Eight RCTs (6,188 participants) were included. Study size ranged from 103 to 2,521 participants. Four trials were on people with coronary artery disease, three were on people with non-ischaemic heart disease and one was on both.
Tests showed no evidence of publication bias.
Pooling of data on all-cause mortality in people with coronary artery disease revealed considerable heterogeneity (I2=74.9%, five trials). Heterogeneity was reduced (I2=61.5%) when two trials that recruited participants who were recovering from acute myocardial infarction and participants who underwent coronary artery bypass surgery were removed. Results then showed an association between ICD therapy and a reduction in mortality (RR 0.67, 95% CI 0.51 to 0.88; three trials). In people with dilated cardiomyopathy, ICDs were associated with a reduction in all-cause mortality (RR 0.74, 95% CI 0.59 to 0.93, I2=0%; four trials). Pooled data from all six trials showed a reduction in mortality with ICDs (RR 0.73, 95% CI 0.64 to 0.82, I2=0%).
Compared to controls, ICDs were associated with a reduction in arrhythmic mortality (RR 0.40, 95% CI 0.31 to 0.50, I2=0%; eight trials).
The mean proportion of people who received appropriate ICD therapy from implanted devices was 23% (range 17.8% to 31.4%). Inappropriate therapy was observed in 16.5% of participants (six trials).