Development of Novel Multivariable Logistic Regression Model for Predicting Graft Failure within 2 Weeks and 4 Weeks after Liver Transplantation Using Post-Transplantation Laboratory Values
Transplantation, Samsung Medical Center, Seoul, Korea, Republic of
Meeting: 2020 American Transplant Congress
Abstract number: A-162
Keywords: Area-under-curve (AUC), Graft failure, Liver transplantation
Session Information
Session Name: Poster Session A: Liver Retransplantation and Other Complications
Session Type: Poster Session
Date: Saturday, May 30, 2020
Session Time: 3:15pm-4:00pm
Presentation Time: 3:30pm-4:00pm
Location: Virtual
*Purpose: This study designed a prediction model for graft failure after liver transplantation.
*Methods: Multivariable logistic regression including aspartate aminotransferase (AST), total bilirubin (TB), and international normalized ratio (INR) of prothrombin time for predicting graft failure within 2 and 4 weeks, respectively, were performed. The predictive models were evaluated with various performance measures including the area under receiver operating characteristic curve (AUC) and net reclassification improvement (NRI). Five-fold cross validation was applied for internal validation. A decision curve analysis was performed to evaluate the clinical usefulness of the models.
*Results: Data of 1539 patients were included for modeling. Logistic regression with single variable of maximum INR≥2.53 during 3 to 7 (AUC=0.9224,CI=0.8726-0.9721,P<0.0001) and maximum INR≥2.6 during 3 to 14 post-transplantation day (AUC=0.900,CI=0.8518-0.9482,P<0.0001) showed relationship to graft failure within 2 and 4 weeks, respectively. Multivariable model including log2-scaled maximum AST during 0 to 7, log2-scaled maximum TB and log2-scaled maximum INR during 3 to 7 post-transplantation day were related to graft failure within 2 weeks (AUC=0.9715,CI=0.9428-0.9993,P<0.0001) while model with log2-scaled maximum AST during 0 to 14, log2-scaled maximum TB and log2-scaled maximum INR during 3 to 14 post-transplantation day, were related to graft failure within 4 weeks (AUC=0.9669,CI=0.9476-0.9863,P<0.0001). The multivariable model showed 128.1% (CI=99.9-156.2%, P<0.0001) and 110.5% (CI=80.7-140.2%, P<0.0001) of improvement in net reclassification improvement compared to INR model in predicting graft failure within 2 and 4 weeks, respectively. Based on Youden’s index method, predictive probability of 5.62% (sensitivity=92.1%, specificity=95.8%) and 6.69% (sensitivity=87.0%, specificity=95.2%) were cut-offs for graft failure within 2 and 4 weeks, respectively. Net benefit analysis showed that 72.7% and 76.9% were the cut-offs with a net-zero benefit for graft failure within 2 and 4 weeks, respectively.
*Conclusions: Predicted probability of 5.62% and 6.69% calculated by our model can be considered as cut-offs where decision for re-transplantation shows highest advantages.
To cite this abstract in AMA style:
Kwon J, Kim S, Choi G, Joh J, Park J, Rhu J, Kim K, Lee O, Lim M, Jeong E, Yang J. Development of Novel Multivariable Logistic Regression Model for Predicting Graft Failure within 2 Weeks and 4 Weeks after Liver Transplantation Using Post-Transplantation Laboratory Values [abstract]. Am J Transplant. 2020; 20 (suppl 3). https://atcmeetingabstracts.com/abstract/development-of-novel-multivariable-logistic-regression-model-for-predicting-graft-failure-within-2-weeks-and-4-weeks-after-liver-transplantation-using-post-transplantation-laboratory-values/. Accessed May 16, 2025.« Back to 2020 American Transplant Congress