To the Editor:
I read with great interest and attention the original article published recently by Griesinger et al. (1) entitled “Progesterone elevation does not compromise pregnancy rates in high responders: a pooled analysis of in vitro fertilization patients treated with recombinant follicle-stimulating hormone/releasing hormone antagonist in six trials.” In this article, Griesinger et al. conclude that a progesterone (P) elevation > 1.5 ng/mL on the day of hCG administration is associated with a lower ongoing pregnancy rate (OPR) in the general population, but not in patients with a high ovarian response, defined as an oocyte yield > 18 oocytes.
These authors performed excellent statistical work, pooling data from six multicenter randomized controlled trials relatively similar in their design, and presented the results as if all came from one single study. However, the study deserves several comments that question the unequivocal conclusion that is presented even in the title of the manuscript.
First, this study attempts to demonstrate the lack of impact of a particular event (P elevation on the day of hCG) in a particular population (high responders), from data obtained from trials in which this type of patient was excluded in most cases. Indeed, and as the authors explain, patients with cycles of more than 35 days and/or basal LH > 12 IU/L (that is, polycystic ovary syndrome patients) were not included in these studies. Moreover, in two of the studies, patients with an antral follicle count above 20—and therefore at risk of having a high response—were specifically excluded.
The consequence of this situation is that only unexpected high responders could be included in the sub-analysis, and therefore the sample is expected to be small.
Unfortunately, the authors do not describe how many patients were high responders. However, a careful analysis of the data provided allows estimating that there were a total of 195 cases (10.4% of total), of which 37 (19%) had high P on the day of hCG. Since the ongoing pregnancy rate in women with P<1.5 ng/mL was 33.2%, and the odds ratio of OPR for patients with high P was 0.55, the statistical power of this study for this sample size is only 19.3%, considering a confidence level of 95% (error alfa=0.05). In other words, the likelihood that the findings of this study are due to random is 80.7%.
Our group published an analysis of this event in a series of more than 4,000 cycles (2). The impact of elevated P on the OPR was almost identical to the one observed in the present study (OR: 0.53; 95% CI: 0.38-0.72). Nevertheless, the stratified analysis according to oocyte yield showed that OPR was lower when P was high in all the strata of ovarian response. If we perform a sub-analysis including only those patients in whom more than 18 oocytes were obtained (n=578, 14.3%), we observe that P was high in 94 of them (16.3%). The OPR of these patients was 18.1%, versus 40.3% in patients with P <=1.5 ng/mL (OR: 0.33; 95% CI: 0.19-0.57; p<0.001). The statistical power of this analysis for a 95% confidence level is 76%, and therefore much more reliable than that of the study under comment.
This result should not be surprising, as two studies have shown that in the presence of high P at the end of the follicular phase, there are significant variations of the endometrium gene profile described from the day of oocyte retrieval (3) to the implantation window (4) that impair endometrial receptivity.
After all, the clinical relevance of this analysis is small. Once accepted that in case of high P on the day of hCG the recommendation is to freeze all the embryos and transfer them in a subsequent natural or artificial cycle, this practice should not be different in high responders in whom there are other stronger reasons (such as the risk of ovarian hyperstimulation syndrome) to freeze all the embryos.
Finally, and based on the results of their study, Griesinger et al. suggest that trying to prevent an increase of P on the day of hCG has very little impact on the overall pregnancy rate. Again, the particular population included in the studies, basically young and normo-ovulatory and therefore of good prognosis patients, influences this assumption, since the probability of having high P in this population is low (8.4% in this study). However, other studies have shown an incidence of high P on the day of hCG up to 38%, especially if high doses of FSH are used for stimulation (5). Being the OR of OPR of 0.55 when P is high, the impact of this event should not be underestimated. To evaluate the impact on the general population, the analysis has to be performed on the general population, and not only in good prognosis patients.
In summary, the conclusions of this study should be taken with caution, as the characteristics of the population included and the small sample size of the target subpopulation of the study could underestimate the real impact of elevated P in high responders. Anyhow, it cannot be dismissed that good quality embryos—usually available in high responders—could implant in an endometrium with reduced receptivity, and therefore, the impact on OPR would be lower, or only be observed at higher P levels than in the general population. However, this hypothesis needs to be demonstrated in large cohort studies, as the data currently available are insufficient.
