PURPOSE We evaluated circulating tumor DNA (ctDNA) for detecting tumor burden in melanoma and examined whether early changes in the number of ctDNA copies predict response to treatment. PATIENTS AND METHODS We included 12 patients with stage III and 50 patients with stage IV melanoma with BRAF exon 15 or NRAS exon 3 mutations in tumor tissue. We used droplet digital polymerase chain reaction to retrospectively analyze serial plasma samples for mutation-positive ctDNA. RESULTS Matched plasma and serum samples were positive for ctDNA, lactate dehydrogenase, and S100 in 113 (45.8%), 108 (43.7%; not significant), and 58 (23.5%; P , .0001) of 247 samples from 50 patients with stage IV melanoma, and in 17 (63%), eight (29.6%; P = .014), and five (18.5%; P , .0001) of 27 samples from 12 patients with stage III melanoma. The number of mutant ctDNA copies correlated with concentrations of lactate dehydrogenase (r = 0.50) and S100 (r = 0.64), tumor volume (r2 = 0.58), and tumor metabolic activity (r2 = 0.83). Within 30 days before surgery, initiation of treatment, or change in treatment, ctDNA, LDH, and S100 were positive in 76.8%, 53.6% (P = .01), and 46.4% (P,.001) of patients, respectively. In patients with stage III or IV melanoma, early changes in ctDNA within 1 month after initiation of treatment correctly predicted RECIST response categories in 19 of 20 patients. Detectable ctDNA within 30 days after surgery or initiation of systemic treatment predicted inferior progression-free survival in patients with stage III disease (P = .018). In patients with stage IV disease, 10 or more copies of ctDNA per mL at first follow-up indicated shorter progression-free survival (3.8 v 9 months; hazard ratio, 4.05; 95% CI, 1.56 to 10.53). CONCLUSION ctDNA indicated active tumor and was an adverse prognostic marker for tumor progression. Dynamic changes in ctDNA allowed prediction of response early after initiation of treatment. These data support the use of ctDNA to guide treatment in melanoma.