Outcome | No. of RCTs | No. of participants | Length of follow-up, d median (IQR) | Risk of bias | Inconsistency (I2) | Indirectness | Imprecision | Small-study effects | SMD (95% CI) | Certainty of evidence |
---|---|---|---|---|---|---|---|---|---|---|
Primary | 51 | 5384 | 180 (90–252) | Serious* | No serious inconsistency† (52%) | No serious indirectness | No serious imprecision | Undetected‡ Egger p = 0.37 | −0.02 (−0.11 to 0.07) | Moderate |
Note: CI = confidence interval, IQR = interquartile range, SMD = standardized mean difference.
↵* All RCTs administered the same intervention in both treatment arms (CBT); however, patients and health care providers were unblinded to the method of delivery (remote or in-person CBT).
↵† Although the I2 value showed moderate heterogeneity, we did not rate down the certainty of evidence because the magnitude and direction of effects were largely consistent across trials, and a substantial proportion of between-study variability was contributed by 1 trial (56) that contributed less than 2% of the weight to our pooled estimate.
↵‡ A contoured-enhanced funnel plot showed no evidence of small study effects (Appendix 1, eFigure 3), and Egger’s test was nonsignificant.