Debunking CBT, Part 2: What's It Good For?
What is cognitive-behavioral therapy good for?
Posted Aug 18, 2008
What is cognitive-behavioral therapy good for? Is it an all-purpose treatment, useful for the “neurotic” states for which patients have traditionally consulted psychotherapists? Or is CBT most like its more mechanical parent, behaviorism, a tailored intervention best suited to afflictions in which disordered actions play a prominent role?
In an earlier posting, I looked at the summary results from an authoritative meta-analysis, or statistical amalgamation, of research on CBT in the treatment of anxiety disorders. My read was that while CBT had been shown to work, the integrated findings exposed the therapy as either untested or fairly disappointing in the treatment of the very conditions it had been designed for.
Today, I propose to return to the meta-analysis and ask: where exactly has CBT been shown to work?
Looking at specific diagnoses, Stefan Hofmann and Jasper Smits, the authors of the overview, found that CBT was most effective for two diagnoses: obsessive-compulsive disorder (or OCD) and acute stress disorder (ASD). But that conclusion is based on meager data.
No study of OCD met Hofmann and Smits’s strictest criteria for scientific merit, and only one study met their second-level, less rigorous standards. That research did not really employ CBT. It tested a behavioral method in which patients were confronted with an anxiety-provoking stimulus (like touching a dirty object) and then prevented from enacting their compulsive response (like hand-washing). "Exposure and response-prevention" is a known, effective treatment for OCD, although to be fair, training patients not to enact their main symptoms and then (as an outcome criterion) measuring symptomatic behaviors is a fairly sure way of achieving statistical significance.
Then, too, because there are few placebo responses in OCD, it is an illness in which many standard treatments, including antidepressant medications, are shown to good effect. And of course, a meta-analysis that is based on only one study does not add much to the scientific literature; there is no data that needs combining. Instead of announcing “meta-analysis confirms that CBT treats OCD,” it would be as well to say that there is still one fairly good study that says a related treatment works for the indication.
Similarly, only one ASD study met the authors’ inclusion criteria.
For those unfamiliar with ASD, the condition is a troubled response to recent stress. Its main importance is as a risk factor for a more substantial affliction, post-traumatic stress disorder or PTSD. ASD is one of those diagnosable conditions that lead to criticism of the Diagnostic and Statistical Manuals. Is it an illness or not?
Only one research group seems to have looked at CBT for ASD. Not surprisingly, the scientists found that focusing on a person’s distorted perceptions of an event diminishes the event’s impact. Still, no one knows whether ASD sufferers resemble traditional candidates for psychotherapy.
It turns out that the only anxiety disorders that have been at all well studied, in terms of response to CBT, are PTSD and panic anxiety. For panic, Hofmann and Smits found two or three top-flight research trials; for PTSD, one or two. Two other studies, one of social anxiety disorder and one of generalized anxiety disorder, met the authors’ second-rank quality criteria. In other words, when it comes to the treatment of anxiety, there is surprisingly little basis for assessing CBT. As for outcomes, the efficacy for these common conditions was mostly at the weaker end of the range, significantly less than what was reported for the behavioral treatment of OCD. For conditions like PTSD, the strongest results came in the less rigorous studies.
To be fair, the problem here is mostly “rigor.” Hofmann and Smits are looking for “intention to treat,” or ITT, analyses: if you enter a study, what are the odds that you will respond to CBT? Most early trials were reported via “completer” analyses: if you follow through with all the sessions and fill out all the questionnaires, what are the odds that you will have improved with treatment?
One reason that both psychotherapy and psychopharmacology have looked good, over most of the past half-century, is that scientists accepted “completer” studies. After all, what you as a consumer want to know is, if I follow my doctor’s recommendation, will I get a good result?
Unfortunately, completer studies do not quite answer that question. People who are floundering are more likely to drop out of the study; perhaps they are especially likely to drop out of the more onerous arm, the one that (in the case of psychotherapy) makes psychological demands or (in the case of medication) causes side effects. You could make the opposite argument, that people who believe they are in active treatment are more likely to see things through. But generally, completer trials are thought to be biased in favor of the intervention under study. If completer studies are looking at a select sample — of people who seem to be making progress throughout the trial — then of course they will show that the treatment works.
So if you are stoical, if you would stick with any treatment right up to the end of an eight- or twelve-week trial, then the result you achieve will likely sit somewhere between the completer and intention-to-treat outcomes. Seen through the lens of ITT trials, CBT is unimpressive. If you believe that completer trials contain some of the truth, then you are likely to think more highly of CBT; but then, you will also think more highly of other psychotherapies and of other approaches to anxiety disorders, like medication.