Medications are more effective than therapy.”
“Therapy is just as effective, but lasts longer.”

Depending on your biases, you can probably find evidence to support either of the above propositions.

Dozens of trials compare cognitive behavior therapy (CBT) to a waiting-list control group, and hundreds of trials examine how antidepressants compare to placebo. Yet not many trials directly compare CBT to antidepressant medication. Meta-analytic reviews generally find that both antidepressants and psychotherapy offer similar efficacy in the short-term, but that after treatment discontinuation, results are better with CBT.

However, nearly all prior meta-analyses suffer from a major problem: the included trials mostly used older (vs newer) drugs. This choice could give an “unfair” advantage to psychotherapy, because the heavy side effect burden of older medications might cause early dropouts, which are less common in newer antidepressants. Given that 90% of patients on antidepressants are taking newer drugs now, it seems dicey to assume that findings comparing psychotherapy with older medications would yield similar results to trials comparing therapy with newer medications (Olfson M et al, Arch Gen Psychiatry 2009;66(8):848–856). Until recently, only one meta-analysis specifically examined how psychotherapy, mostly CBT, fared versus SGAs, SSRIs in particular, and it found a very small but statistically significant advantage for SSRIs (Cuijpers P et al, J Clin Psychiatry 2008;69(11):1675–1685).

However, two problems remain: 1) several relevant trials have been published subsequent to the Cuijpers meta-analysis, and 2) the quality of the treatments used in clinical trials requires close consideration. To address these concerns, my research team conducted a meta-analysis examining only studies comparing psychotherapy, mostly CBT, to newer antidepressants. While these medications had a slight advantage over therapy at the end of treatment, when therapies provided by primary care physicians, nurses, or therapists with unclear training were removed from consideration, psychotherapy and newer antidepressants had equivalent outcomes in the short term across 14 comparisons (Spielmans GI et al, J Nerv Ment Dis 2011;199:142–149). At longer-term follow-up, psychotherapy by properly trained therapists outperformed newer antidepressants by a small, but statistically significant, margin.

One study we did not include, because it allowed patients taking newer antidepressants to take other drugs, found CBT superior to medication for prevention of depressive relapse. In this study, 240 patients with severe depression were randomly assigned to either CBT or a newer antidepressant (augmented with other medication, if needed). After 16 weeks of treatment, the response rates were nearly identical: 58.3% on CBT vs 57.5% for the antidepressant. So far, no big surprise—CBT and meds are usually equivalent over the short term in depression.

Here’s where things got interesting. Patients who responded to treatment continued in the study, and were randomized to three possible conditions: Those who responded to the antidepressant were randomly assigned to continue on it, or were switched to placebo, while those who responded to CBT stopped therapy, aside from three “booster sessions,” over the next 12 months. The majority of patients (76%) withdrawn from the antidepressant relapsed to depression over the year, while only 31% of those withdrawn from CBT relapsed. Patients who had prior CBT even had lower relapse rates than a group of patients randomized to continuous antidepressant treatment (relapse rate of 47%), though the difference was not statistically significant. At least for these patients, it appeared that a brief course of CBT taught them skills to fend off future misery, while medication did not (Hollon SD et al, Arch Gen 2005;62(4):417–422).