Between January 1985 and June 1992, 263 consecutive patients had mitral valve reconstruction (133 patients) or replacement (130 patients) for degenerative or ischemic mitral regurgitation. The two groups were similar in sex, age, prior infarctions or cardiac operations, hypertension, angina, and functional class. Both groups were similar in mean ejection fraction, pulmonary artery pressure, cardiac index, and incidence of coronary artery disease. More reconstruction than replacement patients had ischemic etiology (22 [16%] versus 12 [9%]; p = not significant), and fewer reconstruction patients had ruptured anterior leaflet chordae (9 [7%] versus 39 [30%]; p < 0.01). More reconstruction than replacement patients had concomitant cardiac procedures (67 [50%] versus 59 [45%]; p = not significant). Hospital death occurred in 4 reconstruction patients (3%) and 15 (12%) replacement patients (p < 0.01). Median postoperative stay was shorter in reconstruction patients (10 versus 12 days; p = 0.02). Late valve-related death occurred in 3 reconstruction patients (2%) and 8 (6%) replacement patients (p = 0.08). Six-year actuarial freedom from thromboembolism was 92% for the reconstruction group and 85% for the replacement group (p = 0.12). Freedom from all valve-related morbidity and mortality was 85% for the reconstruction patients and 73% for the replacement patients (p = 0.03). Significant multivariate predictors of hospital death were age, mitral valve replacement, functional class, congestive heart failure, no posterior chordal rupture, and nonelective operation. Mitral valve reconstruction, when technically feasible, is the procedure of choice for degenerative or ischemic mitral regurgitation because of significantly lower hospital mortality and late valve-related events.