Exposure information for the study was obtained from each subject's birth record. Preliminary evaluation of risk estimates was conducted by stratified analyses. Subsequently, we used multivariable logistic regression to estimate the odds ratios (ORs) and 95% confidence intervals (CIs) for the associations of factors related to the hygiene hypothesis, including mothers' age (<18, 18-24, 25-34, and ≥35 years), race (white, black, Asian, Hispanic, and “other nonwhite”), education (<12, 12, and ≥13 years), marital status, cesarean section delivery, prenatal smoking, number of prior births and pregnancies (0, 1, 2, and ≥3), number of older siblings (estimated by number of prior births now living), use of prenatal care based on the Kotelchuck index of adequacy of prenatal care (inadequate, intermediate, adequate, and adequate plus),27 and trimester prenatal care began (first, second, and third or none). Additional information about the subject's insurance status (Medicaid or charity insurance at the birth hospitalization) was obtained from the linked Comprehensive Hospital Abstract Reporting System record for the mother's delivery hospitalization (this linkage has been routinely performed annually in Washington since 1987). Maternal factors potentially related to the overload hypothesis included age; DM status (established or gestational); body mass index (BMI) calculated from prepregnancy weight and height (calculated as weight in kilograms divided by height in meters squared) and based on World Health Organization classification (<18.5, undernourished; 18.5-24.99, normal; 25-29.99, overweight; and ≥30, obese)28; and prepregnancy weight (<100, 100-149, 150-199, and ≥200 lb); cesarean section; and prenatal smoking. Potentially related infant factors included gestational length (<37, 37-42, and >42 weeks), birth weight (<2500, 2500-3999, and ≥4000 g), and size for gestational age (small for gestational age, appropriate for gestational age, and large for gestational age, with upper and lower 10th percentiles calculated29 using Washington State data 1989-2002 as a standard). Factors evaluated for their potential effects on the relationships of interest included maternal age, race, educational level, marital status, medical insurance at the birth hospitalization, BMI, prenatal smoking, number of prior live births, and infant sex and birth weight. Subanalyses were also conducted to determine if results varied by birth year categories; they did not. Levels of missing data were generally similar for cases and controls; missing data for all infant variables, maternal age, race, and marital status were less than 5%. For 3 variables (maternal education, prepregnancy weight, and BMI), information was available only for birth records from 1992 or later. The greatest level of missing data was for BMI (33% of cases, 38% of controls). Among subjects with missing data for BMI, prepregnancy weight, and maternal education, maternal and infant characteristics were distributed similarly to these distributions in the overall study population. Analyses were restricted to subjects with known relevant information for each risk estimate.