Predicting overall survival from tumor dynamics metrics using parametric statistical and machine learning models: application to patients with RET-altered solid tumors

• DOI: 10.3389/frai.2024.1412865
• Publication Date: 2024-06-12T13:12:17Z
• Authors: Erick Velasquez; Nastya Kassir; Sravanthi Cheeti; Denison Kuruvilla; Rucha Sane; Steve Dang; Dale Miles; James Lu
• Keywords: Solid tumor

In oncology drug development, tumor dynamics modeling is widely applied to predict patients' overall survival (OS) via parametric models. However, the current paradigm, which assumes a disease-specific link between and survival, has its limitations. This particularly evident in development scenarios where clinical trial under consideration contains patients with types for there little no prior institutional data. this work, we propose use of pan-indication solid machine learning (ML) approach whereby all three metrics (tumor shrinkage rate, regrowth rate time growth) are simultaneously used OS type independent manner. We demonstrate utility cancer treated tyrosine kinase inhibitor, pralsetinib. compared ML models results showed that proposed able adequately patient across RET -altered tumors, including non-small cell lung cancer, medullary thyroid as well other tumors. While findings study promising, further research needed evaluating generalizability model types.