Background: Adjuvant dabrafenib plus trametinib reduced the risk of relapse versus placebo in patients with resected, BRAFV600-mutant, stage III melanoma in the phase 3 COMBI-AD trial. This prespecified exploratory biomarker analysis aimed to evaluate potential prognostic or predictive factors and mechanisms of resistance to adjuvant targeted therapy. Methods: COMBI-AD is a randomised, double-blind, placebo-controlled, phase 3 trial comparing dabrafenib 150 mg orally twice daily plus trametinib 2 mg orally once daily versus two matched placebos. Study participants were at least 18 years of age and underwent complete resection of stage IIIA (lymph node metastases >1 mm), IIIB, or IIIC cutaneous melanoma, per American Joint Committee on Cancer 7th edition criteria, with a BRAFV600E or BRAFV600K mutation. Patients were randomly assigned (1:1) to the two treatment groups by an interactive voice response system, stratified by mutation type and disease stage. Patients, physicians, and the investigators who analysed data were masked to treatment allocation. The primary outcome was relapse-free survival, defined as the time from randomisation to disease recurrence or death from any cause. Biomarker assessment was a prespecified exploratory outcome of the trial. We assessed intrinsic tumour genomic features by use of next-generation DNA sequencing and characteristics of the tumour microenvironment by use of a NanoString RNA assay, which might provide prognostic and predictive information. This trial is registered with ClinicalTrials.gov, number NCT01682083, and is ongoing but no longer recruiting participants. Findings: Between Jan 31, 2013, and Dec 11, 2014, 870 patients were enrolled in the trial. Median follow-up at data cutoff (April 30, 2018) was 44 months (IQR 38–49) in the dabrafenib plus trametinib group and 42 months (21–49) in the placebo group. Intrinsic tumour genomic features were assessed in 368 patients (DNA sequencing set) and tumour microenvironment characteristics were assessed in 507 patients (NanoString biomarker set). MAPK pathway genomic alterations at baseline did not affect treatment benefit or clinical outcome. An IFNγ gene expression signature higher than the median was prognostic for prolonged relapse-free survival in both treatment groups. Tumour mutational burden was independently prognostic for relapse-free survival in the placebo group (high TMB, top third; hazard ratio [HR] 0·56, 95% CI 0·37–0·85, p=0·0056), but not in the dabrafenib plus trametinib group (0·83, 95% CI 0·53–1·32, p=0·44). Patients with tumour mutational burden in the lower two terciles seem to derive a substantial long-term relapse-free survival benefit from targeted therapy (HR [versus placebo] 0·49, 95% CI 0·35–0·68, p<0·0001). However, patients with high tumour mutational burden seem to have a less pronounced benefit with targeted therapy (HR [versus placebo] 0·75, 95% CI 0·44–1·26, p=0·27), especially if they had an IFNγ signature lower than the median (HR 0·88 [95% CI 0·40–1·93], p=0·74). Interpretation: Tumour mutational burden alone or in combination with IFNγ gene expression signature or other markers for an adaptive immune response might be of relevance for identifying patients with stage III melanoma who might derive clinical benefit from targeted therapy. Further validation in prospective clinical trials is warranted. Funding: Novartis Pharmaceuticals.