Intraoperative Hypotension and Perioperative Ischemic Stroke after General Surgery: A Nested Case-control Study

• The etiology of perioperative stroke is multifactorial, however, the recent POISE trial findings suggested that hypotension might contribute to the risk of stroke

• The results of this case-control study indicated that the duration of hypotension, defined as 30% reductions in mean blood pressure from baseline, was significantly associated with postoperative stroke in patients undergoing non-cardiac, non-neurosurgical surgery

• Further investigation is required to determine if specific perioperative blood pressure management strategies might mitigate such risk

STROKE is a rare but serious postoperative complication. Depending on the type and complexity of surgery, an ischemic stroke occurs in 0.1–3% of patients undergoing general surgery and as many as 10% of patients after complex cardiac surgery.1,– 3 

Embolism often is considered the primary cause of a postoperative ischemic stroke. It may be related to postoperative atrial fibrillation or surgery-induced hypercoagulability in combination with vulnerable plaques in carotid or major cerebral arteries.1Hypoperfusion, defined as any combination of extracranial stenosis and/or systemic hypotension, is reported to be responsible for only 9% of all postoperative strokes in cardiac surgery patients.4In other types of surgery, no association between intraoperative hypotension (IOH) and postoperative stroke has been found.5,–,7However, the recent results of the Perioperative Ischemic Evaluation Study (POISE) trial (which investigated the effect of metoprolol vs.  placebo on cardiovascular events in 8,351 noncardiac surgery patients) have renewed interest in hypotension as a possible cause for the increase in postoperative stroke.8Hypotension may result in low flow or so-called watershed infarcts, but this type of stroke occurs only occasionally during surgery.9,–,11Still, it is likely that the mechanism of a postoperative stroke is multifactorial and that even in embolic strokes, IOH may aggravate the clinical course by increasing the size of the infarcted area.12 

In the current study, we explore the hypothesis that the duration of IOH is associated with the occurrence of a postoperative ischemic stroke and that this association depends on the chosen definition of IOH. To this aim, using a range of frequently used definitions of IOH, we conducted a case-control study in a large cohort of patients.

The current study was designed as a case-control study nested in a large retrospective cohort of patients undergoing anesthesia for noncardiac and nonneurosurgical procedures. The local hospital ethics committee (University Medical Center Utrecht, Utrecht, The Netherlands) approved the study protocol. Because patients were not subjected to investigational actions and all patients were treated with care as usual, no written informed consent was necessary.

All consecutive patients who underwent anesthesia for inpatient surgical procedures at the University Medical Center Utrecht (a 1,042-bed academic hospital) in the period from January 1, 2002, to June 30, 2009, were included. Patients undergoing cardiac or neurosurgical procedures were excluded.

Preoperative patient data were collected from preoperative evaluation clinic electronic patient records. These consisted of both biometric data (such as age, gender, weight, and preoperative blood pressure) and medical history. Intraoperative data from the patient monitor and anesthesia machine (such as blood pressure and heart rate) were stored every minute in an electronic anesthesia record-keeping system. Postoperative data were collected from the hospital information system and patient charts. An investigator blinded to the status of the patients (case or control) collected all patient data.

Intraoperative hypotension was defined according to a range of threshold values designated a priori . These blood pressure thresholds, both absolute (in mmHg) and relative to a baseline (in percentage decrease from baseline), were based on the most frequently used definitions from the anesthesia literature.13The selected range of blood pressure thresholds consisted of a systolic blood pressure less than 100, 90, 80, and 70 mmHg, a mean blood pressure less than 70, 60, 50, and 40 mmHg, and a decrease in systolic or mean blood pressure of 10%, 20%, 30%, and 40% from baseline. The continuous IOH variable was expressed as the number of minutes that the blood pressure was below the designated threshold.

