Get your full text copy in PDF
Benedikt Reichert, Thomas Becker, Tobias J. Weismüller, Moritz Kleine, Lea Zachau, Kai Johanning, Frank Lehner, Jürgen Klempnauer, Harald Schrem
Ann Transplant 2012; 17(2): 11-17
Background: The SOFT-score, P-SOFT-score, SALT-score and labMELD-score have been applied for the prediction of survival of liver transplant recipients after transplantation. We analysed the value of these scores for the prediction of short-term survival in high-risk liver transplant recipients with a labMELD-score ≥30.
Material/Methods: Retrospective single-centre analysis including 88 consecutive liver transplants in adults between 01.01.2007 and 31.12.2010 with a pretransplant labMELD-score ≥30 and follow-up until the 31.12.2011. Combined and living-related liver transplants were excluded. ROC-curve analysis was used to calculate sensitivity, specificity and overall model correctness of prognostic models.
Results: The P-SOFT-score demonstrated a significant influence on 1-year patient survival (p=0.045, Mann-Whitney-U test). Multivariate Cox regression analysis showed a significant influence of the P-SOFT-score on patient (p=0.013; Exp(B)=1.050; 95%CI: 1.010-1.091) and on graft survival (p=0.023; Exp(B)=1.042; 95%CI: 1.006-1.080). ROC-curve analysis showed areas under the curve (AUROCs) <0.5 for the SOFT-score, P-SOFT-score, SALT-score and the labMELD-score ≥30 for the prediction of 3-month patient and graft survival as well as 1-year patient and graft survival.
Conclusions: Our results imply that the SOFT-score, P-SOFT-score, SALT-score and labMELD-score ≥30 all have a sensitivity, specificity and overall model correctness that is unable to discriminate short-term survivors from non-survivors in a collective of high-risk liver transplant recipients sufficiently in order to guide clinical decision making in the current German transplant situation with decreasing numbers of deceased liver donors, decreasing donor organ quality and increasingly sick transplant candidates.
Keywords: Liver Transplantation, predictive models, Mortality