Secondary Logo

Journal Logo

The Association Between Mild Intraoperative Hypotension and Stroke in General Surgery Patients

Hsieh, Jason K. BS; Dalton, Jarrod E. PhD; Yang, Dongsheng MS; Farag, Ehab S. MD; Sessler, Daniel I. MD; Kurz, Andrea M. MD

doi: 10.1213/ANE.0000000000001526
Trauma: Original Clinical Research Report

BACKGROUND: Intraoperative hypotension may contribute to perioperative strokes. We therefore tested the hypothesis that intraoperative hypotension is associated with perioperative stroke.

METHODS: After institutional review board approval for this case-control study, we identified patients who had nonneurological, noncardiac, and noncarotid surgery under general anesthesia at the Cleveland Clinic between 2005 and 2011 and experienced a postoperative stroke. Control patients not experiencing postoperative stroke were matched in a 4-to-1 ratio using propensity scores and restriction to the same procedure type as stroke patients. The association between intraoperative hypotension, measured as time-integrated area under a mean arterial pressure (MAP) of 70 mm Hg, and postoperative stroke was assessed using zero-inflated negative binomial regression.

RESULTS: Among 106 337 patients meeting inclusion criteria, we identified 120 who had confirmed postoperative stroke events based on manual chart review. Four-to-one propensity matching yielded a final matched sample of 104 stroke cases and 398 controls. There was no association between stroke and intraoperative hypotension. Stroke patients were not more likely than controls to have been hypotensive (odds ratio, 0.49 [0.18–1.38]), and among patients with intraoperative hypotension, stroke patients did not experience a greater degree of hypotension than controls (ratio of geometric means, 1.07 [0.76–1.53]).

CONCLUSIONS: In our propensity score–matched case-control study, we did not find an association between intraoperative hypotension, defined as MAP < 70 mm Hg, and postoperative stroke.

Supplemental Digital Content is available in the text.

From the *Cleveland Clinic Lerner College of Medicine, Cleveland, Ohio; Departments of Quantitative Health Sciences, Outcomes Research, and §General Anesthesiology, Anesthesiology Institute, Cleveland Clinic, Cleveland, Ohio.

Accepted for publication May 6, 2016.

Funding: This work was directly funded by internal sources only. J.K.H. was supported by a Foundation for Anesthesia Education and Research Medical Student Anesthesia Research Fellowship. None of the authors has a personal financial interest in this research.

The authors declare no conflicts of interest.

Supplemental digital content is available for this article. Direct URL citations appear in the printed text and are provided in the HTML and PDF versions of this article on the journal’s website.

Reprints will not be available from the authors.

Address correspondence to Andrea M. Kurz, MD, Departments of General Anesthesiology and Outcomes Research, Cleveland Clinic, 9500 Euclid Ave./Mail Code E3-97A, Cleveland, OH 44195. Address e-mail to

Perioperative stroke is a devastating complication of surgery, with an apparent incidence of approximately 0.1% in noncardiac, nonneurological surgery (general surgery).1 Although commonly defined (as well as by American College of Surgeons National Surgical Quality Improvement Program) as occurring within 30 days after surgery,1,2 most perioperative strokes occur postoperatively (rather than intraoperatively) with the majority occurring within 7 days.3,4 In contrast to strokes in the general population, which have a mortality rate of 12.6%, strokes during or after general surgery have a mortality rate that ranges from 26% to as high as 87% in patients who have had a previous stroke.2,5

The incidence of perioperative stroke varies according to the nature and complexity of the surgical procedure, in addition to other factors such as advanced age, sex, and comorbidities.5,6 Perioperative strokes are most common in patients having cardiac, neurological, and carotid artery surgery.7 For instance, the rate of perioperative stroke in cardiac surgery ranges from 1.9% to 9.7% depending on the complexity of the surgery,7–9 whereas, in general surgery, the risk is considerably lower.1,7,10,11 The pathophysiology of perioperative stroke tends to be ischemic rather than hemorrhagic. In cardiac surgery, strokes are mainly embolic, whereas strokes in general surgery tend to be thrombotic.7 It is unclear why general surgery is associated with thrombotic strokes; however, factors such as endothelial dysfunction and surgically induced hypercoagulability—both of which may persist for days after surgery—may play a role.2

