Validating a novel score based on interaction between ACLF grade and MELD score to predict waitlist mortality

Background & Aims: Among candidates listed for liver transplant (LT), the model for end-stage liver disease (MELD) score may not capture acute-on-chronic liver failure (ACLF) severity. Data on the interaction between ACLF and MELD score in predicting waitlist mortality are scarce. Methods: We analyzed the UNOS database (01/2002 to 06/2018) for LT listings in adults with cirrhosis and ACLF (without hepatocellular carcinoma). ACLF grades 1, 2, 3a, and 3b- were defined using the modified EASL-CLIF criteria. Results: Of 18,416 candidates with ACLF at listing (mean age 54 years, 69% males, 63% Caucasians), 90-day waitlist mortality (patient death or being too sick for LT) was 21.6% (18%, 20%, 25%, and 39% for ACLF grades 1, 2, 3a, and 3b, respectively). Using a Fine and Gray regression model, we identified an interaction between MELD and ACLF grade, with ACLF having a higher impact at lower MELD scores. Other variables included candidate's age, sex, liver disease etiology, listing MELD, ACLF grade, obesity, and performance status. A score developed using parameter estimates from the interaction model on the derivation cohort (n = 9,181) stratified the validation cohort (n = 9,235) into quartiles: Q1 (score <10.42), Q2 (10.42-12.81), Q3 (12.82-15.50), and Q4 (>15.50). Waitlist mortality increased with each quartile from 13%, 18%, 23%, and 36%, respectively. Observed vs. expected waitlist mortality deciles in the validation cohort showed good calibration (goodness of fit p = 0.98) and correlation (R = 0.99). Conclusion: Among selected candidates who have ACLF at listing, MELD score and ACLF interact in predicting cumulative risk of 90-day waitlist mortality, with higher impact of ACLF grade at lower listing MELD score. Validating these findings in large prospective studies will support consideration of both MELD and ACLF when prioritizing transplant candidates and allocating liver grafts. Lay summary: In patients with cirrhosis listed for liver transplantation, the presence of multiorgan failure, a condition referred to as acute-on-chronic liver failure, is associated with high waiting list mortality rates. Current organ allocation policy disadvantages patients with this condition. This study describes and validates a new scoring method that performs better than the currently available scoring systems. Further validation of this approach may reduce the deaths of patients with cirrhosis and acute-on-chronic liver failure on the transplant waiting list.