Modality-Specific Prognostic and Predictive Factors in Metastatic Uveal Melanoma

Introduction Treatment selection in metastatic uveal melanoma (mUM) among immune checkpoint blockade (ICB), liver-directed therapy (LDT), and tebentafusp is largely empiric, and modality-specific biomarkers are lacking. We aimed to identify modality-specific prognostic and predictive factors to inform individualized selection. Methods Retrospective cohort of 192 adults with mUM at a tertiary institution (01/2010-03/2026), contributing 500 treatment courses (194 ICB, 119 LDT, 35 tebentafusp, 27 ICB+LDT). Modality-specific bootstrap validated LASSO-Cox models and pooled gamma-frailty Cox models assessed overall survival (OS), time to next treatment (TTNT), and treatment-free interval (TFI); response rate and adverse events (AEs) were analyzed with logistic and mixed-effects models. A treatment-by-driver-oncogene interaction tested GNA11 versus GNAQ. Results Each modality had a distinct prognostic profile: ICB driven by performance status and lactate dehydrogenase (LDH); LDT by alkaline phosphatase, LDH, and time to recurrence; tebentafusp by alkaline phosphatase alone. No modality independently predicted OS, but tebentafusp extended TTNT (HR=0.29; 95% CI 0.18-0.47) and LDT extended TFI (HR=0.50; 95% CI 0.38-0.68). GNA11 mutation was predictive for LDT (interaction p=0.004; adjusted HR=0.12; 95% CI 0.03-0.58) but not ICB. Combination ICB showed no efficacy advantage over anti-PD-1 monotherapy but greater grade ≥3 toxicity (odds ratio 5.60; P=0.003). Discussion Disease biology, rather than modality, is the primary survival determinant, yet each modality carries a distinct prognostic profile. GNA11 is a candidate predictive biomarker for LDT-the first modality-specific genomic biomarker in mUM. Conclusions Modality-specific profiling may enable personalized selection. Combination ICB should not be the default given harm without benefit; prospective GNA11-LDT validation is the priority.