J Clin Epidemiol Vol. 50, No. 12, pp. 1311–1318, 1997 0895-4356/97/$17.

00
Copyright  1997 Elsevier Science Inc. PII S0895-4356(97)00203-5

The Powerful Placebo Effect: Fact or Fiction?

Gunver S. Kienle* and Helmut Kiene
Institut für Angewandte Erkenntnistheorie und Medizinische Methodologie, D-79112 Freiburg, Germany

ABSTRACT. In 1955, Henry K. Beecher published the classic work entitled ‘‘The Powerful Placebo.’’ Since
that time, 40 years ago, the placebo effect has been considered a scientific fact. Beecher was the first scientist
to quantify the placebo effect. He claimed that in 15 trials with different diseases, 35% of 1082 patients were
satisfactorily relieved by a placebo alone. This publication is still the most frequently cited placebo reference.
Recently Beecher’s article was reanalyzed with surprising results: In contrast to his claim, no evidence was
found of any placebo effect in any of the studies cited by him. There were many other factors that could account
for the reported improvements in patients in these trials, but most likely there was no placebo effect whatsoever.
False impressions of placebo effects can be produced in various ways. Spontaneous improvement, fluctuation
of symptoms, regression to the mean, additional treatment, conditional switching of placebo treatment, scaling
bias, irrelevant response variables, answers of politeness, experimental subordination, conditioned answers, neu-
rotic or psychotic misjudgment, psychosomatic phenomena, misquotation, etc.
These factors are still prevalent in modern placebo literature. The placebo topic seems to invite sloppy method-
ological thinking. Therefore awareness of Beecher’s mistakes and misinterpretations is essential for an appropriate
interpretation of current placebo literature. j clin epidemiol 50;12:1311–1318, 1997.  1997 Elsevier Science
Inc.

KEY WORDS. Placebo, placebo effect

Placebo effects have gained great popularity. Within the effect we have analyzed [21,22]). In the following article,
last three years hardly any major medical journal failed to the results of the analysis of the 15 trials reported in Bee-
have publications about placebo effects and their scientific cher’s article are described. The analysis is based on two
basis. The tradition of placebo research goes back to the questions: 1. Is the existence of the placebo effect demon-
fifties. It was in 1955 that Henry K. Beecher, with his fa- strated in those 15 trials that Beecher had surveyed in ‘‘The
mous and seminal article ‘‘The Powerful Placebo’’ [1], was Powerful Placebo’’? 2. If not, what are the factors that can
the first author to quantify the effects of placebos in a vari- create the false impression of a placebo effect?
ety of diseases. He claimed that the symptoms of 35% of
1082 patients in 15 studies [2–16] were satisfactorily re-
DEFINITION AND METHODS
lieved by placebos alone [1]. Today placebos are supposed
to be effective in almost every disease, and estimates of the It seems easy to define placebos: They are imitations of spe-
extent of the placebo effect even go far beyond Beecher’s cific treatments, with the absence of the specific therapeutic
35% [17–20]. constituents. However, defining placebos is a very contro-
This paper fundamentally questions the claimed extent versial topic [22–24]. Gøtzsche even concluded, ‘‘The pla-
of the placebo effect. A reanalysis of placebo literature was cebo concept as presently used cannot be defined in a logi-
carried out, with surprising results: A wide range of errors cally consistent way and leads to contradictions’’ [25]. From
was found in the placebo literature, which produced false reading Beecher’s own article, he refers to ‘‘pharmacologi-
impressions of placebo effects. cally inert substances’’ [1], the administration of which he
To illustrate these errors it is most appropriate to refer to considers can have ‘‘real therapeutic effects’’ [1]. Based on
the classic ‘‘The Powerful Placebo’’ [1] itself, because it is this, the criteria for acknowledging a placebo effect taken
still the most frequently cited paper on placebo and because for this present paper are as follows: (1) A placebo had to
its mistakes are still prevalent in placebo literature today be given. (2) The event had to be an effect of the placebo
(as far as we can judge from 800 articles on the placebo treatment, i.e., the event would not have happened without
placebo administration. (3) The event had to be relevant
*
Address for correspondence: Dr. med. Gunver S. Kienle, Institut für an- for the disease or symptom, i.e., it had to be a therapeutic
gewandte Erkenntnistheorie und medizinische Methodologie, Muselgasse
10, D-79112 Freiburg, Germany. event.
Accepted for publication on 20 August 1997. Besides these three criteria there were no other prede-
1312 G. S. Kienle and H. Kiene