The baseline blood pressure was defined as the mean of the blood pressure measured at the preoperative evaluation clinic (measured on either arm with the patient in the sitting position) and all available blood pressure measurements in the operating room before induction of anesthesia. The time of induction of anesthesia was estimated using a previously published algorithm implemented with LabView software (version 8; National Instruments Corporation, Austin, TX) and defined as either the moment of administration of induction medication or 3 min before the first appearance of continuous expired carbon dioxide registration, whichever came first.13In cases of spinal or epidural anesthesia, the time of puncture was taken as the time of induction.

Cases were patients with an ischemic stroke (defined as the acute onset of new focal neurologic deficit of cerebral origin persisting more than 24 h without hemorrhage on computed tomography [CT]) within 10 postoperative days. Thus, potential stroke cases were selected as all patients in whom a CT scan of the brain was performed within 10 days after surgery. The diagnosis of ischemic stroke was made by chart review, and all CT scans were reviewed again by two neurologists (SP and LJK), who were blinded to the intraoperative course. Diagnoses made on clinical grounds (i.e. , without CT abnormalities) were also included as cases. The condition of the stroke patients at discharge from the hospital was scored on the modified Rankin scale.14 

For every patient with a postoperative stroke, six control patients were selected who underwent surgery during the same period (as close as possible to the date of surgery of the case) but who did not experience a postoperative stroke. A ratio of approximately 1:5 is known to yield the best results taking into account the effort to collect the information of the controls.15Because most data were stored electronically and relatively easy to collect, we chose a ratio of 1:6. Cases and controls were matched on age and type of surgery. For age, we used frequency matching, using age groups covering a decade (the control had to have the age of the case plus or minus 5 yr). When no control with an identical type of surgery was available, a patient who underwent the most similar procedure was selected.

Variables potentially confounding the association between the duration of IOH and postoperative stroke were predefined based on existing knowledge and the literature and included: age, gender, comorbidity, antihypertensive medication, type and duration of surgery, and type of anesthesia.1,–,3,16,17Comorbidity was included both as the American Society of Anesthesiologists class and as a history of diabetes, atrial fibrillation, previous stroke, or hypertension. Type of surgery was categorized into risk groups according to previous studies: general surgery, peripheral vascular surgery, resection of head and neck tumors, carotid endarterectomy (CEA), and aortic repair.1 

Differences in patient characteristics and potential confounders between the cases and controls were compared using the Student t  test for normally distributed continuous variables and the Mann–Whitney U test for continuous variables that were not normally distributed. Categorical variables were compared with the Pearson chi-square test or the Fisher exact test where appropriate.

Conditional logistic regression analysis was used to estimate univariate associations between the duration of IOH and postoperative stroke for the different definitions. Subsequently, we adjusted for the above-defined potential confounders except for type of surgery because cases and controls were matched on this variable. Age was adjusted for because we used frequency matching using 10-yr intervals.

Because a range of blood pressure thresholds was used for subsequent models, correction for multiple testing was deemed necessary. Thus, in the conditional logistic regression analyses we tested against a P  value of 0.001 and used 99.9% confidence intervals.18 

All analyses were performed using R (release 2.13.1; R Foundation for Statistical Computing, Vienna, Austria).

During the study period, 48,241 patients underwent a noncardiac, nonneurosurgical procedure. The median duration between preoperative evaluation and surgery was 26 days. In 53 patients, the CT report was suggestive of a stroke. After the hospital charts and original CT scans were reviewed, 42 patients (0.09%) were identified as definitive stroke cases. Of these, 32 had the diagnosis based on CT abnormalities (26 on initial CT scan and 6 on delayed CT scan) and 10 on clinical signs without CT abnormalities. The 11 excluded patients had a diagnosis of hyperperfusion syndrome after CEA (n = 7), postanoxic encephalopathy secondary to a cardiac arrest 1 day after CEA (n = 1), epileptic insult (n = 1), and an anterior spinal cord syndrome after aortic surgery (n = 1). During the study period, a temporal trend in the occurrence of postoperative strokes could not be observed. The characteristics of the cases with postoperative ischemic stroke are presented in table 1.

