Practical Value of Serum Biochemical Indexes for Predicting Liver Fibrosis

ZHANG Zhan-qing,ZHANG Zhui,LU Wei,TONG Hai-tao

DOI: https://doi.org/10.3969/j.issn.1005-8982.2007.03.010

2007-01-01

Abstract:[Objective] To build mathematical models for diagnosing liver fibrosis, and to evaluate the practical value of serum biochemical indexes for predicting liver fibrosis. [Methods] 74 patients with chronic hepatitis were enrolled into present study. The specimens of liver biopsy were diagnosed for pathological staging (S). Serum biochemical indexes were determined with automated biochemical analyzer. Taking pathological staging as the grouping variables, and serum biochemical indexes as the independents, the Fisher′s linear discriminant functions were built by Bayes′s stepwise discriminant analysis. [Results] The pathological staging was correlated evidently with serum aspartate transaminase, total protein, albumin, globulin, prealbumin, total bile acids, fibronectin, choline esterase, albumin/globulin ratio, aspartate transaminase/alanine transaminase ratio. Only albumin/globulin ratio entered the discriminant functions accord to the entry criteria of the independents. The correctly classified rates of the original grouped cases by the discriminant models were 81.82% for S1, 21.05% for S2, 57.69% for S3 and 14.29% for S4, respectively; and the correctly classified total rate of the original grouped cases was 51.35%. The susceptibility, specificity, positive and negative predictive value by the discriminant models for diagnosing cirrhosis were 85.71%, 100%, 100% and 91.78%, respectively. [Conclusions] The discriminant models are sensitive for predicting the liver fibrosis, but are not sensitive for differentiating different pathological staging. The discriminant models are very valuable for affirming and negating cirrhosis.