significant associations were observed between MDA or TCPy and any Bayley or CBCL outcomes (Table 6).

We have previously reported in this cohort a relationship between DAP metabolites during pregnancy and both shortened gestation and poorer neonatal reflexes. These associations were seen primarily with dimethyl phosphate metabolites. Results from the present investigation suggest that DAP metabolite levels during pregnancy, particularly from dimethyl phosphate pesticides, may be negatively associated at 24 months with mental development (MDI) on the Bayley Scales and an increase in risk of maternally reported PDD. We also observed a curious positive association of postnatal DAPs with MDI; however, these postnatal metabo-lites were associated with an increased risk of PDD. Results persisted after controlling for other chemical exposures and were independent of previously observed associations of shortened gestation and poorer reflexes.

Although we observed associations between neurodevelopment and DAPs, we saw no such associations with metabolites specific to malathion (MDA) or chlorpyrifos (TCPy). Malathion is one of several pesticides devolving to DMs, and chlorpyrifos is one of several pesticides devolving to DEs; MDA and TCPy are only moderately correlated with DMs and DEs, respectively (MDA and DMs: r = 0.2 to 0.3, p < 0.001; DE and TCPy: r = 0.1 to 0.2, p < 0.04). Thus, the observed associations with DAPs may be attributed to compounds other than just these two. For example, oxydemeton-methyl, for which we cannot measure a specific metabolite, is a widely used DM in the Salinas Valley that is several times more neurotoxic than malathion (Castorina et al. 2003).

Few previous studies have examined low-level exposure to OP pesticides and children’s neurodevelopment using exposure biomarkers. Direct comparison with existing studies is complicated by differences in the exposure scenario or measurements and neurodevelopmental assessment methods. For example, a recently published cross-sectional study from Ecuador lacked biomarker-based pregnancy exposures, but DAP metabolites in 6- to 9-year-olds were associated with increased reaction times (Grandjean et al. 2006). In comparison, our study was larger and longitudinal, allowing for two pregnancy and three early childhood DAP measurements, but we assessed different outcomes and in younger children. Grandjean et al. (2006) reported that their DAP levels were similar to levels for 6- to 11-year-old U.S. children participating in NHANES (CDC 2006). DAPs in our children were lower than those in NHANES, but our children were considerably younger than those sampled in the national study. We observed a rise in DAPs from 6 months to 2 years of age, which is likely to continue as the children interact more with their environment. As in our study, Rauh et al. (2006) used the Bayley Scales and CBCL to assess neurodevelopment in a cohort of children up to 3 years of age, but measured chlor-pyrifos, a DE pesticide, in pregnancy bloods. Rauh et al. found inverse associations of chlor-pyrifos with mental development and pervasive developmental disorder, results that we found with total DAP and DM, but not DE or TCPy, metabolites. Unlike that study, we did not observe associations with PDI or attentional deficits. However, the exposure scenario in an agricultural community is likely quite different from that in an urban environment. Additionally, Rauh et al. (2006) found no effect at 2 years of age, only at 3 years.

We observed that DAPs in children were positively associated with neurodevelopment. We have no ready explanation for this finding. One possible explanation is that children with higher cognitive functioning may be more interactive with their environment, leading to higher exposure to pesticide residues. The pre-natal measurements do not suffer from the same issues of unknown temporal order in that they clearly preceded the measures of child development. Another possibility is that children who eat more fruits and vegetables may have higher DAP levels, but because of better diets have higher functioning.

DAP metabolites are limited as biomarkers of exposure. Recent studies suggest that urinary metabolite levels may reflect exposure not only to OP parent compounds, but also to the potentially less toxic ambient metabolites (Lu et al. 2005). Also, because exposure to OPs varies considerably from day to day, DAPs from spot urine samples may not represent average exposure over time. Although we observed significant correlations between some DAP measures over time, these correlations were weak. Preliminary results from our Center indicate that children’s DAP measurements more than a few days apart are uncorrelated, suggesting considerable intraperson variability (A. Bradman, personal communication). Any measurement of OPs in urine or blood reflect exposure during the brief (usually < 48 hr) antecedent period and therefore may not accurately reflect exposure throughout the entire critical period of neurodevelopment. While adjustment for creatinine muted the positive associations we observed with child metabo-lites and 24-month MDI, such an adjustment warrants caution. Although widely accepted in occupational studies of nonpregnant adults, creatinine adjustment may not be appropriate for metabolite levels in populations undergoing rapid physiologic changes, such as pregnant women and young children, due to high intraindividual variability in creatinine excretion (Boeniger et al. 1993).

