JOURNAL OF THE ROYAL SOCIETY OF MEDICINE Volume 92 October 1999

Placebos and placebo effects in medicine:

historical overview
Anton J M de Craen PhD Ted J Kaptchuk OMD1 Jan G P Tijssen PhD J Kleijnen MD PhD2

J R Soc Med 1999;92:51 1-515

In modern medical research placebos constitute an PLACEBOS IN CLINICAL PRACTICE BEFORE

WORLD WAR 11
important methodological tool. Placebos are given to
controls in a clinical trial with the intention of mimicking Until the first half of the 20th century the use of placebos
some experimental intervention. Although the most seems to have been widespread in medicine. In 1807
frequently used placebo is the 'sugar pill' in drug trials, Thomas Jefferson, recording what he called the pious fraud,
placebos can be and have been used for all kinds of observed that 'one of the most successful physicians I have
interventions, ranging from placebo ultrasound in the ever known has assured me that he used more bread pills,
treatment of pressure ulcers and placebo surgery in the treat- drops of colored water, and powders of hickory ashes, than
ment of osteoarthritis to sham traction in the treatment of of all other medicines put together'4. About a hundred
low back pain and placebo oestrogen implants in the years later, Richard Cabot, of Harvard Medical School,
prevention of menopause symptoms. described how he 'was brought up, as I suppose every
physician is, to use placebo, bread pills, water subcu-
taneously, and other devices'5.
ETYMOLOGY Only a few physicians considered the bread pill a threat
The word placebo (Latin, 'I shall please') was first used in to the integrity of medicine, and most ethical codes
the 14th centuryl. In that period, it referred to hired endorsed 'necessary deception'. A 'polychromatic assort-
mourners at funerals. These individuals often began their ment of sugar pills' was routinely quaffed by patients6.
wailings with Placebo Domino in regione vivorum, the ninth However, placebos were thought to bring only comfort to
verse of psalm cxiv, which in the Latin Vulgate translation the patient, with no impact on pathophysiology7. The value
means 'I shall please the Lord in the land of the living'. of placebo was thought inversely related to the intelligence
Here, the word placebo carries the connotation of of the patient; the use of a medical ritual was more effective
depreciation and substitution, because professional mourn- and necessary for 'unintelligent, neurotic, or inadequate
ers were often stand-ins for members of the family of the patients'8.
deceased. Around the same time, in the late 1 300s,
Geoffrey Chaucer in his Canterbury Tales (Merchant's Tale) PLACEBOS IN CLINICAL RESEARCH
depicts a man named Placebo. Like the hired mourners, the
Until 1950, most therapies were judged to be efficacious on
man is associated with wicked behaviour and is portrayed as
the basis of pathophysiological rationales provided by
a sycophant.
authoritative experts rather than by documented observa-
The first documented medical use of the word placebo
tions and comparative research9. The bulk of clinical
dates from the late 18th century1. In the 1785 New Medical
knowledge was based on noncomparative research, though
Dictionary, placebo is described as 'a commonplace method
there are some exceptions10O
or medicine'2. In 181 1, the revised Quincy's Lexicon-Medicum
In 1801, John Haygarth reported the results of what
defines placebo as 'an epithet given to any medicine adapted
more to please than to benefit the patient'3.
may have been the first placebo-controlled trial11. A
common remedy for many diseases at that time was to apply
metallic rods, known as Perkins tractors, to the body.
