Expectation and Dopamine Release: Mechanism of the Placebo Effect in Parkinson's Disease

Abstract

The power of placebos has long been recognized for improving numerous medical conditions such as Parkinson's disease (PD).
Little is known, however, about the mechanism underlying the placebo effect. Using the ability of endogenous dopamine to compete
for [11C]raclopride binding as measured by positron emission tomography, we provide in vivo evidence for substantial release of endogenous
dopamine in the striatum of PD patients in response to placebo. Our findings indicate that the placebo effect in PD is powerful
and is mediated through activation of the damaged nigrostriatal dopamine system.

Full text

Expectation and Dopamine
Release: Mechanism of the
Placebo Effect in Parkinson’s
Disease
Rau´l de la Fuente-Ferna´ndez,
1
Thomas J. Ruth,
2
Vesna Sossi,
2
Michael Schulzer,
1
Donald B. Calne,
1
A. Jon Stoessl
1
*
The power of placebos has long been recognized for improving numerous
medical conditions such as Parkinson’s disease (PD). Little is known, however,
about the mechanism underlying the placebo effect. Using the ability of en-
dogenous dopamine to compete for [
11
C]raclopride binding as measured by
positron emission tomography, we provide in vivo evidence for substantial
release of endogenous dopamine in the striatum of PD patients in response to
placebo. Our findings indicate that the placebo effect in PD is powerful and is
mediated through activation of the damaged nigrostriatal dopamine system.
The simple act of receiving any treatment
(active or not) may, in itself, be efficacious
because of expectation of benefit (1). This is
the placebo effect—a potential confounder in
assessing the efficacy of any therapeutic in-
tervention (2, 3). Placebo-controlled studies
were designed precisely to control for such an
effect (4). It has been assumed that the pla-
cebo response is not mediated directly
through any physical or chemical effect of
treatment (5). In Parkinson’s disease (PD),
the placebo effect can be prominent (6, 7 ).
We asked whether the placebo effect in
PD is produced by activation of the pathway
primarily damaged by degeneration [i.e., the
nigrostriatal dopaminergic system (8, 9)]. To
answer this question, we took advantage of
the ability of positron emission tomography
(PET) to estimate pharmacologically or be-
haviorally induced dopamine release based
on the competition between endogenous do-
pamine and [
11
C]raclopride (RAC) for bind-
ing to dopamine D
2
/D
3
receptors (10–14).
We hypothesized that if the placebo effect is
mediated through the activation of the path-
way relevant to the disorder under study, we
should be able to detect placebo-induced re-
lease of endogenous dopamine in PD.
We examined the striatal RAC binding
potential of six patients with PD (group 1,
placebo group) under two conditions (15):
Condition 1, a placebo-controlled, blinded
study in which the patients did not know
when they were receiving placebo or active
drug (apomorphine) (16)—all patients re-
ceived both placebo and active drug; and
condition 2, an open study in the same pa-
tients without placebo.
We found a significant decrease in striatal
RAC binding potential [17% for the caudate
nucleus (range, 8 to 25%); 19% for the puta-
men (range, 8 to 28%); P ⬍ 0.005 for both,
two-tailed paired t test] when the patients
received placebo compared with open base-
line observations (Table 1). This placebo-
induced change in RAC binding potential
was present in each patient and in each stri-
atal subregion, although it was greatest in the
posterolateral part of the putamen (Table 1).
The magnitude of the placebo response was
comparable to that of therapeutic doses of
levodopa (17 ), or apomorphine (see below)
(18). There were no differences in the striatal
RAC binding potential between this group of
patients when studied without placebo and a
second group of patients matched by age and
severity of parkinsonism studied exclusively
in an open fashion (group 2, open group) (15)
(Fig. 1).
These observations indicate that there is
placebo-induced release of endogenous dopa-
mine in the striatum (19). The estimated re-
lease of dopamine was greater in patients
who perceived placebo benefit than in those
who did not (20). This suggests a “dose-
dependent” relation between the release of
endogenous dopamine and the magnitude of
the placebo effect.
