Commentary

In our trial we asked patients to consent to follow up but not to consent to randomisation itself. There were several reasons for adopting this approach, which was approved by our local ethics committee. Firstly, we did not expect our intervention to be harmful, though whether this expectation was fulfilled must be judged from our results. Secondly, patients and their carers could refuse to see our stroke family care worker or follow up psychologist whenever they wished. Thus half the patients and their carers were asked to consent to the intervention and all were asked to consent to follow up after randomisation. Thirdly, we were concerned that if we tried to obtain informed consent this might bias our results. For instance, if we made patients and their families aware of the help they might receive from the stroke family care worker and then randomised them to the control group this might have had a detrimental effect on their morale. This could have led to a false positive result simply by having an adverse effect on the controls.

In addition, as our patients and their carers were not aware that they were in a randomised trial to assess our stroke family care worker and that our follow up was attempting to assess her effectiveness, they were in effect partially blinded. We might imagine that loyalty to the care worker might have biased their responses had they known the precise purpose of our follow up. Fourthly, our approach allowed patients or carers to decide to see the stroke family care worker when it was relevant to them. Some patients might not consent to randomisation shortly after their stroke, when they are unlikely to foresee the possible psychosocial impact of the stroke on them and their families. They might then regret the decision not to be randomised when the potential benefits of the intervention become more evident. Lastly, as our intervention was applied to patients and their families it was unclear who might most appropriately give consent.

Increasingly, purchasers and providers of health care are looking for evidence from methodologically sound randomised controlled trials and systematic reviews to guide their practice. In a trial whose outcome measures reflect the feelings or opinions of the subjects a detailed knowledge of the trial and its exact purpose are likely to influence or bias responses. Thus responses may reflect either a control subject's disappointment or dissatisfaction with not receiving a potentially beneficial treatment or a treated patient's appreciation or loyalty to those providing the treatment. Those who review such studies will be unable to judge whether this source of bias might account for any difference in outcomes between treated and control groups. Thus no studies would be regarded as methodologically watertight. Is it ethical to randomise patients into trials which because of an inherent methodological weakness cannot provide a definite answer to the main question?