Objective - To examine the validity of estimates of coronary heart disease (CHD) risk by the Framingham risk function, for European populations. Design - Comparison of CHD risk estimates for individuals derived from the Framingham, prospective cardiovascular Munster (PROCAM), Dundee, and British regional heart (BRHS) risk functions. Setting - Sheffield Hypertension Clinic. Patients - 206 consecutive hypertensive men aged 35-75 years without preexisting vascular disease. Results - There was close agreement among the Framingham, PROCAM, and Dundee risk functions for average CHD risk. For individuals the best correlation was between Framingham and PROCAM, both of which use high density lipoprotein (HDL) cholesterol. When Framingham was used to target a CHD event rate > 3% per year, it identified men with mean CHD risk by PROCAM of 4.6% per year and all had CHD event risks > 1.5% per year. Men at lower risk by Framingham had a mean CHD risk by PROCAM of 1.5% per year, with 16% having a CHD event risk > 3.0% per year. BRHS risk function estimates of CHD risk were fourfold lower than those for the other three risk functions, but with moderate correlations, suggesting an important systematic error. Conclusion - There is close agreement between the Framingham, PROCAM, and Dundee risk functions as regards average CHD risk, and moderate agreement for estimates within individuals. Taking PRO-CAM as the external standard, the Framingham function separates high and low CHD risk groups and is acceptably accurate for northern European populations, at least in men.