Evaluation of <i>KRAS</i> -mutant circulating tumor DNA as a surgical biomarker in localized pancreatic cancer treated with multi-agent neoadjuvant chemotherapy.

Abstract

e16471 Background: Circulating tumor DNA (ctDNA) is a promising biomarker for risk stratification in localized pancreatic ductal adenocarcinoma (PDAC), but its ability to predict surgical resection and overall survival (OS) after multi-agent neoadjuvant chemotherapy (NAC) is not well defined. High-sensitivity digital droplet PCR (ddPCR) enables detection of low-frequency KRAS mutations in plasma with greater sensitivity than standard approaches, allowing noninvasive, dynamic assessment of tumor burden. We evaluated whether KRAS -mutant ctDNA detected by ddPCR predicts surgery and OS in localized PDAC. Methods: Patients with localized PDAC were enrolled in a prospective study at an academic health system from October 2020 to July 2025. Peripheral blood was collected at diagnosis, after NAC, and after surgery. Plasma was isolated for cell-free DNA extraction and probed for KRAS G12D, G12V, and G12R ctDNA using ddPCR. Multivariable logistic regression assessed the association between ctDNA detection and receipt of resection, adjusting for baseline anatomic resectability and receipt of NAC, and OS was evaluated by Kaplan-Meier. Results: This study included 247 blood samples from 123 patients, with a median follow-up of 18.6 months. Fifty-nine patients received FOLFIRINOX, 35 gemcitabine-based therapy, and 9 both; 63 underwent surgery after a median of 2.5 months of NAC. KRAS ctDNA detection rates were 66.3% at diagnosis, 61.3% post-NAC, and 56.4% post-surgery. Detection of KRAS ctDNA at diagnosis was associated with lower odds of surgery, independent of baseline resectability status (aOR 0.29, CI 0.08–0.82, p = 0.028). Among FOLFIRINOX-treated patients who underwent surgery, post-surgery KRAS ctDNA detection was associated with shorter OS (median OS [mOS] 30.2 vs not reached [NR], p = 0.012), whereas no association was observed in gemcitabine-treated patients (mOS 24.2 vs 37.4 months, p = 0.567). Among 30 patients completing adjuvant chemotherapy (AC), ctDNA detection after FOLFIRINOX and surgery remained associated with inferior OS (30.2 months vs NR, p = 0.023), with ctDNA clearance after AC observed in only a third of patients after a median of 2.7 months of treatment. Conclusions: In this prospective cohort of localized PDAC, the prognostic value of KRAS -mutant ctDNA detected by ddPCR varied by NAC regimen. KRAS ctDNA detection at diagnosis predicted lower odds of surgery after NAC, regardless of resectability status, consistent with poor tumor biology. The prognostic impact of post-surgery KRAS ctDNA differed by NAC regimen, with inferior OS among ctDNA-positive patients treated with FOLFIRINOX. The use of AC did not change prognosis, suggesting ddPCR identifies patients with persistent KRAS ctDNA after NAC who are unlikely to benefit from additional chemotherapy and may be better served by treatment with novel KRAS -targeted agents.