Distribution of In Vivo Insulin Action in Pima Indians as Mixture of Three Normal Distributions

If a single gene produced insulin resistance, with environmental effects creating some additional variance, insulin action might be distributed as a mixture of two normal distributions if the gene is dominant or recessive or as a mixture of three normal distributions if the gene is codominant. To estimate maximal insulin-stimulated glucose uptake rates (MaxMs), hyperinsulinemic-euglycemic clamps were performed on 245 nondiabetic Pima Indians (126 men, 119 women). Five models (for 1, 2, 3, 4, or 5 components each, normally distributed with a common variance) were fitted to the frequency distribution of MaxM by iterative maximum-likelihood estimation. The three-component model fit the data significantly better than a single normal distribution (χ 2 = 14.3 with 4 df P 2 = 9.9 with 2 df, P