RESUMO
BACKGROUND: A substantial minority of women undergoing IVF will under-respond to controlled ovarian hyperstimulation. These women-so-called 'poor responders'-suffer persistently reduced success rates after IVF. Currently, no single intervention is unanimously accepted as beneficial in overcoming poor ovarian response (POR). This has been supported by the available research on POR, which consists mainly of randomized controlled trials (RCTs ) with an inherent high-risk of bias. The aim of this review was to critically appraise the available experimental trials on POR and provide guidance towards more useful-less wasteful-future research. METHODS: A comprehensive review was undertaken of RCTs on 'poor responders' published in the last 15 years. Data on various methodological traits as well as important clinical characteristics were extracted from the included studies and summarized, with a view to identifying deficiencies from which lessons can be learned. Based on this analysis, recommendations were provided for further research in this field of assisted conception. RESULTS: We selected and analysed 75 RCTs. A valid, 'low-risk' randomization method was reported in three out of four RCTs. An improving trend in reporting concealment of patient allocation was also evident over the 15-year period. In contrast, <1 in 10 RCTs 'blinded' patients and <1 in 5 RCTs 'blinded' staff to the proposed intervention. Only 1 in 10 RCTs 'blinded' ultrasound practitioners to patient allocation, when assessing the outcome of early pregnancy. The majority of trials reported an intention-to-treat analysis for at least one of their outcomes, with an improving trend in the recent years. Substantial variation was noted in the definitions used for 'poor responders', the most popular being 'low ovarian response at previous stimulation'. The preferred cut-off value for defining previous low response has been 'less or equal to three retrieved oocytes'. The most popular tests used for diagnosing diminished ovarian reserve have been antral follicle count and FSH. Although the Bologna criteria for POR were only recently introduced, they are expected to become a popular definition in future 'poor responder' trials. Numerous interventions have been studied on 'poor responders'. Most of these have been applied before/during controlled ovarian hyperstimulation. The antagonist protocol, the microdose flare protocol and the long down-regulation protocol have been among the most popular interventions. The analysis of outcomes revealed a clear improving trend in reporting live birth. In contrast, only 10% of RCTs reported significant improvement in reproductive outcomes among tested interventions. Twelve 'significant' interventions were reported, each supported by a single 'positive' RCT. Finally, trials of higher methodological quality were more likely to have been published in a high-impact journal. CONCLUSIONS: Overall, the majority of published trials on POR suffer from methodological flaws and are, thus, regarded as being high-risk for bias. The same trials have used a variety of definitions for their poor responders and a variety of interventions for their head-to-head comparisons. Not surprisingly, discrepancies are also evident in the findings of trials comparing similar interventions. Based on the identified deficiencies, this novel type of 'methodology and clinical' review has introduced custom recommendations on how to improve future experimental research in the 'poor responder' population.