Allocation bias occurs when there is a systematic difference in how participants are assigned to groups in a trial. The trial is unfair because the groups are not fairly balanced from the outset.
Appropriate randomization is the key to preventing allocation bias. Allocation bias can result if the investigators know, or are able to predict, what intervention the next participant is supposed to receive (insufficient allocation concealment or a lack of blinding). Knowledge of the next intervention can influence whether the physician wants to enroll the patient in the trial, and therefore can lead to a situation in which healthier patients are selected towards one group in the trial.
Key point: Always check the methods section for a description of the allocation concealment protocol, which is the primary defense against allocation bias.
Sedgwick P. Selection bias versus allocation bias BMJ. 2013; 346:f3345-f3345.