We thank Campbell et al. for their interest in our paper. Their critiques about our methodological approach are interesting and give us the opportunity to underline more in-depth the interest of the rate of fibrosis progression model.
The authors are right to state that we did not provide a true longitudinal observation of fibrosis progression using serial liver biopsy samples. However, our study was designed to avoid all selection biases and was more informative than a cross-sectional study in which the endpoint would be the fibrosis stage at a single time. Indeed, it simulated a follow-up examination with two liver biopsy samples: the first one (virtual) at the time of HCV infection with a fibrosis stage at F0, the second one (real) performed at a known date.
More important than a debate on semantic is the doubt of Campbell et al. about the validity of our results. The majority of our analyses were focused on the rate of fibrosis progression (a continuous quantitative variable), allowing t tests and multiple regression models for statistical analyses. Survival analyses were only used to illustrate our findings and to test the robustness of our results by using different statistical models. When we grouped patients with fibrosis stage F2, F3, and F4 in the same analysis we did not generate artificial data by using extrapolation of estimated dates of development of any stage of liver fibrosis; only true histological findings were considered. However we agree with Campbell et al that the concept of “fibrosis stage F2 or greater” is somewhat vague since F2 and F4 have not the same clinical significance. To address this issue, the best approach is not to extrapolate the estimated time of development of stage F2 in patients with F3 or F4 stage, but rather to perform separate analyses considering each stage of liver fibrosis (F1, F2, F3 and F4) as different censoring events.1 Patients in whom the fibrosis stage is greater than the stage considered as censoring event must be excluded from these analyses. We did not report these results in the paper because they should have appeared as redundant, but we are pleased to report them in the present letter. The comparison of nulliparous women with women with past history of pregnancy using the log-rank test showed an accelerated progression to F1 (P = .02), F2 (P = .09), F3 (P = .004) or F4 (P = .07) in nulliparous women (Fig. 1). A type 2 error may account for the non significant results. Similar analyses of the impact of hormonal replacement therapy (HRT) on the fibrosis progression in HCV-infected menopausal women showed an accelerated progression to F1 (P = .02) and F2 (P = .07) stages in women who did not receive HRT; the difference was not further significant for the F3 and F4 stages, suggesting that HRT may impact only on the early development of liver fibrosis.
In conclusion, we understand the remark on semantic but disagree with Campbell et al. when they talk about misuse of statistical methods.