Seven RCTs were included in the review (633 caregivers, range 45 to 157). Follow-up ranged from four to twelve months. All RCTs were considered to have a low risk of bias; all RCTs used blinded outcome assessment.
There was no statistically significant difference in Zarit score between intervention and control groups when all studies were pooled together (MD -1.79,95% CI -4.27 to 0.69; seven RCTs), or in the subgroup analysis of studies using ITT analysis (MD -1.10,95% CI -2.25 to 0.05; five RCTs) or the subgroup analysis of studies that did not use ITT analysis (MD -4.46,95% CI -15.54 to 6.62; two RCTs). There was significant heterogeneity for each of these results.
In the sensitivity analysis, excluding three clinically heterogeneous RCTs, there was a statistically significant decrease in Zarit score, favouring the educational intervention group (MD -1.62,95% CI -2.16 to -1.08; four RCTs, no significant heterogeneity).