fined criteria for the analysis. Basic medical knowledge and TABLE 1. Factors that can cause the false impression of a
common sense were the only scientific tools. placebo effect
Natural course of a disease
Spontaneous improvement
RESULT Fluctuation of symptoms
Regression to the mean
For 14 out of the 15 trial publications [2–16] detailed analy- Habituation
sis was possible. (One publication [4] did not give account Additional treatment
of the study design.) The overall result was that for none Observer bias
of these trials was there any reason to assume the existence Conditional switching of treatment
Scaling bias
of the slightest placebo effect. These studies were placebo- Poor definition of drug efficacy
controlled drug trials. Although they were not carried out in Irrelevant response variables
order to investigate placebo effects, Beecher retrospectively Subsiding toxic effect of previous medication
attributed the improvements in the placebo groups to effects Patient bias
of the placebo administration. However, on the basis of the Answer of politeness and experimental subordination
Conditioned answers
published data, in all of these trials the reported outcome Neurotic or psychotic misjudgment
in the placebo groups can be fully, plausibly, and easily ex- No placebo given at all
plained without presuming any therapeutic placebo effect. Psychotherapy
The published data of these trials make it quite obvious that Psychosomatic phenomena
there were a variety of reasons for the reported results, such Voodoo medicine
Uncritical reporting of anecdotes
as spontaneous improvements, additional treatments, meth- Misquotation
odological artifacts, etc. In some of the original trial publica- False assumption of toxic placebo effects created by
tions even the authors themselves had explicitly written Everyday symptoms
that there were no placebo effects. Misquotation
Beecher completely neglected all obvious reasons for the Persistence of symptoms
outcome in the placebo groups, simply calling the reported
results ‘‘real therapeutic effects’’ of placebo administration.
Thus, he totally misinterpreted the trials. receiving placebos felt better within 6 days (2 days after
Factors that have caused false impressions of placebo ef- the onset of placebo administration) [2] Beecher interpreted
fects—not only in Beecher’s but in other publications as these improvements as an effect of the placebo administra-
well—are listed in Table 1. Most of these factors are rele- tion [1]. However, he did not consider that many patients
vant in the 15 studies surveyed by Beecher; their distribu- with a mild common cold improve spontaneously within 6
tion is shown in Table 2 [1–16,21]. days (as already pointed out in the original publication [2]).
Beecher’s ‘‘The Powerful Placebo’’—presenting a quanti- Other examples: Four [8,9,12,14] of the trials in Beecher’s
tative ‘‘proof’’ of the existence of real therapeutic placebo list evaluated treatment of post-operative pain. Reanalyzing
effects—created a cognitive framework for further placebo these trials, it was possible in two [8,9] of these studies to
research in which all kinds of phenomena were registered determine the spontaneous diminishing of postoperative
as therapeutic placebo effects in a rather uncritical fashion pain on the basis of published data on the decrease of pa-
(further details see [21,22]). Therefore, in order to avoid tients’ demand for analgesics. This diminishing rate was
such obvious misinterpretations, it is important to know equal to that of Beecher’s claimed ‘‘placebo effect’’ [21].
those factors that can create illusions of placebo effects. Therefore, there is no reason to assume a placebo effect.
They will be described in the following. Examples will be Spontaneous improvement was a major factor in Bee-
taken from Beecher’s ‘‘The Powerful Placebo’’; for further cher’s misinterpretation of 10 of the 15 trials. This error is
illustration a few examples will be taken from a similarly wide-spread in the placebo literature.
classic German placebo survey [17].
Fluctuation of Symptoms
FACTORS THAT CAN CREATE FALSE In chronic diseases (or with chronic pain [26,27]) fluctuation
IMPRESSIONS OF PLACEBO EFFECTS of symptoms should be taken into account. Patients feel bet-
Spontaneous Improvement ter one day and worse the next. Therefore, looking at a
Spontaneous improvement of a disease does not occur as a number of chronically ill patients, one will simply always
result of a placebo administration; it is not an effect of a see some patients improving. Because of this, it is a mistake
placebo. This often seems to be disregarded in placebo liter- to forget to mention the rate of deterioration, and only re-
ature. port the rate of improvement and call the latter a placebo
In a placebo-controlled drug trial on acute common cold, ‘‘effect.’’
described as mild and of short duration, 35% of the patients For example, Beecher referred to patients with diseases
The Powerful Placebo Effect 1313

TABLE 2. Factors that created the illusion of a placebo effect in H. K. Beecher’s study list (quoted from [21])
Study [see references]
[2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16]

Percent of patients who were ‘‘satisfac-

torily relieved by a placebo,’’
according to Beecher [1] 35 38 52 58 26 38 21 26 31 37 26–40 30 15–53 36–43 30
Factors creating illusions of placebo effect
Spontaneous improvement x x x x x x x x x x
Spontaneous fluctuation of symptoms x x x x x
Conditional switching of treatment x x x x x x
Scaling bias x x x
Additional treatment x x
Irrelevant response variables x x x
Answer of politeness x x x
Conditioned answers x x
Neurotic or psychotic misjudgment x
Misquotation x x x x x x x x x x
Everyday symptoms misinterpreted as
placebo side effects x x
Habituation x x
Poor definition of drug efficacy x x x
Subsiding toxic effect of previous
medication x x
Demonstration of placebo effect? No No / a No No No No No No No No No No No No
a
The publication [4] gives no account of the study design.

