The population of this study consisted of 1 024 patients with mitral stenosis who underwent percutaneous mitral commissurotomy (PMC) between March 1986 and 1995 in our institution. Mean age was 49+/-14 years. Most patients were highly symptomatic in NYHA class III or IV (77%) and 314 patients (31%) had unfavourable anatomical caracteristics with a Cormier score of 3. Pre-procedural mean valve area was 1.1+/-0.2 cm(2). Finally, 163 patients (16%) had a history of commissurotomy. PMC was systematically performed using the antegrade transvenous approach. Follow-up was complete in 749 patients and reaches 20 years. Procedural mortality was 0.4 % and the most frequent complication was traumatic severe mitral regurgitation >= grade 3 of Sellers' classification in 3.4 % of cases. Good immediate results defined as a final mitral valve area >= 1.5 cm(2) with mitral regurgitation <= 2 was obtained in 912 patients (89%). Good late results defined as cardiovascular survival without reintervention on the mitral valve and in NYHA class I or II, were obtained in 29% of patients at 20 years and in 32% of patients with good immediate results of PMC. Using a multivariate Cox analysis, predictive factors of good late results were: age<50 years (p<0.0001), female sex (p=0.009), NYHA class I or II (p=0.001), Cormier's score 1 (p=0.0009), no history of commissurotomy (p=0.005) for pre-procedural variables, and large mitral valve area (p<0.0001) and low mean transmitral gradient (p<0.0001) for post-procedural variables. PMC is safe and provides good results both immediate and late. Even when late deterioration of results occurred, PMC was useful to delay surgery and its inherent complications. This study is therefore in favour of the large use of PMC in selected patients according to the identified predictive factors of good late results.