Exploration and machine learning model development for T2 NSCLC with bronchus infiltration and obstructive...

Area of Science:

• Oncology
• Thoracic Surgery
• Cancer Staging

Background:

• The 8th edition AJCC staging system reclassified Non-Small Cell Lung Cancer (NSCLC) with main bronchial infiltration (MBI) and pneumonia/atelectasis (P/ATL) from T3 to T2.
• These reclassifications impact prognostic assessment and treatment strategies for NSCLC patients.

Purpose of the Study:

• To evaluate the survival outcomes and optimal treatment strategies for T2 NSCLC patients with MBI or P/ATL.
• To assess the efficacy of predictive models for prognostication and treatment allocation in these specific NSCLC subgroups.

Main Methods:

• Analysis of the SEER database (2007-2015) using Propensity Score Matching (PSM) to compare survival between patients with and without MBI/P/ATL.
• Evaluation of treatment approaches, including surgery, induction therapy, and adjuvant therapy, for T2 NSCLC with MBI/P/ATL.
• Application of the XGBoost model for 5-year survival prediction and treatment determination.

Main Results:

• Patients with P/ATL (median OS 12 months) and MBI (median OS 22 months) had significantly inferior overall survival compared to controls.
• Both MBI and P/ATL were strongly correlated with lymph node metastasis.
• Surgery was identified as the optimal treatment strategy. For MBI patients, surgery combined with adjuvant or induction therapy significantly improved prognosis, a benefit not observed in P/ATL patients.
• The XGBoost model demonstrated high efficacy (AUC 0.853 for P/ATL, 0.814 for MBI) in predicting survival and guiding treatment.

Conclusions:

• The reclassification of MBI and P/ATL to T2 NSCLC highlights their significant negative impact on patient survival.
• Surgery, particularly when combined with multimodal therapy for MBI, offers improved outcomes.
• The XGBoost model provides a valuable tool for personalized prognostication and treatment selection in these T2 NSCLC subtypes.