Perioperative CA19-9 and CEA Trajectories After Pancreatic Cancer Resection

BACKGROUND: The prognostic utility of CA19-9 may vary based on the clinical and genetic factors (e.g., FUT2/FUT3 mutation or UGT1A1 alleles). We sought to examine early post-operative changes in CA19-9 relative to prognosis following resection of PDAC. METHODS: Patients undergoing resection of PDAC were identified from the All of Us Research Program. Patients were categorized into neoadjuvant therapy followed by resection (NAT→resection) and surgery-first. The primary endpoint was real-world time-to-treatment-failure (rwTTF). Genomic and sensitivity analyses excluded post-treatment draws with total bilirubin >2mg/dL and restricted to CA19-9 producers. RESULTS: Among 283 resected patients, 48 underwent NAT→resection and 235 underwent surgery-first. Index years ranged from 1999 to 2023. Baseline CA19-9 distributions differed markedly by pathway: NAT→resection patients had median pre-treatment CA19-9 273U/mL with 10.0% normal values, whereas surgery-first patients had median pre-treatment CA19-9 19U/mL with 68.7% normal values. CEA was available pre-treatment in 88 patients and post-operatively in 41 patients; only 1 of 8 phenotypic CA19-9 non-producers had elevated CEA. In CA19-9 trajectory analysis, median rwTTF was 1185.5 days for normal→normal, 693 days for elevated→normalized, and 138 days for elevated→persistently elevated patients. After excluding planned adjuvant therapy from the rwTTF endpoint, higher pre-treatment CA19-9 and greater post-operative CA19-9 change were independently associated with rwTTF, whereas binary post-operative normalization alone was not. CONCLUSIONS: Basing prognosis on a single CA19-9 value or percent drop may be insufficient, particularly because many surgery-first patients had normal baseline CA19-9. Perioperative CA19-9 trajectories provided a more clinically informative framework, with persistent postoperative elevation identifying the highest-risk group after PDAC resection.