Ischemic stroke after cardiac surgery is a devastating complication affecting approximately 2% of patients
The relationship between hypotension occurring before, during, and after cardiopulmonary bypass and stroke remains unclear
Mean arterial pressure less than 65 mmHg for 10 or more min during cardiopulmonary bypass is associated with an increased risk of stroke
Even mild relative hypotension, defined as a less than 10% decrease from preinduction baseline during bypass, was also associated with an increased risk of stroke
Stroke is a leading cause of morbidity, mortality, and disability in patients undergoing cardiac surgery. Identifying modifiable perioperative stroke risk factors may lead to improved patient outcomes. The association between the severity and duration of intraoperative hypotension and postoperative stroke in patients undergoing cardiac surgery was evaluated.
A retrospective cohort study was conducted of adult patients who underwent cardiac surgery requiring cardiopulmonary bypass at a tertiary center between November 1, 2009, and March 31, 2015. The primary outcome was postoperative ischemic stroke. Intraoperative hypotension was defined as the number of minutes spent within mean arterial pressure bands of less than 55, 55 to 64, and 65 to 74 mmHg before, during, and after cardiopulmonary bypass. The association between stroke and hypotension was examined by using logistic regression with propensity score adjustment.
Among the 7,457 patients included in this analysis, 111 (1.5%) had a confirmed postoperative diagnosis of stroke. Stroke was strongly associated with sustained mean arterial pressure of less than 64 mmHg during cardiopulmonary bypass (adjusted odds ratio 1.13; 95% CI, 1.05 to 1.21 for every 10 min of mean arterial pressure between 55 and 64 mmHg; adjusted odds ratio 1.16; 95% CI, 1.08 to 1.23 for every 10 min of mean arterial pressure less than 55 mmHg). Other factors that were independently associated with stroke were older age, hypertension, combined coronary artery bypass graft/valve surgery, emergent operative status, prolonged cardiopulmonary bypass duration, and postoperative new-onset atrial fibrillation.
Hypotension is a potentially modifiable risk factor for perioperative stroke. The study’s findings suggest that mean arterial pressure may be an important intraoperative therapeutic hemodynamic target to reduce the incidence of stroke in patients undergoing cardiopulmonary bypass.
Stroke and cognitive impairment are the second leading cause of postoperative morbidity, mortality, and long-term disability.1 Strokes complicate 1.3 to 2.2% of coronary artery bypass graft (CABG) surgeries2–4 and up to 8.7% of other major cardiac procedures.5
The lack of a clear understanding of the pathophysiology of postoperative stroke led to the lack of evidence-based strategies to reduce the incidence of this complication. The traditional putative mechanism for perioperative cerebral ischemia is thromboembolism.1 The unique physiologic vasomotor changes associated with cardiopulmonary bypass (CPB) suggest that hypotensive or distributive events may compound or even surpass embolic events in explaining strokes after cardiac surgery; however, this hypothesis has not been well studied. Additionally, there are no mean arterial pressure (MAP) maintenance guidelines for patients undergoing CPB, and the associations between hypotension thresholds, duration, and stroke remain unclear. Previous analyses have been limited by blood pressure measurements during select phases of surgery such as the average MAP on CPB6 and lack of minute-to-minute blood pressure variations throughout the surgical procedure. Optimal blood pressure thresholds for stroke prevention have been described as MAP within 30% of baseline in a noncardiac, nonneurologic surgery cohort, and ranges from systolic pressure of greater than 90 mmHg at any point,7 MAP of greater than 66 mmHg,8 and between 80 and 100 mmHg during CPB in the context of cardiac surgery.9
This study aims to evaluate the association between the severity and duration of intraoperative hypotension and postoperative stroke. We hypothesize that hypotension during and after CPB is associated with the risk of postoperative stroke.
Materials and Methods
Design and Selection Criteria
The Research Ethics Board at the University of Ottawa Heart Institute (Ottawa, Ontario, Canada) approved this protocol and waived the need for individual patient informed consent. We conducted a retrospective cohort study of 7,457 consecutive adult patients who underwent cardiac surgery requiring CPB at the University of Ottawa Heart Institute between November 1, 2009, and March 31, 2015. The University of Ottawa Heart Institute is a high-volume, university-based tertiary care center that performs a full scope of cardiac procedures. Patients undergoing off-pump procedures were excluded because we aimed to study the association of hypotension and stroke during the distinct periods before, during, and after CPB. In addition, thoracic aortic surgeries, cardiac transplantation, and ventricular assist devices were excluded.
