The placebo response in medicine: Minimize, maximize or personalize?

Abstract

Our understanding of the mechanisms mediating or moderating the placebo response to medicines has grown substantially over the past decade and offers the opportunity to capitalize on its benefits in future drug development as well as in clinical practice. In this article, we discuss three strategies that could be used to modulate the placebo response, depending on which stage of the drug development process they are applied. In clinical trials the placebo effect should be minimized to optimize drug-placebo differences, thus ensuring that the efficacy of the investigational drug can be truly evaluated. Once the drug is approved and in clinical use, placebo effects should be maximized by harnessing patients' expectations and learning mechanisms to improve treatment outcomes. Finally, personalizing placebo responses - which involves considering an individual's genetic predisposition, personality, past medical history and treatment experience - could also maximize therapeutic outcomes.

Full text

For many years, randomized double-blind
placebo-controlled trials (RCTs) have been
conducted to disentangle the specific effects
of a therapeutic intervention — such as the
administration of a drug in pill form —
from unspecific effects that could be due
to the general nature of the intervention.
For example, in the case of trials evaluating
the effects of a drug, this typically involves the
administration of an inert substance —
the placebo — in the same form and manner
to patients in the control arm of the trial.
It is well known that clinical improvements
occur following placebo treatments in RCTs,
and these placebo effects can be substantial.
Indeed, in some therapeutic areas, such as
psychiatric disorders or pain, large placebo
responses can be an important factor
contributing to the failure of clinical trials
to demonstrate a meaningful therapeutic
effect for adrug.
Given the importance of placebo
responses in medicine, in recent years there
has been considerable research focused on
elucidating the mechanisms underlying
these effects. The knowledge that is
being gained regarding the mechanisms
orchestrating the placebo response in
different physiological systems and diseases
is paving the way towards the systematic
utilization of the placebo response in clinical
research and patient care in order to achieve
the following outcomes: minimization of the
placebo response to improve the sensitivity
of clinical trials; maximization of the placebo
response to enhance patient benefit from
any medical treatment strategy; and person-
alization of placebo responses by consider-
ing the individual patient’s characteristics,
including genetics, underlying disease,
personality, past medical history and
treatment experience.
In this article, we briefly review the
evidence in each of these areas and propose
recommendations for the application of the
placebo response in the different stages of
drug development and clinicaluse.
Mechanisms of placebo responses
From a psychological point of view, placebo
responses can be triggered by various inter-
related environmental and psychosocial
factors that make up the treatment context.
The principal underlying mechanisms that
are best supported by empirical evidence
are as follows: expectancy (that is, patients’
expectations of the benefit of a treatment)
(FIG.1a); behavioural conditioning (FIG.1b);
and the quality of the patient–physician
relationship1–6.
Converging evidence from research over
the past few decades has revealed that these
psychological factors trigger complex
neuro biological phenomena involving the
contribution of distinct central nervous
system mechanisms as well as peripheral
physiological mechanisms, including the
release of endogenous substrates (BOX1).
For instance, placebo analgesia — one of
the most robust and best studied placebo
responses — is mediated by changes in
neural activity in the dorsolateral prefrontal
cortex, the anterior cingulate cortex,
the amygdala and the periaqueductal grey.
All of these regions of the brain are key
players in the so-called ‘descending pain
modulatory network’7,8 that amplifies or
inhibits incoming pain signals, even at the
level of the spinal cord9. During placebo
analgesia, increased activity in this network
triggers decreased activity in somatosensory
pain areas and the subsequent analgesic
effect. Alterations in brain activity are
accompanied by changes in brain neuro-
chemistry, particularly in the endogenous
opioid and the endogenous cannabinoid
systems that, intriguingly, seem to differen-
tially contribute to different types of placebo
analgesia10–16.
Similarly, the involvement of system-
specific changes in the brain as well as
peripheral changes seems to underlie
placebo responses in many other physio-
logical systems or medical conditions.
In Parkinson’s disease, placebos trigger
endogenous dopamine release in the basal
ganglia17,18, which leads to improvements
in motor function. In anxiety and depres-
sion, placebo responses are associated with
increased activity in neural networks related
to emotional regulation19–21 and can be
predicted by genetic polymorphisms
modulating monoaminergic tone; for
example, the tryptophan hydroxylase 2
(TPH2) polymorphism is a significant
predictor of clinical placebo responses
in social anxiety22.
OPINION
The placebo response in medicine:
minimize, maximize or personalize?
Paul Enck, Ulrike Bingel, Manfred Schedlowski and Winfried Rief
Abstract | Our understanding of the mechanisms mediating or moderating the
placebo response to medicines has grown substantially over the past decade and
offers the opportunity to capitalize on its benefits in future drug development as
well as in clinical practice. In this article, we discuss three strategies that could be
used to modulate the placebo response, depending on which stage of the drug
development process they are applied. In clinical trials the placebo effect should
be minimized to optimize drug–placebo differences, thus ensuring that the
efficacy of the investigational drug can be truly evaluated. Once the drug is
approved and in clinical use, placebo effects should be maximized by harnessing
patients’ expectations and learning mechanisms to improve treatment outcomes.
Finally, personalizing placebo responses — which involves considering an
individual’s genetic predisposition, personality, past medical history and treatment
experience — could also maximize therapeutic outcomes.
PERSPECTIVES
NATURE REV IEWS
|
DRUG DISCOVERY VOLUME 12
|
MARCH 2013
|
191
© 2013 Macmillan Publishers Limited. All rights reserved

Nature Reviews | Drug Discovery
a Expectations b Behavioural conditioning
Before conditioningHidden application Open application
Pain relief
Pharmacological effect
Expectancy-related
(placebo) effect
Hidden Open
Neutral stimulus No response
Unconditioned stimulus Unconditioned response
HO
O
HO
NCH3
Evocation
Conditioned stimulus Conditioned response
Acquisition
Unconditioned stimulus + neutral stimulus
+
Unconditioned response
HO
O
HO
NCH3
Medication
is administered
by a machine
(unbeknown
to the patient)
Medication is
administered
by a physician
In addition, behaviourally conditioned or
learned placebo responses modulate periph-
eral immune responses (such as lymphocyte
functions and cytokine production)23–27 as
well as neuroendocrine functions (such as
the release of glucocorticoids and growth
hormone)28,29. Similar placebo-induced
changes have been reported in other physio-
logical systems, including respiratory, cardio-
vascular and gastrointestinal systems (BOX1).
Research on the neurobiological mecha-
nisms underlying placebo responses is still
at a very early stage, and many unanswered
questions remain30. For example, the shared
and distinct contributions of different
neural networks to different types of placebo
responses, as well as the communication
between central and peripheral nervous and
end-organ systems, still need to be eluci-
dated. Nevertheless, these data indicate
that the placebo response has neurobio-
logical underpinnings that can be targeted
by specific interventions to influence
therapeutic outcome.
Minimizing placebo responses
Minimizing placebo responses and/or
optimizing drug–placebo differences (that
is, improving ‘assay sensitivity’) are scientific,
ethical and regulatory requirements when
testing new compounds during drug develop-
ment. These requirements serve to prevent an
overestimation of the potency of a drug under
evaluation in comparison with other effects
that occur with all medical treatment, such as
spontaneous variations in disease symptoms,
regression to the mean and methodological
Figure 1 | Mediation of placebo responses by cognitive factors and
behavioural conditioning with pharmacological stimuli. a | The pivotal
role of cognitive factors, such as patients’ expectations, in mediating
placebo responses is best illustrated by the so‑called open/hidden drug
paradigm68. In this paradigm, identical concentrations of active drugs are
administered by a physician in a visible (open condition) or hidden manner,
in which the patient is unaware of the timing of administration of the medi‑
cation (for example, a computer is used to control infusion timing). This
permits the dissociation of the pure pharmacodynamic effect of the treat‑
ment (hidden treatment) from the additional benefit of the psychological
context that comes from knowing that the treatment is being administered.
The difference between the outcomes following the administration of the
expected and unexpected therapy can be seen as the placebo or psycho‑
logical component, even though no placebo treatment has been used.
b | In the context of behavioural conditioning (associative learning), the
unconditioned stimulus (for example, a pharmacological agent) induces
a response in the central nervous system (termed the unconditioned
response); a neutral stimulus (for example, environmental stimuli or an
inert substance) induces no such response. During the acquisition phase,
the neutral stimulus is paired with the unconditioned stimulus. After one
or several pairings of the neutral stimulus with the unconditioned stimu‑
lus, the neutral stimulus becomes the conditioned stimulus. During the
evocation period, the conditioned stimulus is able to mimic the effects
formerly induced by the unconditioned stimulus.
PERSPECTIVES
192
|
MARCH 2013
|
VOLUME 12 www.nature.com/reviews/drugdisc
© 2013 Macmillan Publishers Limited. All rights reserved

Nature Reviews | Drug Discovery
Mood (anxiety,
depression, addiction) Motor functions
Respiratory functions
Cardiovascular functions Endocrine functions
Immune functions Gastrointestinal functions
Pain
DLPFC
ACC
Am
Expectations,
conditioning
PAG
biases, which are collectively termed as
unspecific effects31 (FIG.2a). To address these
unspecific effects, RCTs have become the
standard study design since the 1950s32.
The legal and scientific principles for
conducting RCTs (for example, the number
of patients to include, the duration of the
trials and the registration, monitoring and
statistical evaluation of RCTs) have been
specified and tested. Many design variants
and features have been developed for meth-
odological and/or ethical reasons over the
past 30years, and their suitability for mini-
mizing the placebo response and optimizing
drug–placebo differences is discussed below
and summarized in BOX2.
Strategies utilized. One strategy that has
been used to improve assay sensitivity is to
increase the number of patients who are
randomized to the active treatment group
(called enrichment trials or multidosing
studies). However, this strategy resulted in
even higher placebo responses in clinical
trials of migraine33, depression34 and schizo-
phrenia35, compared to trials in the same
indications with equal numbers of patients
in both the drug and placebo groups.
This finding is in line with studies showing
that the chance of being in a treatment
group increases the magnitude of placebo
responses18,36–38.
These clinical trials were, however,
mainly based on patient-reported outcomes
(PROs); accordingly, further meta-analyses
are needed to investigate the impact of the
randomization ratio in other medical condi-
tions, particularly with respect to objective
outcome measures such as biochemical or
physiological markers of disease. Together
with statistical considerations, the current
status strongly argues in favour of equal
sample sizes (where there is a 50% chance
of being assigned to either group when
randomizing patients into two groups) to
optimize drug–placebo differences38.
Approaches that identified and excluded
placebo responders at an early stage of
an RCT (for example, by placebo run-in
periods or by repeated treatment phases
with re-randomization) could not prevent
the occurrence of placebo responses at a
later phase of the RCT39. However, placebo
responders required less dose adjustment
than placebo non-responders when dose
escalation was offered40. Randomized run-in
and withdrawal periods appear to be more
effective at optimizing drug–placebo differ-
ences41 (FIG.2b) but so far they have rarely
been tested for their efficacy in improving
assay sensitivity.
Box 1 | Placebo responses and underlying mechanisms
Placebo responses are primarily mediated via cognitive factors, such as patients’ expectations
regarding treatment benefits, and by behavioural conditioning or associative learning processes.
These psychological factors trigger complex neurobiological phenomena involving distinct central
nervous system (CNS) as well as system-specific, peripheral, physiological and end-organ changes
(see the figure). The CNS mechanisms initiating and mediating placebo responses are best
characterized for placebo analgesia and involve the descending pain modulatory network,
which includes the dorsolateral prefrontal cortex (DLPFC), the anterior cingulate cortex (ACC),
the amygdala (Am) and the periaqueductal grey (PAG). Similar regions of the brain have been shown
to contribute to emotional placebo responses. The shared and distinct contributions of different
brain networks in other types of placebo responses is currently unknown. Nevertheless, some of the
mechanisms steering the placebo response in different physiological systems or disease states have
been described and are listed below.
Mood (anxiety, depression, addiction). Mechanisms include: changes in metabolic or electric
activity in the ACC, orbitofrontal cortex, ventral striatum and amygdala22,19–21; genetic variants of
tryptophan hydroxylase 2 (REF.22), monoamine oxidase107 and catechol-O-methyltransferase
(COMT)107; as well as changes in metabolic activity in different regions of the brain during the
open/hidden administration of methylphenidate123.
Respiratory functions. Mechanisms include: conditioning of opioid receptors in the respiratory
centres124; and expectation-induced changes in bronchoconstriction and bronchodilation125.
Cardiovascular and autonomic functions. Mechanisms include: reduction of β-adrenergic receptor
activity in the heart126; changes in coronary diameter127; changes in systolic blood pressure128;
and changes in neuronal excitability in specific regions of the brain during open/hidden deep-brain
stimulation, which evoke expectancy-related differences in autonomic responses (heart rate)129.
Immune functions. Mechanisms include: behaviourally conditioned suppression of cytokine release
(for example, interleukin-2 and interferon-γ) and lymphocyte activity24,25,27,90 mediated via the insular
cortex and the amygdala; and sympathetic β-adrenergic receptor-dependent mechanisms26.
Gastrointestinal functions. Mechanisms include: behaviourally conditioned reductions in symptoms;
changes in cingulate and prefrontal activity130,131; and changes in gastric motility132.
Endocrine functions. Mechanisms include: behaviourally conditioned release of growth hormone
and cortisol28,29; and conditioned corticosteroid effects in patients with psoriasis82.
Pain. Mechanisms include: activation of endogenous opioids10–15 and dopamine133,134; activation of
cannabinoid receptors in non-opioid placebo analgesia11; and genetic variants of COMT135.
Motor functions. Mechanisms include activation of dopamine in the striatum17,18 and changes in the
activity of neurons in the basal ganglia and thalamus in patients with Parkinson’s disease136, as well as
changes in neuronal excitability in limbic areas of the brain induced by deep-brain stimulation127.
PERSPECTIVES
NATURE REV IEWS
|
DRUG DISCOVERY VOLUME 12
|
MARCH 2013
|
193
© 2013 Macmillan Publishers Limited. All rights reserved

