The hyperventilation test has been used as a clinical tool to induce coronary spasm. However, its diagnostic and prognostic values have not been fully elucidated. This study was designed to establish the sensitivity and specificity of the hyperventilation test and to clarify the characteristics of hyperventilation test-positive patients. We examined 206 patients in whom coronary spasm was documented by angiography (spasm group), and 183 patients without angina at rest in whom acetylcholine failed to induce spasm (nonspasm group). All patients performed vigorous hyperventilation for 6 minutes in the early morning. Of the spasm group patients, 127 showed positive responses to the test, including ST elevation (n = 111), ST depression (n = 15) and negative U wave (n = 1). None in the nonspasm group showed any ischemic electrocardiographic change. Thus, the sensitivity and specificity of this test for diagnosis of coronary spasm were 62% and 100%, respectively. In the spasm group, there were no significant differences between hyperventilation test-positive and test-negative patients in age, sex, the prevalence of hypertension, diabetes mellitus, obesity, smoking, and the number of diseased vessels. When clinical characteristics were compared, the proportions of the patients with high disease activity >> or =5 attacks a week), with severe arrhythmias (second- or third-degree atrioventricular block and/or ventricular tachycardia) during attacks, and with multivessel spasm were significantly higher in the hyperventilation test-positive patients than in the negative patients (69% vs 20%, p <0.0001; 31% vs 11%, p <0.005; and 58% vs 34%, p <0.01, respectively). These findings imply that hyperventilation is a highly specific test for the diagnosis of coronary artery spasm, and that hyperventilation test-positive patients are likely to have life-threatening arrhythmias during attacks and multivessel spasm.