Why the debate over this question never seems to end.
That presumption, however, is often called into question and in the last few months, several articles and studies have appeared that challenge the notion that antidepressants work.
First, in a study first published online in the journal The Lancet Psychiatry last September, investigators from several British universities reported that “we did not find convincing evidence that [the antidepressant medication] sertraline [brand name: Zoloft] led to a greater reduction in depressive symptoms compared with placebo within 6 weeks.” Curiously, in the discussion of their findings, the authors chose to emphasize a secondary data analysis that showed that sertraline was superior to placebo in reducing anxiety symptoms, but the failure to find an effect for a commonly prescribed and FDA-approved antidepressant for actual depression was the primary finding of the study.
Second, in a review of placebo-controlled studies of antidepressants also published earlier this year, the authors concluded: “The benefits of antidepressants seem to be minimal and possibly without any importance to the average patient with major depressive disorder.” Their review challenged the notions that antidepressants are more effective for more severely ill depressed patients and that adding antidepressants to psychotherapy meaningfully improves outcomes.
Third, a research group has been funded to reanalyze data, some of it never published, from a huge antidepressant trial funded by the federal government referred to as STAR*D. Already one of the investigators involved in the reanalysis has concluded that mistakes were made in the way the data were originally presented.
All of this doubt about antidepressant medication efficacy feeds into a general concern that psychiatry has overemphasized the usefulness of medications in general. In a searing indictment of the state of psychiatry that appeared in October in The New England Journal of Medicine, Harvard’s Caleb Gardner and Arthur Kleinman wrote that the use of medication to treat psychiatric illnesses like depression “has taken over practice to an alarming degree.” They go on to assert that “psychiatry finds itself plagued by overprescription of psychiatric medication for a large segment of the population.”
First They Work, Then They Don’t, Then They Do
This recent pushback against antidepressant medication seems part of a cycle in which papers alternatively appear showing that they do work followed some time later by papers saying they don’t and then back to papers showing efficacy. In 2018, for example, a group led by researchers from the University of Oxford performed what they called the “most comprehensive currently available evidence base to guide the initial choice about pharmacological treatment for acute major depressive disorder in adults.” They presented clear evidence that antidepressant medications work better than placebo, some more than others. “We found that all antidepressants included in the meta-analysis were more efficacious than placebo in adults with major depressive disorder,” they concluded. Note that sertraline—a.k.a. Zoloft—which was said not to work for depression by six weeks in the primary care study mentioned above—is one of the antidepressants that does appear to work in this study. In reporting on the Oxford analysis, which included a huge number of studies and study participants, The Guardian noted that “doctors hope [this study] will finally put to rest doubts about the controversial medicine.
Apparently not. Little more than a year later we again have articles analyzing many of the same studies and declaring that antidepressants don’t work. What is going on here?
It is generally understood in the field of psychopharmacology that in order to obtain approval from the FDA to market a new antidepressant it is necessary to present at least two large, randomized controlled trials (RCTs) in which the new medication meets the criteria of a statistical test of superiority at lowering the score on a depression symptom rating scale compared to placebo. The FDA does take a look at “failed” studies—studies in which the drug does not turn out to be superior to placebo—but has often considered that two “positive” trials are sufficient to declare the drug worthy of FDA approval. Because these studies, called phase III trials, are very expensive to conduct in order to meet exacting FDA regulatory standards, it is unusual for a drug company to continue to study a drug if it gets one or two failed studies. Thus, it is unusual for there to be just one or two positive large RCTs for an antidepressant and many times that many failed studies.
The Benefits and Risks of Meta-analysis
According to a 2017 rule made by the Obama administration, all clinical trial data from drug company-sponsored trials must be made public. This means that research groups are now able to amass the data from multiple studies, some conducted by drug companies and some funded by other sources like the National Institutes of Health, and perform a meta-analysis on these data. Meta-analysis is a technique that is used when it is felt that individual studies may not be strong enough to answer an important research question. They are an excellent tool for finding important information that individual studies with small sample sizes are not powerful enough to reveal.
For example, let’s say we are worried that a potentially toxic ingredient is present in orange juice produced by one method but not by another. One research group tests 10 samples from each type of production and finds that two of 10 samples from the suspect method indeed have a potentially toxic ingredient, but only one of the 10 samples in the more standard method has it. That’s worrisome, but doesn’t make it over a traditional statistical hurdle and the difference could be due to mere chance. But if 10 research groups studying the same number of samples all find the same thing, it may be possible to combine all the data and reanalyze the results. A meta-analysis putting the data together from all 10 studies would give us 20 out of 100 samples in the suspect method of production that have the toxic ingredient compared to 10 in 100 samples in the standard method. That is a statistically significant difference and makes clear that the suspect method indeed is twice as likely to contain the toxic ingredient. In this case, performing a meta-analysis provides vital information.
