How important are placebo effects? In 1955, when randomized trials and, with them, placebo pills, were just coming into widespread use in research, an anesthesiologist named Henry Beecher gave a high estimate. In “The Powerful Placebo,” an essay in JAMA, Beecher reported that, for conditions ranging from post-operative pain to the common cold, the response rate to placebo hovers at 35 percent, close to what many useful drugs offer. In the wake of Beecher’s article, “placebo effect” became a commonplace shorthand in daily speech for the belief that upbeat expectations can work wonders for our health.
But skeptics called each of Beecher’s examples into question. Finally, in 1997, noting sources of error such as misreported data, critics concluded that, in the cases Beecher cited, “most likely there was no placebo effect whatsoever.”
It is tempting to say that this oscillation — between strong claims for placebo and thorough dismissal — is unending, but that’s not quite true. Researchers have largely settled on the conclusion that placebo effects play only a minor role in medical care — although in certain areas the impact of expectancy can be impressive.
In 1972, in my first year of medical school, I saw the placebo effect, or something like it, in action. To illustrate Parkinson’s disease, a lecturer in neurology brought a patient into the lecture hall. The poor man stood before us rigidly motionless but for head shaking and hand tremor, which are hallmarks of the disease. From the blackboard tray, the teacher took a piece of chalk. Holding it in his outstretched hand, he asked the patient to come and grasp it. The man strained to comply but stayed rooted in place
Now the neurologist stood beside the patient and offered encouragement. “Let’s go!” The doctor began walking. So did the patient, haltingly. “Come on! Faster!” The patient began jogging around the laboratory table. Soon he was running athletically.
Parkinson’s, our teacher explained, begins as a defect in the ability to initiate motion. The sufferer wants to act, but his brain cannot supply the oomph to get him started. This difficulty is specific to Parkinson’s. Similar-looking movement disorders — one is called frontotemporal dementia — do not respond to cheerleading.
What the professor demonstrated was not exactly a placebo effect — there was no sham treatment — but dummy pills might well have been able to inspire a similar improvement in the patient’s mobility.
By the early ’70s, it was known that the neurotransmitter dopamine is deficient in the brains of Parkinson’s patients. Treatment had, not long before, been transformed by the introduction of L-DOPA, a medicine that enhances dopamine production. L-DOPA is the drug whose effects Oliver Sacks later depicted in his book Awakenings. He had given it to patients afflicted by sleepy sickness, an extreme form of Parkinsonism, and it had unfrozen them, at least for a while.
Wondering how encouragement achieves parallel effects, researchers turned to PET scans, which track chemical action in the brain. In 2001, scientists demonstrated that sham treatment — injections of saltwater — can cause patients (led to believe that they are receiving medication) to squeeze dopamine from failing neurons. Encouragement works in Parkinson’s disease because of a highly particular balance between disability and a residual capacity to overcome it.
In medicine, the stars rarely align in this favorable fashion.
Fifteen years ago, a pair of Danish researchers, Asbjørn Hróbjartsson and Peter C. Gøtzsche, published a breakthrough study in the New England Journal of Medicine. The title “Is the Placebo Powerless?” identified the paper as a further answer to Beecher. Seeking to put placebo to the test required of active drugs, Hróbjartsson and Gøtzsche had asked whether, in randomized trials, sham treatments could outperform a lesser intervention, like time on a waitlist.
For pain relief, the answer was yes. As with Parkinsonism, the pain paradigm is partly worked out. In the right context, dummy pills can cause the brain to release analgesic molecules, including endorphins. Some of this response is “conditioned,” that is, based on past responses to inherently effective drugs. Because real morphine worked then, sham morphine works now. (Similarly, in the Parkinson’s experiment the participants had previously taken L-DOPA.) In part, improvement on sham treatment is a bodily tribute to the successes of medication.
Beyond pain relief, the mere fact of pill taking — or of interventions like sham acupuncture — rarely adds benefit. Hróbjartsson and Gøtzsche included studies of disorders ranging from anemia to epilepsy. When the outcome measured was objective — for example, blood sugar levels — placebo contributed nothing. Nor did sham treatment help for subjective outcomes, like joint flexibility, that doctors could observe directly. With all-or-none outcomes — such as smoking or managing to abstain — placebo never outperformed placement on a waitlist.
Only for patient-reported subjective outcomes that occur on a spectrum — nausea is an example — might placebo do better than “no treatment,” and those data, in Hróbjartsson’s and Gøtzsche’s view, were suspect: the statistical pattern suggested that trials with contrary outcomes remained unpublished. Also, Hróbjartsson and Gøtzsche speculated that patient-reported symptoms were subject to “goodbye effects,” the tendency for participants to report progress in order to satisfy caregivers’ expectations.
