Outcomes after treatment of 17 378 patients with locally advanced (T3N0–2) non-small-cell lung cancer†

OBJECTIVES: Treatment patterns and outcomes in a population-based database were examined to identify patients likely to benefit from surgical resection of locally advanced (T3N0–2) non-small-cell lung cancer (NSCLC). METHODS: Factors predicting the use of surgery for patients with T3N0–2M0 NSCLC in the Surveillance, Epidemiology and End Results (SEER) database from 1988 to 2010 were assessed using a multivariable logistic regression model. Survival was analysed using the Kaplan– Meier approach and Cox proportional hazard models. Propensity matching was used to compare outcomes after surgery and outcomes in patients who refused surgery and underwent radiation therapy (RT). RESULTS: Of 17 378 patients identified for study inclusion [8597 (50%) T3N0, 2304 (13%) T3N1 and 6477 (37%) T3N2], surgery was used in 7120 (41%). Only female sex and being married predicted the use of surgery, while older age, black race and N2 nodal disease predicted non-surgical management. Surgical patients overall had better long-term survival than non-surgical patients [odds ratio (OR) 0.42, 95% confidence interval (CI): 0.41–0.45, P < 0.001]. After propensity adjustment, patients who refused surgery and instead were treated with RT had significantly worse long-term survival than matched surgery patients (OR 0.65, 95 % CI: 0.48–0.89, P = 0.0074). Sublobar resection and pneumonectomy predicted worse survival in patients who had surgery. Nodal disease also predicted worse survival after surgery, but surgery maintained an association with better overall survival compared with non-operative therapy among patients with both N1 (OR 0.53, P < 0.001) and N2 disease (OR 0.50, P < 0.001) in separate analyses stratified by nodal status. Older age also predicted worse surviva