[e-med] Le placebo aurait un effet zéro (2)

["Trouiller, Patrice" <PTrouiller@chu-grenoble.fr>]

E-MED: Le placebo aurait un effet zéro (2)
-------------------------------------------------------------------------


Quel que soit l'intérêt des commentaires du Pr J.P. Bader dans "Le Figaro",
je pense qu'il est intéressant d'avoir l'article original (en anglais) paru
la semaine dernère dans le "New England Journal of Medicine" (cf. référence
3), publication qui a été précédée par un éditorial. Le tout est consultable
sur le site de la revue:  http://www.nejm.org .

Ceci permettra à chacun de lire la publication princeps et de se faire sa
propre opinion - vraisemblablement plus nuancée - en la matère.

Patrice Trouiller
Praticien hospitalier
CHU de Grenoble
ptrouiller@chu-grenoble.fr

The New England Journal of Medicine -- May 24, 2001 -- Vol. 344, No. 21
The Powerful Placebo and the Wizard of OzSome myths really ought to be true.

We react with surprise and pleasure when we encounter them and then believe
them when they neatly and comfortably help to explain some confusing aspect
of our world. Thereafter, evidence against them is unwelcome and not to be
trusted. But some such myths are flawed and misleading.  John Snow has been
widely credited with stopping a cholera epidemic in 1854.

He noticed that the disease was prevalent among Londoners who drank from a
well supplied by one water company but not among those who drank from the
well of another company, and he removed the handle of the offending pump.
Alas, the record shows that the number of new cases of cholera had already
decreased sharply, and Snow's insight (though important for establishing the
cause of cholera) had little effect as the epidemic completed its course.

Studies of five workers at Western Electric's Hawthorne plant in Illinois
were cited for decades as having shown that productivity increased each time
the investigators changed the working conditions, including the last
change -- back to the original conditions. This "Hawthorne effect" was said
to show how a research setting itself could change behavior. Again, though,
the records belie this simple and appealing story; the last working
conditions differed from the original conditions in several important ways,
including the addition of a break for rest and tea as well as the
progressive transformation of the workers who were being studied from
passive subjects into active study participants.

The potential benefits of placebos as treatments for diseases have rarely
been questioned since Beecher reported in 1955 that they could relieve
symptoms and otherwise contribute to the well-being of patients. Many
readers will remember the Wizard of Oz, who was powerful because others
thought he was powerful -- until they found that the curtain hid a very
ordinary man.  Is the placebo powerful because we have not looked behind the
curtain? The question is involuted; layers of meaning surround the words
"powerful" and "because." In this issue of the Journal, Hrobjartsson and
Gotzsche  help to remove at least one layer of the mystery. They conducted a
systematic review of clinical trials in which patients were randomly
assigned to either placebo or no treatment. They found little evidence in
general that placebos had powerful clinical effects.

The basic problem is that the patient who has had a bad day with cancer or
emphysema or a headache does not need a placebo to feel better the next day.
Is the improvement in patients given a placebo a result of the placebo
itself, of natural fluctuations in the progression of the disease, or of how
the patient responds to the symptoms? The primary comparison in a
placebo-controlled trial is usually of the placebo with a possibly active
therapy, not with no treatment. This design cannot distinguish an effect of
placebo from the natural course of the disease, regression to the mean (the
tendency for random increases or decreases to be followed by observations
closer to the average), or the effects of other factors.

Clinical trials generally include blinding to improve objectivity in the
assessment of outcomes if the patient or the observer might otherwise be
able to tell which treatment was being given. Effective blinding may require
the use of a placebo. Some trials include a placebo for other reasons, such
as to strengthen the bond between the patient and the study. Some studies do
not include a placebo at all. Few, however, include both a placebo and a
nonplacebo (untreated) group for which outcomes can be compared directly. It
is remarkable that Hrobjartsson and Gotzsche found 114 randomized clinical
trials that included both a placebo group and an untreated group and in
which other treatment, if any, was held constant.

Placebo is surprisingly hard to define precisely. These authors defined
placebo in operational terms as an intervention labeled as such in the
report of a clinical trial. They report only the main outcome of each trial,
so no patients were counted more than once. When the original report did not
specify the main outcome, the reviewers selected the outcome they considered
most relevant to patients. The placebos they reviewed were pharmacologic
(e.g., tablet), physical (e.g., manipulation), or psychological (e.g.,
conversation).

