Regular readers of the Mental Elf may remember my previous blog which looked at a series of papers examining the placebo effect in relation to depression and the use of antidepressant medication. You can find it here along with a placebo primer by Ben Goldacre here.

To recap, the placebo effect refers to the effect of a simulated or otherwise ineffectual treatment. It has been known about for more than 200 years. Placebos are often used as control treatments in trials of medicine, the rationale being that the medicine should outperform the placebo (commonly a sugar pill) in order for it to be regarded as efficacious.

In the paper I’ve been asked to review for the Mental Elf (Rutherford et al, 2014) the authors chose to examine the placebo response in antipsychotic clinical trials. Rutherford and colleagues note at the beginning of the paper that placebo response rates in trials of antipsychotic medications are increasing. The authors sought to determine the causes of this increase in placebo response rate. They argued that this phenomenon increases the cost of drug development, delays the availability of new antipsychotic medications and means that pharmaceutical companies spend less money on psychiatric research.

Rutherford et al. sought to investigate the causes of placebo response in a selection of randomised controlled trials (RCTs) of antipsychotic medication.

Methods

The authors used a complicated statistical technique called hierarchical linear modeling. This formed the basis of a meta-analysis.

They did not examine all possible predictors of treatment change and instead selected possible variables based on a pre-existing model of the causes of placebo response. This seems like an appropriate strategy.

The model that Rutherford et al used divides causes for placebo response into three categories:

1. Placebo effects based on patient expectancy of improvement and therapeutic contact
2. Measurement error due to rater bias and sampling error
3. Spontaneous improvement

Databases were searched using what seem to be appropriate inclusion and exclusion criteria although these appear to be limited to published studies written in English. Trials were required to last between 4 and 24 weeks but no rationale for these thresholds was given.

A standardized mean change score was calculated because different studies used different measures of pre-post change.

Results

Following application of their inclusion and exclusion criteria, the authors were left with 105 studies analyzing 257 different treatment groups.

The mean change in patients receiving placebo increased significantly over time (Spearman r = 0.52, n = 39, p = 0.01). There was a greater placebo effect in more recent studies.

The mean change in patients receiving “effective dose medication” decreased significantly over the same time period (Spearman r = -0.26, n = 208, p <0.001).

Discussion

• The placebo response has been significantly increasing from 1960 to the present.
• In 1960 the average placebo treated patient worsened by 3 Brief Psychiatric Rating Scale (BPRS) points. In the 2000s the average placebo treated patient improved by 3.2 BPRS points.
• The treatment change associated with “effective dose” medication decreased significantly over the same time period.
• The benefit of effective dose medication over placebo was greater for more severely ill patients.
• The mean change for placebo treated patients was associated with sample size.
• The authors suggest that the placebo response for antipsychotic medication shares similarities with placebo responses for other mental health difficulties, notably depression. They explain that trials have generally evolved over time. Older trials tend to be smaller, academic, single-site trials while newer ones are large, commercial and multi-centre. The authors suggest that there are pros and cons to these developments.

Limitations

The authors mention that one limitation concerns publication bias. The discussion of this limitation in the paper lasts one sentence but, in my view, this is a very significant limitation. Studies funded by pharmaceutical companies that fail to demonstrate a difference between placebo and antipsychotic medication are unlikely to be published. There is a risk that this meta-analysis does not then tell the whole story because it only includes a proportion of all the trials conducted in this field.

I’m reminded of my previous Mental Elf blog on the placebo response in antidepressant medication. Here, the reviewers asked for data on 33 unpublished studies from pharmaceutical companies and were denied access to data on 30 occasions (Naudet et al, 2013). Naudet al. suggested that no firm conclusions could be drawn from their results as “antidepressant research is completely controlled by the pharmaceutical industry”.

It is difficult to have much confidence in the results of the Rutherford et al study in the absence of such data for antipsychotic medication.

Links

Rutherford, B.R., Pott, E., Tandler, J.M., Wall, M.M., Roose, S.P. & Lieberman, J.A. (2014). Placebo response in antipsychotic clinical trials: A meta-analysis. JAMA Psychiatry, 71(12), 1409-21

Judge J. Placebo for depression: are some sugar pills better than others? The Mental Elf, 4 Sep 2014.

Naudet F, Millet B, Charlier P, Reymann JM, Maria AS, Falissard B. (2013). Which placebo to cure depression? A thought-provoking network meta-analysis. BMC Medicine, doi: 10.1186/1741-7015-11-230.