AI Pathomics Tool Beats PD-L1 at Predicting Lung Cancer Immunotherapy Benefit

Overview

• Path-IO, presented at the AACR Annual Meeting in April 2026, accurately predicted outcomes for patients with metastatic non‑small cell lung cancer treated with immune checkpoint drugs.
• The team trained on 797 cases from UT MD Anderson and validated on about 280 more from Mayo Clinic, Gustave Roussy, and the phase III Lung‑MAP S1400I trial, bringing the total test population to over 1,000 patients.
• Measured by the C‑index, a gauge of how well a marker separates patient outcomes, Path‑IO reached up to 0.69 for overall survival and 0.65 for progression‑free survival, compared with PD‑L1 often near 0.58 and at times near chance.
• Blending Path‑IO’s slide‑based predictions with CT radiomics and clinical variables raised discrimination to a C‑index of 0.70 for progression‑free survival and 0.75 for overall survival.
• The study was retrospective, so the team is preparing prospective validation and deeper molecular integration, and the slide‑based, biology‑guided design aims for explainable use in routine workflows if later trials confirm benefit.