How Effective Are Antidepressant Drugs?
Antidepressants, one of the major weapons in the biological therapeutic arsenal, illustrate well the largely unacknowledged uncertainty that exists in the biological approach to psychopathology. We suggest that, at present, no one actually knows how effective antidepressants are. Confident declarations about their potency go well beyond the existing evidence.
To get an understanding of the scientific status of antidepressants, we analyzed how much more effective the antidepressants are than inert pills called "placebos." That is, if antidepressants are given to one depressed group and a placebo to another group, how much greater is the recovery of those taking the active drug as compared to those taking the inactive placebo? Generous claims that antidepressants usually produce improvement in about 60 to 70 percent of patients are not infrequent, whereas placebos are said to benefit 25 to 30 percent. If antidepressants were, indeed, so superior to placebos, this would be a persuasive advertisement for the biological approach.
We found 15 major reviews of the anti-depressant literature. Surprisingly, even the most positive reviews indicate that 30 to 40 percent of studies show no significant difference in response to drug versus placebo! The reviews indicate overall that one-third of patients do not improve with anti-depressant treatment, one-third improve with placebos, and an additional third show a response to medication they would not have attained with placebos. In the most optimistic view of such findings, two-thirds of the cases (placebo responders and those who do not respond to anything) do as well with placebo as with active medication.
We also found two large-scale quantitative evaluations (meta-analyses) integrating the outcomes of multiple studies of antidepressants. They clearly indicated, on the average, quite modest therapeutic power.
We were particularly impressed by the large variation in outcomes of studies conducted at multiple clinical sites or centers. Consider a study that compared the effectiveness of an antidepressant among patients at five different research centers. Although the pooled results demonstrate that the drug was generally more effective than placebo, the results from individual centers reveal much variation. After six weeks of treatment, every one of the six measures of effectiveness showed the anti-depressant (imipramine) to be merely equivalent to placebo in two or more of the centers. In two of the settings, a difference favoring the medication was detected on only one of 12 outcome comparisons.
In other words, the pooled, apparently favorable, outcome data conceal that dramatically different results could be obtained as a function of who conducted the study and the specific conditions at each locale. We can only conclude that a good deal of fragility characterized the apparent superiority of drug over placebo. The scientific literature is replete with analogous examples.
Incidentally, we also looked at whether modern studies, which are presumably better protected against bias, use higher doses, and often involve longer treatment periods, show a greater superiority of the antidepressant than did earlier studies. The literature frequently asserts that failures to demonstrate antidepressant superiority are due to such methodological failures as not using high enough doses, and so forth.
We examined this issue in a pool of 16 studies assembled by psychiatrists John Kane and Jeffrey Lieberman in 1984. These studies all compare a standard drug, such as imipramine or amitriptyline, to a newer drug and a placebo. They use clearer diagnostic definitions of depression than did the older studies and also adopt currently accepted standards for dosage levels and treatment duration. When we examined the data, we discovered that the advantage of drug over placebo was modest. Twenty-one percent more of the patients receiving a drug improved as compared to those on placebo. Actually, most of the studies showed no difference in the percentage of patients significantly improved by drugs. There was no indication that these studies, using more careful methodology, achieved better outcomes than older studies.
Finally, it is crucial to recognize that several studies have established that there is a high rate of relapse among those who have responded positively to an anti-depressant but then are taken off treatment. The relapse rate may be 60 percent or more during the first year after treatment cessation. Many studies also show that any benefits of antidepressants wane in a few months, even while the drugs are still being taken. This highlights the complexity of evaluating antidepressants. They may be effective initially, but lose all value over a longer period.
Are Drug Trials Biased?
As we burrowed deeper into the antidepressant literature, we learned that there are also crucial problems in the methodology used to evaluate psychotropic drugs. Most central is the question of whether this methodology properly shields drug trials from bias. Studies have shown that the more open to bias a drug trial is, the greater the apparent superiority of the drug over placebo. So questions about the trustworthiness of a given drug-testing procedure invite skepticism about the results.
