Skip navigation
Link to Back issues listing | Back Issue Listing with content Index | Subject Index

That old placebo feeling

Bandolier continuously gets into trouble over placebo because people have their own very fixed views about what it is. The extremes of definition are shown in Figure 1.

Figure 1

One definition is that use of a placebo describes what happens when you do nothing, so that in the context of a clinical trial, for instance, a placebo group could describe the natural history of a disorder without the intervention under test. It is the effect we see when we do nothing.

This rather clumsy phrase is abbreviated to the placebo effect . The trouble is that placebo effect can be taken to mean that any effect seen is caused by placebo. The definition has now changed to mean that doing nothing causes things to happen. The idea that placebo does make things happen is reinforced by much interesting academic work, almost always in contrived situations and with small numbers of people. Extrapolating that to clinical practice is a big step, but one often assumed.

Is it possible to say that one definition is right and the other wrong? A systematic review and analysis comparing clinical trials in which patients were randomly assigned to placebo or no treatment suggests that the evidence for placebo having an effect is at best weak [1].


The review sought studies comparing placebo with no treatment that were properly randomised and blinded. Exclusions were healthy volunteers, studies with high dropout rates, or if the alleged placebo had some clinical effect.


The analysis included 114 trials in 40 clinical conditions.

For dichotomous outcomes in 32 trials with 3,800 patients (average 119 per comparison) there was no difference between placebo and no treatment overall, or with subjective or objective outcomes.

For continuous outcomes in 82 trials with 4,800 patients (average 59 patients per comparison) there was a small overall effect, all for subjective outcomes.

The only condition in which there was an effect was for pain for continuous outcomes in 27 trials with 1,600 patients (average 59 trials per comparison).


The interesting thing is that any possible effect of placebo was found with continuous outcomes for pain in small trials. This is likely to be an artefact. Firstly we know [2] that continuous pain outcomes are often reported as a mean of data with highly skewed distribution. Analysis of means is therefore meaningless [2]. Smaller trials in pain have also been shown to have more effect than larger ones, probably because they are of poorer quality [3].

Not much evidence there, then, that doing nothing makes things happen.


  1. A Hröbjartsson & PC Gøtzsche. Is the placebo powerless? An analysis of clinical trials comparing placebo with no treatment. New England Journal of Medicine 2001 344: 1594-1602.
  2. HJ McQuay et al. Variation in the placebo effect in randomised controlled trials of analgesics: all is as blind as it seems. Pain 1996 64:331-335.
  3. RA Moore et al. Quantitative systematic review of topically-applied non-steroidal anti-inflammatory drugs. British Medical Journal 1998 316: 333-8.
previous or next story in this issue