Selection bias in randomised trials

Selection bias can occur in a randomised trial if recruiters selectively enrol patients into a trial based on expectations of what the next treatment allocation will be. It is therefore important to have a concealed treatment allocation. However, in certain circumstances, the patient recruiter may be able to guess the next patient’s assignment. This can, for example, happen in an unblinded trial with a randomisation procedure forcing the number of patients in each arm or within each centre to be the same.

Kahan et al. (1) reviewed 152 trials. Most of them (98 %) did not provide any information on whether recruiters were blind to previous treatment allocations. While only 3 % of the trials used simple randomisation, 63 % used restricted randomisation and 35 % did not describe the method of randomisation. The risk of selection bias could for many trials not be ascertained because of poor reporting, but many trials which did provide sufficient details on the randomisation procedure were at risk of selection bias because of poorly chosen randomisation methods.

The authors conclude that techniques to reduce the risk of selection bias should be more widely implemented.


1. Kahan BC, Rehal S, Cro S. Risk of selection bias in randomised trials. Trials 2015;16:405