Another limitation of this study is that we conducted many analyses and did not adjust for multiple comparisons. However, because of the interrelationships among the exposures and among the outcomes, the comparisons we made were not entirely independent. Specifically, the urinary DM and DE metabo-lites are subsets of the total DAPs and are therefore highly correlated with total DAPs at each time (DAPs and DMs: r = 0.9 to 0.96. p < 0.001; DAPs and DEs: r = 0.6 to 0.7. p < 0.001). In addition, MDI and PDI were moderately correlated at each of the three time points (r = 0.3–0.5; p < 0.001) and, although not as strongly, across ages (r = 0.1–0.2; p < 0.01), and the three CBCL outcomes we considered were associated with one another (chi-square p < 0.001 for all combinations). Thus, a Bonferroni correction might be considered unduly conservative (Bland and Altman 1995).

This study also has a number of strengths. Because children, especially those from economically impoverished environments, may be exposed to other environmental agents, we collected extensive information about additional environmental exposures and measured many with biomarkers, including lead, OC pesticides, and PCBs. We previously found that the CHAMACOS mothers had high serum levels of certain OCs, such as DDT and DDE, because many women had originated from Mexico where these pesticides were used more recently (Bradman et al. 2006). We have determined in our population that both the previously reported association of maternal DDT with MDI (Eskenazi et al. 2006) and the present association of maternal DAPs with MDI persisted when these exposures were modeled simultaneously.

Because our population was also relatively homogeneous, we could eliminate the potential for other covariates to explain the observed associations. Yet because the CHAMACOS population consisted mostly of children from low-income families, they were already at risk for poorer neurodevelopment (Black et al. 2000). Although we observed < 5% with MDI scores < 85 at 6 and 12 months of age, at 2 years of age almost 50% showed possible deficits. Although this drop in scores at 24 months might be attributed to a lack of developmentally stimulating environments, it is also possible that the Bayley Scales, although having internal validity in our population, may not be a clinically valid tool in Spanish or in Latino immigrant communities. Similar findings of drops in MDI scores at 24 months of age have been seen in a cross-sectional study of children in semiurban Mexico (Fernald et al. 2006) and in African-American and Dominican children in New York City (Rauh et al. 2006). The proportion of children with delays may decrease as they are exposed to more enriched environments such as preschool.

We also observed a very high proportion of CHAMACOS children with PDD. Although DSM-oriented scales based on maternal report are not directly equivalent to a DSM diagnosis (Achenbach and Rescorla 2000), this high report of PDD merits further investigation and follow-up, especially in light of a recent study that hypothesized that OP exposure during critical prenatal periods of neuronal migration may precipitate autism in a genetically vulnerable population (D’Amelio et al. 2005). In a case–control study of autism, D’Amelio et al. (2005) reported a significant association among Caucasian Americans between autism and certain variants of the paraoxonase gene (PON1) that encodes the enzyme responsible for OP detoxification. Thus, in future investigations, we will examine the interaction between OP exposure and PON1 gene variants and enzyme expression on the observed associations with PDD (Furlong et al. 2006; Holland et al. 2006). Follow-up studies of this cohort as they enter school will determine the clinical significance and persistence of the observed deficits and may shed light on the positive relationships with postnatal exposure measurements.

In summary, we report an adverse association of prenatal organophosphate pesticide exposure as measured by DAPs with mental development and pervasive developmental problems at 24 months of age. This study is one of the first to examine the associations of both prenatal and postnatal organophosphate exposure on early neurodevelopment. The negative associations of pregnancy DAP measures with mental development should be interpreted with caution given the positive postna-tal relationships we observed and should be replicated in other populations.