Utility of circulating tumor DNA in patients with unresectable pancreatic cancer using a patient-specific panel: ARTEMIS-PC study.

684 Background: Circulating tumor DNA (ctDNA) analysis has proven to be a highly accurate method for assessing treatment efficacy and detecting molecular residual disease (MRD) in cancer patients, outperforming tumor markers and imaging modalities. However, the ctDNA detection rate is reported to be low, and its utility remains unclear in pancreatic cancer (PC). We conducted the prospective “ARTEMIS-PC” study to evaluate the utility of a personalized, tumor-informed MRD assay in patients with unresectable PC. Methods: Eligible patients had histopathologically confirmed unresectable PC, were previously untreated, and had disease measurable per Response Evaluation Criteria in Solid Tumors v1.1. A patient-specific tumor-informed assay (Invitae Personalized Cancer Monitoring) was used for the detection of ctDNA. Blood samples were collected at structured pre- and post-systemic therapy timepoints. Results: A total of 99 patients were eligible for the ARTEMIS-PC study, and personalized panels were successfully created for 92 patients. The median age of patients was 70 years (range 45-81 years), with 43/92 being female. Thirty patients were stage III, and 62 stage IV. Median size of the primary tumor was 38 mm (range 13-100 mm). MRD positivity at enrollment was 88.0%, with 73.3% in stage III patients, and 95.2% in stage Ⅳ (p=0.007). During follow-up, ctDNA clearance was achieved in 33 patients (40.7%). ctDNA clearance was associated with a significantly higher objective response rate (61.5% vs. 17.6%, p=0.001) and disease control rate (100% vs. 64.7%, p=0.002) compared to those who did not achieve clearance. Patients who achieved ctDNA clearance had significantly longer progression-free survival (PFS) than those who did not (9.0 vs. 3.5 months, hazard ratio 0.2, p<0.001). The predictive performance of select biomarkers (variant allele frequency [VAF], CEA, CA19-9) was evaluated using area under the curve (AUC) for treatment response and disease control. For disease control, VAF showed superior predictive performance with AUCs of 0.84 at enrollment and 0.97 at both Week 4 and Week 8, compared to CEA (AUCs: 0.65, 0.73, 0.60) and CA19-9 (AUCs: 0.51, 0.57, 0.56). The predictive performance for treatment response was lower than AUC 0.75 for any of the biomarkers. Among patients who were MRD negative before treatment or became negative during follow-up (n=44), 45.5% (n=20) subsequently turned MRD positive. Of these, disease progression was confirmed by CT in 14 patients. The median time from MRD turning positive to CT-confirmed progressive disease was 88.5 days (range: -28 to 385 days). Conclusions: ctDNA clearance is a useful predictor of improved objective response, disease control, and PFS, with VAF emerging as the more accurate biomarker for disease control, underscoring its potential utility in treatment monitoring in patients with unresectable PC. Clinical trial information: UMIN000043561 .