Randomised controlled trials evaluating endometrial scratching: assessment of methodological issues

STUDY QUESTION: Do randomised controlled trials (RCTs) evaluating endometrial scratching suffer from methodological issues including insufficient trial registration, statistical errors or irreproducibility, randomisation errors or miscellaneous issues? SUMMARY ANSWER: The majority of RCTs investigating endometrial scratching have methodological issues. WHAT IS KNOWN ALREADY: A large number of small RCTs investigating the effectiveness of endometrial scratching prior to in vitro fertilisation (IVF) and intrauterine insemination (IUI)/intercourse have reported favourable findings. Subsequently, systematic reviews incorporating these RCTs yielded meta-analyses in favour of endometrial scratching. Endometrial scratching has been widely adopted by infertility specialists around the world. Recently, an international RCT including 1364 women reported no benefit from endometrial scratching before IVF. STUDY DESIGN, SIZE, DURATION: We evaluated several methodological issues of RCTs investigating the effectiveness of endometrial scratching prior to IVF and IUI/intercourse. We identified 25 RCTs for IVF and 12 RCTs for IUI/intercourse with full-text publication. PARTICIPANTS/MATERIALS, SETTING, METHODS: We assessed the RCTs on the following criteria: adequacy of trial registration, statistical issues (description of statistical methods and reproducibility of univariable statistical analysis), excessive similarity or difference in baseline characteristics that is not compatible with chance (Monte Carlo simulations and Kolmogorov-Smirnov test) and miscellaneous methodological issues. MAIN RESULTS AND THE ROLE OF CHANCE: Of 25 RCTs evaluating endometrial scratching prior to IVF, only eight (32%) had adequate trial registration. In total, 10 (40%) RCTs had issues regarding statistical methods. Nine (69%, 13 applicable) RCTs had at least one inconsistency between reported and reproduced univariable statistical analysis for categorical baseline/intermediate characteristics. Statistical results of at least one outcome were not reproducible in 14 (74%, 19 applicable) RCTs. Only two (8%) RCTs had none of the above issues. Suggested by the simulations, these RCTs did not significantly violate the null hypothesis that the baseline characteristics were the results of a properly conducted randomisation process (P = 0.4395). Of 12 IUI/intercourse RCTs, only 2 (17%) had adequate trial registration. In total, five (42%) studies had issues of statistical methods. Inconsistency between reported and reproduced univariable analysis for baseline/intermediate categorical variable(s) was found in four (57%, 7 applicable) RCTs. Statistical analysis was not reproducible for at least one outcome in eight (80%, 10 applicable) studies. All RCTs had at least one of the above issues. These RCTs were inconsistent with the null hypothesis that their baseline characteristics were the results of proper randomised allocation (P = 1.659(*)10(-7)). LIMITATIONS, REASONS FOR CAUTION: We were unable to assess RCTs which were not published as full-text papers. We could not analyse individual participant data to investigate possible reasons for statistical inconsistencies. The method to infer the likelihood of proper random sampling rests on assumptions including independent baseline characteristics, simple randomisation and no publication bias. WIDER IMPLICATIONS OF THE FINDINGS: The methodological issues common to RCTs evaluating endometrial scratching may have biased the results of the trials. Further development and validation of these novel methods may be helpful for the critical appraisal of RCTs.