The question of potential bias first came to our attention in studies comparing inactive placebos to active drugs. In the classic double-blind design, neither patient nor researcher knows who is receiving drug or placebo. We were struck by the fact that the presumed protection provided by the double-blind design was undermined by the use of placebos that simply do not arouse as many body sensations as do active drugs. Research shows that patients learn to discriminate between drug and placebo largely from body sensations and symptoms.
A substance like imipramine, one of the most frequently studied antidepressants, usually causes clearly defined sensations, such as dry mouth, tremor, sweating, constipation. Inactive placebos used in studies of antidepressants also apparently initiate some body sensations, but they are fewer, more inconsistent, and less intense as indicated by the fact that they are less often cited by patients as a source of discomfort causing them to drop out of treatment.
Vivid differences between the body sensations of drug and placebo groups could signal to patients as to whether they are receiving an active or inactive agent. Further, they could supply discriminating cues to those responsible for the patients's day-to-day treatment. Nurses, for example, might adopt different attitudes toward patients they identify as being "on" versus "off" active treatment—and consequently communicate contrasting expectations.
The Body Of Evidence
This is more than theoretical. Researchers have reported that in a double-blind study of imipramine, it was possible by means of side effects to identify a significant number of the patients taking the active drug. Those patients receiving a placebo have fewer signals (from self and others) indicating they are being actively treated and should be improving. By the same token, patients taking an active drug receive multiple signals that may well amplify potential placebo effects linked to the therapeutic context. Indeed, a doctor's strong belief in the power of the active drug enhances the apparent therapeutic power of the drug or placebo.
Is it possible that a large proportion of the difference in effectiveness often reported between antidepressants and placebos can be explained as a function of body sensation discrepancies? It is conceivable, and fortunately there are research findings that shed light on the matter.
Consider an analysis by New Zealand psychologist Richard Thomson. He reviewed double-blind, placebo-controlled studies of antidepressants completed between 1958 and 1972. Sixty-eight had employed an inert placebo and seven an active one (atropine) that produced a variety of body sensations. The antidepressant had a superior therapeutic effect in 59 percent of the studies using inert placebo—but in only one study (14 percent) using the active placebo. The active placebo eliminated any therapeutic advantage for the antidepressants, apparently because it convinced patients they were getting real medication.
How Blind Is Double-Blind?
Our concerns about the effects of inactive placebos on the double-blind design led us to ask just how blind the double-blind really is. By the 1950s reports were already surfacing that for psychoactive drugs, the double-blind design is not as scientifically objective as originally assumed. In 1993 we searched the world literature and found 31 reports in which patients and researchers involved in studies were asked to guess who was receiving the active psychotropic drug and who the placebo. In 28 instances the guesses were significantly better than chance—and at times they were surprisingly accurate. In one double-blind study that called for administering either imipramine, phenelzine, or placebo to depressed patients, 78 percent of patients and 87 percent of psychiatrists correctly distinguished drug from placebo.
One particularly systematic report in the literature involved the administration of alprazolam, imipramine, and placebo over an eight-week period to groups of patients who experienced panic attacks. Halfway through the treatment and also at the end, the physicians and the patients were asked to judge independently whether each patient was receiving an active drug or a placebo. If they thought an active drug was being administered, they had to decide whether it was alprazolam or imipramine. Both physicians (with an 88 percent success rate) and patients (83 percent) substantially exceeded chance in the correctness of their judgments. Furthermore, the physicians could distinguish alprazolam from imipramine significantly better than chance. The researchers concluded that "double-blind studies of these pharmacological treatments for panic disorder are not really 'blind.'"
Yet the vast majority of psychiatric drug efficacy studies have simply assumed that the double-blind design is effective; they did not test the blindness by determining whether patients and researchers were able to differentiate drug from placebo.
