In women with compromised ovarian reserve selected for IVF, clomiphene citrate and a short protocol with high-dose gonadotropins are both associated with disappointing results. The use of the latter regimen increased the number of oocytes retrieved and the proportion of women performing embryo transfer but not the delivery rates. In fact, a regimen of ovarian stimulation using exclusively clomiphene citrate may be considered in the management of these women since the chances of success did not markedly differ and this scheme is associated with lower costs.
The disappointing chances of success in women with compromised ovarian response are important data emerging from our trial. In both arms of the study, the delivery rate was below 5%, corresponding to the definition of “very poor prognosis” recently proposed by the Ethics Committee of the American Society for Reproductive Medicine . One may argue that this may be at least in part due to the low quality of the units involved. Even if we cannot robustly refute this criticism, some contrary arguments deserve to be mentioned. Firstly, during the study period, the participating centres reported delivery rates per started cycle for patients not included in the trial in line with those of the national Italian ART register. Secondly, all four units involved handled a consistent number of couples annually (>1000 IVF cycles per year) and all of them had a long-standing clinical and scientific experience in IVF. Thirdly, the observed low rate of success appears to be in agreement with data from the literature on women with a compromised ovarian reserve . Finally, it should be emphasised that the criteria used for the definition of compromised ovarian reserve in this study were very stringent and we may have selected a subgroup of women whose ovarian reserve is particularly damaged. From the health-provider point of view, the appropriateness of this intervention is inevitably questionable in these cases. Of note, our economical model under-estimated the overall costs per delivery since it did not include the costs of pregnancy assistance. More in-depth economic analyses are required prior to drawing firm indications for guiding health provider’s decisions in these cases.
Some aspects of our trial warrant clarification and discussion. Of utmost relevance here is that the trial was closed prior to completing the scheduled sample size and the power of the study is 60% in contrast to the initial calculation. Once the sample size has not been achieved one cannot make definite conclusions. This is especially true when the there is no difference between the two arms indicating the possibility of a beta error. In fact, the upper limit of the absolute reduction difference in favour of the high-dose gonadotropins group (7%) exceeded the 5% limit postulated as clinically relevant. The incapacity to achieve the scheduled sample size occurred despite the number of eligible patients being adequate to fulfil recruitment within a two-year period (Figure 1) and despite the initial assumptions for sample size calculation not being stringent. In our view, the difficulty in recruitment reflects the negative prejudgements women may have on an innovative approach using exclusively clomiphene citrate. External conditioning from other infertile women, the media or the internet may have played a role on this. We also cannot rule out that at least some physicians engaged in recruitment were somewhat reluctant to our study design, thus unconsciously influencing the woman’s decision to decline participation. Future studies aimed at testing mild approaches for women with compromised ovarian reserve should consider in advance this difficulty. Results of the present study may be of help in this regard since they can be used to persuade physicians and eligible subjects about the appropriateness of alternative, less aggressive approaches. In this study, we planned a two-year period of recruitment in order to maintain the conditions of the studied population as stable as possible. This point is of particular relevance in the field of IVF since notions and practices are in rapid and constant evolution. However, given the difficulties that we encountered, in future trials, one may consider the possibility of extending the recruitment period beyond the conventional limit of two years despite the intrinsic risk of biases associated with prolonged enrolments. Unfortunately, difficulties in recruitment do not allow us to draw definite conclusions regarding the financial aspects. The sensitivity analysis using the limits of the 95%CI of the DRs was not conclusive because at the two extremes, there were relevant economic advantages of one of the two groups. As such, our data has to be intended as preliminary and further study is warranted to clarify whether clomiphene citrate should become the first-line option in women with compromised ovarian reserve.
Some further aspects of the trial deserve to be noted. Firstly, we decided to test clomiphene citrate rather then natural cycle IVF because this is currently the cheapest available option and because we had encouraging results in a preliminary study with this approach . The underlying hypothesis tested is that, in a situation of compromised ovarian reserve, the regimen of stimulation is much less relevant than the patient herself in determining the success of IVF. Natural cycle IVF was a plausible alternative since it may appear at prima facie cheaper but monitoring these cycles is more troublesome because of the risk of premature LH surge and this may actually increase costs. In fact, some authors have suggested employing a natural-cycle-guided IVF adding GnRH antagonists and low-dose gonadotropins during the last days of stimulation to prevent this deleterious event . However, this is inevitably more expensive. Overall, we deemed that clomiphene citrate represented the less expensive option. It is noteworthy that premature LH surge is less likely with this drug because of its long-lasting anti-oestrogen effects . Accordingly, the rate of premature luteinisation was similar in the two arms of our study. The interest of clomiphene citrate in this context is further supported by the observation that the rate of cancellation was identical in the two arms of our trial. In fact, although discarded in clinical practice, IVF using exclusively clomiphene citrate can yield reasonable results. A clinical pregnancy rate per started cycle and per embryo transfer of 18% and 34% have been reported .
Secondly, the choice of a short protocol regimen as a comparator may also be a concern. As mentioned earlier, the most appropriate regimen for poor responders has yet to be established. Evidence on this point is confusing. Nevertheless, there is some data suggesting that the short protocol regimen may be more suitable in poor responders .
Thirdly, the inclusion criteria used to define poor responders may also be a point of debate. Until recently, a shared definition of this condition was lacking. However, in 2010, opinion leaders in this area met within an ESHRE consensus meeting in order to propose a unique definition . The presence of at least two of the following features are now necessary to define poor responders: 1) advanced maternal age (≥40 years) or any other risk factors for poor response;2) previous poor ovarian response (≤3 oocytes retrieved with a conventional stimulation protocol);and 3) an abnormal ovarian reserve test (AFC<5-7 follicles or AMH<0.5-1.1 ng/ml). Two episodes of poor response after maximal stimulation is an additional condition to define poor responders. In our study, we could not use these criteria since they were published after the study was implemented. However, it has to be emphasised that at least 278 women (96%) fulfilled this definition. An additional questionable point in this regard is the use of serum FSH as a unique selection criterion for a subgroup of included subjects. More than one criterion to characterise a woman as a poor responder is generally recommended . However, it has to be pointed out that the use of serum FSH to predict poor responsiveness is arguable because of a high rate of false negatives (normal FSH but poor response) rather than the opposite (high FSH but normal response). Serum FSH is actually a late marker of compromised ovarian reserve. In fact, our decision led us to exclude some women with compromised ovarian reserve rather than including subjects with normal ovarian reserve. The analysis focussing exclusively on women recruited based on elevated serum FSH (see Table 3) confirmed this view. On the other hand, the selection criteria has to be clearly kept in mind in the interpretation of our results and it has to be remembered that inferences should be done exclusively to the specific population in our study.
Fourthly, the study was un-blinded. However, the impact of this potential bias is presumably small because the funding source was impartial to the treatments studied and a relevant role of the placebo effect in this kind of context is questionable . Finally, it is worth mentioning that baseline characteristics of the two groups were similar, thus supporting the view that randomisation effectively prevented biases in allocation in our trial. Only duration of infertility slightly differed. However, this variable was not found to influence the chances of success by logistics regression analysis, thus excluding a critical confounding effect.