Venous plasma glucose levels two hours after a 75 gm. carbohydrate load were determined on over 2,900 Pima Indians, a population known to have an extremely high prevalence of diabetes mellitus. In each sex and in each decade above twenty-five years of age, the frequency distributions of the logarithms of the glucose levels were clearly bimodal, but below this age a single symmetrical unimodal distribution was found. A maximum likelihood procedure was used to derive the best fitting theoretical gaussian distributions for each group of data, together with the parameters of each distribution.
The observed bimodal distributions were found to be in satisfactory agreement with a model of two overlapping gaussian distributions, indicating that a logical separation between those with normal and high levels of glucose is possible, although the presence of overlap indicates that some misclassification will occur if any finite level is used to subdivide the population.
The data indicate that among the Pima: 1. The frequency distributions of two-hour glucose tolerance levels can be used to identify objectively and describe a hyperglycemic population without recourse to other criteria for diabetes. 2. There are small changes in the parameters of “normal” glucose tolerance between the ages of twenty-five and sixtyfour years. 3. The increase in mean glucose level found with rising age in this population is mainly the result of an increasing proportion of subjects who are in the group characterized by marked glucose intolerance.
The bimodal distributions of plasma glucose levels The bimodal distributions of plasma glucose levels among the Pima Indians contrast with those described so far in other groups. It seems likely that differences are attributable to the lower prevalence of diabetes elsewhere which would obscure the identification of bimodality.