Aims/hypothesis. The degree of glycaemia has been shown to be associated with all-cause and cardiovascular mortality in diabetic subjects. Whether this association also exists in the general population is still controversial. We studied the predictive value of fasting plasma glucose, 2-hour post-load glucose and HbA1 c in a population-based cohort of 2363 older (50–75 years) subjects, without known diabetes. Methods. Relative risks (RR) of all-cause and cardiovascular mortality were estimated by Cox proportional hazards model, adjusting for age and sex, and additionally for known cardiovascular risk factors. Results. During 8 years of follow-up, 185 subjects died; 98 of cardiovascular causes. Fasting plasma glucose was only predictive in the diabetic range, although the risks started to increase at about 6.1 mmol/l. Post-load glucose and HbA1 c values were, even within the non-diabetic range, associated with an increased risk (p for linear trend < 0.05). These increased risks were mostly, but not completely, attributable to known cardiovascular risk factors. After exclusion of subjects with newly diagnosed diabetes or with pre-existent cardiovascular disease (n = 551), a 5.8 mmol/l increase of post-load glucose (corresponding to two standard deviations of the population distribution) was associated with a higher age-adjusted and sex-adjusted risk of all-cause (RR 2.24) and cardiovascular mortality (RR 3.40) (p < 0.05). After additional adjustment for known cardiovascular risk factors, these relative risks were still statistically significant, with values of 2.20 and 3.00 respectively (p < 0.05). Conclusion/interpretation. High glycaemic variables, especially 2-h post-load glucose concentrations and to a lesser extent HbA1 c values, indicate a risk of all-cause and cardiovascular mortality in a general population without known diabetes. [Diabetologia (1999) 42: 926–931]