Uni and multifactorial (Cox) statistical analysis of the results of surgery in a series of 247 patients operated between 1969 and 1988 for pure, non-ischaemic mitral regurgitation was undertaken to determine the factors influencing operative and late mortality. All but 3 cases were adults, average age 51 years, and very symptomatic (75 per cent Class III et IV of NYHA Classification). Dystrophic or degenerative lesions were responsible for 53 per cent of cases of regurgitation whilst rheumatic valvular disease was only observed in 30 per cent of cases. Mitral valve replacement was performed in 137 patients (96 mechanical and 41 bioprostheses) and conservative surgery was possible in 110 cases. There were 12 operative deaths (4.9%); the operative risk increased with age and with the practice of valve replacement. Eleven of the 235 survivors (4.7%) were lost to follow-up. During the follow-up period (average 58 months) there were 37 late deaths of which nearly a half were due to left ventricular dysfunction. Multifactorial analysis identified cardiothoracic ratio and atrial fibrillation as predictive factors of late mortality whilst left ventricular ejection fraction and the type of surgery were related to the development of left ventricular dysfunction. The postoperative left ventricular ejection fraction was significantly lower after valve replacement than after valvuloplasty (11 +/- 13% vs 3 +/- 13%; p less than 0.01). These results are therefore in favour of early correction of severe mitral regurgitation by conservative surgery whenever possible.