Recently, hypotension has regained interest as a potential risk factor for perioperative stroke. The Perioperative Ischemic Evaluation Study (POISE) trial, completed in 2007, found an increased rate of death in patients receiving metoprolol versus placebo for noncardiac surgery. Increased mortality was associated with a higher rate of postoperative stroke, for which clinically important hypotension had the largest attributable risk.12 However, the contribution of intraoperative versus postoperative hypotension, and hypotension in general, to perioperative stroke remains unclear. A recent study by Bijker et al,13 for example, reported that the duration of time that mean arterial pressure (MAP) was decreased >30% from baseline was significantly associated with postoperative stroke. In contrast, several older studies identified no relationship between intraoperative hypotension and stroke, while maintaining that postoperative hypotension may contribute.3,10,14

We therefore tested the hypothesis that intraoperative hypotension is associated with perioperative stroke. Specifically, we evaluated (1) whether patients with postoperative stroke were more likely to become hypotensive (as defined by MAP < 70 mm Hg at any point during surgery) and (2) whether hypotensive events in stroke patients were more severe than in patients without stroke.

Back to Top | Article Outline


The study was approved by the institutional review board, which waived the requirement for individual consent. From the Cleveland Clinic Perioperative Health Documentation System registry, we obtained data on 106 337 adult, American Society of Anesthesiologists Physical Status I–IV patients who had elective noncardiac, nonneurological, and noncarotid surgery with general anesthesia at Cleveland Clinic between January 2005 and December 2011. This was the maximum number of patients available to us at the time of study design. Only the most recent visit per patient was included, and the index surgery within a given visit was the first in which there was an anesthesia record. A patient flow diagram for our study is given in Figure 1.

Figure 1

Figure 1

We first identified potential postoperative stroke cases by searching for procedure codes possibly associated with a postoperative stroke. A manual chart review then ensued to identify which of these patients indeed suffered a postoperative stroke. We defined postoperative stroke as those diagnosed up to 30 days after surgery. This was a critical step because the procedure codes often referred to previous strokes rather than postoperative events during the index visit.

Potential controls were identified after excluding those with incomplete demographic, morphometric, or perioperative data (in addition to the exclusion criteria given above) and were matched in a 4-to-1 ratio to stroke cases using propensity scores,15 restricting successful matches to those with common principal procedure (according to the US Agency for Healthcare Research and Quality’s Clinical Classifications Software) and propensity scores within 0.01 propensity score units of one another. Partial matches, that is, those stroke patients for whom 4 suitable controls were not available, were allowed.

Back to Top | Article Outline

Statistical Analysis

Propensity scores were estimated based on patients’ baseline comorbidities using multivariable logistic regression. Balance on baseline comorbidities was assessed using standardized difference scores (defined as the between-group difference in means, mean rankings, or proportions, as appropriate, divided by a combined estimate of standard deviation); any baseline variable exhibiting a standardized difference of >0.2 among the matched sample was used for adjustment in all subsequent analyses. Duration of surgery was explicitly entered into our final models as we sought to ensure strict adjustment for duration of surgery.

The primary hypotension exposure was defined as the integrated area above the MAP-versus-time curve and below the threshold of MAP = 70 mm Hg. This area under the threshold (AUT) metric, depicted in Figure 2, combines both duration and severity of hypotensive episodes into a singular measure. For example, a hypotensive episode of MAP = 60 for 10 minutes, as well as an episode of MAP = 65 for 20 minutes, would each represent an AUT of 100 mm Hg-minutes.

Figure 2

Figure 2

The distribution of the AUT measure was summarized separately for matched stroke cases and controls using standard numerical and graphical techniques. Because we expected many patients to have no hypotension below the MAP = 70 mm Hg threshold, we used a modeling technique that simultaneously models the following parameters: (1) the odds ratio for any positive AUT value comparing stroke patients with controls; and (2) the ratio of geometric mean AUT between stroke patients and controls among those with positive AUT values. The first parameter addresses the hypothesis that stroke patients may have been more likely to become hypotensive, whereas the second parameter addresses the hypothesis that, among hypotensive patients, those with perioperative stroke had more severe exposures to low MAP values. Zero-inflated negative binomial regression—adjusting for any imbalanced baseline characteristics—was used to estimate these parameters.16

To interrogate the effect of our choice of MAP threshold, as well as the timing of postoperative stroke on our results, sensitivity analyses were conducted using MAP thresholds of 65 and 60 mm Hg, as well as by restricting our analysis to strokes that occurred within 3 postoperative days and within 9 postoperative days. R statistical software version 3.0.0 (The R foundation for Statistical Computing, Vienna, Austria) was used for all analysis. The type I error rate for all hypothesis tests was fixed at 5%.

