We thank Berg et. al. for the interest expressed on our article.1

We are grateful that many investigators like Berg et. al. had tested our prognostic model in their patient populations.2, 3 Using the aspartate aminotransferase to platelet count ratio index (APRI), significant fibrosis could be identified or excluded in 57.5% of Berg's patients with 82% sensitivity and 93% specificity, and cirrhosis could be identified or excluded in 82.5% of patients with 69% sensitivity and 93% specificity. The specificities were almost identical to those in our study, while the sensitivities were lower. The proportion of patients who fell into the classifiable group was higher than in our study. We agree that the diminished sensitivities may be related to the use of a different fibrosis scoring system, as well as multiple versus single pathologists scoring the biopsies. It is also possible that the inclusion of stage 3 in the Scheuer4 scoring system (fibrosis with architectural distortion but no obvious cirrhosis) as cirrhosis may have affected the accuracy of prediction of cirrhosis.

We agree that interobserver or intraobserver variability and size of the biopsies may affect histological interpretation. Because liver biopsies are not necessarily gold standards for assessing liver histology, noninvasive models will not have complete concordance with histological staging. We agree that neither liver biopsy nor noninvasive model is necessary when the severity of the patients' chronic liver disease is obvious, as in the case of patients with decompensated cirrhosis. However, differentiating mild from moderate fibrosis, and moderate fibrosis from well-compensated cirrhosis may be difficult even for experienced hepatologists. Thus, there remains a role for simple noninvasive models that will help in providing more reliable estimates of the stage of liver disease than a clinician's intuition.

Chun-Tao Wai*, Anna S. F. Lok*, * University of Michigan, Ann Arbor, MI.