Quality criteria

We stated in the protocol our intention to collect information for the following quality criteria. These were collected but the headings for the quality criteria have now been updated in line with the quality criteria headings in the 'Risk of bias' table.

1. Adequate allocation concealment and sequence generation

2. Clear inclusion and exclusion criteria

3. No potential confounding variables

4. Intention‐to‐treat analysis used

5. < 25% lost to follow‐up

6. Blinded outcome assessment

Skewed data

In the protocol, it was stated that "if the standard deviation multiplied by two was greater than the mean it will not be added to the meta‐analysis". However, based on statistical advice this was thought to be overly stringent. Therefore, we have included these data (and carried out a sensitivity analyses as described above) but excluded data where the standard deviation was greater than the mean.

Meta‐analytic model

In the protocol for this review, it was stated that meta‐analysis of binary and continuous would be calculated using a fixed‐effect model but if there was significant heterogeneity, a random‐effects model was to be used. However, during the editorial process, the editors concluded that it was preferable to use random‐effects throughout.

Categorisation of psychological therapies

It became apparent to us during the review process that some of the psychological therapies described in included studies were not based on any clear theoretical domain, we made the decision to include these under the heading of 'Other psychological therapies'.