Enhanced patient selection with quantitative continuous scoring of PD-L1 expression for IO treatment in metastatic NSCLC
Year: 2025
Type: article
Abstract: Immune checkpoint inhibitors targeting PD-1/PD-L1 are approved for metastatic non-small-cell lung cancer treatment. In clinical practice, the treatment choice depends on visual scoring of PD-L1, which is subjective and semi-quantitative. In this work, we present PD-L1 Quantitative Continuous Scoring (PD-L1 QCS), a computer vision system for granular cell-level quantification of PD-L1 staining intensity in digitized whole slide images (WSI). We derived a biomarker which captures the percentage of tumor cells (TC) with medium to strong staining intensity (PD-L1 QCS-PMSTC) and classifies patients with ≥0.575% as biomarker positive (BM+). Its effectiveness was examined against visual scoring of %TC ≥ 50 in 768 WSI from the MYSTIC trial (NCT02453282). Considering anti-PD-L1 treatment (N = 256) vs. chemotherapy (N = 246), visual scoring resulted in a hazard ratio (HR) of 0.69 (CI 0.46-1.02) with a 29.7% prevalence of the BM+ group. With PD-L1 QCS-PMSTC, a similar HR of 0.62 (CI 0.46-0.82) with an increased prevalence of 54.3% was obtained.
Source: npj Precision Oncology
Authors: J. Lesniak, M. Schick, T. Kunzke, F. Pollastri, J. P. Vigueras-Guillén, H. Hessel, S. Haneder, P. Sontakke, K. DaCosta, R. Alleze, J. Zimmermann, A. Kapil, N. Brieu, A. Shumilov, H. Sade, J. C. Barrett, G. Schmidt, R. Stewart
Language: English
Topic: AI in cancer detection
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: gold
Estimated APC paid: $4,090
Funder: AstraZeneca
OpenAlex ID: W4414308186
