The results of the 19 studies examining the association between estrogen replacement therapy and cerebrovascular disease are inconsistent. Although all 7 studies of death from stroke found a 20% to 60% reduction in risk among estrogen users relative to nonusers, studies of incident stroke and subarachnoid hemorrhage in particular are conflicting, with RRs from 0 to 2.3. Although a protective effect of estrogen on stroke is biologically plausible, information regarding effects of dose, duration, and discontinuation of estrogen is limited and contradictory. Additionally, selection bias, recall bias, and confounding cannot be completely discounted. Further studies are needed to determine estrogen's biological effects; the effects of varying doses, durations, and types of estrogen; the consequences of combined estrogen/progestogen regimens and of other routes of administration; and whether other factors (such as age and smoking) modify or confound estrogen's effect on stroke risk.