Nature Reviews | Drug Discovery
a Factors influencing outcomes
Treatment outcome (arbitrary units)
0
20
40
60
80
Effect
Placebo Drug
1
PPatient 1
Days
Patient 2
2
D
3
D
4
D
5
D
6
D
7
Trial start
b Run-in and withdrawal
c Zelen design
End of trial
D
P PDDDDD ...
...
...
...
PPatient 3
Patient 4
P P D D D D
...
...
PPatient 5 P P P P D D ...
P P P P D D D
t
D
u
D
v
P
w
P
x
P
y
P
z
P
D D D P P P P
D D D D P P P
D D D D D D P
DDDDDP P
Individual x...
PIndividual yP P P P P P ...
PPPPPPP
P P P P P P P
PPPPPPP
Natural course
Patient population
Patient recruitment
Approval for an observational study
Approval for an intervention study
Regression to the mean
Methodological bias
Contextual factors
(such as expectations,
conditioning)
d RCTs: the additive model
Treatment outcome (arbitrary units)
0
20
40
60
80
Effect
Placebo Drug
Additive drug-specific effects
Placebo effect
Additive drug-specific effects
Placebo effect
Additive drug-specific effects
Interaction
Placebo effect
RCTs: the interactive model
Treatment outcome (arbitrary units)
0
20
40
60
80
Effect
Placebo Drug
Random selection
Health monitoring:
No-treatment controls
Study evaluation:
Drug, placebo, comparator
Crossover designs do not seem to opti-
mize assay sensitivity either. If different
treatment periods are applied to the same
patients, carry-over effects occur and an
adequate interpretation of trial results is
hindered by within-subject variability of
the placebo response. The utility of cross-
over trials in minimizing placebo responses
has been questioned because treatments in
the first phase may generate behavioural
conditioning effects during the second
phase3,42. Crossover designs may also lead to
un-blinding of the study owing to perceived
differences in side effects43. Although the
risk of un-blinding could be controlled by
using an active placebo that mimics the side
effects of the drug under evaluation, active
placebos are difficult to develop and there-
fore used only occasionally in a few clinical
conditions — for example, in the treatment
of depression44. Without active placebos,
however, it is difficult to determine whether
drugs have beneficial effects because of their
genuine pharmacological action or because
they induce side effects that trigger expec-
tations of a positive outcome45. Although
active placebos are likely to increase the
placebo response, they may help to reduce
false-positive results of drug testing.
We therefore propose that the use of active
placebos should be reconsidered as they
are a promising strategy for detecting valid
drug–placebo differences inRCTs.
Recent approaches to improve assay
sensitivity that have been reported include
identifying drug responders rather than
placebo responders during a run-in phase46
or pre-selecting patients who were previ-
ously exposed to a similar drug47. This may
Figure 2 | Design features to illustrate and control components of the
placebo effect in clinical trials. a | In comparison with the effect in the
placebo group in a randomized double‑blind placebo‑controlled trial (RCT),
the effect in the drug group is composed of drug‑specific (orange) and
unspecific effects (blue), with the latter including the natural course of the
disease, statistical effects (regression to the mean), methodological biases
and contextual factors (the placebo response). The relative contribution of
each component may vary depending on the clinical condition, study
design and the nature of the primary and secondary end points.
b|Schematic drawing of the randomized run‑in and withdrawal design41:
patients 1 to 5 start treatment at the same time but receive placebo (P)
initially for a variable period of time before being switched to the drug (D)
in a double‑blinded manner. Similarly, at the end of a set period of the study
patients are switched from the drug to placebo at variable time points.
Individuals x and y receive placebo throughout the entire study. c|The
(modified) Zelen design49 separates the recruitment of patients for an obser‑
vational study from the recruitment of patients for an intervention study.
Provided the sample size for the observational study is large enough and
the recruitment of patients for the intervention study does not create a
selection bias, this allows the natural course of the disease to be monitored
without randomizing patients to a ‘no‑treatment controls’ group. d | In com‑
parison with the additive model50 of RCTs, which assumes unspecific effects
of equal size in both the placebo and the drug groups of studies (left plot),
the interactive model (right plot) assumes that drug‑specific effects may
interact with the placebo responses to result in unequal placebo effects in
the two study groups.
PERSPECTIVES
194
|
MARCH 2013
|
VOLUME 12 www.nature.com/reviews/drugdisc
© 2013 Macmillan Publishers Limited. All rights reserved

improve assay sensitivity by increasing the
drug–placebo difference35, but it carries the
risk of biasing for a selected drug indication:
if a drug works only in a predetermined or
preselected patient group — for example,
in women — it may receive only a limited
indication when it is being considered for
regulatory approval. Any previous treat-
ment will affect a new therapy through the
effects of behavioural conditioning as well as
the positive or negative expectations of the
individuals participating in the trial. As a con-
sequence, documentation of the patient’s dis-
ease history, prior disease management and
the patient’s previous participation in RCTs
needs to become part of the baseline docu-
mentation in any RCT (as discussed below).
Variables to control for. A major challenge of
most RCTs is the fact that spontaneous remis-
sion and fluctuation of disease symptoms need
to be controlled for. A separate ‘no-treatment
controls’ group (in which the natural course
of the disease is followed) in clinical trials is
ethically questionable and demotivating for
patients during recruitment48, and ‘treatment
as usual’ as an alternative is difficult to stand-
ardize and monitor. However, controlling for
the natural course of the disease is essential
and the only possibility for dissecting spe-
cific placebo responses from general placebo
effects. Therefore, natural course conditions
should be incorporated more frequently
into RCTs. A recently favoured alternative to
standard RCTs is the ‘Zelen’ design49 (FIG.2c),
in which recruitment and consent to partici-
pate in an observational group (that is, treat-
ment as usual) is separate from recruitment
and consent to participate in an RCT (either
a placebo-controlled trial or a CER trial).
The rationale of RCTs has been ques-
tioned because of their basic assumption that
the placebo effects in the placebo group are
identical to the placebo effects in the drug
group50 and that both combine in an additive
manner (FIG.3a). In recent years, accumulating
evidence has indicated that this assumption
may not be true under all conditions; indeed,
placebo responses can differ between drug
groups and placebo groups51 (FIG.2d).
Another basic assumption is that minimizing
placebo responses perse would improve the
assay sensitivity. However, other associations
between the size of placebo mechanisms
and the potential to detect significant drug–
placebo differences are possible under certain
circumstances (FIG.3b).
One approach for circumventing the
placebo dilemma in RCTs is to abandon or
discourage the use of placebos in clinical
trials; this approach has recently been
Box 2 | Strategies to optimize drug–placebo differences
The advantages and limitations of strategies used to minimize the placebo response and/or to
optimize drug–placebo differences are outlined below.
Traditional design features
•Extending the trial duration may eventually decrease the placebo response, but long-term
(longerthan1year)randomizeddouble-blindplacebo-controlledtrials(RCTs)alsoshowhigh
placebo response rates137
•Crossover trials potentiate the risk of learning processes (conditioning) and carry-over effects,
and increase the placebo response42
•The use of placebo run-in phases can exclude placebo responders but does not prevent placebo
responses during the medication phase138
•Using treatment-naive patients blunts the placebo response but does not affect the drug response47
•Increasing the number of trial groups increases acceptance owing to higher chances of receiving
medication, but also increases the placebo response34
•Single-case (N = 1) studies require a very high number of individuals, but the placebo response is
difficult to assess139
•Using active placebos is difficult for most drugs but would substantially increase drug–placebo
differences44
•Usingacomparatordrug(inacomparativeeffectivenessresearch(CER)trial)increasesthe
acceptance of a candidate to enter a drug study (as they are guaranteed to receive a treatment)
but also substantially increases the placebo response55
•Repetitivetreatmentperiodsallowre-randomizationbutdonotpreventtheplaceboresponse
and do not predict the placebo response during all periods39
Novel design features
•The use of drug run-in phases helps to identify drug non-responders and augments drug–placebo
differences, but this approach is biased113
•Randomizedrun-inandwithdrawalperiodsmayidentifysomebutnotallplaceboresponses,
and allow better discrimination between the drug and placebo140
•In adaptive dosing and group allocation studies placebo responders tend to demand fewer dose
adjustments40, and enrichment of groups increases the placebo response34
•In two-way enriched design (TED) studies placebo non-responders and drug responders are
re-randomized to receive the drug or placebo, but these have studies not yet been evaluated140
•The preference design reduces disappointment and drop-out rates specifically in the placebo
group,whichisoptimalforCERtrials141
•The step-wedge design (waiting lists) induces disappointment and drop-out, but improvements
in the placebo response may be observed during the waiting period142
•The Zelen design (classical and modified) allows the use of ‘no-treatment controls’ without
randomization, and allows improved separation of the placebo response from the natural course
of the disease49
•Cluster randomization of health service delivery units that either provide drug or placebo may
allow better patient recruitment but still requires fully informed consent143
Other measures of quality
•Patient-reportedoutcomes(PROs)insteadofphysicianreportsmayincreasetheplacebo
response to some extent59
•UsingbiomarkersofdrugefficacyinsteadofPROsmayreducetheplaceboresponse,
but biomarkers of the placebo response would be preferred60
•Standardization of symptom severity; placebo responders are frequently patients with less
severe symptoms when they enter the study61
•Controlling for centre effects by standardizing recruitment and separating it from study
conductance may reduce the placebo response63
•Personality profiling of the placebo response is possible but currently no valid psychometric
test exists that reliably predicts the placebo response108,144
•Controlling for patient expectations is seldom done, which may eliminate patients with
inappropriate expectations and a high placebo response74
•Increasing medication adherence is associated with higher responses in the placebo group64
•Assessing and controlling effective blinding; incomplete blinding is frequent and a common
reason for drop-out in the placebo group65
•Responderanalysisallowspost-hocidentificationoftheresponderstoend-pointdefinitions
and may improve future patient selection66
PERSPECTIVES
NATURE REV IEWS
|
DRUG DISCOVERY VOLUME 12
|
MARCH 2013
|
195
© 2013 Macmillan Publishers Limited. All rights reserved

