This article originally appeared in the March 2003 Harvard Mental Health Letter and is provided courtesy of Harvard Health Publications.

Are antidepressants placebos?

A persistent controversy could affect the choice of psychiatric treatments.

In a familiar joke, a well-attended funeral service for a prominent member of the community is repeatedly disrupted by a voice from the back shouting, “Give him some chicken soup!” “Sir!” the master of ceremonies declares, after one too many interruptions, “Of what possible use could chicken soup be to the deceased?” Comes the reply: “It wouldn’t hurt him.”

In light of recently published articles suggesting that placebos are a fairly good treatment for depression, we may all wonder whether antidepressants are more effective than chicken soup. And given that antidepressants are more expensive than chicken soup and probably have more undesirable side effects, is the benefit worth the costs?

In mid-2002, Irving Kirsch and his colleagues published a meta-analysis of placebo-controlled clinical trials of antidepressants reported to the FDA. The report covered six drugs: fluoxetine (Prozac, others), paroxetine (Paxil), sertraline (Zoloft), venlafaxine (Effexor), nefazodone (Serzone), and citalopram (Celexa). The controversial and widely reported result was that as much as 80% of the positive response to these medications may be a placebo effect. The authors concluded that the popular drugs “may have no meaningful pharmacological effect at all.”

Response to an antidepressant is often measured by the Hamilton Rating Scale for Depression (HAM-D), a 17- or 21-item scale that is completed by a mental health professional. The items cover mood, thoughts, and other symptoms of depressive illness, such as sleep and appetite disturbance. The Kirsch study showed an average drug-placebo difference of less than 2 points on the HAM-D.

The authors grant that this difference, though small, was too great to be attributed to chance alone, but they label its clinical (practical) significance as “dubious.” This assertion has provoked fierce rebuttals from mainstream pharmacology researchers.

Placebo defined
To evaluate the arguments, it’s necessary to understand the purpose of a placebo and define the placebo effect. The ultimate test of a treatment’s effectiveness is its ability to induce measurable change in a double-blind, randomized controlled trial. In such an experiment, the subjects are divided into two or more groups—at random, to make the groups as similar as possible. The subjects are then given treatments that are ideally identical except for the treatment under study. Thus, conditions are controlled, and any group differences in the outcome can be attributed to one variable—the treatment under study.

The term double-blind means that neither the investigators nor the subjects know which treatment the subjects receive. This requirement is designed to limit any bias resulting from the interests, ambitions, hopes, and fears of the participants.

The placebo serves to keep the study double-blind. It is an inactive substance—one that is believed to have no physical or chemical effect on the condition being studied. Ideally, neither the subject nor the researcher should be able to distinguish it from the active drug.

What is called the placebo effect is the tendency for almost any treatment to make people feel better. In practice, the placebo effect also stands for several non-specific factors associated with treatment. Some subjects recover spontaneously. Others may come to regard their symptoms differently over time or simply change the way they describe them. If the study groups are similar enough, their chances of spontaneous recovery or simply having a better day should be equal.

Placebos look better and better
These non-specific responses are becoming stronger in trials of antidepressant drugs—a situation that creates problems for researchers. B.T. Walsh and his colleagues looked at 75 randomized controlled trials published between 1981 and 2000 and found that on average, the later the study, the higher the response to placebo. The proportion of people who respond to antidepressants in these trials is also rising, but not as fast. When placebo responses become stronger, there is less room at the margins where the antidepressant’s effect can be seen.

Some think the trend is due to changes in the kinds of patients who participate in clinical trials. Twenty years ago, more of these patients were suffering from the most severe forms of depression. These are the patients most likely to respond to antidepressants and least likely to respond to a placebo. Today, many depressed patients are already taking antidepressants. Since treatment is initiated earlier, fewer patients are referred to studies. To cope with this shrinking pool of subjects, researchers have turned to advertising, and that means they are often recruiting people with less severe symptoms. Researchers may also feel pressure to inflate HAM-D scores in order to boost participation in their studies.

The result, one argument goes, is that study subjects today have milder forms of depression that are more responsive to non-specific influences, including the hope or anticipation of improvement, the opportunity to talk about their distress, and regular meetings with enthusiastic professionals who are trying to find better treatments. Some scientists believe these factors, which are also common to all types of psychotherapy, account for much of the placebo effect.

