A Nomogram-based Model to Predict Neoplastic Risk for Patients with Gallbladder Polyps

Abstract

Background and aims: Gallbladder polyp (GBP) assessment aims to identify the early stages of gallbladder carcinoma. Many studies have analyzed the risk factors for malignant GBPs. In this retrospective study, we aimed to establish a more accurate predictive model for potential neoplastic polyps in patients with GBPs.

Methods: We developed a nomogram-based model in a training cohort of 233 GBP patients. Clinical information, ultrasonographic findings, and blood test findings were analyzed. Mann-Whitney U test and multivariate logistic regression analyses were used to identify independent predictors and establish the nomogram model. An internal validation was conducted in 225 consecutive patients. Performance and clinical benefit of the model were evaluated using receiver operating characteristic curves and decision curve analysis (DCA), respectively.

Results: Age, cholelithiasis, carcinoembryonic antigen, polyp size, and sessile shape were confirmed as independent predictors of GBP neoplastic potential in the training group. Compared with five other proposed prediction methods, the established nomogram model presented better discrimination of neoplastic GBPs in the training cohort (area under the curve [AUC]: 0.846) and the validation cohort (AUC: 0.835). DCA demonstrated that the greatest clinical benefit was provided by the nomogram compared with the other five methods.

Conclusions: Our developed preoperative nomogram model can successfully be used to evaluate the neoplastic potential of GBPs based on simple clinical variables that maybe useful for clinical decision-making.

Keywords: Gallbladder polyps; Neoplastic polyp; Nomogram model; Preoperative diagnosis.

Fig. 1. Patient selection flowchart.

GBPs, gallbladder polyps; GBC, gallbladder carcinoma.

Fig. 2. Developed nomogram presented with ROC.

(A) The nomogram was established due to the training cohort, with age, cholelithiasis, CEA, polyp size and sessile incorporated. (B) Comparison of ROC curves between our model, US-reported, JSHBPS guideline, ESGAR guideline, CCBS guideline, and Korean model in the training and validation. ROC, receiver operating characteristic; CEA, carcinoembryonic antigen; US-reported, ultrasonic report diagnosis.

Fig. 3. DCA for each prediction method in the training (A) and validation (B) dataset.

The y-axis measures the net benefit. DCA, decision curve analysis; US-reported, ultrasonic report diagnosis.