Estimation of sample size in randomized controlled trials in multiple sclerosis studying annualized relapse rates: A systematic review - Archive ouverte HAL
Article Dans Une Revue Multiple Sclerosis Journal Année : 2021

Estimation of sample size in randomized controlled trials in multiple sclerosis studying annualized relapse rates: A systematic review

Résumé

Background: In multiple sclerosis (MS) studies, the most appropriate model for the distribution of the number of relapses was shown to be the negative binomial (NB) distribution. Objective: To determine whether the sample-size estimation (SSE) and the analysis of annualized relapse rates (ARRs) in randomized controlled trials (RCTs) were aligned and compare the SSE between normal and NB distributions. Methods: Systematic review of phase 3 and 4 RCTs for which the primary endpoint was ARR in relapsing remitting MS published since 2008 in pre-selected major medical journals. A PubMed search was performed on 30 November 2020. We checked whether the SSE and ARR analyses were congruent. We also performed standardized (fixed α/β, number of arms and overdispersion) SSEs using data collected from the studies. Results: Twenty articles (22 studies) were selected. NB distribution (or quasi-Poisson) was used for SSE in only 7/22 studies, whereas 21/22 used it for ARR analyses. SSE relying on NB regression necessitated a smaller sample size in 21/22 of our calculations. Conclusion: SSE was rarely performed using the most appropriate model. However, the use of an NB model is recommended to optimize the number of included patients and to be congruent with the final analysis.
Fichier non déposé

Dates et versions

hal-04721469 , version 1 (04-10-2024)

Identifiants

Citer

Louis Poncet-Megemont, Bruno Pereira, Fabien Rollot, Maria Pia Sormani, Pierre Clavelou, et al.. Estimation of sample size in randomized controlled trials in multiple sclerosis studying annualized relapse rates: A systematic review. Multiple Sclerosis Journal, 2021, 28 (9), pp.1457-1466. ⟨10.1177/13524585211052400⟩. ⟨hal-04721469⟩
0 Consultations
0 Téléchargements

Altmetric

Partager

More