Nature Reviews | Drug Discovery
a Linear associations
Treatment outcome (arbitrary units)
0
20
40
60
80
Placebo mechanisms
Minimized Maximized
1
XDrug
Day
Placebo
2
X
3
X
4
X
5
X
6
X
7
Acquisition period
c Placebo-controlled dose reduction
Maintenance or evocation period
X
8
X
9
X
10
X
11
X
12
X
13
X
...
...
...
Drug group
Placebo group
i (additive model)
iii
ii
b Optimal window (interactive)
Treatment outcome (arbitrary units)
0
20
40
60
80
Placebo mechanisms
Minimized Maximized
Drug group
Placebo group
favoured by regulatory authorities, boards of
medical societies and ethics committees52–54.
Placebo-controlled drug trials withhold
effective treatment in a substantial subgroup
of patients for scientific reasons, whereas
ethical standards (the Declaration of Helsinki)
require that all patients receive the best treat-
ment available. This dilemma is avoided in
CER trials, which deliver active treatment
to all patients. CER trials compare novel
compounds to approved drugs or standard
th e r apy.
However, as shown in several meta-
analyses in depression34 and schizophrenia55,
comparing a new drug to an active compara-
tor in a head-to-head trial results in a higher
clinical efficacy of the drug than when it is
compared to placebo in a conventional RCT.
The drug response is thus increased solely
because, compared to placebo-controlled
RCTs, in CER trials there is a higher chance
(100%) that patients receive active treat-
ment. Consequently, CER trials increase the
placebo response without being able to con-
trol for it. Another methodological problem
with CER trials is the selection of a ‘fair’ and
adequate comparator 56.
Finally, CER studies raise specific
statistical problems with the test for the
‘non-inferiority’ of the compound under
investigation57. In a trial intending to
demonstrate equal efficacy of a novel drug
with a comparator, the difference between
both drugs should not be lower than a
certain degree called the equivalence
margin (set either by the investigator a
priori or by regulatory authorities), which
is based on clinical data of available drugs
for the same indication. A non-inferiority
trial tests the null hypothesis that the novel
drug is inferior by the equivalence margin.
If this null hypothesis is rejected, the novel
compound can be regarded as clinically
equivalent. This, on average, requires a
fourfold larger sample size compared to
classical RCTs58 and is one of the reasons
why drug companies are increasingly
concerned by the growing costs of drug
development57.
Placebo responses seem to be more
pronounced in trials in which the outcomes
are measured by PROs compared to those
in which they are assessed by physicians59.
However, placebo responses are higher in
outcomes based on questionnaires than in
outcomes based on clinical biomarkers60.
Therefore, developing biomarkers of dis-
eases with sufficient sensitivity to detect
clinical improvements during therapy
may improve the assay sensitivity of future
RCTs. However, biomarkers of therapeutic
efficacy are not readily available for most
diseases. Moreover, variables of subjective
wellbeing are crucial from a patient’s point
of view and cannot be fully replaced by
biomarkers.
Finally, there are several general indices
used to measure the quality of the design of
clinical trials that affect placebo responses
and subsequent assay sensitivity. These
measures include standardizing for symp-
tom severity61, avoiding physician’s selection
bias during recruitment62, controlling for
centre effects (physician training, separation
of recruitment and study conductance)63,
controlling for patient adherence64, control-
ling for and ensuring effective blinding65 as
well as performing (post-hoc) assay sensitivity
analyses of RCTs66. Researchers should
carefully consider these measures in order
to minimize the placebo response and
maximize drug–placebo differences.
Future study designs. Further exploration is
needed to determine whether an ideal study
design exists or whether it requires disease-
specific adaptation. However, such a design
should incorporate the features discussed
above to optimize assay sensitivity; our
interpretation of an ideal study design is
provided in FIG.4. Importantly, an ideal
study should systematically assess and/
or standardize the key factors moderating
placebo responses, such as the past medi-
cal history and treatment expectations of
Figure 3 | Is there an optimal window for the detection of drug–placebo differences? Clinical trial
designs can either be enriched (for example, through increased patient–physician contact and
optimized treatment expectations; see BOX2) or impoverished (for example, through Internet‑
administered treatment schedules and limited social contacts associated with study participation) in
terms of placebo mechanisms. To date, it is unclear how these factors influence the detection of drug‑
specific effects. a | If drug‑specific effects are constant factors that contribute to placebo responses
(as represented by the additive model50; see also FIG.2d), the detection of drug–placebo differences
would be independent of the level of the placebo mechanisms associated with the study design (paral‑
lel orange line (i) to the blue line). However, increased placebo responses could make it more difficult
to detect drug–placebo differences (represented by the dashed line ii). Vice versa, placebo mecha‑
nisms could also enhance the effects of the drug, thus leading to larger drug–placebo differences in
trials that are enriched with placebo mechanisms. For example, positive expectations further increase
after patients experience positive drug responses following the first dose, and this effect can be
further amplified through conditioning processes that are specific for the drug group (represented by
the dashed line iii). b | A defined window representing an optimized consideration of placebo mecha‑
nisms in clinical trial designs could exist, which could facilitate the detection of drug efficacy.
Accordingly, both over‑enriched paradigms (activating many placebo mechanisms) and impoverished
study designs (with very limited placebo mechanisms) could hinder the detection of specific treatment
effects and thereby impede drug discovery. c | Placebo‑controlled dose reduction as a partial rein‑
forcement strategy could be applied to a drug regimen. After an acquisition and conditioning period
(days 1 to 6), placebo administration could be interspersed (for example, on day 7, day 10, day 12, and
so on) with the aim of achieving dose reduction while maintaining drug efficacy.
PERSPECTIVES
196
|
MARCH 2013
|
VOLUME 12 www.nature.com/reviews/drugdisc
© 2013 Macmillan Publishers Limited. All rights reserved

Nature Reviews | Drug Discovery
a
b
50:50 randomization
Inclusion of a no-treatment control group (Zelen design)
Randomized and double-blinded run-in and withdrawal
Two duration strategies: repetitive short-term and long-term studies
Adaptive dosing (step-up/step-down) at fixed time points
Primary end point: a combination of biomarkers and patient-reported outcomes*
Build-in ‘placebo modules’: variation in patient information, communication‡
Patient recruitment via the treating physician
Patients documented for experience with drugs
Optimized physician–patient communication; training of empathic behaviour‡
Optimization of patient information
Primary end point: patient-reported outcomes; secondary end point: biomarkers
Adaptive dosing (step-up/step-down) at variable time points
Open application of drugs; salient context stimuli when applying the drug
Use of active placebos
Drug run-in and withdrawal at fixed time points
Inclusion of a no-treatment control group (Zelen design)
More patients randomized to drug and/or comparator than to placebo
Preference first, then randomization
Preference first, then randomization
Assessment of effectiveness of blinding Pre- and post-assessment of preferences
Inclusion of treatment-naive patients
Multicentre trial
Post-hoc responder analysis including patient and physician data
Separation of patient recruitment and study conductance
Patients recruited via a patient registry§ to allow inclusion of individual patient data (past medical history,
psychometrics, previous trial participation, and so on) into trial evaluation without violating anonymity
Patient training: repeated assessment and standardization of expectations; verification of
patient
expectations aer pre-trial information; assessment of patient’s rating about blinded
group allocation
Centres selected from a physician registry§:
core physician and centre characteristics documented in registry
Physician training: explicit definition of communication rules and training to harmonize and
standardize physician’s behaviour
Optimization of patients’ expectations at trial start and during treatment (for example, optimizing
interventions; inclusion of patients only aer successful optimization of expectations
Powerful pretreatments to optimize learning processes; selective analyses
of subgroups with different treatment pre-experiences
Patients stratified for experience with drugs
Three-arm trials (drug, comparator, placebo)
Placebo-controlled trial Comparator trial
individual patients. Managing treatment
expectations of patients (as discussed in
more detail below) may affect many aspects
of RCTs, including the design of patient
information leaflets, improvements in the
understanding and acceptance of informed
consent and the training of study nurses
and physicians to standardize the patient–
clinician relationship.
Experimental designs that have mostly
been used in laboratory conditions in
healthy volunteers (for example, the bal-
anced placebo design and the balanced
crossover design)67, strategies such as
hidden treatment68 (FIG.1) or novel designs
currently without empirical proof69 need to
be explored for their potential applicability
in the testing of new drugs specifically in
chronically ill patients. However, these
designs may have serious limitations as they
require a level of deception that may be
acceptable in healthy volunteers but not
in studies involving patients70.
In summary, many of the designs and
study features that have been developed
for scientific, methodological, ethical or
other reasons do not appear to reliably
minimize the placebo response and optimize
drug–placebo differences. By contrast,
they seem to increase, obscure or even
ignore the placebo response. However,
a critical re-evaluation of these innova-
tions in clinical trial design, with a view
to enhancing drug–placebo differences,
may contribute to improvements in drug
development.
In addition to the selection of a general
study design, investigators have the oppor-
tunity to directly exploit specific placebo
mechanisms. Patients’ expectations are not
only intimately linked to the characteristics
of the study design but can also be directly
manipulated by the information that is given
to them. Pretreatments with low efficiency
(or placebo run-in trials) may reduce
placebo responses to subsequent treatments,
but they also run the risk of reducing the
effects of the drug under evaluation. A poor
physician–patient relationship and reduced
direct interaction time can also hamper
placebo responses. These factors can vary
substantially between trials, and the opti-
mum level of physician–patient interaction
required to detect drug–placebo differences
remains unclear. Although further research
is needed to answer the question of what the
optimum level physician–patient interaction
is, the potential of utilizing these mecha-
nisms to intentionally increase placebo
responses in clinical care is highlighted in
the next section.
Figure 4 | What are the features of an ideal study design? For the evaluation of drug efficacy, two
variants of an ideal study design are appropriate. One involves testing novel drugs during drug develop‑
ment (PhaseII and III studies), with an emphasis on minimizing the placebo response and optimizing
drug–placebo differences (assay sensitivity) (part a); and the second involves testing approved drugs
under routine clinical practice (‘PhaseIV’ studies), which could be associated with maximizing the
placebo response (part b). Although the regulatory approval of drugs requires clear causality in highly
specific scientific designs, their broader dissemination in health‑care systems requires the proof that a
drug contributes to the efficiency of optimized placebo factors in clinical practice. Both trial designs
must aim to fully control for all mechanisms that can contribute to placebo responses. If variations of
placebo mechanisms are left uncontrolled, it will be more difficult to document any specific effects of
a drug. Therefore, all aspects affecting patients’ expectations, learning processes, differences in drug
pretreatments or interactions between patients and medical personnel involved in the trial should be
clearly defined in the study protocol. *Subjective assessment of therapy outcome instead of biomarkers.
‡To allow controlled assessment of unspecific effects; for example, the level of physician–patient
contact. §The ethical and legal rules of such registries are unknown but need to be explored.
PERSPECTIVES
NATURE REV IEWS
|
DRUG DISCOVERY VOLUME 12
|
MARCH 2013
|
197
© 2013 Macmillan Publishers Limited. All rights reserved

