The three study groups and each neurologic and neurobehavioral outcome were compared by demographic (including ethnicity, place of birth, child's age, child's gender, maternal and paternal education, and household income), lifestyle (parental smoking and alcohol use) and medical factors (including congenital abnormalities and type of insurance) to assess potential confounding, using chi-square tests for categorical variables and ANDV A tests for continuous variables. Initial unadjusted logistic regression models were performed to assess differences in categorical outcomes (e.g., out of the normal range. i.e., <10th or >90th percentile for weight, height. weight for height and head circumference), and linear regression models were used to assess differences in continuous outcomes (i.e., neurobehavioral test scores, height, weight and head circumference) among the three study groups. Certain neurologic outcomes (tests of muscle strength and visual acuity) were integer scores for which cut-offs for non-normal outcomes were determined and analysis was conducted using logistic regression to compare the proportion affected with such outcomes by group. Final multiple logistic and linear regression models were constructed by including potential confounders found to be related to study group and outcome at the p < 0.20 level or as suggested by the relevant literature. In addition, selected multivariate analyses were performed, stratifying on age of the child when warranted by large differences in the outcome or when the accuracy of the assessment of outcomes was affected by the child's age. Due to skewing of results, neurobehavioral scores were excluded from our analyses when the child did not understand the instructions or the child made an attempt at completing the test but did not perform any item on a given test correctly. This type of result was more frequent among younger children and children in the group in which neither parent worked in agriculture (group III) (see Table I). Since more than 40% of children overall had such a result for the story recall (immediate and delayed) component, with the exception of the delayed recognition exam, of the neurobehavioral test battery, we did not analyze further the results for this set of tests.