We next asked whether there might be an
interaction between the effects of the placebo
and the active drug (21). The placebo re-
sponse could synergistically enhance the ben-
efit of an active drug, in which case double-
blind, placebo-controlled studies would over-
estimate the active drug effect. Alternatively,
the placebo effect could mask (or decrease)
the specific effect of an active drug, which
would lead to the opposite conclusion in the
interpretation of a placebo-controlled study.
After adjusting for differences in “base-
line” RAC binding potential, we found no
significant differences in the response to apo-
morphine between the open group and the
placebo group (combining patients who per-
ceived a placebo effect and those who did
not) (22). However, the degree of apomor-
phine-induced change in RAC binding poten-
tial tended to be lower in patients who per-
ceived a placebo effect compared with those
who did not and with patients studied in an
open fashion (Fig. 2). We explored whether
this observation could reflect a floor effect in
the placebo group (i.e., whether the technique
was insensitive for further reductions in RAC
binding), but this did not appear to be the case
(Fig. 3) (23). We conclude that the placebo
response does not potentiate the effect of an
active drug. Indeed, our results suggest that in
some patients, most of the benefit obtained
from an active drug might derive from a
placebo effect.
The dopaminergic system is involved in
the regulation of several cognitive, behavior-
al, and sensorimotor functions, and particu-
larly in reward mechanisms (24–28). Howev-
er, our experiments did not involve a direct
reward. We conclude that dopamine release
in the nigrostriatal system is linked to expec-
tation of a reward—in this case, the anticipa-
tion of therapeutic benefit (29, 30). All pa-
tients were familiar with the effect of an
active drug (levodopa), and such previous
experience may have enhanced their expec-
tation. We found that the level of expectation
may determine experience (20) —patients
who perceived a placebo effect had higher
release of dopamine than those who did not.
Our observations indicate that the placebo
effect in PD is mediated by an increase in the
1
Neurodegenerative Disorders Centre,
2
TRIUMF, Uni-
versity of British Columbia, Vancouver, BC, Canada
V6T 2B5.
*To whom correspondence should be addressed. E-
mail: jstoessl@interchange.ubc.ca
Table 1. Striatal RAC binding potential (mean ⫾ SD) of PD patients (group 1) scanned at open baseline
and after receiving placebo (n ⫽ 6).
Site Open baseline Placebo
Mean percent change
(range)
Head of caudate 1.964 ⫾ 0.221 1.638 ⫾ 0.230 16.6 (8.4–25.1)
Putamen
Rostral 2.398 ⫾ 0.342 1.976 ⫾ 0.321 17.6 (5.3–26.3)
Intermediate 2.621 ⫾ 0.438 2.142 ⫾ 0.389 18.2 (7.4–27.0)
Caudal 2.095 ⫾ 0.269 1.646 ⫾ 0.261 21.2 (8.8–32.6)
R EPORTS
10 AUGUST 2001 VOL 293 SCIENCE www.sciencemag.org1164

synaptic levels of dopamine in the striatum.
Expectation-related dopamine release might
be a common phenomenon in any medical
condition susceptible to the placebo effect.
PD patients receiving an active drug in the
context of a placebo-controlled study benefit
from the active drug being tested as well as
from the placebo effect. By contrast, in the
usual clinical practice setting, active drugs
may be devoid of placebo effect. We found
no evidence to suggest that the placebo effect
synergistically augments the action of active
drugs (in fact, a trend for the opposite was
observed), so positive conclusions derived
from placebo-controlled studies are not im-
pugned by our findings.
References and Notes
1. D. G. Altman, Practical Statistics for Medical Research
(Chapman & Hall, London, 1991), pp. 450 – 451.
2. H. K. Beecher, JAMA ( J. Am. Med. Assoc.) 159, 1602
(1955).
3. E. Ernst, K. L. Resch, Br. Med. J. 311, 551 (1995).
4. T. J. Kaptchuk, Lancet 351, 1722 (1998).
5. L. D. Fisher, G. van Belle, Biostatistics: A Methodology
for the Health Sciences (Wiley, New York, 1993), p.
22.
6. N. Shetty et al., Clin. Neuropharmacol. 22, 207
(1999).
