. 2006 Jul 1;333(7557):15. doi: 10.1136/bmj.38875.675486.55

Table 2.

Subgroup analysis of active treatment arms considered beneficial

			Tests for heterogeneity
Analysis group	No of studies	Pooled odds ratio (95% CI)	P value (Q statistic)	I²(%)
Active treatment arm considered beneficial	19	0.55 (0.49 to 0.62)	0.71	0
Post-myocardial infarction studies only^{w1-w3 w5-w8}	7	0.52 (0.41 to 0.66)	0.96	0
HIV studies only^w9-w15	7	0.53 (0.41 to 0.69)	0.06	50.2
Primary prevention studies only^{w16 w17}	2	0.58 (0.46 to 0.73)	0.99	0
Method used to measure adherence:
Objective method (pill count, pharmacy refill, blood level, medication event monitor system)^{w1-w3 w5-w8 w11 w13-w15 w17 w19-w21}	15	0.53 (0.46 to 0.60)	0.99	0
Subjective method (patient self report, clinician impression)^{w9 w10 w12 w16}	4	0.55 (0.37 to 0.83)	0.05	61.0
Threshold used to define good adherence group:
≥75%^{w1-w3 w5 w7 w11 w13-w17 w19}	12	0.54 (0.48 to 0.61)	0.97	0
≥80%^{w1 w5 w7 w11 w14 w16 w19}	7	0.58 (0.50 to 0.68)	1.00	0
≥90%^{w11 w14 w16}	3	0.60 (0.49 to 0.73)	0.91	0
Study design:
Randomised controlled studies only^{w1-w3 w8 w16 w17}	6	0.55 (0.46 to 0.65)	0.95	0
Cohort studies only^{w5-w7 w9-w15 w19-w21}	13	0.55 (0.47 to 0.64)	0.36	8.6