To the Editor:
I read with interest the article by Doody, et al (1). The literature support for the need for luteal support in IVF was presented by Daya and Gunby.(2) A recent reanalysis (3) of this meta-analysis suggests the use of vaginal progesterone to be as effective as IM Progesterone. In order to better understand this publication I seek a few points of clarification.
First, the trial was designed as a “non-inferiority” trial with a lower boundary of 10% on the difference in the ongoing pregnancy rates between Endometrin® and Crinone® usage. In Table 2 the lower boundary of the 95% confidence interval for the difference between Endometrin BID and Crinone on the primary endpoint of ongoing pregnancy is a minus 10.3%. Exceeding this defined limit (10.3% vs. 10%) makes the conclusion unfounded, as would a claim of statistical significance be in a superiority trial with any p-value greater than 0.05.
Second, in the ‘clinicaltrials.gov’ listing for this trial, it was stated that “This…study will be performed in approximately 990 healthy females…” The publication report is on “1211 ART patients.” An explanation for the 22% (1211 vs. 990) increase in sample size is needed.
Third, are there deviations from CONSORT guidelines (4,5) since the disposition of subjects is not provided? Both the manuscript and the clinicaltrials.gov list the study dates as starting June 2005 and completion May 2006. In reporting live birth rates it seems unlikely that the last patients initiating luteal support and enrolled would have been completed prior to September, 2005. Was “live birth rate” a post-hoc analysis? The manuscript states that live birth data was obtained by mail or telephone contact. How many lost-to-follow-up cases were there?
Fourth, “assessor-blinded” should be defined. While performing TVUS, the sonographer would routinely see the type of vaginal formulation.
Fifth, the manuscript reports “These data demonstrate that the use of Endometrin for luteal support in ART patients results in excellent single-cycle pregnancy rates ….” This would seem to be a misleading post-hoc analysis of the data, since the study was designed as a non-inferiority trial to another FDA approved product and in several planned but undisclosed analyses there were remarkable differences.
Lastly, in the FDA approved PI for Endometrin (6), planned analyses not described in the manuscript are included in the Clinical Studies Section (page 4, section 14). They show that the trial failed to prove Endometrin BID and TID are as effective as Crinone QD in patients age >34 and in patients with FSH between 10 and 15 IU/L. In these subgroups the lower boundary of the 95% confidence interval was -20.3 versus -14.9 and -31.0 versus -26.6 for the age and FSH subgroup comparisons of Endometrin BID and TID, respectively). The Endometrin PI states that “Efficacy in women 35 years of age and older has not been clearly established” and “In women age 35 and older and in women with serum FSH levels between 10 and 15 IU/L, the results with respect to ongoing pregnancy rate for both dosing regimens of Endometrin did not reach the criteria for non-inferiority.”
1. Doody KJ, Schnell VL, Foulk RA, Miller CE, Kolb BA, Blake EJ, Yankov VI. Endometrin for luteal phase support in a randomized, controlled, open-label, prospective in-vitro fertilization trial using a combination of Menopur and Bravelle for controlled ovarian hyperstimulation. Fertil Steril. 2008 Mar 25. [Epub ahead of print]
2. Daya, S. and Gunby, J Luteal Phase support in assisted reproduction cycles (review) The Cochrane Library 2006; CD004830 (1) 1-125.
3. Zarutskie PW, Phillips JA. Re-analysis of vaginal progesterone as luteal phase support in assisted reproduction cycles. Fertility & Sterility; vol 88, suppl 1, P10, September 2007.
4. Mohr D, Schulz KF, Altman DG, for the CONSORT Group. The CONSORT Statement: Revised Recommendations for Improving the Quality of Reports of Parallel-Group Randomized Trials. Annals of Internal Medicine, 17 April 2001, vol 134, no 8, pg 657-662.
5. CONSORT Transparent Reporting of Trials Website. http://www.consort-statement.org/index.aspx?o=1011. Accessed 23 April 2008.
6. FDA Approved Endometrin Label: http://www.fda.gov/cder/foi/label/2008/022057s001lbl.pdf accessed April 23, 2008.
Published online in Fertility and Sterility: doi 10.1016/j.fertnstert.2008.12.133
The Authors Respond:
We are pleased that Dr. Zarutskie’s reanalysis of a meta-analysis, supported by a grant from Columbia Laboratories, concluded that the use of vaginal progesterone is as effective as IM progesterone. We offer the following points of clarification regarding the pivotal trial that was used to support FDA approval of Endometrin®:
1. The defined limit in ongoing pregnancy rates between Endometrin® and Crinone® was set at 10%, not 10.0%, so statistical significance was met, as reflected by the FDA approval.
2. As stated in the statistical section of the manuscript, “a sample size of no less than 330 evaluable patients per treatment group” would be required for a 30% pregnancy rate. This sample size would “increase as the pregnancy rate improves.” The higher sample size reflects the higher expected pregnancy rate and the unexpected speed of enrollment.
3. The initial analysis compared ongoing pregnancy rates, but live birth data analysis (required by the FDA) was always planned. The manuscript was prepared after all birth outcome data were available, so for completeness those analyses were included in the paper. Of the 1211 patients, eight were lost to follow-up: three in Endometrin BID; two in Endometrin TID; and three in the Crinone arm.
4. Assessor-blinded refers to the person performing the ultrasound. To our knowledge, vaginal sonography will not distinguish between these two products.
5. Comparison of the study pregnancy rates with SART data for 2005 (the year the study was conducted) found superior results in the study, described as “excellent single-cycle pregnancy rates.” This is especially meaningful, since the study minimized patient selection bias by having very generous inclusion criteria. There were no other “planned, but undisclosed analyses” considered, and there is no attempt or desire to withhold any results.
6. We acknowledge that the data presented in Endometrin’s package insert is inconsistent with what we report from the study. This trial was powered to look at pregnancy rates for all enrolled women combined. It was never intended or powered, a priori, for evaluation of any subgroups (such as age or FSH). The FDA performed a post-hoc analysis after the trial was completed and the data gathered, using age brackets that had not been specified in the trial. Since sub-analyses were not powered for, or anticipated, there is no validity to conclusions for these small sub-sample slices. Thus, the overall study was positive on all parameters for which it was powered for which analyses were intended.
Emily J. Blake, M.D.
Ferring Pharmaceuticals Inc.