Six RCTs (n=1,202) were included in the review. Sample sizes ranged from 80 to 397.
The pooled hazard ratio for cardiovascular events showed a statistically significant difference in favour of antihypertensive drugs (0.69, 95% CI 0.56 to 0.84 using a fixed-effect model). However, there was evidence of significant statistical heterogeneity (I2=50.4%, p=0.073). Risk ratios for cardiovascular events and all-cause mortality also significantly favoured antihypertensive drugs.
Evidence from sensitivity analyses suggested that there was a significant difference (p<0.006) between studies that assessed only hypertensive patients and those that assessed populations that also included normotensive patients (HR 0.49, 95% CI 0.35 to 0.67; three studies for hypertensive patients only and HR 0.86, 95% CI 0.67 to 1.12; three studies for hypertensive and normotensive patients). Similar findings were reported when risk ratios were assessed for cardiovascular events, but differences were not statistically significant for all-cause mortality.
Egger's test and funnel plot suggested that there was a significant risk of publication bias.