This study adhered to the Strengthening the Reporting of Observational Studies in Epidemiology guidelines. The primary outcome and the statistical analysis plan were made before accessing the data.
We performed a retrospective analysis of prospectively collected data from the University of Ottawa Heart Institute perioperative database, which is a multimodular database managed by a multidisciplinary committee and which undergoes regular, scheduled quality assurance audits. This database contains patient demographics, comorbidities, intraoperative management and hemodynamics, postoperative interventions, and in-hospital outcomes.10
All intraoperative invasive blood pressure measurements were recorded automatically every 15 s in an electronic patient record (CompuRecord, Philips Medical Systems, The Netherlands), with any artifacts removed by using an automated algorithm as previously described.11 Specifically, because there were up to four recorded MAP values per minute, the median of these values was selected for the analysis. Time periods corresponding to absent (no MAP readings) or aberrant MAP values (an isolated MAP value that differed more than 50% from both preceding and subsequent values) were deleted. This approach effectively removed artifacts in invasive MAP values related to manipulation of the arterial line including blood sampling, clamping, flushing, and zeroing.11 In the event when arterial tracings were dampened, femoral or direct aortic pressure monitoring was provided by the surgeon via a three-way stopcock, connected to the original arterial line transducer. The new arterial recordings automatically replaced the dampened MAP recordings in the electronic record. MAP readings were analyzed from the onset of end-tidal carbon dioxide (i.e., induction) until the last end-tidal carbon dioxide reading (i.e., the conclusion of anesthesia and transfer of patient from the operating room to the intensive care unit).11 We routinely place arterial lines before induction at our institution with the exception of critical emergencies such as aortic rupture, which were excluded from this analysis. Intraoperative hemodynamic data was processed by using R (version 3.2.1; https://cran.r-project.org/bin/windows/base/old/3.2.1/; accessed on June 1, 2015).
Outcome and Exposures
The primary outcome was in-hospital postoperative ischemic stroke during the index surgical admission, defined as new focal or global neurologic deficit of cerebrovascular origin lasting 24 h or longer that was not present before surgery. Potential strokes were identified in the perioperative database by a trained data abstractor. Such cases were then confirmed as strokes by clinician members of the study team (L.Y.S., A.M.C.) after reviewing the physician notes, consults, and postoperative brain computed tomography and magnetic resonance imaging studies.
Three definitions were used a priori to define intraoperative hypotension, namely, the number of minutes spent within MAP bands of less than 55, 55 to 64, and 65 to 74 mmHg before, during, and after CPB. During CPB, MAP is driven by the pump flow and systemic vascular resistance. Although MAP before CPB and MAP after CPB are both driven by systemic vascular resistance and intrinsic cardiac function, the post-CPB pulsatile flow and systemic vascular resistance are also influenced by factors such as CPB duration, degree of myocardial preservation, and air embolism. We therefore separated the periods before, during, and after CPB in our analyses due to their physiologic distinctness. The selected MAP thresholds were based on thresholds shown to be associated with harm in recent studies of hypotension during noncardiac surgery.11–13
Continuous variables were analyzed by using analysis of variance and presented as mean (SD). Categorical variables were analyzed by using chi-square test and presented as number (proportion).
Association of Hypotension and Stroke.
The possible association between hypotension and stroke was examined by using logistic regression with propensity score adjustment based on the theoretical framework proposed by Rosenbaum and Ruben14 to estimate the average treatment effect among treated. Propensity scores were derived for each of the predefined MAP bands (i.e., MAP levels of less than 55, 55 to 64, and 65 to 74) to represent the likelihood of an intraoperative hypotensive event lasting for 1 min or longer. The propensity of each threshold of hypotension was calculated by using a nonparsimonius multivariable logistic regression model based on biologically plausible risk factors of hypotension (age, sex, body surface area, left ventricular ejection fraction less than 35%, hypertension, heart failure, peripheral vascular disease, atrial fibrillation, pulmonary hypertension, active endocarditis, recent myocardial infarction, previous stroke, transient ischemic attack, carotid disease, diabetes, renal insufficiency, dialysis dependence, surgery type, operative priority, preoperative cardiogenic shock or cardiac arrest, CPB duration, lowest hematocrit on CPB, intraoperative blood transfusion, and intraoperative tranexamic acid dose of 5g or more). Because current knowledge of hypotension risk factors is limited, these covariates were derived in consultation with the main components of the Society of Thoracic Surgeons15 and EuroScore II16 risk scores. Each of the propensity-adjusted models consisted of four continuous variables: the propensity score corresponding to one of the predefined MAP thresholds and three covariates representing the total duration of hypotension below that MAP threshold before, during, and after CPB, respectively. The propensity score models had moderate discriminating abilities (c-statistics ranging from 0.66 to 0.75). The covariates and their definitions are provided in appendix 1.
