A

RECENT review of between 500 and 600 publications on a leading tranquilizer states that only 37 of these studies meet certain minimum standards of scientific acceptability.' Many clinical investigators and journal editors do not appear dismayed by this overwhelming evidence of substandard research. These individuals appear unduly convinced of the value of uncontrolled drug studies, viewing double-blind and other controlled procedures as merely gilding the lily. Can we afford this attitude, particularly in dealing with disorders of human behavior in which subjectivity, bias, and the personal element are all too prominent? The present experiment was designed to shed some light on the answer to this question.

As its etymology suggests, the word "placebo" originally referred to a pharmacologically inactive substance which was given to placate the patient, presumably when no effective treatment was available or indicated. In more recent times placebo has also come to refer to the inert substance given as the "dummy" in objective studies of medications, to control for suggestibility or expectation on the part of the patient and observer.

In drug investigations classes of phenomena have been referred to as placebo effects. 1. Strictly speaking, the term refers to the changes produced in an individual by the set, expectancy, or sug gestibility which may accompany the taking of medication. 2. A broader definition recognizes the fact that the one observing or evaluating the individual receiving medication may be influenced by his own set or suggestibility; in other words, an observer may perceive changes in a patient, when in fact the patient is no different than he was before taking the medication. 3. Finally, there is a contaminated usage which tends to consider most changes which follow the administration of placebo as placebo effects. Those who use the term in this third sense fail to take cognizance of the fact that changes may occur in a patient due to endogenous or environmental events which just happen to coincide with the ingestion of placebo, and which have nothing to do with expectancy, From the New York Hospital, Westchester Division.

A drug investigation involving 120 hospitalized psychiatric patients Was simulated. It appeared to participating patients, psychiatrists, and nurses, that a new tranquilizer and a new energizer were to be evaluated, but both "drugs" were actually placebos. According to uncontrolled and subjective methods pf evaluation, 53% to 80% of the patients benefited from the new "drugs." When matched control groups and objective rating scales were used in the evaluation, a temporary improvement was caused by the tranquilizer but not by the energizer. The tendency to attribute improvement to what were thought to be active drugs, and the discrepancy in one experiment between controlled and uncontrolled methods of evaluation, dramatically il lustrates the dubious value of studies which do not employ double-blind and other controlled procedures.

suggestibility, or the ritual of pill-taking. In a drug study the error introduced by such extraneou events is actually removed by a control gro and not by a placebo. However, the present es periment is concerned with all 3 uses of the term

Method

This investigation was designed to measur placebo effect in a psychiatric hospital env ment, where not only the patients but also the psychiatrists and nurses would be unaware of the real nature of the study. Only the present invest gators and the medical director of the hospital knew the actual purpose of the experiment Th psychiatrists and nurses were given a standast description of what the study was supposed to be about. They were told that under a grant im the National Institute of Mental Health the bur pital was going to undertake a detailed cimc evaluation of a new tranquilizer, NIMH-Jan. 457 and a new energizer, NIMH-Mar. 2056. They wan

further informed that the chemical composition of the 2 drugs was to be withheld to prevent bias in the staff's evaluation. Such bias, they were told, might arise if they were familiar, through their own experience or the literature, with drugs of similar chemical characteristics. However, infornation was provided about the general properties of the drugs, such as their indications, dosage -egimens, and possible side-effects. The staff was eassured that, although both compounds were new, they had been used by several investigators, ad shown promise, and appeared to be free from oxicity.

In reality, neither of the 2 drugs was an active gent. The tranquilizer (NIMH-Jan. 4057) was a lacebo consisting of an orange tablet of lactose with quinine added to give it a bitter taste. The nergizer (NIMH-Mar. 2056) was a yellow tablet If lactose with di-calcium phosphate added to proide a bland taste. The 2 placebos also differed in ze, and both "drugs" were available in “25-mg." nd "50-mg." strengths.

To guard against the possibility that some of he staff might raise the question of whether placeos as well as active agents were to be used in e experiment, the following statement was conuned in the description given to the staff: “In e judgment of those directing this experiment, e double-blind placebo design can be misleading nd generate a false confidence in the results. It natural for the staff to indulge in guessing as to ho is getting the placebo. A transitory side effect ke drowsiness is all that is needed to reveal the entity of drug patients. The use of a carefully atched control group, base lines, and statistical chniques is considered a superior procedure c)." The study lasted for one year, and there as little or no evidence that any psychiatrist, irse, or patient suspected either that placebos ere being employed, or, especially, that no active ents whatsoever were being used.

