MRI-Derived Liver Score Flags Failure Risk Before Hepatocellular Carcinoma Surgery

A higher score on a specialized magnetic resonance imaging measure of liver function is linked to markedly lower odds of a dangerous surgical complication in patients having liver tumors removed, according to a new systematic review and meta-analysis published in European Radiology Abdomen. The functional liver imaging score, or FLIS, is calculated from gadoxetic acid–enhanced MRI scans already performed as part of the routine diagnostic workup for hepatocellular carcinoma, the most common primary liver malignancy and a major cause of cancer-related death worldwide. The review pooled data from five retrospective studies totaling 1,832 patients, of whom 1,740 were included in analyses specific to hepatocellular carcinoma, and found that each additional point on the six-point scale was associated with roughly 45 percent lower odds of post-hepatectomy liver failure.

Post-hepatectomy liver failure is a leading cause of death after liver surgery. When surgeons remove a hepatocellular carcinoma, the safety of the operation depends on how well the remaining liver tissue, the future liver remnant, can carry out its metabolic work. If that residual reserve is inadequate, the remnant may fail, and the consequences can be fatal. Conventional preoperative risk stratification relies on clinical evaluation, biochemical scores such as the Child–Pugh classification, the Model for End-Stage Liver Disease, the albumin-bilirubin grade, and the indocyanine green retention rate at 15 minutes, together with volumetric measurement of the future remnant. These tools reflect global hepatic function but contain no imaging-derived information, and the review’s authors argue they may miss the regional variability in hepatocyte performance that characterizes advanced chronic liver disease.

FLIS was introduced in 2016 as a way to quantify liver function directly from imaging. Gadoxetic acid is a liver-specific MRI contrast agent taken up by functioning hepatocytes through the organic anion-transporting polypeptides OATP1B1 and OATP1B3 and excreted into bile via multidrug resistance-associated protein 2, which makes the hepatobiliary phase of the scan a window into cellular function. The score ranges from 0 to 6 and combines three hepatobiliary phase components, each graded 0, 1, or 2: the quality of hepatic parenchymal enhancement, judged against the signal intensity of the right kidney; the quality of biliary contrast excretion, from none to contrast reaching the common hepatic duct or beyond; and the portal vein sign, which compares portal vein signal to liver parenchyma. Higher composite values indicate better preserved hepatocyte function. Although originally developed to predict graft function in liver transplantation, FLIS has since been studied as a prognostic factor for liver failure, overall survival, and liver regeneration after hepatocellular carcinoma resection.

The new review, registered prospectively with the Open Science Framework and conducted according to PRISMA 2020 guidelines, searched PubMed, Embase, and Web of Science from database inception through February 2026. After screening, five retrospective cohort studies published between 2022 and 2025 met the inclusion criteria. Four were conducted in Asia, three in China and one in Japan, and one in Italy. Hepatitis B virus was the predominant underlying etiology in the Asian cohorts, accounting for 84 to 85 percent of patients, while hepatitis C virus predominated in the Italian cohort at 56.9 percent. All five studies used gadoxetic acid at the standard dose of 0.025 mmol/kg with hepatobiliary phase imaging 20 minutes after injection. Sample sizes ranged from 60 to 885 patients, and all studies were single-center.

The central quantitative result comes from two of the three studies that examined the association between FLIS and post-hepatectomy liver failure. In multivariate logistic regression, Luo and colleagues reported an odds ratio of 0.452 per FLIS point, and Zheng and colleagues reported 0.680. Pooling these with a DerSimonian-Laird random-effects model yielded a combined odds ratio of 0.55, with a 95 percent confidence interval of 0.37 to 0.83 and a p value of 0.004, meaning each one-point increase in the score was associated with a 45 percent reduction in the odds of liver failure. Both individual estimates pointed in the same protective direction. Heterogeneity between the two studies was substantial, with an I² of 82.5 percent, likely attributable to differences in how liver failure was defined: Luo et al counted all International Study Group of Liver Surgery grades A through C, with an incidence of 17.9 percent, whereas Zheng et al restricted the outcome to clinically significant grades B and C, with an incidence of 7.6 percent.

The review also assessed how well FLIS discriminates between patients who will and will not develop liver failure. The two contributing studies reported areas under the receiver operating characteristic curve of 0.752 and 0.722, and the pooled AUC was 0.73 with a 95 percent confidence interval of 0.70 to 0.76 and low heterogeneity of 29.3 percent. In head-to-head comparisons within the same studies using DeLong’s test, FLIS consistently outperformed the conventional scores: MELD achieved AUCs of 0.557 to 0.558, the albumin-bilirubin score 0.609, and indocyanine green retention at 15 minutes 0.605 to 0.624, with all pairwise differences statistically significant. A sensitivity analysis adding a third study that used a composite endpoint of liver failure or death barely changed the pooled AUC, which remained 0.73. The authors characterize this as moderate discriminative ability.