We take the somewhat radical view that this means most past studies of the efficacy of psychotropic drugs are, to unknown degrees, scientifically untrustworthy. At the least, we can no longer speak with confidence about the true differences in therapeutic power between active psychotropic drugs and placebos. We must suspend judgment until future studies are completed with more adequate controls for the defects of the double-blind paradigm.
Other bothersome questions arose as we scanned the cascade of studies focused on antidepressants. Of particular concern is how unrepresentative the patients are who end up in the clinical trials. There are the usual sampling problems having to do with which persons seek treatment for their discomfort, and, in addition, volunteer as subjects for a study. But there are others. Most prominent is the relatively high proportion of patients who "drop out" before the completion of their treatment programs.
Numerous dropouts occur in response to unpleasant side effects. In many published studies, 35 percent or more of patients fail to complete the research protocol. Various procedures have been developed to deal fairly with the question of how to classify the therapeutic outcomes of dropouts, but none can vitiate the simple fact that the final sample of fully treated patients has often been drastically reduced.
There are still other filters that increase sample selectivity. For example, studies often lose sizable segments of their samples by not including patients who are too depressed to speak, much less participate in a research protocol, or who are too disorganized to participate in formal psychological testing. We also found decisions not to permit particular racial or age groups to be represented in samples or to avoid using persons below a certain educational level. Additionally, researchers typically recruit patients whose depression is not accompanied by any other type of physical or mental disorder, a situation that does not hold for the depressed in the general population.
So we end up wondering about the final survivors in the average drug trial. To what degree do they typify the average individual in real life who seeks treatment? How much can be generalized from a sample made up of the "leftovers" from multiple depleting processes? Are we left with a relatively narrow band of those most willing to conform to the rather rigid demands of the research establishment? Are the survivors those most accepting of a dependent role?
The truth is that there are probably multiple kinds of survivors, depending upon the specific local conditions prevailing where the study was carried out. We would guess that some of the striking differences in results that appear in multicenter drug studies could be traced to specific forms of sampling bias. We do not know how psychologically unique the persons are who get recruited into, and stick with, drug research enterprises. We are not the first to raise this question, but we are relatively more alarmed about the potential implications.
Researched Motivation And Outcome
We conducted an analysis that further demonstrates how drug effectiveness diminishes as the opportunity for bias in research design wanes. This analysis seized on studies in which a newer antidepressant is compared (under double-blind conditions) with an older, standard antidepressant and a placebo. In such a context the efficacy of the newer drug (which the drug company hopes to introduce) is of central interest to the researcher, and the effectiveness of the older drug of peripheral import. Therefore, if the double-blind is breached (as is likely), there would presumably be less bias to enhance the efficacy of the older drug than occurred in the original trials of that drug.
We predicted that the old drug would appear significantly less powerful in the newer studies than it had in earlier designs, where it was of central interest of the researcher. To test this hypothesis, we located 22 double-blind studies in which newer antidepressants were compared with an older antidepressant drug (usually imipramine) and a placebo. Our meta-analysis revealed, as predicted, that the efficacy rates, based on clinicians's judgments of outcome, were quite modest for the older antidepressants. In fact, they were approximately one-half to one-quarter the average size of the effects reported in earlier studies when the older drug was the only agent appraised.
Let us be very clear as to what this signifies: When researchers were evaluating the antidepressant in a context where they were no longer interested in proving its therapeutic power, there was a dramatic decrease in that apparent power, as compared to an earlier context when they were enthusiastically interested in demonstrating the drug's potency. A change in researcher motivation was enough to change outcome. Obviously this means too that the present double-blind design for testing drug efficacy is exquisitely vulnerable to bias.
Another matter of pertinence to the presumed biological rationale for the efficacy of antidepressants is that no consistent links have been demonstrated between the concentration of drug in blood and its efficacy. Studies have found significant correlations for some drugs, but of low magnitude. Efforts to link plasma levels to therapeutic outcome have been disappointing.
