Intention-to-treat analysis

摘要

Trials that randomly allocate participants to groups tend toproduce groups with similar characteristics, especially when manyparticipants are randomised.1 Because the groups have similarcharacteristics, their outcomes are also expected to be similar. If onlyone of the groups then receives an intervention, the difference inoutcomes between the groups can therefore be attributed to thatintervention. If each group receives a unique intervention, thedifference in outcomes between the groups instead reflects thedifference in the effects of the two interventions.2 Before recruitment of participants into a clinical trial begins, the final protocol forthe trial should be agreed upon by all the investigators, approved byan ethics committee, and registered on a publicly available register.3Ideally, the trial is then conducted exactly according to that protocol,with all participants who were randomised receiving all doses oftheir allocated intervention and undergoing all scheduled measurement procedures.4 Unfortunately, the conduct of many clinical trialsdeviates from the protocol, to some degree, which can introduceunwanted systematic differences (ie, bias) between the groups afterrandomisation.