Nomogram model for predicting early recurrence for resectable pancreatic cancer：A multicenter study

Abstract

Abstract This study aimed toestablish an accurate preoperative nomogram model for predicting early recurrence (ER) for resectable pancreatic adenocarcinoma. We retrospectively analyzed patients who underwent pancreatectomy for pancreatic ductal adenocarcinoma between January 2011 and December 2020. The training set consisted of 604 patients, while the validation set included 222 patients.Survival was estimated using Kaplan–Meier curves. The factors influencing early recurrence of resectable pancreatic cancer after surgery were investigated, then the predictive model for early recurrence was established, and subsequently the predictive model was validated based on the data of the validation group. The preoperative risk factors for ER included a Charlson age-comorbidity index ≥4 (odds ratio [OR]: 0.628), tumor size >3.0 cm on computed tomography (OR: 0.628), presence of clinical symptoms (OR: 0.515), carbohydrate antigen 19-9 >181.3 U/mL (OR 0.396), and carcinoembryonic antigen >6.01 (OR: 0.440). The area under the curve (AUC) of the predictive model in the training group was 0.711 (95% confidence interval: 0.669–0.752), while it reached 0.730 (95% CI: 0.663–0.797) in the validation group. The predictive model may enable the prediction of the risk of postoperative ER in patients with resectable pancreatic ductal adenocarcinoma, thereby optimizing preoperative decision-making for effective treatment.