such as ulcer, migraine, muscle tension, or headache who ous fluctuation of symptoms are special forms of regression
suffered from anxiety and tension and were treated for eight to the mean, i.e., the tendency of extreme values to move
2-week periods alternately with mephenesin and placebo closer to the average on repeated measurement. In their in-
[13]. Beecher claimed a placebo effect of 30% since teresting article, ‘‘How much of the placebo ‘effect’ is really
‘‘roughly’’ 20–30% of the patients improved. However, 10– statistical regression?’’ McDonald et al. [30] have argued,
20% of the patients deteriorated. As can be seen in a pub- that ‘‘most improvements attributed to the placebo effect
lished figure, there was only a net improvement of 5–10% are actually instances of statistical regression.’’
[21]. This seems a rather low rate, considering the observa-
tion period (16 weeks), the kind of diseases (ulcer, muscle
tension, headache, etc.) and possible improvement through Additional Treatment
the intermediate mephenesin treatment. Therefore, there So-called placebo effects often occur under additional treat-
was no reason to assume any placebo effect. (Besides, no ment. Of course there is no justification to call such im-
information about patient compliance was supplied in the provements a ‘‘placebo effect.’’
publication, and it was not ruled out that patients had other In one of the angina pectoris trials in Beecher’s list [6]
medical support.) the placebo group additionally received nitrates. In another
Neglecting spontaneous fluctuations of symptoms was the trial, concerning the common cold [2], the patients were
main reason why Beecher also misinterpreted three other allowed to take rest, hot baths, gargles, diets, etc.
trials [3,6,7]. This is a very common mistake also in other Many other examples can be found in the literature about
literature about placebos: A 20% placebo effect is claimed placebo effects. For instance, a study [31] supposedly shows
[17] for a placebo-controlled drug trial on patients with an- placebo effects in irritable colon treatment [17], but all pa-
gina pectoris. However, in the same trial, 72% [28] of the tients had been put on a special diet. In another study [32],
placebo-treated patients deteriorated. taken as a show case for placebo effects in alcoholism [17],
A 21% placebo effect is claimed [17] for a trial on cerebral patients in the placebo group received specialized medical
infarction, because 21% of the patients improved in the pla- and psychosocial support.
cebo-group. However, in that trial [29], 53% of the patients
on placebo died, even though every patient received the
best supportive medical care. (Of course, neither the im- Conditional Switching of Treatment
provement of 21%, nor the death of 53% of the patients In some of the trials in Beecher’s list the ‘‘placebo effect’’
can be reasonably attributed to placebo administration.) was further amplified by selecting patients in the following
Spontaneous improvement of diseases and the spontane- manner: When the patients felt well, they received a pla-
1314 G. S. Kienle and H. Kiene

cebo; when they felt worse, they were switched to active They had observed that patients often claimed pain relief
treatment, or they were excluded from evaluation until they in contrast to the physician’s impression. This observation
felt better again. is only peripherally related to the trials surveyed by Beecher,
In a study on angina pectoris [3], patients got placebos but it is of great importance in many other placebo reports
as long as they only had a few episodes of angina; when the [22]. The issue was recently described by Roberts [35]: ‘‘The
episodes increased, they received one of the test drugs. Thus word ‘placebo’ means ‘to please’ but this applies to both the
good periods were selected for placebo treatment, and bad patient and the doctor. For example, patients may report
periods were selected for drug treatment. Consequently, the positive outcomes to their physicians out of a need to ‘be
extent of the placebo effect was grossly overestimated. Simi- polite’ to them.’’ The same issue was addressed by Sackett
larly in one of the trials concerning treatment of postopera- [36]: ‘‘Finally, when the patient is grateful for clinician’s
tive pain [12], patients were only included when they had time and effort in trying to help them, this gratitude (plus
already improved to the degree that they could take oral simple good manners) often is reflected in an exaggeration
medication. When patients got worse again (pain increase, of the benefits of the latest prescription when they are asked
regurgitation, etc.), they were excluded from evaluation un- ‘Did that medicine help you?’ ’’
til they improved again. The same phenomenon was called a verbal (in contrast
to a real therapeutic) placebo effect by Kiene [37]. He also
mentioned the phenomenon of experimental subordination,
Scaling Bias i.e., in an experiment the subject says what he thinks he
In three of Beecher’s trials [2,3,7] there were false augmen- is expected to say, rather than what he really observes or
tations of placebo effects due to asymmetrical measurement experiences [37]. Similar phenomena have been described
scales [21]. The scales included two or more categories for by several other authors [38–41].
improvement, and only one or even none for deterioration. To differentiate polite answers or experimental subordi-
Thus the scales tempted patients to falsely give too many nation from true therapeutic placebo effects and to develop
positive reports. appropriate methods for this differentiation [22] is a key is-
sue in placebo research.
Irrelevant or Questionable Response Variables
Immense placebo effects can be claimed when they are Conditioned Answers
based on response variables which are irrelevant for the It seems difficult to differentiate therapeutic placebo effects
condition in question [21]: and conditioned effects. Numerous authors closely associate
There is the claim of a 73% placebo effect in multiple them or even presume that conditioning is the basic constit-
sclerosis [17]. The facts in the original publication [33] were uent of placebo effects [42–46]. However, a differentiation
that no objective change in the neurological condition was is necessary. Conditioned effects need specific presupposi-
found in any patient on placebo, yet 73% of the patients tions: First a specific unconditioned stimulus and second a
had the subjective feeling of increased euphoria, strength, specific setting, which is a very close temporal pairing of
and agility. However, euphoria is itself a symptom of multi- the unconditioned and the conditioned stimulus. In many
ple sclerosis; therefore an increase of euphoria is not neces- instances, conditioning even seems to work only when it
sarily a sign of improvement. Spontaneous variation of eu- superimposes biological rhythms. These specific presupposi-
phoric and optimistic answers are typical for this disease and tions are usually not present in clinical placebo situa-
therefore are inappropriate response variables for demon- tions.
strating placebo effects. Since Pavlov, many experiments on drug conditioned re-
Supposedly there is a 61% placebo effect in hypertension sponses in animals were carried out. But from these experi-
[17]. The facts in the original trial [34] were that there was ments one cannot conclude that healing or a real therapeutic
no significant change in blood pressure under placebo, but drug effect also can be provoked as a conditioned reflex.
61% of the patients subjectively felt better. However, all Surely, in cancer patients nausea and vomiting can be con-
patients had first received veratrum, which caused severe ditioned by repeated chemotherapy. But this does not mean
toxic symptoms in 64% of the patients. It was then substi- that tumor remissions can be conditioned as well. Unfortu-
tuted by placebos. Therefore the relief of symptoms in 61% nately, it is just the other way round: While conditioned
of placebo-treated patients can be explained by the cessa- vomiting often increases during chemotherapy cycles, there
tion of veratrum toxicity [21]. There is no reason to assume is generally a decrease in the therapeutic sensitivity of the
any placebo effect. tumor.
In fact, clinical experience contradicts the assumption
that healing can be conditioned. Episodes of chronic disease
Answers of Politeness and Experimental Subordination are usually more difficult to treat than the acute or first man-
In one of the trials on postoperative pain [8] the authors ifestation of an illness, even if this first manifestation has
discuss the ‘‘exceedingly difficult’’ criteria of pain relief. been treated successfully. (Classical conditioning paradigm
The Powerful Placebo Effect 1315

