ACTUALLY, IT WAS otherwise. With the newspaper article’s appearance, all hell broke loose. Scientists who served as advisers to the trial complained that the preliminary report misrepresented the research findings. Because of disputes over statistical methods, it would be 1989 before the study’s results appeared in a professional journal.
The Treatment of Depression Collaborative Research Program had one purpose, planners had written: “to advance the scientific study of the effectiveness of different forms of psychotherapy.” The organizers looked for therapies standardized for use as brief treatments. Two were available: interpersonal psychotherapy, or IPT, and cognitive behavioral psychotherapy, or CBT.
IPT represented the Freudian strain in psychology, therapies aimed at emotional self-understanding. CBT, with roots back to Pavlov, relied on educational approaches to help patients correct maladaptive thoughts. This one trial, of CBT and IPT for depression, would stand in for Morris Parloff’s eighteen thousand experiments. The overarching question was, Does psychotherapy work?
The planners argued over the selection of a control condition. Ideally, IPT and CBT would have been compared to fifty-minute conversations with experts who had been trained not to bolster patients’ interpersonal functioning or reasoning processes, but no one had designed a placebo psychotherapy. Also, having pressed psychotherapists to test their practices, pharmacologists now worried that the NIMH trial would be a put-up job, enlisting a healthy population, not real patients. The pharmacologists asked for what would later be called a comparator arm.
A comparator—you can stress either the first or second syllable—is an established treatment that serves to “validate the sample.” In a trial with properly diagnosed patients, the comparator will produce an expected level of change. If the group is unrepresentative, the comparator will fail, and the trial results can be dismissed.
The obvious comparator was imipramine—in the organizers’ words, “The best reference drug to use would be the drug with the longest history of use and for which a large amount of efficacy data exists with this patient population.” The control condition would be a placebo pill. But in the 1980s, “brief psychotherapy” lasted at least sixteen weeks, an eternity for depressed patients. To minimize suffering and prevent suicides, the researchers decided to supplement placebo administration with emotional support. The patients on pills (placebo or imipramine) would attend sixteen or more twenty-to-thirty-minute “advice and encouragement” sessions with experienced psychiatrists.
The NIMH researchers wrote that this clinical management was “not a no-treatment condition or an ‘inactive’ placebo condition.” They considered it “minimal supportive therapy.” The question under study was whether IPT and CBT could outperform modest nonspecific counseling.
The answer was, yes and no. When the experiment had been run and the numbers crunched, the NIMH researchers reported that the data showed a “consistent ordering of treatments at termination, with imipramine plus clinical management generally doing best, placebo plus clinical management worst, and the two psychotherapies in between but generally closer to imipramine plus clinical management.” The details told a more particular story. Interpersonal psychotherapy had performed better than placebo and almost as well as imipramine. Cognitive behavioral therapy had failed. CBT had been contrasted to placebo-plus-support through twenty or thirty statistical analyses. On none did the psychotherapy demonstrate superiority at a level that achieved statistical significance.
One of the trial’s advisers was Donald Klein, a pharmacology pioneer I had met while working for Gerry Klerman and whose theories I would rely on in Listening to Prozac. Klein summarized the NIMH results in provocative fashion: “It is not simply that imipramine is better, faster, and cheaper than CBT, but that the whole basis for the belief that cognitive psychotherapy is doing anything specific has been placed in jeopardy.”
The TDCRP had been designed as the finest outcome trial that research technology would allow. Do we imagine that, in response, cognitive behavioral therapists began (à la Gerry Klerman) to warn patients that CBT might work no better than supportive conversations? We do not.
Psychologists argued that the trial did not contrast psychotherapy to “treatment as usual” but rather to a supportive therapy that was not, in the end, all that minimal. Some of what CBT provided (expert attention, reassurance, direction) was duplicated in the placebo arm, and when the benefits of that care were subtracted away, CBT proved to offer little extra.
Then there was the matter of enrollment. Interviews conducted during the sign-on found that what volunteers wanted was free psychotherapy. They needed a Hamilton score of 14 to enter the trial. Perhaps people who were not depressed had listed false symptoms and then given the impression that they were “responding” to placebo (or any offering) by giving honest answers once the trial began. Consultants who had seen the raw data found evidence of “baseline score inflation,” as this problem is called, with high Hamilton scores plummeting within days. The resulting puffed-up response rate in the placebo arm was hard to compete with.
Defenders of CBT also questioned the study’s implementation. At two sites, CBT appeared to have been performed in rote fashion. Perhaps early critics of the trial had been right: manualized therapies could not duplicate the real thing, therapy built around creative or spontaneous efforts with patients.
But finally, cognitive therapists set the results in context and went about their business. Although the enrollment, sixty patients per arm, was large for the time, TDCRP was hardly of gold standard dimensions, and prior small studies had been favorable. Adherents of CBT might be advocates of controlled trials, but in defense of a treatment they found useful in practice, they listed endless reasons to keep the faith, and likely they were right. Subsequent research has suggested that CBT performs as well as other psychotherapies.
The NIMH trial had paradoxical results. It was designed to test psychotherapies, and CBT bombed. But previously, cognitive behaviorism had been a bit player, a curiosity dear to hyperrational types for whom psychoanalysis represented mysticism. Now, because a team of experts had taken CBT seriously, the field did. CBT failed to outperform placebo—and gained legitimacy.
Imipramine’s fate was equally paradoxical. The drug did its job, validating the patient sample, but in time the results of the collaborative study would be used to foster doubt about antidepressants.