The outcomes of emergent noncardiac, nonintracranial surgery in patients with previous stroke remain unknown.
All emergency surgeries performed in Denmark (2005 to 2011) were analyzed according to time elapsed between previous ischemic stroke and surgery. The risks of 30-day mortality and major adverse cardiovascular events were estimated as odds ratios (ORs) and 95% CIs using adjusted logistic regression models in a priori defined groups (reference was no previous stroke). In patients undergoing surgery immediately (within 1 to 3 days) or early after stroke (within 4 to 14 days), propensity-score matching was performed.
Of 146,694 nonvascular surgeries (composing 98% of all emergency surgeries), 5.3% had previous stroke (mean age, 75 yr [SD = 13]; 53% women, 50% major orthopedic surgery). Antithrombotic treatment and atrial fibrillation were more frequent and general anesthesia less frequent in patients with previous stroke (all P < 0.001). Risks of major adverse cardiovascular events and mortality were high for patients with stroke less than 3 months (20.7 and 16.4% events; OR = 4.71 [95% CI, 4.18 to 5.32] and 1.65 [95% CI, 1.45 to 1.88]), and remained increased for stroke within 3 to 9 months (10.3 and 12.3%; OR = 1.93 [95% CI, 1.55 to 2.40] and 1.20 [95% CI, 0.98 to 1.47]) and stroke more than 9 months (8.8 and 11.7%; OR = 1.62 [95% CI, 1.43 to 1.84] and 1.20 [95% CI, 1.08 to 1.34]) compared with no previous stroke (2.3 and 4.8% events). Major adverse cardiovascular events were significantly lower in 323 patients undergoing immediate surgery (21%) compared with 323 successfully propensity-matched early surgery patients (29%; P = 0.029).
Adverse cardiovascular outcomes and mortality were greatly increased among patients with recent stroke. However, events were higher 4 to 14 days after stroke compared with 1 to 3 days after stroke.
There is a steep decline and stabilization of risks within the first 9 months after stroke among patients undergoing elective surgery, but risks after emergency surgery are unknown
After emergency noncardiac, nonintracranial surgery, risks of 30-day major adverse cardiovascular events (acute myocardial infarction, ischemic stroke, or cardiovascular death) were high for patients with stroke less than 3 months before surgery (odds ratio [OR] = 4.7), 3 to 9 months (OR = 1.9) and more than 9 months (OR = 1.6) compared with no previous stroke
Risks of death (1.6, 1.2, and 1.2) in the same period were also increased
Risk of major adverse cardiovascular events was significantly lower after immediate (1 to 3 days after stroke) compared with early surgery (4 to 14 days)
These patterns were similar to that observed in poststroke patients having elective surgery
EMERGENCY surgery is one of the most high-risk situations encountered in clinical practice, and risks become especially pronounced in patients with weighty comorbidities, such as cerebrovascular disease.1 Because time is an important factor in the clinical setting, detailed information on perioperative risks is important to guide decision-making and to inform the patient and relatives about realistic perioperative expectations. More than 800,000 patients experience a stroke in the United States every year, with the majority being ischemic strokes,2 and the World Health Organization estimates that incidence rates of stroke in Europe will increase by approximately 1.5 million per year from 2000 to 2025.3 Emergency surgeries in patients with previous stroke are becoming more frequent due to advances in treatment and an increased willingness to treat older and more fragile patients. Stroke has been included in widely used preoperative risk evaluation schemes, such as the revised cardiac risk index by Lee et al.,4 but the importance of time elapsed between stroke and emergency surgery has not previously been reported and is not considered in current perioperative guidelines.5,6 A recent study demonstrated a steep decline and stabilization of risks within the first 9 months after stroke among patients undergoing elective surgery.7 In patients with previous myocardial infarction, a similar decrease in risks followed by stabilization at 6 months has been demonstrated, and perioperative guidelines do address the importance of time between stenting and noncardiac surgery.8,9 The mechanism for increased risks in patients with recent stroke are thought to be at least partly mediated by deteriorating cerebral autoregulation within the first 5 days after stroke and impaired autoregulation for up to 3 months.10,11 This condition renders patients vulnerable to secondary neuronal injury during anesthesia, surgical stress, and perioperative care. Indeed, a recent strategy to improve outcomes from emergency surgery includes very early operation, and a recent guideline presented several arguments in favor of surgical repair of hip fractures within 48 h, but does not provide specific guidance on whether this is appropriate in patients with very recent strokes.12 Herein we hypothesized that very early or more delayed surgery would be associated with better outcomes than surgery conducted at an intermediate time point when autoregulation may be maximally dysregulated.
We investigated the association between time elapsed after ischemic stroke and the risk of adverse events after emergency noncardiac surgery. Although we recognize that postponing emergency surgeries is rarely an option, choosing to operate earlier is sometimes a possibility. Furthermore, our results may provide additional guidance for clinical decision-making and lead to a better understanding of expected perioperative outcomes in this growing and critically ill patient population.
Materials and Methods
This study was approved by the Danish Data Protection Agency, Copenhagen, Denmark (reference No. GEH-2014–019 and I-Suite No. 02737). Register-based studies using depersonalized data do not require ethics committee approval or patient consent in Denmark.
