Placebo Washouts Inflate Antidepressants in General Practice: Selection Bias
Geddes and Cipriani aim to restore some balance in the debate about
the risks of antidepressants.1 They claim that based on the evidence, the
suicide risk of selective serotonin reuptake inhibitors (SSRIs) is blown
out of proportion by the media. The methodological question they raise is:
what can be considered ‘best evidence’ in antidepressant trials? They fear
that regulatory bodies may overrate the importance of placebo-controlled
trials. Their head to head comparison to the older tricyclic
antidepressants indicates that SSRIs are of comparable efficacy.2
We disagree with Geddes and Cipriani and agree with the regulatory bodies.
The placebo controlled randomised trial is still the best test of efficacy
available. To put the methodological quality of the evidence about the
risks of antidepressants into perspective, Geddes and Cipriani describe
several already known types of biases, such as publication bias.
Recent meta-analyses show that the list of factors, and methodological
flaws, which may bias the contrast of antidepressant drugs versus placebo
is already quite long.3;4 However, differential selection bias by using a
placebo washout period has not yet been addressed.
Most antidepressant trials and especially the older ones have used
such a placebo washout period before randomisation. To demonstrate the
effects of placebo washouts: say 200 patients are available for a study,
and 20 placebo responders are excluded. The remaining 180 patients are
then randomised to antidepressant or placebo treatment. Without the
placebo wash out period, the ten excluded placebo-responders in the
placebo arm would all have been considered to have been recovered.
However, as it is not known whether all placebo responders in the
antidepressant arm would recover, the direction of this differential
selection favours antidepressants. Clinically, exclusion of placebo
responders appears logical. Our contention is however, that this custom
may also help to inflate the effects of antidepressants. Randomisation is
meant to provide sufficient evidence for the doctor in the surgery for the
decision whether this class of patients (with major depression) benefits
from an antidepressant prescription. Few GPs use a placebo washout in real
life practice.
Other known potential biases are: a) time: over the years, placebos become
more effective;3 4 b) type of placebo: the use of ‘active’ placebo’s,
which allow better blinding as they contain substances that mimic the
specific side-effects of antidepressants, diminishes the effectiveness of
antidepressant drugs, whereas most conventional trials use inert
placebos,5 c) a lack of concealment of allocation, yielding larger
estimates of treatment effect6 7, d) a lack of independent outcome
assessment, e) exclusion effects: many relevant groups such as older
patients are excluded from trials, and f) type of analysis: consistent use
of the more conservative and therefore more real-life intention-to-treat
(ITT) analyses in stead of per-protocol (PP) analyses decreases the risk
difference between antidepressant drugs and placebo from 28% to 19%.8
Several other potential biases have been suggested, such as setting
effects, as remarkably few studies have been performed in primary care,
and funding effects as almost all studies were sponsored by pharmaceutical
industries.6
A large, independently funded, good quality general practice trial of
antidepressants versus placebo is much needed now, without placebo
washouts.
2. Geddes JR, Freemantle N, Mason J, Eccles MP, Boynton J. SSRIs versus
other antidepressants for depressive disorder. Cochrane Database Syst Rev
2000;CD001851.
3. Stolk P, ten Berg MJ, Hemels ME, Einarson TR. Meta-Analysis of Placebo
Rates in Major Depressive Disorder Trials. Ann Pharmacother 2003;37:1891-
9.
4. Walsh BT, Seidman SN, Sysko R, Gould M. Placebo response in studies of
major depression: variable, substantial, and growing. JAMA 2002;287:1840-
7.
5. Moncrieff J, Wessely S, Hardy R. Active placebos versus antidepressants
for depression. Cochrane.Database.Syst.Rev. 2004;CD003012.
6. MacGillivray S, Arroll B, Hatcher S, Ogston S, Reid I, Sullivan F et
al. Efficacy and tolerability of selective serotonin reuptake inhibitors
compared with tricyclic antidepressants in depression treated in primary
care: systematic review and meta-analysis. BMJ 2003;326:1014.
7. Schulz KF, Chalmers I, Hayes RJ, Altman DG. Empirical evidence of bias.
Dimensions of methodological quality associated with estimates of
treatment effects in controlled trials. JAMA 1995;273:408-12.
8. Begg C, Cho M, Eastwood S, Horton R, Moher D, Olkin I, et al. Improving
the quality of reporting of randomized controlled trials: the CONSORT
statement. JAMA 1996;276:637-9.