7. C. G. Goetz, S. Leurgans, R. Raman, G. T. Stebbins,
Neurology 54, 710 (2000).
8. J. M. Fearnley, A. J. Lees, Brain 114, 2283 (1991).
9. S. J. Kish, K. S. Shannak, O. Hornykiewicz, N. Engl.
J. Med. 318, 876 (1988).
10. P. Seeman, H. C. Guan, H.B. Niznik, Synapse 3,96
(1989).
11. N. D. Volkow et al., Synapse 16, 255 (1994).
12. M. J. Koepp et al., Nature 393, 266 (1998).
13. A. J. Stoessl, T. J. Ruth, NeuroScience News 2,53
(1999).
14. M. Laruelle, J. Cereb. Blood Flow Metab. 20, 423
(2000).
15. All PET scans were performed in three-dimensional
(3D) mode using an ECAT 953B/31 tomograph. We
obtained 16 sequential frames over 60 minutes,
starting at the time of injection of 5 mCi of [
11
C]ra-
clopride (mean ⫾ SEM specific activity ⫽ 4692 ⫾
349 Ci/mmol at ligand injection). A time-integrated
image with 31 planes, each 3.37 mm thick, was made
from the emission data (from 30 to 60 minutes) for
each subject. The five axial planes in which the
striatum was best visualized were summed. On this
time- and spatially summed image, one circular re-
gion of interest (ROI) of 61.2 mm
2
was positioned on
the head of each caudate nucleus (Caud), and three
circular ROIs of the same size were placed without
overlap along the axis of each putamen (from rostral
to caudal putamen: P1, P2, and P3); ROI position was
adjusted to maximize the average radioactivity. The
ROIs were replicated on the spatially summed image
of each time frame. The background activity was
averaged from a single elliptical ROI (2107 mm
2
)
drawn over the cerebellum on the summed image of
two contiguous axial planes. The binding potential
(BP ⫽ f NS䡠B
max
/K
d
, where f NS is the free fraction of
tracer) was determined using a tissue input graphical
approach [J. Logan et al., J. Cereb. Blood Flow Metab.
16, 834 (1996)]. Further details of the PET scan
protocol are reported elsewhere (17). We studied two
groups of PD patients, of six patients each, under two
different protocols as described below. Both groups
were matched by age and severity of parkinsonism as
measured by the Modified Columbia Scale (MCS)
[R. C. Duvoisin, in Monoamines noyaux gris centraux
et syndrome de Parkinson, J. de Ajuriaguerra, G. Gau-
thier, Eds. (Georg and Cie SA, Geneva, 1971), pp.
313–325]. Clinical details can be found on Science
Online at www.sciencemag.org/cgi/content/full/293/
5532/1164/DC1. After being pretreated with domp-
eridone for 48 hours to prevent side effects, all
patients underwent three consecutive RAC PET scans
on the same day according to the following protocol:
(i) either baseline or placebo scan 12 to 18 hours
after withdrawal of medications; (ii) after subcutane-
ous injection of 0.03 mg of apomorphine per kilo-
gram of body weight; and (iii) after subcutaneous
injection of 0.06 mg/kg of apomorphine. The treat-
ment order was maintained constant for all patients.
Group 1 (the placebo group) was studied in a blind
fashion—patients did not know when they were
receiving placebo (subcutaneous injection of saline)
or apomorphine (all patients received all three treat-
ments). This group also received a fourth injection,
consisting of 0.12 mg/kg of apomorphine on the
same day, to explore the possibility of a floor effect
Fig. 1. Placebo-induced changes in RAC binding potential in the striatum ipsilateral (A) and
contralateral (B) to the more affected body side of patients with PD. The ROIs are on the head of
the caudate nucleus (Caud) and on the putamen, from rostral to caudal, P1, P2, P3 (15).
