Meta-analyses of chronic disease trials with competing causes of death may yield biased odds ratios.