Background/Aims:
Bayesian statistics in stroke trials have become more popular in recent years. Perceived advantages include: 1) incorporating information across treatments, participants subgroups, and prior trial stages, and 2) the ability to estimate Bayesian probability of hypothesized treatment effects. We conducted a scoping review of reported Bayesian stroke trials to describe the extent and motivation to use Bayesian methods and their practical use.
Methods:
Completed in accordance with PRISMA-ScR for scoping reviews. Bayesian stroke trials were identified from clinical trials registries (Australian and New Zealand Clinical Trial Registry and clinicaltrials.gov) and PubMed Central up to date. Data were descriptively collated.
Results:
From 359 records, 20 stroke trials (14 completed, 5 ongoing, 1 proposed) used Bayesian methods. Trials spanned stroke prevention (2), acute (14) and recovery (4), and Phases I-IV of clinical trial development. Fifteen trials used Bayesian adaptive designs. The motivation for Bayesian design was: presumed efficiency gains associated with adaptive designs, ability to make Bayesian confidence statements about treatment effect, and ability to borrow information from other studies to increase power. From 16 studies providing sufficient detail, 14 trials did not substantively incorporate prior information and made decisions predominantly based on the observed data.
Conclusion:
The use of Bayesian methods in stroke trials is typically associated with adaptive designs. The ability to make probability statements about hypothesized treatment effects is the key instrument that enables adaptive decision making. Prior information is incorporated in a manner to have minimal impact on the decisions about the trial conduct and outcomes.