Adaptive clinical trials

Explain, in simple lay language, the outcomes and significance of the research. You should include what the research is trying to achieve and why this is important, and brief notes on the context of the research and the methods (see Guidance Notes).
Classically designing a clinical trial involves making assumptions about various attributes, such as the treatment effect, before any data are collected. The aim of our programme is to introduce novel designs that are more flexible. The rigidity of the classic design can lead to failed studies due to wrong assumptions such as using an optimistic treatment effect to limit the size of the study. Our novel designs will allow the use of accrued information in many ways; this includes stopping the study early for futility or efficacy, re-assessing the size of the study and recruiting more patients if need be altering dose choices during the study and identifying subgroups of patients who respond the best to treatment. Our overall goal is to minimise the number of people needed within a study whilst maintaining the statistical rigour needed to correctly answer the scientific questions.

Given the increase in costs and high risk of developing new interventions smarter clinical designs are needed to make the best use of limited resources. The aim is that the leaner development plans will require fewer patients and shorter timelines. Our programme builds on that of the past where we have contributed to the design of numerous trials and continue to build collaborations in the area of adaptive designs. Most of the Hub Network for Trials Methodology Research stakeholders identified the need for more methodological research in the area of adaptive designs. This program can expand its research, training and software provision to meet this need and enable the translation of theory into future clinical trial practice. We aim to develop new trial designs to handle multiple treatments and multiple interim analyses that minimise the expected number of patients. There is an increasing need to consider diseases as multiple phenotypes and to find treatments that are effective for particular subgroups. Our methods aim to minimise the time of the clinical development and the amount of exposure time for patients whilst obtaining robust evidence for treatment safety and/or efficacy.

Each adaptive design is tailor-made for each clinical situation and the methodological problems are diverse. A clear example of this is the increasing need to design phase I trials for drug combination products. Currently phase I trial designs are very conservative using inefficient rule-based designs. We aim to not only develop better methods but to apply them in real trials, promote them in the clinical community and support clinical trials statisticians. We propose to develop more generalised methodology, such as methods that are robust to model assumptions, that can be applied across a broad range of trials to improve accessibility.

Of particular interest are study designs that can stop early and thus achieving a smaller expected sample size. Sample size re-estimation is another key area, where uncertain sample size parameters can be estimated part-way through a trial; we want to find optimal ways of how to do sample size re-estimation and how to use intermediate outcomes in these calculations.

Future clinical development will involve using more information from intermediate markers of efficacy and/or predictive/prognostic markers to help reduce the time and resources needed to get treatments to bedside. Many of our future proposals are looking to extend current methodology to optimally incorporate these new developments.