We appreciate the letters of Tsochatzis et al. and Degos et al. regarding our recent review. We essentially agree with their comments but would like to stress a few points.
It is obviously very relevant to maximize the diagnostic accuracy of any diagnostic test. In the case of noninvasive markers of liver fibrosis, the latter applies particularly to confirm or exclude the presence of cirrhosis, because there are important clinical implications derived from this finding (i.e., screening for hepatocellular carcinoma and gastroesophageal varices). We all would agree that in this setting most of the noninvasive methods widely used in clinical practice perform well. Regarding the identification of significant fibrosis, the diagnostic accuracy is certainly lower. However, their performance in the area of viral hepatitis C is probably acceptable. As an example in our own field, we are using the Model for End-Stage Liver Disease (MELD) score to predict the 3-month survival in patients awaiting liver transplantation (with the obvious consequences in a patient's life—or death) with diagnostic accuracies not far from those used to exclude significant fibrosis. The MELD score has been widely validated in different cohorts of patients with a prediction in 3-month survival ranging from 0.76-0.87.1, 2 Identifying or excluding significant fibrosis has limited practical implications in real life and, in the field of viral hepatitis, it may become less relevant once antiviral treatment efficacy increases.
Regarding the comment by Tsochatzis et al. on the use of the collagen proportionate area as a histological standard (based on its quantitative nature) we believe this is a good approach.3 The prognostic value regarding patient outcome was excellent in their study, although the results need validation in other cohorts. Using such methodology, however, will not preclude one of the main problems of liver biopsy, which is sampling error.
Finally, as we stressed in our review and as was noticed by Degos et al., using noninvasive methods to predict clinical outcomes is probably the most important goal. It is in this particular area where we need more and well-designed studies.