DHEA administration in poor responders

Sir, We read with interest the randomized control trial by Wiser et al. (2010) published in Human Reproduction regarding the benefit of administering DHEA to poor responders. Since the conclusion of this trial is that administration of DHEA is associated with a higher probability of live birth, we think it is probably useful to the reader to highlight some important aspects of this trial, which show that the data presented do not seem to support such a conclusion. In the current trial, following randomization, patients were allowed to perform two IVF cycles. In this way, 17 and 16 patients in the DHEA and in the control group performed 26 and 25 cycles, respectively. Such a design makes it necessary to analyze the probability of pregnancy or live birth by using appropriate methods, which in this case is the cumulative probability of pregnancy or live birth as assessed by life table analysis or by Kaplan–Meier survival analysis. A comparison of the probability of pregnancy can be performed by the use of the log rank test (Kaplan–Meier). The authors, however, have used, according to the statistical methods described in their manuscript, the Fisher's Exact test. It should be noted though that, this test requires that the observations are independent from each other, which, is not the case here, since some patients performed two cycles. What is more important is that even by accepting the authors (inappropriate) choice of statistical methods (Fisher's Exact test), the numbers they present in Table IV, do not support their claims of an improved probability of live birth, as shown by a P-level of 0.05. The application of a Fisher's Exact test with the numbers provided in Table IV leads to a P-level of 0.099 and not 0.05 as the authors report. Thus, even when analyzed by using inappropriate statistical methods, the data provided in this report cannot support the conclusion that administration of DHEA is associated with a higher probability of live birth. Adopting DHEA as a beneficial intervention for the management of poor ovarian response should be guided by appropriately analyzed data originating from rigorously designed studies.