Two RCTs (n=130) were included.
Both studies were of poor quality. Methodological flaws included: lack of description of treatment regimens; inadequate information on patient characteristics; lack of clarity about the blinding of assessments; potential detection bias in methods used to assess compliance; potential observer bias in the assessment of symptoms; lack of information on the uptake of booster sessions; lack of power calculations; inadequate information on the scales used to assess compliance; lack of clarity about the handling of drop-outs; and potentially biased methods used to deal with drop-outs.
Both studies recruited patients from limited geographical areas and the generalisability of the results was limited.
One unblinded RCT (n=74 at baseline, n=66 at 12 months) reported that compliance therapy significantly improved compliance at 12 months compared with control. The ratings on a scale of 1 to 7 were 3.7 (SD=1.2) at baseline, improving to 5.5 (SD=1.8) at 12 months, with compliance therapy versus 4.1 (SD=1.2) at baseline, worsening to 3.6 (SD=2.1) at 12 months, with control; the mean difference the 19%. One subsequent report stated that the P-value for treatment difference was less than 0.001.
The other single-blinded RCT (the review author stated there were inconsistencies in the number analysed at 1 year: n=56 at baseline and 6 dropped out during follow-up, but 56 were analysed) reported no statistically significant difference between compliance and control for compliance. The OR of being compliant at baseline was 2.267 (95% CI: 0.47, 11.41), favouring the group allocated to compliance therapy, while the OR of being compliant at 12 months was 0.65 (95% CI: 0.2, 2.11), favouring the compliance treatment group.
Neither study reported any statistically significant difference in symptoms between treatments. One study reported a mean BPRS (range: 7 to 49; reduction suggests improvement) of 20.3 (SD=7.6) at baseline, improving to 13.8 (SD=6.3) at 12 months, for the group allocated to compliance therapy versus 19.2 (SD=6.6) at baseline, improving to 15.3 (SD=6.2) at 12 months, for the group allocated to control. The second study reported a mean PANSS (range: 7 to 210; reduction suggests improvement) of 71 (SD=22) at baseline, improving to 58.2 (SD=17) at 12 months, for the group allocated to compliance therapy versus 66 (SD=17) at baseline, improving to 52.1 (SD=21) at 12 months, for the group allocated to control; this resulted in a non significant difference of 6.1 (95% CI: -4.7, 16.9, P=0.26) between treatment groups at 12 months.