Predicting Response to Immunotherapy in NSCLC: AI-Powered PD-L1 Tumor Proportion Score Analyzer

By: Jenna Carter, PhD
Posted: Monday, August 26, 2024

An article published in the Journal of Clinical Oncology highlighted findings from a study that developed an artificial intelligence (AI)-powered analyzer to assess tumor proportion score for the prediction of response to immune checkpoint inhibitors in patients with advanced non–small cell lung cancer (NSCLC). Hyojin Kim, MD, PhD, of the Seoul National University Bundang Hospital, Seongnam, Republic of Korea, and colleagues trained and validated their AI analyzer and compared their AI tumor proportion scores with three different pathologists. Their correlation analyses revealed a significant positive correlation between tumor proportion scores from the pathologists and the AI analyzer.

A total of 393,565 tumor cells were included in this study. Cells were annotated by board-certified pathologists for PD-L1 expression in 802 whole-slide images, and images were stained by 22C3 pharmDx immunohistochemistry. The AI-powered PD-L1 analyzer scored the whole-slide images by counting PD-L1–positive tumor cells and PD-L1–negative tumor cells, and calculations were used to create grids for comparisons.

There was a significant positive correlation between the AI-generated PD-L1 tumor proportion score compared with the pathologists’ reading (Spearman coefficient = .925; P < .001). Assessments of median progression-free survival revealed that the AI-based tumor proportion score was able to predict prognosis in the 1% to 49% score group and in the < 1% score group better than the pathologists’ reading. Using the ≥ 50% tumor proportion score group as a reference, the hazard ratios (HRs) were as follows: for the 1% to 49% group: HR = 1.49 (95% confidence interval [CI] = 1.19–1.86) vs 1.36 (95% CI = 1.08–1.71); for the < 1% group: HR = 2.38 (95% CI = 1.69–3.35) vs 1.62 (95% CI = 1.23–2.13). Based on these findings, the authors concluded that their AI model may prove to be useful in accurately predicting PD-L1 tumor response and progression-free survival in patients with advanced NSCLC.

Disclosure: Dr. Kim reported no conflicts of interest. For full disclosures of the other study authors, visit coi.ascopubs.org.

Journal of Clinical Oncology

---