Human clinical trials should be based on the best possible evidence. However, changing the design used in early-phase trials has proved challenging, even when newer methods increase the chance that the trial will get the correct answer. The NIHR Statisticians Research Group on Early Phase Clinical Trials, which includes CSM statisticians, is on a mission to help clinical trialists to underpin their phase I trials with best-practice designs. They target clinicians, nurses and statisticians in an open-access motivational awareness-raising paper published in the British Journal of Cancer this week.
The first time a promising new medical treatment leaves the laboratory and is tested in a person is in a phase I dose-finding trial. These trials give a series of human participants different doses of the drug to find out which doses are not too toxic for humans.
Phase I trials typically use rules set before the trial starts to decide what dose each participant gets. However, 20 years of statistics research has overwhelmingly shown that it is better to use mathematical models that decide each participant’s dose using the information gathered in the trial up until that participant. Although pharmaceutical companies have switched to using these models, academic trials still tend to use rules. Only 6.7% of all dose-finding trials use best-practice model-based designs!
CSM’s Sharon Love is a member of the NIHR Statisticians Research Group and a co-author on the motivational paper. She says, “This is the start of a concerted effort by UK statisticians and trialists using these designs to make them accessible to all. This first paper explains the benefits of using model-based designs and how to overcome barriers to using them. In collaboration with the MRC Network of Hubs for Trials Methodology Research's Adaptive Designs Working Group, we have submitted a second paper that will take readers through each step of the design of model-based design using non-statistical language.”
The NIHR Statisticians Research Group plans to offer courses in 2018 to train trialists in using model-based designs for phase I trials. As these best-practice methods are used by the trials research community, the Group plans to conduct methodology research on the challenges that trialists discover when using these methods.