Prediction of clinical outcomes of immune checkpoint inhibitors in advanced head and neck cancer directly from histopathology slides using inferred transcriptomics

Introduction: Immune checkpoint inhibitors (ICI) prolong survival in advanced head and neck squamous cell carcinoma (HNSCC), yet response remains widely varied, necessitating more accurate and applicable biomarkers. We present a retrospective analysis of ENLIGHT-DP, a novel transcriptome-based biomarker applied directly on histopathology slides, in HNSCC patients treated with first-line programmed death (PD)-1 inhibitors. Methods: We retrospectively scanned high-resolution hematoxylin and eosin (H&E) slides from pre-treatment tumor-tissue samples of advanced HNSCC treated with first-line ICI and applied our ENLIGHT-DP pipeline to generate an individual prediction score. ENLIGHT-DP is composed of two steps: (i) predict individual mRNA expression directly from H&E scanned slides using DeepPT, a digital-pathology based algorithm. (ii) Use these values as input to ENLIGHT, a transcriptome-based platform that predicts response to ICI and targeted therapies. We then unblinded the clinical outcomes and evaluated the predictive value of ENLIGHT-DP in comparison to combined positive score (CPS). Results: We evaluated 25 patients with advanced HNSCC treated with first-line PD-1 inhibitors as monotherapy (15/25) or in combination with chemotherapy (10/25). In patients treated with monotherapy, ENLIGHT-DP predicts response to ICI with ROC AUC of 0.74 while CPS is not predictive. Utilizing a predetermined binary cutoff (established on independent cohorts), ENLIGHT-DP achieves 75% positive predictive value (PPV), which is superior to CPS. ENLIGHT-DP was also predictive of response to treatment in the entire patient cohort. Conclusion: ENLIGHT-DP accurately predicts response to PD-1 inhibitors treatment in HNSCC, especially in patients receiving ICI monotherapy and relying solely on easily accessible H&E scanned slides.