The learning resources hub is a central place for clinicians, investigators, community members, funders and ethics committee members to access educational materials about adaptive clinical trials and resources on Bayesian Network models.
For access to these resources, please click on the links below:
Sample size for adaptive trials
The sample size is the patient recruitment target that the trial needs to avoid an inconclusive result. This sample size calculation is typically based on a minimum 80% power. This means that if researchers successfully enrol the target sample size, the researchers have an 80% probability of detecting a true difference between the trial groups. It is important to ensure that there are enough participants to detect a clinically meaningful effect (e.g. reduction of hospital stay by 24 hours). Usually, researchers need to enrol large numbers of participants to increase the precision of the results and to minimise the risk that any observed differences have occurred by chance. Sample size calculations require estimates of the underlying rate of disease, the expected treatment effect and variability in the population being studied. Inaccurate estimation can result in researchers undershooting (recruiting too few participants to answer the research question) or overshooting (recruiting too many patients); both scenarios have ethical implications.
Designing a trial with an adaptive sample size means that we are more likely to enrol just enough participants to answer our research question (a ‘Goldilocks’ sample size). All high-quality trials have stopping rules. These rules may include stopping the trial earlier than expected for success (if we have answered the research question) and stopping for futility (when we can conclude the treatments are all the same, ineffective or we are unlikely to answer the research question within the resources available). Other stopping rules may be based on safety, recruitment and research costs.
Stopping the trial early for success or futility is more ethical (reducing continued exposure to ineffective or unsafe treatment), saves research costs (funding is extremely competitive) and participant and researcher time.
Adaptive clinical trials allow researchers to embed stopping rules in a series of interim analyses (timepoints when the accumulating data are analysed) to ensure trials can stop early for success or futility. We use statistical methods (mathematical formulae) to estimate the probability of a treatment effect based on the accumulating data and use pre-defined thresholds for deciding whether to continue or stop enrolment. In adaptive clinical trials, researchers present a plausible sample size range in their protocol (e.g. 500 to 5000 patients), which is often based on thousands of computer simulations of the trial design. Trial simulations are data generated assuming different trial outcomes, recruitment rates, disease rates and other factors.