Describing an adaptive design
A trial whose design adapts to emerging data asks more of its protocol: its planned changes have to be described across the whole document.
What an adaptive design is
An adaptive design changes aspects of a trial on the basis of emerging data analysed at one or more interim analyses, and it does so without undermining the trial's validity and integrity. Its changes are made by design rather than ad hoc, so adaptation is a design feature meant to enhance the trial, not a remedy for inadequate planning.
Common adaptations re-estimate the sample size by comparing the treatment effect seen at the interim analysis with the effect anticipated at the start, abandon treatments or doses, change the allocation ratio between trial arms, enrich the patient population, or stop the trial early for success or for lack of efficacy. Changing the allocation ratio can favour the arms that show better results, and patients might be quite interested in that. Many adaptations are possible, among them changes to the dose, the treatment, the randomisation, the sample size estimation, the early stopping rules, the hypothesis, the primary outcome variable, the eligibility criteria or the statistical analysis plan. Each adaptation has its pitfalls, so an adaptive design requires careful and extensive planning.
One protocol writer held in 2023 that adaptive designs had become more embedded in the clinical research landscape. In the same writer's view, they are used more where the number of participants exposed to a treatment has to be minimised, and in rare diseases, where the population is limited and the information received has to be used as fully as possible.
Describing the adaptations in the protocol
Although its underpinnings are statistical, an adaptive design is a matter of content in the protocol. It has to be described clearly throughout the protocol, not mentioned briefly in one section with a note that an interim analysis will be done and the sample size may change.
One protocol writer holds that ethics committees by and large accept adaptive designs, provided there is adequate justification and a reason for adapting, the design is well done, and it is well described. In that writer's view, medical writers can support that acceptance by making the description clear and simple to understand. Its reader does not necessarily need to understand the statistical reasons for the adaptations, which are set out in the statistical analysis section or elsewhere.
The protocol makes plain when the interim analysis will happen, who will look at the data and what measures will keep the analysis from biasing the study. It also says whether the reviewing committee will be independent or part of the sponsor, whether it will have a charter, and what its roles and remit will be, including whether it can decide anything or works to specific rules. Stopping for success, for example, has to be very tightly controlled and predefined ahead of any interim analysis, so that the approach is not biased.