Individualized prediction of risk of ascending aortic syndromes.

Abstract

Although ascending aortic diameter changes acutely after dissection, recommendation for prophylactic surgery of thoracic aortic aneurysms rely on data from dissected aortas. In this case-control study we aim to identify risk markers for acute and chronic aortic syndromes of the ascending aorta (ACAS-AA). Furthermore, to develop a predictive model for ACAS-AA. We collected data of 188 cases of ACAS-AA and 376 controls standardized to age- and sex of the background population. Medical history and CT-derived aortic morphology were collected. For the dependent outcome ACAS-AA, potential independent risk factors were identified by univariate logistic regression and confirmed in multivariate logistic regression. As post-dissection tubular ascending aortic diameter is prone to expand, this factor was not included in the first model. The individual calculated adjusted odds ratios were then used in ROC-curve analysis to evaluate the diagnostic accuracy of the model. To test the influence of post-ACAS-AA tubular ascending aortic diameter, this was added to the model. The following risk factors were identified as independent risk factors for ACAS-AA in multivariate analysis: bicuspid aortic valve (OR 20.41, p = 0.03), renal insufficiency (OR 2.9, p<0.01), infrarenal abdominal aortic diameter (OR 1.08, p<0.01), left common carotid artery diameter (OR 1.40, p<0.01) and aortic width (OR 1.07, p<0.01). Area under the curve was 0.88 (p<0.01). Adding post-ACAS-AA tubular ascending aortic diameter to the model, negated the association of bicuspid aortic valve, renal insufficiency, and left common carotid artery diameter. Area under the curve changed to 0.98 (p<0.01). A high performing predictive model for ACAS-AA, free of ascending aortic diameter, can be achieved. Furthermore, we have identified abdominal aortic ectasia as an independent risk factor of ACAS-AA. Integration of potential biomarkers and morphologic variables, derived from undissected aortas, would probably improve the model.