Subsequently, 252 control patients were selected. One stroke patient underwent surgery of the brachiocephalic artery. This patient could be matched to only one patient with an identical procedure. The other five controls were matched using CEA patients. Similarly, two stroke patients who underwent surgery of the subclavian artery could not be matched to patients who underwent identical procedures. Again, CEA patients were used for matching. All these cases and their matched controls were analyzed as CEA patients. The median difference between the surgery date of the case and control was 20 months (interquartile range 9–36 months).

The characteristics of the stroke and control patients are presented in table 2. None of the crude associations between duration of IOH and postoperative stroke reached statistical significance (data not shown). After accounting for confounding and multiple testing, the durations that the mean blood pressure was decreased more than 30% from baseline were associated with the occurrence of a postoperative stroke (odds ratio = 1.013/min hypotension, 99.9% confidence interval = 1.000–1.025; table 3). None of the associations for the other IOH threshold values reached statistical significance at the 0.001 level. The adjusted odds ratios are illustrated in figure 1, where the 99.9% confidence intervals for the thresholds relative to a baseline blood pressure (panels C and D) are much narrower than those for the absolute blood pressure thresholds (panels A and B).

Post hoc  sensitivity analyses revealed no effect from type of surgery (data not shown), but the odds ratios for decreasing mean blood pressures relative to a baseline appeared larger (not statistically significant) in the 22 patients (52%) whose stroke occurred within 24 h after surgery (fig. 2). This effect could not be observed for other thresholds.

In this explorative case-control study, we investigated the possible association of IOH and the occurrence of a postoperative ischemic stroke within 10 days after general surgery. The observed stroke rate was 0.09%. After correction for potential confounding and multiple testing, the duration that the mean blood pressure was decreased more than 30% from baseline blood pressure remained statistically significantly associated with the occurrence of ischemic stroke.

Some limitations of the study need to be discussed. First, in the CEA patients, no routine preoperative ultrasound examination of the cerebropetal arteries was performed. Occlusion of the contralateral internal carotid artery is associated with an increased risk of postoperative stroke.19However, it is unlikely that preoperative contralateral carotid occlusion is associated with IOH, so we do not believe this significantly influenced the relation between IOH and postoperative stroke.

In addition, the period in which a stroke is considered a perioperative stroke varies between 3 and 30 postoperative days.3We chose a period of 10 postoperative days to find a compromise between finding too few cases with a very short period and a questionable association with postoperative events when using a very long period. To investigate the influence of IOH on the occurrence of stroke early after surgery, a post hoc  sensitivity analysis with stroke cases within 24 h after surgery was performed.

Moreover, as with IOH, it is likely that postoperative hypotension plays a comparable role. Postoperative hypotension might be even more important than hypotension in the highly monitored and controlled intraoperative and early postoperative period, especially because a postoperative stroke most commonly occurs after an asymptomatic interval.5It is possible that IOH predicts postoperative hemodynamic instability. However, postoperative blood pressures were measured at highly variable intervals, ranging from continuous or at least hourly at the intensive care unit to once or twice per shift on the ward. Thus, finding no hypotensive measurements in the patient charts does not imply that no postoperative hypotension occurred. On the contrary, when a low blood pressure reading was found in the patient chart, no information on the length or severity of the hypotensive episode was provided. Thus, we regarded the postoperative blood pressure data from the patient charts as unreliable and did not use them in the analyses to prevent significant bias.

In theory, it could also be possible that IOH is a consequence and not the cause of a postoperative stroke. However, it is more common for the blood pressure to increase after a stroke. In addition, IOH could have been preceded by a stroke only when the stroke occurred during surgery, which is very rare. Thus, we think this is an unlikely explanation for the possible association between IOH and postoperative stroke.

