In many population and screening studies of diabetes, the onehour glucose level of the GTT has been used to define the diabetic status of subjects. The one-hour postglucose load determination has been preferred over the two-hour value by many investigators primarily because of convenience and justified on the basis of the high correlation between the two values.

Venous plasma glucose levels, one and two hours after a 75-gm. carbohydrate load, were determined on over 1,600 Pima Indians. In most sex and age groups, the frequency distributions of both the one-hour and two-hour glucose levels were clearly bimodal. By the logarithms of the glucose values these distributions were consistent with a model of two overlapping Gaussian distributions. The data indicate that for the Pima the amount of overlap of the distributions was greater for the one-hour than for the two-hour values. For each sex and decade the probabilities of misclassification of a normal as “hyperglycemic” and vice versa were smaller for the two-hour than for the one-hour values. Such misclassifications for the two-hour levels averaged 6.6 per cent and 11.6 per cent for the one-hour levels.

The reproducibility of the GTT taken one to three weeks apart in a sample of ninety-nine Puna Indians showed that the two-hour level was superior to the one-hour level as measured by the mean values of the absolute difference between log GTT levels for test and retest values. The one-hour measurements also gave more disagreements between the classifications of diabetic status than the two-hour test values.

If a single measure of glucose tolerance is to be selected for the diagnosis of diabetes among Pima Indians, these data provide a mathematical rationale for preferring the two-hour level to the one-hour determination.