| Thorsten Reineke at MSOURCE considers adaptive designs within the clinical research arena
For a statistician the planning stage of a clinical study is one of the most challenging phases. Planning issues as well as analysis considerations are discussed with the clinical experts in order to translate the clinical questions into an optimal statistical design. During these discussions the critical question regarding the right sample size will arise sooner or later. Often the answer may not be easy to determine, because essential information to perform the calculation is missing or uncertain. How can these challenges be overcome? Is it really necessary to plan a clinical study with a fixed sample size? Could it be possible to become more flexible?
This article presents two alternative design strategies: sequential and adaptive designs, with a focus on the adaptive approach. The core concepts will be discussed as well as the possibilities and limitations of these flexible approaches to study design.
SAMPLE SIZE IN CLINICAL STUDIES
When planning a clinical study there are many general design issues to be discussed by the study team:
- What should be the primary endpoint?
- Are we aiming at demonstrating superiority, equivalence or non-inferiority?
- How many different arms are needed?
- Is a placebo-control necessary?
- Should the study be conducted in (simple) parallel groups or can the advantages of a crossover design be exploited?
- Will a parametric or a non-parametric model be appropriate?
One important question will ultimately have to be answered by the statistician: how many patients (or subjects) do we need? The answer to this question may not be easy, since it depends not only on the general design issues, but also on the type I error (α ‘consumers’ risk’), the type II error (β ‘sponsor’s risk’), the expected effect size (Δ) and its variability (δ). |