Other Stroke Risk Factors.
We identified other stroke risk factors through a nonparsimonious multivariable logistic regression model with variables that were selected a priori based on perioperative stroke literature,1,5,17 the key components of the Society of Thoracic Surgeons score,15 and the EuroScore II.16 These risk factors were age, sex, history of transient ischemic attack, stroke, or carotid stenosis; hypertension, diabetes, reoperative procedure, CPB duration, postoperative atrial fibrillation, mixed CABG and valve/other procedure and thoracic aortic surgery, as compared to CABG-only or single-valve procedures.5 We tested for the presence of any interaction between MAP levels before, during, and after CPB of less than 55 mmHg and each of these covariates using multiplicative interaction terms.
Several sensitivity analyses were performed post hoc. First, we examined the association of longest hypotensive episode durations before, during, and after CPB with stroke as a means for assessing consecutive hypotensive minutes. Although we address the association between consecutive minutes of hypotension and stroke via a sensitivity analysis of longest hypotensive episodes, more rigorous analyses of the time-varying patterns of hypotensive episodes contributing to a total cumulative hypotensive duration were beyond the scope of this study. Second, as the pre- and post-CPB periods had in common the pulsatile flow that was driven by intrinsic cardiac function, we repeated the above-described propensity-adjusted analyses by combining the total durations of hypotension pre- and post-CPB into a single variable. Thus, two distinct variables, one representing the combined pre- and post-CPB hypotension duration and the other representing hypotension duration on CPB, were studied. Third, we combined the total durations of hypotension throughout the entire case into a single variable to examine the impact of the cumulative duration of hypotension on stroke. Fourth, we determined the association of stroke with relative hypotension thresholds. The relative thresholds examined were decreases in MAP of less than 10%, 10 to 20%, and 20 to 30% from the preinduction value. The preinduction MAP was defined as the mean of three consecutive MAP measurements immediately before induction of anesthesia and 3 min before the first appearance of continuous expired carbon dioxide registration.7
We used the odds ratio (95% CI) to describe the measure of association. We defined a minimum clinically meaningful effect as more than 1.05 per 10 min of hypotension, and more than 1.5 for other covariates. In this primarily exploratory study, we did not correct for multiple testing. Statistical analyses were conducted using SAS 9.4 (SAS Institute, USA), with statistical significance defined by a two-tailed P < 0.05.
Main outcome and exposure variables were complete for all included subjects. Left ventricular ejection fraction was imputed using the group mean for 112 (1.5%) of patients. Weight was imputed with the group mean for 15 patients. The proportion of absent and artifactual MAP values removed was less than 1% of the total. No other data were missing.
Among the 7,457 patients included in this analysis, 111 (1.4%) had a confirmed postoperative diagnosis of stroke. The absolute risk of stroke was 0.7% (n = 22) in isolated CABG, 1.2% (n = 23) in valve-only, and 2.8% (n = 66) in combined CABG/valve. Table 1 summarizes the demographic and perioperative characteristics of patients with and without strokes. Stroke patients were more frequently older and female; had preexisting hypertension, carotid or cerebrovascular disease, renal insufficiency, active endocarditis, or preoperative cardiogenic shock; underwent surgeries that were reoperative, emergent, or more complex (i.e., combined CABG/valve); had longer CPB durations, lowest CPB hematocrit less than 0.22, or new-onset atrial fibrillation postoperatively. Pre-CPB MAP less than 55 mmHg modified the effect of new-onset postoperative atrial fibrillation on stroke (interaction P = 0.05).
