A New Survival and Prognosis Predictive Model for Combined-Small Cell Lung Cancer (C-SCLC): A Machine Learning Approach

Abstract

Combined Small Cell Lung Cancer (C-SCLC), a rare variant subtype of lung cancer, has its distinct characteristics and prognosis challenges. Despite its clinical significance, there exists a knowledge gap in the diagnosis, treatment, and prognosis of C-SCLC. Utilizing the SEER database, an authoritative source of cancer data in the United States, we applied advanced machine learning techniques to analyze the overall survival (OS) of C-SCLC patients from 2004 to 2020, across multiple staging systems. Through rigorous data preprocessing and analysis, we developed predictive models that highlight the prognosis factors of C-SCLC but also underscore the subtype’s important features and analysis across stages. Our findings provide significant insights and contributions into the survival outcomes of C-SCLC patients, for their distinction within lung cancer classifications. Our study of C-SCLC falls among the very few longitudinal studies and analyses. We presented a model to predict the OS (Overall Survival) for patients with this rare subtype from the year 2004 to 2015, using the American Joint Committee on Cancer (AJCC) 6th edition, alongside visualizations from the analysis, and the code to reproduce these results. We only use the 2004-2015 data to model to ensure consistency in the corresponding staging system (AJCC 6th Edition). We presented a model with 81% recall for patients at high risk (less than 9 months of survival) with the most contributing factors to this prediction, which are Metastasis, Chemotherapy, Radiation, Surgery, and Tumor Size.