In a prospective population-based study of middle-aged Caucasian men, performed in Malmö, Sweden, specifically designed to evaluate physical fitness, early and late insulin response as predictors of non-insulin-dependent diabetes mellitus (NIDDM), 4,637 non-diabetic men underwent oral glucose tolerance tests at the ages of 48 and 54 years. At the baseline examination, physical fitness was measured in terms of lung vital capacity and oxygen uptake during ergometry; early insulin response in terms of the 40-min insulin increment during an oral glucose tolerance test (a correlate of acute insulin response to an intravenous glucose tolerance test), and late insulin response were measured in terms of the 2-h insulin value during the oral glucose tolerance test (a correlate of glucose disposal during euglycaemic clamp testing). Of the subjects studied 116 developed NIDDM (0.4% annually), and when compared with non-diabetic men at baseline, they were found to have an 11% higher mean body mass index (p < 0.001), a higher frequency of family history of diabetes (31 vs 18%, p < 0.001), 16% lower mean physical activity index (p < 0.05), 16% lower mean estimated maximal oxygen uptake (p < 0.001), 10% lower mean vital capacity (p < 0.001), 26% lower 40-min to total insulin response ratio (p < 0.001), and a 2.7 times higher mean 2-h insulin value during an oral glucose tolerance test (p < 0.001). Regression analysis (using Cox's proportional hazards model) showed both low vital capacity, and impaired early insulin response but late hyperinsulinaemia to be independent predictors of NIDDM, in addition to body mass index and fasting blood glucose level (p = 0.05-0.0001). Among subjects with impaired glucose tolerance at baseline (44 of 278 developed NIDDM), fasting glucose level alone predicted diabetes in this model. The findings suggest that in this age group in a Caucasian population, not only does insulin resistance precede glucose intolerance and NIDDM, but also loss of early insulin response indicating impaired beta-cell function to be an early feature of the process culminating in diabetes. As both physical fitness [which correlates inversely with late insulin response (r= -0.42, p < 0.0001)], and the level of physical activity were shown to correlate with diabetes development in this large series, measures to correct these adverse features should be included in future strategies for preventing NIDDM.