These rods were supposed to relieve symptoms through the
Department of Clinical Epidemiology and Biostatistics, Academic Medical
Centre-University of Amsterdam, 1100 DE Amsterdam, The Netherlands;
electromagnetic influence of the metal. Haygarth treated
'Center for Altemative Medicine Research, Beth Israel Deaconess Medical five patients with imitation tractors made of wood and
Center, Harvard Medical School, Boston, MA 02215, USA; 2Centre for Reviews found that four gained relief. He used the metal tractors on
and Dissemination, University of York, York YO1 5DD, UK
the same five patients the following day and obtained
Correspondence to: Dr Anton JM de Craen, Department of Clinical identical results: four of five subjects reported relief. It is
Epidemiology, Leiden University Medical Centre, PO Box 9600, 2300 RC Leiden,
The Netherlands clear that Haygarth had the notion of a placebo effect when
E-mail: AJMdeCraen@epi.azl.nl he stated that 'an important lesson in physic is here to be
 JOURNAL OF THE ROYAL SOCIETY OF MEDICINE Volume 92 October 1 999

learnt, the wonderful and powerful influence of the passions In 1938, the word placebo was first applied in reference
of the mind upon the state and disorder of the body. This is to the treatment given to concurrent controls in a trial16. In
too often overlooked in the cure of diseases'. He further previous years, uncontrolled observations had given
wrote that the experiment 'clearly prove[d] what wonderful promising results with vaccines in preventing colds.
effects the passions of hope and faith, excited by mere Controlled experiments, with persons in the control group
imagination, can produce on disease'. In fact, Haygarth had receiving no treatment, had also given favourable results.
not abandoned the Renaissance idea that imagination was The efficacy of cold vaccines was evaluated in several
the major mediator between body and mind12. placebo-controlled trials. The authors reported that
In 1863, Austin Flint tried to understand whether drugs
given for articular rheumatism changed the natural history 'the students in the control groups were treated in
of disease13. Flint placed 13 patients 'on the use of a exactly the same manner as those in the experimental
placebo which consisted, in nearly all the cases, of the groups but received placebos instead of vaccine. The
tincture of quassia, very largely diluted. This was given subjects in this group were given lactose-filled capsules
regularly, and became well known ... as the placeboic remedy which were indistinguishable from the capsules contain-
for rheumatism'. Flint, in accord with the understanding at ing the vaccine. They were prescribed with exactly the
his time, had no idea of concurrent control, but concluded same instructions as the capsules containing the vaccine.'
that 'the disease does end from self limitation' and that
inert treatment worked. The explanation was fitted into The trial gave negative results, although the results in the
prevailing dogma. Placebo was a fraud and deception that vaccine-treated groups were comparable with those in
had the 'moral effect of a remedy given specially for the previously reported experiments. It was the substantial
disease', but placebos did not affect the natural course of improvement in the placebo group that made the findings
disease; they were a priori excluded from having such an negative. The conclusion reads
impact. Placebos were therapeutic duds to manage patients,
or, as in the Flint investigation, a camouflage behind which 'one of the most significant aspects of this study is the
to watch nature take its course. great reduction in the number of colds which the mem-
In the beginning of this century, the German physician bers of the control groups reported during the
Adolf Bingel performed a large-scale comparative clinical experimental period. In fact these results were as good
trial to assess the specific effect of diphtheria antitoxin as many of those reported in uncontrolled studies which
serum in the treatment of diphtherial3. Bingel was con- recommended the use of cold vaccines'.
cerned whether the antitoxin in the serum was responsible
for the effect or whether treatment with serum not The placebo effect was born.
