Abstract 2787: Evaluation of imaging-based prognostication (IPRO) using artificial intelligence (AI) in stage IV colorectal cancer (CRC) patients treated with first-line (1L) systemic therapy.

Abstract INTRODUCTION: Accurate prognostication in stage IV CRC informs treatment decisions and stratifies patients in clinical trials. Tumor, node, metastasis (TNM) staging derived from pre-treatment computed tomography (CT) imaging classifies extent of disease but may not fully characterize prognosis. IPRO-α is an AI-generated prognostic score derived from pre-treatment CT imaging, with higher scores representing improved survival. IPRO-α was trained and validated to predict survival in advanced non-small-cell lung cancer. This study evaluates generalizability and prognostic utility of IPRO-α in real-world stage IV CRC patients receiving 1L systemic therapy, compared to TNM substage. METHODS: We retrospectively evaluated IPRO-α and TNM substage in a real-world dataset of stage IV CRC patients treated with 1L systemic therapy between 2010-2018 at 17 cancer centers. We evaluated median overall survival (mOS) and hazard ratios (HR) using Cox proportional hazards models across stage IV substages (A, B, C) and matched relative distributions for IPRO-α. RESULTS: 372 patients had available pre-treatment CT and known TNM substage (IVA=141, IVB=162, IVC=69). The median age was 61 years (IQR 52-69), with 32.8% (n=122) females. TNM substage IVA mOS was significantly better than IVB, with no statistically significant difference between stage IVB and IVC (Table 1). Distribution-matched IPRO-α groups showed significant survival differences across high, intermediate and low scores. CONCLUSIONS: IPRO-α may stratify survival with greater prognostic discrimination than TNM substage in metastatic CRC patients. IPRO-α, trained on lung cancer data, learned shared prognostic features allowing generalizability of survival predictions to entirely different tumour sites. Future work will evaluate IPRO-α’s ability to stratify survival in various CRC treatment subsets. Citation Format: Mohammed A. Alvi, Ronald Bridges, Marina Salluzzi, Felipe Soares Torres, Kartik Jhaveri, Natasha B. Leighl, John Riskas, Shahid Haider, Vignesh Sivan, Oleksandra Samodorova, Jay Hennesy, Duoaud Shah, FELIX BALDAUF-LENSCHEN, Omar F. Khan. Evaluation of imaging-based prognostication (IPRO) using artificial intelligence (AI) in stage IV colorectal cancer (CRC) patients treated with first-line (1L) systemic therapy abstract. In: Proceedings of the American Association for Cancer Research Annual Meeting 2026; Part 1 (Regular Abstracts); 2026 Apr 17-22; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2026;86(7 Suppl):Abstract nr 2787.