A total of 120 patients were employed, 60 in the anquilizer study and 60 in the energizer study. ithin both the tranquilizer and energizer experients patients were assigned at random to either 'drug" or a control group, with an attempt made match the 2 groups in an experiment on the sis of age, sex, diagnosis, length of hospitalizan, and severity of illness. In the tranquilizer dy the patients in the "drug" group received drug during 2-week pretreatment and 2-week st-treatment base-line periods, while during the ervening 6 weeks they received the tranquilizer MH-Jan. 4057. The design of the energizer study s identical, the only exception being that the ug" group received the energizer NIMH-Mar. 6. At no time during the 10-week study did ients in the control groups receive placebo, ive agents, or any other somatotherapy. All 4 ups of 30 patients received psychotherapy and

daily occupational and recreational therapies. During the study, biweekly blood counts were performed on the patients in the 2 "drug" groups. Blood pressure recordings were made twice daily during the first 3 days of "drug" administration and twice weekly thereafter.

The dosage was gradually increased during the first week of "drug" administration and decreased during the final week. The maintenance dose for the tranquilizer was "50 mg.," 4 times daily, with provision that it could be increased to "75 mg." if a patient's physician considered the response inadequate. The energizer maintenance dosage was set at "50 mg.." 3 times daily. The difference in dosage regimen between tranquilizer and energizer was employed to reinforce the illusion that 2 different agents were being studied. The more conservative approach to the dosage of the energizer was also intended to reflect the fact that at the time of the study, the development of energizers or antidepressants lagged far behind that of tranquilizers. Capital was made of the fact that the present investigation coincided with the early clinical reports on the new psychic energizer iproniazid. A few physicians inquired whether the new energizer was iproniazid, and they were told that the only available information concerned the clinical properties of the drug, and that these appeared somewhat similar to those attributed to iproniazid. At the time the experiment was near completion, some of the staff became aware of reports that some patients who were taking iproniazid had died of hepatitis. They expressed concern over the possible toxicity of NIMH-Mar. 2056, and were then informed that the "drug" definitely was not iproniazid.

Patient Sample

Patients were selected over a one-year period (July, 1957-June, 1958) from the population of the New York Hospital-Westchester Division, a 326bed voluntary nonprofit hospital for the care and treatment of psychiatric disorders. Patients from the 2 extremes of the hospital population were not employed in the experiment. The severely agitated or extremely depressed were given the treatment of choice in the hospital for such conditions, which is electroshock therapy or, in a few instances, the immediate administration of tranquilizers. Patients approaching convalescent status were not considered, because they were likely to leave the hospital before completing the study, and in most cases had reached a level of improvement at which medication did not appear to be indicated. Patients who had received electroshock treatments within 6 weeks or tranquilizing or stimulant medication within 4 weeks also were not eligible. Patients were assigned to the tranquilizer study if they had a recent history or predominant symptomatology of anxiety, agitation, hostility, or hyper

activity; those with a recent history or predominant symptoms of depression, apathy, seclusiveness, or fatigue were placed in the energizer study.

Most of the patients were acutely ill, recent or comparatively recent admissions, 80% having been in the hospital one year or less. The length of time that had elapsed since admission ranged from 2 days to 4.6 years, the median being 3 months. Twenty-eight per cent had a history of previous hospitalization prior to the one-year period preceding their present admission. The sample included 70 women and 50 men. The patients ranged in age from 13 to 77, the mean being 37 and the standard deviation 16. More than half (55%) had been educated beyond high school, and the average patient was from a middle to upper-middle socioeconomic level. In the tranquilizer experiment two-thirds of the patients had a diagnosis of schizophrenia, and the remainder that of manic-depressive psychosis, involutional psychosis, or psychoneurosis. In the energizer experiment half the patients were schizophrenics, and the remainder manic-depressives, involutional psychotics, psychoneurotics, or sociopaths.

Once a patient began the 10-week period of study, every effort was made to have him complete the experiment. Nevertheless, for unavoidable reasons 57 patients were lost to the experiment at various times during the 10 weeks. Nineteen patients were given electroshock treatments or medication when they did not improve; 14 refused to take the "drugs"; 9 were discharged without the approval of the hospital; 8 were released on trial visits; 5 were discharged with approval; 1 was transferred to another hospital; and 1 developed a physical illness. Consequently, to obtain 120 patients it was necessary to draw from a larger sample of 177 patients. This attrition rate is slightly higher than but quite comparable to that reported in a recent well-designed large-scale Veterans Administration chemotherapy study. The possibility that the drop-outs may have biased the results of the study is considered in the statistical analysis.

