Bimodality in the frequency distribution of plasma glucose 2 h after a 75-g oral glucose challenge has been demonstrated in only a few population groups, and the generalizability of the phenomenon remains unclear. Therefore, we have studied the distribution of 2-h glucose in 1813 Micronesians, aged 20 years and over, from an urbanized community of the Pacific Ocean Republic of Kiribati. The 2-h plasma glucose distributions were consistent with a mixture model comprising two lognormal components for age groups 30-39, 40-49 and 50 + years. The parameters of the mixture distribution were estimated by the method of maximum likelihood. The prevalence of non-insulin-dependent diabetes mellitus was 7.3% as defined by World Health Organization diagnostic criteria, and 4.7% when defined according to age-specific optimal cut-off values based on the minimization of the total number of misclassified individuals. Assuming that the fitted mixture distribution for each age group was the true plasma glucose distribution, the sensitivity and specificity of the World Health Organization diagnostic criteria and the age-specific optimal cut-off values were determined. Higher sensitivity with relatively little impairment of specificity resulted with the use of the former criteria. The prevalence of non-insulin-dependent diabetes mellitus increased with age as did the mean and standard deviation of the lower lognormal glucose component ("nondiabetics"). However, the mean and standard deviation of the "diabetic" upper component remained relatively constant and varied little with age. Considered together with results from other studies, our findings suggest that bimodality of 2-h plasma glucose distributions is probably present in most populations, but the ability to demonstrate it is a function of the number of subjects for whom data are available, the proportion of subjects in the upper component of the distribution, and the method of assessing its presence.