Table 5.

Statistical analysis methodology (n = 43 unless otherwise stated)

	Number	Percentage
Analysis groups according to randomised groups
Yes	42	98
No	1	2
Unclear	0	0
Analysis adjusting for non-compliance carried out
Yes	8	19
No	35	81
Unclear	0	0
Compliance analysis type
As treated groups	3	7
Intervention completer subset analysis (full compliance)	2	5
Intervention completer subset analysis (partial and full compliance)	2	5
Randomisation-based analysis (method unclear)	1	2
N/A	35	81
Analysis adjusting for clusteringa
Yes	2b	5
No	41	95
Unclear	0	0

N/A not available

^aAnalyses carried out were mixed-effect models: one with two levels (intervention deliverer and) and one with three (site, intervention deliverer, and participant). ^bIn one study, the sample size was adjusted to account for clustering by using an intracluster correlation coefficient (ICC) of 0.02; for the other study the observed ICC of the higher level(s) was reported for the two primary outcomes (retention and engagement) under a two level (therapist and participant) and a 3 level model (site, therapist and participant) as 0.180 and 0.22 and 0.099 and 0.110 respectively