Peripheral blood T-lymphocyte subsets provide prediction of early postoperative recurrence in elderly patients with lung cancer: model development and external validation.

Background: With the accelerating aging of the population, the number of elderly patients with lung cancer continues to increase. Although surgery remains an important treatment modality for lung cancer, early recurrence (ER) is still a major factor affecting long-term survival in elderly patients. Purpose: This study aimed to integrate peripheral blood T-lymphocyte subset indicators with clinicopathological characteristics to develop and externally validate a machine learning-based prediction model for ER in elderly patients with lung cancer. Methods: This study was designed as a multicenter retrospective cohort study. ER was defined as local recurrence or distant metastasis within 2 years after surgery. A total of 23 candidate predictors were included, and dual feature selection was performed using the Boruta algorithm and LASSO regression, followed by a systematic comparison of 14 machine learning algorithms and 7 ensemble optimization strategies. Model performance was evaluated using the area under the receiver operating characteristic curve (AUROC), area under the precision-recall curve (AUPRC), calibration curve, Brier score, and decision curve analysis. SHAP was used to interpret the optimal model. Results: A total of 893 elderly patients with lung cancer were included in this study. After dual feature selection, five predictors were ultimately retained: tumor stage, total T-lymphocyte percentage, body mass index (BMI), Karnofsky performance scale (KPS), and suppressor/cytotoxic T-lymphocyte percentage. After ensemble optimization, SMOTE-Stacking achieved the best performance in internal validation. In external validation, among the four models belonging to the statistically equivalent top-performing tier, Soft Voting Ensemble (SV) demonstrated the most robust performance. SHAP analysis indicated that tumor stage and total T-lymphocyte percentage were the two core drivers of model prediction, both exhibiting clear nonlinear threshold effects. Conclusion: ER in elderly patients with lung cancer results from the combined effects of tumor burden, host functional status, and T-lymphocyte immune imbalance. A machine learning model integrating peripheral blood T-lymphocyte subsets has noninvasive and clinically translational potential and may provide evidence for ER risk stratification, individualized follow-up, and early intervention decision-making in elderly patients with lung cancer.