The Steatosis-associated fibrosis estimator (SAFE) score: validation in the general US population

BACKGROUND: Noninvasive tests are important in the initial risk stratification of people at risk of fibrosis. The recently developed steatosis-associated fibrosis estimator (SAFE) score may have such potential but awaits external validation. METHODS: We analyzed 6973 participants aged 18-80 in the National Health and Nutrition Examination Survey 2017-2020 cycle with data on liver stiffness and SAFE score without prevalent heart failure. Fibrosis was defined as liver stiffness ≥8.0 kPa. Accuracy was evaluated by AUC and assessment of test characteristics at the prespecified cutoffs for ruling out/ruling in fibrosis. RESULTS: The SAFE score categorized 14.7% of the population as high risk, 30.4% as intermediate risk, and 54.9% as low risk for fibrosis. The actual fibrosis prevalence in these groups was 28.0%, 10.9%, and 4.0%, respectively, translating into a positive predictive value of 0.28 at the high-risk cutoff and a negative predictive value of 0.96 at the low-risk cutoff. The AUC of the SAFE score (0.748) was significantly higher than the fibrosis-4 index (0.619) or NAFLD fibrosis score (0.718). However, test performance strongly depended on age categories: 90% of participants aged 18-40 years were considered at low risk for fibrosis, including 89/134 (66%) of clinically significant fibrosis cases. In the oldest group (60-80 y), fibrosis could only be safely ruled out among 17%, corresponding to a high referral rate of up to 83%. The best SAFE score performance was found in the middle-aged group (40-60 y). The results were consistent in target populations with metabolic dysfunction or steatosis. CONCLUSIONS: The SAFE score has overall good diagnostic accuracy in detecting fibrosis but was highly dependent on age. The SAFE score lacked sensitivity in younger populations and the ability to rule out fibrosis in older populations.