We found that gains with psychotherapy were maintained for up to 24 months. We also showed that patients who stopped medication remained stable. This is good news for the affected patients. However, as patients in the placebo groups also did not show deterioration we concluded that enduring effects observed in follow-up studies might be superimposed by spontaneous remission or effects of concurrent treatments.

For detailed inclusion criteria, we had referred to our previous meta-analysis.Reference Bandelow, Reitt, Rover, Michaelis, Görlich and Wedekind¹ As there are only a few head-to-head follow-up comparisons of psychotherapy and pharmacotherapy, we decided to calculate pre–post effects. Thus, we were able to include as many as 93 follow-up studies, which also comprised all head-to-head comparisons.

Pre–post effect sizes do not only measure ‘true’ treatment effects, but also natural course and placebo effects. However, when conditions are the same in psychotherapy and pharmacotherapy studies, this comparison is fair. Patients are mainly interested in a substantial decline in their anxiety scale scores, and this is what we measure with pre–post effects. Follow-up studies for psychotherapy mostly lack a control group, because a waitlist is used as a control during the active treatment period. After termination, the waitlist patients cannot serve as controls anymore because they will now receive the active treatment. Therefore, the pre–post effect size is the only option to compare the results of follow-up studies.

In every meta-analysis, not only in the ones using pre–post effect sizes, populations may be heterogeneous. However, all the included medication studies were comparisons with psychotherapy. Therefore, it is unlikely that our results were biased because of differences in patient populations.

It is also unlikely that allegiance effects in favour of pharmacotherapy have influenced our results, for 4 reasons: (1) In the data set of our previous meta-analysis,Reference Bandelow, Reitt, Rover, Michaelis, Görlich and Wedekind¹ on which the present analysis was performed, we found possible allegiance effects in both psychotherapy and pharmacotherapy studies but were unable to find significant differences between the average effect sizes of studies with or without allegiance effects for both treatment modalities. (2) The vast majority of the studies, including those involving medications, were published by behavioural psychotherapists. (3) The patents of all drugs mentioned in the study have expired for long. (4) We published the raw data of our analysis so that anyone who has the feeling that results might be biased can re-calculate the effect sizes. Allegiance effects are also possible in psychotherapy studies.Reference Munder, Brutsch, Leonhart, Gerger and Barth² We frankly disclosed our conflicts of interest, but this should also be expected from authors publishing in the field of psychotherapy, in particular when they are strongly promoting certain forms of psychotherapy, such as Dr Leichsenring, who is a fervent advocate for psychoanalysis and has been criticised for possibly biased meta-analyses in the literature.Reference Coyne, Bhar, Pignotti, Tovote and Beck³ Further, the authors seem to have overlooked the part in which we mentioned adverse effects of drugs. Enduring side-effects of medications that are used for anxiety disorders are rare, however.Reference Bandelow, Michaelis and Wedekind⁴ Furthermore, medication does not lessen cognitive–behavioural therapy (CBT) gains; we found much higher average effects for CBT plus medication (Cohen's d = 2.12) than for CBT alone (d = 1.22).Reference Bandelow, Reitt, Rover, Michaelis, Görlich and Wedekind¹