The relationship between the level of habitual physical activity and glucose intolerance was examined cross-sectionally and during a 2-year follow-up among a sample of 388 subjects in Malta. At baseline, the subjects were classified into three categories of physical activity, which was inversely related to the 2-h post challenge blood glucose (P = 0.02). In a multivariate analysis, age (standardized regression coefficient 0.23; P less than 0.001), family history of diabetes (0.20; P less than 0.001), and physical activity (-0.18; P = 0.002) were the strongest predictors of the 2-h blood glucose at baseline. The age standardized 2-year risk of glucose intolerance, i.e. impaired glucose tolerance or diabetes was consistently and inversely related to the level of physical activity. Among subjects with normal glucose tolerance at baseline (n = 127) those with low physical activity had a 2.7 times higher risk of glucose intolerance during follow-up than those with high physical activity (P = 0.1), and even a 3.7-fold risk of glucose intolerance at baseline (n = 196) when both the subjects with normal and impaired glucose tolerance at baseline were considered together (P = 0.005). Similar trends were observed for the risk of diabetes. The suggested protective effect of physical activity was independent of body mass, a family history of diabetes and gender. Within the limits of this small study we conclude that physical activity may have some importance in the primary prevention of impaired glucose tolerance and, possibly, non-insulin-dependent diabetes mellitus.