In Denmark, all of the citizens have a unique personal identification number, which enables identification and linkage of individual patients across several Danish registers. For this study, we retrieved information on diagnoses at hospital discharge and surgical procedures from the Danish National Patient Register, complete since 1977.13 All of the information is coded according to international standards including the International Classification of Diseases, Tenth Revision (ICD-10) and the Nordic Medico-Statistical Committee (NOMESCO) Classification of Surgical Procedures.14 From the Danish Anesthesia Database, complete since 2005, we retrieved information on whether a surgery was elective or emergent, as well as information on body mass index, type of anesthesia, and duration of anesthesia, which was collected as part of the daily clinical work. Information on pharmacotherapy before surgery was retrieved from the Danish Register of Medicinal Product Statistics, which holds information on all claimed prescriptions from Danish pharmacies. The specific drug, coded according to the Anatomical Therapeutic Chemical classification system, and the date of dispense were available. The register is linked to government reimbursement and has proven to be accurate.15 The Central Population Register provided information on date of birth and sex. The National Causes of Death Register provided information on cause of death and death date.
Our study population included all emergency noncardiac and nonintracranial surgeries from 2005 to 2011 for patients more than 20 yr of age. As several procedures were closely linked to a preceding stroke, we excluded cardiac surgeries, surgeries to the carotid arteries, gastrostomies, tracheostomies, and surgical procedures to the arteries of the aortic arch, as done previously.16 Patients with repeat surgery within the 30-day follow-up period were only included at first surgery. Surgeries beyond a completed 30-day follow-up period were eligible for analyses. See figure 1, Supplemental Digital Content (http://links.lww.com/ALN/B461), for a flow diagram of the inclusion and exclusion criteria for the study population.
We undertook two analytic approaches. Patients with stroke 14 days or less before surgery were stratified as immediate surgery (1 to 3 days between stroke and surgery) and early surgery (4 to 14 days between stroke and surgery) and underwent propensity score–matched analysis and restricted cubic splines analysis to investigate the short-term relationship between time interval to surgery and perioperative outcomes. Longer time intervals were analyzed by logistic regression and restricted cubic spline analyses with patients, also stratified by time between stroke and surgery as patients with no previous stroke, patients with stroke less than 3 months before surgery (stroke less than 3 months), patients with stroke 3 to 9 months before surgery (stroke 3 to 9 months), and patients with stroke more than 9 months before surgery (stroke more than 9 months). Previous ischemic strokes were defined as cerebral infarction (ICD-10: I63) or unspecified stroke (ICD-10: I64) occurring within 5 yr of surgery, because the majority of unspecified strokes in the Danish National Patient Registry have been shown to be of ischemic origin.17 Transient ischemic attacks and hemorrhagic strokes were not considered because the frequency is low, and the validity and pathophysiology of these diagnoses are different from those of ischemic strokes. Vascular and nonvascular surgeries were analyzed separately in the main analyses, and only nonvascular surgeries were included in sensitivity analyses due to low numbers of vascular surgeries. Vascular surgery was defined as surgery to both arterial and nonarterial vessels according to the NOMESCO classifications (see table 1, Supplemental Digital Content, http://links.lww.com/ALN/B461, which is a list of surgery types according to the NOMESCO classifications).
Study Variables and Outcomes
The uses of pharmacotherapy within 120 days before surgery were grouped according to the Anatomical Therapeutic Chemical classification system. These included lipid modifying agents (C10A), β-blocking agents (C07), agents acting on the renin–angiotensin system (C09), potassium-sparing agents (C03D), thiazides (C03A), calcium channel blockers (C08), digoxin (C01AA05), vitamin K antagonists (B01AA0), glucose-lowering agents (A10), loop diuretics (C03CA01), and antithrombotic therapy as low-dose acetylsalicylic acid (B01AC06), dipyridamole (B01AC07), clopidogrel (B01AC04), or a combination of acetylsalicylic acid and dipyridamole (B01AC30).
Based on ICD-10 coding, we identified previous comorbid conditions as myocardial infarction, chronic obstructive pulmonary disease, anemia, cancer with metastases, renal disease, rheumatic disease, peripheral artery disease, liver disease, diabetes mellitus, chronic heart failure, ischemic heart disease, and atrial fibrillation (coding details are available in table 1, Supplemental Digital Content, http://links.lww.com/ALN/B461). We only considered comorbidities diagnosed or treated within 5 yr before surgery. The diagnoses for comorbidities used in this study have been validated with excellent positive predictive values of 97 to 100% in the Danish National Patient Registry.17
Surgeries were divided into 16 categories in accordance with previously published work14,18 (coding details are available in table 1, Supplemental Digital Content, http://links.lww.com/ALN/B461) and defined as low-, intermediate-, or high-risk surgery in agreement with the European Society of Cardiology perioperative guidelines.5 Our primary outcomes were 30-day all-cause mortality and a combined endpoint of 30-day major adverse cardiovascular events (MACEs), which included nonfatal ischemic stroke, nonfatal myocardial infarction, and cardiovascular death (any cause of death with ICD-10 code I). Nonfatal ischemic stroke was evaluated separately as a secondary endpoint.
