Patients who do not receive any of their allocated treatment

In most randomized trials a (usually small) proportion of patients will never receive the allocated treatment. The reasons for this may include: the patient changes their mind, or the patient's condition deteriorates (or improves) rapidly, or an administrative error means the patient receives the wrong treatment. In this circumstance they may receive the alternative treatment in the trial, another treatment altogether or no treatment at all. It may be tempting to exclude such patients from the analysis. However, in general patients who do not receive any of their allocated treatment are likely to be those with a poorer prognosis. This is perhaps most obvious in trials comparing an active treatment against a 'no-treatment' control group - in this situation the active treatment group may have patients receiving none of their allocated treatment. It may be tempting to exclude these patients from the analysis. However, it is almost impossible to identify and exclude a similar group of patients from the control arm, because we cannot predict who would have failed to receive the treatment if they had been allocated it. Thus exclusion of such patients on the experimental arm from the analysis means that we are introducing systematic differences between the arms and are likely to introduce bias in favour of the more aggressive, usually new, treatment. It should be noted that including these patients in the analysis according to their allocated treatment group, does itself introduce a form of bias, because the difference between the experimental and control arms will be underestimated (as patients who receive none of the allocated treatment cannot benefit from it). Nevertheless, such an approach is inherently conservative and is to be preferred to overestimating the difference (by an unknown amount) in favour of the experimental therapy.

0 0

Post a comment