Table 2 demonstrates the average total and longest episode durations of hypotension exposure before, during, and after CPB. Compared to nonstroke patients, stroke patients were more likely to have had MAP less than 55 pre-CPB and MAP less than 64 during and after CPB. The absolute risk of stroke increased with decreasing MAP thresholds, such that those who were exposed to longer than the average duration of MAP less than 55 had the highest risk of stroke. Table 3 illustrates the association between stroke and various thresholds of hypotension before, during, and after CPB. Stroke was associated with longer durations of hypotension during CPB. Specifically, in the analysis using total hypotension duration, every additional 10 min of CPB MAP less than 55 was associated with a 16% increased odds of stroke (adjusted odds ratio 1.16; 95% CI, 1.08 to 1.23), and every additional 10 min of MAP between 55 and 64 was associated with a 13% increased odds of stroke (adjusted odds ratio 1.13; 95% CI, 1.05 to 1.21).
In addition to intraoperative hypotension, other independent predictors of stroke risk included older age, hypertension, combined CABG/valve surgery, emergent operative status, prolonged CPB duration, and postoperative new-onset atrial fibrillation (table 4). This multivariable model had excellent discriminative ability (c-statistic 0.81). Post hoc, we determined that we achieved 92% power at a 0.05 significance level to detect an odds ratio of 1.5 for risk factors with a prevalence of 20% or greater and 73% power to detect an odds ratio of 1.5 for risk factors with a prevalence of 10% or greater. On testing multiplicative interaction terms, we found that the effect of pre-CPB MAP less than 55 on stroke was amplified by complex surgery (combined CABG and valve) and new-onset postoperative atrial fibrillation.
The association between hypotension and stroke remained robust in sensitivity analyses where the longest hypotension episodes were considered in place of total duration (table 3). In addition to similar increases in the odds of stroke in the lower MAP bands, every additional 10 min of CPB MAP between 65 and 74 was associated with an 8% increased odds of stroke (adjusted odds ratio 1.08; 95% CI, 1.01 to 1.16).
Table 5 illustrates the association between stroke and total hypotension duration throughout the entire surgical case (i.e., before, during, and after CPB) and combined pre- and post-CPB hypotension durations. In both of these sensitivity analyses, more than 10 min of MAP less than 64 was associated with stroke. In the sensitivity analysis that examined relative MAP thresholds (table 6), small decreases (less than 10%) in MAP pre-CPB appeared protective, whereas any MAP thresholds below the preinduction value during CPB were associated with stroke.
Our findings suggest that intraoperative blood pressure may be a modifiable risk factor for stroke after cardiac surgery. We identified critical MAP thresholds and durations for postoperative ischemic stroke to be MAP less than 64 mmHg for more than 10 min during CPB. This finding is important because stroke is associated with mortality, quality of life, and healthcare cost.
The present study is novel in its exploration of the association between hypotension and stroke throughout all phases of cardiac surgery on a minute-to-minute basis. One novel aspect of this study is the continuous high-fidelity intraoperative recording of invasive blood pressure measurements for each physiologically distinct stage of surgery. Published studies evaluating the association between hypotension in cardiac surgery and stroke were not performed in contemporary surgical populations, were descriptive in nature, and/or did not explore minute-to-minute blood pressure thresholds throughout the intraoperative period. Our findings are nonetheless consistent with previous reports. In a randomized control study of 248 patients undergoing CABG and valve surgery in the early 1990s, the high-MAP group (CPB MAP target of 80 to 100 mmHg) had a lower incidence of strokes (2.4% vs. 7.2%) and mortality (1.6% vs. 4.0%) at 6 months, compared to the control group whose CPB MAP target was 50 to 60 mmHg.9 Two historical cohort studies of CABG patients demonstrated that intraoperative hypotension defined as a systolic blood pressure less than 90 mmHg for more than 30 min or intraoperative systolic blood pressure less than 40 mmHg for more than 5 min were each associated with an up to fourfold higher risk of stroke.18,19 In addition, in patients with established cardiac surgery-related strokes, those with at least 10-mmHg decrease in MAP from preoperative value had four times the odds of bilateral watershed infarcts.6
