Development of a multivariable prediction model for patient-adjusted aortic risk morphology

Abstract

Abstract OBJECTIVES Preventing type A aortic dissection requires reliable prediction. We developed and validated a multivariable prediction model based on anthropometry to define patient-adjusted thresholds for aortic diameter and length. METHODS We analysed computed tomography angiographies and clinical data from 510 control patients, 143 subjects for model validation, 125 individuals with ascending aorta ectasia (45–54 mm), 58 patients with aneurysm (≥55 mm), 206 patients with type A aortic dissection and 19 patients who had received a computed tomography angiography ≤2 years before they suffered from a type A aortic dissection. Computed tomography angiographies were analysed using curved planar reformations. RESULTS In the control group, the mean ascending aortic diameter was 33.8 mm [standard deviation (SD) ±5.2 mm], and the length, measured from the aortic valve to the brachiocephalic trunk, was 91.9 mm (SD ±12.2 mm); both diameter and length were correlated with anthropometric parameters and were smaller than the respective values in all pathological groups (P < 0.001). Multivariable linear regression analysis of the control group revealed that age, sex and body surface area were predictors of ascending aorta diameter (R2 = 0.40) and length (R2 = 0.26). Bicuspidity of the aortic valve was not included in the model; its prevalence was only 3.2% in the control group but >25% in the ectasia and aneurysm groups. CONCLUSIONS The regression model provides a patient-adjusted prediction of the thresholds for aortic diameter and length. In our retrospective data, the model resulted in better identification of aortas at the risk of dissection than the conventional 55-mm diameter threshold. The model is available as an Internet calculator (www.aorticcalculator.com).