Table 3.
Standard mean differences for studies with a “low risk” of bias and studies with “any risk” of bias within each of the seven risk domains.
Low risk | Any risk | |||||||
---|---|---|---|---|---|---|---|---|
EFFECT | 95% CI | k | EFFECT | 95% CI | k | X2 | P | |
Short term | ||||||||
Selection bias | −0.08 | −0.28 to 0.11 | 7 | −0.19 | −0.49 to 0.11 | 10 | 0.35 | 0.55 |
Performance bias | – | – | – | −0.11 | −0.28 to 0.07 | 17 | – | – |
Treatment fidelity | −0.10 | −0.20 to 0.01 | 4 | −0.14 | −0.40 to 0.11 | 13 | 0.12 | 0.73 |
Detection bias | −0.07 | −0.27 to 0.13 | 13 | −0.23 | −0.59 to 0.13 | 4 | 0.58 | 0.45 |
Statistical bias | −0.11 | −0.52 to 0.29 | 6 | −0.10 | −0.28 to 0.07 | 11 | <0.01 | 0.96 |
Reporting bias | −0.14 | −0.35 to 0.06 | 15 | 0.11 | −0.11 to 0.33 | 2 | 2.66 | 0.10 |
Generalisability bias | −0.02 | −0.14 to 0.09 | 8 | −0.25 | −0.62 to 0.12 | 9 | 1.35 | 0.25 |