WHEN PEOPLE HEAR the word depression, they tend to think of two different conditions. The first is acute and intense, a variant of Irma’s profound and paralyzing melancholy. The second is milder but persistent, a bleak state composed of sadness, lethargy, apathy, emotional vulnerability, and a poor sense of self. In my medical-school years, this lesser disorder was understood as a form of neurosis; for most of my time in practice, it has gone by the name dysthymia.
The current American diagnostic manual has eliminated dysthymia as an independent entity, integrating it into a group of chronic depressions. But the World Health Organization still recognizes dysthymia as a diagnosis, and the major research on low-level depression—thirty years’ worth—concerns dysthymia. I will stick with the term here.
Dysthymia entails sadness or emotional depletion, with a scattering of other symptoms, on most days, for years. Dysthymia is grave enough. Sitting with a dysthymic person, you feel yourself very much in the territory of depression.
Dysthymic patients are walled off from pleasure and vulnerable to disappointment. They feel defeated by minor humiliation. Dysthymia may wax and wane, but its worst phases are never far from consciousness. For many, the symptoms seem to extend to the core of personality. Dysthymia is debilitating. Studies show that in the long run, endless minor symptoms are as disabling as severe episodes spaced out in time.
Often, my dysthymic patients are plagued by writer’s block or the equivalent condition for artists, mathematicians, scientists, planners, and homemakers. The patients complain of loneliness, although I might have difficulty understanding why. They can be insightful and incisive, empathetic and wise—regarding others, anyway.
In psychotherapy, dysthymic patients can be especially engaging. Because their struggles are chronic, they dig deep. Although, as we shall see, the objective evidence is thin, I have always considered psychotherapy to be especially helpful in dysthymia.
That said, the psychoanalytic literature makes note of a frequent problem: neurotic patients who gain insights may have trouble translating them into action, trouble with “working through.” On this front, antidepressants can be especially useful—so I found increasingly as I began to prescribe. By the late 1990s, if a patient showed improved self-understanding but still could not set brush to canvas, and if depressive symptoms persisted, I might recommend a trial of medication. Often, an antidepressant supplied the final element—hope, drive, calm—that set recovery in motion.
For some patients, medication did the whole job: They gained energy and direction. They felt fine. For others, the initial uplift was subtle, a diminution in the fixed sense of inadequacy. Small changes allowed for big ones. To mount an exhibition, to risk moving toward marriage, to stay afloat and then succeed in the office workplace—these achievements create new and supportive contexts, ones that enhance mood and bring life meaning. I came to think of SSRIs as catalysts for working through.
In a general sense, research results confirmed these observations. In the early years of evidence-based medicine, in psychiatry, dysthymia research was at the forefront.
I have referred to evidence-based medicine without saying much about it. The movement announced its presence through a manifesto in the Journal of the American Medical Association in 1992. The goal of EBM, as doctors called it, was to get practitioners to rely less on clinical experience and more on findings from formal outcome trials.
Standard definitions of evidence-based practice are innocuous. One in wide use refers to “the integration of best research evidence with clinical expertise and patient values.” The devil is in the application. Notwithstanding the lip service given expertise and values, from the start devotees emphasized randomized trial results to the exclusion of all else. Too often, EBM entailed an uncritical embrace of meta-analyses. The practical effect was to disqualify most evidence, from biological plausibility to less formal experiments to clinical experience and observation.
As a result, clinicians had mixed feelings—I did—about evidence-based medicine. To the extent that it implied the existence of an opposite—superstition-based medicine?—the term was insulting. The more compact expression for evidence-based medicine is medicine.
That said, evidence-based medicine has had its victories, instances in which the application of meta-analysis in underexplored areas proved illuminating. The first major EBM-based study in psychiatry—it appeared in 1999—was of this sort. It concerned persistent low-level depression. The authors—the lead was a psychiatrist-epidemiologist, Mauricio Silva de Lima—would go on to substantial research careers in evidence-based medicine. In this pioneering paper, reviewing controlled trials, they found that commonly used antidepressants work for dysthymia and more vaguely defined chronic minor depressive disorders.
