In clinical trial design, adaptive methods are gaining traction for their potential to optimize resource allocation and improve trial efficiency. A recent study published in BMC Medical Research Methodology explores sample size recalculation within the framework of three-stage clinical trials, offering insights into how interim results can inform adjustments to the remaining sample size.
The study focuses on comparing an intervention group to a control group, assuming normally distributed endpoints with a common variance. The goal is to test the alternative hypothesis that the intervention is superior to the control. The researchers employ a two-sample t-test statistic at each stage of the trial to assess the evidence for a treatment effect.
Statistical Methods
The core of the methodology involves recalculating the sample size based on interim results. Specifically, the conditional rejection probability (CRP) is calculated. CRP represents the probability of rejecting the null hypothesis at a future stage, given the data observed up to the current interim analysis. This calculation is crucial for determining whether the trial is on track to demonstrate efficacy or if adjustments to the sample size are necessary.
Conditional power (CP) is also a key concept. It considers the probability of rejecting the null hypothesis in the remainder of the trial, taking into account both the second and third interim analyses. The researchers emphasize that uncertainty in the effect size must be considered when using conditional power for recalculation rules.
Trial Designs
The study contrasts two main types of trial designs: group sequential designs and designs with sample size recalculation. In group sequential designs, the sample sizes for all three stages are fixed in advance. In contrast, designs with recalculation allow the sample sizes for the second and third stages to be adjusted based on the interim results.
The authors focus primarily on recalculation at the first interim analysis, noting its similarity to sample size calculation in a two-stage adaptive trial. They highlight that in a three-stage trial, the remaining sample size is stochastic, influenced by whether the trial continues to the third stage.
Impact of First-Stage Sample Size
The choice of the first-stage sample size (n1) is shown to significantly impact both the sequential testing procedure and the potential benefits of recalculation. A small n1 may provide limited information for sample size adjustment, while a large n1 may reduce the flexibility to adapt. The researchers note that there is currently no consensus on the optimal choice of sample sizes for interim analyses in three-stage trials.
Implications for Clinical Trials
This research provides valuable insights into the design and analysis of three-stage clinical trials with sample size recalculation. By carefully considering the conditional rejection probability and conditional power, researchers can make informed decisions about adjusting sample sizes during the trial, potentially improving the efficiency and success rate of clinical investigations.
