Abstract
INTRODUCTION: Fatty liver index (FLI) is widely used for detection of steatotic liver disease (SLD) in population studies. While frequently used to study SLD in lean individuals, its performance in this subgroup has not been validated. We hypothesized that FLI, which incorporates body-size measures, is not suitable for studies focusing on lean SLD. METHODS: Data from the National Health and Nutrition Examination Survey 2017-20 were used. Adults with available imaging and variables for FLI calculation were included. SLD was defined by controlled attenuation parameter ≥263 dB/m. Lean status was determined using ethnicity-adjusted body mass index cutoffs. The diagnostic performance of FLI was assessed overall and in lean participants using standard rule-in (≥60) and rule-out (<30) thresholds. RESULTS: The study population included 7,191 individuals, of whom 3,602 (50%) had SLD. SLD was seen in 238 of 1,739 lean individuals (13.7%). While FLI had adequate performance in the overall population, it performed poorly in lean participants, in whom FLI ≥60 detected only 15 of 238 SLD cases (sensitivity 6.3%, positive predictive value 46%) with only modest improvement using the rule-out threshold. Lean individuals predicted to have SLD by FLI markedly differed from imaging-confirmed lean SLD; they were older and had higher body mass index, waist circumference, gamma-glutamyl transferase, and triglyceride levels and a greater burden of cardiometabolic comorbidities. DISCUSSION: With its inherent dependence on body-size measures, FLI does not identify lean SLD accurately. Moreover, applying FLI to lean individuals overestimates disease severity. The misclassification and misrepresentation biases imply that FLI should not be used in population-based studies of lean SLD.