Pooled hazard ratios and 95% confidence intervals (CI) were calculated using a DerSimonian and Laird random-effects model.
For studies that directly compared total mortality and appropriate ICD therapy between men and women, HRs were adjusted for the maximum number of confounders and covariates and entered into multiple Cox regression analysis.
For the analysis of ICD survival benefit where there was no direct comparison between men and women, meta-regression within each gender was used to examine the influence of the potential confounders that were not accounted for in the primary studies: age, aetiology of cardiomyopathy, New York Heart Association functional class, left ventricular ejection fraction, race, baseline drug therapy and length of study follow-up.
Sensitivity analysis was conducted using a fixed-effect model and after excluding: each study in turn; studies that involved only ischaemic dilated cardiomyopathy patients; studies that involved a mix of ischaemic and nonischaemic patients; and the study that involved only nonischaemic patients.
Heterogeneity was assessed using the Cochrane Q statistic and the I2 statistic. The possibility of publication bias was explored using a funnel plot and Egger’s test.