Molecular profiling to predict the first site of recurrence in pancreatic cancer patients receiving neoadjuvant therapy followed by surgery

Pancreatic ductal adenocarcinoma (PDAC) remains one of the most lethal cancers, with a five-year survival rate of only 13%. Although neoadjuvant chemotherapy and radiotherapy have improved resectability and survival outcomes, most patients ultimately experience disease recurrence. Identifying biomarkers that can stratify recurrence patterns is therefore critical for optimizing patient selection and tailoring postoperative management. We retrospectively analyzed data from PDAC patients who underwent neoadjuvant therapy followed by surgery at Cedars Sinai Medical Center between May 2013 and November 2024. Uni- and multivariable binary logistic regression including clinicopathologic variables and NGS-based molecular features was used to predict the first site of recurrence. Among 244 patients receiving neoadjuvant therapy for PDAC, 138 had complete clinicopathologic and NGS-based molecular data. Most common mutated genes included TP53 (69.6%), KRAS G12D (43.5%), KRAS G12V (31.2%), CDKN2A (24.6%), SMAD4 (20.3%), KRAS G12R (12.3%), CDKN2B (10.1%), KRAS Q61X (7.2%), ATM (5.1%), and BRCA2 (4.3%). First-site recurrence patterns were locoregional (25.5%), liver (32.6%), lung (21.0%), peritoneum (14.5%), and bone (7.2%). Molecular features independently and significantly predicted the first site of recurrence. Specifically, KRAS Q61X mutation was associated with increased risk of locoregional recurrence (OR, 5.6, 95% CI 1.4-22.4, p=0.01), CDKN2B mutations with liver recurrence (OR, 4.4, 95% CI 1.2-16.2, p=0.02), ATM mutations with both lung (OR, 7.7, 95% CI 1.3-44.9, p=0.02) and peritoneal recurrence (OR, 5.0, 95% CI 1.0-24.4, p=0.04), and BRCA2 mutations with bone recurrence (OR, 7.7, 95% CI 1.2-48.9, p=0.03). Our study demonstrates that molecular profiling can predict the first site of recurrence in PDAC patients who have undergone neoadjuvant therapy followed by surgery. While prospective validation in larger cohorts is warranted, this work provides a foundation for developing risk-adapted postoperative surveillance strategies in PDAC.