Purpose: To predict cardiovascular disease (CVD) in a clinical care population by using prevalent subclinical ancillary aortic findings detected on chest computed tomographic (CT) images.
Materials and methods: The study was approved by the medical ethics committee of the primary participating facility and the institutional review boards of all other participating centers. From a total of 6975 patients who underwent diagnostic contrast material-enhanced chest CT for noncardiovascular indications, a representative sample population of 817 patients plus 347 patients who experienced a cardiovascular event during a mean follow-up period of 17 months were assigned visual scores for ancillary aortic abnormalities--on a scale of 0-8 for calcifications, a scale of 0-4 for plaques, a scale of 0-4 for irregularities, and a scale of 0-1 for elongation. Four Cox proportional hazard models incorporating different sum scores for the aortic abnormalities plus age, sex, and chest CT indication were compared for discrimination and calibration. The prediction model that performed best was chosen and externally validated.
Results: Each aortic abnormality was highly predictive, and all models performed well (c index range, 0.70-0.72; goodness-of-fit P value range, .45-.76). The prediction model incorporating the sum score for aortic calcifications was chosen owing to its good performance (c index, 0.72; goodness-of-fit P = .47) and its applicability to nonenhanced CT scanning. Validation of this model in an external data set also revealed good performance (c index, 0.71; goodness-of-fit P = .25; sensitivity, 46%; specificity, 76%).
Conclusion: A derived prediction model incorporating ancillary aortic findings detected on routine diagnostic CT images complements established risk scores and may help to identify patients at high risk for CVD. Timely application of preventative measures may ultimately reduce the number or severity of CVD events.
© RSNA, 2010.