In order to determine the influence of tobacco smoking on the course and long-term prognosis of idiopathic dilated cardiomyopathy, we conducted a retrospective study of 111 patients (95 men and 16 women, mean age 45.5 +/- 8.1 years) who had undergone cardiac catheterization between January 1970 and December 1979 and had been followed up for 6 to 16 years. The criteria of inclusion was diffuse hypokinesia of the left ventricle with an ejection fraction of 50 per cent or less, normal coronary arteriography and cardiomyopathy of unknown origin. The overall mortality rates at 1, 5 and 10 years were 10, 50 and 66 per cent respectively, and the main predictive factor was the left ventricular ejection fraction. Forty-six per cent of these patients were smokers, 19 per cent were non-smokers and 35 per cent had undetermined smoking habits. A univariate analysis showed a favourable predictive effect of smoking on survival (P less than 0.01), and this was confirmed by the statistical hypothesis of maximum bias for patients with undetermined smoking habits. On multivariate analysis, this predictive effect was superseded by the left ventricular ejection fraction, but after stratification of the sample according to the mean value of ejection fraction (30 per cent), the predictive value of smoking reappeared clearly in the group with a less than 30 per cent ejection fraction (P less than 0.003).