Predicting outcome after liver transplantation: utility of the model for end-stage liver disease and a newly derived
Niraj M Desai1, Kevin C Mange, Michael D Crawford
1Department of Surgery, Washington University School of Medicine, St. Louis, Missouri, USA.
Background:
The Model for End-Stage Liver Disease (MELD) has been found to accurately predict pretransplant mortality and is a valuable system for ranking patients in greatest need of liver transplantation. It is unknown whether a higher MELD score also predicts decreased posttransplant survival.
Methods:
We examined a cohort of patients from the United Network for Organ Sharing (UNOS) database for whom the critical pretransplant recipient values needed to calculate the MELD score were available (international normalized ratio of prothrombin time, total bilirubin, and creatinine). In these 2,565 patients, we analyzed whether the MELD score predicted graft and patient survival and length of posttransplant hospitalization.
Results:
In contrast with its ability to predict survival in patients with chronic liver disease awaiting liver transplant, the MELD score was found to be poor at predicting posttransplant outcome except for patients with the highest 20% of MELD scores. We developed a model with four variables not included in MELD that had greater ability to predict 3-month posttransplant patient survival, with a c-statistic of 0.65, compared with 0.54 for the pretransplant MELD score. These pretransplant variables were recipient age, mechanical ventilation, dialysis, and retransplantation. Recipients with any two of the three latter variables showed a markedly diminished posttransplant survival rate.
Conclusions:
The MELD score is a relatively poor predictor of posttransplant outcome. In contrast, a model based on four pretransplant variables (recipient age, mechanical ventilation, dialysis, and retransplantation) had a better ability to predict outcome. Our results support the use of MELD for liver allocation and indicate that statistical modeling, such as reported in this article, can be used to identify futile cases in which expected outcome is too poor to justify transplantation.
