Steatotic Liver Disease Risk Scores to Predict Cirrhosis and Hepatocellular Carcinoma

Importance Most individuals with steatotic liver disease (SLD) do not develop advanced liver disease. Multiple risk scores for SLD-associated cirrhosis and hepatocellular carcinoma (HCC) have been proposed, but their utility to inform decisions about surveillance for advanced liver disease is unknown. Objective To compare the clinical utility of SLD risk scores for predicting cirrhosis and hepatocellular carcinoma in patients with noncirrhotic steatotic liver disease. Design, Setting, and Participants This cohort study used data from the national US Veterans Affairs health system of adults with imaging-confirmed SLD without viral hepatitis or primary liver disease from 2008 and 2020. Nine clinical risk scores were calculated at the time of the first imaging results that demonstrated hepatic steatosis. Individuals were followed up until cirrhosis, HCC, death, or 6 months after last follow-up visit. Data were analyzed between November 1, 2025, and May 11, 2026. Exposures Nine risk scores evaluated for prediction of cirrhosis or HCC within 10 years were: (1) ALBI, albumin-bilirubin; (2) aMAP, age-male-albumin-bilirubin-platelet; (3) APRI, the aspartate aminotransferase-to-platelet ratio; (4) BARD, body mass index, age, alanine aminotransferase-to-aspartate aminotransferase ratio, and diabetes; (5) FIB-4, fibrosis-4 index; (6) NFS, nonalcoholic fatty liver disease fibrosis; (7) SAFE, steatosis-associated fibrosis estimator; and 2 additional scores. Main Outcomes and Measures First diagnosis of cirrhosis and first diagnosis of HCC identified using VA and Medicare diagnosis codes and data from the VA Cancer Registry. Cox proportional hazards models were used to calculate time-dependent 10-year probabilities of each outcome predicted by 9 risk scores. Decision curves were generated for each outcome by calculating the net benefit, a weighted measure of true and false positives over a prespecified range of risk thresholds at which a clinician may reasonably recommend intervention. Nine risk scales were used to quantify the risk of cirrhosis or HCC within 10 years. Results The analysis included 853 131 patients (median [IQR] age, 61 [51-68] years; 62 168 females [7.3%] and 790 965 males [92.7%]) whose median (IQR) BMI was 31.3 (27.6-35.5) and 275 898 (32.3%) had diabetes; of these, 33 794 (3.96%) developed cirrhosis and 2978 (0.35%) developed HCC within 10 years. The FIB-4, APRI, and SAFE scores demonstrated the greatest discrimination of cirrhosis risk, while the SAFE, Tate, and FIB-4 scores demonstrated the greatest discrimination of HCC risk. The SAFE score demonstrated the greatest net benefit predicting cirrhosis. A score of 29.5 corresponded to a 10-year cirrhosis risk of 2.5% and yielded a net benefit of 0.019, or 1.9 additional individuals who develop cirrhosis per 100 individuals classified as at-risk patients. At a 0.25% 10-year risk of HCC, the SAFE score had a net benefit of 0.0016 (1.6 additional true positives per 1000 individuals). Conclusions and Relevance The findings of this cohort study show that the SAFE score may inform decisions to repeat screening for cirrhosis in patients with SLD. Clinical risk scores for HCC demonstrated minimal utility to inform HCC screening decisions for patients with noncirrhotic SLD.