Accuracy of a predictive model for severe hepatic fibrosis or cirrhosis in chronic hepatitis C.

Aim: To assess the accuracy of a model in diagnosing severe fibrosis/cirrhosis in chronic hepatitis C virus (HCV) infection. Methods: The model, based on the sequential combination of the Bonacini score (BS: ALT/AST ratio, platelet count and INR) and ultrasonography liver surface characteristics, was applied to 176 patients with chronic HCV infection. Assuming a pre-test probability of 35%, the model defined four levels of post-test probability of severe fibrosis/cirrhosis: <10% (low), 10-74% (not diagnostic), 75-90% (high) and >90% (almost absolute). The predicted probabilities were compared with the observed patients' distribution according to the histology (METAVIR). Results: Severe fibrosis/cirrhosis was found in 67 patients (38%). The model discriminated patients in three comparable groups: 34% with a very high (>90%) or low (<10%) probability of severe fibrosis, 33% with a probability ranging from 75% to 90%, and 33% with an uncertain diagnosis (i.e., a probability ranging from 10% to 74%). The observed frequency of severe fibrosis/cirrhosis was within the predefined ranges. Conclusion: The model can correctly identify 67% of patients with a high (>75%) or low (<10%) probability of cirrhosis, leaving only 33% of the patients still requiring liver biopsy.