Ethics of placebo-controlled drug trials
Some critics have questioned whether it is ethical to use placebos in antidepressant research, presuming that effective treatments are available. They believe that study participants should not be deprived of the proven benefits of medication. New drugs, such writers contend, should be compared only to older drugs that are known to be effective. Instead of proving a new drug’s efficacy by demonstrating its superiority to placebo, studies could try to show that a new drug is at least as good as an old one. That approach minimizes the number of patients who receive no active treatment.

But others warn that abandoning rigorous placebo comparisons could needlessly expose patients to antidepressant side effects. They also fear that antidepressant use may encourage a tendency to view ourselves as passive victims of our biology and to ignore the need for psychological coping. To the extent that drug responses and placebo responses are similar, these critics have a stronger argument. In that case, the ethical problems with the use of placebos become less serious.

I n the Declaration of Helsinki, the World Medical Association (WMA) sets guidelines for ethical human experimentation (see box, right). The WMA recognizes that risks to subjects in clinical trials must be considered along with a study’s scientific value.

But what about the risks of withholding an effective treatment? Another recent meta-analysis suggests that these risks are small. The worst outcome, suicide, was equally common in patients taking drugs and those taking a placebo. At least in short-term studies (those lasting no more than eight weeks), placebos do not appear to create any risks beyond the persistence of depressive symptoms. Longer studies may expose patients to greater risk of bad outcomes, but the rate of spontaneous improvement also rises as time passes.

The National Depressive and Manic-Depressive Association sponsored a consensus conference to tackle the questions arising from this debate. The participants came to the following conclusions:

Because people with mood disorders have an increasingly strong response to placebo in clinical trials, study results would be difficult to interpret without placebo comparisons.

Placebo trials are especially important in testing a new antidepressant with a unique mechanism of action.

Informed consent by patients is vital. Monitoring and safety practices are required. It may not be appropriate to include patients who are psychotic, at high risk for suicide, or otherwise in need of prompt treatment.

 Placebo is acceptable for people with bipolar (manic-depressive) disorder, but researchers must guard against the possibility of a manic episode induced by antidepressants.

The consensus group also made recommendations on research in children and on the further development of ethical guidelines, good documentation, and informed-consent procedures.

Limits of the evidence
Randomized controlled trials have their limitations. Depression is not a single, simple condition. It has many causes and is expressed in many ways. In a group of 20 people with similar symptoms, 20 different underlying processes could be at work. This enormous variety makes it almost impossible to control all the variables even in a single trial, and still more difficult to compare one trial with another.

The Hamilton scale so commonly used in this research presents another problem of interpretation. The scale includes only one direct question about depression. All the other items are less specific, and people with other problems, for instance, anxiety or personality disorders, may score high. Hamilton developed the scale for the most severely ill, hospitalized patients with depression. The scale is not as useful in less severely depressed patients. People with middling scores are a diverse group and may improve for a variety of reasons.

Furthermore, any scale is only a crude approximation of a person’s experience. No researcher would begin to argue that a single number could completely describe what is happening—biologically, psychologically, and socially—when a person becomes depressed.

Meta-analysis too has pitfalls. It is useful for identifying broad trends that may not show up in individual studies, and especially for combining the results of high-quality studies that would otherwise involve too great an element of chance (because, for example, the number of subjects is small).

But combining many results into a single figure can be misleading because studies vary so much in their design. They have different types of subjects with different types of depression. Researchers use different doses of a drug and continue treatment for different lengths of time. In some studies the subjects receive other treatments as well, and in some they do not. The analyst may be biased in choosing which studies to include, and may find it difficult to incorporate measures of the quality of studies.

The Kirsch meta-analysis suggests that antidepressants are not universally effective. Nevertheless, many trials do show a greater than chance difference from a placebo, and there are many individual reports of sustained positive response to antidepressants.

Group responses to different treatments may look similar on the surface. Because depression is not a single condition, some forms may respond to antidepressants, some to psychotherapy, some to neither, some to both, and some to a placebo. Thus, even treatments that seem equally effective may help different patients. Sixty-five out of 100 depressed people might respond to an antidepressant drug and 65 out of 100 to a certain kind of psychotherapy—but they would not be the same 65.

