A diagnostic strategy for pulmonary embolism based on clinical and perfusion lung scan findings

Purpose: To elaborate a diagnostic strategy for pulmonary embolism (PE) by combining clinical evaluation with perfusion lung scan interpretation. Methods: We prospectively evaluated 890 consecutive patients with suspected PE. Prior to lung scanning, each patient was assigned a clinical probability of PE based on risk factors, symptoms and signs, ECG, chest radiography, and arterial blood gas data. Perfusion scans were independently classified as foflows: 1) normal, 2) near-normal, 3) abnormal compatible with PE (PE+, single or multiple wedge-shaped perfusion defects), 4) abnormal not compatible with PE (PE-, perfusion defects other than wedge-shaped). By protocol, pulmonary angiography was required in patients with abnormal scans. Results: Most patients with angiographically proven PE had PE+ scans (sensitivity 92%). Conversely, most patients without emboli on angiography had PE- scans (specificity 87%). Observed sensitivity and specificity were corrected for selection bias in referring patients to angiography by means of Bayes' theorem. Debiased sensitivity and specificity of PE+ scan were 86 and 93%, respectively. These values were used to compute posterior probabilities (probabilistic estimates) of PE from clinical probabilities (prior probabilities) according to Bayesan theorem. When clinical probabilities of 10, 50, and 90% (indicating low, intermediate and high likelihood of PE, respectively) were associated with PE+ scan, posterior probabilities of PE were 58, 93 and 99%, respectively. When the same clinical probabilities were paired with PE- scan, posterior probabilities of PE were 2, 13 and 57%, respectively. Probabilistic estimates of PE were in close agreement with actual prevalence of PE observed in patients with the above combinations of clinical and lung scan findings. Conclusions: A PE+ scan paired with intermediate or high clinical probability yields posterior probabilities of PE sufficiently high to warrant anticoagulant therapy without the need for angiography. A low clinical probability associated with PE- scan yields a posterior probability of PE sufficiently low to withhold anticoagulation without further testing. Angiography is required when clinical and perfusion scan findings are discordant. In our study, angiography would have been required only in 24% of patients with abnormal scans. Clinical implications: combining clinical assessment with perfusion scan interpretation helps to restrict the need for angiography to a minority of patients with suspected PE thereby reducing the cost of diagnostic procedures for PE.