The Cochrane Collaboration, an international consortium of scholars, serves as an arbiter of “evidence-based medicine,” the movement that champions high-quality research as a basis for clinical practice. In 2010, the organization commissioned Hróbjartsson and Gøtzsche to update their overview of “placebo interventions for all clinical conditions.” The verdict stood: “no major health benefits” outside pain management.
Ted Kaptchuk, arguably our premier researcher on placebos, confirmed the “powerless” finding. In a 2015 review article for the New England Journal of Medicine, he wrote that placebos “primarily address subjective and self-appraised symptoms.” Overall, responses are modest: “though placebos may provide relief, they rarely cure.” Kaptchuk’s own research provides a case in point. Studying asthma, he has shown that with sham acupuncture or dummy inhalers patients may report an impression of improvement. Still, their lungs move no more air.
The contemporary understanding of placebo casts a shadow on the new anthology, Placebo Talks: Modern Perspectives on Placebos in Society. The editors’ goal is to extend the concept of the placebo into the social sciences. Contributors relate the placebo effect to our notions of flavor and diet, fetishism and sexual imagery, suggestion and false memory, politics and public values, and choice and decision-making in general. The essays explore ways in which context and expectation influence judgment. But is placebo the right metaphor for this concept?
In concrete terms, Placebo Talks reflects the prevailing consensus — limited efficacy for placebo. The index lists only four non-psychiatric diseases: Parkinson’s, neuropathic itch, irritable bowel syndrome, and peptic ulcer. The text refers also to migraine and to a form of heart disease, obstructive hypertrophic cardiomyopathy — but, for both, pain is a leading symptom.
Given its diminished range, how has placebo become a stand-in for general beliefs about the power of the mind to heal or deceive itself? My impression, reinforced by the book at hand, is that the revival comes via excitement over the claim that antidepressants are little more than “placebos with side effects.”
The prime mover in this drama is the psychologist Irving Kirsch, now a lecturer in Harvard’s Program in Placebo Studies. Early in his career, Kirsch was an innovator in the theory of the placebo. In the 1990s, he turned his attention to antidepressants and became a gadfly, producing debunking analyses that, while subject to criticism, had the salutary effect of pushing the mental health professions to undertake clarifying research. Unfortunately, the results have not made their way into Kirsch’s repertory. His contribution to the volume at hand is a stump speech larded with zombie arguments, contentions long since refuted that will not die.
Kirsch’s line on antidepressants has always had a weak point. He approaches the data with idiosyncratic statistical methods — reanalyses by other scientists often show stronger drug superiority — but even in Kirsch’s overviews, medication outperforms placebo. To dismiss this remaining edge, Kirsch invokes what he calls “nonantidepressant drugs,” medicines that ought not to treat depression but that — because they produce convincing side effects — seemingly do. Since irrelevant pills match antidepressants, the last bit of medication efficacy (so Kirsch reasons) may be placebo action, too.
But the list of nonantidepressant drugs that Kirsch provides now — “lithium, barbiturates, and thyroid medication” — comes from a 1998 paper that had been thoroughly dismissed by critics at the time. In the study of barbiturate that Kirsch cited there, the drug, an old-fashioned sedative, had failed to show efficacy, while a full dose of Elavil (an antidepressant) had outperformed placebo significantly. As for the lithium and thyroid hormone, they had proved helpful when they were given along with a full dose of antidepressant. Nonetheless, Kirsch repeated the litany of drugs in writings from 2000, 2005, and 2009 — and now in 2016.
Kirsch climaxes the current essay with a discussion of a medicine called tianeptine. When first tested in the 1980s, tianeptine was found to impede serotonin-based transmission in the brain. Kirsch writes that scientists’ faith in antidepressants is grounded in the “chemical imbalance” theory of mental illness, the idea that antidepressants restore harmony by making the brain’s use of transmitters like serotonin and norepinephrine more efficient. How, then, asks Kirsch, can tianeptine, which pushes the opposite way, also treat depression? Kirsch calls tianeptine “the last nail in the coffin,” the deceased being the notion that antidepressants work via their pharmacologic influence.
But scientists had continued to elucidate tianeptine’s mechanism of action. Tianaptine may lower serotonin use in one part of the brain, but it increases it in others. Administered for 14 days — the conventional estimate of the interval required for substantial antidepressant action — tianeptine, like Zoloft or Prozac, slows the breakdown of serotonin in the brain more generally. And tianeptine increases norepinephrine levels or slows norepinephrine turnover in a half dozen brain centers. As regards neurotransmission, tianeptine resembles the antidepressants of its era, drugs (like Elavil) that make the brain’s use of norepinephrine more efficient, with lesser or later effects on the handling of serotonin — also mostly in the expected direction.
