A Bayesian reevaluation of randomized controlled trials in assisted reproductive technology: quantifying evidence strength for null and alternative hypotheses

Introduction: Many randomized controlled trials (RCTs) in assisted reproductive technology (ART) face widespread misinterpretation of statistically non-significant results within the conventional frequentist framework. This study aimed to reevaluate high-quality ART RCTs using Bayesian methods to quantify the strength of the evidence for both the null and alternative hypotheses. Methods: We systematically searched ART-related RCTs published in JAMA, The Lancet, The BMJ, and NEJM between 2010 and 2025. A total of 35 trials were ultimately included for Bayesian reanalysis, which used Bayes factors (BF₁₀) to evaluate the primary outcomes. Sensitivity analyses were performed to confirm the robustness of our primary findings. Results: The Bayesian and frequentist results were consistent in 85.7% of the studies. Among these consistent results, 62.9% supported the null hypothesis and 22.9% supported the alternative hypothesis. In 11.4% of the studies, non-significant P-values were paired with inconclusive Bayes factors, indicating data insensitivity. Another 2.9% showed significant P-values but inconclusive BF10. Overall, 71.4% of studies reported non-significant primary outcomes, with an increasing trend observed over time. Conclusion: Bayesian analysis offers a useful framework for interpreting "negative results" in ART RCTs. It effectively complements traditional statistics to improve the interpretation of evidence and inform future trial design in ART.