Comparison of fatty liver index with noninvasive methods for steatosis detection and quantification

To compare noninvasive methods presently used for steatosis detection and quantification in nonalcoholic fatty liver disease (NAFLD).

Cross-sectional study of subjects from the general population, a subgroup from the First Israeli National Health Survey, without excessive alcohol consumption or viral hepatitis. All subjects underwent anthropometric measurements and fasting blood tests. Evaluation of liver fat was performed using four noninvasive methods: the SteatoTest; the fatty liver index (FLI); regular abdominal ultrasound (AUS); and the hepatorenal ultrasound index (HRI). Two of the noninvasive methods have been validated vs liver biopsy and were considered as the reference methods: the HRI, the ratio between the median brightness level of the liver and right kidney cortex; and the SteatoTest, a biochemical surrogate marker of liver steatosis. The FLI is calculated by an algorithm based on triglycerides, body mass index, γ-glutamyl-transpeptidase and waist circumference, that has been validated only vs AUS. FLI < 30 rules out and FLI ≥ 60 rules in fatty liver.

Three hundred and thirty-eight volunteers met the inclusion and exclusion criteria and had valid tests. The prevalence rate of NAFLD was 31.1% according to AUS. The FLI was very strongly correlated with SteatoTest (r = 0.91, P < 0.001) and to a lesser but significant degree with HRI (r = 0.55, P < 0.001). HRI and SteatoTest were significantly correlated (r = 0.52, P < 0.001). The κ between diagnosis of fatty liver by SteatoTest (≥ S2) and by FLI (≥ 60) was 0.74, which represented good agreement. The sensitivity of FLI vs SteatoTest was 85.5%, specificity 92.6%, positive predictive value (PPV) 74.7%, and negative predictive value (NPV) 96.1%. Most subjects (84.2%) with FLI < 60 had S0 and none had S3-S4. The κ between diagnosis of fatty liver by HRI (≥ 1.5) and by FLI (≥ 60) was 0.43, which represented only moderate agreement. The sensitivity of FLI vs HRI was 56.3%, specificity 86.5%, PPV 57.0%, and NPV 86.1%. The diagnostic accuracy of FLI for steatosis > 5%, as predicted by SteatoTest, yielded an area under the receiver operating characteristic curve (AUROC) of 0.97 (95% CI: 0.95-0.98). The diagnostic accuracy of FLI for steatosis > 5%, as predicted by HRI, yielded an AUROC of 0.82 (95% CI: 0.77-0.87). The κ between diagnosis of fatty liver by AUS and by FLI (≥ 60) was 0.48 for the entire sample. However, after exclusion of all subjects with an intermediate FLI score of 30-60, the κ between diagnosis of fatty liver by AUS and by FLI either ≥ 60 or < 30 was 0.65, representing good agreement. Excluding all the subjects with an intermediate FLI score, the sensitivity of FLI was 80.3% and the specificity 87.3%. Only 8.5% of those with FLI < 30 had fatty liver on AUS, but 27.8% of those with FLI ≥ 60 had normal liver on AUS.

FLI has striking agreement with SteatoTest and moderate agreements with AUS or HRI. However, if intermediate values are excluded FLI has high diagnostic value vs AUS.