Personalized ctDNA monitoring to assess the recurrence of early stage pancreatic cancer after surgery.

774 Background: Up to 80% of patients with resected pancreatic cancer recur within 2 years. Here, we assessed the feasibility and accuracy of a personalized, tumor-informed circulating tumor DNA (ctDNA) test for early warning of recurrence risk during long-term surveillance. Methods: This study retrospectively analyzed 43 patients who underwent radical surgery for pancreatic cancer between November 2021 and June 2023 at the Comprehensive Cancer Centre of Drum Tower Hospital, Clinical Cancer Institute of Nanjing University. A personalised panel was constructed to detect ctDNA in plasma based on whole-exon mutation information from tumour tissue. A total of 139 ctDNA samples were obtained from periods including postoperative, during adjuvant therapy, and post-treatment. Results: Until the last follow-up time (May 13, 2024), a personalised ctDNA monitoring panel was successfully constructed for 35 (81.4%) stage I-III patients with a median follow-up of 11.82 months (range: 2-17.8 months). A total of 16 patients experienced recurrence within 15.7 months (range: 5.4-30 months), with a median lead time from first positive ctDNA to imaging recurrence of 3.6 months (range: 1-9.3 months). ctDNA positivity was detected in 25.8% (8/31) of ctDNAs at the landmark time point (postoperative prior to adjuvant therapy), with a positive predictive value (PPV) of 75% (6/8) and a negative predictive value (NPV) of 69.6% (16/23). Patients with positive ctDNA at the landmark time point had inferior DFS (6.9 months vs undefined. HR 3.955, P=0.007) and OS (15 months vs undefined. HR 5.599, P=0.001) compared to those with negative ctDNA. When integrating longitudinal time points, PPV and NPV were 68.2% (15/22) and 91.7% (11/12), respectively. For all 16 relapsed patients, the sensitivity of positive ctDNA detected longitudinally was 93.75% (15/16). The ctDNA-MRD status at longitudinal time points had a more significant difference with DFS (undefined vs 12.6 months. HR 5.096, P=0.002) and OS (undefined vs 21 months. HR 6.487, P=0.038). Conclusions: Personalized ctDNA surveillance is effective in identifying high recurrence risk after pancreatic surgical resection. Long-term longitudinal warning of arbitrary ctDNA positivity suggests a recurrence probability of up to 93.75%; therefore, full ctDNA dynamic monitoring is essential. And ctDNA monitoring will be further used to guide the decision-making in the clinic to prolong patient survival.