References Agency for Health Care Policy and Research. “Treatment of Depression: Newer Pharmacotherapies” (1999). http://www.ahcpr.gov/clinic/epcsums/deprsumm.htm Charney DS, et al. “National Depressive and Manic-Depressive Association Consensus Statement on the Use of Placebo in Clinical Trials of Mood Disorders,” Archives of General Psychiatry (March 2002): Vol. 59, No. 3, pp. 262–70. Khan A, et al. “Symptom Reduction and Suicide Risk in Patients Treated with Placebo in Antidepressant Clinical Trials: An Analysis of the Food and Drug Administration Database,” Archives of General Psychiatry (April 2000): Vol. 57, No. 4, pp. 311–17. Kirsch I, et al. (2002). “The Emperor’s New Drugs: An Analysis of Antidepressant Medication Data Submitted to the U.S. Food and Drug Administration,” Prevention & Treatment (2002), Vol. 5, Article 23. http://www.journals.apa.org/prevention/volume5/pre0050023a.html Leuchter AF, et al. “Changes in Brain Function of Depressed Subjects During Treatment with Placebo,” American Journal of Psychiatry (January 2002): Vol. 159, No. 1, pp. 122–29. Mayberg HS, et al. “The Functional Neuroanatomy of the Placebo Effect,” American Journal of Psychiatry (May 2002): Vol. 159, No. 5, pp. 728–37. Moncrieff J. “Are Antidepressants Overrated? A Review of Methodological Problems in Antidepressant Trials,” Journal of Nervous and Mental Disease (May 2001): Vol. 189, No. 5, pp. 288–95. Moncrieff J. “The Antidepressant Debate,” British Journal of Psychiatry (March 2002): Vol. 180, No. 3, pp. 193–94. Quitkin FM, et al. “Validity of Clinical Trials of Antidepressants,” American Journal of Psychiatry (March 2000): Vol. 157, No. 3, pp. 327–37. Walsh BT, et al. “Placebo Response in Studies of Major Depression: Variable, Substantial, and Growing,” Journal of the American Medical Association (April 10, 2002): Vol. 287, No. 14, pp. 1840–47

The authors suggest that there are two routes to recovery from depression. The response to medications travels a bottom-up or inside-out route, altering neurochemical processes at a deep level (such as the brainstem and striatum) first. The placebo effect—and possibly other non-drug treatments such as psychotherapy that change thoughts and emotions—may take a top-down or outside-in route, influencing the cerebral cortex first.

In another study (see Harvard Mental Health Letter, June 2002), A.F. Leuchter and colleagues, using a technique called quantitative electroencephalography (QEEG), measured brain activity during a nine-week trial comparing the antidepressants fluoxetine and venlafaxine (Effexor) with a placebo. All the subjects whose depression improved showed changes in activity at the front of the brain (the prefrontal cortex), but the direction of change depended on the treatment. Placebo responders showed an increase in prefrontal activity, and medication responders showed a decrease. (Patients who did not improve showed no change.)

One lesson from these studies is that similar responses on a global rating scale like the Hamilton do not necessarily reflect the same underlying biological processes. And the quality of the experience for the patient is probably quite variable too. A symphony by Mozart and a ballet by Stravinsky may have about the same number of notes, and listening to either may be enjoyable, but the experiences are not the same.

What’s more, there are competing analyses that defend the vast literature demonstrating the effectiveness of antidepressants. The Agency for Health Care Policy and Research, reviewing 80 controlled trials in 1999, concluded that antidepressants are an effective treatment for major depression and the milder, more chronic form of depression known as dysthymia. Also, many scientists believe that larger, more statistically powerful clinical trials will be more useful than meta-analyses for answering questions about antidepressant efficacy.

It is lost on no one that this debate may be tainted by guild allegiances and economic interests. Psychiatrists—especially psychopharmacologists, who make a living prescribing antidepressants—defend the drugs’ efficacy. Psychologists and other mental health professionals who do not prescribe drugs are more skeptical.

Personal experience also creates biases. Most busy psychotherapists and psychopharmacologists have seen patients improve, sometimes dramatically, in the course of treatment. A clinician’s last success story may be more convincing to her than a dozen controlled clinical trials.

Professional experience is essential, but anecdotes alone should not rule treatment decisions. Mental health professionals should also disregard ax-grinding by pro- and anti-drug special interests. They should be aware of the best available evidence, while remembering that the studies do not necessarily apply to an individual case. Professionals must evaluate the patient’s circumstances and offer hope, a sympathetic ear, practical advice, and insight. In many cases, after weighing the options, they may also decide that an antidepressant will help. With or without chicken soup.

—Michael Craig Miller, MD Editor in Chief, Harvard Mental Health Letter