Maximizing placebo responses
Placebo responses contribute substantially
to clinical outcome in most — if not all —
medical treatments (BOX1). However, the
potential of placebo mechanisms is far from
being systematically used in daily clinical
practice. From the perspective of a patient
and the physician, maximal drug efficacy
is desirable irrespective of whether the
improvements are based on specific pharma-
cological effects, placebo mechanisms or a
combination of both. Accordingly, although
placebo responses should be controlled for
and reduced in clinical trials to improve
assay sensitivity, in the clinical setting the
mechanisms underlying placebo responses
(that is, treatment outcome expectations,
conditioning and learning processes as well
as the physician–patient relationship) should
be systematically exploited to maximize
treatment benefits (BOX3).
The use of placebos in clinical practice.
It seems that prescribing pure placebos
(inert substances) and impure placebos
(drugs that are not indicated for the medical
problem in question) is common in clinical
practice71. However, the use of pure and
impure placebo treatments without trans-
parent disclosure to patients is limited by
legal and ethical issues such as patient auto-
nomy and informed consent; a discussion
of these issues is beyond the scope of this
article but they have been comprehensively
reviewed in the literature72.
Nevertheless, in a recent RCT, open-label
placebo application substantially reduced
symptom severity in patients with irritable
bowel syndrome compared to ‘no-treatment
controls’; this indicates, against conventional
wisdom, that even the use of pure placebos
without deception is possible73. These semi-
nal findings support the promising notion
that it might be possible to circumvent
the legal and ethical constraints of current
approaches for prescribing placebos, but
these results require replication in large-
scale clinical trials in different patient popu-
lations. To date, pure placebo prescriptions
have been addressed in only a few studies71
and are thus not discussed here. We also
exclude the issue of nocebo effects, although
they are highly relevant. However, BOX4
offers a brief overview on nocebo effects,
with suggested interventions on how to
preventthem.
Managing expectations. Expectations of
patients contribute to treatment efficacy;
for example, open-label applications
of analgesics substantially enhance the
analgesic effect compared to hidden
applications68 (FIG.1a). Even for invasive
medical interventions such as cardiac
surgery, the level of patients’ expectations
is a major predictor of their disability
3months after surgery74. In patients who
suffered a myocardial infarction, modifying
their treatment expectations resulted in
improved functionality, earlier return to
work and an improved quality of life75.
Moreover, a recent study showed that the
pretreatment recovery expectations of
patients with coronary artery disease
predicted long-term survival over 15years,
and that this effect persisted even after
controlling for medical, demographic and
other psychosocial variables76.
Similar results regarding the relevance of
patients’ expectations have been reported for
orthopaedic surgery77, deep brain stimula-
tion in Parkinson’s disease78 and treatments
with antidepressants79. Conversely, the verbal
induction of negative expectations can abol-
ish the effect of potent drugs such as opioids1.
Therefore optimizing patients’ expectations
before and during medical interventions may
contribute to improved clinical outcomes.
However, more research is needed to
investigate how expectations can be assessed
and modified in the context of complex
medical settings. Brief psychological inter-
ventions are promising tools that could be
used by medical personnel in daily routine
scenarios to optimize patients’ expectations.
As inadequate over-optimistic attitudes are
also not desired, the level of modification
of outcome expectations should be adjusted
based on the individual patient’s expectations
and disease severity. Patients with inadequate
treatment expectations (that is, overly nega-
tive or over-optimistic cognitions) should
undergo re-attribution training to develop
more positive and realistic expectations.
In addition, adequate expectations should be
consolidated by strengthening the cognitive
and emotional impact of positive treatment
results.
Influencing patients’ beliefs by the careful
use of language and the provision of appro-
priate information regarding the expected
effect of the drug should be considered as an
important feature of every pharmacological
treatment. It has been observed that in the
United States 50% of patients do not have an
adequate understanding of what their physi-
cian has told them following a visit80; this
highlights a need to improve this element of
the patient–physician interaction. However,
in addition to merely providing information,
manipulation of patients’ treatment expecta-
tions is considered to be most effective if
patients develop a ‘mental map’ that clearly
and adequately promotes an optimistic per-
spective regarding the treatment outcome.
As many variables can determine whether
intervening in patients’ expectations is
useful, psychological and physiological
predictors of successful interventions must
be defined and evaluated to ensure their
appropriate application in each patient.
Furthermore, non-medical variables such as
the physical environment and information
provided through leaflets and consent forms
should be optimized to support the develop-
ment of positive outcome expectations.
Conditioning responses. Classical
(Pavlovian) conditioning of drug responses
is another promising tool that could be
used to improve treatment outcomes.
However, drug intake is rarely analysed
from the perspective of associative learning
processes. Experimental studies have shown
that learned placebo responses can be used
to achieve analgesia and to modulate neuro-
endocrine and immunological functions
(BOX1; FIG.1b). Reframing long-term drug
intake as a learning process opens up a
new avenue for maximizing treatment
efficacy that could also decrease drug
dosages, reduce unwanted side effects
and lower treatmentcosts.
One strategy used to achieve this is
to use full-dose medication for a set period
of time (acquisition period) followed by
a maintenance or evocation period with
interspersed placebo treatment3,31 (FIG.3c).
Using this ‘partial reinforcement’ paradigm81,
drug efficacy can be maintained while drug
dosage is reduced. This has been dem-
onstrated in patients with psoriasis
(who were treated with corticosteroids)82
and in patients with attention-deficit
hyperactivity disorder (who were treated
with mixed amphetamine salts)83.
Although these examples confirm the
positive short-term effects that can be
achieved with the conditioning of drug
responses, the potential for reducing the
negative consequences of long-term drug
applications must be further analysed.
In particular, there is a need to understand
which physiological systems are particularly
susceptible for conditioning, and which
reinforcement schedules should be used
to achieve optimal effects and prevent
habituation or extinction of the learned
pharmacological responses84. For some dis-
ease conditions, partial reinforcement with
full-dose treatment may result in the most
stable effects, whereas for other disease
conditions, reconsolidation of the learned
PERSPECTIVES
198
|
MARCH 2013
|
VOLUME 12 www.nature.com/reviews/drugdisc
© 2013 Macmillan Publishers Limited. All rights reserved

pharmacological response may be achieved
through the use of subtherapeutic drug
doses85. Research is also needed to investi-
gate which psychological and physiological
trait and state variables predict facilitated
conditioning.
Optimizing physician–patient relationships.
The expectation- and conditioning-based
approaches should be combined with
approaches to optimize physician–patient
interactions. Characteristics of the
physician–patient relationship can predict
outcomes in various medical conditions86.
Experimental trials have shown that
switching from a brief, technical interaction
style to an empathic, emotionally warm
interaction style increased the efficacy of
a (placebo) intervention from 42% to 82%
in patients with irritable bowel syndrome4.
Further components of empathic behaviour
include expression of interest in the patient’s
health, validating the patient’s perceptions
and coping attempts, and addressing emo-
tional issues. It has been postulated that
these self-focusing and emotion-related
variables facilitate the activation of neural
mechanisms that are relevant to placebo
responses87. Moreover, describing an inter-
vention as valuable further increases the
efficacy of placebo mechanisms. This has
been demonstrated experimentally: the
effects of placebo pills that were presented as
being normally priced were compared with
pills that were described as inexpensive88;
the pill that was said to be normally priced
led to greater pain reduction. A valuable
description of treatment interventions is
thus relevant not only for self-administered
therapies but also for referrals and interven-
tions provided by other specialists. Indeed,
empirical data have underlined that the
treatment outcomes in clinical practice and
in clinical trials are substantially affected by
these interactional features.
All of these approaches demand a
rethinking of clinical trial designs. As well
as determining whether a drug shows
superiority over an inert placebo pill under
artificial scenarios, the optimal conditions
for achieving maximal drug efficacy with
a minimal risk of side effects and reduced
treatment costs in a real-life scenario should
be evaluated. Therefore, we strongly pro-
pose that after the principal efficacy of a
drug (against placebo) has been determined
using RCTs, subsequent phases of drug
development should explore the disease-
and drug-specific treatment context that
optimally enhances the overall treatment
outcome, which may provide a better esti-
mate of the drug’s therapeutic potential.
Finally, an improved understanding of the
neurobiological mechanisms of patients’
expectations and conditioning in specific
medical conditions, as well as the potential
interactions of the underlying biochemical
pathways of the placebo response with
pharmacological drugs, will open up a
new avenue of research that promises to
optimize pharmacotherapy.
Risk of abusing placebo mechanisms. The
ability to enhance and maximize placebo
responses by using one or more of the
measures discussed above carries the risk
that these strategies may also be used to
demonstrate the effectiveness of ineffective
and inadequate therapies in RCTs. The most
frequent example is a violation of the blinding
procedure caused by the development of
minor- or major-onset effects and/or side
effects that only occur in the drug group and
therefore trigger positive outcome expecta-
tions selectively in this group. Minor-onset
effects that are typically observed with drugs,
but not with placebos, can result in substan-
tial differences in efficacy45. Examples from
the use of complementary and alternative
medicines also indicate that the treatment
context can amplify placebo mechanisms to
Box 3 | Strategies to optimize placebo responses
The optimization of placebo responses via the management of patients’ expectations, the
application of learned responses and a positive physician–patient relationship can help to
ensure that the most efficacious therapeutic outcome is attained.
Expectations
•Select patients with inadequate negative outcome expectations and improve their expectations75:
patients’ expectations should be screened before treatment, and those patients with outcome
expectations that are more negative than the medical prognosis should undergo an expectation-
modification procedure
•Use open application of treatments and give positive instructions68: rituals (for example,
gestures and words that focus the patient’s attention on the drug and its potential benefits)
in the context of drug intake as well as positive instructions on how the drug will help the
patient can amplify treatment benefits
•Reduceexpectationsofnegativeevents(forexample,duringdrugwithdrawal)throughhidden
applications145 (see also BOX4 on nocebo effects): if negative effects are expected (for example,
withdrawal of drugs or negative onset effects), hidden application can reduce negative
expectations; options for hidden applications include covered injection of a drug to an infusion
or mixing drops of a drug into ‘fruit cocktails’ (these options allow the patient to be informed
about the change to a drug regimen without informing them about the details and time points
of changes)
•Promote social learning of the positive effects of the medication (for example, by observing
other patients showing benefits caused by the specific treatment)112: patients could talk to
other patients who received the same treatment successfully, or videos of other patients
could be shown
•Reducepatients’anxiety146
Conditioning
•Useplacebo-controlleddrugreduction(PCDR)ifapplicable:PCDRprovidestheoptionof
starting treatment with repeated full doses to establish associative learning processes and
replacing medication by placebos later on; despite the later reduction in drug doses,
treatmen t effects can be maintained3
•Use effective pretreatments if possible147: some medical interventions allow or require
pretreatments; these should be designed to be highly effective, and the patient should
receive feedback about the positive effects of the pretreatment
•If conditioning is desired, use salient treatment stimuli (for example, colour or odour) and
a constant context with substantial contextual features (for example, fixed time of day or
same room)148
•Avoid extinction processes during long-term treatments148: attention focusing and repeated
motivation strategies can hinder extinction processes; if no conditioning processes are desired
(see above), then changes of situational cues can also hinder extinction processes
•Enhance the physician–patient relationship: switch from a brief technical interaction style
to an empathic and emotionally warm interaction style4 (see main text for more details)
•Value medical interventions88: a short description of how the medical intervention will be
helpful, how positive past experiences with this intervention were, and so on, is warranted
PERSPECTIVES
NATURE REV IEWS
|
DRUG DISCOVERY VOLUME 12
|
MARCH 2013
|
199
© 2013 Macmillan Publishers Limited. All rights reserved

