Reply:^†

Poynard et al. propose several explanations for the discrepancies between our study and their data. Two can be easily dismissed, as FibroTest values were indeed calculated by BioPredictive; moreover, the five serum markers used in this fibrosis index were measured following strict requirements specified by this company. With regard to the choice of the cutoff, we adopted the 0.31 value. This is the most sensitive to exclude significant fibrosis (Metavir F2 or higher)1; all our patients, including all 14 with Metavir stage F2 or F3, remained well below the 0.71 value indicated as the most specific, with a positive predictive value of 76%. Using the restrictive presumption that liver biopsy results are always correct, BioPredictive considers “more prudent to advise the use of FibroTest-ActiTest only at extreme values.”1 Nonetheless, re-analyzing the data according to the 0.48 cutoff, the diagnostic value of FibroTest in the studied population can be expressed as follows: sensitivity, 21%, specificity, 81%, positive predictive value, 38%, negative predictive value, 66%. Finally, ours is undeniably a small study, and we concede it is possible that future, larger studies might come to different conclusions. However, in such future studies the selection of patients with persistently normal alanine aminotransferase (ALT) will have to rely on strictly defined criteria, such as those we used (no value >1.2 times the upper normal limit in a minimum of 17 determinations, along more than 5 years; hence, the small sample size) and which Poynard et al. did not follow. In fact, these authors considered HCV carriers with normal ALT those who, on the very day a liver biopsy was performed, had ALT within the normal limits (baseline normal ALT),2, 3 a very loose surrogate for persistently normal ALT.

Castera et al. are not satisfied with the cutoff we have chosen for liver elasticity (8.7 kPa), originally reported by Ziol et al.4 It is most obvious that, in the trade-off between sensitivity and specificity, at the lower cutoff (7.1 kPa) they suggest, Fibroscan sensitivity cannot improve, being already 100%, but specificity necessarily decreases. Using a cutoff that worsens the overall diagnostic value of a test is hardly justifiable. With regard to the second issue raised by Castera et al., the proportion of patients in whom liver elasticity measurements could not be obtained was 0/40 (vs. 3/100 in their series; P = .56 by Fisher's exact test). Failure to measure liver elasticity is usually due either to abdominal obesity or ascites,5 both absent in our patients. Unfortunately, Castera et al. neither specify why their patients could not have liver elasticity measured, nor define the HCV carriers with normal ALT they studied in terms of length of follow-up and number of ALT determinations. Finally, to equate the specificity of FibroTest in blood donors without hepatitis C (and no liver biopsy) to that in HCV carriers with normal ALT does not seem appropriate, since these two populations should not be considered equivalent.