Random versus predictable risks of mortality after thoracotomy for lung cancer

Over a period of 12 1/2 years, 476 patients underwent thoracotomy for lung cancer at two affiliated hospitals. Hospital mortality for all patients was 5.25% and for those undergoing pulmonary resection, 5.67%. Hospital mortality is more indicative of true risk than is the 30 day mortality figure, which we regard as arbitrary and misleadingly low. Thirty-seven preoperative risk factors were analyzed for their effects on both morbidity and mortality, and 12 classes of postoperative complications were analyzed for their effect on mortality. All preoperative risk factors together accounted only for 12% of the risk of mortality (R2 by multiple regression analysis). Only three of these factors bore a significant association with mortality: patient age 60 years or over (p less than 0.05), need for pneumonectomy (p less than 0.005), and premature ventricular contractions on the admission electrocardiogram (p less than 0.05). All the listed postoperative complications together accounted for only 28% of the risk of mortality. Of these complications, four showed a significant association with postoperative death: infectious complications (pneumonia and empyema) and cardiovascular accidents (pulmonary embolism and myocardial infarction). In both analyses, the remainder of the risk of death must be attributed either to factors not considered or to purely random factors. It follows that much the greater part of the risk of death from surgical treatment of lung cancer could not be predicted from the preoperative status of the patients.