Validation of five hepatic steatosis algorithms in metabolic‐associated fatty liver disease: A population based study
Yuxiu Liu, Shiying Liu, Jiaofeng Huang, Yueyong Zhu, Su Lin
Abstract
BACKGROUND AND AIM: Non-invasive hepatic steatosis algorithms are recommended in detecting metabolic-associated fatty liver disease (MAFLD) in epidemiological studies. However, the diagnostic accuracy of these models is unclear. This study aimed to evaluate the diagnostic efficiency of five common models in a national survey population. METHODS: The Third National Health and Nutrition Examination Survey (NHANES III) datasets were used in this study. The fatty liver index (FLI), hepatic steatosis index (HSI), non-alcoholic liver disease-liver fat score (NAFLD-LFS), the steato text (ST), and visceral adiposity index (VAI) were evaluated. RESULTS: The prevalence of MAFLD in the general population was 31.2%. The proportion of MAFLD estimated using the NAFLD-LFS (30.8%) was the closest to the real number, whereas the ST model (66.1%) significantly overestimated the prevalence of MAFLD in this cohort. The FLI (36.9%) and HSI models (38.5%) also slightly overestimated the prevalence of MAFLD in the study population. The FLI had the highest area under the receiver operating characteristic (AUROC) value (0.793) among all models, with a sensitivity of 57.0%, a specificity of 83.8%, a positive predictive value (PPV) of 67.3%, and a negative predictive value (NPV) of 77.0%. The combination of the original algorithm with additional metabolic dysfunction criteria did not improve the diagnostic efficiency. The discriminative ability for MAFLD in all models was lower in participants with a normal body mass index (BMI). CONCLUSIONS: Non-invasive models, especially the FLI, have satisfactory diagnostic performance in detecting MAFLD. However, models in people with normal BMIs require further development.