Although sleep apnea (SA) appears to be a cardiovascular risk factor, little is known about its frequency in patients with transient ischemic attack (TIA) and stroke. We prospectively studied 59 subjects (26 women and 33 men; mean age, 62 years) with stroke (n = 36) or TIA (n = 23) with the use of a standard protocol that included assessment of snoring and daytime sleepiness (Epworth Sleepiness Score [ESS]), a validated SA score (Sleep Disorders Questionnaire [SDQ-SA]), and a severity of stroke score (Scandinavian Stroke Scale [SSS]). SA was considered clinically probable (P-SA) when habitual snoring was associated with an ESS of > 10 or when SDQ-SA score was > or = 32 in women and > or = 36 in men. Polysomnography (PSG) was obtained in 36 subjects (group 1) a mean of 12 days after TIA or stroke. In 23 subjects (group 2), PSG was not available (n = 11), refused (n = 10), or inadequate (n = 2). Clinical and PSG data were compared with those obtained in 19 age- and gender-matched control subjects. Groups 1 and 2 were similar in mean age (61 versus 64 years), type of event (36% versus 44% TIA), reported habitual snoring (58% versus 52%), and P-SA (58% versus 50%). PSG showed SA (Apnea-Hypopnea Index [AHI], > or = 10) in 25 of 36 subjects (69%). The proportion of subjects with SA was similar in the TIA and stroke groups (69% versus 70%) and was well above the frequency found in our control group (15%). An AHI of > or = 20 and a minimal oxygen saturation of < 85% were each found in 20 of 36 subjects (55%). Gender and age did not correlate with severity of SA. Subjects with habitual snoring, P-SA, or severe stroke (SSS of < 30) had a significantly higher AHI (p < 0.05). The sensitivity of P-SA for SA was 64%, and the specificity was 67%. We conclude that SA has a high frequency in patients in the acute phase of TIA and stroke and SA cannot be predicted reliably on clinical grounds alone but is more likely in patients with habitual snoring, abnormal SDQ-SA, or severe stroke.