Comparisons were made between the group of patients studied in an open fashion (group 2, open
group; open bars) and the group of patients studied both with (solid bars) and without (hatched
bars) placebo intervention (group 1, placebo group). Within-subject placebo-induced changes in
RAC binding potential tended to be greater in the striatum contralateral to the more affected body
side (20%) than in the ipsilateral striatum (17%). The placebo group and the open group did not
differ in their baseline placebo-free RAC binding potential values [for the caudate nucleus, 1.96 ⫾
0.22 (SD) versus 2.07 ⫾ 0.40, respectively; two-tailed t test, t ⫽ – 0.55 (df ⫽ 10), P ⫽ 0.59; for
the putamen, 2.37 ⫾ 0.34 versus 2.42 ⫾ 0.42, t ⫽ –0.20 (df ⫽ 10), P ⫽ 0.84]. Error bars, SEM.
Fig. 2. Apomorphine-induced changes in RAC binding potential in the caudate nucleus (A) and
putamen (B) before (APO_0) and after (APO_1 ⫽ 0.03 mg/kg, and APO_2 ⫽ 0.06 mg/kg)
subcutaneous injection of apomorphine. Patients studied in an open fashion (open bars) had higher
RAC binding potential values than those included in the placebo group [independently of whether
they did not (hatched bars) or did (solid bars) perceive a placebo effect]. The decline in RAC binding
potential induced by an incremental dose of apomorphine tended to be less pronounced in patients
who perceived a placebo effect as compared with those who did not, and with patients studied in
an open fashion: interaction term (group ⫻ apomorphine dose) evaluated by repeated measures
ANCOVA, F ⫽ 4.66 (df ⫽ 2, 9), P ⫽ 0.041 for the caudate nucleus; F ⫽ 3.40 (df ⫽ 2, 9), P ⫽ 0.079
for the putamen. Error bars, SEM.
Fig. 3. Linear regression plots for patients without (n ⫽ 3; open symbols, thin lines) and with (n ⫽
3; solid symbols, thick lines) perceived placebo effect: (A) caudate and (B) putamen RAC binding
potential values against apomorphine dose (APO_dose). The four slopes were significantly different
from zero (P ⬍ 0.01), but they did not differ significantly between patients with and without
perceived placebo effect (for the caudate nucleus, –3.2 versus –5.1, respectively, P ⫽ 0.28; for the
putamen, –3.8 versus –6.5, P ⫽ 0.15).
R EPORTS
www.sciencemag.org SCIENCE VOL 293 10 AUGUST 2001 1165

(see below). Group 2 (open group) was studied in an
open fashion for comparison purposes (e.g., to investi-
gate the effect of novelty on dopamine release). Here,
recipients were scanned under all three conditions but
knew explicitly if they were receiving no medication or
which dose of apomorphine they were receiving at any
given time. The advantages of this design are threefold:
(i) It minimizes potential carry-over effects from the
active drug (apomorphine) (17). (ii) It helps maintain
the level of expectation throughout the study, which is
crucial to this experiment. For example, the occurrence
of apomorphine-induced side effects could “unblind”
the study. (iii) It maximizes the tolerability of the pro-
cedure. In total, there was a 2.5-hour interval between
scans (1-hour scan plus 1.5-hour break), sufficient to
allow for decay of radioactivity, as well as for dopamine
receptor recovery after apomorphine injection (16, 17).
An additional open baseline scan was performed on
group 1 patients on a different day to obtain placebo-
free baseline values (interval between both sets of
scans, 1 to 4 months). All patients had been contacted
1 month before the scans, and details of the protocol in
which they were included were explained; they were
reminded of these details 3 days before the scans. We
avoided anticipation bias (e.g., patients’ knowledge of
the fact that the placebo effect can determine measur-
able changes in dopamine release might alter the re-
sults) by keeping the patients and the clinical staff
unaware of the purpose of the study. In all cases, care
was taken to optimize patient positioning in the scan-
ner. Motion within and between scans was minimized
by the use of a molded thermoplastic mask. All subjects
gave written informed consent. The study was approved
by the U.B.C. ethics committee.
16. S. T. Gancher, W. R. Woodward, B. Boucher, J. G. Nutt,
Ann. Neurol. 26, 232 (1989).
17. R. de la Fuente-Ferna´ndez et al., Ann. Neurol. 49, 298
(2001).
18. The placebo-induced change in striatal RAC binding
potential is much higher than the reported within
subject scan-rescan variation expected to occur with-
in subject for scan and rescan (mean, 5%) [N. D.
