Oral (pre-recorded)
Pancreas 
 
OP02 Pancreas: Pancreatic Cysts 
Selection of Presentations from Abstract Submissions
OP02-02 Recurrence Patterns after Surgical Resection of Intraductal Papillary Mucinous Neoplasm (IPMN) of Pancreas; a Multicenter, Retrospective Study of 1,074 IPMN Patients by Japan Pancreas Society
Seiko Hirono, Japan

S. Hirono1, Y. Shimizu2, T. Ohtsuka3, T. Kin4, S. Hijioka5, A. Yanagisawa6, M. Nakamura3, K. Okazaki7, H. Yamaue1, The Japan Pancreas Society
1Second Department of Surgery, Wakayama Medical University, School of Medicine, Japan, 2Aichi Cancer Center Hospital, Japan, 3Kyushu University, Japan, 4Teine-Keijinkai Hospital, Japan, 5National Cancer Center Hospital, Japan, 6Kyoto Prefectural University of Medicine, Japan, 7Kansai Medical University, Japan

Introduction: Although there are numerous reports focusing on surgical indication for intraductal papillary mucinous neoplasm (IPMN), the recurrence patterns following surgery are less widely reported. To ascertain optimal treatment and postoperative surveillance for IPMN patients, we analyzed patterns and risk factors for recurrence after surgery for IPMN.
Methods: This study is a retrospective, multi-institutional, observational study, including 1,074 patients undergoing surgery for IPMN at 11 academic institutions. We analyzed risk factors for recurrence after classifying postoperative recurrences into metachronous high-risk lesions (malignant progression of IPMN and/or metachronous pancreatic ductal adenocarcinoma) in the remnant pancreas and extra-pancreatic recurrence.
Results: Of 1,074 patients undergoing surgery for IPMN, 155 patients (14.4%) developed postoperative recurrence. We found that 34.3% of 70 high-risk lesions in the remnant pancreas occurred over 5 years after surgery, and survival of 36 patients undergoing second operation for high-risk lesions was better than that of 34 patients who did not (P=0.04). We found four independent risk factors for metachronous high-risk lesions in remnant pancreas: symptoms (P=0.005, hazard ratio [HR]: 1.988), location of pancreatic body/tail (P< 0.001, HR: 3.876), main duct size ≥10 mm (P=0.021, HR: 1.900), and high-grade dysplasia/invasive intraductal papillary mucinous carcinoma (IPMC) (P< 0.001, HR: 3.204). Although six patients (0.7%) with low- or high-grade dysplasia IPMN developed extra-pancreatic recurrence, invasive IPMC was the strongest risk factor for extra-pancreatic recurrence (P< 0.001, HR: 39.667).
Conclusion: We suggest that life-time continuous surveillance might be necessary for IPMN patients. Second surgery for metachronous high-risk lesions in remnant pancreas should be considered.
OP02-03 The Preoperative Dysplasia Grade Prediction Nomogram for Main-duct/Mixed Type Pancreatic Intraductal Papillary Mucinous Neoplasms (MD-IPMNs): The Development and Validation Based on a 7-year Single Center Database
Han Yu Zhang, China

H.Y. Zhang, Y.T. Li, M.H. Dai
Peking Union Medical College Hospital, Chinese Academy of Medical Sciences & Peking Union Medical College, China

Introduction: Surgical indications for MD-IPMN remain controversial due to the risk of pancreatic surgical procedure. Previous guidelines and predictive models were mainly focused on branch duct type IPMNs(BD-IPMNs) and have been relatively insufficient for both development and independent validation.
Method: IPMN patients who had undergone resection were collected retrospectively between 2013 and 2019. Patients were separated into MD-IPMN and BD-IPMN based on preoperative imaging. 177 MD-IPMN patients were finally enrolled and divided into training subset and test subset. Logistic regression modeling was used to develop model for identifying low-risk(low-/intermediate-grade dysplasia) or high-risk(high-grade dysplasia or invasive carcinoma) disease. The test subset was used for validating model by AUC, Brier's Score and C-index.
Results: We identified 203 IPMN patients underwent resection[MD:193(95.1%), BD:10(4.9%)]. 12 loss-pathology-patients were excluded. In MD-group, high-risk was present in 103 patients(53%) which is comparable to 74 low-risk patients(47%). MD-IPMN predictive nomogram was developed on the training set(70%, 124 patients) and validated on the test set(30%, 53 patients). The significant risk factor associated with high-risk disease including the presence of enhancing nodule(P=0.036), mural nodule(P=0.03), main-duct diameter(P=0.004) and the abnormal of CA19-9 value(P=0.028). Other potential risk factors including age, time since detection and cyst size. Brier's Score were 0.092 and 0.152, C-indexes were 0.94 and 0.85 on training and independent validation sets, respectively; AUCs were 0.916 and 0.856, respectively.
Conclusions: We present an independently validated nomogram for the prediction of MD-IPMNs malignant risk which thereby improve the identification of high-risk precursor and help avoid unnecessary damage by redundant surgical procedure.
[Figure1]
OP02-05 Comparison of Performance between Machine Learning Technique and Logistic Regression in Terms of Risk Prediction for Malignancy of the Intraductal Papillary Mucinous Neoplasm of Pancreas
Jae Seung Kang, Korea, Republic of

J.S. Kang1, C. Lee2, Y. Han1, Y.J. Choi1, Y. Byun1, H. Kim1, W. Kwon1, T. Park2, J.-Y. Jang1
1Surgery and Cancer Research Institute, Seoul National University College of Medicine, Korea, Republic of, 2Interdisciplinary Program in Bioinformatics, Seoul National University, Korea, Republic of

Background: Most nomograms predicting malignant intraductal papillary mucinous neoplasm (IPMN) of pancreas were developed based on the logistic regression (LR) analysis. This study was to develop a prediction model using machine learning (ML) and compare the performances between ML and LR model.
Method: This was a multi-national, multi-institutional, retrospective study. Malignant IPMNs were defined as those with high grade dysplasia and associated invasive carcinoma. Auto ML technique was utilized in R program. Six algorithms of ML (XG boost, deep learning, distributed random forest, generalized linear mode, gradient boosting machine, stacked ensemble [SE]) were utilized and compared. The algorithm which had the best performance was selected, and the performances of ML algorithm and LR model were compared.
Result: The total of 3,096 patients were enrolled. The patients were divided into model development set and external validation set with ratio of 2:1. In a multivariate LR, age, sex, main duct diameter, cyst size, mural nodule, and tumor location were independent risk factors for malignant IPMN. LR model consisted of these factors. Of the six algorithms, SE had the highest area under the receiver operating curve (AUC) in the internal validation (AUC, 0.742). The performances were comparable between ML and LR models in the external validation (AUC, 0.725 vs. 0.721).
Conclusion: The performance of LR model was comparable to that of ML. The LR model would be more practical because of its clinical convenience.