Predicting death from surgery for lung cancer: a comparison of two scoring systems in two European countries

Objectives: Current British guidelines advocate the use of risk prediction scores such as Thoracoscore to estimate mortality prior to radical surgery for non-small cell lung cancer (NSCLC). A recent publication used the National Lung Cancer Audit (NLCA) to produce a score to predict 90 day mortality (NLCA score). The aim of this study is to validate the NLCA score, and compare its performance with Thoracoscore. Materials and methods: We performed an internal validation using 2858 surgical patients from NLCA and an external validation using 3191 surgical patients from the Danish Lung Cancer Registry (DLCR). We calculated the proportion that died within 90 days of surgery. The discriminatory power of both scores was assessed by a receiver operating characteristic (ROC) and an area under the curve (AUC) calculation. Results: Ninety day mortality was 5% in both groups. AUC values for internal and external validation of NLCA score and validation of Thoracoscore were 0.68 (95% CI 0.63–0.72), 0.60 (95% CI 0.56–0.65) and 0.60 (95% CI 0.54–0.66) respectively. Post-hoc analysis was performed using NLCA records on 15554 surgical patients to derive summary tables for 30 and 90 day mortality, stratified by procedure type, age and performance status. Conclusions: Neither score performs well enough to be advocated for individual risk stratification prior to lung cancer surgery. It may be that additional physiological parameters are required; however this is a further project. In the interim we propose the use of our summary tables that provide the real-life range of mortality for lobectomy and pneumonectomy