Risk of death calculation after percutaneous or open surgical aortic valve replacement in high-risk patient group: a multicenter retrospective study

Based on the long-term results of treatment of patients with acquired aortic valve disease, identify predictors of an unfavorable outcome depending on the method of prosthesis implantation. Create regression models for prediction of the outcomes of surgical treatment. Materials and Methods: the main group included 48 patients with aortic stenosis who underwent percutaneous aortic valve implantation (TAVI). The control group included 72 patients who were operated on using frameless bioprostheses (AVR) in the period from 2002 to 2008. All patients underwent echocardiographic assessment of the parameters of the left ventricle and aortic valve. In order to assess the influence of individual factors, the method of multiple regression was used. Results: the overall mortality rate in TAVI group was 16.6% and 11.1% in AVR group, with no statistically significant differences revealed. There was a direct correlation between the presence of COPD and poor outcomes of moderate strength (R = 0.55). A regression model was made based on a combination of the EuroScore risk, and the left ventricular ejection fraction. This model predicted an adverse surgical outcome with an accuracy of 92%, sensitivity of 83%, specificity of 78% at x2 = 12 and p = 0.01. An ROC analysis was performed for each group. For the first group, the AuC was 0.789 (95% CI: 0.72-0.95) at p=0.03, for the second group, the AUC was 0.918 (95% CI: 0.834-0.99) at p = 0.001. For all patients included in the study, the AUC was 0.817 (95% CI: 0.755-0.91) at p = 0.02. The critical value of the probability of death was calculated for each group: in the TAVI group - 75% (sensitivity 85%, specificity 77%), in AVR group 37% (sensitivity 83%, specificity 85%). Conclusions: the regression model obtained in this study allows us to calculate the risk of unfavorable outcome after intervention on the aortic valve. The identified critical values for this model can serve as an indication in determining the method of treating pathology of the aortic valve.