Development and Validation of a Comprehensive Model to Estimate Early Allograft Failure Among Patients Requiring Early Liver Retransplant

Abstract

Importance: Expansion of donor acceptance criteria for liver transplant increased the risk for early allograft failure (EAF), and although EAF prediction is pivotal to optimize transplant outcomes, there is no consensus on specific EAF indicators or timing to evaluate EAF. Recently, the Liver Graft Assessment Following Transplantation (L-GrAFT) algorithm, based on aspartate transaminase, bilirubin, platelet, and international normalized ratio kinetics, was developed from a single-center database gathered from 2002 to 2015.

Objective: To develop and validate a simplified comprehensive model estimating at day 10 after liver transplant the EAF risk at day 90 (the Early Allograft Failure Simplified Estimation [EASE] score) and, secondarily, to identify early those patients with unsustainable EAF risk who are suitable for retransplant.

Design, setting, and participants: This multicenter cohort study was designed to develop a score capturing a continuum from normal graft function to nonfunction after transplant. Both parenchymal and vascular factors, which provide an indication to list for retransplant, were included among the EAF determinants. The L-GrAFT kinetic approach was adopted and modified with fewer data entries and novel variables. The population included 1609 patients in Italy for the derivation set and 538 patients in the UK for the validation set; all were patients who underwent transplant in 2016 and 2017.

Main outcomes and measures: Early allograft failure was defined as graft failure (codified by retransplant or death) for any reason within 90 days after transplant.

Results: At day 90 after transplant, the incidence of EAF was 110 of 1609 patients (6.8%) in the derivation set and 41 of 538 patients (7.6%) in the external validation set. Median (interquartile range) ages were 57 (51-62) years in the derivation data set and 56 (49-62) years in the validation data set. The EASE score was developed through 17 entries derived from 8 variables, including the Model for End-stage Liver Disease score, blood transfusion, early thrombosis of hepatic vessels, and kinetic parameters of transaminases, platelet count, and bilirubin. Donor parameters (age, donation after cardiac death, and machine perfusion) were not associated with EAF risk. Results were adjusted for transplant center volume. In receiver operating characteristic curve analyses, the EASE score outperformed L-GrAFT, Model for Early Allograft Function, Early Allograft Dysfunction, Eurotransplant Donor Risk Index, donor age × Model for End-stage Liver Disease, and Donor Risk Index scores, estimating day 90 EAF in 87% (95% CI, 83%-91%) of cases in both the derivation data set and the internal validation data set. Patients could be stratified in 5 classes, with those in the highest class exhibiting unsustainable EAF risk.

Conclusions and relevance: This study found that the developed EASE score reliably estimated EAF risk. Knowledge of contributing factors may help clinicians to mitigate risk factors and guide them through the challenging clinical decision to allocate patients to early liver retransplant. The EASE score may be used in translational research across transplant centers.

Conflict of interest statement

Conflict of Interest Disclosures: Dr Burra reported receiving personal fees and nonfinancial support from Biotest and from Kedrion; nonfinancial support from Chiesi farmaceutici; and personal fees and nonfinancial support from Astellas Pharma outside the submitted work. Dr Carraro reported receiving personal fees from Novartis outside the submitted work. No other disclosures were reported.

Figures

Figure 1.. Patient Flow Diagram

Patients accrued in the derivation and external validation sets are displayed separately. AST represents aspartate aminotransferase; PLT, platelet count. aIncluded were 7 high-volume transplant centers and 7 intermediate-volume centers. During the 11- to 90-day period, 1 patient in the derivation set was lost to follow-up. During the 91- to 730-day period (24 months), 5 patients were lost to follow-up. bIncluded were 1 high-volume transplant center and 1 intermediate volume center.

Figure 2.. Receiver Operating Characteristic (ROC) Curves

A, B. The ROC curves for the Early Allograft Failure Simplified Estimation (EASE) score (final model 9) and other models (5, 6, 7, and 8) at 90 days in the derivation set and in the external validation set. C, The ROC curves in the derivation set. D, The ROC curves for the EASE score developed at 90 days and applied at 30 days and for other estimated scores in the derivation. D-MELD indicates donor age × Model for End-stage Liver Disease; DRI, Donor Risk Index; EAD, Early Allograft Dysfunction; L-GrAFT, Liver Graft Assessment Following Transplantation; MEAF, Model for Early Allograft Function; and New ET-DRI, New Eurotransplant Donor Risk Index.

Figure 3.. Early Allograft Failure Simplified Estimation (EASE) Score and Kaplan-Meier Survival Curves

A, Sigmoidal day 90 early allograft failure (EAF) distribution according to the EASE score in 1609 evaluated patients. Five different risk classes are identified, with the dashed central line denoting the threshold for an unsustainable EAF risk. The constant obtained by logistic regression analysis was increased by 0.3060 to calibrate the unsustainable risk cutoff at the 0 threshold. B and C, Early allograft failure–free graft survival and patient survival according to the 5 EASE score risk classes are shown. The dashed line between classes 4 and 5 in panel C representing the extremely high-risk threshold (unsustainable risk cutoff) indicates the poor survival of patients in extremely high-risk class 5. The extremely high-risk class and the unsustainable risk cutoff indicate the threshold that mandates prompt retransplant.