A Brief History of Placebos and Clinical Trials in Psychiatry

Abstract

The history of placebos in psychiatry can be understood only in the context of randomized controlled trials (RCTs). Placebo treatments are as old as medicine itself, and are particularly effective in dealing with psychosomatic symptoms. In psychiatry, placebos have mainly been featured in clinical drug trials. The earliest controlled trial in psychiatry (not involving drugs) occurred in 1922, followed by the first crossover studies during the 1930s. Meanwhile the concept of randomization was developed during the interwar years by British statistician Ronald A Fisher, and introduced in 3 trials of tuberculosis drugs between 1947 and 1951. These classic studies established the RCT as the gold standard in pharmaceutical trials, and its status was cemented during the mid-1950s. Nevertheless, while the placebo became established as a standard measure of drug action, placebo treatments became stigmatized as unethical. This is unfortunate, as they constitute one of the most powerful therapies in psychiatry. In recent years, moreover, the dogma of the placebo-controlled trial as the only acceptable data for drug licensing is also being increasingly discredited. This backlash has had 2 sources: one is the recognition that the US Food and Drug Administration has been too lax in permitting trials controlled with placebos alone, rather than also using an active agent as a test of comparative efficacy. In addition, there is evidence that in the hands of the pharmaceutical industry, the scientific integrity of RCTs themselves has been degraded into a marketing device. The once-powerful placebo is thus threatened with extinction.

It is possible to understand the history of placebo treatments in psychiatry only in the context of controlled trials. Placebos have always been used clinically, though in the Freudian era agents of any kind were frowned on, whether placebo or active. The most intriguing aspect of placebos in clinical trials in psychiatry, however, is the extent to which RCTs themselves have shifted from a positive to a negative valence.

Placebo treatment has been known in medicine—and in psychiatry—since time out of mind, given that placebos are most effective in treating symptoms of psychological origin, or somatoform illnesses. As John T G Nichols¹ of Harvard University wrote in 1893,

The average patient listens with much more interest to the prescription of his physician than to his directions about hygiene. Expecting good results from the drug, he often imagines that he feels them. So great is the power of hope that, even in incurable diseases, a temporary improvement often follows each new prescription. This power of hope … is sometimes used by the educated physician, who calls it ‘expectant attention.’^{p 28}

Nichols¹ said pills that aided expectant attention were sometimes referred to by “pharmaceutists” as “blank tablets,” citing the example of “Harvard Experimental Diagnostic Tablets”^{p 29} which had recently done big business. Blank tablets were another term for placebo, and they are as old as medical practice itself.

But how about in psychiatry?

In psychiatry, placebo has played a role mainly in the clinical trials of new agents. The mind has always been vulnerable to the influence of suggestion,² in psychiatry as in other fields. To be sure, major mental illnesses such as psychotic depression are quite unresponsive to placebo treatment.³ Nevertheless, the rest of the disorders are fair game. Inevitably, many psychopharmacologic agents used in psychiatry before the Second World War had significant placebo components. Nevertheless, we have no way of knowing that. What we do know about is the systematic use of placebos in psychiatric drugs trials.

The first controlled trial in psychiatry seems to have occurred in 1922, when Nicholas Kopeloff, a bacteriologist at the New York State Psychiatric Institute, then on Ward’s Island, and Clarence O Cheney, a staff psychiatrist, attempted to evaluate the popular theory of focal infection as a cause of major psychiatric illness.⁴ Focal infection meant infections of the teeth and tonsils, and so firmly implanted was the view that toxins from such foci caused dementia praecox and manic-depressive illness that it had become standard at some mental hospitals to remove patients’ teeth and tonsils to relieve the symptoms of melancholia and chronic psychosis. Kopeloff and Cheney⁴ divided their 60 infected patients into a control group of 33, who received no operation (the placebo group) and 27 operated patients. There was no difference in therapeutic outcome: “The removal of infected teeth and tonsils from 27 cases … has been followed by no more mental benefit than was shown by a comparable group of 33 patients from whom such supposed foci of infection were not removed.”^{4, p 156} The patients were not randomized.⁴

The pieces for what we would recognize as a modern RCT in psychiatry were starting to fall into place. Was benzedrine or ephedrine more effective in the treatment of narcolepsy? In 1935 Myron Prinzmetal, a Los Angeles internist, and Wilfred Bloomberg, a Boston neuropsychiatrist, undertook a trial in 9 in- and outpatients, starting all on a sodium chloride placebo, then giving all benzedrine, followed by ephedrine, followed again finally by benzedrine. The patients were blinded, as all 3 substances have a similar salty taste. The responses to sodium chloride and ephedrine were equally weak. Benzedrine, by contrast, “afforded complete relief from symptoms.”^{5, p 2052} The patients were not randomized and served as their own control subjects (a crossover study), yet of the results of the trial there could be no doubt.⁵