In all, data from about 7500 patients with 40 different clinical conditions
were included in the comparison of placebo with no treatment. Binary
outcomes and continuous outcomes, both subjective and objective, were
examined separately. Statistical tests of the pooled data showed no
significant effect on subjective or objective binary outcomes or continuous
objective outcomes. There was a significant effect of placebo with respect
to pain, but the difference as compared with no treatment appeared to
diminish with increasing sample size. This finding suggests that the
observed effect may be a product of publication bias or other bias in
reporting. Even this possible effect of placebo was small, an average of 6.5
mm on a 100-mm visual-analogue scale (an effect the authors state is
approximately one third that of nonsteroidal antiinflammatory drugs as
compared with placebo in double-blind trials).

These results held across numerous types of trials. For example, the results
did not depend on whether physicians were aware of the treatment
assignments, whether standard treatments were also given, whether
determining the effect of placebo was an explicit objective of the study, or
who specified the main outcome (the original investigators or Hrobjartsson
and Gotzsche).

Hrobjartsson and Gotzsche conclude that there is no justification for the
use of placebos outside the setting of clinical trials. Their findings are
impressive, but is their conclusion too sweeping? First, they did find some
evidence of an effect in the important subgroup of trials in which the main
outcome was pain. Second, despite the large sample, the statistical power to
examine many subgroups of interest was low. Their data may have failed to
demonstrate a small but clinically useful benefit of placebo for some
patients and for some outcomes other than pain. Third, they found
statistical evidence of heterogeneity of results in studies with binary
outcomes. The results could not be heterogeneous unless at least one trial
differed from the others, which would require a real (though unidentified)
effect. Fourth, they studied patients in randomized clinical trials, many of
which focused on serious conditions whose clinical consequences may have
overshadowed small but useful effects of placebo. Fifth, they noted that the
low methodologic quality of some trials might explain a lack of effect,
though they found no association between dimensions of trial quality and
significant effects of placebo.

Finally, there is that pesky, utterly unscientific feeling that some things
just ought to be true. Perhaps most important is that the research setting,
with its generally intense methods of observation and precise measurement of
outcomes, may obscure a real effect of placebo that would be evident in
nonresearch settings. However, it is not clear how one could study and
compare the effects of placebo in research and nonresearch settings, since
that would of course require a research study.

Few physicians would argue against using innocuous means that might relieve
their patients' symptoms or reverse the course of illness. Unfortunately,
placebos may not be entirely innocuous. They may divert patients from
seeking more effective treatments, they may mask symptoms that need
attention, they add to the cost of treatment, and they may have unexpected
physiological effects. There may also be some reason for concern that
regular reminders of illness (in the form of placebos) may make a person
less rather than more comfortable. The deception that is inherent in the use
of placebos troubles some physicians as well as ethicists. This deception
may damage the doctor-patient relationship in subtle ways. There is thus
reason for caution about the casual acceptance of the notion that placebos
cannot hurt.

Overall, the uncompromising condemnation of placebos advocated by
Hrobjartsson and Gotzsche seems to me just a bit too sweeping. In
particular, the evidence that placebos might contribute to pain relief may
merit their continued therapeutic use when there is reason to think that a
patient may benefit. The Wizard of Oz did give each of the travelers
something of great value -- a heart, a brain, courage, hope. It was Toto the
dog, unimpressed by all the magical trappings, who ran behind the curtain
and brought down the whole scheme. However, I believe there should be a
sharp reduction in the prescription of placebos and careful justification
for each continued use. Future studies may show either that placebos have
benefits not yet documented or that the appearance of small benefits -- for
example, for pain relief -- is, in fact, illusory. At present, I would not
want to prescribe or receive a placebo without some reason that was far more
specific than weak evidence of some general "placebo effect."

John C. Bailar III, M.D., Ph.D., University of Chicago - Chicago, IL 60637

References
1. Tufte ER. Visual explanations: images and quantities, evidence and
narrative. Cheshire, Conn.: Graphics Press, 1997.
2. Gillespie R. Manufacturing knowledge: a history of the Hawthorne
experiments. Cambridge, England: Cambridge University Press, 1991.
3. Beecher HK. The powerful placebo. JAMA 1955;159:1602-6.
4. Baum LF. The wonderful Wizard of Oz. Chicago: G.M. Hill, 1900.
5. Hrobjartsson A, Gotzsche PC. Is the placebo powerless? An analysis of
clinical trials comparing placebo treatment with no treatment. N Engl J Med
2001;344:1594-602.


--
Adresse pour les messages destinés au forum E-MED:
<e-med@usa.healthnet.org>
Pour répondre à un message envoyer la réponse au forum
ou directement à l'auteur.
Pour toutes autres questions addresser vos messages à :
<owner-e-med@usa.healthnet.org>