Video-Assisted Thoracoscopic Surgery versus Thoracotomy in Non-Small Cell Lung Cancer: Survival Analysis and Exploratory Predictive Modelling.

Background: This study retrospectively compared outcomes of minimally invasive and open surgical approaches for non-small cell lung cancer (NSCLC) in a cohort of 255 patients operated on between 2014 and 2018. A machine-learning analysis was performed as an exploratory, hypothesis-generating component. Method: Survival was analysed by disease stage using Kaplan-Meier curves and log-rank tests. Because the groups were not randomised and were imbalanced at baseline, comparisons are unadjusted and should be interpreted as associations. Exploratory predictive models were developed using five machine-learning algorithms (random forest performing best), with feature contributions interpreted using Shapley Additive Explanations (SHAP). Missing values were handled by multiple imputation, continuous variables were standardised, and correlated features were excluded. Results: Video-assisted thoracoscopic surgery (VATS) was used more frequently for adenocarcinoma, in women, and at stage I, with zero postoperative mortality (0% versus 5.8% for thoracotomy). At stage I, there were no differences in median overall survival (65.3 versus 67.7 months; p=0.768); at stage II, VATS was associated with longer survival (75.8 versus 48.8 months; p=0.003); at stage III, a similar direction did not reach significance. Recurrent predictors in the exploratory models were coronary artery disease, chronic obstructive pulmonary disease, intervention volume, tumour histology, and vascular invasion; however, test accuracy remained low (0.39-0.48), precluding clinical use. SHAP analysis highlighted resection completeness, lack of conversion, and tumour morphology. Conclusions: In this unadjusted retrospective cohort, VATS was associated with lower perioperative mortality, and at stage II with longer overall survival. Given the retrospective design and baseline imbalances, these findings are hypothesis-generating and require confirmation in prospective, adjusted studies before informing surgical indications.