Meta-analysis of Blood Pressure Reduction

All statistical analyses were done using Stata software. We combined relative risk estimates of disease events from individual trials using a random effects model31 (link) (which avoids assuming that participants in the individual trials in the meta-analysis are sampled from populations in which the intervention has the same quantitative effect). Summary relative risk estimates from blood pressure difference trials were standardised to a blood pressure reduction of 10 mm Hg systolic or 5 mm Hg diastolic, by raising the relative risk estimate in each trial to the appropriate power (10 divided by the observed reduction in systolic blood pressure or 5 divided by the observed reduction in diastolic pressure)—for example, if the relative risk was 0.7 and the reduction in systolic blood pressure was 8 mm Hg, the standardised relative risk estimate was 0.64 (0.7^1.25, since 10/8=1.25). If reductions in both systolic and diastolic blood pressures were reported (as in most trials), we took the average of the two risk estimates (more strongly predictive than either alone25 (link)). As the reduction in blood pressure was not reported in most trials of people with a history of CHD, we estimated the average reduction from the average blood pressure before treatment and the average drug dose (as a multiple of standard dose32 (link)
33 ), using results from a meta-analysis in which the effect of pretreatment blood pressure and dose on blood pressure reduction was quantified.32 (link) The estimated blood pressure reduction was 5.9 mm Hg systolic and 3.1 mm Hg diastolic, close to the median reduction in the 27 trials in which blood pressure reduction was reported, which was 6 mm Hg systolic and 3 mm Hg diastolic.