Detailed Abstract
[E-poster]
[EP106] Development of a Nomogram Based on Radiologic Findings for Predicting Malignancy in Intraductal Papillary Mucinous Neoplasms of the Pancreas: An International Multicenter Study
Hyeong Seok KIM1, Wookyeong SONG2, Wonho CHOO2, Sungyoung LEE3, Youngmin HAN1, Hongbeom KIM1, Wooil KWON1, Taesung PARK2, Jin-Young JANG*1
1Department of Surgery and Cancer Research Institute, Seoul National University College of Medicine, Korea
2Department of Statistics, Seoul National University College of Natural Sciences, Korea
3Center for Precision Medicine, Seoul National University Hospital, Korea
Introduction : Although we previously proposed a nomogram and validated it in an external cohort, it is difficult to apply without tumor markers data. This study aimed to develop a nomogram to predict individual risks of malignancy in intraductal papillary mucinous neoplasms (IPMN) based on radiologic findings using an international cohort. Background:
Methods : A total of 3463 patients who underwent surgical resection at 31 tertiary institutes in 8 countries were enrolled and clinicopathological data were retrospectively analyzed. Patients were randomly allocated 2:1 into model development and test sets to construct the nomogram. Using the development set, the area under the receiver operating characteristics curve (AUC) was calculated using 10-fold cross validation by exhaustive search. Performance of the nomogram was assessed via Hosmer-Lemeshow goodness-of-fit test.
Results : Among patients, 188 had main duct type, 1991 had branch duct type, and 1284 had mixed type IPMN. Pathologic results showed 2094 (60.5%) had low grade dysplasia, 667 (19.3%) had high grade dysplasia, and 702 (20.3%) had invasive carcinoma. The exhaustive search resulted in four variables (location, cyst size, duct dilatation, and mural nodule) selected to construct the nomogram. AUC was 0.726 for the model development set (range, 0.707-0.743) and 0.726 for the test set (range, 0.691-0.762); the Hosmer-Lemeshow test did not show any evidence of model violation.
Conclusions : Although the new nomogram based on radiologic findings has slightly lesser diagnostic power than the previous nomogram including tumor markers, it is clinically helpful in circumstances where tumor markers are not available. The nomogram is freely available at http://statgen.snu.ac.kr/software/nomogramIPMN.
Methods : A total of 3463 patients who underwent surgical resection at 31 tertiary institutes in 8 countries were enrolled and clinicopathological data were retrospectively analyzed. Patients were randomly allocated 2:1 into model development and test sets to construct the nomogram. Using the development set, the area under the receiver operating characteristics curve (AUC) was calculated using 10-fold cross validation by exhaustive search. Performance of the nomogram was assessed via Hosmer-Lemeshow goodness-of-fit test.
Results : Among patients, 188 had main duct type, 1991 had branch duct type, and 1284 had mixed type IPMN. Pathologic results showed 2094 (60.5%) had low grade dysplasia, 667 (19.3%) had high grade dysplasia, and 702 (20.3%) had invasive carcinoma. The exhaustive search resulted in four variables (location, cyst size, duct dilatation, and mural nodule) selected to construct the nomogram. AUC was 0.726 for the model development set (range, 0.707-0.743) and 0.726 for the test set (range, 0.691-0.762); the Hosmer-Lemeshow test did not show any evidence of model violation.
Conclusions : Although the new nomogram based on radiologic findings has slightly lesser diagnostic power than the previous nomogram including tumor markers, it is clinically helpful in circumstances where tumor markers are not available. The nomogram is freely available at http://statgen.snu.ac.kr/software/nomogramIPMN.
SESSION
E-poster
E-Session 7/27 ~ 7/29 ALL DAY