containing the antitoxin would give comparable results. He
alternately allocated 937 patients to either diphtheria
antitoxin serum or normal horse serum (the placebo) and FROM 'HUMBLE HUMBUG' TO 'POWERFUL
assessed the effect. He concluded that treatment with PLACEBO'
normal horse serum achieved a similar clinical outcome to The view on placebos until the 1950s was that 'it cannot
that with the antitoxic serum. harm and may comfort the patient'17. The placebo was
In the 1930s, several important papers were published considered a 'humble humbug'18. While using placebos in
with regard to the introduction of placebos in clinical research, clinicians began to recognize the therapeutic value
research. Evans and Hoyle'4 and Gold and colleagues'5 of administering inert preparations to patients in control
actually used the word placebo for the inert treatment groups of trials. Henry Beecher was one of the first
given to controls in an experimental situation. Both papers researchers to note this phenomenon. In his 1955 landmark
assessed the value of drugs used in the treatment of angina article 'The Powerful Placebo', he reviewed 15 placebo-
pectoris in cross-over experiments and deceptively controlled trials and concluded that, on average, the
administered placebos to the 'no-treatment' comparison magnitude of the placebo effect was 35.2%19. In retrospect,
group. Various drugs were given, interspersed with periods it is difficult to understand the large impact this paper had
of placebo administration. In both trials the drugs were since 13 of the 15 papers reviewed did not include no-
judged to exert no specific action that might be useful in the treatment groups. They could therefore not distinguish
treatment of angina. Gold and colleagues tried to explain between changes caused by the natural course of disease and
why inert interventions might work: their points included those caused by placebo. Remarkably, in the two studies
'confidence aroused in a treatment', the 'encouragement that included no-treatment controls no differences were
afforded by a new procedure' and 'a change of medical observed between the no-treatment group and the placebo
512 advisor'15. group. Beecher made the mistake many still make: effects
 JOURNAL OF THE ROYAL SOCIETY OF MEDICINE Volume 92 October 1 999

observed in the placebo-treated group were solely defined placebo as 'any therapeutic procedure which has an
attributed to the placebo. Anyhow, the paper has certainly effect on a patient, symptom, syndrome or disease, but
had great impact on the concept of placebo'4, and may be which is objectively without specific activity for the
responsible for the misconception that a fixed fraction (one- condition being treated'26. Brody defined it as 'an
third) of patients respond to placebos. Kienle and Kiene intervention designed to simulate medical therapy, that at
lately pointed to at least nineteen other possible reasons for the time of use is believed not to be a specific therapy for
the changes in the placebo-treated groups in Beecher's the condition for which it is offered'27. Hornung has
review20. pragmatically defined placebo as 'an empty preparation or
intervention imitating an effective preparation or interven-
SURGERY AS PLACEBO tion where one must decide on the "emptiness" of the
At the end of the 1950s, several reports suggested that a preparation in each particular situation'28.
surgical procedure, ligation of the internal mammary There has been considerable debate over the definition
artery, alleviated heart disease. At that time collateral of the placebo effect. Shapiro defined it as 'the psychological
vessels were believed to originate from the internal or psychophysiological effect produced by placebos'26.
mammary artery; thus, ligation of the artery would increase Brody proposed a wider definition. He considered the
coronary blood flow through collateral vessels proximal to placebo effect as 'a change in a patient's illness attributable
the point of ligation. Some researchers were sceptical about to the symbolic import of a treatment rather than a specific
the efficacy of the operation, and the technique was pharmacologic or physiologic property'27. Note that,
evaluated in two very small randomized clinical trials21'22 according to this definition, a placebo effect does not
All patients in both trials were actually operated on, but require a placebo. G0tzsche defined the placebo effect as
the artery was ligated in only half of them in one trial21, 'the difference in outcome between a placebo treated group
one-third in the other22. In both trials the rates of and an untreated control group in an unbiased experi-
improvement were the same in ligated patients and those ment'29. Anderson prefers to define the placebo effect by
who received skin incision. Subsequently the treatment was inclusion rather than by exclusion and suggests 'an effect in
abandoned, although neither trial had the statistical power which individually or culturally based expectations for a
to demonstrate equivalence. In 1961, Beecher used the data treatment cause are contributory to physical or psycho-
of these trials to calculate an overall improvement rate23. logical improvement after such a treatment'30.
Interestingly, he found the average rate to be 37%, and From a theoretical point of view, we think that
stated that 'its average magnitude is the same as other G0tzsche's definition of the placebo effect is correct.
average placebo effects encountered in disease'. However, it is difficult, if not impossible, to conduct
More recently, clinical trials of fetal dopamine cell unbiased experiments with a placebo-treated and an
implants for Parkinson's disease have been conducted. In untreated control group because the treatment allocation
these studies, one half of patients received the cell implants can not be masked. Brody's definition is conceptually clear,
while the other half had placebo surgery. but presents difficulties in operation. Methodologically, the
purpose of masking treatment allocation is to make
extraneous factors influencing the clinical course compar-
PLACEBO RESPONDERS AND PLACEBO able between groups. These extraneous factors include the
NON-RESPONDERS
placebo effects as defined by Brody, but also include
In the 1950s, researchers became interested in whether lifestyle adjustments and use of co-medication. G0tzsche
there were personality factors that would identify placebo clearly refers to the latter concept. In psychologically
responders (i.e. those who react exceptionally well to oriented research on placebo effects, Brody's definition is
placebo treatment24). If it were possible to separate placebo frequently used. However, in the methodological work,
responders from non-responders before the start of a trial, G0tzsche's definition is more common. This has led to
on the basis of personality characteristics, trials would some confusion in the past. We consider Brody's definition
become more efficient. Numerous experiments were the more appropriate.
