The goal of this study was to estimate the effects of childbearing on subsequent glucose tolerance and non-insulin-dependent diabetes mellitus (NIDDM) prevalence. A sample of subjects from 64 different locations in the United States were recruited for inclusion in the Second National Health and Nutrition Examination Survey. A complex survey design was used to select a probability sample of subjects from each location. A total of 4577 women were recruited, of whom 3057 underwent clinical and laboratory evaluation for the presence of diabetes mellitus. Women were classified as to their glucose tolerance based on the results of an oral glucose tolerance test or previous physician diagnosis of diabetes mellitus combined with current use of hypoglycemic medication. Childbearing was defined as number of live births experienced by each woman at the time of the interview. Fasting plasma glucose increased linearly with increasing number of live births (coefficient 0.009, 95% confidence interval [CI] 0.006-0.012), as did the 2-h value (coefficient 0.015, 95% CI 0.009-0.021). Adjustment for age, body mass index (BMI), education, and income substantially reduced the magnitude of the association between childbearing and either plasma glucose measurement. When the prevalence of NIDDM in relation to childbearing was examined with logistic regression analysis, a significant linear increase in diabetes prevalence was seen with increasing number of live births (relative prevalence of NIDDM, 1 birth vs. 0 = 1.73, 95% CI 1.39-2.15), but adjustment for age, BMI, education, and income greatly reduced the magnitude of this association (relative prevalence of NIDDM, 1 birth vs. 0 = 1.07, 95% CI 0.98-1.17).(ABSTRACT TRUNCATED AT 250 WORDS)