Design: Retrospective national cohort with decision-curve analysis of nine scores
Population: 853,131 US veterans with steatotic liver disease and no cirrhosis
Primary outcome: Cirrhosis and hepatocellular carcinoma within 10 years
Effect: Cirrhosis 3.96%, HCC 0.35%; SAFE greatest net benefit for cirrhosis (0.019 at 2.5% risk); minimal utility for HCC

A US Veterans Affairs cohort followed 853,131 adults with imaging-confirmed steatotic liver disease and no cirrhosis or other liver disease, and compared nine risk scores for predicting cirrhosis and liver cancer within 10 years. Over that time 3.96% developed cirrhosis and 0.35% hepatocellular carcinoma.

FIB-4, APRI and SAFE (the steatosis-associated fibrosis estimator) discriminated cirrhosis risk best. On decision-curve analysis SAFE added the most net benefit: a score of 29.5 corresponded to a 2.5% 10-year cirrhosis risk and identified about two extra future cases per 100 people flagged. For liver cancer, no score was useful enough to guide screening decisions in people without cirrhosis.

For a general physician seeing fatty liver on every second ultrasound, the useful message is that a simple blood-based score helps decide who needs elastography or hepatology follow-up, and that cancer surveillance belongs to those with cirrhosis rather than to fatty liver in general. The cohort was 93% men, so check fit for women.

Calculate a fibrosis score such as FIB-4 or SAFE for every patient with fatty liver on imaging
Use a raised score to prompt elastography or referral, and a low score to support routine follow-up
Do not start liver cancer surveillance for fatty liver without cirrhosis on a risk score alone
The cohort was 93% men, so apply the thresholds to women with some caution

Why it matters

Most fatty liver never progresses, and a score is how a busy clinic finds the few that will.

The statistics, in plain English

Net benefit weighs correct identifications against false alarms at a chosen risk threshold. A net benefit of 0.019 at a 2.5% threshold means about 1.9 extra true future cirrhosis cases found per 100 people screened, compared with treating nobody as high risk.