|
Back
Placebo Response Rates in Randomized Controlled Trials (RCTs). A meta-analysis
Principal Investigators: C. Sadaghiani, PhD
Affiliations: University Hospital Freiburg, Department of Environmental Medicine
and Hospital Epidemiology, Germany
Background: Since Beecher (1955) pointed to the importance of the placebo as a
potentially powerful therapeutic intervention, the debate has not ceased. The
scientific community is divided into two groups concerning the therapeutic
relevance of placebo or unspecific effects: those who believe that placebos can
influence physical health and those who dont. Part of the discussion is
possibly based on confusion in defining and measuring placebo effects. However,
the question still remains whether the success in placebo groups is solely due
to artifacts (e.g., regression to the mean, time course of the illness,
measuring artifacts), or if there really do occur clinically relevant changes?
Method: We searched the medical literature from 1992 to 1997 (Medline, Cochrane
Library, 1997) for long-term RCTs (double-blind, placebo-controlled, duration >
12 weeks). Search terms were 'placebo', 'double-blind', 'RCT', 'meta-analysis',
'long', 'studies', and 'treatment outcome'. These search terms were combined
with different strategies. 141 studies fulfilled our inclusion criteria. Formal
characteristics were coded: disease treated (ICD-10), duration of study,
improvement with placebo and with medical treatment (% patients improved),
attrition rate, and evaluation according to intent-to-treat. Two independent
raters coded study quality.
Hypotheses: We assumed if a large part of the placebo response rates in RCTs are
due to artifacts (e.g., study quality), we should find some indication for this
in correlations between study characteristics representing these variables and
the amount of placebo response rates.
Results: We found a highly significant correlation between placebo and treatment
response rate across diseases (r=.78; p<.001). Also, there were significant
relationships between placebo response rates and duration of study (r=.41),
prevention trials (r=.59) and trials of ischemic heart disease (r=.45). Placebo
response rates were lower in trials of psychotropic substances, dementia and
anti-epileptic trials. Surprisingly, study quality and other characteristics of
the studies were unrelated. In a regression analysis, the treatment response
rate was the only substantial predictor (ß=.75) and explained approx. 60% of the
variance. These results are a strong argument against the artifact hypothesis of
placebo effects. Natural history also is an unlikely explanation for this
phenomenon, which warrants further scrutiny.
|