Meanwhile, pharmacologist Harry Gold at the Cornell University Medical College, together with 2 cardiologist colleagues, was studying the effect of the xanthines (theobromide and aminophylline) on cardiac pain. In a study beginning in 1932—and reported in 1937⁶—in 209 courses of treatment they alternated the xanthines and a placebo (or “some other agent”^{p 2178}). The patients were blinded and served as their own control subjects. The results showed the xanthines to be ineffective:

Patients with cardiac pain are unable to distinguish the effects of a placebo from those of a xanthine [when blinded]. It is concluded that the xanthines exert no specific action which is useful in the routine treatment of cardiac pain.^{6, p 2178}

Although not a psychiatrist, Gold is often seen as the progenitor of placebo therapy in American medicine, which is inexact in view of previous psychiatric contributions. Nevertheless, he did contribute significantly to a psychiatric discussion after the Second World War.

By the beginning of the Second World War, such crossover studies alternating drug and placebo had become common. In 1938, Leonard A Dub and Louis A Lurie at the Longview State Hospital in Cincinnati, following Gold’s example, introduced the doubly blinded trial into psychiatry. The investigators used the patients as their own control subjects, in a study of the effectiveness of benzedrine in depression. Forty-eight female inpatients, who had been depressed (but had varying diagnoses) from between 10 months to 26 years, were given a trial of benzedrine, substituted at intervals with a lactose placebo; “86 per cent of the group under observation showed a marked improvement when benzedrine was administered.”^{7, p 39}

Thus, by the Second World War, the use of placebo was well established in psychiatric trials, and the blinding of patients and trialists was familiar. Not yet established was the concept of the randomization of patients to parallel treatment groups, important to avoid treatment selection bias.

After the war, the term placebo entered the common parlance of medicine. At a 1946 “Conference on Therapy” sponsored by the Departments of Pharmacology and Medicine of Cornell University Medical College, Eugene F DuBois, a New York physiologist, pointed out that, “although placebos are scarcely mentioned in the literature, they are administered more than any other group of drugs … Although few doctors admit that they give placebos, there is a placebo ingredient in practically every prescription.”^{8, p 1718} He noted that in fact, “the placebo is a potent agent and in its actions can resemble almost any drug.”^{8, p 1718} At the conference, Harry Gold added, “The placebo is a specific psychotherapeutic device with values of its own.”^{8, p 1723} Gold had reviewed some of his patients’ charts for effects attributed to medication that were in fact owing to a placebo: “Here they are: I sleep better; my appetite is improved; my breathing is better … I can walk further without pain in the chest; my nerves are more steady.”^{8, p 1724} Even though he himself was not a psychiatrist, Gold affirmed a role for the placebo in psychiatry: “I think the placebo as a chemical device for psychotherapy has a definite place which cannot be filled by anything else in many cases.”^{8, p 1724} Henry Richardson, a psychiatrist at the Peter Bent Brigham Hospital in Boston, pointed out the dangers of the clinical use of placebos in psychiatry: “I don’t like the element of deception in a placebo, apart from the fact that it is disastrous to get found out. I think that if there is to be a deception it should be the one which the patient demands and which, also, I think he needs.”^{8, p 1725} The diversion of the discussion, from Gold’s interest in placebos in trials to Richardson’s in clinical use, was prescient: in trials, the placebo was an instrument of the science of measuring drug action; in the clinic, it was an instrument of deception with a negative moral valence. In psychiatry, this tension between trial science and clinical use would first grow, and then in our own time reverse.

