Association between trial registration and reporting large treatment effects in randomized controlled trials of infertility interventions: a meta-epidemiological study

IMPORTANCE: While prospective registration is a key measure to prevent selective reporting, excluding unregistered or retrospectively registered trials would greatly reduce the statistical power of systematic reviews. OBJECTIVE: We aimed to explore the association between trial registration status and the reporting of large treatment effects on pregnancy or live birth. EVIDENCE REVIEW: This is a meta-epidemiologic study of infertility randomized controlled trials (RCTs) published between 2012 and 2023, identified through systematic searches of Embase, Medline, and CENTRAL. Eligible RCTs involved infertile couples and reported outcomes on biochemical, clinical, ongoing pregnancy, or live birth. Conference abstracts, secondary analyses of RCTs, or trials with unclear registration timing were excluded. The primary outcome was reporting of large treatment effects, defined as a statistically significant relative risk for pregnancy or live birth below 0.80 or above 1.25. The association between trial registration status and reporting of large treatment effects was analyzed using a Poisson regression with a log link, adjusted for potential confounders, with prospective registration as the reference group. FINDINGS: Among 1,369 infertility RCTs, 19.8% were prospectively registered, 29.6% were retrospectively registered, and 50.6% were unregistered. Compared with prospectively registered trials, there was no statistically significant difference in the likelihood of reporting large treatment effects among retrospectively registered trials (adjusted relative risk [aRR] = 0.77; 95% confidence interval [CI]: 0.49-1.20) or unregistered trials (aRR = 1.07; 95% CI, 0.71-1.66). These findings remained consistent in analyses restricted to trials published after 2018. In trials published before 2018, retrospective registration was associated with a significantly lower likelihood of reporting large treatment effects (aRR = 0.40; 95% CI, 0.21-0.75). CONCLUSION AND RELEVANCE: The reporting of large treatment effects has not been shown to be greater in unregistered or retrospectively registered infertility trials. Our findings do not support the blanket exclusion of all unregistered or retrospectively registered RCTs from systematic reviews.