In comparative randomised clinical trials, the purpose of random allocation of patients is to create groups with similar characteristics. The only difference between the groups should be the intervention being tested. As in routine practice, not all patients in a clinical trial receive the treatment exactly as planned. In a trial, this is referred to as deviation from the protocol. In intention to treat analysis, the results from all patients assigned to a group are taken into account, including those who deviated from the protocol. In conjunction with randomisation, it is the best guarantee that the groups of patients being compared have similar characteristics. In addition, it usually best reflects the effects of treatment in everyday practice. Per protocol analysis excludes patients who deviated from the protocol. It can introduce a form of bias called attrition bias, in which the groups of patients being compared no longer have similar characteristics. The results of per protocol analysis usually provide a lower level of evidence but better reflect the effects of treatment when taken in an optimal manner. Per protocol analysis is particularly useful for interpreting non-inferiority trials and, under given conditions, for analysing the adverse effects of treatments. Be wary of "modified" intention to treat analyses that do not include all patients assigned to a group by randomisation: they are actually "extended" per protocol analyses.