Patients presenting with TAAAD aged 70 years or above (n = 1449) were stratified according to presence or absence of CVA before surgery (CVA: n = 110, 7.6%). Additionally, we analyzed patients presenting with CVA in whom a strategy of medical management was followed (n = 35).
In the IRAD, preoperative CVA is defined as a loss of neurological function caused by a disturbance in cerebral blood supply with residual symptoms 24 hours after onset. Among baseline characteristics, aortic valve disease is defined as presence of either severe aortic valve stenosis or regurgitation. Among variables characterizing patients’ clinical presentation, coma is defined as complete or partial mental unresponsiveness beyond that expected from anesthesia or no evidence of psychological or physiologically appropriate responses to stimulation. Presenting hypertension indicates blood pressure values > 150 of 90 mm Hg on admission. Shock indicates a maximum systolic blood pressure < 80 mm Hg for at least 30 minutes, pump failure, or signs of hypoperfusion. Signs of congestive heart failure include paroxysmal nocturnal failure, dyspnea on exertion or at rest, and pulmonary congestion on x-ray. Ischemic lower extremity indicates pain, pulselessness, pallor of the foot on elevation, rubor on dependency, necrosis, paralysis, paresthesia, intermittent claudication, or rest pain. Among preoperative imaging results, arch vessel involvement is present if the aortic pathology includes the level of the left subclavian artery or any more proximally originating arch vessels. Among in-hospital outcomes, post-operative CVA is defined as loss of neurological function (loss or slurring of speech, altered state of consciousness) caused by an ischemic event that is confirmed by either computed tomography or magnetic resonance imaging. Spinal cord ischemia is defined as evidence of occlusion of the radicular arteries of the spinal cord with loss of function to the lower extremities with or without bowel/bladder involvement. Post-operative hypotension indicates a systolic blood pressure that has decreased below 90 mmHg from an earlier higher recording.
Continuous variables with non-normal distributions are reported using medians and interquartile ranges. Categorical variables are presented using absolute frequencies and percentages. The Kruskal-Wallis test was used for group-wise comparison of non-normally distributed continuous variables. Categorical variables were compared using the chi-square test or Fisher's exact test where appropriate.
Because the CVA and non-CVA groups presented with marked differences in underlying comorbidities that could have had a confounding influence on postoperative outcomes, a binary logistic regression was generated to identify the independent association of particular comorbidities associated with CVA at presentation. To avoid the confounding influence of missing data, multiple imputation was utilized to first generate missing data for the variables in the model. After considering clinical relevance, variables exhibiting significant differences between the 2 groups during univariate analysis were used to generate the conditional probability of CVA calculated for each case, adjusting for the covariates in the model. This conditional probability generated was utilized to create both a balancing score as well as a propensity score for use in analysis of outcomes. The variables used to generate the balancing score were 4 baseline characteristics (chronic obstructive pulmonary disease, peripheral arterial disease, atherosclerosis, aortic valve disease) and 9 clinical characteristics (chest pain, head/neck pain, radiating pain, coma, syncope, hypertension, shock, signs of congestive heart failure, and ischemic lower extremity).
Primary outcome variables were defined as in-hospital mortality and the rates of postoperative complications (neurological deficit [CVA, coma, spinal cord ischemia], mesenteric ischemia, acute renal failure, extension of dissection, hypotension, cardiac tamponade, limb ischemia, and discharge to tertiary hospital). To investigate the independent influence of CVA on in-hospital mortality, a multivariable logistic regression model was created based on a matched dataset. Propensity matching using a nearest neighbor technique without replacement was used to compare groups for in-hospital mortality. A caliper of 0.2 was specified for the maximum difference between propensity scores for the matched pairs. The following variables were considered for introduction into the multivariable model: CVA (forced), gender, atherosclerosis, prior cardiac surgery, head/neck pain, coma, syncope, preoperative hypotension, cardiac tamponade, mesenteric ischemia, limb ischemia, proximal extension of dissection to aortic root, proximal extension of dissection to ascending aorta, post-operative hypotension, and acute renal failure.
The secondary outcome variable was mortality up to 5 years after discharge. A multivariable Cox proportional hazards regression model was generated to analyze independent associations between CVA and the secondary outcome. Because groups had fewer numbers at follow-up, instead of propensity matching the balancing score for CVA was used to maximize the number of cases included in this analysis. The balancing score was forced into the model to balance for group differences in comorbidities at presentation. After evaluating clinical relevance, variables with p-values of less than 0.15 were considered for introduction into the multivariable model. These variables included age, diabetes mellitus, atherosclerosis, aortic valve disease, prior cardiac surgery, family history of aortic disease, preoperative spinal cord ischemia, distal extension of dissection to descending aorta, time between initial admission and surgery, post-operative acute renal failure, chronic beta blocker therapy at discharge, and angiotensin converting enzyme inhibitor therapy at discharge. A backward stew-wise method was utilized as a tool leading to the creation of the final model.
Simple survival probabilities were also compared by the Kaplan-Meier method using the log-rank test to evaluate statistically significant differences between survival curves. Date of admission was defined as starting point for Kaplan-Meier survival curves.
Additionally, in-hospital outcomes of patients presenting with CVA who received medical management were described (in-hospital mortality, and the rates of in-hospital complications). Moreover, reasons for medical management were reported.
Two-sided P-values < 0.05 were defined as statistically significant. All statistical analyses were performed using Statistical Package for the Social Sciences (SPSS) for Windows, version 25.0 (IBM Corporation, Armonk, NY).