In the current study, only 42 stroke cases were found. Because this was an observational, hypothesis-generating study, we used all stroke cases that we could find and did not perform a priori  sample-size calculations. These stroke cases were matched to control subjects with regard to age and type of surgery. The control was then chosen so that the date of surgery was as close to that of the case as possible. Still, the median time between the surgery of the case and control was 20 months (interquartile range 9–36 months). Nevertheless, because no temporal trend in the occurrence of postoperative strokes was observed, it is not likely this has influenced our findings to a great degree.

Another factor to be taken into account is the need to correct for multiple testing because of the range of IOH definitions that were used. Because the different definitions of IOH are correlated, traditional methods such as the Bonferroni correction are likely to be too conservative. Thus, we chose to consider a P  value <0.001 to be statistically significant. It should be noted that although the range of cutoff values for IOH might give the impression that the blood pressure was included as a dichotomized variable, this was not the case. In all analyses, IOH was used as a continuous variable, expressed as the number of minutes the blood pressure was below the designated IOH threshold.

After publication of the results of the POISE trial, IOH received renewed interest as a possible explaining factor for increased stroke rate (from 0.5% in the control group to 1.0% in the metoprolol group).8The results of the current study are consistent with this hypothesis, although the magnitude of the effect is much less. Our results suggest that IOH accounts for an increase in stroke risk of approximately 1.3% per minute hypotension (i.e. , the risk is increased 1.013 times for every minute of hypotension), depending on the definition of IOH that is used (in this case a decrease in mean blood pressure more than 30% from baseline). For example, a cumulative duration of 10 min of hypotension will result in a 1.14 times increased stroke risk (1.013¹⁰). If applied to the POISE trial, this would mean an increase in absolute stroke risk from 0.5% (POISE trial control patients) to 0.57%. This difference with the POISE trial might be explained by the wide variety of β blockers in different dosages that patients used in the current study as opposed to the high-dose metoprolol patients received in the POISE trial.20 

The most-often proposed mechanism of a postoperative stroke is an embolism originating from the heart or great vessels. Hypoperfusion, leading to a watershed infarction, is thought to be responsible for only 9% of ischemic strokes after cardiac surgery, and deliberate hypotension does not seem to adversely affect cerebral perfusion.1Nevertheless, in the current study we found seven (17%) strokes in watershed areas. In addition, we hypothesized that even in the presence of an embolic stroke, hypotension might aggravate the clinical course by compromising blood flow to potential ischemic, but still viable, brain areas (the so called penumbra). Support for this hypothesis is found in studies on cross-clamping of noncritical segmental arteries of pigs, which makes the spinal cord more vulnerable to hypotension.21Thus, the current study suggests that IOH may indeed play a role, albeit limited, in the occurrence of a postoperative stroke.

Defining IOH using blood pressure thresholds relative to a patient's baseline blood pressure offers more information (association with postoperative stroke when mean blood pressure decreases more than 30% from baseline and smaller confidence intervals) on associations with postoperative stroke than does using fixed blood pressure levels. When generalized, this supports the clinical reasoning that defining IOH relative to a baseline blood pressure is a better way to assess a patient's risk on adverse outcome than using fixed, arbitrarily chosen threshold values, disregarding a patient's comorbidities.13 

However, the combination of a low stroke incidence, the complexity of the postoperative stroke mechanism, and the absence of a uniform definition of IOH makes it challenging to study the association between IOH and postoperative stroke. This is further hampered by the limited effect that IOH appears to have on the occurrence of a postoperative stroke. Thus, these associations should be interpreted cautiously and do not automatically imply causality. Nevertheless, although case-control studies generally are considered to be hypothesis generating, these hypotheses are of valuable clinical importance because perioperative blood pressures are controllable, potentially providing an opportunity to alter the postoperative stroke risk, especially in high-risk patients.

In conclusion, the most widely proposed mechanism of a postoperative stroke is arterial embolism. Nonetheless, the results of the current study support the hypothesis that hypotension can influence the evolution of a postoperative stroke by compromising (collateral) blood flow to ischemic areas. In this context, hypotension is best defined as a decrease in mean blood pressure relative to a preoperative baseline, rather than an absolute low blood pressure value.