conducted and many indicated certain personality char-
acteristics in responders; however, such findings could not
be replicated in separate studies and the personality of a HAWTHORNE EFFECT
placebo responder was never determined25. The Hawthorne effect was described in the 1930s after
investigations on the relation between illumination and
DEFINITION OF PLACEBO AND PLACEBO EFFECT industrial efficiency at the Hawthorne plant of the Western
Because of the scientific interest in placebos and placebo Electric Company in Chicago. According to legend, worker r
effects, definitions were needed. In the early 1960s, Shapiro productivity at the plant improved not only when the 51
 JOURNAL OF THE ROYAL SOCIETY OF MEDICINE Volume 92 October 1 999

illumination was increased but also, later, when it was physiological mechanism, whereby the placebo effect is
decreased. The reason was supposed to be the attention manifested. There is some experimental evidence that the
paid to the workers by the researchers and not the lighting placebo response in experimental pain is associated with
itself. Although subsequent investigations revealed that conditioning39 while other work points to the response
improvements in efficiency had probably resulted from expectancy model as more plausible40. Moreover, there is
other factors31, the term 'Hawthorne effect' has survived to some evidence that endogenous opioids are implicated in
describe the phenomenon whereby a subject's performance placebo analgesia41, though how they would act in the
changes simply because he or she is being studied. The proposed mechanisms remains unclear. In other disease
Hawthorne effect is not part of the placebo effect32. models, a psychoneuroimmunological response has been
suggested42.
NOCEBOS AND NOCEBO EFFECT
In 1961, Kennedy introduced the term nocebo to CURRENT STATUS OF PLACEBOS IN CLINICAL
distinguish the pleasing and salubrious effects of an empty TRIALS
preparation from its noxious effects33. A few years later, A frequent misconception is that a placebo-controlled trial
the concept of nocebo was elaborated by Kissel and is equivalent to a trial in which the control group receive no
Barrucand34. Hahn has lately proposed that nocebo effects treatment. If experimental treatment is complementary to
are in fact placebo side effects35. Kennedy and Kissel and standard care, it can be given on top of standard treatment.
Barrucand distinguished placebos from nocebos only in Then, blinding is implemented by giving placebo-controlled
terms of positive and negative outcomes. Hahn proposes a treatment. If the treatment under study is competitive with
definition of nocebo effect in which the expectation of the established therapy, experimental treatment is contrasted
recipient is accounted for. This means that, if one expects a with regular treatment. In this situation, blinding is
negative effect and this effect indeed occurs, it will be called established via a double-dummy technique. This means that
a true nocebo effect. Knowledge of nocebo effects is patients in the experimental group receive a placebo for
hindered by ethical concerns36. First, as with studies into control treatment, and patients in the reference group
the placebo effect, experimental studies of nocebo effects receive a placebo for the experimental treatment. In the
involve deception of participants. Thus, participants cannot evaluation of treatment strategies, placebos are generally
participate with full information, which makes informed not indicated. Here, cointerventions and behavioural
consent difficult. Second, such studies into nocebo effects changes are part of the strategy under study. This situation,
are expected to result in detrimental outcomes. Both these however, is prone to bias in assessment of patient outcome.
drawbacks make research into nocebo effects difficult. Measures such as outcome adjudication by a blinded
adjudication committee should then be taken.
WORKING MECHANISMS OF THE PLACEBO Before a drug can be registered, the legal requirement is
EFFECT proof of efficacy in 'adequate and well-controlled trials'.
