Artificial Intelligence-Informed Adaptive Clinical Trial of Adjuvant Immunotherapy for High-risk Oral Cavity Cancer (AI-ACTIVATE)
Head and neck cancer is the 6th most frequently diagnosed cancer worldwide. Oral cavity cancer is the most common cancer type from the head and neck, and overall survival of these patients remains poor with aggressive multimodality treatment. Immunotherapy has been approved for recurrent or metastatic oral cavity cancer, but its role in definitive treatment is an active area of clinical investigation. We show that sophisticated histomorphometric analysis with artificial intelligence (AI)-based approaches computationally capture the spatial architecture of both tumor cells and tumor-infiltrating lymphocytes (TILs) and integrate into a risk classifier based on digital pathology to inform prognosis. A comprehensive understanding of pathomics and other predictive tools is essential to guide treatment decision-making. Using digital pathology, we seek to develop histomorphometric analysis of TILs (HistoTIL) as a risk classifier to identify patients of oral cavity cancer with high risk for recurrence. Next, we will validate the predictive power of random forest classifier with 16 input features (RF16) for response to immunotherapy. Integration of HistoTIL with RF16 increases the global impact of our study by creating a tiered solution for companion diagnostics that can be applied into different resource settings. Through the proposed AI-ACTIVATE trial, we aim to seamlessly integrate HistoTIL and RF16 into the clinical management for high-risk oral cavity cancer to guide the decision for adjuvant immunotherapy. The success of this project will lay the foundation for prospective validation and eventual routine use of artificial intelligence-informed biomarkers as companion diagnostics for risk prognostication and response prediction.
This project aims to optimize and validate a quantitative biomarker, Histomorphometric analysis of Tumor-Infiltrating Lymphocytes (HistoTIL), to better risk stratify survival in oral cavity cancer. We will further validate a machine learning model to predict therapeutic response to immunotherapy. Successful integration of both data science tools into the adaptive phase 1 clinical trial in oral cavity cancer will enable the precision delivery of adjuvant immunotherapy to an optimally selected group of patients, who are at high risk for disease recurrence and high probability of therapeutic response.