Thoracic ultrasound for diagnosing pulmonary embolism: a prospective multicenter study of 352 patients.

BACKGROUND: Pulmonary embolism (PE) continues to be a major challenge in terms of diagnosis, as evidenced by the fact that many patients die undiagnosed and/or untreated. The aim of this multicenter study was to determine the accuracy of thorax ultrasound (TUS) in the diagnosis of PE (TUSPE).

METHODS: From January 2002 through September 2003, 352 patients with suspected PE were examined in seven clinics. The patients were investigated prospectively by TUS according to the following criteria: (1) PE confirmed: two or more typical triangular or rounded pleural-based lesions; (2) PE probable: one typical lesion with pleural effusion; (3) PE possible: small (< 5 mm) subpleural lesions or a single pleural effusion alone; or (4) normal TUS findings. In all cases, CT pulmonary angiography (CTPA) was used as the reference method. In the event of discrepant findings, a combination of duplex sonography of the leg veins, echocardiography, ventilation/perfusion scintigraphy, and a quantitative enzyme-linked immunosorbent assay or latex d-dimer, or a biopsy/autopsy was performed.

FINDINGS: PE was diagnosed in 194 patients. On TUS, 144 patients had a total of 333 subpleural lesions (mean, 2.3 lesions per patient) averaging 15.5 x 12.4 mm in size. Additionally, a narrow pleural effusion was found in 49% of the patients. TUS yielded the following results under application of the strict criteria 1 and 2: PE true-positive, n = 144; PE false-positive, n = 8; PE true-negative, n = 150; and PE false-negative, n = 50. The sensitivity was 74%, specificity was 95%, positive predictive value was 95%, negative predictive was value 75%, and accuracy was 84%, at a prevalence of 55%. The sensitivity in patients with criterion 1 was 43% and a specificity of 99%.

INTERPRETATION: TUS is a noninvasive method to diagnose peripheral PE. In the absence of CTPA, TUS is a suitable tool to demonstrate a PE at the bedside and in the emergency setting.