The presence of a clot-- even asymptomatic-- in the proximal lower limb veins of a patient with clinically suspected pulmonary embolism (PE) provides evidence for venous thromboembolism and indicates anticoagulant therapy in such patients. We aimed at assessing the diagnostic performance of compression ultrasonography as compared to multi-slice computed tomography (MSCT) for the diagnosis of PE. We analyzed data from a large outcome management study that included consecutive outpatients referred to the emergency ward with clinically suspected PE. All high clinical probability patients, and all non-high clinical probability patients with a positive D-dimer test underwent both MSCT and CUS. Of the 756 included patients, 232 had PE ruled out on the basis of a negative D-dimer test, and 524 patients underwent both MSCT and CUS. PE was found in 187 out of the 511 patients with a conclusive MSCT. The sensitivity of CUS for the presence of PE on MSCT was 39% (95% confidence interval: 32 to 46%), and its specificity was 99% (95% CI:97 to 100%). Positive and negative likelihood ratios were 42.2 (95% CI: 13.5 to 131.9) and 0.6 (95% CI: 0.5 to 0.7), respectively. We conclude from that large study of unselected patients that CUS has high specificity but low sensitivity, for the diagnosis of PE at MSCT in suspected patients. It allows ruling in the diagnosis of PE without further invasive and/or expensive testing in suspected patients.