Helical Computed Tomography and the Diagnosis of Pulmonary Embolism

• Pulmonary embolism
• Tomography, computed, spiral
• Angiography
• Sensitivity and specificity
• Anticoagulants

Acute pulmonary embolism is a common and potentially life-threatening disorder (1). Treatment is highly effective (2) but not without complications, and the diagnosis of pulmonary embolism therefore requires a high degree of certainty. Conversely, if untreated, pulmonary embolism can be fatal.

The clinical diagnosis alone is inaccurate for patients with suspected pulmonary embolism. Pulmonary angiography remains the diagnostic reference standard, but it is invasive and expensive, is not universally available, and is associated with morbidity (in 2% to 5% of patients) and death (in <1% of patients) (3). Ventilation-perfusion scanning, the most commonly used screening technique, provides a probability estimate for pulmonary embolism on the basis of the size and number of perfusion defects. Normal results on ventilation-perfusion lung scanning essentially exclude pulmonary embolism, and a high-probability lung scan is virtually diagnostic when accompanied by high or moderate clinical suspicion (4). However, because approximately two thirds of lung scans are nondiagnostic (4), pulmonary angiography or serial venous ultrasonography (done over 14 days to detect deep venous thrombosis) (5) is often required. Both of these approaches are inconvenient and costly, and angiography can be dangerous.

Recently, tremendous interest and enthusiasm have surrounded the use of helical (spiral) computed tomography (CT) as the initial screening test for patients with suspected acute pulmonary embolism. In fact, some centers use the results of helical CT to manage such patients (6). As shown by Rathbun and colleagues (7) in this issue, however, this approach is risky.

Several criteria should be met before physicians use a new diagnostic test to make clinical decisions. The studies done to evaluate the test should be methodologically rigorous. In addition, they should involve three stages: The interobserver and intraobserver variability and the criteria for positive and negative results should be established; the sensitivity and specificity …