Back to Top | Article Outline


Among 106 337 patients meeting our inclusion criteria, we initially identified 933 with International Classification of Diseases, Ninth Revision (ICD-9) codes potentially related to strokes, of which we confirmed perioperative strokes in 120 based on our manual chart review. Among 105 404 potential control patients, we excluded 8100 because of incomplete data, leaving 97 304 available for matching. Patient flow is shown in Figure 1, and the timing of postoperative stroke is represented in Figure 3.

Figure 3

Figure 3

The 4-to-1 propensity matching procedure yielded a final matched sample of 502 patients: 104 stroke cases and 398 controls. Balance was generally excellent among the matched sample, as shown in Tables 1 and 2. Thus, no variables (other than duration of surgery) were included as covariates in our regression modeling.

Table 1

Table 1

Table 2

Table 2

Distributions of the AUT measure among matched stroke cases and among matched controls were similar (Figure 3). Hypotension (ie, any MAP < 70 mm Hg) was observed in 77 of 104 (74%) stroke cases and 310 of 398 (78%) controls. Median (first quartile, third quartile) AUT values for hypotensive cases and controls were 19 (4, 55) mm Hg-minutes and 19 (6, 48) mm Hg-minutes, respectively.

Based on our zero-inflated negative binomial model, we found no associations between stroke and intraoperative hypotension as measured by AUT. The odds ratio (95% confidence interval) for any positive area (ie, any MAP < 70 mm Hg) comparing stroke cases with controls was estimated at 0.49 (0.18–1.38), which was not statistically significant (P = .18, Wald test for regression model coefficients). Among patients experiencing any intraoperative hypotension (MAP < 70 mm Hg), severity of the hypotension did not differ significantly in patients who did and did not have perioperative strokes (ratio of geometric means [95% confidence interval] of 1.07 [0.76–1.53]; P = .69).

Table 3

Table 3

Sensitivity analyses were performed using threshold MAP values of 65 and 60 mm Hg to define the upper bound of our AUT hypotension measure. These also failed to uncover any statistically significant relationship between intraoperative hypotension and postoperative stroke (Table 3; Supplemental Digital Content 1, Supplemental Figure 1,.

Back to Top | Article Outline


We did not identify any statistically significant or clinically important relationship between intraoperative hypotension and perioperative stroke. This differs from a previous study by Bijker et al,13 which found that time spent >30% below baseline was associated with an increased risk of postoperative stroke. We note, however, that those investigators also found no relationship between intraoperative hypotension and postoperative stroke when using several different definitions of hypotension including raw systolic blood pressures below thresholds of 100, 90, 80, and 70 mm Hg; MAPs <70, 60, 50, and 40 mm Hg; and decreases in both systolic and mean blood pressure by 10%, 20%, and 40% from baseline.

We elected a priori not to use measures of hypotension defined relative to patient baseline blood pressure. The definition of “baseline” varies among the literature and, depending on the definition used, may affect the degree of intraoperative hypotension recorded.17 The most consistently available baseline blood pressure values—those immediately before induction and at preoperative clinic visits—may be affected by factors such as perioperative medications, anxiety or white coat hypertension, and acute disease-related suffering; as such, they may poorly reflect patients’ true baseline blood pressure levels. Rather than incorporating these variable and potentially unreliable baseline levels into our measure of hypotension, we accept the limitation that substantial interindividual differences in baseline blood pressure may decrease our ability to detect a relationship between hypotension and postoperative stroke.

Our study differs from Bijker et al17 in not only matching patients and controls by age and type of surgery but also in the same principal procedure. By using propensity scores, which incorporate patient comorbidities, we were able to tightly match patients who did and did not experience perioperative strokes. We also excluded procedures involving the carotid arteries or proximal aorta because they have substantial potential for atheroembolic events. Finally, our study included approximately 3 times as many stroke cases.

At rest, the brain receives approximately 15% of cardiac output and is able to regulate cerebral blood flow across a range of cerebral perfusion pressures (referred to as cerebral autoregulation).18 Traditionally, the range of blood pressures across which cerebral autoregulation is thought to maintain stable blood flow spans approximately 60 to 150 mm Hg. However, recent evidence suggests that the autoregulation range may be smaller and that cerebral blood flow may be more sensitive to hypotension than hypertension.19

Studies of cerebral autoregulation in humans are complicated by the need for pharmacological agents to achieve large perturbations in blood pressure homeostasis—which may affect cerebrovascular autoregulatory mechanisms—and by difficulties in accurate quantification of cerebral blood flow.19,20 Furthermore, autoregulation during hypotension may be compromised by chronic hypertension, increased intracranial pressure, atherosclerosis, and other disease states.21–24 A recent study of cerebral blood flow during cardiopulmonary bypass, for example, reported that the lower limit of autoregulation varied widely between patients and had no relationship with preoperative MAP.25 These factors suggest that determination of a critical lower limit of blood pressure is complicated and highly variable between patients. In our study, we used a measure of hypotension that combined both the severity and the duration of hypotension experienced. In this way, our measure maintains a degree of sensitivity to differences in severity of hypotension between patients even if the threshold under investigation is higher than the “true” critical lower limit of blood pressure.