During cardiac surgery, hypotension occurs as a result of decreased venous return from cardiac manipulation, arrhythmias, reduced ventricular function, and/or decreased systemic vascular resistance.20 Despite the importance of maintaining systemic perfusion pressure,4,13 no guidelines provide perfusion pressure targets or define what duration of hypotension can be harmful. The range of cerebral autoregulation is thought to be between MAP of 50 and 150 mmHg.21,22 In practice, common CPB MAP targets vary between 50 and 70 mmHg.23 The determination of an optimal CPB perfusion pressure is often based on the perceived lower limit of autoregulation. Early studies suggested 50 to 55 mmHg to be acceptable in healthy young volunteers and animal models.21 Subsequently, a MAP of no less than 70 mmHg has been suggested in normotensive, nonanesthetized adults,24–26 and a MAP of no less than 66 mmHg has been suggested in anesthetized elderly patients during CPB.8 Despite decreased cerebral metabolic demand under general anesthesia, the commonly accepted CPB MAP target of 50 mmHg is likely inappropriate in elderly patients with a rightward cerebral autoregulatory shift.27,28
Neurologic complications are a cause of major morbidity and mortality after cardiac surgery. Although the pathogenesis of stroke is multifaceted, current evidence points to patient-related risk factors as a more important contributor than surgical technique, such as the use of cardiopulmonary bypass.29 We identified independent stroke risk factors after adjustment for intraoperative hypotension. Our finding of older age,18,29–36 preexisting hypertension,1,37,38 combined valvular and CABG surgery,5,39–41 and prolonged CPB duration41 are consistent with multiple previous reports. Most importantly, in addition to defining critical thresholds of hypotension as a modifiable stroke risk factor, we also confirmed CPB duration and postoperative atrial fibrillation37,38 as strong and potentially modifiable risk factors. Of note, MAP less than 55 mmHg pre-CPB amplified the effect of new-onset postoperative atrial fibrillation on the development of stroke. Careful perioperative planning may mitigate the effects of nonmodifiable stroke risk factors such as age and complex surgery, but careful perfusion pressure management during CPB and employing preventative measures against postoperative atrial fibrillation are potential therapeutic strategies to reduce postoperative stroke that merit evaluation in a prospective multicenter trial.
Limitations and Strengths
Our study has several limitations. First, there are limitations that are inherent of its retrospective and observational design. Our study demonstrates a strong association between stroke and hypotension. However, further studies are needed to demonstrate whether stroke results directly from hypotension or also indirectly through low cardiac output, hypovolemia, and vasopressor use. Second, we are limited in our ability to collect certain data elements that may be related to the occurrence of hypotension, such as precise end-tidal concentrations of volatile anesthetics and maintenance medications. Nonetheless, this study is novel in its attempt to define hypotension on a minute-to-minute basis during physiologically distinct periods of cardiac surgery (i.e., before, during, and after CPB) and therefore provides important insight for blood pressure management throughout the surgical period. Third, our findings are to be interpreted in the context of multiple testing. Furthermore, although we address the association between consecutive minutes of hypotension and stroke via a sensitivity analysis of longest hypotensive episodes, more rigorous analyses of the time-varying patterns of hypotensive episodes contributing to a total cumulative hypotensive duration were beyond the scope of this study. However, our refined MAP artifact removal algorithm allowed for reliable definition of hypotensive episodes. Fourth, the present study is single center in nature. Further studies are needed to confirm the generalizability of our findings.
In conclusion, we observed an increased risk of perioperative stroke when intraoperative MAP fell below 64 mmHg for more than 10 min during CPB. Our findings suggest that intraoperative MAP may be a modifiable hemodynamic therapeutic target for stroke prevention in patients undergoing cardiac surgery and provide a rationale for prospective studies of hemodynamic goal-directed therapies in these patients.
Supported in part by the Research Funds of the Division of Cardiac Anesthesiology of the University of Ottawa Heart Institute, Ottawa, Ontario, Canada. Dr. Chung was supported by the Heart and Stroke Foundation (Ontario, Canada) Hannah Pherril Scholarship.
Dr. Farkouh receives research support from Amgen (Mississauga, Ontario, Canada). The other authors declare no competing interests.
Appendix 1. Covariates and Their Definitions
All covariate information is prospectively collected in our institutional perioperative database. These data are provided by attending anesthesiologists on the day of surgery, undergo weekly audits, and are of research quality.10