Rational use of adjuvant anti-PD-1: Multi-omics model of recurrence in stage III melanoma.

9569 Background: Stage III melanoma patients (pts) undergoing adjuvant anti-PD-1 immunotherapy have a high risk of recurrence (43% within one year) and treatment-related adverse events (25% severe). This study developed multi-omics models that accurately identify pts at high risk of recurrence who may benefit from alternative treatment strategies or closer surveillance. Methods: We analyzed a cohort of 131 pts with stage III melanoma (47% IIIA/B and 53% IIIC/D) who received adjuvant anti-PD-1 therapy. The nodal burden was micrometastatic (35%) and macrometastatic (54%), with in-transit metastases in 11% of pts. Comprehensive multi-omics profiling included DNA sequencing (tumor mutational burden [TMB]), whole-transcriptome sequencing (gene expression profiling [GEP]), and multiplex immunohistochemistry (tumor microenvironment [TME]) of the baseline tumor sample. We developed predictive models for 12-month recurrence using multivariable penalized logistic regression with consensus-nested cross-validation incorporating clinical, TMB, GEP, and TME features. Internal validation was performed using optimism bias through 500 bootstrap iterations. Results: Clinical factors (nodal burden, stage, and site of primary melanoma) alone achieved a modest AUC of 0.66 (95% CI: 0.56-0.75) for predicting recurrence. Addition of TME features, particularly CD16+ cells interacting with PD-L1+CD16+ macrophages, significantly improved predictive accuracy (AUC: 0.81, 95% CI: 0.72-0.91). Further enhancements were observed with TMB and BRAF mutation status (AUC: 0.83, 95% CI: 0.74-0.92) and GEP-derived natural killer (NK) cell and interferon-gamma (IFNg) signatures (AUC: 0.83, 95% CI: 0.73-0.93). A consensus model integrating these features achieved an optimal AUC of 0.86 (95% CI: 0.78-0.94). This model demonstrated robust performance across macroscopic (AUC: 0.88, 95% CI: 0.78-0.98) and microscopic (AUC: 0.86, 95% CI: 0.70-1.00) nodal diseases. Conclusions: This study demonstrates the potential of multi-omics profiling to significantly enhance recurrence risk prediction in stage III melanoma pts receiving adjuvant anti-PD-1 therapy. We developed a robust model with high predictive accuracy by integrating clinical data with TME, TMB, and GEP features (AUC, 0.86). This model can help identify pts at high risk of recurrence who may benefit from alternative treatment strategies or closer surveillance. AUC scores of various models. Model AUC Clinical 0.66 Clinical+TME 0.81 Clinical+TMB 0.83 Clinical+GEP 0.83 Consensus Model 0.86