Similarly, few data show a relationship between antidepressant dosage levels and their therapeutic efficacy. That is, large doses of the drug do not necessarily have greater effects than low doses. These inconsistencies are a bit jarring against the context of a biological explanatory framework.
We have led you through a detailed critique of the difficulties and problems that prevail in the body of research testing the power of the antidepressants. We conclude that it would be wise to be relatively modest in claims about their efficacy. Uncertainty and doubt are inescapable.
While we have chosen the research on the antidepressants to illustrate the uncertainties attached to biological treatments of psychological distress, reviews of other classes of psychotropic drugs yield similar findings. After a survey of anti-anxiety drugs, psychologist Ronald Lipman concluded there is little consistent evidence that they help patients with anxiety disorders: "Although it seems natural to assume that the anxiolytic medications would be the most effective psychotropic medications for the treatment of anxiety disorders, the evidence does not support this assumption."
Biological Versus Psychological?
The faith in the biological approach has been fueled by a great burst of research. Thousands of papers have appeared probing the efficacy of psychotropic drugs. A good deal of basic research has attacked fundamental issues related to the nature of brain functioning in those who display psychopathology. Researchers in these areas are dedicated and often do excellent work. However, in their zeal, in their commitment to the so-called biological, they are at times overcome by their expectations. Their hopes become rigidifying boundaries. Their vocabulary too easily becomes a jargon that camouflages over-simplified assumptions.
A good example of such oversimplification is the way in which the term "biological" is conceptualized. It is too often viewed as a realm distinctly different from the psychological. Those invested in the biological approach all too often practice the ancient Cartesian distinction between somatic-stuff and soul-stuff. In so doing they depreciate the scientific significance of the phenomena they exile to the soul-stuff category.
But paradoxically, they put a lot of interesting phenomena out of bounds to their prime methodology and restrict themselves to a narrowed domain. For example, if talk therapy is labeled as a "psychological" thing—not biological—this implies that biological research can only hover at the periphery of what psychotherapists do. A sizable block of behavior becomes off limits to the biologically dedicated.
In fact, if we adopt the view that the biological and psychological are equivalent (biological monism), there is no convincing real-versus-unreal differentiation between the so-called psychological and biological. It all occurs in tissue and one is not more "real" than the other. A patient's attitude toward the therapist is just as biological in nature as a patient's response to an antidepressant. A response to a placebo is just as biological as a response to an anti-psychotic drug. This may be an obvious point, but it has not yet been incorporated into the world views of either the biologically or psychologically oriented.
Take a look at a few examples in the research literature that highlight the overlap or identity of what is so often split apart. In 1992, psychiatrist Lewis Baxter and colleagues showed that successful psychotherapy of obsessive-compulsive patients results in brain imagery changes equivalent to those produced by successful drug treatment. The brain apparently responds in equivalent ways to both the talk and drug approaches. Even more dramatic is a finding that instilling in the elderly the illusion of being in control of one's surroundings (by putting them in charge of some plants) significantly increased their life span compared to a control group. What could be a clearer demonstration of the biological nature of what is labeled as a psychological expectation than the postponement of death?
Why are we focusing on this historic Cartesian confusion? Because so many who pursue the so-called biological approach are by virtue of their tunnel vision motivated to overlook the psychosocial variables that mediate the administration of such agents as psychotropic drugs and electroconvulsive therapy. They do not permit themselves to seriously grasp that psychosocial variables are just as biological as a capsule containing an antidepressant. It is the failure to understand this that results in treating placebo effects as if they were extraneous or less of a biological reality than a chemical agent.
Indeed, placebos have been shown to initiate certain effects usually thought to be reserved for active drugs. For example, placebos clearly show dose-level effects. A larger dose of a placebo will have a greater impact than a lower dose. Placebos can also create addictions. Patients will poignantly declare that they cannot stop taking a particular placebo substance (which they assume is an active drug) because to do so causes them too much distress and discomfort.
