Background: Tebentafusp, a bispecific (gp100 × CD3) ImmTAC showed significant overall survival (OS) benefit (HR 0.51) versus investigator’s choice in a Phase (Ph) 3 trial in first line (1L) HLA-A*02:01+ adult patients (pts) with metastatic uveal melanoma (mUM) [Nathan NEJM 2021]. In Ph2 and Ph3 trials of tebentafusp, OS was improved regardless of RECIST best response and the degree of early reduction in ctDNA was a better predictor of OS [Carvajal Nat Med 2022; Sullivan AACR 2023]. Here we explored whether early on-treatment reduction in ctDNA could distinguish pts with stable disease (SD) and long OS vs those with SD and short OS in a Ph3 trial of tebentafusp in 1L mUM pts. Methods: 1L HLA-A*02:01+ pts with mUM received 68 mcg tebentafusp weekly intravenously after intra-patient dose escalation (NCT03070392). Response was assessed by investigators per RECISTv1.1. Sera collected at baseline (BL) and week 9 on tebentafusp were analyzed for ctDNA using targeted mPCR-NGS assay for mutations in 15 genes including GNAQ, GNA11, SF3B1, CYSLTR2, PLCB4 and EIF1AX. Landmark OS from week 9 was analyzed in subsets with ≥0-3 log reductions in mean tumor molecules (MTM) per ml of serum on treatment. Data cutoff 11-Nov-2022. Results: 76/245 (36%) tebentafusp-treated pts had best response of SD, with a median OS of 29 months and were evaluable for ctDNA. Baseline tumor burden as assessed by sum of longest lesion diameters and percentage of pts with elevated BL serum LDH were greater in the subset with < 1 year (yr) OS (median 70 mm and 64%, respectively) than in subsets with OS ≥ 1 yr (median 42 mm, 11%) or OS ≥ 2 yrs (median 31 mm, 9%). Almost half (36/76; 47%) of all SD pts had detectable ctDNA mutations in one or more UM genes at baseline. Baseline ctDNA levels were similar across survival groups. By week 9 on-treatment, ctDNA reduction was observed in 34/36 (94%) evaluable SD pts, including 29/36 (81%) with ≥ 0.5 log reduction, 18/36 (50%) with ≥ 2 log reduction and 16/36 (44%) with undetectable ctDNA (clearance). Deeper reductions in ctDNA were associated with better OS (Table). Conclusions: In this Ph3 trial, ctDNA reduction by week 9 on tebentafusp was strongly associated with improved OS in pts with best RECIST SD. Early ctDNA reduction may predict SD patients with long OS on tebentafusp.

*HRs estimated from Kaplan-Meier analysis for subsets above vs below ctDNA reduction threshold.