A systematic review of sample size determination in Bayesian randomized clinical trials: full Bayesian methods are rarely used

Le résumé fourni par la source

BACKGROUND: Utilizing Bayesian methods in clinical trials has become increasingly popular, as they can incorporate prior information into the design, and allow for smaller sample sizes while providing reliable and robust statistical results. Various Bayesian methods for sample size determination are available, and while these methods are well justified and understood, it is unclear how they are being used in practice. This study aims to understand how sample sizes for Bayesian efficacy randomized clinical trials (RCTs) are determined and inform future designs of Bayesian trials. METHODS: A systematic literature review was conducted in May 2023 and updated in July 2025. We included completed RCTs which (a) assessed the efficacy of interventions in humans; (b) utilized a Bayesian framework for the primary data analysis; (c) published in English; and (d) enrolled participants between December 2009 – July 2025. RESULTS: The literature search produced 74,833 records, of which 27,890 were duplicates, and 46,943 were screened using manual and automated screening. 283 full texts were screened and 164 studies moved to extraction. Our findings demonstrate a slow increase in RCTs using Bayesian methods to analyse primary efficacy data from 2012 onwards, with a sharp increase during the COVID-19 pandemic (42%). The most common method for sample size determination in Bayesian RCTs was a hybrid approach (58%) in which elements of Bayesian and frequentist theory are combined. Bayesian RCTs predominantly took place in North America (34%) and mainly focused on adult study populations (85%). Bayesian trials were used in a variety of disease areas; the most common being COVID-19 (31%). CONCLUSION: Fully Bayesian methods for sample size determination are rarely used in practice, despite significant theoretical development. Our review revealed a lack of standardized reporting across Bayesian RCTs, making it challenging to review the sample size determination. The CONSORT statement indicates that RCTs must report sample size calculations; adhered to by only 84% of included RCTs. Among RCTs that reported sample size determination, relevant information was frequently omitted from reports and discussed in poorly structured supplementary materials. Thus, there is a critical need for greater transparency, standardization and translation of relevant methodology in Bayesian RCTs.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.