International multicenter study to characterize the individual risk of malignancy in branch duct IPMN and proposal of nomogram to predict malignancy

Background: Although many clinical factors are used to predict malignancy in branch duct type intraductal papillary mucinous neoplasms (BD-IPMNs) according to consensus guideline, several variables have different statistical value and individual prediction of malignancy risk is impossible. The purpose of this study was to elucidate the malignant predictor and evaluate individual risk for malignancy and finally suggest nomogram for malignancy prediction of BD-IPMN using world largest DB of IPMN by Korea-Japan collaboration study group. Methods: A total of 2,258 BD-IPMNs were included from 22 tertiary hospitals in Korea and Japan. Using these patients, nomogram was made and externally validated. Result: Malignant IPMNs were defined as those with IPMN with high grade dysplasia and associated invasive carcinoma. 986 patients had low, 443 intermediate, 398 high grade dysplasia, and 431 invasive carcinoma. For nomogram construction and validation, we randomly assigned the patients into the training and validation set with keeping ratio of benign and malignancy. Through exhaustive searching using multiple logistic regression, five variables (cyst size, duct dilatation, mural nodule, serum CA19-9/CEA) were finally selected to construct nomogram. In validation set, this nomogram showed excellent discrimination power through 1000 times bootstrapped calibration test, and all predictions lie within a 10% margin of error. Conclusions: We propose malignancy predicting nomogram for BD-IPMN using meaningful variables through logistic regression model. This nomogram could be useful to select optimal treatment method considering individual risk of malignancy.