Pan-cancer analysis of tissue-plasma genomic concordance reveals enhanced concordance with liver metastasis and plasma clonality as predictor of poorer overall survival.

Circulating tumor DNA (ctDNA) enables noninvasive tumor genotyping, yet its concordance with tissue remains unclear. Using a 437-gene panel in 1111 pan-cancer patients, we compared somatic variants between ctDNA and matched tissues. ctDNA detection sensitivity (61.5%) correlated with advanced stage (r = 0.955, p = 0.045) and tumor size (r = 0.955, p = 0.045). Actionable alterations were detected in 49.2% (ctDNA) and 77.1% (tissue) of patients. Both shared frequently mutated genes (TP53, APC, KRAS, LRP1B, PIK3CA) and pathways (RTK-RAS, p53, DNA repair). ctDNA-specific mutations were predominantly subclonal (61.5% vs. 9.7% in tissue-concordant variants) and less frequently drivers. Machine learning linked elevated concordance to progressive disease, liver metastasis, and larger tumors. ctDNA-positivity predicted worse prognosis (HR = 2.019, p < 0.001), exacerbated by subclonal enrichment. These findings underscore ctDNA's capacity to reveal subclonality for risk stratification. While tissue remains superior for initial detection, ctDNA complements biopsies by capturing clonal heterogeneity.