a degree that exceeds the regular, evidence-
based effects of medication. Accordingly,
placebo acupuncture has demonstrated
higher efficacy in the treatment of migraine
than standard medical care using well-
evaluated drugs89. This highlights the fact
that a potential misuse of placebo mecha-
nisms can only be identified in relation to an
adequate comparator condition. Although
a full comparability of placebo mechanisms
between drug and placebo groups must
be required for the approval of new drugs,
optimized placebo mechanisms are desirable
from a patient’s perspective. However, only
an improved understanding and systematic
analyses of placebo mechanisms can help to
detect the risk of misusing the placebo effect
to obscure inefficient therapies.
Personalizing placebo responses
In both experimental and clinical scenarios,
the placebo response varies tremendously
among individuals. The individual
placebo response can range from no effect
(non-responders) to profound changes in
symptoms or disease severity (responders)90–92.
Given that placebo responses fundamentally
contribute to the overall treatment outcome,
knowledge about the magnitude of their
occurrence in individual patients could
guide therapeutic decisions regarding the
selection and dose of adrug.
This is best illustrated in an open/hidden
study of local anaesthesia in patients with
Alzheimer’s disease93. Patients who displayed
a reduced neural connectivity of the pre-
frontal lobes, which are crucially involved
in the initiation of placebo analgesia, expe-
rienced less additional pain relief from an
open application of lidocaine compared
to a hidden application of the anaesthetic.
This study demonstrates that the individual
contribution of placebo mechanisms to
therapeutic outcome is crucially determined
by the neurobiological make-up of an indi-
vidual and underscores the necessity to
adjust drug treatment approaches depending
on the individual patient’s predisposition for
placebo responses.
Psychological, neuroendocrine and genetic
predictors of placebo responsiveness.
In order to systematically utilize placebo
responses, it is essential to predict the
capacity of an individual to develop pla-
cebo responses in a context-, physiological
system- and disease-specific manner. Thus,
identifying predictor variables for placebo
responders or non-responders would be of
great value for the effective use of placebo
responses in the optimization of treatment
Box 4 | Nocebo effects: a brief summary
Nocebo effects are defined as the development of negative effects that are attributed to a
medication, albeit the drug itself does not cause the provocation of these symptoms. Although the
typical example of nocebo effects refers to the development of unwanted symptoms, the
reduction of efficacy caused by contextual factors is also included under the definition of nocebo
mechanisms. These effects have to be distinguished from mere attribution effects (for example,
symptoms that existed before treatment was started, but are re-attributed to the medication later
on). A thorough summary of these aspects can be found in REF.149.
Nocebo effects in clinical trials
In clinical trials, nocebo effects are best illustrated as the development of side effects in the
placebo groups. The development of side effects after placebo intake has been reported for
several medical conditions: for example, in clinical trials of patients with hypertension150, high
cholesterol (using statins)151, depression152 and cancer153. Many patients in the placebo groups of
these clinical trials discontinued pill intake explicitly because of symptoms that were attributed
to the medication. Nocebo effects are estimated to account for 72% of drop-outs in drug groups
of fibromyalgia trials154. Notably, the side-effect profiles of placebo groups seem to partially
reflect the expected side effects of the drug. This has been shown for migraine treatments155,
antidepressants152, treatments for multiple sclerosis156 and other drugs. For example, in trials
investigating migraine treatments, the placebo groups of non-steroidal anti-inflammatory drug
(NSAID) trials reported more gastrointestinal symptoms, whereas the placebo groups of
anticonvulsant trials reported more neurological symptoms155.
Mechanisms of nocebo responses
Observationsfromclinicaltrialsalreadyindicatetheroleofpatients’expectationsinthe
development of nocebo effects, but stronger evidence from experimental designs confirms the
causal role of expectations as a trigger of nocebo effects. If patients were intensively informed
about the potential side effects of a specific drug, they reported more symptoms than patients
who were given limited information about potential side effects145,157. Neurophysiological studies
have indicated that the expectation of symptoms stimulates similar areas of the brain as the
experience of symptoms, thus preparing a facilitated and amplified perception of side
effects158,159. These results are in agreement with the experimental study of Bingel etal.1,
showing that a negative instruction can even block the expected analgesic effects of opioids.
Moreover, it seems that conditioning and associative learning can also have a role in the
development of nocebo effects, although there is weaker experimental evidence for their
involvement in nocebo effects compared to their role in the development of placebo responses.
A frequently cited clinical example for the conditioning of side effects is the development of
anticipatory nausea in patients undergoing chemotherapy. Further studies have indicated that
vicarious learning can also have a substantial role through the observation of other patients
experiencing symptoms160. The role of learning in the development of nocebo effects is further
outlined in the study by Colloca and colleagues112. Finally, there is also some evidence that the
physician–patient relationship influences the development of side effects in medical treatments121,
and has the potential to both protect and amplify side effects.
Predictors of nocebo effects
In clinical trials as well as in clinical practice, it is crucial to anticipate who will develop side effects.
In line with the mechanisms described above, the expectation of patients to develop side effects
is a strong predictor of patient-reported side effects a few weeks or months later161. Moreover,
patients who reported symptoms in the past seemed to be more prone to developing new
symptoms in future treatments162. The same study has confirmed that worries about the
development of symptoms and general anxieties had an additional role as potential predictors
of the development of future symptoms.
Avoiding nocebo effects
A major challenge for the future of drug discovery and development is the design of prevention
programmes to avoid nocebo effects and the associated discontinuation of medication intake.
Considering the fact that most patients with indicated long-term treatments discontinue drug
intake163, nocebo-induced symptoms have a tremendous impact on clinical practice. Prevention
programmes could include information on nocebo mechanisms leading to symptoms that are not
caused by the drug itself, a s well as coping strategies for benign nocebo symptoms, thereby
reducing the risk of drug discontinuation.
Health-care specialists should be aware that every interaction with the patient has the potential
to induce either helpful or unhelpful expectations. Consequently, the information that patients
receive before medical treatments is of particular relevance. Unfortunately, information leaflets
of medications primarily induce negative side-effect expectations164, whereas positive outcome
expectations and positive coping expectations for side effects are omitted. Future formats of
medication leaflets should try to integrate the latest findings on nocebo mechanisms with legal
and ethical regulations for the content of medication leaflets.
PERSPECTIVES
200
|
MARCH 2013
|
VOLUME 12 www.nature.com/reviews/drugdisc
© 2013 Macmillan Publishers Limited. All rights reserved

outcomes in clinical practice, as well as for
the design and interpretation of clinical
trials. Much effort is currently being dedi-
cated to the identification of psychological,
neuroendocrine or genetic variables that
moderate an individual’s placebo respon-
siveness in different physiological systems
and diseases.
There is evidence supporting the
putative relevance of several psychosocial
variables in mediating placebo responses.
These variables include: trait and state
anxiety94–96; patients’ expectations97,98,
dispositional optimism99–101 and hypnotic
suggestibility in placebo analgesia102; health
locus of control103; cultural variations in
anxiety disorders104; and coping abilities
in irritable bowel syndrome105.
Evidence supporting the role of genetic
traits in moderating placebo responses is best
documented in anxiety disorders and depres-
sion. Serotonin-related gene polymorphisms
have been found to influence individual pla-
cebo responses in social anxiety and in pain,
both at the behavioural and neural level22,106.
In addition, genetic polymorphisms modu-
lating monoaminergic tone have been related
to the degree of placebo responsiveness in
major depressive disorder107. More recently,
anxiety state and plasma noradrenaline levels
were identified as predictors of a learned
placebo response in levels of interleukin-2
release90. The latest evidence from diffu-
sion tensor imaging in healthy volunteers
supports the notion that individual brain
anatomy also predicts an individual’s capacity
to develop placebo analgesia92.
Harnessing individual placebo responses.
A better knowledge of the variables predicting
placebo responses in the various clinical
conditions will enable the targeting of
those patients who are particularly prone to
developing placebo responses while adjust-
ing the dose and regimen of active phar-
macological treatments for those patients
who show a reduced capacity for placebo
responses. Given that some of the variables
moderating placebo responses are modifi-
able (for example, anxiety, expectations or
learning experiences), the consideration
and potential therapeutic modulation of
these variables might become a useful tool
for improving the placebo responsiveness
and thereby overall treatment outcome
of individual patients. This may involve
both psychological and pharmacological
approaches.
The search for predictor variables in
placebo responses is still at a very early
stage. Many of the available studies used
only small sample sizes, which might
explain the often conflicting results108,109.
In addition, little is known regarding the
stability of potential placebo predictors
over time as well as the influence of
symptoms and different underlying
con ditions on these placebo predictors.
Future large-scale experimental and
clinical studies must therefore address
the important issue of identifying bio-
logical and psychological predictors
of placebo responsiveness.
However, the data already demonstrate
that personalized medicine, in the truest
sense of the word, is not limited to genetic
approaches110,111. Indeed, an individual’s
prior experience with a certain treatment
is a key determinant of their placebo
response and thereby modulates overall
treatment outcome. This is evident not
only in experimental studies evaluating
placebo treatments with a conditioning
component112 but also in clinical trials3,47,113.
Consequently, a systematic appreciation
of a patient’s prior treatment history,
various personal characteristics, vulner-
ability to adverse events and other
disease-specific information should guide
therapeutic decisions regarding not only
the selection and dose of a drug treatment
but also the use of specific psychological
interventions.
Glossary
Active placebo
A substance or treatment that mimics the side effects
of the active compound under investigation and is thus,
by definition, not an inert substance. In clinical trials,
active placebos are administered to avoid un-blinding
owing to different side-effect profiles of drugs and
placebo treatments.
Assay sensitivity
The ability of a clinical trial to differentiate between an
effective treatment (for example, a drug) and a less
effective or ineffective treatment (for example, placebo).
CER trial
A comparative effectiveness research (CER) trial
is performed to analyse the efficacy of a novel
pharmacological agent or treatment in comparison
with standard treatments or approved drugs. Patients
are therefore randomly allocated to receive the
treatment under investigation or one or more
standard treatments.
Declaration of Helsinki
A statement, developed by the World Medical
Association (WMA), of ethical principles for medical
research involving human participants, identifiable
human material and data.
Health locus of control
The extent to which individuals believe that they
can control events that affect their personal health.
Open/hidden study
An experimental approach undertaken to separate the
effects of the psychosocial context (placebo) from the
pharmacodynamic effects of a drug under investigation.
The pharmacological agent is administered either in an
open condition (by a physician in a visible way) or in
a hidden condition, in which the patient is unaware
of the timing of the administration of the medication
(for example, the drug is administered using computer-
controlled infusion).
Open-label
A method of application in which both the patients
(or participants) and the investigators know which
pharmacological agent or treatment is being
administered. This design contrasts the single- or
double-blind study designs.
Patient-reported outcomes
(PROs). A method of measuring treatment efficacy via the
states of symptom severity and health from the patient’s
perspective, instead of physician’s reports or biomarkers
of clinical outcome. PROs are typically analysed via
questionnaires or interviews, providing insight into how
patients perceive the impact of a treatment on their health
and quality of life.
Placebo
Latin term for “I shall please”. Used to indicate sham
treat ments or inert substances such as sugar pills or
saline infusions.
Placebo effects
Defined as any improvements in a symptom or physiological
condition of individuals following a placebo treatment. There
are different mechanisms underlying this phenomenon,
including spontaneous remission, regression to the mean,
natural course of a disease, biases and placebo responses.
Placebo responses
The outcomes caused by a placebo manipulation. The
placebo response reflects the neurobiological and psycho-
physiological response of an individual to an inert substance
or sham treatment and is mediated by various factors that
make up the treatment context. Importantly, placebo
responses are not restricted to placebo treatments and
can also modulate the outcome of any active treatment.
Randomized double-blind placebo-controlled trials
(RCTs). The most commonly used clinical trial design
for testing the efficacy of a treatment within a patient
population. Patien ts are randomly allocated to a treatment
or placebo group. Patients and investigators are blinded to
group allocation. The design aims to control for confounding
factors such as suggestion, imagination and biases for the
patient and investigator, as well as spontaneous fluctuation
of diseases and symptoms.
Regression to the mean
A statistical phenomenon; individuals tend to have extreme
values in symptom severity or physiological parameters
when enrolled into a clinical trial. These values tend to be
lower and closer to the average at subsequent assessments,
because they are more likely to change in the direction of
the mean score, instead of developing even more extreme
scores. This phenomenon in part explains the improvement
observed in placebo groups in clinical trials.
PERSPECTIVES
NATURE REV IEWS
|
DRUG DISCOVERY VOLUME 12
|
MARCH 2013
|
201
© 2013 Macmillan Publishers Limited. All rights reserved

