You are assuming the pre-intervention incidence of each type of stroke is a guide to the effect size of the intervention on each, which isn't, as a rule, a justified assumption.