Interobserver Agreement for the Diagnosis of Venous Thromboembolism on Computed Tomography Chest Angiography and Indirect Venography of the Lower Extremities in Emergency Department Patients

Abstract

To determine interobserver agreement between radiologists for computed tomography (CT) angiography and venography. CT venography of the lower extremities combined with standard CT angiography of the chest may result in an increased overall diagnosis rate of venous thromboembolism (pulmonary embolism or deep venous thrombosis). The study had a retrospective cohort design. The population consisted of emergency department patients who were evaluated for suspected pulmonary embolism. A random sample of 50 patients diagnosed and treated for venous thromboembolism and 50 age- and gender-matched patients whose CT angiograms and venograms were read as negative were enrolled. The original reading (R1) was compared with readings of two study radiologists: R2, a general radiologist, and R3, a radiologist with fellowship training in cross-sectional imaging. All readers were blinded to each other. Both R2 and R3 found both CT angiogram and venogram components technically adequate in 95% (95% CI = 89% to 98%) and 86% (95% CI = 78% to 92%) of studies, respectively. The agreement was very good for CT angiography (lowest agreement = 92%; lowest kappa = 0.83) and was good for CT venography (85%, kappa = 0.65). In nine cases, R1 read the CT angiogram as negative but the venogram as positive for DVT, whereas both R2 and R3 read both components as negative in four of these nine, suggesting a false-positive isolated DVT rate of 44% (95% CI = 19% to 73%). In no case did R1 read both scan components as negative when R2 and R3 agreed on presence of pulmonary embolism or DVT. Diagnosis of pulmonary embolism on CT angiography is more reliable than diagnosis of isolated DVT on CT venography.