san diego – A new AI model applied to routine pathology slides accurately predicts outcome and response to immunotherapy in patients with metastatic non-small cell lung cancer (NSCLC). This study was presented at the American Association for Cancer Research (AACR) annual meeting.
“Immunotherapy has revolutionized cancer treatment, but only some patients benefit from it, and it remains difficult to predict who will respond,” said Dr. Rukmini Bandyopadhyay, a postdoctoral fellow at the University of Texas MD Anderson Cancer Center in Utah.
“To our knowledge, this study is the first deep learning-based pathomics biomarker rigorously validated across an international real-world cohort and a phase III randomized clinical trial, directly addressing one of the most urgent unmet needs in precision oncology: reliable patient selection and stratification for immunotherapy,” he continued.
Pathomics applies computational and machine learning techniques to perform high-throughput analysis of digital pathology images to extract large-scale data related to cellular and tissue structures relevant to disease outcome.
Bandyopadhyay and colleagues developed a deep learning survival prediction model called Pathology-driven Immunotherapy Optimization (Path-IO). This can study patterns across tissues to identify patients most likely to benefit from immunotherapy. The model then combines imaging and clinical data to estimate whether a patient is at high or low risk for poor outcomes from immunotherapy.
Researchers tested the platform in a study of 797 patients with NSCLC treated with immune checkpoint inhibitors at UT MD Anderson, external validation in an additional 280 patients at Mayo Clinic, Gustave Roussy, and in the Phase III Lung MAP S1400I trial treating immunotherapy-naïve patients with immune checkpoint inhibitors in lung squamous cell carcinoma, a subtype of NSCLC.
This model reliably stratified patients into high- and low-risk groups. In the MD Anderson cohort at UT, patients had high levels of‑The risk group had more than twice the risk of death or disease progression compared to patients in the low-risk group‑risk group.
Model performance was assessed using the concordance index (C-index), which measures how well each biomarker discriminates between patients with different outcomes. Of note, Path-IO consistently outperformed PD-L1, a standard-of-care biomarker validated by the U.S. Food and Drug Administration to guide the use of immunotherapy in NSCLC patients, in both discovery and trial cohorts.
PD-L1 alone had limited prognostic results, with a C-index of 0.58 for overall survival (OS) and 0.57 for progression-free survival (PFS) in the discovery cohort, which decreased to 0.50 and 0.51, respectively, in the test cohort. In contrast, Path-IO showed stronger discriminatory ability, achieving a C-index of 0.69 for OS and 0.65 for PFS in the discovery cohort and 0.63 for OS and 0.58 for PFS in the testing cohort.
Combining pathology-based predictions with radiomics and clinical data further improved model performance, increasing the C-index from 0.58 to 0.70 for PFS and from 0.63 to 0.75 for OS.
Considering that this approach is designed to be applied to routine pathology slides, this platform can be integrated into existing clinical workflows without significant expense compared to other emerging database technologies.
As this study is retrospective, further investigation is needed to go beyond identifying patients who will benefit from immunotherapy to help predict what type of immunotherapy will benefit. FFuture directions include prospective validation and integration of paired more comprehensive molecular profiling to improve predictive performance.