It remains possible that intraoperative hypotension contributes to perioperative strokes. Indeed, severe and prolonged intraoperative hypotension as a complication of surgery inevitably results in hypoxic damage to end organs, including the brain. A recent study reported that even short durations of hypotension (defined as MAP < 55 mm Hg) during noncardiac surgery were associated with acute kidney injury (AKI) and myocardial injury.26 The relative scarcity of perioperative strokes (compared with postoperative AKI or myocardial infarction) and small number of patients in our sample experiencing severe degrees of intraoperative hypotension preclude the use of this threshold in our analysis. Our sensitivity analyses failed to find any relationship between hypotension and stroke at more severe degrees of hypotension (<65 and <60 mm Hg). Our results suggest that within the typical range of intraoperative blood pressures experienced at our institution, factors other than hypotension contribute more.

We also defined postoperative stroke as occurring up to 30 days after surgery. Although severe hypotensive events or surgical complications may result in strokes that manifest immediately after surgery, the majority of strokes occur between 1 day and 1 week postoperatively, and we sought to include these in our investigation. Any proposed pathological mechanism connecting intraoperative hypotension and postoperative stroke becomes more tenuous with an increasing number of postoperative days; we thus also conducted analyses restricted to strokes only occurring within 3 and 9 postoperative days. There was no substantial change in the results (Supplemental Digital Content 3, Supplemental Table 1, Postoperative hypotension may be a greater contributor to later-occurring strokes than intraoperative hypotension. Postoperative blood pressure measurement is comparatively infrequent and, during the routine assessment of vital signs, patients are often stimulated, which might mask ongoing hypotensive events. The incidence of postoperative hypotension remains poorly characterized and is the subject of current investigation.

We are also limited by our definition of hypotension as <70 mm Hg and our power to detect an association at progressively greater degrees of intraoperative hypotension. A threshold of 70 mm Hg likely represents normal cerebral perfusion for a supine patient; yet, as demonstrated in Figure 4, a substantial proportion of our patients did not experience any hypotension even under this threshold. Because our measure of hypotension includes both depth and length of exposure, this renders it somewhat resilient to the choice of threshold (ie, patients with more severe hypotension will have greater exposure in the model under any chosen threshold regardless of how high); however, the relative scarcity of stroke and further reduction in the number of exposed patients under the threshold decrease our power.

Figure 4

Figure 4

We also conducted sensitivity analyses at lower thresholds for hypotension, but for these analyses the number of exposed patients—and, therefore, our power—is reduced even further (Supplemental Digital Content 1, Supplemental Figure 1,; and Supplemental Digital Content 2, Supplement Figure 2, It may be that we are underpowered to detect an association for this rare event using any of our thresholds. Nonetheless, our sample of over 100 postoperative strokes is the largest ever studied in this population.

Our study incorporated an observational, case-control design; it is thus possible that our results might be influenced by confounding variables unavailable in our registry. However, a randomized trial of intraoperative hypotension and postoperative stroke would be ethically challenging.

In summary, we did not find evidence of a relationship between intraoperative hypotension <70 mm Hg and postoperative stroke in adults having nonneurological, noncardiac, and noncarotid surgeries. It thus seems likely that factors other than blood pressure contribute more to the risk of postoperative stroke.

Back to Top | Article Outline


Name: Jason K. Hsieh, BS.

Contribution: This author helped design the study, collect the data, participate in data analysis, and prepare the manuscript.

Name: Jarrod E. Dalton, PhD.

Contribution: This author helped to design the study, participated in data collection, and performed the data analysis.

Name: Dongsheng Yang, MS.

Contribution: This author helped design the study and participate in data collection.

Name: Ehab S. Farag, MD.

Contribution: This author helped design the study.

Name: Daniel I. Sessler, MD.

Contribution: This author helped design the study and prepare the manuscript.

Name: Andrea M. Kurz, MD.

Contribution: This author helped design the study and prepare the manuscript.

This manuscript was handled by: Gregory Crosby, MD.

Back to Top | Article Outline


1. Mashour GA, Shanks AM, Kheterpal S. Perioperative stroke and associated mortality after noncardiac, nonneurologic surgery. Anesthesiology. 2011;114:1289–1296.
2. Ng JL, Chan MT, Gelb AW. Perioperative stroke in noncardiac, nonneurosurgical surgery. Anesthesiology. 2011;115:879–890.
3. Hart R, Hindman B. Mechanisms of perioperative cerebral infarction. Stroke. 1982;13:766–773.
4. Parikh S, Cohen JR. Perioperative stroke after general surgical procedures. N Y State J Med. 1993;93:162–165.
5. El-Saed A, Kuller LH, Newman AB, et al. Geographic variations in stroke incidence and mortality among older populations in four US communities. Stroke. 2006;37:1975–1979.
6. Wong GY, Warner DO, Schroeder DR, et al. Risk of surgery and anesthesia for ischemic stroke. Anesthesiology. 2000;92:425–432.
7. Selim M. Perioperative stroke. N Engl J Med. 2007;356:706–713.
8. Bucerius J, Gummert JF, Borger MA, et al. Stroke after cardiac surgery: a risk factor analysis of 16,184 consecutive adult patients. Ann Thorac Surg. 2003;75:472–478.
9. McKhann GM, Grega MA, Borowicz LM Jr, Baumgartner WA, Selnes OA. Stroke and encephalopathy after cardiac surgery: an update. Stroke. 2006;37:562–571.
10. Kam PC, Calcroft RM. Peri-operative stroke in general surgical patients. Anaesthesia. 1997;52:879–883.
11. Bateman BT, Schumacher HC, Wang S, Shaefi S, Berman MF. Perioperative acute ischemic stroke in noncardiac and nonvascular surgery: incidence, risk factors, and outcomes. Anesthesiology. 2009;110:231–238.
12. Devereaux PJ, Yang H, Yusuf S, et al. Effects of extended-release metoprolol succinate in patients undergoing non-cardiac surgery (POISE trial): a randomised controlled trial. Lancet. 2008;371:1839–1847.
13. Bijker JB, Persoon S, Peelen LM, et al. Intraoperative hypotension and perioperative ischemic stroke after general surgery: a nested case-control study. Anesthesiology. 2012;116:658–664.
14. Limburg M, Wijdicks EF, Li H. Ischemic stroke after surgical procedures: clinical features, neuroimaging, and risk factors. Neurology. 1998;50:895–901.
15. Rosenbaum PR, Rubin DB. The central role of the propensity score in observational studies for causal effects. Biometrika. 1983;70:41–55.
16. Lambert D. Zero-inflated Poisson regression, with an application to defects in manufacturing. Technometrics. 1992;34:1–14.
17. Bijker JB, van Klei WA, Kappen TH, van Wolfswinkel L, Moons KG, Kalkman CJ. Incidence of intraoperative hypotension as a function of the chosen definition: literature definitions applied to a retrospective cohort using automated data collection. Anesthesiology. 2007;107:213–220.
18. Lassen NA. Cerebral blood flow and oxygen consumption in man. Physiol Rev. 1959;39:183–238.
19. Willie CK, Tzeng YC, Fisher JA, Ainslie PN. Integrative regulation of human brain blood flow. J Physiol. 2014;592:841–859.
20. Hamner JW, Tan CO. Relative contributions of sympathetic, cholinergic, and myogenic mechanisms to cerebral autoregulation. Stroke. 2014;45:1771–1777.
21. Strandgaard S, Paulson OB. Cerebral blood flow and its pathophysiology in hypertension. Am J Hypertens. 1989;2:486–492.
22. Guo ZN, Liu J, Xing Y, et al. Dynamic cerebral autoregulation is heterogeneous in different subtypes of acute ischemic stroke. PLoS One. 2014;9:e93213.
23. Brady KM, Lee JK, Kibler KK, et al. The lower limit of cerebral blood flow autoregulation is increased with elevated intracranial pressure. Anesth Analg. 2009;108:1278–1283.
24. Pesek M, Kibler K, Easley RB, et al. The upper limit of cerebral blood flow autoregulation is decreased with elevations in intracranial pressure. Neurosurgery. 2014;75:163–170.
25. Joshi B, Ono M, Brown C, et al. Predicting the limits of cerebral autoregulation during cardiopulmonary bypass. Anesth Analg. 2012;114:503–510.
26. Walsh M, Devereaux PJ, Garg AX, et al. Relationship between intraoperative mean arterial pressure and clinical outcomes after noncardiac surgery: toward an empirical definition of hypotension. Anesthesiology. 2013;119:507–515.

Supplemental Digital Content

Back to Top | Article Outline
© 2016 International Anesthesia Research Society