CHARACTERIZING MILD MAJOR depression, I mentioned Maggie, who could barely drag herself to work until medication restored her morale. I think equally of Josh, whose chief complaint was a dulling of thought. After a minor business setback, one that should hardly have derailed him, he found himself unable to analyze alternatives or make choices. He had never before been in this state.
Psychotherapy with a psychologist whose work I admire had reached a standstill. But an antidepressant, Lexapro, lifted the impediment, and in short order.
These stories, Maggie’s and Josh’s, are among the commonest in psychiatry and, for that matter, in all of medicine. The JAMA paper’s conclusion—minimal antidepressant efficacy in mild and moderate depression—is contrary to clinical experience. That meta-analysis, so particular in its inclusion criteria, should have left the field and the media hungry for a contrasting overview, one built around patients with mild-to-moderate major depression. Ideally, the research would test adequate doses of medication. Ideally, the enrollment design would encourage honesty and accuracy.
That study exists. In 2012, in the wake of the JAMA paper, researchers at the New York State Psychiatric Institute, mostly Columbia University faculty, looked at thirty-two years’ worth of research from the hospital’s depression-evaluation service, where Donald Klein and Fred Quitkin, pioneers in outcome assessment, had worked.
Sifting through records on 1,440 patients, the Columbia group located 825 with nonsevere major depression, more patients in the relevant range than there were patients altogether in the JAMA collection. The average initial Hamilton score was below 14—mild. The files came from six randomized trials, testing four types of antidepressant, all in adequate doses. Imipramine was the drug most often tested, typically at 200 to 300 milligrams—realistic prescribing. Except that the trial designs included a run-in phase, this overview contained the research that the JAMA team had been seeking.
Here’s the beauty part: In their entry criteria, five of the six studies had no lower limit to the Hamilton score—none—and no minimum symptom number. All were welcome. The sixth study, like the Dartmouth paper in the JAMA collection, required a Hamilton score of 10, what NICE called “subthreshold” depression. The Columbia researchers noted, “Clinicians should have had no incentive to apply diagnostic criteria loosely or inflate entry [Hamilton] scores.”
Where baseline score inflation is not at issue, how do antidepressants measure up in the treatment of mild-to-moderate major depression? They work well. In trials that used the standard Hamilton scale, drug outperformed placebo with an effect size just over 0.5, medium. Looking at response rates, the number needed to treat was 4. On medication, typical patients ended up in remission, or close to it.
The Columbia meta-analysis has two important weaknesses.
Only forty-nine patients were on an SSRI—Prozac, at a high dose. This limitation is serious. But in head-to-head tests, SSRIs match imipramine, and the field has always believed that if imipramine outperforms the SSRIs anywhere, it is in the treatment of severe depression. SSRIs test out especially well with dysthymia. They mute neuroticism. They may affect people with no mood disorder at all. If other antidepressants treat mild-to-moderate major depression, almost certainly SSRIs do, too.
The second shortcoming—in some regards, it is a strength—is that the Columbia studies come from a single site, an urban university clinic. The trials tended to draw on a different population from the one recruited by drug-testing mills. Most had at least part-time jobs or sporadic employment and some college education. Patients might get vouchers for travel, but there was no van. And the experiment has the virtue of completeness. All relevant trials are included.
The central result is so strong that the severity hypothesis becomes moot. If antidepressants work yet better in sicker patients, then bless us all. But it may be that medication effects are uniform. In the Penn-Vanderbilt analysis, the number needed to treat for “very severe” depression was 4 or 5, effectively the same as the result for mild-to-moderate depression here. And we know that dysthymia responds at that level.
The Columbia results suggest that the problems we have been discussing—baseline score inflation, inadequate dosing, and inconsistencies in diagnosis—are substantial. Avoid the confounds, and uniform treatment effects emerge.
When the Columbia study appeared, no headlines read “Popular Drugs May Help in Mild-to-Moderate Depression” or “Antidepressant Lift Not All in Your Head.” Yet, if one possible conclusion—no utility for most patients—is important, then so is the other: medications are broadly effective.
* * *
It would be nice if other studies of this type existed—university-based trials testing Zoloft, Lexapro, and the rest, studies where, at entry, patients’ level of depression does not matter, studies that avoid baseline score inflation, studies that give us a shot at understanding whether low levels of depression respond to medication. If you think about it, we have heard mention of another sort of research that takes on all comers—in the discussion of my friend Alan and the usefulness of antidepressants for stroke survivors. There, everyone who suffers a certain sort of brain injury is invited to sign on. Patients don’t need to be depressed to participate. No one, patient or staff, has an incentive to exaggerate mood symptoms.