Competing interests:
None declared
Competing interests:
No competing interests
02 December 2004
Harm WJ van Marwijk
senior researcher
Herman J. Adèr
VU University Medical Centre, Van der Boechorststraat 7, 1081BT Amsterdam, The Netherlands
Rapid Response:
Placebo Washouts Inflate Antidepressants in General Practice: Selection Bias
Geddes and Cipriani aim to restore some balance in the debate about
the risks of antidepressants.1 They claim that based on the evidence, the
suicide risk of selective serotonin reuptake inhibitors (SSRIs) is blown
out of proportion by the media. The methodological question they raise is:
what can be considered ‘best evidence’ in antidepressant trials? They fear
that regulatory bodies may overrate the importance of placebo-controlled
trials. Their head to head comparison to the older tricyclic
antidepressants indicates that SSRIs are of comparable efficacy.2
We disagree with Geddes and Cipriani and agree with the regulatory bodies.
The placebo controlled randomised trial is still the best test of efficacy
available. To put the methodological quality of the evidence about the
risks of antidepressants into perspective, Geddes and Cipriani describe
several already known types of biases, such as publication bias.
Recent meta-analyses show that the list of factors, and methodological
flaws, which may bias the contrast of antidepressant drugs versus placebo
is already quite long.3;4 However, differential selection bias by using a
placebo washout period has not yet been addressed.
Most antidepressant trials and especially the older ones have used
such a placebo washout period before randomisation. To demonstrate the
effects of placebo washouts: say 200 patients are available for a study,
and 20 placebo responders are excluded. The remaining 180 patients are
then randomised to antidepressant or placebo treatment. Without the
placebo wash out period, the ten excluded placebo-responders in the
placebo arm would all have been considered to have been recovered.
However, as it is not known whether all placebo responders in the
antidepressant arm would recover, the direction of this differential
selection favours antidepressants. Clinically, exclusion of placebo
responders appears logical. Our contention is however, that this custom
may also help to inflate the effects of antidepressants. Randomisation is
meant to provide sufficient evidence for the doctor in the surgery for the
decision whether this class of patients (with major depression) benefits
from an antidepressant prescription. Few GPs use a placebo washout in real
life practice.
Other known potential biases are: a) time: over the years, placebos become
more effective;3 4 b) type of placebo: the use of ‘active’ placebo’s,
which allow better blinding as they contain substances that mimic the
specific side-effects of antidepressants, diminishes the effectiveness of
antidepressant drugs, whereas most conventional trials use inert
placebos,5 c) a lack of concealment of allocation, yielding larger
estimates of treatment effect6 7, d) a lack of independent outcome
assessment, e) exclusion effects: many relevant groups such as older
patients are excluded from trials, and f) type of analysis: consistent use
of the more conservative and therefore more real-life intention-to-treat
(ITT) analyses in stead of per-protocol (PP) analyses decreases the risk
difference between antidepressant drugs and placebo from 28% to 19%.8
Several other potential biases have been suggested, such as setting
effects, as remarkably few studies have been performed in primary care,
and funding effects as almost all studies were sponsored by pharmaceutical
industries.6
A large, independently funded, good quality general practice trial of
antidepressants versus placebo is much needed now, without placebo
washouts.
Reference List
1. Geddes JR, Cipriani A. Selective serotonin reuptake inhibitors.
BMJ 2004;329:809-10.
2. Geddes JR, Freemantle N, Mason J, Eccles MP, Boynton J. SSRIs versus
other antidepressants for depressive disorder. Cochrane Database Syst Rev
2000;CD001851.
3. Stolk P, ten Berg MJ, Hemels ME, Einarson TR. Meta-Analysis of Placebo
Rates in Major Depressive Disorder Trials. Ann Pharmacother 2003;37:1891-
9.
4. Walsh BT, Seidman SN, Sysko R, Gould M. Placebo response in studies of
major depression: variable, substantial, and growing. JAMA 2002;287:1840-
7.
5. Moncrieff J, Wessely S, Hardy R. Active placebos versus antidepressants
for depression. Cochrane.Database.Syst.Rev. 2004;CD003012.
6. MacGillivray S, Arroll B, Hatcher S, Ogston S, Reid I, Sullivan F et
al. Efficacy and tolerability of selective serotonin reuptake inhibitors
compared with tricyclic antidepressants in depression treated in primary
care: systematic review and meta-analysis. BMJ 2003;326:1014.
7. Schulz KF, Chalmers I, Hayes RJ, Altman DG. Empirical evidence of bias.
Dimensions of methodological quality associated with estimates of
treatment effects in controlled trials. JAMA 1995;273:408-12.
8. Begg C, Cho M, Eastwood S, Horton R, Moher D, Olkin I, et al. Improving
the quality of reporting of randomized controlled trials: the CONSORT
statement. JAMA 1996;276:637-9.
Competing interests:
None declared
Competing interests: No competing interests