would predict just the opposite.) Moreover, there are many hand. (This differentiation is difficult, but not impossible;
severe symptoms that are treated effectively by regular and in fact, it is the psychiatrist’s daily work.) Neurotic or psy-
repeated drug administration. These therapeutic settings are chotic misjudgments can hardly give any valid evidence for
similar to conditional settings, and therefore should be ade- the existence of placebo effects.
quate for the conditioning of therapeutic effects. Yet when
interrupting such regular therapies, a rapid deterioration of
patients is observed in practice.
Nevertheless, conditioning may be important when giv- No Placebo Given at All
ing placebos, in that it can produce answers of politeness, There is a class of anecdotal reports in the placebo litera-
verbal placebo effects, and experimental subordination ture, which have nothing to do with placebos, because no
rather than effecting a true placebo effect. This realm of placebos were given at all [21].
communication seems far more susceptible to influences The purpose of these anecdotes is to demonstrate the pos-
such as conditioning than the realm of effective healing. sible power of ‘‘nonspecific’’ causes. Beecher himself re-
It is easier to provoke such communicative and behavioral ported adventurous episodes from the voodoo culture, when
reactions than true therapeutic placebo effects. This seems supposedly dying people recovered immediately, or when
to have happened in one of the pain trials [14] in Beecher’s magic rituals brought about the death of apparently healthy
list. It had a typical conditioning setting: Morphine and pla- people [53].
cebo were either alternated, or series of morphine adminis- Another classic example is an anecdote in Stewart Wolf ’s
trations were interrupted by placebos. In this setting the well known ‘‘The Pharmacology of Placebos’’ [54]: A
‘‘placebo effect’’ decreased after repeated placebo adminis- woman with a gastric ulcer could not respond with gastric
tration. One can find an easy explanation for this decrease, acid production during provocative tests with even the most
because it was just like in Pavlov’s classic experiments. powerful secretory drugs. Yet, immediate acid secretion oc-
When, in Pavlov’s experiments, the ringing of the bell re- curred when she was asked about her husband who, as she
peatedly was not combined with real food, the salivation had just recently discovered, had been sexually abusing her
decreased. Similarly, when in those patients the drug appli- 12-year-old daughter. Wolf used this story to demonstrate
cation (‘‘ringing of the bell’’) was not combined repeatedly the possible range of placebo effectiveness. However, this
with real pain relief (‘‘food’’), the patients’ positive answers is misleading. This was an example of a psychosomatic ef-
(‘‘salivation’’) decreased. This means that the patients grad- fect, not the effect of placebo application. The example
ually recognized that they were receiving inactive treat- does not show that the mere ritual of giving a pill can be
ments, and that only verbal placebo effects had been condi- equated with the effect of discovering the sexual abuse of
tioned, not real ones [21]. A key issue when judging placebo one’s daughter by one’s husband.
effects is to decide whether a patient’s report is true or
not.
In an example of a crossover placebo-controlled study on
hypertension, a conditioned reduction of blood pressure was Uncritical Reporting of Anecdotes
shown, however, it was short term (a few days). Notably, One needs to be cautious about claims of placebo effects
when placebos were given as first treatment within this not only in clinical trials, but also in case reports. While
crossover design, no antihypertensive effect occurred, al- most scientists are reasonably skeptical regarding therapeu-
though 83% of the patients had previously been treated tic benefits from drugs, they welcome reports about placebo
with antihypertensive remedies [47]. Thus, in this trial only effects with uncritical enthusiasm [55]. For example, Bee-
a short-term conditioned effect occurred, due to the specific cher demonstrates the power of nonspecific effects by the
conditioning setting, while there was no placebo effect. following story [53]: A middle-aged woman underwent sur-
These findings concur with several trials on placebo in hy- gical exploration because of cancer, which was then found
pertension [48–51]; they did not show any placebo effect to be inoperable. When she had recovered from anesthesia,
either. one of her relatives told her the truth about her illness.
Within the next hour the woman went into cardiovascular
shock and died after a few hours.
Neurotic or Psychotic Misjudgment This story, however, does not testify for nonspecific ef-
The reliability of a patient’s report is often particularly dif- fects. Before diagnosing such a mysterious ‘‘placebo’’ death,
ficult to assess in neurotic or psychotic disturbances [21]. every rational doctor must first rule out the most likely
Here the placebo literature offers fascinating stories [52]. causes: postoperative complications, such as bleeding or pul-
However, one should not forget that a common feature in monary embolism. These are frequent hazards after opera-
psychosis or neurosis is disturbed interpretation of reality. tions and in cancer patients.
Therefore one clearly has to differentiate between a psy- Many such uncritical placebo anecdotes, although im-
chotic or neurotic misjudgment on the one hand, and a pressive, come to nothing when they are looked at a little
correct observation of a therapeutic effect on the other closer [21].
1316 G. S. Kienle and H. Kiene

