Comprehensive Summary
This study investigates whether AI-based detection of tumor-infiltrating lymphocytes (AI-TILs) can predict treatment outcomes in patients with metastatic melanoma receiving first-line anti-PD-1 immunotherapy. The researchers applied an open source AI model (Hover-NeXt) to determine the percentage of lymphocytes on standard H&E slides, using pretreatment metastatic tumor samples from 1202 patients across multiple Dutch centers. The researchers compared AI-TIL levels with objective response, progression-free survival, and OS. The results indicated that higher AI-TIL percentages were associated with improved treatment response and longer PFS and OS. Compared to manually scored TIL, AI-TILs showed stronger correlations with outcomes, proving their potential as a biomarker. However, the authors note that AI-TILs could complement, but not replace existing pathology assessments.
Outcomes and Implications
AI-based analysis of tumor immune infiltration could become a practical and less costly tool to guide melanoma treatment decisions. Counting TILs could help clinicians identify which patients would benefit from anti-PD-1 therapy and could help in determining personalized treatments plans. This approach could improve prognostic accuracy and reduce unnecessary combination therapies for the future.