Development and validation of a new model for predicting malignant pancreatic cystic lesions based on clinical and EUS characteristics.

Abstract

Objective: This study aimed to develop and validate a predictive model integrating clinical and Endoscopic Ultrasound (EUS) morphological characteristics to improve the diagnosis of malignant pancreatic cystic lesions (PCLs) in radiologically challenging cases.

Methods: A retrospective analysis was conducted on patients with pathologically confirmed PCLs at The First Affiliated Hospital of Naval Medical University between January 2006 and June 2023. Model performance was evaluated using the area under the receiver operating characteristic curve (AUC) and decision curve analysis (DCA).

Results: Among 126 enrolled patients, multivariate logistic regression identified six independent predictors of malignancy: weight loss (OR:29.118, 95% CI: 1.894-1031.096, p = 0.010), serum CA19-9 level > 29.3 U/ml (OR: 24.601, 95% CI:7.305-110.886, p < 0.001), location of head-neck (OR: 3.539, 95% CI:1.106-12.857, p = 0.033), thickened cystic wall (OR:3.605, 95% CI:1.106-12.857, p = 0.033), solid component (OR:7.124, 95% CI: 2.113-30.399, p = 0.001) and peripancreatic invasion (OR:8.749, 95% CI: 1.367-78.268, p = 0.021). The model demonstrated high discriminative ability, with an AUC of 0.938 in the training set and 0.951 in the validation set. DCA demonstrated a net benefit (NB) of 0.224 at 10-80% threshold probabilities, potentially avoiding 10% unnecessary surgeries.

Conclusions: This EUS-based nomogram improves malignancy prediction in radiologically challenging PCLs, offering clinical utility for personalized management. External validation is warranted to confirm its generalizability.