| Clinical question How effective is placebo?
Bottom line Although the authors argue against the existence of a placebo effect, this study does not rule out a small placebo effect, particularly for patient-reported continuous outcomes, such as scales for pain or nausea. It may never be possible to separate out completely the contribution of natural history, study bias, treatment effect and placebo effect to the health of our patients.
Synopsis Observed benefit in a treated patient may be caused by the
natural history of disease, the treatment effect, study bias, or the
oft-cited “placebo effect”. The authors previously reported
in a systematic review of 114 randomised controlled trials (RCTs) that
included both placebo and untreated control groups that there was no
such thing as a placebo effect, or at least that it could not be distinguished
from biases inherent in the studies. This report updates that systematic
review with the results of 42 additional such RCTs.
The authors first did a careful search of the literature
for studies that had both a placebo and a no-treatment control group.
They only included
studies with concealed allocation, blinded outcome assessment, and a
dropout rate of less than 50 per cent. Only the primary outcome of each
study was extracted.
The identified RCTs studied a variety of problems, including
pain, nausea, hypertension, jet lag and erectile dysfunction. The results
of the 42
newer trials were similar to those originally reported by the authors,
so the authors combined them. In total there were 156 studies with extractable
data, 38 reporting binary outcomes (ie, treatment success or failure)
and 118 reporting continuous outcomes (ie, a numerical rating of pain).
The authors began by drawing a funnel plot, which is a way
to look for publication bias. Although the funnel plot is difficult to
interpret,
it is possible that smaller studies with negative outcomes for placebo
were under-reported.
However, since looking for a difference between placebo and untreated
controls was unlikely to be the intention of any of these studies, it
is not clear why authors would choose to not submit smaller “negative” trials
for publication. With this caveat in mind, the authors pooled the results.
There was no benefit to placebo over no treatment for success/failure
outcomes, although examination of individual topics and the confidence
interval suggests that there is a possible trend in favour of placebo.
Regarding continuous outcomes, there was a small but significant
benefit to placebo among all studies (standardised mean difference [SMD]
= –0.24,
95 per cent confidence interval [CI] –0.31 to –0.17). The
benefit was especially large among patient-reported continuous outcomes
(SMD = –0.30, CI –0.38 to –0.21) and was of borderline
significance for observer-reported outcomes (SMD = –0.10, CI –0.20
to 0.01). This benefit of placebo for trials with continuous outcomes
persisted even when only the three highest quality studies were included.
It was also consistent for all 10 topics studied with continuous outcomes
(eg, pain, smoking, nausea): each had a point estimate of effect less
than zero, consistent with benefit of placebo over no treatment. The
authors argue that there may be a small placebo effect, but it is probably
clinically insignificant and cannot be distinguished from the bias inherent
in any study.
Level of evidence = 1a (systematic review, with homogeneity, of randomised
controlled trials)
Reference Hrobjartsson A, Gotzsche PC. Is the placebo powerless? Update
of a systematic review with 52 new randomised trials comparing placebo
with no treatment. Journal of Internal Medicine 2004;
256:91–100
POEM (Patient Oriented Evidence that Matters) is a registered trademark
of InfoPOEMs
© infoPOEMs 1995–2004
|
The
Pharmaceutical Journal