Chen Gh

Selection of the most powerful predictors for the evaluation of hepatic steatosis grade: An experimental study

PURPOSE To select the most powerful predictors for the evaluation of hepatic steatosis grade. METHODS AND MATERIALS Forty-five healthy New Zealand rabbits were randomly divided into one normal control group and three experimental groups. Hepatic steatosis models were established by feeding a high-fat, high-sugar diet and drinking water containing 5% ethanol. Twenty-two variable indexes were measured using general observation, biochemical examination, ultrasonography, computed tomography (CT), and proton magnetic resonance spectroscopy (MRS). Univariate analysis, correlation analysis, and stepwise regression analysis were used to make the selection of the most powerful predictors. ROC analysis was used to compare the diagnostic efficacy of single index with combined index (Y) expressed by a regression equation. RESULTS Based on statistical analysis, there were 12 variable indexes with significant differences among groups, which correlated with hepatic steatosis grade: liver weight, hepatic index, liver CT value, liver-to-muscle attenuation ratio, 1H MRS fat peak value, fat peak area, fat-to-water peak area ratio, fat percentage, ultrasound attenuation coefficient, serum aspartate aminotransferase, total cholesterol (TC) and triglycerides. Among them hepatic index, liver CT value and serum TC were selected as the most powerful predictors for hepatic steatosis grade with correlation coefficients of 0.709, -0.764, and 0.886, respectively. The regression equation was: Y=1.975 + 3.906 x 10(-2)X1 + 0.369X2-2.84 x 10(-2)X3, where Y=hepatic steatosis grade, X1=TC, X2=hepatic index, and X3=liver CT value. ROC analysis displayed PPV, NPV, curve area of combined index (Y) were superior to simple index (hepatic index, liver CT value and serum TC) in evaluating hepatic steatosis grade, and they were nearly 1.0000, 1.0000 and 1.000, respectively. CONCLUSIONS Combined application of several diagnostic methods is superior to simple diagnostic method, and could provide comprehensive, rapid, accurate evaluation of hepatic steatosis grade.