OBJECTIVE--To characterise clinical, investigative, and prognostic features of women referred with chest pain who subsequently underwent coronary angiography. DESIGN--Analysis of all women with angina referred to one consultant during 1987-91 who subsequently underwent coronary angiography, with follow up to present day. SETTING--Cardiothoracic centre. SUBJECTS--Women with normal coronary arteries; women with coronary artery disease shown on angiography; men with coronary artery disease matched for age; men referred with chest pain during the same period subsequently found to have normal coronary arteries. MAIN OUTCOME MEASURES--Risk factor analysis; results of exercise testing and coronary angiography; intervention; morbidity and mortality. RESULTS--Women comprised 23% (202/886) of patients referred with chest pain who subsequently underwent angiography. 83/202 women had normal coronary angiograms compared with 55/684 men (41% v 8%, P < 0.01). Diabetes mellitus was the only risk factor more frequently encountered in women with coronary artery disease (P = 0.001). The specificity and positive predictive value of exercise testing before angiography were significantly lower in women than men (71% v 93%, P < 0.001 and 76% v 95%, P < 0.001, respectively). Revascularisation procedures were as common in women with coronary artery disease as in men (81 (68%) v 70 (59%)), and there was no difference in event rate during follow up. Many patients with normal coronary arteries, irrespective of sex, had symptoms during follow up (61 (73%) women, 36 (65%) men) and continued to take antianginal drugs (27 (33%) women, 14 (28%) men); 14 (17%) women and six (11%) men required hospital readmission for severe symptoms. CONCLUSIONS--In this series, although women comprised the minority of patients referred with chest pain, a diagnosis of normal coronary arteries was five times more common in women than men. Risk factor analysis and exercise testing were of limited value in predicting coronary artery disease in women. There was no sex bias regarding revascularisation procedures, and outcome was similar. A diagnosis of non-cardiac chest pain in patients with normal coronary arteries was of little benefit to the patient with regard to morbidity.