De Lima introduced readers to a measure favored in evidence-based medicine, the number needed to treat. Effect size looks at average symptom change—how much help treatment gives a typical patient. In contrast, the number needed to treat counts people—how many enjoy a substantial benefit; that is, how many enter a category like remission or response. De Lima wrote, “The NNT [number needed to treat] is an estimate of the number of patients a clinician would have to treat in order to observe one outcome due to that treatment”—one response or remission more than the doctor would see with the placebo grab bag. As with golf scores, lower means better.
In the treatment of dysthymia with antidepressants, where the measure of success was cutting symptoms by half, de Lima found a number needed to treat of just below 4. The findings in actual studies are more favorable to medication than those in the following example, but a simple way to think about the dysthymia results is this:
If you treat a typical group of four patients, two will do well on medication—but one would also have done well on placebo. You need to treat four patients to get one additional favorable response.
Having encountered additivity, we may mistrust this operation, subtraction. It is true that number needed to treat can be tricky. When our two patients respond to medication, we cannot conclude that only one is being helped by the drug. It may be that in both cases the antidepressant is supplying remoralization directly, biologically. It’s just that one of the patients would have been equally buoyed by attention from doctors and the passage of time.
It may be best to think of number needed to treat as a means of comparison—a tool used so widely that it allows efficacy levels to be contrasted across disciplines. In that case, how good is a number needed to treat of 4?
Matching his results against those found in a standard textbook (Evidence-Based Medicine: How to Practice and Teach EBM, by David L. Sackett), de Lima concluded that a number needed to treat of 4 was favorable.
This conclusion is counterintuitive. How terrific can a one-in-four hit rate be? But the number needed to treat is always humbling. We’re used to taking Excedrin and having our headache go away. Ideally, the number needed to treat is one, and before we give the matter thought, we may expect success to occur at that level.
The figures Sackett provided showed that doctors make a difference with a small proportion of the patients with whom they intervene: Give 250 people a cholesterol-blocking drug for a year and you prevent one bad cardiac outcome, like a heart attack or sudden death. (Today, the number-needed-to-treat estimates for statins run between 25 and 100.) Roto-Rooter ten blocked carotids, the arteries that supply blood to the brain, and you prevent one bad outcome—stroke or death—in the subsequent two years.
In the textbook’s reference chart, the only intervention with a low number needed to treat was medicating patients with very high blood pressure. If the risk was severe enough, you could see a number needed to treat of 3, where what you were forestalling was heart attack, stroke, or death. But treat patients with slightly lower and still abnormal blood pressures, and the figure leaped to 128, for five or six years of daily treatment.
Until the late 1990s, the number-needed-to-treat measure had largely been applied to preventive interventions. Since then, researchers have developed data on treatment of existing illness. Standard medications for heart failure, stroke, and chronic obstructive pulmonary disease have numbers needed to treat of 20 or higher. Only with antibiotics for specific indications, such as bladder infections, do you get numbers needed to treat as low as 3. Oh, and Excedrin in the treatment of headache? If your criterion is reduction of pain by half, the number needed to treat is north of 5. The decision, whether to recommend the treatment, is vastly more complex, but looking at response rates alone, SSRIs for dysthymia work as reliably as Excedrin for headache.
For medications expected to give near-term benefit, a number needed to treat in the mid-single digits signals high efficacy, and numbers needed to treat of up to 10 can identify useful drugs. For grave diseases with few alternative approaches—relentless cancers—doctors may resort to treatments that benefit only one in hundreds of patients.
Dysthymia’s number needed to treat of 4 has an extra advantage. For blocked arteries, you give the full surgical treatment—widening the passage—without knowing who will be helped. Antidepressants are different: dysthymic patients whose symptoms don’t budge can come off the medication and pursue a different remedy. If we think about our hypothetical four patients, in short order two of them, the nonresponders, may no longer be taking the antidepressant.
Effectiveness for dysthymia goes a long way toward explaining Prozac’s rise to popularity. Imipramine might have done the job equally well, but was so hard to live with that it was reserved for highly symptomatic patients.
This property of SSRIs, their efficacy for dysthymia, is the great open secret of the antidepressant debate. I know that I’ve given Marcia Angell a hard time, but I take her opinion as a touchstone, an indicator of where sophisticated nonpsychiatrists stand. Evidently, the critiques that Angell reviewed did not mention research on dysthymia. Speaking to the Boston Globe columnist about my Sunday Review piece, she had complained, “In his article, Kramer says it’s well established that antidepressants work for chronic and recurrent mild depression, but where is the evidence for that? That screams out for a reference. Should we believe it just because he says that?”
