Abstract
BACKGROUND: Several models are currently available for predicting the malignancy of pancreatic intraductal papillary mucinous neoplasm (IPMN), namely, the Pancreatic Surgery Consortium (PSC), the Japan Pancreas Society (JPS), the Johns Hopkins Hospital (JHH), and the Japan-Korea (JPN-KOR) models. However, a head-to-head comparison that shows which model is more accurate for this individualized prediction is lacking. AIM: To perform a head-to-head comparison of the four models for predicting the malignancy of pancreatic IPMN. METHODS: A total of 181 patients with IPMN who had undergone surgical resection were identified from a prospectively maintained database. The characteristics of IPMN in patients were recorded from endoscopic ultrasound imaging data and report archives. The performance of all four models was examined using Harrell's concordance index (C-index), calibration plots, decision curve analyses, and diagnostic tests. RESULTS: Of the 181 included patients, 94 were categorized as having benign disease, and the remaining 87 were categorized as having malignant disease. The C-indexes were 0.842 [95% confidence interval (CI): 0.782-0.901], 0.704 (95%CI: 0.626-0.782), 0.754 (95%CI: 0.684-0.824), and 0.650 (95%CI: 0.483-0.817) for the PSC, JPS, JHH, and JPN-KOR models, respectively. Calibration plots showed that the PSC model had the least pronounced departure from ideal predictions. Of the remaining three models, the JPS and JHH models underestimated the probability of malignancy, while the JPN-KOR model overestimated the malignant potential of branch duct-IPMN. Decision curve analysis revealed that the PSC model resulted in a better clinical net benefit than the three other models. Diagnostic tests also showed a higher accuracy (0.801) for the PSC model. CONCLUSION: The PSC model exhibited the best performance characteristics. Therefore, the PSC model should be considered the best tool for the individualized prediction of malignancy in patients with pancreatic IPMN.