Misquotation they get worse [21]. For example, in a trial on chronic pain
A particular problem of placebo literature seems to be that 13% of the patients in the placebo group improved, and
of misquotations. An example is Beecher’s claim that in a 20% deteriorated. While the improvement was interpreted
study of antitussive agents [15] there was a placebo effect as a therapeutic placebo effect, the deterioration was inter-
in 36% of 22 patients and in 43% of another 22 patients. preted as a toxic placebo effect [19,59].
However, the actual result was, that under none of the pla-
cebo administrations could any significant change be dem- DISCUSSION
onstrated. Besides, there were no 22 placebo-treated pa-
tients (the groups were much smaller), and there were no Beecher’s ‘‘The Powerful Placebo,’’ published in 1955, has
reports about any 36% or 43% of patients. Thus, Beecher’s been a seminal and most influential paper. It is still the most
quotation was wrong (which is amazing, as Beecher himself frequently cited placebo reference. This is amazing, as none
had been one of the authors of the original publication). of the original trials cited by Beecher gave grounds to as-
Beecher misquoted 10 of the 15 trials listed in ‘‘The Pow- sume the existence of placebo effects. The reanalysis of a
erful Placebo.’’ He sometimes inflated the percentage or the similar classic German placebo survey [17] gave the same
number of patients, or he cited as a percentage of patients results. No placebo effects could be found [21].
what in the original publications is referred to as something The conceptual and methodological mistakes of Bee-
completely different, such as the number of pills given, the cher’s classic paper are still prevalent today. Although some
percentage of days treated, the amount of gas applied in modern experimental placebo research is of better method-
an experimental setting or the frequency of coughs after ological quality, a valid demonstration of therapeutic pla-
irritating a patient [21]. The main effects of these errors cebo effects still appears lacking. Having analyzed a total
were false inflations of the alleged placebo effect. A multi- of 800 articles on placebo, we have not found any reliable
tude of misquotations can also be found in other placebo demonstration of the existence of placebo effects. (In bron-
literature [21]. chial asthma effects of suggestion are documented under ex-
perimental conditions. This, however, does not imply the
existence of an efficacious placebo therapy of bronchial
False Assumptions of Toxic Placebo Effects asthma [22]).
Beecher did not only write about ‘‘real therapeutic effects’’ Comparing placebo-treated and untreated patients might
of placebo administration; he also wrote about ‘‘toxic and be a valid method for investigating placebo effects. (In one
other side-effects of placebos.’’ He states that in various tri- of the trials [5] of Beecher’s list there was an untreated con-
als there had been 35 different toxic effects of placebos such trol group; it showed the same result as the placebo group.)
as dry mouth, nausea, headache, drowsiness, warm glow, fa- As these trials, however, do not control for factors such as
tigue, and sleep. The frequency ranged from 8% to 50%. answers of politeness, experimental subordination and addi-
For this, Beecher did not quote any references. tional treatment they can create false positive results. For
When judging toxic placebo effects one needs to take instance, Ernst and Resch [60] systematically collected trials
into account the studies by Green [57] and by Reidenberg that included both a placebo-treated and an untreated
et al. [58]. They demonstrated that many people experience group. They found four trials that showed superior outcomes
everyday symptoms such as dry mouth, headache, drowsiness, in the placebo groups. The best trials were two 5-arm ran-
fatigue, etc. The frequency of these symptoms is similar to domized trials on ultrasound treatment of postoperative
the frequency of Beecher’s so-called ‘‘placebo side-effects.’’ swelling. As there were better results in the placebo group
Therefore, it is very likely that these everyday symptoms (i.e., turned off ultrasound apparatus), than in the untreated
are documented in a trial situation and are then misinter- groups [61,62] the results were categorized as ‘‘true’’ and
preted as ‘‘side-effects’’ of the placebos. ‘‘substantial’’ placebo effects [60]. However, in the placebo
With respect to ‘‘toxic placebo effects,’’ one always has groups, a coupling cream was also applied, the humidity and
to consider the possibility of misquotations. In one of the cooling effect of which possibly reduced the postoperative
publications in Beecher’s trial list, the authors [13] reported swelling. Consequently, the existence of a placebo effect is
an impressive finding that 61% of the placebo patients in questionable in these trials, too. A possibility to do placebo
a streptomycin trial showed the specific toxic effects of research lies within balanced study designs [63,64] (2 3 2
streptomycin, including high-tone and low-tone hearing factorial designs: verum vs. placebo, strong vs. weak sugges-
loss, eosinophilia, and impairment of urea clearance. This tion of efficacy). However, variations of outcome do not
remarkable placebo toxicity has been passed on in the medi- indicate true therapeutic placebo effects as long as experi-
cal literature. However, going back to the original publica- mental subordination has not been ruled out [37].
tion [56] one will find that none of the patients in the strep- There can be no doubt that the extent and frequency of
tomycin trial ever received a placebo. placebo effects as published in most of the literature are
Finally, symptoms are called ‘‘side effects’’ of placebo gross exaggerations. Some placebo experts have had some
treatment, only because they do not disappear or because awareness of these issues. For example, Shapiro and Shapiro
The Powerful Placebo Effect 1317

