Designing trials in smaller populations
Session type: Oral
How should we approach trial design when the ideal sample size is considered larger than the number of participants that can be recruited in a reasonable time frame? We present an ordered framework for designing randomised trials, staying with the frequentist approach well accepted and understood in large trials, that includes small alterations to the design parameters. The first step should always be to attempt to extend collaborations, consider broadening eligibility criteria and increase the accrual time or follow-up time. The second set of ordered considerations are the choice of research arm, outcome measures, power and target effect. If the revised design is still not feasible, in the third step we propose moving from two- to one-sided significance tests, changing the type I error rate, using covariate information at the design stage, re-randomising patients and borrowing external information. We discuss benefits from some proposals and warn against others. This framework would allow appropriate evaluation of treatments when large-scale phase III trials are not possible, but where the need for high-quality randomised data is as pressing as it is for common diseases.