The purpose of our study was to assess feasibility, safety and clinical utility of selective pulmonary angiography in patients with suspected pulmonary embolism and a nondiagnostic lung scan. The design was a prospective, descriptive study. The subjects were consecutive patients with clinically suspected pulmonary embolism and a nondiagnostic lung scintigram in whom pulmonary angiography was considered. Angiography was withheld in cases of manifest heart failure, renal failure, mean pulmonary artery pressure above 40 mmHg, or if there were compelling clinical reasons. All patients were followed-up for 6 months. The outcome measures were successful angiography, morbidity, mortality and recurrent pulmonary embolism in patients with normal angiogram in whom anticoagulants were withheld during 6 months of follow-up. Of 487 patients, 196 (40%) had nondiagnostic lung scan findings. In 46 patients (23%) pulmonary angiography was withheld. Pulmonary embolism was excluded in 105 patients (70%), and proven in 40 (27%) patients. In 5 (3%) patients the angiogram was inadequate for interpretation. No fatal complications were encountered [95% confidence interval (CI) 0-2.4%]. Nonfatal complications occurred in 3 patients (2%; 95% CI 0.4-6.0%); all recovered spontaneously. None of 105 patients with a normal angiogram returned with thromboembolism during follow-up (0%; 95% CI 0-3.4%). Pulmonary angiography is safe, rules out pulmonary embolism in two thirds of patients with a nondiagnostic lung scan and can be performed in almost 80% of these patients. It is safe to withhold long-term anticoagulants if a normal angiogram is obtained in this subgroup of patients with clinically suspected pulmonary embolism.