Table 5.

Risk of Bias Moderator Analyses

Moderator	k	Effect size	Standard error	Qbetween	p (Qbetween)
Allocation to treatment groups				1.322	.250
Randomization of individual subjects	74	0.267	0.047
Subjects not individually randomized	34	0.363	0.069
Allocation concealment				0.080	.777
Allocation concealed	18	0.328	0.112
Did not report allocation concealed	90	0.362	0.048
Data collector masking				2.930	.087
Data collectors masked to group assignment	29	0.210	0.057
Did not report data collectors masked to group assignment	79	0.340	0.500
Intention-to-treat approach				0.061	.805
Reported intention-to-treat approach	23	0.277	0.098
Did not report intention-to-treat approach	85	0.304	0.044
Publication status				0.417	.518
Published articles	96	0.310	0.043
Dissertations, presentations, unpublished reports	12	0.230	0.116
Adherence measured with medication event monitoring system				2.925	.087
Medication event monitoring system	10	0.621	0.202
Did not use medication event monitoring system	98	0.269	0.038
Adherence measured with pharmacy refill data				0.001	.974
Pharmacy refill data	13	0.299	0.090
Did not use pharmacy refill data	95	0.303	0.046
Adherence measured with pill counts				0.001	.979
Pill counts	23	0.299	0.078
Did not use pill counts	85	0.297	0.046
Adherence measured by self-report				2.349	.125
Self-report adherence data	46	0.232	0.059
Did not use self-report data	62	0.352	0.052

k denotes number of comparisons, effect size is standardized mean difference, Q is a conventional heterogeneity statistic.