What is the study about?

Within randomised controlled trials, there are a variety of allocation methods available to decide participant treatments, however there is limited guidance on when these methods perform best, and if the performance of methods is affected by the design of the study.

Aim?

To produce evidence-based guidance to aid researchers in choosing the best randomisation method for their specific trial design.

Summary of work

A literature review of randomisation methodology  

A literature review of trials published in 2019 in the Lancet, the BMJ, NEJM and JAMA investigating 1) What randomisation methods are being used 2) if there is an association between the choice of randomisation method and study characteristics 

Publication: https://bmcmedresmethodol.biomedcentral.com/articles/10.1186/s12874-022-01786-4 

A talk on this work: https://www.youtube.com/watch?v=1nyeLt2z4zM 

Focus groups with trialists 

Focus groups with researchers involved in randomised controlled trials including Statisticians, IT/programmers and other trialists involved in the randomisation process. The research investigates 1) The motivations behind why researchers select specific randomisation methods 2) Features of randomisation methods that researchers consider when selection a randomisation method. 

Publication: https://trialsjournal.biomedcentral.com/articles/10.1186/s13063-024-08005-z 

A talk on this work:  https://www.youtube.com/watch?v=vwu5tw6MmrE 

A survey investigating the prevalence of predictability 

During focus groups, it was clear that for some, the definition of predictability is more closely linked with how often subversion is taking place rather than the potential for prediction. A survey was conducted with people involved in the recruitment and randomisation process (Participant identification, eligibility assessment, taking of consent or accessing the randomisation system) to investigate 1) Whether recruiters are attempting to predict upcoming allocations 2) If so, what methods are they using to make this prediction.  

Development of metrics to assess the predictability and balance of randomisation methods 

From previous research and the conducted focus groups, it was clear that balance and predictability are the main drivers for researchers when selecting a randomisation method. This research aimed to develop easy to implement metrics to assess the balance and predictability of randomisation methods in a variety of ways. 

A simulation study assessing the performance of commonly used randomisation methods 

Using all the previous research, this project involves a simulation study where simple randomisation, block randomisation, stratified block randomisation and minimisation were assessed using the developed metrics, across different sample sizes, numbers of centres and numbers of randomisation variables included to compare the performance of these methods under different study designs 

A taxonomy of randomisation methods 

Alongside the evaluation of commonly used methods, a taxonomy will be developed to describe existing randomisation methodology, classify these methods and improve assess and information around existing randomisation methodology. 

Guidance on randomisation method selection 

Bringing together all this previous work, we will develop a guidance document for researchers to use during study design to help them to select the most appropriate method for their study design.