Twenty-nine articles were initially selected for review (n=8,873). The authors reported that they included 26 studies for meta-analysis; however, inspection of the forest plot indicated that 22 studies were included in the meta-analyses (n=8,362). Twenty-one studies were RCTs (n=8,322) and one trial was a pre-post test controlled study with a waiting list control (n=40). Quality scores were not reported. Methodological limitations identified in included studies were recruitment bias, use of intervention in control group, small sample sizes, high drop-out and lack of randomisation.
Depression: Psychosocial intervention (effect size 0.03, 95% CI -0.42 to 0.35; 15 studies), telephone support (effect size 0.07, 95% CI -2.62 to 2.75; two studies) and case management (effect size -0.32, 95% CI -0.73 to 0.09; three studies) did not significantly decrease feelings of depression in family carers of dementia patients compared with controls. There was evidence of significant statistical heterogeneity for psychosocial interventions (I2=54.4%) and case management (I2=83.1%).
Burden: Respite care significantly increased burden in family carers of dementia patients compared to controls (effect size 0.30, 95% CI 0.12 to 0.48; two studies). Psychosocial intervention did not significantly alter feelings of burden compared to controls (effect size -2.94, 95% CI -6.28 to 0.40; six studies). There was evidence of significant statistical heterogeneity for psychosocial interventions (I2=72.5%).
There was evidence of publication bias for studies of psychosocial interventions for both outcomes. It was not possible to assess publication bias for other interventions due to the small number of studies.