Circulating tumor DNA (ctDNA) holds promise for guiding immune checkpoint inhibitor (ICI) therapy and stratifying responders from non-responders. While tumor-informed ctDNA detection approaches are sensitive and mutation-inclusive, they require tumor tissue, which limits applicability in real-world settings. Conversely, tumor-agnostic methods often have limited genomic coverage. In this study, we evaluated a tumor-agnostic, broad-panel ctDNA assay in patients with advanced melanoma treated with ICI.

We conducted a prospective analysis of 241 longitudinal samples from 39 patients with unresectable stage III/IV melanoma using a SYSMEX targeted NGS panel covering 1,114 COSMIC mutations. Plasma samples were collected at baseline and during ICI therapy. The assay’s sensitivity reached seven mutant molecules, corresponding to a 0.07% mutation allele frequency (MAF). ctDNA profiles were compared with matched tumor tissue and correlated with clinical features and survival.

At baseline, ctDNA was detected in 64.5% of patients. Common mutations included BRAF^V600E (43.8%) and NRAS^G12D (36.4%), followed by KRAS, EGFR, and PIK3CA variants. Overall tissue–plasma concordance was 51.6%, with more extended biopsy–plasma intervals associated with discordance (p = 0.0105). Notably, 12.2% of cases exhibited partial concordance, characterized by shared mutations and additional plasma-only alterations, underscoring the complementary value of blood-based profiling. Persistent or re-emerging ctDNA positivity post-therapy correlated with shorter progression-free survival (PFS, p = 0.003), while ctDNA-negative patients showed significantly improved outcomes. Patients that remained ctDNA-negative had significantly longer progression-free survival (median not reached) compared to those with persistent ctDNA positivity (median 3 months) or those converting to positive (median 7.5 months; p = 0.0073). Early NRAS and KRAS ctDNA levels strongly predicted poor response (p = 0.0069 and p = 0.028). The prognostic impact extended beyond canonical drivers, as non-hotspot variants also correlated with the outcome. Notably, even low-level ctDNA persistence (5–10 MM/mL) carried adverse prognostic implications (p = 0.0054). Concerning a shorter PFS, ctDNA positivity was also associated with elevated S100 levels (p = 0.047). Organ-specific mutation enrichment (e.g., KRAS^G12D in brain, EGFR^G719A in lymph nodes) suggested possible metastatic tropism.

Broad tumor-agnostic ctDNA analysis effectively identified clinically relevant mutations and predicted outcomes in ICI-treated melanoma patients. This approach enables tissue-independent and real-time ctDNA monitoring and may inform patient selection and therapeutic strategies in future interventional trials.