Something different happens, however, when data from positive and failed antidepressant trials are lumped together. Invariably, the effect is to diminish the size of the difference between drug and placebo seen in the positive studies alone. Does that actually mean that we should discount the positive studies? Let’s say, for instance, that we have two studies in which 600 people each are randomized to drug or placebo and in each case the drug turns out to be more effective at reducing depression than the placebo. On the other hand, we have two studies of 600 people each in which neither study shows a statistically significant difference. Note that it almost never happens that placebo works better than an antidepressant, which would be called a “negative study.” If we do all the math for a meta-analysis, it becomes clear that in one group of 1,200 people with depression the antidepressant “worked” but in another group of 1,200, it didn’t beat placebo. Combining all of those results into one analysis of all 2,400 participants will make the size of the difference between drug and placebo in the two positive studies shrink or even disappear.
Why is there so much variability in the results obtained from placebo-controlled trials of antidepressants? The problem usually stems from the fact that placebo actually works for depression. In almost all trials involving psychiatric drugs, the group randomized to get the sugar pill gets better. In positive studies, the people on active drug get “more” better and in failed studies they don’t. The difference usually resides in the size of the placebo response. Some studies have larger placebo response rates than others. No one knows why this is true, but we do know that people who enter clinical trials get a lot of attention from research personnel and it may be this that accounts, at least in part, for the size of the placebo response.
Now one could argue that if giving a depressed patient more attention is sufficient to improve their depression, medication should not be necessary. Experts in psychiatry and psychology seem reluctant to agree that merely having a friend or family member spend more time with a depressed patient is going to work in most cases. Moreover, the kind of attention someone gets when enrolled in a clinical trial would be hard to duplicate in the real world if the burden were placed solely on friends and family. A participant in a research study may be asked to visit the research clinic weekly for several months, spending several hours during each visit seeing doctors, nurses, and research assistants who probe how the patient is doing, administer various psychological tests, and obtain blood and urine samples. Finally, despite the powerful effect of placebo, antidepressant medications do frequently demonstrate greater effects in rigorously conducted clinical trials. With a disorder like depression that causes so much suffering and can be fatal, we are going to want to employ treatments that appear to work at least some of the time.
When Should An Antidepressant Be Prescribed?
Even if antidepressants do work, however, the issue of when to use them remains very much on the table. For a 55-year-old man who stops eating, sleeping, bathing, and going to work, tells family members he sees no reason to go on living, and has difficulty talking about his feelings or paying attention to what others are saying, an antidepressant medication might be the only treatment approach with any chance of working.
But what about the many people who tell their primary care doctors that they feel a little down lately, that things at work have been tough, and that they are having trouble falling asleep? Many such comparatively mild cases of depression remit without any intervention at all, so inviting such people to return for a follow-up visit in a week is a much better idea than prescribing medication. This is where Gardner and Kleinman’s warning about “overprescription of psychiatric medications” seems particularly salient, because primary care physicians too often reach for the prescription pad rather than offering encouragement and a willingness to listen to the patient’s problems.
And what about the more severely depressed person who is suffering but at the same time is able to talk fluently about what is going on with her? Most studies show that evidence-based psychotherapies like cognitive behavioral therapy (CBT) work just as well as antidepressant medications, even for severe depression. Patients with depression generally express a preference for psychotherapy over medication. Psychotherapy, of course, has none of the side effects common to antidepressant medication.
The logical approach to treating depression that requires treatment would seem to be to start with an evidence-based psychotherapy like CBT and reserve antidepressant medication for patients who are too ill to benefit from psychotherapy or who don’t respond to it. Logic is usurped in this case, however, by economics. At least in the short run, a course of CBT costs more than putting someone on an antidepressant. Over and over again, experts lament the inability of patients to actually access evidence-based therapies like CBT, often because of an inadequate supply of properly trained therapists or the refusal of health insurance companies to pay for it.
The continuous cycles of meta-analyses finding that antidepressants either do or don’t work—and the headlines that accompany each version of the story—are really beside the point. We have enough data at this point showing that for some people with severe depression they are effective. The real debate should be about how we can educate physicians to stop prescribing antidepressants indiscriminately and how we can make evidence-based psychotherapies available as first-line treatments for depression. Let’s forgo any more antidepressant meta-analyses and concentrate on solving the public health issues that prevent them from being used to best advantage, that is, with more restraint.