Dynamics of Serum CA19-9 in Patients Undergoing Pancreatic Cancer Resection

Carbohydrate antigen (CA) 19-9 is an established perioperative prognostic biomarker for pancreatic ductal adenocarcinoma (PDAC). However, it is unclear how CA19-9 monitoring should be used during postoperative surveillance to detect recurrence and to guide the initiation of recurrence-focused therapy. This study aimed to elucidate the value of CA19-9 as a diagnostic biomarker for disease recurrence in patients who underwent PDAC resection. Serum CA19-9 levels at diagnosis, after surgery, and during postoperative follow-up were analyzed in patients who underwent PDAC resection. All patients with at least two postoperative follow-up CA19-9 measurements before recurrence were included. Patients deemed to be nonsecretors of CA19-9 were excluded. The relative increase in postoperative CA19-9 was calculated for each patient by dividing the maximum postoperative CA19-9 value by the first postoperative value. Receiver operating characteristic analysis was performed to identify the optimal threshold for the relative increase in CA19-9 levels to identify recurrence in the training set using Youden's index. The performance of this cutoff was validated in a test set by calculating the area under the curve (AUC) and was compared to the performance of the optimal cutoff for postoperative CA19-9 measurements as a continuous value. In addition, sensitivity, specificity, and predictive values were assessed. In total, 271 patients were included, of whom 208 (77%) developed recurrence. Receiver operating characteristic analysis demonstrated that a relative increase in postoperative serum CA19-9 of 2.6× was predictive of recurrence, with 58% sensitivity, 83% specificity, 95% positive predictive value, and 28% negative predictive value. The AUC for a 2.6× relative increase in the CA19-9 level was 0.719 in the training set and 0.663 in the test set. The AUC of postoperative CA19-9 as a continuous value (optimal threshold, 52) was 0.671 in the training set. In the training set, the detection of a 2.6-fold increase in CA19-9 preceded the detection of recurrence by a mean difference of 7 months ( P