Background: The advent of effective adjuvant therapies for patients with resected melanoma has highlighted the need to stratify patients based on risk of relapse given the cost and toxicities associated with treatment. Here we assessed circulating tumor DNA (ctDNA) to predict and monitor relapse in resected stage III melanoma.Patients and methods: Somatic mutations were identified in 99/133 (74%) patients through tumor tissue sequencing. Personalized droplet digital PCR (ddPCR) assays were used to detect known mutations in 315 prospectively collected plasma samples from mutation-positive patients. External validation was performed in a prospective independent cohort (n ¼ 29).Results: ctDNA was detected in 37 of 99 (37%) individuals. In 81 patients who did not receive adjuvant therapy, 90% of patients with ctDNA detected at baseline and 100% of patients with ctDNA detected at the postoperative timepoint relapsed at a median follow up of 20 months. ctDNA detection predicted patients at high risk of relapse at baseline [relapse-free survival (RFS) hazard ratio (HR) 2.9; 95% confidence interval (CI) 1.5-5.6; P ¼ 0.002] and postoperatively (HR 10; 95% CI 4.3-24; P < 0.001). ctDNA detection at baseline [HR 2.9; 95% CI 1.3-5.7; P ¼ 0.003 and postoperatively (HR 11; 95% CI 4.3-27; P < 0.001] was also associated with inferior distant metastasisfree survival (DMFS). These findings were validated in the independent cohort. ctDNA detection remained an independent predictor of RFS and DMFS in multivariate analyses after adjustment for disease stage and BRAF mutation status. Conclusion:Baseline and postoperative ctDNA detection in two independent prospective cohorts identified stage III melanoma patients at highest risk of relapse and has potential to inform adjuvant therapy decisions.

No abstract