RESUMEN
PURPOSE: Pancreatic ductal adenocarcinoma (PDAC) is associated with a poor prognosis. Multianalyte signatures, including liquid biopsy and traditional clinical variables, have shown promise for improving prognostication in other solid tumors but have not yet been rigorously assessed for PDAC. MATERIALS AND METHODS: We performed a prospective cohort study of patients with newly diagnosed locally advanced pancreatic cancer (LAPC) or metastatic PDAC (mPDAC) who were planned to undergo systemic therapy. We collected peripheral blood before systemic therapy and assessed circulating tumor cells (CTCs), cell-free DNA concentration (cfDNA), and circulating tumor KRAS (ctKRAS)-variant allele fraction (VAF). Association of variables with overall survival (OS) was assessed in univariate and multivariate survival analysis, and comparisons were made between models containing liquid biopsy variables combined with traditional clinical prognostic variables versus models containing traditional clinical prognostic variables alone. RESULTS: One hundred four patients, 40 with LAPC and 64 with mPDAC, were enrolled. CTCs, cfDNA concentration, and ctKRAS VAF were all significantly higher in patients with mPDAC than patients with LAPC. ctKRAS VAF (cube root; 0.05 unit increments; hazard ratio, 1.11; 95% CI, 1.03 to 1.21; P = .01), and CTCs ≥ 1/mL (hazard ratio, 2.22; 95% CI, 1.34 to 3.69; P = .002) were significantly associated with worse OS in multivariate analysis while cfDNA concentration was not. A model selected by backward selection containing traditional clinical variables plus liquid biopsy variables had better discrimination of OS compared with a model containing traditional clinical variables alone (optimism-corrected Harrell's C-statistic 0.725 v 0.681). CONCLUSION: A multianalyte prognostic signature containing CTCs, ctKRAS, and cfDNA concentration outperformed a model containing traditional clinical variables alone suggesting that CTCs, ctKRAS, and cfDNA provide prognostic information complementary to traditional clinical variables in advanced PDAC.