This theme develops novel statistical methods to increase study design efficiency across all stages of treatment development and evaluation. Scientific focus will be on the following five areas and supplemented by activities translating the methods into practice.
Master protocols seek to answer multiple research questions within a single trial
We will develop methods for (i) platform trials that consider the uncertainty around the treatments (which to include, how to add or remove treatments) and uncertainty around the endpoints in emerging diseases, and (ii) basket and umbrella trials that allow information borrowing between the trial subsets.
Personalized dose-finding.
Focus of this work will be on intra-patient dose-escalation methods. The designs will make efficient use of information from various cycles of treatments and consider informative data (e.g., continuous outcomes, longitudinal data). Another aspect will be the development of seamless dose-finding designs progressing from a monotherapy trial to a combination trial with proof-of-concept expansion phase.
Response adaptive designs.
(i) Develop novel hypothesis tests with improved efficiency for final analysis.
(ii) Improve robustness including considerations around missing data and time trends.
(iii) Exploit baseline covariates and short-term outcomes in response adaptive designs when the time to observe the primary outcome renders adaptation on it impractical.
Evaluation of algorithms and devices.
When developing a pharmacological intervention, high-quality evidence is required, and commonly agreed standards apply. When developing an algorithm (or device) for use in healthcare, no such standards exist. We will use ideas for adaptive decision making to obtain an acceptable body of evidence around the utility of the tool and seek to overcome the issue of repeated evaluations when the device or algorithm changes.
Trial emulation
Using observational data to emulate a target trial to estimate the causal effect of an intervention is particularly valuable for post-marketing surveillance analyses and to inform future trials. We aim to improve the methods for trial emulation and promote its implementation by developing new software. Additionally, we will explore ‘online’ multiple testing which has potential utility in settings where a large (and unspecified) number of tests are undertaken.
Translation
We will engage with different stakeholders to disseminate new methods and provide training and tools for implementing them in practice. We will seek to increase capacity in efficient study design by developing software for innovative designs as well as providing short courses for methodologists and applied researchers alike.