Table 1.

Questions in the checklist for assessing moderators and predictors of treatment effects. Note that questions 10–12 are listed also as questions 13–15, as they are applicable both to individual studies and bodies of evidence covering the same moderator or predictor

Criteria for individual studies  Design

1. A priori plausibility: was there sufficient empirical or theoretical support for the moderator or predictor that was examined?   2. Was the moderator or predictor specified a priori?   3. Was the moderator or predictor variable measured before the allocation or start of the intervention?   4. Was measurement of the moderator or predictor reliable and valid in the target population?

Analysis

5. In case of a moderator, was an interaction test used?   6. Was a limited number of moderators and predictors tested?   7. Was sample size adequate for the moderator or predictor analysis?

Results

8. Were results presented for all candidate moderators or predictors that were examined?   9. Did statistical tests or confidence intervals indicate that observed moderator or predictor effects were unlikely to be merely due to chance variation?   10. Was the moderator or predictor effect consistent with related moderators or predictors, or across related outcomes measured within the study?

Transferability

11. Were the setting and study population comparable to the setting and population in which the information would be used?   12. Is the moderator or predictor effect clinically important?

Criteria for bodies of evidence (systematic review or related sets of studies)

13. Was the moderator or predictor effect consistent with related moderators or predictors, or across related outcomes measured between the studies?   14. Were the setting and study population comparable to the setting and population in which the information would be used?   15. Is the moderator or predictor effect clinically important?   16. Was the moderator or predictor effect reasonably homogenous across studies?   17. Was the moderator or predictor measured similarly across the included studies, or was an adequate conversion performed?