Abstract 14815: Clinical Prediction Model of Growth Rate in Type B Aortic Dissection Using 4D Flow MRI

Introduction: Chronic type B aortic dissection is often associated with aneurysmal degeneration and rupture of the descending aorta (DAo). Early identification of patients at risk for aneurysmal degeneration could identify candidates for prophylactic surgical intervention. Clinical and static imaging measures (age, early aortic dilation, etc.) have been unable to reliably predict DAo growth rate (GR). Importantly, these measures do not account for dissection hemodynamics, an important driver of GR. We therefore aimed to build a multivariate linear regression model that combines baseline clinical, morphologic, and hemodynamic metrics to predict GR. Methods: Patients who underwent 4D-flow MRI and had ≥9 months of imaging follow-up were identified. Baseline clinical features were obtained from medical records. Morphologic metrics were assessed on baseline imaging and GR (mm/y) was calculated by comparison to follow-up scans. Baseline 4D-flow MRI was used to measure the following parameters in the true lumen: pulse wave velocity (PWV), peak velocity (PV), and forward flow. False lumen thrombus (FLT) was assessed in percentage quartiles (0, 1-25%, 26-50%, 51-75%, >75%) and scored from 0 to 4. Stepwise multivariate regression was conducted to identify a model for best prediction of GR (Figure 1). Results: Thirty-four patients were included (70.0 ± 16.5y; 21 M). On multivariate analysis, we detected a model that significantly correlates with GR (r 2 = 0.55 p < 0.001) by using PWV, FLT, and PV as predictors. In fact, PWV was the strongest predictor overall (r = - 0.54 p = 0.001). Clinical features and other imaging measures showed no association with GR and hence were excluded by stepwise regression analysis. Conclusions: The regression model identified two hemodynamic parameters and FLT as the strongest predictors of GR. This finding suggests that a multitude of hemodynamic factors impact GR and that 4D flow parameters have the potential to improve clinical risk models.