Stroke after transcatheter aortic valve implantation: new insights into risk prediction

Abstract Background Stroke is still a feared complication following transcatheter aortic valve implantation (TAVI), associated with increased mortality and severe impact on patients' daily living. Despite technological development and knowledge improvement, cerebrovascular events (CVE) are still not predictable so far and simple risk scores are lacking. Purpose This study aimed to evaluate different preprocedural factors that may favour stroke after TAVI, especially regarding severity of aortic calcification. Methods From May 2011 to January 2018 a total of 1365 patients underwent TAVI with a balloon-expandable (n=442, 32.4%) or self-expandable (n=923, 67.6%) device at our institution. All patients underwent multi-slice computed tomography (MSCT) before TAVI. To obtain the significant factors for stroke risk prediction we performed multivariate regression analysis and selected the optimal cut-off values according to the maximum Youden index for dichotomization of the risk score items. Finally, we used receiver operating characteristics (ROC) analysis and areas under the ROC curves (AUC) to validate our risk score in comparison to other existing models. Results 60 of 1365 patients (4.4%) had new neurological impairment after TAVI during hospital stay (mean 11.2±6.7 days). We performed propensity score matching (1:10) to balance baseline characteristics between patients with and without stroke following TAVI, resulting in 56 patients with in-hospital stroke and 521 patients without. Preprocedural factors associated with stroke turned out to be history of prior stroke (OR 1.94; 95% CI 0.85–4.43; p=0.114), aortic valve area ≥0.545 cm2 (OR 3.11; 95% CI 1.16–8.34; p=0.024), atrioventricular angle ≥48.5° (OR 2.32; 95% CI 1.20–4.49; p=0.013), RCC Agatston Score ≥447.2 AU (OR 1.8; 95% CI 0.94–3.44; p=0.077), LVOT Agatston Score ≥262.4 AU (OR 2.01; 95% CI 1.08–3.75; p=0.028) and ascending thoracic aorta Agatston Score ≥116.4 AU (OR 2.21; 95% CI 1.17–4.17; p=0.015). ROC analysis showed that our risk model had an AUC of 0.73 (95% CI 0.66–0.80; p<0.001), a sensitivity of 70.6%, a specificity of 69.0%, a positive predictive value of 19.5% and a high negative predictive value of 95.7%. Our risk model appeared to be the best for stroke prediction after TAVI compared to other risk scores used in literature before like EuroSCORE II (AUC 0.50; 95% CI 0.43–0.58; p=0.950) or CHA2DS2-VASc Score (AUC 0.62; 95% CI 0.55–0.70; p=0.004). Conclusion Especially aortic root calcium volume assessed by MSCT predicts CVE after TAVI and could be integrated into a six items risk model for preprocedural prediction of stroke after TAVI. This model could guide us in identifying those patients who are most likely to benefit from transcatheter cerebral embolic protection devices. Funding Acknowledgement Type of funding source: None