Meanwhile in Britain, events were marching toward the introduction of randomization in controlled trials, which would result in the term RCT. This powder train began in 1925, when English statistician Ronald Aylmer Fisher published Statistical Methods for Research Workers,⁹ followed in 1935 by The Design of Experiments,¹⁰ which began Fisher’s decades-long efforts to provide a statistical basis for random samples of the population. Randomization was first used in trials of a chemotherapeutic agent in England during the period from 1947 to 1951 as Marc Daniels, a member of the Tuberculosis Research Unit of the Medical Research Council, and A S Bradford Hill, Professor of Medical Statistics at the London School of Hygiene and Tropical Medicine, conducted 3 trials of various tuberculosis drugs, randomizing the patients to different treatment groups (not to a placebo). This epochal triad of trials introduced randomization as the gold standard in pharmaceutical trials.¹¹

There is little placebo effect in the treatment of tuberculosis. In psychiatry, however, a huge placebo effect often exists, and here the new clinical pharmacology meant randomizing patients to placebo to counter the effect of suggestion (and to compensate for the spontaneous recovery that many patients might experience in the course of a trial). The first randomized psychiatric trials occurred in North America and Denmark, and the former have remained largely ignored in writing on the history of clinical trials.

With a flip of the coin in 1952, Mogens Schou,¹² staff psychiatrist at the Aarhus University Psychiatric Institute in Risskov, Denmark, initiated the first randomized clinical trial in psychiatry, employing lithium on 38 manic patients (For Schou’s recollection of the timing, see his account in Healy.¹² Schou et al¹³ submitted the manuscript in 1954.¹²) A portion of these patients were enrolled in a proper doubly blinded RCT. “Every two weeks the medication was shifted in a random manner from lithium to placebo and vice versa, and changes from lithium to lithium and from placebo to placebo were also used.^{13, p 250} The outcome of the trial was stunningly in favour of lithium, establishing it as the preferred treatment for mania.¹³

Louis Lasagna, who was to become the dean of American pharmacology, was right on Schou’s heels. At some point in 1952 or 1953, Lasagna initiated a clinical trial of 2 standard hypnotic agents (chloral hydrate and pentobarbital sodium) vis-à-vis the recently introduced agent methylparafynol (Dormison) and placebo. Assignment to all 4 arms of the doubly blinded trial was randomly determined. Submitted to a journal in November 1953 and published in 1954,¹⁴ this may be considered the first published RCT in psychiatry, even though Lasagna was fellowing in the Anesthesia Laboratory of the Harvard Medical School at the Massachusetts General Hospital.

In 1954, a group of investigators at the Saskatchewan Hospital in Weyburn undertook a randomized, placebo-controlled study of various nucleotides (such as adenylic acid) in the treatment of schizophrenia. This was the same group, including Abram Hoffer, Humphrey Osmond, and John Smythies, that shortly thereafter embarked on a clinical study of lysergic acid diethylamide (commonly referred to as LSD). As for the nucleotide study, “We were advised that a series of 40 would be statistically valid, half of these patients being controls.”^{15, p 149} They also were among the first investigators to use a rating scale in evaluating end points. In this particular case the results were negative: nucleotide treatment contributed nothing to outcomes in schizophrenia.¹⁵

The above North American studies are today little known. Even less known is a doubly blinded randomly controlled placebo trial that John L Hampson, David Rosenthal, and Jerome Frank¹⁶ conducted in 1954 at the Continued Treatment Clinic—given over mainly to outpatient psychotherapy—of the Department of Psychiatry at Johns Hopkins University. They evaluated the new muscle relaxant and potential tranquilizer mephenesin (the forerunner of meprobamate). The findings were negative, in that the patients got better on both drug and placebo, and set the authors to musing about the “placebo effect.”^{16, p 176} “The high value which our culture places on pills and medicines may be involved in this phenomenon whereby even inert substances become endowed with physiologic potency when they are presented to the patient as therapeutic agents.”^{16, p 176}

In the United Kingdom, David Lewis Davies, Dean of the Institute of Psychiatry at the Maudsley Hospital, and Michael Shepherd, then a senior registrar, conducted the first RCT. In The Lancet, on July 16, 1955, Davies and Shepherd¹⁷ described their trial of reserpine “in the treatment of anxious and depressed patients,”^{p 117} who had been randomized to a treatment group and “dummy tablets,”^{p 117} the investigators and the patients both blinded. Shepherd is considered among the pioneers of British psychopharmacology (despite his scornful disbelief in the efficacy of lithium in mania), and this trial lighted a pathway to the future.¹⁷ (W Linford Rees at the Royal Bethlem and Maudsley Hospitals, also considered a pioneer trialist, did not randomize the patients in a 1955 trial of chlorpromazine for anxiety [see Rees and Lambert¹⁸].)

