Objective: To evaluate, in a prospective design, whether migraine is a
n independent risk factor for subsequent stroke. Design: Evaluated as
part of the Physicians' Health Study, a randomized, double-blind, plac
ebo-controlled trial of aspirin and beta-carotene in the primary preve
ntion of cardiovascular disease and cancer begun in 1982. The aspirin
component of the study was terminated in 1988, with average follow-up
of 60.2 months. Setting: Conducted by mail among male physicians throu
ghout the United States.Participants: A total of 22 071 US male physic
ians aged 40 to 84 years in 1982 with no prior history of cancer or ca
rdiovascular diseases who were enrolled in the Physicians' Health Stud
y. Interventions: Participants were randomized to receive 325 mg of as
pirin or aspirin placebo every other day and to receive 50 mg of beta-
carotene or placebo on alternate days. Main Outcome Measures: The prim
ary outcomes of the Physicians' Health Study were cardiovascular disea
se and cancer. Because stroke was a main outcome, this provided the op
portunity to evaluate the association between migraine headaches and s
troke. Results: Physicians reporting migraine (n=1479) had significant
ly increased risks of subsequent total stroke and ischemic stroke comp
ared with those not reporting migraine. After adjustment for age, aspi
rin and betacarotene treatment assignment, and a number of cardiovascu
lar risk factors, the relative risks were 1.84 (95% confidence interva
l, 1.06 to 3.20) for total stroke and 2.00 (95% confidence; interval,
1.10 to 3.64) for ischemic stroke. There were too few hemorrhagic stro
kes in the study to evaluate this end point. No associations were seen
between ordinary nonmigraine headache and subsequent stroke or betwee
n migraine and subsequent myocardial infarction or cardiovascular deat
h. Conclusion: These data raise the possibility that vascular events a
ssociated with migraine map also have causative importance in stroke b
ut require confirmation in other studies specifically designed to eval
uate this question.