Introduction
Determining the sample size for adaptive designs is more complex than for fixed designs, as it requires additional information and considerations. Transparent reporting at the planning stage enables reviewers to understand both the rationale behind the sample size and the adaptive features, facilitating trial planning, conduct, and interpretation. However, existing guidance provides limited support for reporting in trial grant applications and protocols.
Research Question
This thesis addressed the question: How should sample size be reported in trial grant applications and protocols for randomised adaptive designs?
Methods
A pragmatic, mixed-methods approach was adopted to integrate evidence, interpret current practices, and develop practical solutions. The theoretical and methodological foundation was established through a scoping and systematic review of methodological literature, reporting guidance, and adaptive trial documents. Semi-structured interviews with clinical trial stakeholders were conducted to explore their experiences. A two-round Delphi study, followed by a consensus meeting, was conducted to develop and refine a checklist providing structured guidance for sample size reporting. Each checklist item was then explained and elaborated. Finally, the checklist was piloted using two hypothetical adaptive trial examples to assess its applicability.
Results
Stakeholders agreed that reporting should be transparent and reproducible, with justification for all assumptions and parameters. While reporting was generally sufficient for simpler designs such as group sequential trials, it was often incomplete for more complex designs (e.g., multi-arm multi-stage trials). Adaptive-specific aspects, particularly those concerning operating characteristics and sample size variability introduced by adaptations, were frequently underreported due to a lack of practical guidance. The Delphi process achieved consensus on a 14-item checklist, which was found to be applicable across various adaptive design types and suitable for both grant applications and protocols.
Conclusions
The proposed checklist supports comprehensive, transparent, and reproducible reporting of sample size for adaptive designs, enhancing methodological rigour and interpretability.