Volkow et al., J. Nucl. Med. 34, 609 (1993)]. The
administration of 0.03 and 0.06 mg/kg of apomor-
phine led to a 14% and 26% decrease, respectively, in
putamen RAC binding potential in the open group
(see Fig. 2).
19. The increasing rostrocaudal gradient of the placebo
effect ( Table 1) eliminates the possibility that the
results could be due to down-regulation of presyn-
aptic D
2
/D
3
receptors. Partial volume effects cannot
explain the gradient in BP
open baseline
–BP
placebo
re-
ported here. Other considerations supporting our in-
terpretation of the results can be found elsewhere
(17).
20. Because the clinical benefit from apomorphine lasts
typically about 1 hour (16), which is the duration of
RAC PET scans, no objective measurements on
changes in the clinical status after placebo or apo-
morphine injection were made (motor activity might
confound the assessment of changes in striatal RAC
binding potential). However, only half of the patients
reported placebo-induced clinical improvement
(comparable in magnitude to the clinical benefit ob-
tained when they were on their regular treatment
with levodopa). Although the number of subjects is
small, those patients who perceived the placebo ef-
fect (n ⫽ 3) had higher changes in RAC binding
potential than those who did not (n ⫽ 3) [for the
caudate nucleus, 22% versus 12%; for the putamen,
24% versus 14%; P ⬍ 0.05 and P ⬍ 0.01, respective-
ly, by analysis of covariance (ANCOVA)] (Fig. 2).
21. J. Kleijnen, A. J. M. de Craen, J. van Everdingen, L. Krol,
Lancet 344, 1347 (1994).
22. Repeated measures ANCOVA gave the following re-
sults: For the caudate nucleus, between-group differ-
ences, F ⫽ 0.03 (df ⫽ 1, 9), P ⫽ 0.87; interaction
term (group ⫻ apomorphine dose), F ⫽ 0.09 (df ⫽ 1,
10), P ⫽ 0.77. For the putamen, between-group
differences, F ⫽ 0.71 (df ⫽ 1, 9), P ⫽ 0.42; interaction
term, F ⫽ 1.81 (df ⫽ 1, 10), P ⫽ 0.21. The power for
the interaction terms may not have been sufficient.
23. An apomorphine dose of 0.12 mg/kg led to a further
decrease in RAC binding potential in the placebo
group (Fig. 3). The total reduction in RAC binding
potential (compared with placebo-free baseline val-
ues) was 42% in the caudate nucleus (range, 19 to
59%) and 46% in the putamen (range, 24 to 60%).
24. R. A. Wise, Trends Neurosci. 3, 91 (1980).
25. H. C. Fibiger, A. G. Phillips, in Handbook of Physiology:
The Nervous System, vol. 4, Intrinsic Systems of the
Brain, V. B. Mountcastle, F. E. Bloom, S. R. Geiger, Eds.
(American Physiological Society, Bethesda, MD,
1986), pp. 647– 675.
26. T. W. Robbins, B. J. Everitt, Semin. Neurosci. 4, 119
(1992).
27. W. Schultz, J. Neurophysiol. 80, 1 (1998).
28. S. Ikemoto, J. Panksepp, Brain Res. Brain Res. Rev. 31,
6 (1999).
29. I. Kirsch, Ed., How Expectancies Shape Experience
(American Psychological Association, Washington,
DC, 1999).
30. J. M. Fish, Science 284, 914 (1999).
31. We thank J. McKenzie, S. Jivan, J. Leighton, T. Dobko,
and members of the UBC-TRIUMF PET team for
assistance with the scans. This study was funded by
the Canadian Institutes of Health Research, the Brit-
ish Columbia Health Research Foundation (R.F.-F. and
V.S.), the Pacific Parkinson’s Research Institute ( Van-
couver, B.C., Canada) (R.F.-F.), and a TRIUMF Life
Science grant. A.J.S. is supported by the Canada Re-
search Chairs program.
22 March 2001; accepted 5 June 2001
R EPORTS
10 AUGUST 2001 VOL 293 SCIENCE www.sciencemag.org1166