Conclusions and future directions
Research into the neurobiology underlying
placebo responses is still at a very early stage,
which limits their applicability for drug
development and testing in addition to their
use in clinical settings. Although placebo
responses are not restricted to PROs114 and
have also been demonstrated for various
physiological and pathophysiological func-
tions115 (BOX1), a major challenge relates to
the differential susceptibility of medical
conditions and physiological systems to
respond to placebo.
Not all diseases and symptoms are sub-
ject to potent placebo responses, and some
conditions are not suitable for assigning
placebo groups (for example, antibiotics
in bacterial infections, or chemotherapy in
stage II or stage III cancer). Therefore, our
knowledge of the involvement and potential
of placebo mechanisms in these conditions
is — and will probably remain — limited.
Moreover, although long-term placebo
effects of interventions have been shown in
some conditions116–119, the sustainability of
these effects over several months to years,
which is required in the treatment of chronic
conditions such as pain and depression, must
be further explored in larger samples across
different diseases.
A comprehensive discussion of the
mechanisms underlying the nocebo
effect16,31,120,121 and its prevention in RCTs
and in the clinical setting is urgently needed
but beyond the scope of this article (see
BOX4 for a brief overview). In addition,
some of the open questions remaining
at this stage include whether or not the
approaches for maximizing the placebo
response can also be used to identify and
develop better drugs, to choose the optimal
dosage of a therapy and to find the best
treatment for a specific clinical condition.
A final question concerns the reasons why
the evolution of mankind has shaped these
placebo mechanisms122. Answers to this
question may provide valuable insights
into the sociocultural aspects of healing
and self-healing processes.
Ultimately, the utilization of placebo
mechanisms by minimizing, maximizing
and personalizing placebo responses will
require context-, patient- and disease-
specific decisions based on neurobiological,
psychological and methodological evidence.
Recent experimental and clinical evidence
demonstrates in detail how patients’ expec-
tations, the quality of physician–patient
communication and associative learning
processes all affect the efficacy of a drug
or treatment.
This detailed knowledge can form the
basis for the systematic utilization of these
strategies in clinical practice. A consideration
of placebo mechanisms could help to better
detect drug-specific effects and optimize
drug treatment regimens, drug efficacy, drug
adherence and context-specific characteristics
of the general medical setting. So, harnessing
placebo mechanisms in an empirically based
and systematically applied manner could
fundamentally improve the effectiveness of
drug development as well as the effective-
ness of treatment strategies. Therefore, these
approaches may ultimately reduce health-care
costs and improve patientcare.
Paul Enck is at the Department of Internal
Medicine VI, University Hospital Tübingen,
72076 Tübingen, Germany.
Ulrike Bingel is at the Department of Neurology,
University Medical Center Hamburg-Eppendorf,
20246 Hamburg, Germany.
Manfred Schedlowski is at the Institute of Medical
Psychology and Behavioural Immunobiology,
University Hospital Essen, University of
Duisburg-Essen, 45122 Essen, Germany.
Winfried Rief is at the Division of Clinical Psychology,
University of Marburg, 35032 Marburg, Germany.
All authors contributed equally to this work.
Correspondence to P.E.
e-mail: paul.enck@uni-tuebingen.de
doi:10.1038/nrd3923
1. Bingel,U. etal. The effect of treatment expectation
on drug efficacy: imaging the analgesic benefit of the
opioid remifentanil. Sci. Transl. Med. 3, 70ra14
(2011).
2. Colloca,L. & Benedetti,F. Placebos and painkillers:
is mind as real as matter? Nature Rev. Neurosci. 6,
545–552 (2005).
3. Doering,B.K. & Rief,W. Utilizing placebo
mechanisms for dose reduction in pharmacotherapy.
Trends Pharmacol. Sci. 33, 165–172 (2012).
4. Kaptchuk,T.J. etal. Components of placebo effect:
randomised controlled trial in patients with irritable
bowel syndrome. BMJ 336, 999–1003 (2008).
5. [No authors listed.] An Audience With… Ted Kaptchuk.
Nature Rev. Drug Discov. 7, 554 (2008).
6. Benedetti,F. Placebo Effects: Understanding the
mechanisms in health and disease (Oxford Univ.
Press, 2008).
7. Fields,H. State-dependent opioid control of pain.
Nature Rev. Neurosci. 5, 565–575 (2004).
8. Millan,M.J. Descending control of pain.
Prog. Neurobiol. 66, 355–474 (2002).
9. Eippert,F., Finsterbusch,J., Bingel,U. & Buchel,C.
Direct evidence for spinal cord involvement in placebo
analgesia. Science 326, 404 (2009).
10. Amanzio,M. & Benedetti,F. Neuropharmacological
dissection of placebo analgesia: expectation-activated
opioid systems versus conditioning-activated specific
subsystems. J.Neurosci. 19, 484–494 (1999).
11. Benedetti,F., Amanzio,M., Rosato,R. & Blanchard,C.
Nonopioid placebo analgesia is mediated by CB1
cannabinoid receptors. Nature Med. 17, 1228–1230
(2011).
12. Eippert,F. etal. Activation of the opioidergic
descending pain control system underlies placebo
analgesia. Neuron 63, 533–543 (2009).
13. Wager,T.D., Scott,D.J. & Zubieta,J.K.
Placebo effects on human μ-opioid activity during
pain. Proc. Natl Acad. Sci. USA 10 4, 11056–11061
(2007).
14. Guo,J.Y., Wang,J.Y. & Luo,F. Dissection of placebo
analgesia in mice: the conditions for activation of
opioid and non-opioid systems. J.Psychopharmacol.
24, 1561–1567 (2010).
15. Levine,J.D., Gordon,N.C. & Fields,H.L.
The mechanism of placebo analgesia. Lancet 2,
654–657 (1978).
16. Tracey,I. Getting the pain you expect: mechanisms of
placebo, nocebo and reappraisal effects in humans.
Nature Med. 16, 1277–1283 (2010).
17. de la Fuente-Fernandez,R. etal. Expectation and
dopamine release: mechanism of the placebo effect
in Parkinson’s disease. Science 293, 1164–1166
(2001).
18. Lidstone,S.C. etal. Effects of expectation on
placebo-induced dopamine release in Parkinson
disease. Arch. Gen. Psychiatry 67, 857–865
(2010).
19. Leuchter,A.F., Cook,I.A., Witte,E.A., Morgan,M.
& Abrams,M. Changes in brain function of
depressed subjects during treatment with placebo.
Am. J.Psychiatry 159, 122–129 (2002).
20. Mayberg,H.S. etal. The functional neuroanatomy of
the placebo effect. Am. J.Psychiatry 159, 728–737
(2002).
21. Petrovic,P. etal. Placebo in emotional processing-
induced expectations of anxiety relief activate
a generalized modulatory network. Neuron 46,
957–969 (2005).
22. Furmark,T. etal. A link between serotonin-related
gene polymorphisms, amygdala activity, and placebo-
induced relief from social anxiety. J.Neurosci. 28,
13066–13074 (2008).
23. Ader,R. & Cohen,N. Behaviorally conditioned
immunosuppression. Psychosom. Med. 37, 333–340
(1975).
24. Goebel,M.U. etal. Behavioral conditioning of
immunosuppression is possible in humans. FASEB J.
16, 1869–1873 (2002).
25. Goebel,M.U., Meykadeh,N., Kou,W.,
Schedlowski,M. & Hengge,U.R. Behavioral
conditioning of antihistamine effects in patients
with allergic rhinitis. Psychother. Psychosom. 77,
227–234 (2008).
26. Schedlowski,M. & Pacheco-Lopez,G. The learned
immune response: Pavlov and beyond. Brain Behav.
Immun. 24, 176–185 (2010).
27. Wirth,T. etal. Repeated recall of learned
immunosuppression: evidence from rats and men.
Brain Behav. Immun. 25, 1444–1451 (2011).
28. Benedetti,F. etal. Conscious expectation and
unconscious conditioning in analgesic, motor, and
hormonal placebo/nocebo responses. J.Neurosci.
23, 4315–4323 (2003).
29. Sabbioni,M.E. etal. Classically conditioned changes
in plasma cortisol levels induced by dexamethasone
in healthy men. FASEB J. 11, 1291–1296 (1997).
30. Benedetti,F. The Patient’s Brain: The neuroscience
behind the doctor–patient relationship (Oxford Univ.
Press, 2011).
31. Rief,W., Bingel,U., Schedlowski,M. & Enck,P.
Mechanisms involved in placebo and nocebo
responses and implications for drug trials.
Clin. Pharmacol. Ther. 90, 722–726 (2011).
32. Hill,A.B. Suspended judgement. Memories of the
British streptomycin trial in tuberculosis. The first
randomized clinical trial. Control. Clin. Trials 11,
77–79 (1990).
33. Diener,H.C., Dowson,A.J., Ferrari,M., Nappi,G. &
Tfelt-Hansen,P. Unbalanced randomization influences
placebo response: scientific versus ethical issues
around the use of placebo in migraine trials.
Cephalalgia 19, 699–700 (1999).
34. Papakostas,G.I. & Fava,M. Does the probability of
receiving placebo influence clinical trial outcome?
A meta-regression of double-blind, randomized clinical
trials in MDD. Eur. Neuropsychopharmacol. 19,
34–40 (2009).
35. Mallinckrodt,C.H., Zhang,L., Prucka,W.R. &
Millen,B.A. Signal detection and placebo response
in schizophrenia: parallels with depression.
Psychopharmacol. Bull. 43, 53–72 (2010).
36. Sinyor,M. et al. Does inclusion of a placebo arm
influence response to active antidepressant treatment
in randomized controlled trials? Results from pooled
and meta-analyses. J.Clin. Psychiatry 71, 270–279
(2010).
37. Nakamura,Y. etal. Investigating dose-dependent
effects of placebo analgesia: a psychophysiological
approach. Pain 153, 227–237 (2012).
38. Rutherford,B.R., Sneed,J.R. & Roose,S.P.
Does study design influence outcome? The effects
of placebo control and treatment duration in
antidepressant trials. Psychother. Psychosom. 78,
172–181 (2009).
PERSPECTIVES
202
|
MARCH 2013
|
VOLUME 12 www.nature.com/reviews/drugdisc
© 2013 Macmillan Publishers Limited. All rights reserved