Many groups have interpreted this rule as mandatory
Though sound scientific evidence is lacking, we believe that comparison with an untreated placebo group, but this
placebo effects exist across all disciplines of medicine. It interpretation is at odds with the ethical requirement that
seems therefore unlikely that a single universal theory can
established therapy may not be withheld from patients.
explain all placebo effects. Currently, several theories are Attention has been drawn to the unjustified use of placebos
taken seriously as possible explanations for the placebo in some clinical research43.
effect, among them classic conditioning, response ex-
pectancy and a psychoneuroimmunological response.
Classic conditioning considers the placebo response as a METHODOLOGICAL CONSIDERATIONS AND
conditioned Pavlovian-type learned response that has its FUTURE DIRECTIONS
basis in experience37. Inert substances, procedures, people A research design that can validly investigate components of
or treatment setting can all act as a conditioned stimulus for the placebo effect is the balanced placebo design44. A good
the alleviation of symptoms, if they have been repeatedly example is the study conducted by the general practitioner
associated with powerful unconditioned stimuli. This theory K B Thomas45. Thomas randomly assigned 200 symptom-
focuses on the input; the placebo effect arises because it is atic patients in whom no definite diagnosis could be made to
stimulus-expected. The second theory, response expect- one of four treatment arms-a consultation conducted in a
ancy38, is the anticipation of one's own automatic reactions 'positive' manner, with and without treatment; and a
to various situations. Response expectancy is different from consultation conducted in a 'negative' manner, with and
stimulus expectancy in that it focuses on the output rather without treatment. Two weeks after consultation, he found
514 than the input. Neither theory, however, offers a a significant difference in patient satisfaction between the
 JOURNAL OF THE ROYAL SOCIETY OF MEDICINE Volume 92 October 19991

positive and negative groups, but not between the treated 20 Kienle GS, Kiene H. The powerful placebo effect: fact or fiction. J Clin
and untreated groups. Epidemiol 1997;50: 1311-18
21 Cobb LA, Thomas GI, Dillard DH, Merendino KA, Bruce RA. An
By use of the balanced placebo design, we shall be evaluation of internal mammary artery ligation by a double-blind
able to identify important factors contributing to the technic. N Engl J Med 1959;260: 1115-18
placebo effect and to assess the impact of extraneous 22 Dimond EG, Kittle CF, Crockett JE. Comparison of internal
factors on specific treatment effects4. Thus, we should mammary artery ligation and sham operation for angina pectoris. Am
J Cardiol 1960;5:483-6
be helped to provide optimal treatment to individual
23 Beecher HK. Surgery as placebo. A quantitative study of bias. JAMA
patients and to discover the extent to which results of 1961;176:1102-7
placebo-controlled trials can be generalized. Knowledge 24 Beecher HK, Keats AS, Mosteller F, Lasagna L. The effectiveness of
of the components of placebo effects in different disease oral analgesics (morphine, codeine, acetylsalicycic acid) and the
models will contribute to both of these issues. Hence, problem of placebo "reactors" and "non-reactors". J Pharmacol Exp
Ther 1953;109:393-400
there is a need for investigations to identify the most 25 Doongaji DR, Vahia VN, Bharucha MP. On placebos, placebo
important non-specific factors. When these factors and responses and placebo responders. J Postgrad Med 1978;24:147-57
the mechanisms of action have been uncovered, we shall 26 Shapiro AK. Factors contributing to the placebo effect. Their
have to debate their implications for medical research implications for psychotherapy. Am J Psychother 1964;18:73-88
and clinical practice. 27 Brody H. Placebos and the Philosophy of Medicine. Clinical, Conceptual, and
