Patient preference trialsBMJ 2011; 342 doi: https://doi.org/10.1136/bmj.d3602 (Published 15 June 2011) Cite this as: BMJ 2011;342:d3602
- Philip Sedgwick, senior lecturer in medical statistics
- 1Section of Medical and Healthcare Education, St George’s, University of London, Tooting, London, UK
Researchers compared the efficacy of antidepressant drugs with generic counselling for the treatment of mild to moderate depression in general practice. A randomised controlled trial, with patient preference arms, was conducted. In total, 52 patients were randomised to counselling and 51 to antidepressant drugs, with 140 recruited to the counselling and 80 to the antidepressant preference arms.1
The outcome measures included the Beck depression inventory score at one year follow-up. No differences were reported in Beck scores between the randomised arms or patient preference groups. Patients who chose counselling did better than those who were randomised to it.
Which of the following statements, if any, are true?
a) All patients were randomised to a treatment
b) Patients could choose which treatment they received
c) Systematic differences in baseline characteristics between the randomised groups were minimised
d) The potential for performance bias was reduced in the randomised treatment groups
Answers b, c, and d are true, whereas a is false.
The randomised trial with patient preference arms, also known as a partially randomised patient preference trial, allows participants to choose their preferred treatment if they wish (b is true). Only those people without a preference are randomised (a is false). The trial had four groups, with a randomised and patient preference group for each treatment. After informed consent was obtained, 103 people did not express a preference for either treatment and consented to randomisation, with 52 patients being allocated to the counselling group and 51 to the antidepressant drugs group. A total of 220 expressed a preference for treatment and did not consent to randomisation—140 were recruited to the counselling group and 80 to the antidepressant drugs group.
Those people without a preference for counselling or antidepressant drugs entered a traditional randomised controlled trial and were randomised to an intervention. Any systematic differences between these two treatment groups at baseline were therefore minimised (c is true). Hence, any difference between the randomised treatment groups in outcomes would be the result of differences in treatment received and not differences between groups in baseline characteristics. The patient preference groups were not similar in baseline characteristics, not least because of treatment preference. Therefore, differences between the patient preference groups in outcome may not have resulted from differences in treatment received but differences in baseline characteristics. Those people who expressed a preference for treatment entered a non-randomised trial.
In a traditional randomised controlled trial, people may express a preference for a treatment and refuse to be randomised. Depending on the numbers of these people and their characteristics, their refusal to enter the trial may restrict the generalisation of the results, because those participants who agree to randomisation may not be representative of the wider population. If a treatment is not available outside the trial, people with a treatment preference may still participate, in the hope that they will receive that treatment. If participants cannot be blinded to treatment allocation, those patients that receive their preferred treatment may be better motivated and show greater compliance with the treatment regimen. On the other hand, participants not receiving their preferred treatment may exhibit resentful demoralisation, whereby they comply poorly and possibly withdraw from the trial. Such patient preference for treatment contributes to performance bias, where factors other than the treatments under investigation influence the outcome measures.
The randomised trial with patient preference arms allowed the study of those people with a strong preference for antidepressant drugs or counselling, who otherwise would have refused randomisation and not entered the trial. Comparison of the patient preference groups provided a pragmatic evaluation of response when patients are able to choose their treatment. The patient preference arms may have minimised performance bias caused by patient preference in the randomised arms, thereby enhancing the rigours of the randomised controlled trial and allowing a more accurate comparison of the therapeutic benefits of the treatments (d is true).
Cite this as: BMJ 2011;342:d3602
Competing interests: None declared.