Functional liver imaging score and relative enhancement index predict transplant-free survival in cirrhosis

To evaluate the prognostic performance of the Relative Enhancement Index (REI) and the Functional Liver Imaging Score (FLIS), derived from gadoxetic acid-enhanced MRI, for predicting transplant-free survival in patients with cirrhosis; to determine whether these imaging biomarkers provide prognostic information after adjustment for clinical severity markers; to assess their interobserver agreement; and, because transplantation and death are competing events, to disentangle the association of FLIS and REI with each outcome using a competing-risks framework.

In this retrospective study, cirrhotic patients who underwent gadoxetic acid-enhanced MRI between September 2014 and August 2018 were included and followed for up to eight years for transplant-free survival. The cohort was assembled using stratified convenience sampling, with 20 patients selected from each Child-Pugh class (A, B, and C), to ensure representation across the full spectrum of liver dysfunction. Two independent readers assessed the Relative Enhancement Index (REI) and the Functional Liver Imaging Score (FLIS). Interobserver agreement was evaluated using kappa statistics for the individual FLIS components and ICC for the total FLIS score and REI. Transplant-free survival was analyzed using Kaplan-Meier curves, the log-rank test, and Cox proportional hazards regression. Multivariable Cox regression models were constructed separately for FLIS and REI, adjusted for age, MELD score, and Child-Pugh score. A sensitivity analysis was additionally performed by treating liver transplantation as a censoring event and all-cause death as the only event of interest. In addition, a competing-risks analysis was performed using cumulative incidence functions, Gray's test, and Fine-Gray subdistribution hazard models, with death and liver transplantation analysed alternately as the event of interest. Discrimination was compared with MELD and Child-Pugh score using DeLong's test, and incremental value was assessed by the change in Harrell's C-index with likelihood-ratio testing. The proportional hazards assumption was tested with Schoenfeld residuals, and a 30-day landmark analysis was performed.

Sixty patients were included (32 men; median age, 58 years; interquartile range [IQR], 51-64 years), with 20 patients in each Child-Pugh class by design. During follow-up, 20 patients underwent liver transplantation (33.3 %), 16 died (26.7 %), and 24 remained alive without transplantation (40 %). Interobserver agreement was excellent for both FLIS (ICC = 0.97) and REI (ICC = 0.97). Because the cohort was deliberately stratified by Child-Pugh class, absolute transplant-free survival estimates were considered cohort-specific and not interpreted as population-level survival rates. AUROC values showed moderate discrimination for the composite endpoint of death or liver transplantation (FLIS, 0.745; REI, 0.703). In univariable Cox analysis, FLIS 0-3 and REI < 100 were significant predictors of adverse outcomes (death or liver transplantation), with hazard ratios of 3.767 (95 % CI: 1.872-7.579; p < 0.001) and 3.883 (95 % CI: 1.679-8.979; p = 0.002), respectively. After adjustment for age, MELD score, and Child-Pugh score, FLIS 0-3 remained independently associated with transplant-free survival (adjusted HR, 2.64; 95 % CI: 1.10-6.34; p = 0.030), whereas the association for REI < 100 was of similar magnitude but did not reach statistical significance (adjusted HR, 2.67; 95 % CI: 0.94-7.58; p = 0.065). In the death-only sensitivity analysis, FLIS remained significantly associated with survival (log-rank p = 0.042), whereas the association for REI was attenuated and did not reach statistical significance (log-rank p = 0.096). Discrimination did not differ significantly from that of MELD (AUROC, 0.731) or Child-Pugh score (AUROC, 0.753; all DeLong p ≥ 0.43), and both indices improved model fit when added to MELD (likelihood-ratio p = 0.004 for both). In competing-risks analysis, FLIS 0-3 and REI < 100 were associated with the cumulative incidence of liver transplantation (subdistribution HR, 3.42 and 6.18, respectively) but not with that of death (1.41 and 1.45).

FLIS and REI demonstrate excellent interobserver agreement and are associated with transplant-free survival in patients with cirrhosis, with moderate discriminative performance by AUROC analysis. FLIS remained independently associated with outcome after adjustment for age, MELD score, and Child-Pugh score, whereas REI showed an association of similar magnitude that did not reach significance after multivariable adjustment. Competing-risks analysis indicates that the association of both indices with the composite endpoint is driven predominantly by the transplantation component, and discrimination did not differ from that of MELD or Child-Pugh score. FLIS and REI should therefore be regarded as reproducible imaging markers of progression to end-stage liver disease rather than as validated predictors of mortality.