Ethical Issues. Chicago: University of Chicago Press, 1980
28 Hornung J. Was ist ein Placebo? Die Bedeutung einer korrekten
REFERENCES Definition fuer die klinische Forschung. Forsch Komplementdrmed
1994; 1: 160-5
1 Shapiro AK. A historic and heuristic definition of the placebo.
Psychiatry 1964;27:52-8 29 G0tzsche PC. Is there logic in the placebo? Lancet 1994;344:925-6
2 Motherby G. A New Medical Dictionary. 2nd ed. London, 1785 30 Andersen LO. Letter. Lancet 1994;344: 1642
3 Quincy J. Quincy's Lexicon-Medicum. A New Medical Dictionary. Revised 31 Parsons HM. What happened at Hawthorne? Science 1974;183:922-32
by R Hooper. London, 1811 32 Fletcher RH, Fletcher SW, Wagner EH. Clinical Epidemiology: the
4 Jefferson T. The Writings of ThomasJefferson (edited by PL Ford, version Essentials. Baltimore, Maryland: Williams & Wilkins, 1984
9). New York: GP Putnam's, 1898 33 Kennedy WP. The nocebo reaction. Med World 1961;91:203-5
5 Cabot RC. The use of truth and falsehood in medicine: an 34 Kissel P, Barrucand D. Placebos et effet placebo en mddecine. Paris:
experimental study. Am Med 1903;5:344-9 Masson, 1964
6 Anonymous. Placebos. Med Rec 1885;27:576-7 35 Hahn RA. The nocebo phenomenon: concept, evidence, and
7 Kaptchuk TJ. Powerful placebo: the dark side of the randomised implications for public health. Prev Med 1997;26:607-11
controlled trial. Lancet 1998;351:1722-5 36 Hahn RA. The nocebo phenomenon: scope and foundations. In:
8 Handfield-Jones RCP. A bottle of medicine from the doctor. Lancet Harrington A, ed. The Placebo Effect: An Interdisciplinary Exploration.
1953;ii:823-5 Cambridge, MA: Harvard University Press, 1997
9 Feinstein AR. Clinical Epidemiology. The Architecture of Clinical Research. 37 Wickramasekera I. A conditioned response model of the placebo
Philadelphia: WB Saunders, 1985 effect. Predictions from the model. Biofeedback Se!f-reg 1980;5:5-18
10 http: /www.rcpe.ac.uk/cochrane 38 Kirsch I. Response expectancy theory and application: a decennial
review. Appl Prev Psychol 1997;6:69-79
11 Haygarth J. Of the Imagination, as a Cause and as a Cure of Disorders of the
Body; Exemplfied by Fictitious Tractors, and Epidemical Convulsions. Bath: 39 Voudouris NJ, Peck CL, Coleman G. The role of conditioning and
Crutwell, 1801 verbal expectancy in the placebo response. Pain 1990;43:121-8
12 McMahon CE. The 'placebo effect' in Renaissance medicine. J Am Soc 40 Montgomery GH, Kirsch I. Classical conditioning and the placebo
Psychosom Dentistry Med 1975;22:3-9 effect. Pain 1997;72:107-13
13 Kaptchuk TJ. Intentional ignorance: a history of blind assessment and 41 ter Riet G, de Craen AJM, de Boer A, Kessels AGH. Is placebo
placebo controls in medicine. Bull Hist Med 1998;72:389-433 analgesia mediated by endogenous opioids? A systematic review. Pain
1998;76:273-5
14 Evans W, Hoyle C. The comparative value of drugs used in the
continuous treatment of angina pectoris. Quartj Med 1933;26:311-38 42 Cohen S, Herbert TB. Health psychology: psychological factors
and physical disease from the perspective of human
15 Gold H, Kwit NT, Otto H. The xanthines (theobromine and psychoneuroimmunology. Ann Rev Psychol 1996;47:113-42
aminophylline) in the treatment of cardiac pain. JAMA 1937;108:
2173-9 43 Rothman KJ, Michels KB. The continuing unethical use of placebo
controls. N Englj Med 1994;331:394-8
16 Diehl HS, Baker AB, Cowan DW. Cold vaccines. An evaluation based
on a controlled study. JAMA 1938;111:1168-73 44 Kleijnen J, de Craen AJM, van Everdingen J, Krol L. Placebo effect in
double blind clinical trials: a review of interactions with medications.
17 Pepper OHP. A note on the placebo. AmJ Pharmacy 1945;117:409-12 Lancet 1994;344: 1347-9
18 Anonymous. The humble humbug. Lancet 1954;ii:321 45 Thomas KB. General practice consultations: is there any point in being
19 Beecher HK. The powerful placebo. JAMA 1955;159:1602-6 positive? BMJ 1987;294:1200-2

515