Predicting Postoperative Lung Cancer Recurrence and Survival Using Cox Proportional Hazards Regression and Machine Learning

Overview

Surgical resection remains the optimal treatment for early-stage lung cancer. However, the recurrence rate after surgery is unacceptably high (30%-50%). Despite the significant efforts, it remains elusive to accurately predict the likelihood and timing of recurrence. In this study, we propose to predict postoperative lung cancer recurrence by identifying novel image biomarkers from preoperative chest CT scans. A cohort of 309 patients was selected from 512 non-small cell lung cancer patients who underwent lung resection. We used Cox proportional hazard regression analysis to identify risk factors associated with lung cancer recurrence and compared its performance with machine learning (ML) methods in predicting lung cancer recurrence. The goal is to improve our ability to predict the risk and time of a seemingly “cured” cancer to come back and facilitate personalized surveillance strategies and thus minimize lung cancer recurrence. Our experimental results showed that surgical procedure, TNM staging, lymph node involvement, body composition, and tumor characteristics are important determinants of the risk of both local/regional and distant recurrence for recurrence-free survival (RFS) and overall survival (OS). ML -based approaches and Cox models exhibited similar performance with an area under the receiver operative characteristic (ROC) curve (AUC) ranging from 0.75- 0.77. ML -based approaches may be a useful analytic approach for survival prediction in lung cancer recurrence. We expect that this computer tool could be an important addition to the clinical practice of lung cancer treatment and may improve decision-making for patients, oncologists, and surgeons, ultimately improving the long-term quality of life for patients.