Are all placebos equal?

I've written twice this year (January and September) about psychological effects that can influence a person's perception of pain or other stimuli. This month I'm looking at some research into the difference between various placebos used in trials of methods to treat the same complaint.

Watching television I see two advertisements which seem to appear on multiple occasions each night. Both are for over-the-counter treatments for osteoarthritis pain in the knees. In one of them a grey-haired gentleman with a short grey beard takes pills containing paracetamol to relieve the pain, enabling him to play with his (grand)children by swimming or running with a kite. The other advertisement shows a woman, apparently much younger than the man in the pill ads, who eases the pain by rubbing her knee with ointment out of a tube. (Another difference is that she doesn't seem to do this so she can play with children but instead it helps her do yoga and also to skip nimbly across a bedroom and jump on the bed, causing the man who is already there to put down the book he is reading. I will leave discussion of the representation of gender roles in advertisements for another time.)

Research published in the current edition of The Annals of Internal Medicine looks at different placebos used in clinical trials of treatments for knee pain caused by osteoarthritis. There are four basic forms of treatment – topical gels and creams, intra-articular injection, oral medication and oral plus topical. Rather obviously any placebo treatment must match the real treatment being evaluated, but what researchers were interested in was whether the different forms of placebo had effects which were different enough to possibly distort the findings of trials.

The paper, "Effectiveness and Implications of Alternative Placebo Treatments: A Systematic Review and Network Meta-Analysis of Osteoarthritis Trials" looked at "149 randomized trials of adults with knee osteoarthritis that reported pain outcomes and compared widely used pharmaceuticals against the following 4 placebos: oral, intra-articular, topical, and oral plus topical".

One non-surprising finding was that treatments with higher levels of intervention showed a stronger placebo response. I had an intra-articular steroid injection when I broke my ankle and this procedure takes quite some time and is surrounded by a certain amount of drama with local anaesthetics and imaging, whereas taking pills is an everyday occurrence for a large number of people. It was no surprise then to find that the placebo effect was strongest for this form of treatment. The authors note that a previous study comparing sham acupuncture with oral placebo found a significantly greater a placebo response to sham acupuncture, reinforcing the idea that the level of intervention itself makes a difference. (That study was not included in this meta-analysis because it was for shoulder pain, not knee osteoarthritis.)

There is also a possibility that placebos may themselves have a clinical effect. The most obvious instance of this is the injection of saline into a joint which can by itself have a lubricating and pain easing effect. It is not therefore a true placebo so the measured difference in outcomes between placebo and control groups might be less than if a true placebo was used, leading to rejection of a treatment which is actually effective.

Even without the possibility of placebos having a clinical effect, the varying amounts of response have to be considered when interpreting trials. Finding significant difference might be more unlikely for treatments with a high level of placebo response, leading to some effective treatments not being applied. This could have ramifications for the convenience and cost of patient treatments. For example, intra-articular injections given once a year could possibly be a far better solution than the daily administration of non-steroidal anti-inflammatory drugs but the high level of placebo response could lead to a rejection of this in favour of something which is actually less effective and could even be dangerous in the long run. (No responsible doctor is going to allow any patient to constantly take large daily doses of paracetamol or ibuprofen, for example.)

There is a well-known flaw of meta-analysis itself of which the authors were obviously quite aware. When the different levels of placebo response to the different types of treatment were ignored the studies in aggregate indicated that NSAIDs are superior to any other form of treatment. This may or may not be the case, but it is a conclusion that could be drawn by someone not understand the limitations of meta-analysis.

I'll finish by quoting the conclusion of the study: "All placebos are not equal, and some can trigger clinically relevant responses. Differential placebo effects can substantially alter estimates of the relative efficacies of active treatments, an important consideration for the design of clinical trials and interpretation of the results".

This article was published as the Naked Skeptic column in the October 2015 edition of Australasian Science