Prognostic value of preoperative inflammatory markers in resectable non-small cell lung cancer: a multi-center retrospective study based on logistic regression and machine learning techniques.

Purpose: This study aimed to assess the prognostic significance of the neutrophil-to-lymphocyte ratio (NLR), platelet-to-lymphocyte ratio (PLR), systemic immune-inflammation index (SII), and systemic inflammation response index (SIRI), measured 15 days prior to surgery, in patients undergoing primary resection for NSCLC. Methods: We retrospectively analyzed data from 460 NSCLC patients treated at two comprehensive hospitals, along with another 50 patients for external validation. Optimal cut-off values for NLR, PLR, SII, and SIRI were determined. Kaplan-Meier analysis was performed to estimate overall survival (OS) and progression-free survival (PFS). Univariate and multivariate Cox regression analyses were conducted to identify independent prognostic factors. To evaluate and validate the reliability of prognostic factor selection, three commonly used machine learning methods-LASSO, Random Forest, and Support Vector Machine (SVM)-were applied. The systemic inflammatory markers were subjected to external validation in an independent cohort, encompassing discriminative analysis, calibration analysis, and clinical decision curve analysis (DCA), with a final evaluation of the SIRI's underlying rationale. Results: The optimal cut-off values for NLR, PLR, SII, and SIRI were 3.470, 186, 853.71, and 1.66, respectively. Kaplan-Meier curves demonstrated that elevated levels of all four markers were significantly associated with poorer OS and PFS. The 3- and 5-year OS rates were 83.6 and 72.6%, while the 3- and 5-year PFS rates were 82.4 and 75.0%, respectively. Univariate analysis identified several factors significantly associated with survival, including inflammatory markers, smoking, antibacterial use, lymphatic metastasis, radiotherapy, intraoperative blood loss, and preoperative albumin levels. Multivariate analysis further revealed that lymphatic metastasis, antibacterial use, NLR, and SIRI were independent predictors of OS, whereas lymphatic metastasis, radiotherapy, antibacterial use, NLR, and SIRI were independently associated with PFS. Consistently, machine learning methods also highlighted NLR and SIRI as reliable independent prognostic indicators in patients with resectable NSCLC. SIRI showed good discrimination with a concordance index of 0.803 (95% CI: 0.644-0.962) for external validation. Conclusion: Preoperative systemic inflammatory markers, particularly SIRI, are strong and independent prognostic indicators of PFS in patients with resectable NSCLC. Among these, a low SIRI may provide superior risk stratification for identifying high-risk patients and informing individualized treatment strategies.