Eight RCTs (total n=5,207; 2,146 with NICM) were included in the review. Five studies were primary prevention (total n=3,244; 1,854 with NICM) and three were secondary prevention (total n=1,963; 292 with NICM).
The authors stated that quality was comparable across the 8 studies. None of the studies described the method of allocation generation. Allocation concealment and blinding were not possible, but all trials used either an independent or blinded committee for the adjudication of events. Follow-up, when reported, ranged from 95 to 100%, and all studies used an intention-to-treat analysis. Wide variations in the crossover rates were reported.
Primary prevention (5 studies).
ICD therapy was associated with a statistically significant reduction in all-cause mortality compared with medication (RR 0.69, 95% CI 0.55, 0.87, P=0.002). There was no evidence of statistical heterogeneity (P=0.74).
Subgroup analyses.
A statistically significant reduction in all-cause mortality with ICD therapy, compared with medication, was still seen when a trial that recruited people with more advanced heart failure was excluded from the analysis (RR 0.74, 95% CI 0.58, 0.96, P=0.02), and when a study that recruited patients with heart failure of short duration and low control group mortality was also removed (RR 0.73, 95% CI 0.55, 0.96, P=0.03). When studies with 3 or more years' follow-up were removed from the analysis, there was still a statistically significant reduction in all-cause mortality with ICD therapy (RR 0.63, 95% CI 0.46, 0.88, P=0.006).
Secondary prevention (3 studies).
One study did not report results for NICM patients separately and was not included in the meta-analysis. Two studies found that ICD therapy was associated with a 31% reduction in all-cause mortality compared with medication, although this was not statistically significant (RR 0.69, 95% CI 0.39, 1.24, P=0.22). There was no evidence of statistical heterogeneity (P=0.66).
The authors reported that there was no evidence of publication bias (Begg adjusted rank correlation test, P=0.81).