For a series of 649 patients who died of coronary heart disease within 24 hours of onset of symptoms, and an equal number of neighborhood controls, information was obtained on a large number of variables, including coffee consumption. An analysis using multivariate risk scores to control for all available variables yields a maximum likelihood estimate of the risk ratio associated with coffee drinking of 1.1 (95 per cent two-sided confidence limits, 0.8 to 1.6). The estimate of the risk ratio depends somewhat on the number and nature of variables controlled for in the analysis. Overall, our findings, limited to low-risk and middle-risk patients, suggest that the risk, if any, of death from coronary heart disease associated with coffee drinking is small.