Diagnosis of hepatic fibrosis in hepatitis B patients by logistic regression modeling based on plasma amino acid ratio and age

ObjectiveTo explore the efficacy of logistic regression modeling based on plasma amino acid profile and patient age, for diagnosing hepatic fibrosis in patients with chronic hepatitis B (CHB). MethodsOne-hundred-and-forty-eight patients (108 males; mean age: 38.1±11.9 years, range: 16-72 years) histologically diagnosed with CHB at our institute between January 2009 and June 2011 were enrolled for study. Pathological staging of fibrosis classified patients as: significant (≥S2), severe (≥S3), or advanced (S4) stages. Plasma molar concentrations of amino acids were determined by high-performance liquid chromatography, and used to establish the top six fractional ratio models according to progressive significance of correlations with pathological stages (by Spearman′s rank correlation coefficient analysis). The abilities of scale models to predict fibrosis were evaluated using the area under the receiver operating characteristic (AUROC) curve. ROC curve analysis allowed for the best cutoff for age to be determined for the pathological stages and calculation of sensitivity, specificity, and accuracy for the corresponding prediction probabilities (PP). ResultsPatient age was postively correlated with pathological stage (rs=0.262, P=0.001). The significant correlation coefficients for the six models were (in descending order): (isoleucine+leucine)/tyrosine (rs=-0.350, P=0.000), leucine/tyrosine (rs=-0.349, P=0.000), (isoleucine+leucine+valine)/tyrosine (rs=-0.345, P=0000), (leucine+valine)/tyrosine (rs=-0.342, P=0.000), (isoleucine+valine)/tyrosine (rs=-0.336, P=0.000), valine/tyrosine (rs=-0311, P=0.000).The logistic regression model based on age and the ratio models of plasma amino acids could not be established. For severe fibrosis, the PP was 0.722 (95% CI: 0.640-0.804), with sensitivity of 0.49, specificity of 0.85, and accuracy of 0.69. For advanced fibrosis, the PP was 0.792 (95% CI: 0.708-0.876), with sensitivity of 0.73, specificity of 0.76, and accuracy of 0.75. ConclusionLogistic regression modeling based on plasma amino acid profile and patient age could have clinical utility for sensitive, specific and accurate diagnosis of severe and advanced fibrosis in CHB patients.