The competing risks methods used yielded results similar to
a traditional Cox analysis of identical data, with the notable
exception that the decrease in risk for MI with rosuvastatin among
participants with ischaemic heart failure was statistically significant
in the competing risks model and of borderline significance in
the traditional Cox analysis. In this competing risks analysis, study
participants who died of a non-atherothrombotic cause without having experienced a previous MI or stroke were incorporated
into the analysis whereas they would have been censored from
traditional analyses assessing atherothrombotic events. Different
rates of censored events between groups being compared (in this
case, rosuvastatin vs. placebo) can amplify the difference between
risks estimated by competing risks and traditional Cox approaches;
we observed this in a previous analysis of racial disparities in first CVD events in which the disparity between black and white men
in non-cardiovascular death was in part responsible for significant
differences in risks for ischaemic heart disease estimated by
competing risks vs. Cox analyses of identical data.20 However,
in this analysis, rates of competing non-atherothrombotic causes
of death were similar between rosuvastatin and placebo groups,
thereby minimizing differences between the competing risks and
traditional Cox approaches.