Answers are grounded in this paper and cite their sources.
Sir, We read with great interest the retrospective cohort study by Lainas et al. (2015), comparing modified natural cycle IVF (MNC) with high-dose FSH (HDFSH) in poor responders according to the Bologna criteria. The authors found that the probability of live birth was significantly higher in the MNC group when compared with the HDFSH group after adjusting for different confounders (OR: 4.01, 95% CI: 1.14–14.09), and concluded that MNC-IVF offers a higher probability of live birth than ovarian stimulation with HDFSH in poor responders who do not wish to undergo oocyte donation. Poor responders are a special difficult group of patients and no clear conclusion has been drawn on which regimen may constitute the ideal therapeutic approach (Polyzos and Devroey, 2011). Although the natural cycle or MNC IVF appear to be attractive options, evidence is still sparse, as published data on the effectiveness of these protocols is limited (Kolibianakis et al., 2004: Morgia et al., 2004; Polyzos et al., 2012). The current study (Lainas et al. 2015) represents a large cohort analysis of poor responders with interesting findings. The authors, by using three different logistic regression models and presenting only adjusted odds ratios, demonstrate a significantly higher live birth rate per cycle in the MNC treated patients compared with the HDFSH treated patients. In the first model adjustments were performed for four independent variables including the type of the protocol (MNC versus HDFSH), patients' age, basal FSH levels and cause of infertility, whereas in the two other regression models five predictors were used by adding one more variable: the number of embryos transferred or the number of good quality embryos transferred. Nevertheless, although the conclusion of the authors appears to be relatively robust, suggesting that MNC results in statistically significant higher live birth rates compared with HDFSH, we have to highlight some important methodological limitations that should raise caution and certainly do not allow us to draw such a conclusion. Although we totally agree that in retrospective studies with heterogeneous populations, as the study by Lainas et al., adjusted odds ratios are preferable compared with unadjusted results, it is well established that the reliability of the adjusted odds ratios derived from logistic regression models is strictly dependent on the number of events per variable included in the model. Logistic regression is commonly used in health research and it is important to safeguard that the parameter estimates can be trustworthy. A common problem occurs when the outcome has few events; in such a case, parameter estimates may be biased or unreliable. Although, current epidemiological evidence suggests that in many situations, logistic regression modeling may pose substantial problems when the number of events per variable is relatively small (Courvoisier et al., 2011), even at the most optimistic scenarios, the number of events per variable should at least be greater than 5 or 10 (Vittinghoff and McCulloch, 2007) in order to reach reliable conclusions. In the study by Lainas et al. the number of events (live births) per variable in the regression models not only does not exceed 10, but on the contrary it is less than 5 in all of the models presented (only 17 live births for 4 or 5 predicting variables). In particular, live birth was achieved in 12 out of 161 MNCs (7.5%) and in 5 of 164 HDFSH antagonist cycles (3.1%). Although the authors preferred to use logistic regression and presented adjusted odds ratios of live births, despite the severely limited number of events per predictor variables, unadjusted odds ratios are not at all presented. Based on the proportions of live births, we calculated the unadjusted odds ratios for MNCs versus HDFSH and the result was not significant (OR: 2.5, CI 0.8–7.4, P = 0.08). Consequently, taking into account the lack of significance in the unadjusted odds ratio and the questionable results of the adjusted analysis, owing to the severely limited number of events (live births) per variable, we believe that the data are not at all sufficient in order to drive such conclusions. For this reason, we disagree with the authors' conclusion that live birth rates are significantly higher in MNCs compared with HDFSH GnRH antagonist cycles in poor responder patients. Although their results appear to favor the MNC group, we believe that the conclusion and the summary answer of the article should be reconsidered, since neither the number of events (live births) nor the analysis of the data justify that MNC offers a higher probability of live birth compared with ovarian stimulation with HDFSH in poor responders. MNC may have a role in the treatment of poor ovarian responders; nevertheless in the absence of robust evidence, high quality and adequately powered cohort and randomized trials are needed prior to guiding clinicians towards the adoption of MNC as a superior treatment modality for the treatment of these patients, especially if we consider that previous studies (Kolibianakis et al., 2004) were completely contradictory concluding that MNC does not offer a realistic chance of parenthood in poor responders.
No takes yet. Share an insight, caveat, or question.
Polyzos et al. (2015) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: