Clinical and radiological risk stratification of patients with acute pulmonary embolism

Background: Acute pulmonary embolism is, to this day, a disease with a significant risk of death. Many attempts have been made to better understand the risk factors for adverse outcome after the diagnosis of pulmonary embolism. While many individual clinical or radiological risk assessment systems have been suggested, there has been no attempt to combine radiological and clinical parameters in a composite risk-score to enhance prediction of adverse outcome. Methods: From a group of 96 patients that were diagnosed with acute pulmonary embolism through CT angiography, we collected clinical data which included information necessary to the Wells and Miniati Score for clinical pretest probability of pulmonary embolism, echocardiographic, radiographic and CT data, the latter being reassessed to determine the radiological obstruction index (Mastora score). Using retrograde multivariate logistic regression modelling four components were identified to form a new composite score. This new score was validated by application to a second group of 132 patients. Both groups of patients were compared to each other in terms of comorbidities, clinical presentation and outcome by using unpaired t-tests and chi-square-tests. Results: Of the 96 patients that constituted the first group, 8 patients died (primary endpoint), and 27 had other adverse outcome such as major hemorrhage or recurring embolism (secondary endpoint). Of the 132 patients of the second group, there were 8 deaths and 33 adverse outcomes, respectively. The four strongest significant predictors of outcome that were included in the final new composite score, were the Wells and the Mastora score, the short axis of the right ventricle (RVsa) on axial CT planes and patient collapse. Four different risk groups could be identified. The composite score showed the best overall performance on ROC analysis as well as the best specificity, the best positive predictive value and discriminated better among surviving and deceased patients than any of the individual components. The overall odds ratio of the composite score was significantly superior to that of the Mastora and Wells score (p<.05), and notably superior to that of the RVsa. Upon validation through the second group of patients, the overall performance was even higher than in the first group (Az = 0,88 vs. 0,85). Sensitivity, specificity and positive predictive value could be confirmed (0,52 vs. 0,75, 0,96 vs. 0,94 and 57,2 vs. 42,9%, respectively for primary endpoint). Conclusion: The composite score is a promising tool to correctly assign patients with pulmonary embolism to four different risk groups with significant difference in outcome without causing diagnostic delay. In comparison to other prediction models, for the first time not only low risk attribution, but also high risk attribution is safe with excellent positive predictive values for primary and secondary outcome.2015-01-2