The aim of the present study was to determine the risk factors and impact on outcome of blood transfusions following pneumonectomy for thoracic malignancies. A retrospective analysis of 432 consecutive patients was carried out, of whom 183 (42.4%) were transfused post-operatively. The associations between blood transfusions and 20 variables were assessed by univariate and multivariate analysis. Survival analysis included log-rank test and Cox regression model. Patient age, neoadjuvant treatment, completion pneumonectomy and extended procedures were independent predictors of transfusion. It was found that 30-day mortality increased significantly from 2.4% (no transfusion) to 10.9 and 21.9% (<or=2 and >2 red blood cell packs, respectively). Blood transfusion was the strongest predictor of 30-day mortality (odds ratio (OR) 10; 95% confidence interval (CI): 3.7-27), respiratory failure (OR 19.2; 95% CI 7.4-49.4) and infectious complications (OR 3; 95% CI 1.5-6.2). In the 367 lung cancer patients, a significantly lower 5-yr survival was observed in univariate analysis of transfused patients (27.8+/-5.4% versus 39.4+/-4.5%). In a Cox regression analysis, blood transfusion was no longer found to be significant. A dose-related correlation is suggested between blood transfusion and early mortality through an increase of infectious and respiratory complications. In contrast, blood transfusion had no independent adverse impact on long-term survival.