The paper I had discussed was easily accessed. I had cited the recovery rates, and they matched those in the first reference to pop up if you plug dysthymia and meta-analysis into Google. The study, from 2011, was an updating of de Lima’s, and the new version found much the same level of benefit. And long before, in 2002, the Cochrane Collaboration had invited de Lima to contribute a version of his overview of the literature. He had worked with Joanna Moncrieff, of University College London, who would go on to write papers with Irving Kirsch and become an outspoken skeptic about the use of antidepressants in the treatment of depression. De Lima and Moncrieff’s summary, posted on the Cochrane website, was straightforward: “Drugs are effective in the treatment of dysthymia.” This indication for drug treatment was at once officially recognized and absent from public discussion.
The dysthymia research is not extensive. De Lima’s 1999 study covered fewer than two thousand patients, and the 2011 follow-up, which excluded older studies, fewer than fifteen hundred. That said, the evidence is compelling. Drug companies have not sought FDA approval for antidepressants in the treatment of dysthymia, so the field has been spared the distortions that commercial trials entail. The research has been university-based, conducted by academics following their own standards. Often, the trials were conducted in clinics where patients received their usual care.
Gathering material for review, statisticians worry about “file-drawer bias”—the tendency for disappointing research to remain unpublished—and they have ways of analyzing data to see whether unfavorable small studies remain hidden. The methods are based on the premise that, because many careers depend on them and because the results tend to be convincing, large trials will find their way into print. Small trials will have a tougher road to publication, and preferences on all sides—editors’, sponsors’—may lead to unfavorable research’s being deep-sixed. Where there’s no publication bias, we expect that the average of many small trials will match the average of the few large trials. The dysthymia literature has that pattern, the one that suggests that publication bias has not been a factor.
Moreover, the studies show strong, consistent results. One research group tried to estimate the number of unpublished, less favorable trials that would need to be unearthed for the dysthymia findings to come into question. The answer was: many more than there are published trials altogether. That same calculation applies as readily to future research. The conclusion, substantial antidepressant efficacy, is likely to hold up indefinitely, an unsurprising prospect since in practice the medicines work with great reliability.
It makes sense that virtually all the dysthymia research should have found its way into print. Throughout the twentieth century, psychotherapists had claimed neurosis as their domain. When antidepressants became available, pharmacologists doubted that they would work for depression that looked like a component of personality. Trials that showed drugs falling short would have been welcome.
The field had been exactly wrong. In 2008, Bruce Wampold and other psychologists at the University of Wisconsin reviewed outcome studies. (I am a great admirer of Wampold’s work on psychotherapy outcome. He finds that factors common to therapies, such as the doctor-patient relationship, account for much of the efficacy and that the specific benefits of cognitive behavioral therapy in particular have been oversold.) Wampold reported that “medication was significantly more efficacious than psychotherapy in the treatment of dysthymia.” His caveat was that trials of the usual duration, often only four months, are brief; psychotherapy’s benefits may appear later. By the same token, medication, if patients stayed on it, continued to work for dysthymia for a year or more.
Later, we will consider the “severity hypothesis,” which holds that antidepressants work best (or only) for highly symptomatic, grave depression. But even exponents of that view make an exception: when they dismiss medication as a treatment for mild and moderate mood disorder, they note that antidepressants work for dysthymia. That is to say, medication works for the sort of mild depression that doctors are most likely to prescribe for, mild depression that persists.
It is hard to know why the dysthymia research has not gotten more play, but inattention to it has clouded the antidepressant debate. Far from being placebos, antidepressants have repeatedly been shown effective for a debilitating mental illness.
As late as the 1990s, doubt remained over antidepressants’ usefulness in the treatment of chronic, low-level depression. Against expectations, meta-analysis based on consistent trials demonstrated that medication can be effective, and with a highly acceptable number needed to treat. Shortly, psychiatrists became adept at combining interventions, medication and psychotherapy. The application of antidepressants in the care of dysthymic patients is a key development in mental health care in the last half century, and EBM was the motive force.