Meanwhile, the chemical imbalance theory has become a straw man. Years ago, it was supplanted by the neuroplasticity hypothesis, the notion that antidepressants restore resilience in the brain, protecting against cell death while restoring the brain’s capacity (impaired in depression) to grow new cells and elaborate new connections between cells. Tianeptine is typical in this regard.
Like the invocation of thyroid and lithium as nonantidepressants, the tianeptine argument is a revenant. Kirsch premiered it in 2009, citing three scientific papers for support. They contain virtually every finding I have just reviewed, about tianeptine’s standard effects on neurotransmission and neuroplasticity.
What makes placebo seductive for social scientists is the notion that medications in widespread use — antidepressants — work largely through hopeful expectancy. That premise is at best controversial and most likely (I would say certainly) wrong.
In practice, the impact of placebo on patient care may be minimal. Let’s think again about Parkinson’s disease. In the 2001 study involving PET scans, the research subjects on the active drug, unlike those on saltwater, did not need to force the last drop of dopamine from dying neurons. Patients on real medication responded because the medication increased brain dopamine fairly directly.
In the experiment, the researchers found placebo effects only for the patients getting saline injections. Medicated patients did equally well whether the drug was given openly or covertly. The medication preempted or overwhelmed any potential impact of hopeful expectations.
This result matches the clinical impression. Learning that placebo can unfreeze Parkinsonian patients does nothing to shake doctors’ conviction that L-DOPA and similar drugs work via their chemical action on the brain. The technical term related to this paradox is failed additivity. Because expectancy adds nothing to the impact of L-DOPA, if we subtract the progress patients make on dummy injections from the progress patients make on medication, we will underestimate the drug’s inherent powers. The PET-scan researchers counseled readers that “in the usual clinical practice setting, active drugs may be devoid of placebo effect.” What neurologists see — an effective drug — is what patients get.
With depression, the additivity question is less cut and dried, but the set-up may be similar. In a recent small trial at UCLA, patients’ stated expectations predicted their response to placebo but not their response to antidepressants. Drug effects, for good and ill, swamped any influence deriving from patients’ beliefs. Results of brain-action measurement in depression are complex and sometimes contradictory, but prior brain-wave work from the UCLA lab had suggested that — as with L-DOPA in parkinsonism — patients on antidepressants do not experience an added placebo response.
To his credit, in various professional papers Kirsch has reminded colleagues of this potential complication:
If antidepressant drug effects and antidepressant placebo effects are not additive, the ameliorating effects of antidepressants might be obtained even if patients did not know the drug was being administered. If that is the case, then antidepressant drugs have substantial pharmacologic effects that are duplicated or masked by placebo.
Contrary to the general assumption, what remains uncertain is not whether antidepressants have meaningful effects on depression but whether, for patients on the medications, the placebo effect plays an important role at all. Certain researchers have argued — and many clinicians believe — that people who report immediate, very marked improvement on antidepressants may be responding to expectations. Some studies suggest that association, and others do not; but even where the pattern emerges in the data, the benefits — any substantial relief from depression — appear to be lasting in at best a small proportion of patients.
The notion that mood disorder is highly placebo-responsive comes largely from analyses of “candidate drug trials,” the tests of new medicines that pharmaceutical houses submit to the Food and Drug Administration. The experiments are rife with problems. Often, trials are conducted at for-profit centers; seeking to fill enrollment quotas, recruiters exaggerate participants’ symptoms at the outset — so that, later, patients seem to recover from problems they never had. Because the FDA disfavors trials with high dropout rates — or, worse, suicides — research centers offer participants extensive emotional support. Progress, real and apparent, in patients on placebo is attributable to a mix of factors — not only hopeful expectations but also implicit psychotherapy effects, distortions inherent to the testing process, the passage of time, and much more.
The research in Hróbjartsson’s and Gøtzsche’s collection suggests that, among the sources of change for depressed patients, faith in dummy pills is a minor contributor. A telling experiment involved research subjects who, in a brief outcome trial, had improved on placebo. They were then randomized to remain on placebo or be withdrawn. The two conditions produced equal and disappointing results — a quick return of depression, on or off pills. That result argues against robust expectancy effects.
If placebo influence is spotty and of uncertain medical significance, then extending the reach of the underlying concept becomes a dicey undertaking. It should come as no surprise that, in Placebo Talks, the chapters that address clinical practice are the least convincing.