[65] wrote: ‘‘In our opinion, the belief that placebos and patients with angina of effort. A study of khellin (visammin).
psychological factors have a specific and clinically meaning- Am J Med 1950; 9: 143–155.
8. Keats AS, Beecher HK. Pain relief with hypnotic doses of bar-
ful effect on physical illness is not supported by a critical, biturates and a hypothesis. J Pharmacol Exp Ther 1950; 100:
data-oriented review of the literature.’’ Even more drasti- 1–13.
cally, Roberts [35] said: ‘‘The so-called placebo effect is a 9. Keats AS, D’Alessandro GL, Beecher HK. A controlled study
myth born of misperception, misunderstanding, mystery and of pain relief by intravenous procaine. J Am Med Assoc 1951;
hope.’’ However, these comments remained isolated even 147: 1761–1763.
10. Beecher HK, Deffer PA, Fink FE, Sullivan DB. Field use of
in the Shapiros’ and Robert’s own publications. methadone and levo-iso-methadone in a combat zone. U S
Undoubtedly, psychosomatic effects exist. Hence, clear Armed Forces Med J 1951; 2: 1269–1276.
differentiation between placebo and non-placebo compo- 11. Hillis BR. The assessment of cough-suppressing drugs. Lancet
nents in therapeutical settings [22] is essential for valid pla- 1952; June 21: 1230–1235.
cebo research and for research in complementary medicine. 12. Beecher HK, Keats AS, Mosteller F, Lasagna L. The effective-
ness of oral analgesics (morphine, codeine, acetylsalicylic
Many factors and phenomena have been summed up under acid) and the problem of placebo ‘‘reactors’’ and ‘‘non-reac-
the terms ‘‘placebo’’ and ‘‘placebo effect,’’ without being pla- tors.’’ J Pharmacol & Exper Therap 1953; 109: 393–400.
cebos or effects of placebo administrations. Those factors and 13. Wolf S, Pinsky RH. Effects of placebo administration and the
phenomena were taken as evidence of ‘‘true therapeutic pla- occurrence of toxic reactions. J Am Med Assoc 1954; 155:
cebo effects,’’ although they are not. Thus, The Powerful 339–341.
14. Lasagna L, Mosteller F, Felsinger JM, Beecher HK. A study
Placebo turns out to be a fiction. of the placebo response. Am J Med 1954; 16: 770–779.
One might consider to substitute the term ‘‘placebo ef- 15. Gravenstein JS, Devloo RA, Beecher HK. Effect of antitussive
fect’’ by the term ‘‘non-specific effects.’’ Changing of words, agents on experimental and pathological cough in man. J
however, does not change the situation that the existence Appl Physiol 1954; 7: 119–139.
of therapeutic effects of placebo administration seems ques- 16. Lasagna L, Felsinger JM, Beecher HK. Drug-induced mood
changes in man. 1. Observations on healthy subjects, chroni-
tionable. Besides, as Peek [66] and Grünbaum [67] have al- cally ill patients, and ‘‘postaddicts.’’ J Am Med Assoc 1955;
ready pointed out, there is no such thing as ‘‘non-specific 157(12): 1006–1020.
effects.’’ The term is a contradictio in adjecto: a contradiction 17. Netter P, Classen W, Feingold E. Das Placeboproblem. In:
in itself. Dölle W, Müller-Oerlinghausen B, Schwabe U, Eds. Grundla-
Finally, these factors that can create false impressions of gen der Arzneimitteltherapie–Entwicklung, Beurteilung
und Anwendung von Arzneimitteln. Mannheim, Wien, Zü-
placebo effects might seem to be compelling reasons for ran- rich: Wissenschaftsverlag; 1986: 355–366.
domization and blinding. However, analyses indicate that 18. Roberts AH, Kewman DG, Mercier L, Hovell M. The power
these factors are not necessarily distributed equally in drug of nonspecific effects in healing: Implications for psychosocial
and control groups, thus challenging the validity of random- and biological treatments. Clin Psychol Rev 1993; 13: 375–
ized double-blind trials [37,68–70]. This issue needs further 91.
19. Turner JA, Deyo RA, Loeser JD, Von Korff M, Fordyce WE.
investigation. The importance of placebo effects in pain treatment and re-
search. J Am Med Assoc 1994; 271(20): 1609–1614.
We are indebted to Prof. Joachim Hornung, Dr. Gerben ter Riet, Dr. 20. Bodem SH. Bedeutung der Placebowirkung in der praktischen
Klaus Linde, Dr. Harald Walach, Dr. Michael Evans, Dr. Gerald Arzneitherapie. Pharm Ztg 1994; 139(51/52): 9–19.
Karnow, Dr. Lou Aventuro, Dr. Frank Mulder, and Dr. Peter van 21. Kienle GS. Der sogenannte Placeboeffekt; Illusion, Fakten, Re-
Leeuwen for their critical commentary on the manuscript. alität. Stuttgart, New York: Schattauer Verlag GmbH; 1995.
22. Kienle GS, Kiene H. Placebo effect and placebo concept: A
critical methodological and conceptual analysis of reports on
the magnitude of the placebo effect. Altern Ther Health Med
References 1996; 2(6): 39–54. [Reprint from: Kienle GS, Kiene H. Pla-
1. Beecher HK. The powerful placebo. J Am Med Assoc 1955; ceboeffekt und Placebokonzept—eine kritische methodolog-
159(17): 1602–1606. ische und konzeptionelle Analyse von Angaben zum Ausmaß
2. Diehl HS. Medicinal treatment of the common cold. J Am des Placeboeffekts. Forsch Komplementärmed 1996; 3: 121–
Med Assoc 1953; 101: 2042–2049. 138.]
3. Evans W, Hoyle C. The comparative value of drugs used in 23. White L, Tursky B, Schwartz GE, Eds. Placebo—Theory, Re-
the continuous treatment of angina pectoris. Quart J Med search, and Mechanisms. New York, London: Guilford Press;
1933; 7: 311–338. 1985.
4. Jellinek EM. Clinical tests on comparative effectiveness of an- 24. Hornung J. Was ist ein Placebo? Die Bedeutung einer korrek-
algesic drugs. Biometrics 1946; 2(5): 87–91. ten Definition für die klinische Forschung. Forsch Komplem-
5. Gay LN, Carliner PE. The prevention and treatment of mo- entärmed 1994; 1: 160–165.
tion sickness. Bull Johns Hopkins Hosp 1949; 84: 470–487. 25. Gøtzsche PC. Is there logic in the placebo? Lancet 1994; 344:
6. Travell J, Rinzler SH, Bakst H, Benjamin ZH, Bobb A. Com- 925–926.
parison of effects of alpha-tocopherol and a matching placebo 26. Whitney CW, Von Korff M. Regression to the mean in
on chest pain in patients with heart disease. Ann New York treated versus untreated chronic pain. Pain 1992; 50: 281–
Acad Sci 1949; 52: 345–353. 285.
7. Greiner Th, Gold H, Cattell M, Travell J, et al. A method 27. Deyo RA. Practice variations, treatment fads, rising disability.
for the evaluation of the effects of drugs on cardiac pain in Spine 1993; 18(15): 2153–2162.
1318 G. S. Kienle and H. Kiene

