Wai et al.1 developed indexes that identified significant fibrosis in patients with chronic hepatitis C. These indexes, including aspartate aminotransferase (AST) and platelets, had good discriminative power, as shown by areas under the receiver operating characteristic curve (AUROC) of 0.82 for formula 1 predicting fibrosis, 0.92 for formula 2 predicting cirrhosis, 0.83 for AST-Platelets Ratio Index (APRI) for fibrosis, and 0.90 for cirrhosis. We also validated a fibrosis index (Fibrotest; Biopredictive, Houilles, France, US Patent Application Serial No. 09/687,459) with high predictive values.2–4
We sought to compare these indices by using data collected retrospectively in 323 patients from our original population with complete biochemical data.2 The prevalence of significant fibrosis (F2–F4) was 41%, and cirrhosis (F4) was 13%. The mean value (SE) for AST expressed in ULN was 1.71 (0.10) and platelets count 192 (3) 109/L. AUROCs for significant fibrosis were 0.75 (0.03) for formula 1 and 0.74 (0.03) for APRI versus 0.83 (0.02) for Fibrotest (P = 0.03 vs. formula 1 and P = 0.02 vs. APRI),5 suggesting that the latter has greater discriminative power. The AUROCs for cirrhosis were 0.82 (0.04) for formula 2 and 0.80 (0.04) for APRI versus 0.92 (0.03) for Fibrotest (P = 0.04 vs. formula 2 and P = 0.02 vs. APRI), suggesting that the latter has greater discriminative power. There was a continuous, almost linear, relationship between Fibrotest and fibrosis stage, with significant differences between stages that were not observed for the Wai1 indexes (Fig. 1).
In addition to superior diagnostic power, Fibrotest has several other advantages. First, Fibrotest is not transaminase-dependent. AST has poor sensitivity for fibrosis detection. When we apply the sensitive cut-off recommended (1.00),1 we observe a 27% false-negative rate for cirrhosis: 11 out of 41 patients with cirrhosis had an APRI below 1.00. Among these 11 patients, 6 had normal AST values. We observed 25 fibrotic patients out of 131 (19% false-negative rate) who had APRI of 0.50 or below, the cut-off recommended for excluding significant fibrosis. Among these 25 patients, 22 had normal AST. The so-called standardization of transaminases using the upper normal limit given by laboratories is hazardous.6 To explain the differences in sensitivities observed in our population versus the Wai population, we wonder if, in their center, they were performing biopsies routinely in patients with normal transaminases, as we did in our center. This type of selection could have overestimated the sensitivity of AST.
Second, we did not include platelet count in Fibrotest to develop an index easily automated using minimal equipment. The components of Fibrotest can be measured using a single autoanalyzer with minimal variability.6
Third, Fibrotest has been validated in patients with human immunodeficiency virus/HCV-coinfection,4 who often have thrombocytopenia. Fibrotest is responsive to changes in fibrosis attributable to interferon-based therapy.3 Finally, a combination of alanine aminotransferase levels and Fibrotest (Actitest) accurately predicts the severity of necroinflammatory activity.2, 3 In conclusion, these comparisons suggest that Fibrotest provides a more accurate estimate of HCV-related fibrosis than the Wai indexes.1 Prospective comparisons in additional centers are warranted to confirm this finding.