Ernesto Bosch, M.D.
Instituto Valenciano de Infertilidad
1. Griesinger G, Mannaerts B, Andersen CY, Witjes H, Kolibianakis EM, Gordon K. Progesterone elevation does not compromise pregnancy rates in high responders: a pooled analysis of in vitro fertilization patients treated with recombinant follicle-stimulating hormone/gonadotropin-releasing hormone antagonist in six trials. Fertil Steril. 2013 Sep 28.[Epub ahead of print].
2. Bosch E, Labarta E, Crespo J, Simón C, Remohí J, Jenkins J, et al. Circulating progesterone levels and ongoing pregnancy rates in controlled ovarian stimulation cycles for in vitro fertilization: analysis of over 4000 cycles. Hum Reprod 2010;25:2092-2100.
3. Van Vaerenbergh I, Fatemi HM, Blockeel C, Van Lommel L, In’t Veld P, Schuit F, et al. Progesterone rise on HCG day in GnRH antagonist/rFSH stimulated cycles affects endometrial gene expression. Reprod Biomed Online 2011;22:263-71.
4. Labarta E, Martínez-Conejero JA, Alamá P, Horcajadas JA, Pellicer A, Simón C, et al. Endometrial receptivity is affected in women with high circulating progesterone levels at the end of the follicular phase: a functional genomics analysis. Hum Reprod 2011;26:1813-25.
5. Bosch E, Valencia I, Escudero E, Crespo J, Simón C, Remohí J, et al. Premature luteinization during gonadotropin-releasing hormone antagonist cycles and its relationship with in vitro fertilization outcome. Fertil Steril 2003;80:1444-49.
Published online in Fertility and Sterility doi:10.1016/j.fertnstert.2013.10.036
The authors respond:
We would like to thank Dr. Bosch and coworkers for their interest in our paper (1) and for providing constructive criticism. Our study showed that late follicular phase progesterone (P) elevation was associated with a reduced chance of ongoing pregnancy although this phenomenon was not present in the subset of patients with a high ovarian response (>18 oocytes retrieved).
Bosch et al. state that we present “the results [from six multicenter randomized trials] as if they all came from one single study.” This is not accurate since, as explicitly outlined in the Materials and Methods section, recognized statistical models based on appropriate meta-analytical approaches were applied.
Furthermore, concerns are expressed regarding the exclusion of expected high responders from the studies pooled in the meta-analysis by Griesinger et al. (1), which might affect the results obtained due to the underrepresentation of high responders in the population analyzed. Although, this might be true, it is clear that both in our study (1) and in that by Bosch et al. (2), the incidence of patients with >18 oocytes is quite similar, for example, 12.4% (1) vs. 14.3% (2). Moreover, an explanation is lacking why the association of P elevation with pregnancy likelihood would differ in expected vs. unexpected hyper responders.
Bosch et al. express their concern as to whether the group of high responders in our study was large enough to allow for solid conclusions to be drawn regarding the association of P elevation and the likelihood of pregnancy. Using the odds ratio of our study, they calculate the incidence of patients with high response and those with elevated P within the high responders and in turn estimate the theoretical chance of a β error in our findings, concluding that “the likelihood that the findings of this study are due to random is 80.7%.” However, the number of high responders cannot be estimated from an adjusted OR stemming from a multivariate analysis. As a consequence, their calculations are incorrect. More importantly, calculating the statistical power from a retrospective analysis (that is, after clinical trials have been conducted) in order to interpret the results obtained is fundamentally flawed (3). The interpretation of data analysis in such circumstances should be based on interval statistics: in our study (restricted to subjects with embryo transfer), 19 of the 44 high responders (>18 oocytes) with high P (43.2%) and 71 of the 181 high responders without high P (39.2%) had an ongoing pregnancy. This corresponds to an estimated OR (95% confidence interval) of 1.18 (0.60 to 2.29) for high P vs. non-high P (1). In the paper by Bosch et al. (2) the estimated OR (95% CI) for high responders (>18 oocytes) was 0.33 (0.19 to 0.57) for high P vs. non-high P.
