Post–Liver Transplant Outcomes: A Comparative Study of 6 Predictive Models

Abstract

Background. We compared the performance of the Liver Transplant Risk Score (LTRS) with the survival outcomes following liver transplantation (SOFT), pretransplant SOFT (P-SOFT), Balance of Risk Score (BAR), donor-age and model for end-stage liver disease (D-MELD), and Organ Procurement and Transplantation Network Risk Prediction Score (ORPS) for the prediction of 90-d mortality, 1-y mortality, and 5-y survival after first-time liver transplantation (LT). Methods. A retrospective analysis of the Scientific Registry of Transplant Recipients was conducted using data collected between 2002 and 2021. Results. A total of 82 696 adult LT recipients with a median age of 56 y were included. The area under the curve for 90-d mortality were 0.61, 0.66, 0.65, 0.61, 0.58, and 0.56 for the LTRS, SOFT, P-SOFT, BAR, D-MELD, and ORPS, respectively (all pairwise comparisons: P < 0.05). The area under the curve for 1-y mortality were 0.60, 0.63, 0.62, 0.59, 0.60, 0.57, and 0.59 for the LTRS, SOFT, P-SOFT, BAR, D-MELD, and ORPS, respectively (all pairwise comparisons: P < 0.05). The c-statistics for 5-y survival were not statistically significant among the models. For 90-d mortality, 1-y mortality, and 5-y survival, the correlation coefficients between the LTRS and P-SOFT (the 2 models requiring only preoperative parameters) were 0.90. 0.91, and 0.81, respectively (P < 0.01). Conclusions. None of the predictive models demonstrated sufficient precision to reliably identify LT recipients who died within 90 d and 1 y after LT. However, all models exhibited strong capabilities in perioperative risk stratification. Notably, the P-SOFT and LTRS models, the 2 models that can be calculated using only preoperative data, proved to be valuable tools for identifying candidates at a significant risk of poor outcomes.