Myocardial damage from heavy alcohol intake can cause the heart failure (HF) syndrome, but the relation of lighter alcohol intake to HF has rarely been studied. We examined the risk of HF hospitalization among 126,236 subjects who supplied data about alcohol during health examinations from 1978 to 1985. Among 2,594 subjects who were subsequently hospitalized for HF, record review established an association between coronary artery disease (CAD) and HF (CAD-HF) in 1,559 patients. Among the remaining 1,035 subjects who had HF (non-CAD-HF), we attempted determination of preponderant etiologic and contributory factors. Analyses used Cox models that were controlled for 7 covariates, with usual alcohol intake studied categorically compared with that in subjects who did not drink alcohol. Heavier drinkers (> or =3 drinks/day) but not light to moderate drinkers had increased risk of non-CAD-HF; e.g., relative risk for subjects who reported > or =6 drinks/day was 1.7 (95% confidence interval 1.1 to 2.6). This association of non-CAD-HF with heavy drinking was limited to subsets with cardiomyopathy or of unclear preponderant etiology. Alcohol drinking was inversely related to risk of CAD-HF (e.g., at 1 to 2 drinks/day, relative risk 0.6, 95% confidence interval 0.5 to 0.7), with consistency across subgroups of age, gender, ethnicity, education, smoking status, interval to diagnosis, and presence or absence of baseline heart disease or systemic hypertension. Moderate drinking was inversely related to non-CAD-HF only in subjects who had diabetes mellitus (n = 252). In conclusion, heavy, but not light, alcohol drinking is associated with increased risk of non-CAD-HF and that apparent protection by alcohol drinking against CAD-HF risk provides confirmation of a protective effect of alcohol against CAD.