Is predictability of randomisation a historic concern: A survey of practice of recruiters to randomised controlled trials

<ns3:p>Background Randomisation in randomised controlled trials (RCTs) is intended to guard against selection bias by creating unpredictable sequences. Some randomisation methods, however, aim to achieve better balance between groups being compared at the expense of predictability. This study aims to investigate the extent to which recruiters into RCTs attempt to predict upcoming allocations, hence how concerned trialists should be about using more predictable randomisation methods. Methods We designed a short survey asking recruiters questions about the practice of predicting allocations before randomisation. The survey was circulated to R&I departments of hospitals in England involved in recruiting to randomised controlled trials. Two versions of the survey were employed, one explicitly referencing “subversion” of the randomisation, and one that did not use this word to investigate whether the framing of the research changed participants responses. A summary of counts and percentages of responses is presented. Results 122 participants responded to the survey. 30% of responders felt that they may be able to predict the next allocation in a sequence dependent on the information available, with 20% responding that they currently work on a trial where themselves or a colleague has attempted to guess the next allocation. 65 received the version that included reference to “subversion”, whilst 57 received the alternative version. There was not a notable difference in responses for the two versions. Conclusion This survey provides evidence that some recruiters do attempt to predict what the upcoming allocation will be. Whilst we do not know the motivations behind these predictions, they highlight the need for researchers to produce unpredictable allocation sequences.</ns3:p>