Bandelow et al recently presented a meta-analysis testing the assumption that the effects of psychotherapy in anxiety disorders are more endurable than those of pharmacotherapy.Reference Bandelow, Sagebiel, Belz, Görlich, Michaelis and Wedekind¹ From non-significant differences between psychotherapy and pharmacotherapy in pre-follow-up effect sizes the authors concluded that ‘… patients who stopped taking a drug showed the same durable improvement as patients who stopped psychotherapy’.Reference Bandelow, Sagebiel, Belz, Görlich, Michaelis and Wedekind¹

Besides the severe (and properly discussed) limitation that an unclear percentage of patients may have started new psychological treatment or taken medications in the follow-up period, this meta-analysis raises further serious concerns.

First, the authors did not clearly specify their inclusion criteria. Apparently, they did not require head-to head comparisons of psychotherapy and pharmacotherapy as an inclusion criterion. Second, as a consequence, Bandelow et al compared pre–post and pre-follow-up effect sizes of psychotherapy, medication and placebo obtained from different randomised controlled trials. Thus, the studies being compared may differ with regard to important treatment moderators such as characteristics of patient populations and setting conditions. For these and other reasons analyses of pre–post and pre-follow-up effect sizes should be avoided in meta-analyses.Reference Cuijpers, Weitz, Cristea and Twisk²

Third, Bandelow et al did not adhere to the logic of equivalence testing that includes the definition of a margin compatible with equivalence and performing two one-sided tests (TOST).Reference Walker and Nowacki³ They apparently applied the more usual two-sided superiority test. However, concluding from a non-significant two-sided superiority test that two treatments (i.e. pharmacotherapy and psychotherapy) are equally efficacious (in the long-term) is questionable.Reference Walker and Nowacki³ The traditional two-sided test and TOST often yield inconsistent results.Reference Barker, Luman, McCauley and Chu⁴ Fourth, furthermore, Bandelow et al seem to have not controlled for researcher allegiance.Reference Mellers, Hertwig and Kahneman⁵ Thus, a bias in favour of pharmacotherapy cannot be excluded given that the first and last authors disclose multifold collaboration with pharmaceutical companies.

Finally and of note, the authors avoid discussing potential long-term negative effects that any type of psychotropic drug treatment, particularly after long-term use, may have, for example by increasing the risk of experiencing additional psychopathological problems that do not necessarily subside with discontinuation of the drug or of modifying responsiveness to subsequent treatments.Reference Fava, Benas and Cosci⁶

The data presented by Bandelow et al suggest that pharmacotherapy may have endurable effects in anxiety disorders as well. However, the authors’ conclusion that in the long-term term psychotherapy and pharmacotherapy are equally efficacious in anxiety disorders is questionable for the reasons given above.