In order to investigate the value of peak oxygen consumption (peak VO2) in predicting mortality in mild to moderately symptomatic patients with left ventricular dysfunction, we studied 103 NYHA II/III class patients with a left ventricular ejection fraction (LVEF) < or = 40%. Heart failure was due to coronary artery disease (CAD) in 39 patients, idiopathic dilated cardiomyopathy in 54, hypertension in eight and surgically corrected valvular disease in two. The following variables were analysed: age, cause of heart failure (CAD vs no CAD), NYHA class, peak VO2, LVEF, left ventricular end-systolic volume index (LVESVI), ventricular tachycardia (VT) on Holter monitoring and the use of antiarrhythmic drugs. Statistical analysis was performed by Cox's proportional-hazards regression model. During a mean follow-up period of 20 months, there were 25 deaths. The estimated cumulative probabilities of survival were 88%, 73% and 58% at 1, 2 and 3 years, respectively. Cox's model identified CAD (P = 0.01), NYHA III class (P = 0.04) and LVEF (P = 0.02) as independent, statistically significant predictors of mortality. Peak VO2 had only a marginal statistical significance (P = 0.07). Age, LVESVI, VT on Holter monitoring and use of antiarrhythmic drugs were not related to mortality. These data can be important in patient clinical management and clinical trial design.