Survival evidence rests on a single study. Sakai and colleagues followed 235 patients and compared those with FLIS scores of 2 to 5 against those with the maximum score of 6. In multivariate Cox regression adjusting for age, tumor multiplicity, tumor size, and microvascular invasion, low FLIS was independently associated with shorter overall survival, with a hazard ratio of 2.558. Five-year overall survival was 76.4 percent for patients with FLIS 6 versus 42.4 percent for those with FLIS 2 to 5, a difference the authors note persisted in a propensity score-matched analysis of 49 patients per group. Microvascular invasion was also independently associated with survival, with a hazard ratio of 2.564. Recurrence-free survival, by contrast, was not significantly associated with FLIS in the multivariate model. Because these findings come from one retrospective study, the review treats them as exploratory.

The evidence base carries important caveats that the authors emphasize repeatedly. All five included studies were retrospective, exposing the results to selection bias and residual confounding. The pooled odds ratio derives from only two studies, and those two, conducted at the same institution in China with overlapping enrollment periods from 2017 to 2019 and at least one shared author, cannot be assumed to involve entirely separate patients. With so few studies per outcome, formal assessment of publication bias was impossible and heterogeneity estimates are unstable. The Quality In Prognosis Studies tool rated two studies as low risk of bias and three as high risk, with common concerns including retrospective design, limited reporting of losses to follow-up, and in one case the exclusion of a patient who died of liver failure, which may introduce survival bias. The predominance of hepatitis B–related disease in the Asian cohorts further limits generalizability to Western populations, where fatty liver disease and alcohol-related liver disease predominate.

One included study also examined liver regeneration, measured as a continuous index from CT volumetry, in 60 patients. FLIS correlated positively with the regeneration index overall and in the minor hepatectomy subgroup, and was an independent predictor in multivariate linear regression, but the association disappeared in the major hepatectomy subgroup of 25 patients. FLIS also correlated inversely with histological fibrosis stage, raising the possibility that the regeneration signal partly reflects the severity of parenchymal disease rather than functional reserve independent of fibrosis. The review labels these findings hypothesis-generating. The authors also note that although FLIS is conceptually capable of capturing spatial heterogeneity in liver function, all included studies applied it as a whole-liver score rather than a remnant-specific assessment, which is a meaningful limitation for surgical decision-making where the function of the future liver remnant is the critical determinant.

Interobserver reliability of the score was consistently high across the evidence base. Two independent radiologists scored FLIS in four studies, and hepatobiliary surgeons in one; intraclass correlation coefficients ranged from 0.914 to 0.953, one study reported a Cohen’s kappa of 0.998, and assessors were blinded to clinical outcomes in all five studies. Median FLIS values ranged from 3 to 6 across cohorts, reflecting differing underlying disease severity. The review’s conclusion is measured: FLIS showed associations with post-hepatectomy liver failure and survival after hepatocellular carcinoma resection, with moderate discriminative performance, but because the pooled estimates rest on two studies for liver failure and one for survival, they should be read as exploratory. The authors call for prospective multicenter validation with standardized liver failure definitions and remnant-specific FLIS assessment before the score is adopted alongside established biochemical scores in preoperative risk stratification.

Subject of Research: Association between the preoperative functional liver imaging score on gadoxetic acid–enhanced MRI and post-hepatectomy liver failure and survival after hepatocellular carcinoma resection

Article Title: Preoperative functional liver imaging score in hepatocellular carcinoma resection: systematic review and meta-analysis

Article References: Barajas Ordonez, F., Martínez Safar, I. E., Westfechtel, S., Rendón-Garavito, A. M., Truhn, D., Kuhl, C., Barabasch, A., & Nebelung, S. (2026). Preoperative functional liver imaging score in hepatocellular carcinoma resection: systematic review and meta-analysis. European Radiology Abdomen, 1(1), Article 4. https://doi.org/10.1007/s44501-026-00006-4

Image Credits: AI Generated

DOI: 10.1007/s44501-026-00006-4

Keywords: hepatocellular carcinoma, functional liver imaging score, gadoxetic acid MRI, post-hepatectomy liver failure, hepatectomy, meta-analysis, liver function, overall survival, risk stratification, future liver remnant, systematic review, prognostic factor