The diagnostic performance of rubidium-82 (Rb-82) positron emission tomography (PET) and thallium-201 (Tl-201) single-photon emission-computed tomography (SPECT) for detecting coronary artery disease was investigated in 81 patients (52 men, 29 women). PET studies using 60 mCi of Rb-82 were performed at baseline and after intravenous infusion of 0.56 mg/kg dipyridamole in conjunction with handgrip stress. Tl-201 SPECT was performed after dipyridamole-handgrip stress and, in a subset of patients, after treadmill exercise. Sensitivity, specificity and overall diagnostic accuracy were assessed using both visually and quantitatively interpreted coronary angiograms. The overall sensitivity, specificity and accuracy of PET for detection of coronary artery disease (greater than 50% diameter stenosis) were 84, 88 and 85%, respectively. In comparison, the performance of SPECT revealed a sensitivity of 84%, specificity of 53% (p less than 0.05 vs PET) and accuracy of 79%. Similar results were obtained using either visual or quantitative angiographic criteria for severity of coronary artery disease. In 43 patients without prior myocardial infarction, the sensitivity for detection of disease was 71 and 73%, respectively, similar for both PET and SPECT. There was no significant difference in diagnostic performance between imaging modalities when 2 different modes of stress (exercise treadmill vs intravenous dipyridamole plus handgrip) were used with SPECT imaging. Thus, Rb-82 PET provides improved specificity compared with Tl-201 SPECT for identifying coronary artery disease, most likely due to the higher photon energy of Rb-82 and attenuation correction provided by PET. However, post-test referral cannot be entirely excluded as a potential explanation for the lower specificity of Tl-201 SPECT.