We investigated the prognostic significance of risk scores and exercise workload with respect to stroke. Background. There are no data on exercise workload combined with European Systematic Coronary Risk Evaluation (SCORE) in the prediction of stroke. Exercise workload was measured by exercise test with an electrically braked cycle ergometer performed at baseline. The study is based on a random population-based sample of 1639 men (42-60 years) without history of type 2 diabetes or atherosclerotic cardiovascular disease including coronary heart disease, stroke or claudication. During an average follow-up of 16 years, a total of 97 strokes occurred, of which 71 were ischaemic strokes. Independent predictors for all strokes were European SCORE [for 1% increment, relative risk (RR): 1.12, 95% CI: 1.02 to 1.22, P=0.017), maximal workload (for 20 W increment, RR: 0.87, 95% CI: 0.80 to 0.95, P=0.003) and body mass index (for 5 kg m(-2) increment, RR: 1.08, 95% CI: 1.03 to 1.14, P=0.004), when adjusted for serum HDL, alcohol consumption, C-reactive protein, family history of coronary heart disease, exercise-induced ST changes and the use of medications for hypertension, dyslipidaemia or aspirin. The risk was 2.54-fold (95% CI: 1.27-5.09, P=0.008) for any strokes and 4.43-fold (95% CI 1.69-11.78, P=0.003) for ischaemic strokes amongst men with exercise capacity less than 162 W when compared with those with high exercise capacity over 230 W, after adjustment for risk factors. Low exercise workload predicts an especially high risk for stroke in the presence of high risk SCORE.