Placebos can produce toxic effects such as rashes, apparent memory loss, fever, headaches, and more. These "toxic" effects may be painful and even overwhelming in their intensity. The placebo literature is clear: Placebos are powerful body-altering substances, especially considering the wide range of body systems they can influence.
Actually, the power of the placebo complicates all efforts to test the therapeutic efficacy of psychotropic drugs. When placebos alone can produce positive curative effects in the 40 to 50 percent range (occasionally even up to 70-80 percent), the active drug being tested is hard-pressed to demonstrate its superiority. Even if the active drug exceeds the placebo in potency, the question remains whether the advantage is at least partially due to the superior potential of the active drug itself to mobilize placebo effects because it is an active substance that stirs vivid body sensations. Because it is almost always an inactive substance (sugar pill) that arouses fewer genuine body sensations, the placebo is less convincingly perceived as having therapeutic prowess.
Drug researchers have tried, in vain, to rid themselves of placebo effects, but these effects are forever present and frustrate efforts to demonstrate that psychoactive drugs have an independent "pure" biological impact. This state of affairs dramatically testifies that the labels "psychological" and "biological" refer largely to different perspectives on events that all occur in tissue. At present, it is somewhat illusory to separate the so-called biological and psychological effects of drugs used to treat emotional distress.
The literature is surprisingly full of instances of how social and attitudinal factors modify the effects of active drugs. Anti-psychotic medications are more effective if the patient likes rather than dislikes the physician administering them. An antipsychotic drug is less effective if patients are led to believe they are only taking an inactive placebo. Perhaps even more impressive, if a stimulant drug is administered with the deceptive instruction that it is a sedative, it can initiate a pattern of physiological response, such as decreased heart rate, that is sedative rather than arousing in nature. Such findings reaffirm how fine the line is between social and somatic domains.
What are the practical implications for distressed individuals and their physicians? Administering a drug is not simply a medical (biological) act. It is, in addition, a complex social act whose effectiveness will be mediated by such factors as the patient's expectations of the drug and reactions to the body sensations created by that drug, and the physician's friendliness and degree of personal confidence in the drug's power. Practitioners who dispense psychotropic medications should become thoroughly acquainted with the psychological variables modifying the therapeutic impact of such drugs and tailor their own behavior accordingly. By the same token, distressed people seeking drug treatment should keep in mind that their probability of benefiting may depend in part on whether they choose a practitioner they truly like and respect. And remember this: You are the ultimate arbiter of a drug's efficacy.
How to go about mastering unhappiness, which ranges from "feeling blue" to despairing depression, puzzles everyone. Such popular quick fixes as alcohol, conversion to a new faith, and other splendid distractions have proven only partially helpful. When antidepressant drugs hit the shelves with their seeming scientific aura, they were easily seized upon. Apparently serious unhappiness (depression) could now be chemically neutralized in the way one banishes a toothache.
But the more we learn about the various states of unhappiness, the more we recognize that they are not simply "symptoms" awaiting removal. Depressed feelings have complex origins and functions. In numerous contexts—for example, chronic conflict with a spouse—depression may indicate a realistic appraisal of a troubling problem and motivate a serious effort to devise a solution.
While it is true that deep despair may interfere with sensible problem-solving, the fact is that, more and more, individuals are being instructed to take anti-depressants at the earliest signs of depressive distress and this could interfere with the potentially constructive signaling value of such distress. Emotions are feelings full of information. Unhappiness is an emotion, and despite its negativity, should not be classified single-mindedly as a thing to tune out. This in no way implies that one should submit passively to the discomfort of feeling unhappy. Actually, we all learn to experiment with a variety of strategies for making ourselves feel better, but the ultimate aim is long-term effective action rather than a depersonalized "I feel fine."