MAF-5 Score and Liver-Related Event Risk in a Screening Cohort

Key Takeaways

• Liver-related event risk rose across low-, intermediate-, and high-risk MAF-5 categories.
• Adding age improved discrimination and further separated risk within the same MAF-5 strata.
• Comparisons with FIB-4, APRI, and NFS plus combined thresholds showed sensitivity-specificity tradeoffs rather than one uniformly superior threshold approach.

At Samsung Medical Center in Seoul, investigators followed 62,625 screened adults for a median 11.2 years. The age-modified MAF-5 score showed stronger discrimination for liver-related events than the original MAF-5, with a C-index of 0.870 in the Samsung Medical Center cohort report. The original score still showed substantial discriminatory ability over follow-up and separated participants into progressively higher-risk groups across its predefined low, intermediate, and high categories.

Researchers assembled a retrospective single-center cohort from health screening data collected in Seoul, South Korea, between January 2001 and December 2016 in adults aged 20 years or older. Of 112,493 screened participants, 62,625 remained after exclusions for chronic hepatitis B or C, heavy alcohol intake, current or prior cancer, organ transplantation, or missing MAF-5 variables. Participants with liver-related events within 6 months of enrollment or with follow-up shorter than 6 months were also excluded. Liver-related events included hepatocellular carcinoma and cirrhosis complications such as ascites, variceal bleeding, or hepatic encephalopathy. During follow-up, 147 participants developed these events, which were the clinical outcome assessed in the cohort.

The original MAF-5 score classified risk as low below 0, intermediate from 0 to below 1, and high at 1 or greater. Observed liver-related event incidence rates were 0.108, 0.576, and 1.520 cases per 1,000 person-years across the low-, intermediate-, and high-risk groups. Discrimination for the original score reached a C-index of 0.818 with a 95% CI of 0.778 to 0.859. Its integrated AUC was 0.784 with a 95% CI of 0.779 to 0.790. The category structure showed a clear stepwise gradient in event risk.

Investigators defined the age-modified score as 0.648×MAF-5 + 0.087×age and set the high aMAF-5 cutoff at 1.419, corresponding to the 85th percentile. This version achieved a C-index of 0.870 with a 95% CI of 0.837 to 0.904. Its integrated AUC reached 0.858 with a 95% CI of 0.854 to 0.861. Risk also separated further within the same MAF-5 strata. Across the tested biomarkers, aMAF-5 had the highest overall C-index and integrated AUC.

Among comparators, FIB-4 showed a higher C-index than the original MAF-5 at 0.835, while NFS showed a slightly higher integrated AUC at 0.795. The MAF-5 and aMAF-5 comparisons with FIB-4, APRI, and NFS also showed sensitivity-specificity tradeoffs at 5 and 10 years rather than a uniform threshold winner. For example, either-positive high-cutoff criteria raised sensitivity from 0.412 to 0.765 while specificity fell from 0.944 to 0.830. Both-positive rules shifted the tradeoff toward higher specificity, and subgroup analyses by age, sex, and metabolic dysfunction showed good performance in participants with and without metabolic dysfunction.

Interpretation is bounded by the retrospective design, baseline-only score assessment, absent biopsy or elastography data, an all-Asian cohort, and an event definition narrower than broader liver outcomes. Overall, the results reflect long-term discrimination for hard liver-related events within this cohort.