28. LeRoy GV. The effectiveness of the xanthine drugs in the 49. Mancia G, Omboni S, Parati G, Ravogli A, Villani A, Zan-
treatment of angina pectoris. J Am Med Assoc 1941; 116(10): chetti A. Lack of placebo effect on ambulatory blood pressure.
921–925. Am J Health 1995; 8: 311–315.
29. Dyken M, White PT. Evaluation of cortisone in the treatment 50. Weber MA, Neutel JM, Smith DHG. Controlling blood pres-
of cerebral infarction. J Am Med Assoc 1956; 162(17): 1531– sure throughout the day: Issues in testing a new anti-hyperten-
1534. sive agent. J Hum Hypertens 1995; 9(Suppl. 5): 29–35.
30. McDonald C, Mazzuca S, McCabe G. How much of the pla- 51. Report of Research Council Working Party on mild to moder-
cebo ‘effect’ is really statistical regression? Stat Med 1983; 2: ate hypertension. Randomized controlled trial of treatment for
417–427. mild hypertension: Design and pilot trial. Br Med J 1977; 1:
31. Lichstein J, DeCosta Mayer J, Hauch EW. Efficacy of methan- 1437–1440.
theline (banthine) bromide in therapy of the unstable colon. 52. Schindel L. Placebo und Placebo-Effekte in Klinik und
J Am Med Assoc 1955; June 25: 634–637. Forschung. Arzneimittelforschung 1967; 17: 892–918.
32. Wells RE. Use of reserpine (Serpasil) in the management of 53. Beecher HK. Die Placebowirkung als unspezifischer Wirkungs-
chronic alcoholism. J Am Med Assoc 1957; 163(6): 426–429. faktor im Bereich der Krankheit und der Krankenbehandlung.
33. Blomberg LH. Treatment of disseminated sclerosis with active In: Gross F, Beecher HK, Eds. Placebo—das universelle Med-
and inactive drugs. Lancet 1957; 1: 431–432. ikament? Paul-Martini-Stiftung der Medizinisch Pharma-
34. Coe WS, Best MM, Kinsman JM. Veratrum viride in the zeutischen Studiengesellschaft e.V. Mainz: Eggebrecht-
treatment of hypertensive vascular disease. J Am Med Assoc Presse; 1984.
1950; 143(1): 5–7. 54. Wolf S. The pharmacology of placebos. Pharmacol Rev 1959;
35. Roberts AH. The powerful placebo revisited: The magnitude 11: 689–704.
of nonspecific effects. Mind/Body Medicine 1995; March; 1– 55. Hollister L. Placebology: Sense and nonsense. Curr Ther Res
10. 1960; 2(9): 477–483.
36. Sackett DL. Randomized trials in individual patients. In: 56. Veterans Administration. Minutes of the Fifth Streptomycin
Antes G, Edler L, Holle R, Köpcke W, Lorenz R, Windeler Conference; Knickerbocker Hotel, Chicago, Illinois; April
J, Eds. Biometrie und unkonventionelle Medizin. Münster– 15–18, 1948.
Hiltrup: Landwirtschaftsverlag GmbH; 1995: pp. 19–33. 57. Green DM. Pre-existing conditions, placebo reactions, and
37. Kiene H. A critique of the double-blind clinical trial. Parts 1 ‘‘side effects.’’ Ann Intern Med 1964; 60(2): 255–265.
1 2. Altern Ther Health Med 1996; 2(1): 74–80 and 1996; 58. Reidenberg MM, Lowenthal DT. Adverse nondrug reactions.
2(2): 59–64. [Reprint from: Kiene H. Kritik der klinischen N Eng J Med 1968; 279(13): 678–679.
Doppelblindstudie. MMW-Taschenbuch. München: MMW 59. Long DM, Uematsu S, Kouba RB. Placebo responses to medi-
Medizin; 1993: 35.] cal device therapy for pain. Stereotact Funct Neurosurg
38. Barber TX. The effects of ‘‘hypnosis’’ on pain. Psychosom 1989; 53: 149–156.
