SAFE Score Most Useful for Predicting Cirrhosis, Researchers Report

Out of nine liver risk scores evaluated in a nationwide study of US veterans, the Steatosis-Associated Fibrosis Estimator score was the most useful for predicting cirrhosis risk within a decade, though its predictive edge remained modest, researchers reported on September 21, 2026, in JAMA Internal Medicine.

Evaluating Nine Risk Scores in US Veterans with Liver Disease

Led by Catherine Mezzacappa, MD, PhD, MPH, of the Department of Internal Medicine at Yale School of Medicine in New Haven, Connecticut, the study analyzed 853,131 patients with imaging-confirmed steatotic liver disease between 2008 and 2020. The median patient age was 61 years, and 92.7% of the cohort were men. Investigators excluded individuals with viral hepatitis, other primary liver diseases, pre-existing cirrhosis, hepatic decompensation, or hepatocellular carcinoma diagnosed before or shortly after steatosis.

The research team calculated nine distinct clinical risk scores around the time of each patient’s initial steatosis imaging. These models included the Fibrosis-4 Index, aspartate aminotransferase-to-platelet ratio, albumin-bilirubin, age-male-albumin-bilirubin-platelet, BMI-age-aspartate aminotransferase-to-alanine aminotransferase ratio-diabetes, Nonalcoholic Fatty Liver Disease Fibrosis Score, SAFE, Sinn, and Tate. Patients were followed for a median duration of 8.5 years.

Ten-Year Cirrhosis and Cancer Outcomes

Over the ten-year observation window, 33,794 patients, representing 3.96% of the cohort, developed cirrhosis. Another 2,978 patients, or 0.35%, developed hepatocellular carcinoma. The Fibrosis-4 Index, aspartate aminotransferase-to-platelet ratio, and SAFE scores performed similarly when distinguishing patients who developed cirrhosis over 10 years from those who did not. For hepatocellular carcinoma prediction, SAFE, Tate, and Fibrosis-4 Index performed best.

The SAFE score yielded the largest net benefit for predicting cirrhosis. At a threshold score of 29.5, reflecting a 2.5% ten-year risk, the model identified one additional case for every 53 patients flagged as at risk. This represented a small improvement over reassessing every patient with steatotic liver disease. For hepatocellular carcinoma, however, the predictive gain was minimal. Among patients without cirrhosis, a SAFE score of 43.8 translated to a 0.25% ten-year risk and identified roughly 1.6 additional cancer cases per 1,000 patients flagged as at risk.

Better Stratification Tools Could Reduce Specialist Referrals

Current guidelines from the American Association for the Study of Liver Diseases recommend the Fibrosis-4 Index to screen for cirrhosis in steatotic liver disease. However, the index can miss at-risk patients while flagging numerous low-risk individuals for extra testing. The new findings suggest that better stratification tools could alter how clinicians manage monitoring intervals.

distinguishing patients at high vs low risk of incident cirrhosis may inform the intensity of risk monitoring and reduce both invasive testing and premature referral to hepatology. Patients at low risk for incident cirrhosis may need less frequent monitoring, which can be managed by their primary care physician, than current recommendations and may not need an early hepatology referral

Melinda Wang, MD, MHS

Study Limitations and Data Gaps

Several methodological limitations affected the findings. Missing clinical data prevented the research team from assessing six other published risk scores. The analysis may have overestimated the predictive performance of the Tate score because investigators developed that specific model using data originating from the same healthcare system. Reliance on imaging report text to identify steatosis may also have introduced selection bias.

The study received financial backing from the US National Institutes of Health, including the National Institute of Diabetes and Digestive and Kidney Diseases, the National Cancer Institute, and the National Institute on Alcohol Abuse and Alcoholism. The US Department of Veterans Affairs provided in-kind support, and the study authors reported no relevant conflicts of interest.