Network meta-analyses and treatment recommendations for obsessive-compulsive disorder

Petros Skapinakis and colleagues conducted an impressive network meta-analysis of 54 randomized controlled trials (RCTs) for obsessive-compulsive disorder (OCD).¹ Their analysis, like previous meta-analyses², supports treating OCD with SSRIs, clomipramine and cognitive-behavioural therapy variants (cognitive therapy, behavioural therapy, or cognitive-behavioural therapy). Psychotherapies alone or combined with medications demonstrated greater effects than medications alone (see their Table 2). However, the authors conclude, “Taking all the evidence into account, the combination of psychotherapeutic and psychopharmacological interventions is likely to be more effective than are psychotherapeutic interventions alone, at least in severe obsessive-compulsive disorder.” We believe this conclusion goes beyond their data and, if misinterpreted, could be harmful to those patients who do not want or need combination treatment.

First, their analysis included few trials comparing psychotherapies to medications or their combination to monotherapies. Although network meta-analysis compensates for an absence of direct comparison trials, direct comparisons offer a higher level of evidence. The authors report that the treatment with the highest efficacy (excluding waiting list-controlled trials) was behavioural therapy plus clomipramine (initiated simultaneously), though this treatment was tested in only a single trial. ³ They cite these data to support their conclusion that combination treatment is likely best, yet in that original RCT, behavioural therapy with clomipramine was directly compared to behavioural therapy alone, producing statistically indistinguishable outcomes.³

Second, the authors bolstered their conclusion that combination treatment is best by noting that many psychotherapy trials included medicated patients, meaning that medications could have contributed to the improvement attributed to psychotherapy. Although possible, this seems less likely in RCTs that randomized only those subjects taking stable medication doses for 3 months or longer. Of greater concern to us, these RCTs are, in fact, augmentation studies—recruiting medicated patients with clinically significant symptoms despite adequate medication trials. These data therefore are not ideal for informing first-line treatment decisions. The authors moderated their statement that combination treatment is best by adding “at least in those with severe” OCD, but they presented no supporting data. More severe OCD has been associated with worse SSRI response, but a meta-analysis of CBT trials (including augmentation studies) did not find baseline severity was associated with CBT effect size.²

Finally, using only efficacy points from a network meta-analysis to identify the so-called best treatment ignores crucial clinical issues such as safety, cost-effectiveness, and patient preferences.⁴ For example, SSRIs are not recommended first for pregnant women or young adults because of safety concerns. Psychotherapy may have greater long-term cost-effectiveness, as more patients achieve remission than with medications, and fewer experience relapse after treatment discontinuation.⁵ Finally, OCD patients have strong treatment preferences—some refusing medications and others psychotherapy. In the final analysis, the “best” treatment for any individual patient is not only efficacious but also safe and acceptable.

Taking all the evidence into account, medications and CBT variants treat OCD effectively, and some patients benefit from their combination. However, the data do not clearly demonstrate superiority of combination treatment over the most effective psychotherapies. Starting with psychotherapy is a rational first step for some patients to minimize unnecessary costs and risks of medication.