Preoperative prediction of futile surgery in patients with radiologically resectable or borderline resectable pancreatic adenocarcinoma

Abstract

Background and Aim The aim of this study is to identify the predictive factors for futile surgery in patients with radiologically resectable or borderline resectable pancreatic cancer and to develop a prediction model. Methods This retrospective study included patients who underwent pancreatic surgery for pancreatic cancer between 2006 and 2017. To identify independent risk factors for futile surgery, logistic regression and random forest analyses were performed in the training cohort, based on which a nomogram was established. The predictive accuracy and discriminative ability of the nomogram were validated in the validation cohort. Results Of 389 patients who underwent pancreatic surgery, the laparotomy was futile in 40 patients (10.3%). In the training cohort, the univariate and multivariate logistic regression analyses revealed that serum carbohydrate antigen 19-9 level of >= 150 U/mL (P = 0.003), the presence of suspicious lymph node (P = 0.013), and more extensive peripancreatic tumor infiltration (P < 0.001) were independent predictive factors for futile surgery. The bootstrap-corrected concordance index of the nomogram was high in the training cohort, 0.826 with a 95% confidence interval of 0.745-0.907. This model also showed a good discriminative performance in the validation cohort, with a concordance index of 0.831. Conclusions We established and validated a novel nomogram that predicts the risk of futile surgery due to occult distant metastasis in patients with radiologically resectable or borderline resectable pancreatic cancer.