Education section – Designing a “placebo” to help define the intervention
Article first published online: 29 NOV 2011
© 2011 Blackwell Publishing Asia Pty Ltd and Chinese Cochrane Center, West China Hospital of Sichuan University
Journal of Evidence-Based Medicine
Volume 4, Issue 4, pages 261–262, November 2011
How to Cite
(2011), Education section – Designing a “placebo” to help define the intervention. Journal of Evidence-Based Medicine, 4: 261–262. doi: 10.1111/j.1756-5391.2011.01153.x
- Issue published online: 29 NOV 2011
- Article first published online: 29 NOV 2011
In an earlier issue of the Journal of Evidence Based Medicine, we discussed the role of binding (or masking) in clinical trials, so that some of the people involved in the study, such as the patients, healthcare providers, outcome assessors or statisticians, would not know which group a patient had been allocated to (1). This might be achieved through the use of a placebo and we return to this topic this month, to see how the designing of a placebo might also reveal things about the actual question being tackled in a randomised trial (2).
There are a variety of reasons for using a placebo in a randomised trial (3). The aim is to reduce bias and to get closer to detecting the true effect of the intervention being tested. For example, a placebo might help to ensure that any benefit experienced by the patient is not simply because of the effect of receiving an intervention, it might reduce the possibility that changes in behaviour for the intervention or the control group will be because of knowledge of the allocated intervention, it might prevent healthcare practitioners from treating or caring for the patients in different randomised groups in different ways, or stop those who are measuring outcomes or analysing the data from being influenced by knowledge of the assigned group for a patient.
One of the commonest ways to create a placebo for comparison in a randomised trial is to remove whatever is believed to be the active ingredient from the intervention group, giving the control group what remains. For example, in a placebo controlled trial of an anti-depressant tablet, the control group would be given an identical tablet which simply lacked the drug being tested. However, difficulties arise when the “active ingredient” is not so clear and a few examples are considered here. In these examples, the “placebo” is not necessarily an identical match for the intervention, but it would be designed to try to blind some or all of people involved with the trial, so that they do not know the allocation for any particular patient.
The first example considers the testing of a new oil based product for treating dry skin. A placebo might be important if there was a concern that it is not anything about the new oil itself which will lead to benefit but simply the use of an oil or the massaging technique which helps the skin. If the new oil was a standard oil into which a small amount of an active ingredient had been added, then it would be like the anti-depressant trial mentioned above. You would simply use the standard oil as the placebo. But, perhaps the ingredients are a more complex mixture or perhaps there are thought to be properties of the oil itself that will help with the dry skin. The control group might then need to use a different oil entirely, or might need to massage the skin without the oil. The fundamental issue becomes is it the massage, the oil or the special ingredients in the oil that are being tested?
Another circumstance could be the use of acupuncture by a physiotherapist to treat neck pain. In one randomised trial, each patient might be randomised to receive acupuncture or sham acupuncture (including the use of needles) from the same physiotherapist. This would test whether the positioning and type of needles was responsible for any effect. However, if the acupuncture was regarded as a more complex intervention than just the use of the needles, perhaps including the approach and time taken by the physiotherapist, the randomisation might need to allocate patients to receive acupuncture from the physiotherapist or to have a consultation with the same physiotherapist but without any use of needles. On the other hand, if the acupuncture was regarded as even more complex, with benefits thought to arise from the physiotherapist herself, the randomisation would need to divide patients into those who would be treated by physiotherapists who practice acupuncture versus those who would be treated by physiotherapists who do not use acupuncture.
Finally, consider an evaluation of whether asking patients to complete a symptoms questionnaire before they meet with their general practitioner improves their experience and their health outcomes if the questionnaire is shared with the general practitioner during the consultation. The hope might be that it would lead to better advice or a more targeted prescription from the general practitioner. In one type of trial, the randomisation would ensure that the answers would be shared with the general practitioner for only half of the patients, even though both groups would have completed the questionnaire. This would test whether the provision of the answers has an effect. But what if completing the questionnaire changed the behaviour of the patient during the consultation? Then, the randomisation might need to compare patients completing the questionnaire with those who did not complete it. And, in such a trial, the patients in the control group might need to complete a different questionnaire entirely, to try to keep them blinded to their allocation.
These three examples highlight the need for careful consideration of what would be an appropriate placebo in a randomised trial. They reveal that a key step at the start of any trial should be clarity about what is actually being tested (2), so that a “placebo” can be designed which provides everything else except that “active ingredient”.
This section was prepared as part of the TENALEA project, which provides an online system for the registration and randomisation of participants into research studies in health care. It has been developed by the National Cancer Institute in Amsterdam in The Netherlands, working with partners in some of the largest cancer research groups in Europe and is now being deployed, with funding from the European Union (contract C029334). More information, including details of how to use the system, is available at http://www.tenalea.com.
- 1Anon. Education section – blinding. Journal of Evidence Based Medicine 2009; 2: 65.
- 3Placebo interventions for all clinical conditions. Cochrane Database of Systematic Reviews 2010, Issue 1. Art. No.: CD003974., .