Med 1963; 25: 303–333. 60. Ernst E, Resch KI. Concept of true and perceived placebo ef-
39. Clark WC. Sensory-decision theory analysis of the effect on fects. Br Med J 1995; 311: 551–553.
the criterion for pain and thermal sensitivity. J Abnorm Psy- 61. Ho KH, Hashish I, Salmon P, Freeman R, Harvey W. Reduc-
chol 1969; 74(3): 363–371. tion of post-operative swelling by a placebo effect. J Psy-
40. Fordyce WE, Lansky D, Calsyn DA, Shelton JL, Stolov WC, chosom Res 1988; 32(2): 197–205.
Rock DL. Pain measurement and pain behavior. Pain 1984; 62. Hashish I, Harvey W, Harris M. Anti-inflammatory effects of
18: 53–69. ultrasound therapy: Evidence for a major placebo effect. Br J
41. Tedeschi JT, Schlenker BR, Bonoma TV. Cognitive disso- Rheumatol 1986; 25: 77–81.
nance: Private ratiocination or public spectacle? Amer Psy- 63. Ross S, Krugman A, Lyerly S, Clyde D. Drugs and placebos:
chol 1971; 26: 685–695. A model design. Psychol Rep 1962; 10: 382–392.
42. Ader R. Conditioned Immunopharmacological Effects in Ani- 64. Marlatt A, Rohsenow D. Cognitive processes in alcohol use:
mals: Implications for a Conditioning Model of Pharmaco- Expectancy and the balanced placebo design. Adv Subst
therapy. In: White L, Tursky B, Schwartz GE, Eds. Placebo— Abuse 1980; 1: 159–199.
Theory, Research, and Mechanisms. New York: Guilford 65. Shapiro AK, Shapiro E. Patient-provider relationships and
Press; 1985. the placebo effect. In: Matarazzo JD, Weiss SM, Herd JA,
43. Peck C, Coleman G. Implications of placebo theory for clini- Miller NE, Eds. Behavioral Health: A Handbook of Health
cal research and practice in pain management. Theor Med Enhancement and Disease Prevention. New York: Wiley-In-
1991; 12: 247–270. terscience; 1984: 371–383.
44. Voudouris NJ, Peck CL, Coleman G. The role of conditioning 66. Peek CJ. A critical look at the theory of placebo. Biofeedback
and verbal expectancy in the placebo response. Pain 1990; Self-Regul 1977; 2: 327–335.
43: 121–128. 67. Grünbaum A. Explications and implications of the placebo
45. Wall PD. Pain and the placebo response. Ciba Found Symp concept. In: White L, Tursky B, Schwartz GE, Eds. Placebo—
1993; 174: 187–211. Theory, Research, and Mechanisms. New York: Guilford
46. Wickramasekera I. A conditioned response model of the pla- Press; 1985.
cebo effect: Predictions from the model. In: White L, Tursky 68. Kirsch I, Weixel LJ. Double blind versus deceptive administra-
B, Schwartz GE, Eds. Placebo—Theory, Research, and tion of a placebo. Behav Neurosci 1988; 102(2): 319–323.
Mechanisms. New York: Guilford Press; 1985. 69. Hornung J. Zur Problematik der Doppelblindstudien. Thera-
47. Suchman AL, Ader R. Classic conditioning and placebo ef- peutikon 1989; 3: 696–701.
fects in crossover studies. Clin Pharmacol Ther 1992; 52(4): 70. Kirsch I, Rosadino MJ. Do double-blind studies with informed
372–377. consent yield externally valid results? Psychopharmacology
48. Iacono P, Drici MD, De Lunardo C, Salimbeni B, Lapalus P. 1993; 110: 437–442.
Placebo effect in cardiovascular clinical pharmacology. Int J
Clin Pharmacol Res 1992; XII(2): 53–56.