JF Boylan, BP Kavanagh, J Armitage



Practising physicians individualise treatments, hoping to achieve optimal outcomes by tackling relevant patient variables. The randomised controlled trial (RCT) is universally accepted as the best means of comparison. Yet doctors sometimes wonder if particular patients might benefit more from treatments that fared worse in the RCT comparisons. Such clinicians may even feel ostracised by their peers for stepping outside treatments based on RCTs and guidelines. Are RCTs the only acceptable evaluations of how patient care can be assessed and delivered? In this controversy we explore the interpretation of RCT data for practising clinicians facing individualised patient choices. First, critical care anaesthetists John Boylan and Brian Kavanagh emphasise the dangers of bias and show how Bayesian approaches utilise prior probabilities to improve posterior (combined) probability estimates. Secondly, Jane Armitage, of the Clinical Trial Service Unit in Oxford, argues why RCTs remain essential and explores how the quality of randomisation can be improved through systematic reviews and by avoiding selective reporting.

Keywords Randomised controlled trials, Bayesian analysis, subgroup analysis, clinical trials, bias, confounders

Declaration of Interests No conflict of interests declared.