Table 1

Meta-analytic strategies and stratifications

Stratified analyses	Number of comparisons	SMD (95% CI)	p Value	τ²	I²*	p Value for interaction
Random effects model (strategy 1)	224	0.58 (0.49 to 0.68)	0.0000	0.41	80%	–
25% largest trials (strategy 2)	57	0.45 (0.26 to 0.64)	0.0000	0.48	82%	–
SE as covariate (strategy 3)	224	0.65 (0.57 to 0.73)	0.0001	0.26	(51%)†	–
Infinite trial size (SE=0) (strategy 4)	224	−0.51 (−0.74 to −0.28)	0.0000	0.26	(51%)†	–
Comparator				0.24		0.004
Sham	76	0.48 (0.35 to 0.61)	0.0000
No intervention	148	0.73 (0.64 to 0.83)	0.0000
Sequence generation				0.26		0.43
Low risk	161	0.67 (0.58 to 0.77)	0.0000
Unclear	55	0.60 (0.43 to 0.76)	0.0000
High risk	8	0.42 (−0.01 to 0.85)	0.0551
Allocation concealment				0.26		0.37
Low risk	114	0.70 (0.59 to 0.82)	0.0000
Unclear	28	0.58 (0.35 to 0.81)	0.0000
High risk	82	0.59 (0.46 to 0.72)	0.0000
Blinding participants				0.23		0.00
Low risk	40	0.34 (0.16 to 0.52)	0.0002
Unclear	47	0.60 (0.43 to 0.77)	0.0000
High risk	137	0.75 (0.66 to 0.85)	0.0000
Blinding personnel				0.25		0.06
Low risk	12	0.31 (−0.03 to 0.64)	0.0728
Unclear	12	0.86 (0.53 to 1.19)	0.0000
High risk	200	0.65 (0.57 to 0.74)	0.0000
Incomplete outcome data				0.26		0.86
Low risk	80	0.63 (0.50 to 0.77)	0.0000
Unclear	50	0.69 (0.52 to 0.86)	0.0000
High risk	94	0.64 (0.51 to 0.76)	0.0000
Selective outcome reporting				0.26		0.27
Low risk	26	0.78 (0.55 to 1.00)	0.0000
Unclear	171	0.65 (0.56 to 0.74)	0.0000
High risk	27	0.51 (0.28 to 0.74)	0.0000

*Calculated in Review Manager.
†I² Derived from the overall I² and τ². All other values were calculated in SAS.
SMD, standardised mean differences; τ² (τ²), estimated between study variance.