Performance of magnetic resonance angiography in suspected acute pulmonary embolism

Pulmonary embolism (PE) is a common and potentially fatal disorder. Non-specific findings make the clinical diagnosis of PE difficult. To assess the diagnostic value and inter-observer agreement of magnetic resonance angiography (MRA) in a cohort of patients with suspected PE, we conducted a prospective clinical study. MRA was compared for sensitivity and specificity to a diagnostic strategy including clinical probability, D-dimer testing, spiral CT, ultrasound leg compression and pulmonary angiography. A total of 89 patients with clinically suspected PE were included: the clinical probability of PE was intermediate or high in 78, and low in the remaining 11. All patients underwent monoor multi-slice spiral CT and MRA with gadolinium injection (both within 24 hours of entry to the study). Anticoagulation was withheld in patients concerned about the strategy. All subjects were followed up for 3 months. MRA was read independently by two experienced teams of radiologists: one local and one from another university centre. Spiral CT was positive in 62 of 63 cases of confirmed PE. No patient with negative CT findings was positive ultrasonographically. Only one patient with a negative CT (and negative ultrasound) had a recurrent thromboembolic event. The first team diagnosed PE with MRA in 47 cases, with a sensitivity of 71% and a specificity of 92%; the second team obtained the diagnosis in 23 cases, with a sensitivity of 31% and a specificity of 85%. Inter-observer agreement between MRA reading was low: Kappa = 0.16 (-0.01 to 0.33); p = 0.07. In conclusion, compared with a non-invasive strategy based on spiral CT, the diagnostic value of MRA is limited by poor inter-observer agreement.