There is limited evidence that raised hematocrit levels may be associated with insulin resistance, which links cardiovascular disease with NIDDM. The association between hematocrit level at screening and the subsequent development of physician-diagnosed NIDDM during 12.8 years of follow-up was examined in a prospective study of 7,735 middle-aged men drawn at random from general practice in 24 British towns. With the exclusion of men with missing hematocrit data and men with diabetes at screening, data were available for 7,193 men, in whom there were 187 new cases of NIDDM during follow-up. The risk of NIDDM increased significantly with increasing hematocrit levels. There was more than a fourfold increase in relative risk (RR) of diabetes among men with a hematocrit of > or = 48% relative to those with a hematocrit <42%, adjusted for age and BMI (RR 4.5; 95% CI 2.5-6.3). On further adjustment for predictors of NIDDM with which hematocrit is correlated, there remained a strong linear association with the risk of diabetes. There was a nearly fourfold increased risk of NIDDM in the highest relative to the lowest hematocrit group in the fully adjusted proportional hazard model (RR 3.6; 95% CI 1.7-7.6). The strong positive association between hematocrit level and risk of diabetes was seen even after exclusion of men with preexisting ischemic heart disease. The findings suggest that a raised hematocrit level, which is a major determinant of whole blood viscosity, should be added to the cluster of risk factors that link NIDDM with atheromatous, vascular disease.