Take, for instance, a commentary by Veronica de Jong and Amir Raz, of McGill University, on the tendency of psychiatrists, more than general practitioners, to prescribe low doses of antidepressants early in the course of treatment. With seeming confidence, de Jong and Raz declare, “The biologically active ingredients in quantities too low to exert therapeutic benefits — although they may still exert side-effects — render sub-therapeutic doses a type of active placebo.” The authors risk slipping into the posture of anthropologists studying a lesser culture: “clinical practice would stand to benefit from a greater appreciation among physicians and policy makers that at low, if not any, dose, antidepressants may be closer to active placebos than efficacious drugs.”
This critique has two problems. Low doses of antidepressants may not induce a placebo response, and they may help patients.
A 20-year survey conducted under the auspices of the National Institute of Mental Health found that only full doses of antidepressants offered protection or relief from depressive episodes. Lower doses did no better than “no treatment.” This outcome suggests that, in real patients, sub-therapeutic doses of antidepressant do not serve as active placebos. Mere pills, even ones with side effects, do not help.
All the same, low-dose treatment may have its uses. Philip Cowen, an Oxford psychiatrist, showed that after a single half-dose of an antidepressant (but not after administration of placebo) patients with mild and moderate levels of depression were better able to read positive emotions in photographs and better able to access their own pleasant memories. This normalization of “emotional processing” predicted later relief from the mood disorder. In Cowen’s view, much of the benefit of medication comes from transmuting a changed perspective into adaptive thoughts and behaviors.
If so, then, when prescribed by psychiatrists who offer psychotherapy, sub-therapeutic medication doses may spark recovery. The colleagues whom de Jong and Raz interviewed may have settled on a wise strategy: amplifying the effects of low-dose medicine when possible, relying on standard treatments where necessary, and retaining a prudent agnosticism about claims for placebo.
Similarly, Bennett Foddy, an ethicist, makes depression treatment the poster child for justifiable deception in patient care. If Kirsch is right, reasons Foddy, “all the competing pharmacologic treatments should be considered placebos” anyway. It follows that “depression provides the clearest case for allowing doctors intentionally and responsibly to prescribe placebos in the clinic” without explicit warning. Why depression, and not heart disease? With different examples, Foddy’s stance — letting doctors take liberties with patients’ trust — might still be disconcerting, but at least stigma would be out of the picture. Perhaps we are most open to benign deception applied to people who — so we think, deep down — suffer only from morbid imaginings.
The canard that antidepressant responses are placebo responses may cause two sorts of harm: depriving patients of needed treatment and arousing doubt about the gravity of their affliction.
As the anthology moves away from medical topics, the essays become more compelling.
Extending the theme of objects that elicit subjective responses, the historian Edward Shorter reviews variations in sexual fetishes over six centuries. Animal attraction is biological, but, despite what each generation may believe, the particulars of eroticism — physiology’s “overlays” — are cultural. The role that leather enjoys today was once held by fur, which Shorter calls “history’s first fetish.” In his 1870 novella Venus in Furs, Leopold von Sacher-Masoch takes his arousal by fetish to be natural and biological.
Michael Orsini and Paul Saurette, political theorists, focus on a hot issue: why citizens act against their apparent self-interest. Examining responses to the cancellation at the University of Ottawa of a talk by the right-wing commentator Ann Coulter, they link her supporters’ Tea Party–like outrage to Darwinist notions of distress in the face of social exclusion.
Cory Harris and Timothy Johns play with food. Context and imposed meaning determine congregants’ response to the communion wafer, whose healing properties are independent of nutritional considerations.
These reviews are great fun, but they derive little added value from calling the “bread” — or fur, or electoral politics — “placebo.”
Placebo effects occur and are worth studying. But increasingly “placebo,” detached from its defined applications, is a metaphor — for the importance of context, the unity of mind and body, the fallibility of perception, the mistrust of industry and professionals, the loss of personhood and meaning, and other popular positions and concepts. What occurs in this volume may be less the translation of medical science into social science than the application of a multipurpose trope in both.
No essay in this collection is without interest. We’re in the presence of imaginative minds. But too many contributions suffer from a complicit tone — they preach contentedly to the choir. Absent is any self-reflective curiosity, any inclination to apply sociology to our trending belief in the ubiquity of placebo effects. We’ve been acquainted with modern antidepressants — seen people benefit from them — for six decades. To my mind, the notion that the drugs work mostly through hopeful expectancy requires explanation on a cultural level. Surely there is significance in our eagerness to hear placebo speak.
Peter D. Kramer’s book on the antidepressant controversy, Ordinarily Well, will be published by Farrar, Straus and Giroux in June 2016. Kramer is the author of Listening to Prozac and a professor of psychiatry at Brown University.