This round of RCTs on both sides of the Atlantic in the mid-1950s began the sanctification of the RCT that continued for decades: RCTs with placebos as the gold standard of clinical pharmacology. By contrast, the stigmatization of the placebo in clinical practice that we observed after the Second World War became even stronger with the Helsinki Declaration of 1964, which elevated to a principle of absolute moral certainty that one must not lie to patients in research studies. This doctrine inevitably migrated transversely to clinical practice, where lying to patients became equally frowned on, thus disqualifying the use of placebos in everyday medicine as well.¹⁹

What has been interesting, however, has been the dulling of the shine on the RCT that has taken place in the past decade. This has had 2 sources. One is the awareness that the US FDA has been too lax in permitting trials controlled with placebos alone, rather than having as a third arm the standard agent. Controlling trials with placebos alone means one is unable to determine if there has been progress in the efficacy of agents: placebo controls are mute as to comparative efficacy. The penny here has been rather slow to drop, but now awareness has dawned that the neuropsychopharmacology branch of the FDA may have presided over a decline in the effectiveness of new agents, rather than enabling a continuing rise.²⁰ Irving Kirsch, a psychologist at the University of Connecticut, has called attention to the lack of efficacy demonstrated in clinical trials of the SSRI agents in particular, the so-called antidepressants that, whatever benefits they may have, do not seem to be effective in serious depression (see Kirsch et al²¹). These defects at the FDA have taken some of the sheen off the RCTs.

Moreover, the whole dogma of placebo-controlled trials as the only acceptable data for licensing is increasingly being called into question. Lasagna²² noted that the big psychiatric drugs of the 1950s, such as chlorpromazine and imipramine, had been developed with “historical controls,”^{p 136} not RCTs, meaning clinical observation of patients whose improvement was undeniable and as plain to see as the betterment of patients on penicillin. In the 1990s, he told an interviewer of the viewpoint he had held in the 1950s: “So here I was on the one hand saying, well, once you identify a compound, if you want to persuade people that it works you really ought to do controlled trials, but having to face the fact that the breakthroughs were not achieved in that way.”^{22, p 137} Elsewhere, Lasagna²³ noted the utility of the unsystematic open trials of an earlier era: “At least … such trials could test the ability of the practicing physician to use a drug with satisfaction, to himself and to his patient.”^{p 21}

So for truly effective drugs, RCTs are probably not indispensable to licensing, especially given the staggering costs of such trials today; if the drug clearly works, vast and expensive trials would seem unnecessary, a way of keeping smaller players who cannot bankroll such trials out of the game.

Other scholars have argued that, in the hands of the pharmaceutical industry, RCTs themselves have become poisoned at the source. David Healy, Professor of Psychiatry at Cardiff University, has documented the degradation of clinical trials to a means of providing copy for pharmaceutical marketing. He notes that using P levels as measures of statistical significance virtually guarantees that a large trial will churn up significant findings, though the differences between drug and placebo may be clinically meaningless. Rating scales often obscure the heterogeneity within clinical populations, vastly expanding the potential market of a given agent (one thinks of the dubious term major depression). The proliferating so-called clinical guidelines, based on the findings of corrupted RCTs, become sluices for pharmaceutical sales rather than bulwarks of scientific integrity (which, I argue, they once were). Healy²⁴ writes:

In the absence of compelling evidence, the erection of guidelines that advocate one set of agents over another, however well meaning, leaves guideline makers open to being captured. Through a combination of apparently novel indications and publication strategies, companies can make diseases fashionable, engineer the appearances of comparative efficacy and enlist academic advocates for particular treatment options.^{p 31}

Healy²⁴ calls this process “guideline capture.”^{p 28}

The poor old placebo, dragged down by bioethical strictures and the corruption of clinical trials that are randomly controlled with placebos, is thus threatened with extinction. This is a shame, in view of its being one of the most powerful treatments that psychiatry has to offer.

Highlights.

• Placebos have always been used in medicine for their effectiveness in treating psychosomatic symptoms, yet, following the Second World War, this valuable psychiatric treatment was banished from clinical use as an instrument of deception.

• The main use of the placebo in modern psychiatry has thus been in the clinical drug trials, particularly with the development of the placebo-controlled RCT. The most intriguing development in the use of placebos for this purpose has been the extent to which the RCT, originally enshrined as the gold standard of clinical pharmacology, has been increasingly discredited in recent decades. This backlash has had 2 sources: growing awareness of the lax standards of the US Food and Drug Administration in mandating trials controlled by placebo alone, rather than also using the standard agent as a test of comparative efficacy; and growing evidence that the scientific integrity of RCTs themselves has been corrupted in the hands of the pharmaceutical industry.

Abbreviations

FDA

Food and Drug Administration

RCT

randomized controlled trial