The relationship of 2-hour (post-75 g oral glucose) serum insulin levels with plasma glucose levels was studied in a population-based random sample comprising 2,627 Hindu Indians, 685 Muslim Indians, 1,351 Creoles (African, European, and Indian admixture), and 415 Chinese from the Indian Ocean island of Mauritius. Known diabetic subjects taking oral hypoglycemic drugs or insulin were excluded from these analyses; 64% of all diabetic subjects had usable glucose and insulin data. Both fasting and 2-hour postload insulin levels were significantly higher in women than in men, and levels in both sexes were significantly greater in Hindu and Muslim Indian subjects than in Creoles or Chinese even after controlling for differences in age, body mass index (BMI), waist to hip ratio (WHR), and plasma glucose level. Levels in Muslims were higher than those in Hindus; it was unclear whether these ethnic differences represented hereditary or unmeasured environmental factors closely associated with ethnicity. All four ethnic groups demonstrated similar inverted U- or V-shaped curves when 2-hour insulin was plotted against either basal or 2-hour glucose. Both quadratic (U) and two-piece (V) regression models improved over linear models for 2-hour insulin versus either fasting or 2-hour glucose in all ethnic groups, although in statistical terms they were good models only for the 2-hour glucose comparison. The two-piece models were associated with modest increases in R2 compared with the quadratic models, but it was not possible to precisely determine optimal turning points with either model. However, in all ethnic groups, 2-hour insulin levels decreased above glucose levels of 7.1 to 7.8 (fasting) and 11.3 to 13.5 mmol/L (2-hour) in quadratic models, and 7.5 to 9.5 (fasting) and 8.5 to 10.5 mmol/L (2-hour) in two-piece models. The shape and point of inflection of the quadratic and two-piece curves were influenced little by gender, obesity, fat distribution, and physical activity. These results are in accord with those observed in cross-sectional and longitudinal studies in other ethnic groups, and support the generality of the plasma glucose levels currently used to define diabetes mellitus, which physiologically correspond with a decrease in beta-cell responsiveness to glucose. Asian Indians appear to have an ethnic propensity to hyperinsulinemia that is not explained by obesity or adverse fat distribution.