With Alan, my concern was motor function, which is why I had searched for the Lancet: Neurology study. On Prozac, stroke victims recovered more arm and leg movement. But the French neurologists also tracked symptoms of depression. Compared to patients given Prozac, those on placebo were four times as likely to develop depression. The antidepressant dose was the standard one, 20 milligrams of Prozac. In the medication group, the average change in the depression rating over three months was zero—no new symptoms, despite the stroke. (A few patients gained enough symptoms to qualify for a diagnosis of depression, but many more patients lost whatever depressive symptoms they had arrived with.) Patients on placebo proved much more vulnerable.
These outcomes speak against the severity hypothesis. At entry, the average patient did not qualify as depressed. Medication helped at the lowest severity—no depression.
The stroke study is typical. If we look beyond psychiatry to fields such as neurology and cardiology, in numerous outcome trials the patients are real and we take them as they are. The ticket in is an affliction, like coronary artery disease, that has a strong association with depression.
In a similar category is research on a medicine, interferon, used to treat infectious diseases (often hepatitis) and certain cancers. Interferon tends to cause depression. Antidepressants can prevent that development or interrupt it just as it emerges. Most interferon trials take on all comers—people with and without depression. There’s no incentive to exaggerate symptoms.
This research is important in its own right. In the face of medical illness, depression is destructive. To return to stroke for a moment: Think of two groups of patients, those who are depressed in the wake of a stroke and those who are not. Ten years later, more than three times as many in the first group will be dead. In life, patients with poststroke depression fail at “activities of daily living,” such as dressing themselves. They demonstrate more cognitive impairment, like problems with memory, than would be expected based on their brain injury. They are less likely to return to work. Even when the depression does not persist, the disability does.
And depression is common in stroke survivors. In the early weeks following the brain injury, a third of the patients will be depressed. Over a five-year period, half of stroke survivors will suffer depression. Once it appears, the depression tends to continue or recur.
Even after antidepressants had come to be widely prescribed, neurologists were cautious about using them in stroke patients. The risks were substantial—seizures, falls, delirium, and perhaps even bleeding and further damage to the brain—and any harm caused by stroke might prove unchangeable, not subject to influence through medication. But the need was great, so doctors undertook the research. On antidepressants, patients had less depression and better brain health overall.
A similar story appears repeatedly in what is called psychosomatics, the study of the overlap between disorders of body and mind: urgent need, substantial risk, tentative research—and then the discovery of widespread benefits from antidepressants.
To generalize about a large literature: In trials where depression is not required for entry, the average patient arrives with mild or no depression. On medicine, patients who start with symptoms tend to lose them. For those who begin free of depression, antidepressants prove protective. Three, four, or five times as many patients on placebo will enter a new episode of depression. When patients on placebo become depressed, the prompt initiation of antidepressants (the intervention is called rescue) ends the emerging episode. Often, what is at issue are “first depressions,” bouts of mood disorder in people who have never before had one.
The antidepressant controversy tends not to encompass these studies because psychosomatics is complex. For instance, SSRIs can act as blood thinners, enhancing heart health. Antidepressants may confer flexibility and resilience in the brain in ways that are only marginally related to depression. Perhaps medication succeeds with cardiac and neurologic patients for incidental reasons—because, mood effects (and side effects) aside, antidepressants protect body organs. As for interferon-induced depression, despite the similar list of symptoms, it may not be identical to run-of-the-mill, wear-and-tear, genetics-and-adversity depression. And preliminary evidence suggests that Prozac can enhance the antiviral effects of interferon—so fighting depression may not be the only action at issue. There’s a lot to debate.
Still, these studies are of high quality. The patients are real and unselected—all comers to a stroke ward or cardiac catheterization lab or (with interferon) hepatitis or cancer clinic. No one’s scouring single-room-occupancy hotels to corral candidates. The placebo conditions are vigorous, with everyone necessarily getting extensive medical attention. The targets for treatment include episodes of depression caught at the outset, as patients on interferon or placebo pills accumulate mood symptoms. SSRIs are the antidepressants most often used. They do not—at all—look like medicines that fail in the face of mild or acute illness.
Instead, it’s the severity hypothesis that fails. The most plausible studies, the ones likeliest to enroll real patients, contradict it. Cast a broad net, and you find that in experimental settings, antidepressants act as they do in clinical settings, helping patients with a wide range of conditions.