To impact on the development of clinical congestive heart failure as a complication of doxorubicin therapy, left ventricular ejection fraction was monitored with serial resting radionuclide angiocardiography in 1,487 patients with cancer over a seven-year period in both university and community hospital environments. A high-risk subset of 282 patients was selected for retrospective analysis of their clinical outcome. High-risk patients were identified by one or two of the following three criteria: decline of 10 percent or more in absolute left ventricular ejection fraction from a normal baseline to 50 percent or less; high cumulative dose of doxorubicin (more than 450 mg/m2); abnormal baseline left ventricular ejection fraction (less than 50 percent). Clinical congestive heart failure occurred in 46 (16 percent) during the treatment period, and in an additional three patients (1.3 percent) at last follow-up examination 11.8 +/- 14.2 months following discontinuation of doxorubicin. Total cumulative dosages of doxorubicin that precipitated congestive heart failure (75 to 1,095 mg/m2) and those that did not (30 to 880 mg/m2) varied widely. Decline of 10 percent or more in absolute left ventricular ejection fraction to a value of 50 percent or less preceded administration of the final dose of doxorubicin that precipitated clinical congestive heart failure in the majority of patients in whom congestive heart failure developed. Clinical congestive heart failure improved in 87 percent given routine therapy with digitalis, diuretics, and/or vasodilators. Criteria for monitoring left ventricular ejection fraction and discontinuing doxorubicin were formulated. The occurrence of clinical congestive heart failure was compared in those patients whose management was concordant with proposed criteria (Group A) and in those whose management was not (Group B). Group A had a lower incidence of congestive heart failure compared with Group B (2.9 percent versus 20.8 percent, p less than 0.001) and had only mild congestive heart failure that resolved with treatment (n = 2) and no deaths due to congestive heart failure. Multivariate analysis with proportional-hazards regression (Cox's model) demonstrated a fourfold reduction in the incidence of congestive heart failure independent of other clinical predictor variables in those patients whose management was concordant with proposed guideline criteria. The incidence, persistence, late development, predictability, and reversibility of clinical congestive heart failure were comparable in university and community hospital settings.(ABSTRACT TRUNCATED AT 400 WORDS)