Thursday 14 march 2019
13:06 - 13:09h
Categories: Klinisch, Postersessie
Parallel session: Postersessies 3 - Clinical
J.D. de Boer1, H. Putter2, J.J. Blok3, I.P.J. Alwayn4, B. van Hoek5, A.E. Braat4.
1Dept. of Transplant Surgery, Eurotransplant, Leiden.2Dept. of Medical Statistics, Leiden University Medical Center, Leiden. 3Dept. of Surgery, Haaglanden Medical Center, Den Haag. 4Dept. of Transplant Surgery, Leiden University Medical Center, Leiden.5Dept. of Gastroenterology and Hepatology, Leiden University Medical Center, Leiden, The Netherlands.
Background: Several risk models have been developed to predict outcome after liver transplantation (LT) in the last decade. This study analyzes the ability of most well-known risk models to predict patient, overall graft and death-censored graft survival at short- and long-term follow-up after transplantation.
Methods: Data included information from the SRTR database on LTs from deceased donors performed in adults (≥18 years old) from January 1st, 2005 until December 31st, 2015. For all LTs the BAR-score, DRI, ET-DRI, DRM, sRRI, SOFT and D-MELD scores were calculated. Model performance was evaluated by the discriminative capacity and area under the ROC-curve (c-statistic) for patient survival, overall graft survival and death-censored graft survival. High-risk transplantations were defined as scores above 80th percentile according to the respective risk models.
Results: In the study period, 62,294 LTs were included. Patient survival at 3 months was best predicted by the SOFT (c-statistic: 0.68) and BAR score (c-statistic: 0.64) while the DRM and SOFT score had the highest predictive capacity at 5 years (c-statistic: 0.59). Overall graft survival was best predicted by the SOFT-score at 3-months (c-statistic: 0.65), and by the SOFT and DRM score at 5-year follow-up (c-statistic: 0.58). Death-censored graft survival at 5-year follow-up is best predicted by the DRI (c-statistic: 0.59) and ET-DRI (c-statistic: 0.58). For patient- and overall graft survival, high-risk transplantations were best defined by the DRM at 5-year follow-up. High-risk transplantations for death-censored graft survival were best defined by the DRI.
Conclusions: This study shows that outcome after liver transplantation is best predicted at short-term follow-up. Models dominated by recipient variables, like the BAR and SOFT score, have best performance for predicting short-term patient survival. Models that also include sufficient donor variables like the SOFT-score and DRM, have better performance for long-term graft survival. The DRI and ET-DRI include solely donor variables and best predict death-censored graft survival.