Assessment of hepatic steatosis algorithms in non-alcoholic fatty liver disease

Background: In order to optimize the identification of persons with non-alcoholic fatty liver disease (NAFLD), several algorithms for hepatic steatosis were developed. These available algorithms, as well as an algorithm, derived using biochemical and anthropometric data of our participants, are compared in a cross-sectional pilot study. Material and methods: We included 77 participants with abdominal obesity: 43 with NAFLD and 33 without NAFLD. Body mass index (BMI), waist circumference (WC) and hip circumference (HC), systolic and diastolic blood pressure were assessed. Fibrinogen, high sensitive C-reactive protein (hsCRP), aspartate aminotransferase (AST), alanine transaminase (ALT), gamma-glutamyl transferase (GGT), uric acid, ferritin, glucose, insulin, homocysteine, lipid status parameters, apolipoprotein A-I, apolipoprotein B and Lp(a)-lipoprotein were measured. Fatty liver was assessed by ultrasound with the presence or absence of hepatic steatosis. Discovering the most significant factor in the presence of NAFLD is assessed through logistic regression modeling. The predictor variables were chosen according to an algorithm derived from conducted factor analysis and other available algorithms for hepatic steatosis. Results: Participants with NAFLD had significantly higher BMI (34.38 +/- 9.73 vs 28.05 +/- 4.79 kg/m(2), p = 0.001), WC (108.05 +/- 11.47 vs 96.15 +/- 14.27 cm, p = 0.001), HC (114.93 +/- 11.01 vs 108.21 +/- 9.82 cm, p = 0.050), systolic (128.98 +/- 8.67 vs 122.42 +/- 10.62 mmHg, p = 0.010) and diastolic blood pressure (83.64 +/- 5.94 vs 78.33 +/- 7.57 mmHg, p = 0.001), AST (23.93 +/- 6.91 vs 21.70 +/- 5.21 U/L, p = 0.014), ALT (30.50 +/- 13.70 vs 23.00 +/- 11.75 U/L, p = 0.007), hsCRP (4.34 +/- 5.56 vs 2.98 +/- 2.34mg/l, p = 0.004) and uric acid (358.02 +/- 83.29 vs 296.78 +/- 84.54 mu mol/l, p = 0.001), in comparison non NAFLD. Logistic regression model with algorithm derived from factor analysis showed the best performance. From other available algorithms, only fatty liver index (FLI) and hepatic steatosis index (HSI) had statistically significant discriminatory power. Conclusions: Elevation of WC, HC, BMI, DBP, SBP, Fbg, hsCRP, glucose, and uric acid, incorporated in our hepatic steatosis prediction model, had the best predictive power among all assessed algorithms.