Pulmonary embolism: Prognostic CT findings

Purpose: To retrospectively determine whether three computed tomographic (CT) findings-ventricular septal bowing (VSB), ratio between the diameters of right ventricle (RV) and left ventricle (LV), and embolic burden-are associated with short-term death, defined as in-hospital death or death within 30 days of CT, whichever was longer, due to acute pulmonary embolism (PE). Materials and Methods: Institutional Review Board approval was obtained, and patient information was reviewed in compliance with HIPAA regulations. A total of 1193 patients with CT scans positive for PE from January 1, 1997, to December 31, 2002, who had given authorization for retrospective research were included. Scans were independently reviewed by two observers. CT findings were compared with risk of death by using univariate analysis (χ² statistic) and multivariate logistic regression. Interobserver variability (κ statistic or intraclass correlation coefficient), sensitivity, and specificity of CT findings for predicting death were calculated. A third observer reviewed discrepant cases post hoc. Results: Fifty-four percent of patients were women and 46% were men (mean age, 63 years ± 16). For observer 1, VSB was associated with death in univariate (odds ratio [OR], 1.98; P = .04) and multivariate modeling (OR, 1.97; P = .05). Interobserver variability was only fair (κ = 0.54) for VSB, and observer 2 found no association with death (OR, 1.52; P = .22). For both observers, VSB had low sensitivity (21% and 18%) and high specificity (88% and 87%) for predicting death. Neither RV/LV diameter ratio nor embolic burden was associated with increased risk of death. For observer 3, VSB was associated with death in univariate (OR, 2.10; P = .05) and multivariate analyses (OR, 2.18; P = .05). Conclusion: CT-depicted VSB is predictive of death due to PE, but with low sensitivity and high interobserver variability. RV/LV diameter ratio and embolic burden are not associated with short-term death due to PE.