Artificial Intelligence-Detected Tumor-Infiltrating Lymphocytes and Outcomes in Anti-PD-1-Based Treated Melanoma

Easy and accessible biomarkers associated with response to immune checkpoint inhibition (ICI)-treated melanoma are limited.

To evaluate artificial intelligence (AI)-detected tumor-infiltrating lymphocytes (TILs) on pretreatment melanoma metastases as a biomarker for response and survival in patients treated with ICIs.

This multicenter cohort study included patients with advanced melanoma treated with first-line anti-programmed cell death 1 protein (PD-1) with or without anti-cytotoxic T-lymphocyte-associated protein 4 (CTLA-4) between January 2016 and January 2023 at 11 melanoma treatment centers in the Netherlands. Data were analyzed from January to July 2025.

All patients received first-line anti-PD-1 with or without anti-CTLA-4.

The percentage of TILs inside manually annotated tumor area in hematoxylin-eosin-stained pretreatment metastases was determined using the Hover-NeXt model trained and evaluated on an independent melanoma dataset containing 161 835 pathologist-verified manually annotated cells. The primary outcome was objective response rate (ORR); secondary outcomes were progression-free survival (PFS) and overall survival (OS). Correlation with manual TILs, scored according to the guidelines stated by the Immuno-Oncology Biomarkers Working Group, was evaluated with Spearman correlation coefficients. Logistic regression and Cox proportional regression were conducted, adjusted for age, sex, disease stage, ICI type, BRAF status, brain metastases, lactate dehydrogenase level, and performance status.

Of 1202 included patients with advanced cutaneous melanoma, 445 (37.0%) were female and 757 (63.0%) were male, and the median (IQR) age was 67.0 (57.0-74.0) years. The median follow-up was 36.3 months (95% CI, 34.0-39.1). Metastatic melanoma specimens were available for 1202 patients, of whom 423 received combination therapy. The median (range) TIL percentage was 9.9% (0.3%-69.4%). A 10% increase in TILs was associated with increased ORR (adjusted odds ratio, 1.40; 95% CI, 1.23-1.59), PFS (adjusted hazard ratio, 0.85; 95% CI, 0.79-0.92), and OS (adjusted hazard ratio, 0.83; 95% CI, 0.76-0.91). Results were consistent for both patients treated with anti-PD-1 monotherapy and patients treated with combination treatment with anti-PD-1 plus anti-CTLA-4. When comparing manual TIL scoring with AI-detected TILs, associations with response and survival were consistently stronger for AI-detected TILs.

In this cohort study, among patients with advanced melanoma, higher levels of AI-detected TILs on pretreatment hematoxylin-eosin slides were independently associated with improved ICI response and survival. Given the accessibility of TIL scoring on routine histology, TILs may serve as a biomarker for ICI outcomes. To facilitate broader validation, the Hover-NeXt architecture and model weights are publicly available.