TY - JOUR
AU - Gunby,, Joanne
AB - Dear Sir, Thank you for the opportunity to respond to the several comments made by Dr Girard regarding our meta-analysis of recombinant FSH (rFSH) versus urinary FSH (uFSH). We would like to start by reiterating that our search to identify all trials (whether published or not) that met our clearly specified inclusion criteria strategy was as comprehensive as possible. Consequently, we are surprised with Dr Girard's assertion that there was selective inclusion of some trials. Since the publication of our study, we have identified several more unpublished trials which are currently being evaluated for possible inclusion in an updated meta-analysis; we are still awaiting responses from the investigators of two studies regarding our questions about their study methods and data and have contacted N.V. Organon for their assistance because follitropin beta was used in both studies. We would be grateful to both Dr Girard and N.V. Organon for drawing to our attention any trials they believe we have missed so that the necessary evaluation of eligibility can be performed. The reliance on only peer-reviewed studies, as implied by Dr Girard (although it is unclear why he singled out for exclusion only one of the two unpublished studies, both of which had non-significant treatment effects) is problematic on two counts. First, for a variety of reasons, only a small proportion of studies presented at scientific meetings is submitted for publication. Thus, the potential for bias by not including the data from these studies is significant; publication bias of this type has been demonstrated to produce inaccurate estimates of the effect of treatment. Publication bias is a major threat to the validity and usefulness of a meta-analysis. The inclusion of data from unpublished trials is necessary to avoid this problem which may result in clinical practice being misdirected. Most meta-analysts are in agreement on this issue i.e. the search for evidence should be comprehensive and combine published data with data from what has been dubbed `fugitive literature' or `grey literature'. This point is clearly evident from our study in which the two unpublished trials, which were completed very recently and are currently undergoing the peer-review process for publication, had the highest rank for methodological quality among all the trials. A difficulty arises in not being certain, despite extensive searching and consultation with the scientific community, that unpublished trials do not remain hidden. Furthermore, the willingness of investigators of located, unpublished trials to share their data can be a challenging obstacle to overcome. Second, the peer-review process is not always successful in ensuring that published results are valid, because it is neither structured nor consistent. Consequently, the published literature is replete with trials that have highly variable methodological quality. In contrast, the selection and evaluation process involved in a systematic review and meta-analysis is much more rigorous and reliable, because the criteria for selecting trials are specified a priori and at least two reviewers are required to independently evaluate the quality of the trials. The issue of heterogeneity deserves clarification. Statistical heterogeneity is observed when the effect size from each trial varies, not only in magnitude (which is expected owing to sampling error) but also in direction. The problem is further magnified by conducting trials with small sample sizes. It may also be caused by known clinical or methodological differences among trials. In our study, the results were pooled using a fixed-effects model only after confirming that statistical heterogeneity was not present (i.e. the observed treatment effects in individual trials were not statistically significantly different from the overall pooled estimate of the treatment effect). Dr Girard argues that the pregnancy rates for rFSH and uFSH among trials comparing follitropin alpha with uFSH were more variable than when follitropin beta was compared to uFSH. We would like to point out the fact that all three trials using follitropin beta (Out et al., 1997a; Hoomans et al., 1999) were multicentre trials comprising three, six and 18 centres, respectively. In the latter two of these trials (Out et al., 1997a), information gleaned from the histograms in the publications showed remarkable variability in the pregnancy rates which ranged from 0% to 50% with rFSH and from 0% to approximately 46% with uFSH. Such wide variability, in part, may be the result of small sample sizes per participating centre, the range extending from three to 116 subjects. There is no question that true randomization is a desirable method for generating comparison groups for clinical trials. However, the debate is still ongoing on whether quasi-randomization as was used in the Manassiev trial (Mannassiev et al., 1997) is also an appropriate method for this purpose. We believe that quasi-randomization methods do have a role but should be given a lower value in a validity scoring system as was done in our study. An issue of higher importance in reducing bias is that of concealing the allocation schedule from health care providers entering subjects into trials. The use of a schedule based on random numbers is no guarantee that allocation to comparison groups will be free from bias, if the allocation is not concealed. Nevertheless, the exclusion of the Manassiev trial did not change the overall conclusions of our meta-analysis. Having observed that rFSH was significantly better than uFSH with respect to pregnancy rates, we performed subgroup analyses to evaluate the effect of the assisted reproductive procedure and the type of rFSH used. Both follitropin alpha and follitropin beta fared better than uFSH. Although the difference in pregnancy rates was statistically significant in the follitropin alpha versus uFSH comparison, it is possible that with larger numbers of subjects, a similar result may be obtained when follitropin beta is compared with uFSH. Nowhere in our manuscript did we state that follitropin alpha was better than follitropin beta, nor was such a comparison an objective of our study. Indeed, in the absence of studies directly comparing the two follitropins, any such conclusion is invalid and inappropriate. The inferences being made by Dr Girard and others, who have interpreted our subgroup analyses as being a direct comparison between the two follitropins, are incorrect. The irony is that when we contacted N.V. Organon in February 1998 for additional data we were advised that because the two follitropin preparation are different, they should be analyzed separately in the meta-analysis reiterating the point they had made in a previous publication (Out et al., 1997b). In conclusion, we stand firmly behind our work, which we believe is a comprehensive and methodologically sound systematic review and meta-analysis that shows rFSH is better than uFSH when used for ovarian stimulation in assisted reproduction. We feel that Dr Girard's comments are invalid and stem from his misinterpretation of our subgroup analyses, his apparent lack of awareness of the wide variability in pregnancy rates in the trials comparing follitropin beta with uFSH, and his view that only peer-reviewed published studies should be included in a meta-analysis. References Hoomans, E.H.M., Andersen, A.N., Loft, A. et al. ( 1999 ). A prospective, randomized clinical trial comparing 150 IU recombinant follicle stimulating hormone (Puregon) and 225 IU highly purified urinary follicle stimulating hormone (Metrodin-HP) in a fixed dose regimen in women undergoing ovarian stimulation. Hum. Reprod. , 14 , 2442 –2447. Manassiev, N.A., Davies, W.A.R., Leonard, et al. ( 1997 ). Initial results from the comparison of recombinant FSH and urinary FSH in an IVF programme. Hum. Reprod. , 12 , (Abstract book 1), p265 (Abstract R-068). Out, H.J., Driessen, S.G.A.J., Mannaerts, B.M.J.L., Coelingh-Bennink, J.H.T. ( 1997a ) Recombinant follicle-stimulating hormone (follitropin beta, Puregon) yields higher pregnancy rates in in vitro fertilization than urinary gonadotropins. Fertil. Steril. , 68 , 138 –142. Out, H.J., Olijve, W. and Coelingh-Bennink, H.J.T. ( 1997b ) The coming of wonders. Fertil. Steril. , 67 , 411 . © European Society of Human Reproduction and Embryology
TI - Letters to the Editor
JF - Human Reproduction
DO - 10.1093/humrep/15.7.1651
DA - 2000-07-01
UR - https://www.deepdyve.com/lp/oxford-university-press/letters-to-the-editor-bb2s6mSOrW
SP - 1651
VL - 15
IS - 7
DP - DeepDyve
ER -