39. Quigley,E.M. etal. Randomised clinical trials:
linaclotide phase 3 studies in IBS-C — a prespecified
further analysis based on European Medicines
Agency-specified endpoints. Aliment. Pharmacol. Ther.
37, 49–61 (2013).
40. Staskin,D.R., Michel,M.C., Sun,F., Guan,Z. &
Morrow,J.D. The effect of elective sham dose
escalation on the placebo response during an
antimuscarinic trial for overactive bladder symptoms.
J.Urol. 187, 1721–1726 (2012).
41. Mallinckrodt,C. etal. A case study comparing a
randomized withdrawal trial and a double-blind long-
term trial for assessing the long-term efficacy of an
antidepressant. Pharm. Stat. 6, 9–22 (2007).
42. Suchman,A.L. & Ader,R. Classic conditioning and
placebo effects in crossover studies. Clin. Pharmacol.
Ther. 52, 372–377 (1992).
43. Hrobjartsson,A. & Boutron,I. Blinding in randomized
clinical trials: imposed impartiality. Clin. Pharmacol.
Ther. 90, 732–736 (2011).
44. Moncrieff,J., Wessely,S. & Hardy,R. Active placebos
versus antidepressants for depression. Cochrane
Database Syst. Rev. 2004, CD003012 (2004).
45. Rief,W. & Glombiewski,J.A. The hidden effects
of blinded, placebo controlled randomized trials:
an experimental investigation. Pain 153, 2473–2477
(2012).
46. Toth,C. etal. An enriched-enrolment, randomized
withdrawal, flexible-dose, double-blind, placebo-
controlled, parallel assignment efficacy study of
nabilone as adjuvant in the treatment of diabetic
peripheral neuropathic pain. Pain 153, 2073–2082
(2012).
47. de la Fuente-Fernandez,R. The powerful
pre-treatment effect: placebo responses in restless
legs syndrome trials. Eur. J.Neurol. 19, 1305–1310
(2012).
48. Tedeschini,E., Fava,M., Goodness,T.M. &
Papakostas,G.I. Relationship between probability
of receiving placebo and probability of prematurely
discontinuing treatment in double-blind,
randomized clinical trials for MDD: a meta-analysis.
Eur. Neuropsychopharmacol. 20, 562–567
(2010).
49. Zelen,M. A new design for randomized clinical trials.
N.Engl. J.Med. 300, 1242–1245 (1979).
50. Kirsch,I. Are drug and placebo effects in depression
additive? Biol. Psychiatry 47, 733–735 (2000).
51. Muthen,B. & Brown,H.C. Estimating drug effects in
the presence of placebo response: causal inference
using growth mixture modeling. Stat. Med. 28,
3363–3385 (2009).
52. Dworkin,R.H. etal. Considerations for improving
assay sensitivity in chronic pain clinical trials:
IMMPACT recommendations. Pain 153, 1148–1158
(2012).
53. US Food and Drug Administration (FDA). Guidance
for Industry: E10 Choice of control group and
related issues in clinical trials. FDA website
[online], http://www.fda.gov/downloads/Drugs/
GuidanceComplianceRegulatoryInformation/
Guidances/UCM073139.pdf (2001).
54. Mulla,S.M., Scott,I.A., Jackevicius,C.A., You,J.J.
& Guyatt,G.H. How to use a noninferiority trial:
users’ guides to the medical literature. JAMA 308,
2605–2611 (2012).
55. Woods,S.W., Gueorguieva,R.V., Baker,C.B. &
Makuch,R.W. Control group bias in randomized
atypical antipsychotic medication trials for
schizophrenia. Arch. Gen. Psychiatry 62, 961–970
(2005).
56. Estellat,C. & Ravaud,P. Lack of head-to-head trials
and fair control arms: randomized controlled trials
of biologic treatment for rheumatoid arthritis.
Arch. Intern. Med. 172, 237–244 (2012).
57. Huitfeldt,B., Hummel, J. & the European Federation
of Statisticians in the Pharmaceutical Industry (EFSPI).
The draft FDA guideline on non-inferiority clinical
trials: a critical review from European pharmaceutical
industry statisticians. Pharm. Stat. 10, 414–419
(2011).
58. Leon,A.C. Comparative effectiveness clinical trials
in psychiatry: superiority, noninferiority, and the
role of active comparators. J.Clin. Psychiatry 72,
1344–1349 (2011).
59. Rief,W. etal. Meta-analysis of the placebo response
in antidepressant trials. J.Affect. Disord. 118, 1–8
(2009).
60. Hrobjartsson,A. & Gotzsche,P.C. Placebo
interventions for all clinical conditions. Cochrane
Database Syst. Rev. 2010, CD003974 (2010).
61. Kirsch,I. etal. Initial severity and antidepressant
benefits: a meta-analysis of data submitted to the
Food and Drug Administration. PLoS Med. 5, e45
(2008).
62. Nijs,J. etal. Recruitment bias in chronic pain
research: whiplash as a model. Clin. Rheumatol. 30,
1481–1489 (2011).
63. Kobak,K.A., Kane,J.M., Thase,M.E. &
Nierenberg,A.A. Why do clinical trials fail?
The problem of measurement error in clinical trials:
time to test new paradigms? J.Clin. Psychopharmacol.
27, 1–5 (2007).
64. Pressman,A., Avins,A.L., Neuhaus,J., Ackerson,L.
& Rudd,P. Adherence to placebo and mortality in
the Beta Blocker Evaluation of Survival Trial (BEST).
Contemp. Clin. Trials 33, 492–498 (2012).
65. Boehmer,J. & Yong,P. How well does blinding work
in randomized controlled trials?: a counterpoint.
Clin. Pharmacol. Ther. 85, 463–465 (2009).
66. Cepeda,M.S., Berlin,J.A., Gao,C.Y., Wiegand,F. &
Wada,D.R. Placebo response changes depending on
the neuropathic pain syndrome: results of a systematic
review and meta-analysis. Pain Med. 13, 575–595
(2012).
67. Enck,P., Klosterhalfen,S. & Zipfel,S. Novel study
designs to investigate the placebo response.
BMC Med. Res. Methodol. 11, 90 (2011).
68. Colloca,L., Lopiano,L., Lanotte,M. & Benedetti,F.
Overt versus covert treatment for pain, anxiety, and
Parkinson’s disease. Lancet Neurol. 3, 679–684
(2004).
69. Enck,P., Grundy,D. & Klosterhalfen,S. A novel
placebo-controlled clinical study design without
ethical concerns — the free choice paradigm.
Med. Hypotheses 79, 880–882 (2012).
70. O’Neil,C.C. & Miller,F.G. When scientists deceive:
applying the federal regulations. J.Law Med. Eth.
37, 344–350 (2009).
71. Fässler,M., Meissner,K., Schneider,A. & Linde,K.
Frequency and circumstances of placebo use in clinical
practice — a systematic review of empirical studies.
BMC Med. 8, 15 (2010).
72. Miller,F.G. & Colloca,L. The placebo phenomenon
and medical ethics: rethinking the relationship
between informed consent and risk–benefit
assessment. Theor. Med. Bioeth. 32, 29–43 (2011).
73. Kaptchuk,T.J. etal. Placebos without deception:
a randomized controlled trial in irritable bowel
syndrome. PLoS ONE 5, e15591 (2010).
74. Juergens,M.C., Seekatz,B., Moosdorf,R.G.,
Petrie,K.J. & Rief,W. Illness beliefs before cardiac
surgery predict disability, quality of life, and
depression 3months later. J.Psychosom. Res.
68, 553–560 (2010).
75. Petrie,K.J., Cameron,L.D., Ellis,C.J., Buick,D.
& Weinman,J. Changing illness perceptions after
myocardial infarction: an early intervention
randomized controlled trial. Psychosom. Med.
64, 580–586 (2002).
76. Barefoot,J.C. etal. Recovery expectations and
long-term prognosis of patients with coronary heart
disease. Arch. Intern. Med. 171, 929–935 (2011).
77. Judge,A. etal. Pre-operative expectation predicts
12-month post-operative outcome among patients
undergoing primary total hip replacement in European
orthopaedic centres. Osteoarthritis Cartilage 19,
659–667 (2011).
78. Mercado,R. etal. Expectation and the placebo effect
in Parkinson’s disease patients with subthalamic
nucleus deep brain stimulation. Mo v. Disord. 21,
1457–1461 (2006).
79. Chen,J.A. etal. Association between patient beliefs
regarding assigned treatment and clinical response:
reanalysis of data from the Hypericum Depression
Tri al Study Group. J.Clin. Psychiat. 72, 1669–1676
(2011).
80. Bodenheimer,T. The future of primary care:
transforming practice. N.Engl. J.Med. 359,
2086–2089 (2008).
81. Acosta,J.I., Thiel,K.J., Sanabria,F., Browning,J.R.
& Neisewander,J.L. Effect of schedule of
reinforcement on cue-elicited reinstatement of
cocaine-seeking behavior. Behav. Pharmacol. 19,
129–136 (2008).
82. Ader,R. etal. Conditioned pharmacotherapeutic
effects: a preliminary study. Psychosom. Med. 72,
192–197 (2010).
83. Sandler,A.D., Glesne,C.E. & Bodfish,J.W.
Conditioned placebo dose reduction: a new
treatment in attention-deficit hyperactivity disorder?
J.Dev. Behav. Pediatr. 31, 369–375 (2010).
84. Ader, R. in The Placebo Effect: An Interdisciplinary
Exploration (ed. Harrington, A.) 138–165
(Harvard Univ. Press, 1997).
85. Hadamitzky, M., Engler, H. & Schedlowski, M.
Learned immunosuppression: extinction, renewal,
and the challenge of reconsolidation. J. Neuroimmune
Pharmacol. 13 Jul 2012 (doi:10.1007/s11481-012-
9388-6).
86. Koudriavtseva,T., Onesti,E., Pestalozza,I.F.,
Sperduti,I. & Jandolo,B. The importance of
physician–patient relationship for improvement of
adherence to long-term therapy: data of survey in a
cohort of multiple sclerosis patients with mild and
moderate disability. Neurol. Sci. 33, 575–584
(2012).
87. Schneider,R. & Kuhl,J. Placebo forte: ways to maximize
unspecific treatment effects. Med. Hypotheses 78,
744–751 (2012).
88. Waber,R.L., Shiv,B., Carmon,Z. & Ariely,D.
Commercial features of placebo and therapeutic
efficacy. JAMA 299, 1016–1017 (2008).
89. Linde,K. etal. Acupuncture for patients with
migraine: a randomized controlled trial. JAMA 293,
2118–2125 (2005).
90. Ober,K. etal. Plasma noradrenaline and state
anxiety levels predict placebo response in learned
immunosuppression. Clin. Pharmacol. Ther. 91 ,
220–226 (2012).
91. Wager,T.D., Atlas,L.Y., Leotti,L.A. & Rilling,J.K.
Predicting individual differences in placebo analgesia:
contributions of brain activity during anticipation
and pain experience. J.Neurosci. 31 , 439–452
(2011).
92. Stein,N., Sprenger,C., Scholz,J., Wiech,K. &
Bingel,U. White matter integrity of the descending
pain modulatory system is associated with
inter-individual differences in placebo analgesia.
Pain 153, 2210–2217 (2012).
93. Benedetti,F. etal. Loss of expectation-related
mechanisms in Alzheimer’s disease makes analgesic
therapies less effective. Pain 121, 133–144 (2006).
94. Baeken,C., Vanderhasselt,M.A. & De Raedt,R.
Baseline ‘state anxiety’ influences HPA-axis sensitivity
to one sham-controlled HF-rTMS session applied to
the right dorsolateral prefrontal cortex.
Psychoneuroendocrinology 36, 60–67 (2011).
95. Lyby,P.S., Aslaksen,P.M. & Flaten,M.A. Variability
in placebo analgesia and the role of fear of pain —
an ERP study. Pain 152, 2405–2412 (2011).
96. Flaten,M.A., Aslaksen,P.M., Lyby,P.S. &
Bjorkedal,E. The relation of emotions to placebo
responses. Phil. Trans. R.Soc. B 366, 1818–1827
(2011).
97. Linde,K. etal. The impact of patient expectations on
outcomes in four randomized controlled trials of
acupuncture in patients with chronic pain. Pain 128,
264–271 (2007).
98. Flaten,M.A., Aslaksen,P.M., Finset,A., Simonsen,T.
& Johansen,O. Cognitive and emotional factors in
placebo analgesia. J.Psychosom. Res. 61, 81–89
(2006).
99. Geers,A.L., Helfer,S.G., Kosbab,K., Weiland,P.E. &
Landry,S.J. Reconsidering the role of personality in
placebo effects: dispositional optimism, situational
expectations, and the placebo response.
J.Psychosom. Res. 58, 121–127 (2005).
100. Morton,D.L., Watson,A., El-Deredy,W. &
Jones,A.K. Reproducibility of placebo analgesia:
effect of dispositional optimism. Pain 146, 194–198
(2009).
101. Geers,A.L., Wellman,J.A., Fowler,S.L., Helfer,S.G.
& France,C.R. Dispositional optimism predicts
placebo analgesia. J.Pain. 11, 1165–1171 (2010).
102. De Pascalis,V., Chiaradia,C. & Carotenuto,E.
The contribution of suggestibility and expectation to
placebo analgesia phenomenon in an experimental
setting. Pain 96, 393–402 (2002).
103. Reynaert,C., Janne,P., Vause,M., Zdanowicz,N.
& Lejeune,D. Clinical trials of antidepressants:
the hidden face: where locus of control appears
to play a key role in depression outcome.
Psychopharmacology 119 , 449–454 (1995).
104. Moerman,D.E. Cultural variations in the placebo
effect: ulcers, anxiety, and blood pressure.
Med. Anthropol. Quarterly 14, 51–72 (2000).
105. Schneider,A. etal. Acupuncture treatment in irritable
bowel syndrome. Gut 55, 649–654 (2006).
106. Lindstedt,F. etal. Conditioned pain modulation is
associated with common polymorphisms in the
serotonin transporter gene. PLoS ONE 6, e18252
(2011).
PERSPECTIVES
NATURE REV IEWS
|
DRUG DISCOVERY VOLUME 12
|
MARCH 2013
|
203
© 2013 Macmillan Publishers Limited. All rights reserved