Uncontrolled Method of Evaluating Treatment Questionnaire.-The patients in the "drug" groups, their psychiatrists, and the nurses completed a simple questionnaire which inquired whether the medicine appeared to have "helped a great deal," "helped quite a bit," "helped a little," "made no difference," or "made the patient worse." Space was provided for comments, and each week during the treatment period the reply was returned in a sealed envelope to the clinical director of the hospital. Admittedly the phrasing of this questionnaire is open to criticism, but it was deliberately selected because it conformed to the evaluation protocol employed in many uncontrolled drug studies.

Statistical Analysis.-The method used to classify

help, then the case was assigned to no help; or more weekly reports indicated that the pa was worse, and during no other week was be ported helped, he was considered worse, if at leas one weekly report indicated help from the "dr then the patient was classified as helped, and cording to the average extent of the help report (a great deal, quite a bit, or a little), he was & signed to the appropriate help category, t report indicated that during one week a patie was helped, but during another week he was wird then he was classified according to the aver over-all effect; a patient was considered to ha received questionable help when there was at i one week of improvement, accompanied by expression of uncertainty as to whether the "dr.. itself was responsible for the change. Six sens psychiatrists, 14 psychiatric residents, and 40 p chiatric nurses participated in the experiment replies of the patients, psychiatrists, and use to the questionnaire are presented in Table

and 2.

id to have been helped by the placebos for an erage of 2.6 to 4.9 of the 6 weeks they were king them.

In the tranquilizer study the 3 groups of inrmants attributed the following effects to the icebo: Patients-less tense, more relaxed (10); phoria, well-being (4); improved sleep (1); retance to fatigue (1); insight (1); improved psoria(1); effect not specified (4); Psychiatrists-more mfortable, less anxious (9); improved mood, less. athetic or flat (7); more energetic, sociable, outing (6); less hostile, more cooperative (4); imwed sleep (2); improved psoriasis (1); effect not cified (3); Nurses-more sociable (13); improved rits (7); more relaxed (6); less hostile, more coerative (3); improved sleep (3); effect not speci1 (4). Six patients (20%), including 2 who denied efiting from the "drug," attributed the follow; side-effects to it: drowsiness, lethargy (3); adache (2); blurred vision (2); insomnia (1); diara (1); polyuria (1); palpitation (1); loss of appe(1).

The nature of the improvement attributed to the rgizer "drug" was as follows: Patients-more xed, less tense (6); sense of well-being, cheerness (3); energy (2); improved appetite (1); oroved digestion (1); effect not specified (5); chiatrists-less depressed, more cheerful (6); re sociable, communicative (6); interest, initiative energy, activity (3); less tense, more relaxed less hypochondriacal (2); less preoccupied (1); re alert (1); improved attitude (1); improved p (1); improved appetite (1); weight gain (1); ct not specified (1); Nurses-more sociable (14); tense (7); more interest (4); less depressed (3); re energetic (2); more alert (2); more affect (1); preoccupied (1); improved sleep (1); improved etite (1); effect not specified (5). Six patients %) who received the energizer, including 2 who med no help from it, reported the following -effects: headache (2); drowsiness (1); jitteri5 (1); elation (1); palpitation (1); epigastric diss (1); insomnia (1); anger (1); morbid thoughts (1). f the present experiments are regarded as 2 ontrolled drug studies, the results are no less ressive than those reported in many similar lies in the literature. There is little doubt that e value of the 2 placebos had been enhanced some such appellations as Quietil and Moodex, ⚫ might soon constitute, as clinical investigators prone to say, "welcome additions to the clinical amentarium."

Controlled Method of Evaluating Treatment ating Scales.-Weekly ratings were recorded for "drug" and control patients by the head nurse ach hall on a specially constructed behavioral ig scale. This scale was developed and revised the assistance of the nurses, and framed in uage that was familiar and meaningful to

them. Separate scores were obtained for 9 different behavioral traits or "target symptoms": cooperation, depression, restlessness, irritability, tension or nervousness, energy and drive, preoccupation with health, sociability, and interest. The scale also provided a total score for over-all behavior, based on all 9 "target symptoms." To determine the reliability of the scales independent ratings were made by pairs of nurses, and inter-rater agreement was measured by intraclass correlation coefficients based on pre-treatment ratings of the 120 patients. The reliability coefficient for the measure of overall behavior was 0.79, and the coefficients for the various "target symptoms" ranged from 0.50 to 0.80. Copies of the rating scale and more detailed psychometric data are available to interested readers.

