The objectives of the present study were: (i) to compare clinical assessment of coronary heart disease (CHD) risk with risk estimation faced with simulated, written case histories; (ii) to observe the risk assessment performed by general practitioners (GPs) in their clinical setting. Thirty-one GPs participating in a multicentre study were asked to invite 20 consecutive male patients aged 30-59 years to an opportunistic screening of CHD risk factors. They assessed the risk status of these patients and of 10 written case histories containing information about corresponding CHD risk factors. A composite 'infarction score' computed from epidemiologic data was used as a gold standard. Diagnostic performance in the clinical setting was compared with that in the simulated setting by Pearson's correlation. A weak, but statistically significant positive correlation was demonstrated when comparing correct estimation in the two settings. No correlation was found for over- and underestimation. Sensitivity was increased faced with clinical patients at the sacrifice of specificity compared to the simulated setting. The impact of a positive family history on clinical assessment parallels the epidemiological estimate. Due to lack of sensitivity, the other factors had a lower impact on risk estimation than an epidemiological estimate would presuppose. We advocate the application of a formal risk estimation to improve risk assessment accuracy. The synergistic effect of multiple risk factors should be emphasized in medical training to improve the clinical risk estimation.