OBJECTIVE We examined whether the changes in clinical practice with time correlated with the changes in the 90-day mortality following pneumonectomy. METHODS The clinical records of consecutive patients undergoing pneumonectomy in two French centres from 1980 to 2009 were prospectively collected. The 90-day postoperative course was retrospectively studied according to clinical characteristics, underlying diseases, type of surgery and time-period (1980-1989; 1990-1999 or 2000-2009). RESULTS Pneumonectomy was performed in 2064 patients (right n = 948, males n = 1758, mean age 60 ± 10 years). Indications were non-small-cell lung cancer (n = 1805, 87%), mesothelioma (n = 39, 1.8%), other tumours (n = 132, 6.3%) and non-tumour disease (n = 88, 4.2%). The 30- and 90-day mortality were 17.4 and 7.2% in the first decade, 22.3 and 9% in the second decade and 26.4 and 7.3% in the third decade, respectively. In multivariate analysis, older age, right-sided resection, T3-T4 and N2 lung cancer disease were significantly associated with increased overall 90-day mortality, whereas surgery during the last decade was associated with a better outcome when compared with the first decade (RR: 0.63, 95% confidence interval: 0.50-0.80, P = 0.045). When focusing on patients with non-small-cell lung cancer (NSCLC), the 90-day mortality following induction therapy and pneumonectomy decreased from 21.9% in the 1980s to 8.2% in the 2000s (P = 0.038), while such decrease was not found in patients without induction therapy or in patients undergoing a lobectomy. CONCLUSIONS The overall 90-day mortality after pneumonectomy was not significantly modified over the last 30 years, while the 90-day mortality after induction therapy followed by pneumonectomy for NSCLC decreased significantly.