Bosch et al. suggest that the conclusion of their study—a negative association of elevated P with ongoing pregnancy likelihood in hyper responders—is more valid than that drawn from our study, due to a larger sample size. Although it is true that a large sample size typically strengthens the conclusions drawn from a study, the validity of the conclusions does not depend solely on that parameter. It is also dependent on the quality of the data analyzed. In that respect, it should be noted that in contrast to our study, which analyzed data from rigorously conducted prospective clinical trials, the data analyzed by Bosch et al. are data that were collected during routine clinical practice, using a multitude of stimulation protocols and FSH dosages, LH activity, hCG triggering timepoints and, assumably, luteal phase support, use of adjunctive therapies, and early pregnancy management. More importantly, in contrast to our analysis, which was based on individual patient data (for example, one patient, one cycle, one observation), in their study (2) patients could contribute more than one observation to the data set. In this way, one of the fundamental prerequisites, the independence of observations, for using the statistical tests employed by Bosch et al. to arrive at their conclusions, is violated; for example, the outcome of a first cycle could be taken into account when planning subsequent cycles in terms of treatment strategy. Unfortunately, this imposes serious limitations to the conclusions of Bosch et al., which cannot be alleviated by the large sample size of their study.
The discrepancy in the results obtained between the two studies might be explained by the conclusions of a recently published large meta-analysis on the effect of P elevation on the probability of pregnancy (4), which took into account different P thresholds and different patient populations. In that meta-analysis, a detrimental effect of elevated P was present even when a very low threshold for defining P elevation (0.8 to 1.1 ng/ml) was used, both in the general IVF population and in the poor responders. However, in high responders, such an effect was present only when the level of serum P on the day of hCG reached 1.9–3.0 ng/ml.
We advocate that the findings from our study caution against the use of one single progesterone threshold across all response groups to predict outcome and to potentially modify management. Our finding should, however, be consolidated by other researchers.
Georg Griesinger, M.D.;a Bernadette Mannaerts, Ph.D.;b Claus Yding Andersen, D.M.Sc.;c Han Witjes, Ph.D.;b Efstratios M. Kolibianakis, M.D.;d and Keith Gordon, Ph.D.e
aDepartment of Reproductive Medicine and Gynecological Endocrinology, University Clinic of Schleswig-Holstein, Luebeck, Germany; bMSD, Oss, The Netherlands; cLaboratory of Reproductive Biology, University Hospital of Copenhagen, Copenhagen, Denmark; dUnit for Human Reproduction, Medical School, Aristotle University of Thessaloniki, Thessaloniki, Greece; eMerck Sharp & Dohme Corp., Whitehouse Station, New Jersey, USA
1. Griesinger G, Mannaerts B, Andersen CY, Witjes H, Kolibianakis EM, Gordon K. Progesterone elevation does not compromise pregnancy rates in high responders: a pooled analysis of in vitro fertilization patients treated with recombinant follicle-stimulating hormone/gonadotropin-releasing hormone antagonist in six trials. Fertil Steril 2013 Sep 28. doi:pii: S0015-0282(13)03011-2.
2. Bosch E, Labarta E, Crespo J, Simón C, Remohí J, Jenkins J, et al. Circulating P levels and ongoing pregnancy rates in controlled ovarian stimulation cycles for in vitro fertilization: analysis of over 4000 cycles. Hum Reprod 2010;25:2092-100.
3. Hoenig JM, Heisey DM. The abuse of power: The persuasive fallacy of power calculations for data analysis. The American Statistician. 2001;55:1–6.
4. Venetis CA, Kolibianakis EM, Bosdou JK, Tarlatzis BC. Progesterone elevation and probability of pregnancy after IVF: a systematic review and meta-analysis of over 60,000 cycles. Hum Rep Update 2013;19:433-57.
Published online in Fertility and Sterility doi:10.1016/j.fertnstert.2013.10.037