107. Leuchter,A.F., McCracken,J.T., Hunter,A.M.,
Cook,I.A. & Alpert,J.E. Monoamine oxidase A
and catechol-O-methyltransferase functional
polymorphisms and the placebo response in major
depressive disorder. J.Clin. Psychopharmacol. 29,
372–377 (2009).
108. Kaptchuk, T. J. et al. Do “placebo responders” exist?
Contemp. Clin. Trials. 29, 587–595 (2008).
109. Whalley,B., Hyland,M.E. & Kirsch,I. Consistency of
the placebo effect. J.Psychosom. Res. 64, 537–541
(2008).
110 . Caponigro,G. & Sellers,W.R. Advances in the
preclinical testing of cancer therapeutic hypotheses.
Nature Rev. Drug Discov. 10, 179–187 (2011).
111. Beckman,R.A., Clark,J. & Chen,C. Integrating
predictive biomarkers and classifiers into oncology
clinical development programmes. Nature Rev.
Drug Discov. 10, 735–748 (2011).
112. Colloca,L., Sigaudo,M. & Benedetti,F. The role of
learning in nocebo and placebo effects. Pain 136,
211–218 (2008).
113. Iovieno,N. & Papakostas,G.I. Does the presence
of an open-label antidepressant treatment period
influence study outcome in clinical trials examining
augmentation/combination strategies in treatment
partial responders/nonresponders with major
depressive disorder? J.Clin. Psychiatry 73, 676–683
(2012).
114. Wechsler,M.E. etal. Active albuterol or placebo,
sham acupuncture, or no intervention in asthma.
N.Engl. J.Med. 365, 119–126 (2011).
115. Finniss,D.G., Kaptchuk,T.J., Miller,F. & Benedetti,F.
Biological, clinical, and ethical advances of placebo
effects. Lancet 375, 686–695 (2010).
116. McRae,C. et al. Effects of perceived treatment on
quality of life and medical outcomes in a double-blind
placebo surgery trial. Arch. Gen. Psychiatry 61,
412–420 (2004).
117. Moseley,J.B. et al. A controlled trial of arthroscopic
surgery for osteoarthritis of the knee. N.Engl. J.Med.
347, 81–88 (2002).
118. Quessy,S.N. & Rowbotham,M.C. Placebo response
in neuropathic pain trials. Pain 138, 479–483
(2008).
119. Khan,A. Redding, N. & Brown, W. A. The persistence
of the placebo response in antidepressant clinical
trials. J.Psychiatr. Res. 42, 791–796 (2008).
120. Colloca,L. & Benedetti,F. Nocebo hyperalgesia: how
anxiety is turned into pain. Curr. Opin. Anaesthesiol.
20, 435–439 (2007).
121. Colloca,L. & Finniss,D. Nocebo effects, patient–
clinician communication, and therapeutic outcomes.
JAMA 307, 567–568 (2012).
122. Sullivan,R., Behncke,I. & Purushotham,A.
Why do we love medicines so much? An evolutionary
perspective on the human love of pills, potions and
placebo. EMBO Rep. 11, 572–578 (2010).
123. Volkow, N. D. etal. Expectation enhances the
regional brain metabolic and the reinforcing effects
of stimulants in cocaine abusers. J.Neurosci. 23,
11461–11468 (2003).
124. Benedetti,F., Amanzio,M., Baldi,S., Casadio,C. &
Maggi,G. Inducing placebo respiratory depressant
responses in humans via opioid receptors.
Eur. J.Neurosci. 11 , 625–631 (1999).
125. Isenberg,S.A., Lehrer,P.M. & Hochron,S. The effects
of suggestion and emotional arousal on pulmonary
function in asthma — a review and a hypothesis
regarding vagal mediation. Psychosom. Med. 54,
192–216 (1992).
126. Pollo,A., Vighetti,S., Rainero,I. & Benedetti,F.
Placebo analgesia and the heart. Pain 102, 125–133
(2003).
127. Ronel,J. et al. Effects of verbal suggestion on
coronary arteries: results of a randomized controlled
experimental investigation during coronary
angiography. Am. Heart J. 162, 507–511 (2011).
128. Amigo,I., Cuesta,V., Fernandez,A. & Gonzalez,A.
The effect of verbal instructions on blood pressure
measurement. J.Hypertens. 11 , 293–296 (1993).
129. Lanotte,M. etal. Expectation enhances autonomic
responses to stimulation of the human subthalamic
limbic region. Brain Behav. Immun. 19, 500–509
(2005).
130. Elsenbruch,S. etal. Neural mechanisms mediating
the effects of expectation in visceral placebo
analgesia: an fMRI study in healthy placebo
responders and non-responders. Pain 153, 382–390
(2012).
131. Price,D.D., Craggs,J., Verne,G.N., Perlstein,W.M.
& Robinson,M.E. Placebo analgesia is accompanied
by large reductions in pain-related brain activity in
irritable bowel syndrome patients. Pain 127, 63–72
(2007).
132. Meissner,K. Effects of placebo interventions on
gastric motility and general autonomic activity.
J.Psychosom. Res. 66, 391–398 (2009).
133. Scott,D.J. etal. Individual differences in reward
responding explain placebo-induced expectations
and effects. Neuron 55, 325–336 (2007).
134. Scott,D.J. etal. Placebo and nocebo effects are
defined by opposite opioid and dopaminergic
responses. Arch. Gen. Psychiatry 65, 220–231
(2008).
135. Hall,K.T. etal. Catechol-O-methyltransferase
Val158Met polymorphism predicts placebo effect in
irritable bowel syndrome. PLoS ONE 7, e48135
(2012).
136. Benedetti,F. etal. Electrophysiological properties of
thalamic, subthalamic and nigral neurons during the
anti-parkinsonian placebo response. J.Physiol. 587,
3869–3883 (2009).
137. Potkin,S. etal. Placebo response trajectories
in short-term and long-term antipsychotic trials in
schizophrenia. Schizophr. Res. 132, 108–113
(2011).
138. Lee,S., Walker,J.R., Jakul,L. & Sexton,K. Does
elimination of placebo responders in a placebo run-in
increase the treatment effect in randomized clinical
trials? A meta-analytic evaluation. Depress. Anxiety
19, 10–19 (2004).
139. Madsen,L.G. & Bytzer,P. Single subject trials
as a research instrument in gastrointestinal
pharmacology. Aliment. Pharmacol. Ther. 16,
189–196 (2002).
140. Ivanova,A. & Tamura,R.N. A two-way enriched
clinical trial design: combining advantages of
placebo lead-in and randomized withdrawal.
Stat. Methods Med. Res. 4Dec 2011
(doi:10.1177/0962280211431023).
141. King,M. et al. Conceptual framework and systematic
review of the effects of participants’ and professionals’
preferences in randomised controlled trials. Health
Technol. Assess. 9, 1–186 (2005).
142. De Allegri,M. etal. Step-wedge cluster-randomised
community-based trials: an application to the study
of the impact of community health insurance.
Health Res. Policy Syst. 6, 10 (2008).
143. Weijer,C. etal. The Ottawa statement on the ethical
design and conduct of cluster randomized trials.
PLoS Med. 9, e1001346 (2012).
144. Kaptchuk,T.J. etal. Do “placebo responders” exist?
Contemp. Clin. Trials. 29, 587–595 (2008).
145. Mondaini,N. etal. Finasteride 5 mg and sexual side
effects: how many of these are related to a nocebo
phenomenon? J.Sex. Med. 4, 1708–1712 (2007).
146. Petrie,K.J. etal. Effect of providing information
about normal test results on patients’ reassurance:
randomised controlled trial. BMJ 334, 352–353
(2007).
147. Colloca,L. & Benedetti,F. How prior experience
shapes placebo analgesia. Pain 124, 126–133
(2006).
148. Eikelboom,R. & Stewart,J. Conditioning of drug-
induced physiological responses. Psych. Rev. 89,
507–528 (1982).
149. Barsky,A.J., Saintfort,R., Rogers,M.P. & Borus,J.F.
Nonspecific medication side effects and the nocebo
phenomenon. JAMA 287, 622–626 (2002).
150. Preston,R.A., Materson,B.J., Reda,D.J. &
Williams,D.W. Placebo-associated blood pressure
response and adverse effects in the treatment of
hypertension. Arch. Intern. Med. 160, 1449–1454
(2000).
151. Rief,W., Avorn,J. & Barsky,A.J. Medication-
attributed adverse effects in placebo groups.
Implications for assessment of adverse effects.
Arch. Intern. Med. 166, 155–160 (2006).
152. Rief,W. etal. Differences in adverse effect reporting
in placebo groups in SSRI and tricyclic antidepressant
trials. A systematic review and meta-analysis.
Drug Safety 32, 1041–1056 (2009).
153. de la Cruz,M., Hui,D., Parsons,H.A. & Bruera,E.
Placebo and nocebo effects in randomized double-
blind clinical trials of agents for the therapy for fatigue
in patients with advanced cancer. Cancer 11 6 ,
766–774 (2010).
154. Häuser,W., Bartram,C., Bartram-Wunn,E. & Tolle,T.
Adverse events attributable to nocebo in randomized
controlled drug trials in fibromyalgia syndrome and
painful diabetic peripheral neuropathy: systematic
review. Clin. J.Pain 28, 437–451 (2012).
155. Amanzio,M., Corazzini,L.L., Vase,L. & Benedetti,F.
A systematic review of adverse events in placebo
groups of anti-migraine clinical trials. Pain 146,
261–269 (2009).
156. Papadopoulos,D. & Mitsikostas,D.D. Nocebo effects
in multiple sclerosis trials: a meta-analysis. Mult. Scler.
16, 816–828 (2010).
157. Wise,R.A. etal. Randomized trial of the effect
of drug presentation on asthma outcomes:
the American Lung Association Asthma Clinical
Research Centers. J.Allergy Clin. Immunol. 124,
436–444.e8 (2009).
158. Koyama,T., McHaffie,J.G., Laurienti,P.J. &
Coghill,R.C. The subjective experience of pain:
where expectations become reality. Proc. Natl Acad.
Sci. USA 102, 12950–12955 (2005).
159. Keltner,J.R. etal. Isolating the modulatory effect
of expectation on pain transmission: a functional
magnetic resonance imaging study. J.Neurosci. 26,
4437–4443 (2006).
160. Craig, K. D. & Prkachin, K. M. Social modeling
influences on sensory decision theory and
psychophysiological indexes of pain. J. Pers.
Soc. Psychol. 36, 805–815 (1978).
161. Nestoriuc,Y., Orav,E.J., Liang,M.H., Horne,R. &
Barsky,A.J. Prediction of nonspecific side effects
in rheumatoid arthritis patients by beliefs about
medicines. Arthritis Care Res. 62, 791–799
(2010).
162. Petrie,K.J. et al. Worries about modernity predict
symptom complaints after environmental pesticide
spraying. Psychosom. Med. 67, 778–782
(2005).
163. Jackevicius,C.A., Mamdani,M. & Tu,J.V.
Adherence with statin therapy in elderly patients
with and without acute coronary syndromes. JAMA
288, 462–467 (2002).
164. Colagiuri,B., McGuinness,K., Boakes,R.A. &
Butow,P.N. Warning about side effects can increase
their occurrence: an experimental model using
placebo treatment for sleep difficulty.
J.Psychopharmacol. 26, 1540–1547 (2012).
Acknowledgements
All authors are participants of a collaborative research group
dedicated to studying placebo and nocebo mechanisms across
different physiological systems in health and disease. This work
was supported by grants from the German Research
Foundation (DFG) for the Research Unit FOR 1328 (BI 89/2-1;
EN 50/30-1; RI 574/21-1; RI 574-22-1; SCHE341/17-1), the
Volkswagen Foundation Germany (P.E.: I/83 805; M.S.: I/83
806) and the German Federal Ministry of Education and
Research (U.B.: 01GQ0808).
Competing interests statement
The authors declare no competing financial interests.
FURTHER INFORMATION
Paul Enck’s homepage: http://www.medizin.uni‑tuebingen.
de/pages/psymed/1/index2.php?cat=sup_&page=enck&supe
rpage=mitarbeiter
Ulrike Bingel’s homepage: http://www.uke.de/institute/
systemische‑neurowissenschaften/index_53792.php
Manfred Schedlowski’s homepage: http://www.uk‑essen.de/
medizinische‑psychologie/en/institute/staff/research‑
groups/93.html
Winfried Rief’s homepage:
http://www.uni‑marburg.de/fb04/ag‑klin/mitarbeiter/wrief
ALL LINKS ARE ACTIVE IN THE ONLINE PDF
PERSPECTIVES
204
|
MARCH 2013
|
VOLUME 12 www.nature.com/reviews/drugdisc
© 2013 Macmillan Publishers Limited. All rights reserved