Statistical Analysis.-According to the format of a controlled drug study, in both experiments the "drug" and control groups were compared on the 10 measures obtained from the nurses' behavioral rating scale. The statistical model employed was analysis of covariance. 2-factor "mixed" type." Comparisons were made between "drug" and control groups; the significance of the difference be tween the trend of the means of the 2 groups for weeks 3 to 8, inclusive, was tested, with use of the scores for the pretreatment period (weeks 1 and 2) to adjust the means of the two groups to equal starting levels. A second set of comparisons was made between "drug" and control groups for the post-treatment period (weeks 9 and 10), again with the pretreatment scores used to adjust the means of the 2 groups to equal starting levels. Thus, in each of the 2 experiments a total of 20 comparisons were made. The 5% level of statistical significance was employed.

Results of the Controlled Studies

The design of the controlled experiments made it possible to evaluate the placebo effect resulting from patient and observer suggestibility (Definitions 1 and 2), since behavioral measurements were made on matched control groups as well as on the "drug" (placebo) groups. In the tranquilizer experiment, during the 6-week treatment period the "drug" group improved significantly more than the control group on over-all behavior and on 2 of the 9 "target symptoms," depression and restlessness. There were no significant differences between the "drug" and control groups during the post-treatment period. In the energizer experiment none of the "drug" and control comparisons yielded significant differences. Similar comparisons between the dropouts in the "drug" and control groups failed to reveal any significant differences, indicating that no differential bias was introduced by the dropouts into the "drug" and control-group comparisons. Tables 3 and 4 contain the probability values of the F ratios which formed the basis of the "drug" and control-group comparisons.

in explanation of this difference in findings. It could be a reflection of the fact that tranquilizers have been much more readily accepted than energizers or antidepressants; this was especially true at the time that the present investigation was under way. This might have been a factor in minimizing placebo effect, although it does not appear to have affected the replies to the questionnaire. The more conservative dosage regimen in the energizer study and the differences in the nature of the illnesses studied in the 2 experiments are other possible explanations. It should be noted that placebo and

that all significant differences occurred during the second, third, and fourth weeks of treatment. The suggests a 1-week lag in onset of placebo effect the presence of the phenomenon for 3 weeks, fol lowed by its disappearance during the final 2 weeks of treatment. The experiment provides no answer as to whether the effect would have returned treatment had continued beyond 6 weeks b should be noted that these data refer to the place bo effect arising from patient or observer suggest bility, or both (Definitions 1 and 2). A contaminated placebo effect (Definition 3) was operative in the data reported in Tables 1 and 2. When the latter were analyzed (Chi-square), no such temporal phe nomenon occurred. For example, the psychiatrists attributed benefit to the placebos for an average of about 3 of the 6 weeks of treatment; however such benefit was as likely to occur during weeks 4, 5, and 6 as it was during weeks 2, 3, and 4 or 1, 4, and 6.

Comment and Conclusions

Doubtless, the placebo effects reported in Tables 1 and 2 reflect the influence of all 3 uses of the term. Obviously, any contaminated placebo c (Definition 3) was in part a function of what eac observer estimated the course of events would have been if no "drug" had been introduced. The observers had to recall how these patients or s lar patients had fared over a comparable 6-we period without somatotherapy, with treatment limited to psychotherapy, occupational and recrea tional therapies, and the protective hospital vironment. This so-called historical control charac terizes every uncontrolled drug study. It involves a series of subjective judgments which are eas influenced by one's general attitude toward the use of medication, the conscious or unconsci need to please (or displease) the research invest gators, and other personal variables.

The results of the present experiment make i clear that the historical control, when emploves with acutely ill psychiatric patients to eval new medications, is not only a poor control t is really no control at all. Proponents of c trolled studies can find little consolation fact that a "pure" placebo effect apparent not occur in the energizer study, and proved be a temporary phenomenon in the tras experiment. When placebo effect was c nated by the improvement due to extraneos "spontaneous" events (Definition 3), it appen in both tranquilizer and energizer express and displayed no temporal course whers The data do not permit us to generalize cally ill patients, in whom "spontaneous" vara and remissions are less common; however would venture the opinion that any investig who would take comfort in this fact has bee lulled into a state of tranquility exceeding t