Differences between groups at baseline for continuous and categorical variables were tested using the Student’s t test and chi-square test. We used three analytical approaches to assess risks of adverse outcomes, including logistic regression models, spline analyses, and propensity-score matching.
Multivariate logistic regression models were used to estimate odds ratios (ORs) with 95% CIs, adjusted for sex, age, body mass index, comorbidities, pharmacotherapy, type of surgery, and surgery risk. Patients with no previous stroke served as the reference. Analyses were performed for the primary and secondary endpoints stratified by vascular and nonvascular surgery. A few patients with missing data on body mass index (approximately 3% were missing for the stroke patients and 2% for no-stroke patients) were included, but considering the very small amount, no attempts were made to impute or otherwise replace the missing data.
Restricted cubic spline regression models adjusted for sex and age were used to analyze time between stroke and surgery as a continuous variable. These analyses included all of the patients with previous stroke undergoing nonvascular surgery. The median time between stroke and surgery for all of the patients (353 days) and patients with stroke 14 days or less before surgery (2 days) served as the reference. Five knots were placed at the 10th, 25th, 50th, 75th, and 90th percentiles of time between stroke and surgery.19
Among patients undergoing nonvascular surgery we used propensity-score matching to estimate risks of MACEs in patients having immediate surgery (1 to 3 days after stroke) and early surgery (4 to 14 days after stroke). Propensity-score matching was carried out in two steps. First, a logistic regression model was used to estimate the propensity score for undergoing early surgery, based on all of the variables from table 1 (C statistic for this model was 0.681). Next, we used the gmatch-macro, based on a greedy matching algorithm,20 to match patients undergoing early surgery in a 1:1 ratio with patients undergoing immediate surgery. Patients were matched on propensity score (maximum deviation = 0.01), sex, and type of surgery in three groups (abdominal, orthopedic, or other surgery). Differences in baseline characteristics between the propensity score–matched groups were assessed using standardized mean differences, with a value of less than 0.10 indicating minimal imbalance between the groups.
We estimated absolute risks of MACEs in clinically relevant patient subgroups undergoing nonvascular surgery, stratified by the presence of stroke, chronic obstructive pulmonary disease, atrial fibrillation, diabetes mellitus, kidney disease, ischemic heart disease, previous stroke, heart failure, sex, and age more than or less than 70 yr, without additional adjustment. In addition, as a sensitivity analysis, we also estimated absolute risks of a MACE for stroke patients, stratified according to the revised cardiac risk index.4
Several sensitivity analyses were performed in the group of patients undergoing nonvascular surgery and subsequently repeated in a subgroup of patients undergoing minor or major orthopedic surgery. We estimated stroke-associated risks of 30-day MACEs stratified by anesthesia time (less than 120 min vs. 120 min or more), type of anesthesia (general vs. other, which included regional anesthesia and/or sedation), and regular versus odd operation hours (regular being 7:00 am to 4:00 pm on weekdays).
Data management and statistical analyses were performed using SAS version 9.4 (SAS Institute Inc, USA). Figures were compiled using R statistical software version 3.2.2 (https://www.r-project.org; accessed February 5, 2017).
Main analyses were decided on a priori, including patient selection, variables, outcomes, and statistical methods. Subanalyses and sensitivity analyses were decided on post hoc and after inputs from peer review. Throughout the study, efforts were made to comply with the Strengthening the Reporting of Observational Studies in Epidemiology guideline for reporting observational studies.21
All Patients: Baseline
We identified 146,694 emergency surgeries with 7,861 patients (5.4%) having previous stroke. Only 3,509 (2.4%) were vascular surgeries, with 365 patients having previous stroke. Eleven percent of the patients had more than one eligible surgery within the study period.
For patients undergoing nonvascular surgery, more than half were women in all of the groups. Previous stroke patients were 7 to 8 yr older, on average, than patients with no previous stroke. All of the comorbidities were more prevalent in patients with previous stroke compared with no-stroke patients (all P < 0.05). Major orthopedic surgery accounted for 51 to 55% of surgeries in patients with previous stroke compared with only 30% in patients with no previous stroke (table 2, Supplemental Digital Content, http://links.lww.com/ALN/B461, which contains details on these surgeries and the proportion of surgeries related to fractures). Abdominal nonbowel surgery was more frequent in no-stroke patients (19%) compared with patients with previous stroke (8 to 12%). At baseline, patients with stroke less than 3 months before surgery were largely comparable with patients with stroke 3 to 9 months and more than 9 months before surgery (baseline characteristics are presented in table 1 for nonvascular surgery and in table 3, Supplemental Digital Content, http://links.lww.com/ALN/B461, for vascular surgery).
Long-term Interval between Stroke and Surgery in All Patients: Crude Events and Adjusted Models
The crude number of events for nonvascular surgeries is shown in table 2. Thirty-day MACEs occurred in 20.7% of patients with stroke less than 3 months before surgery and 8.8% of patients with stroke more than 9 months before surgery compared with only 2.3% of patients with no previous stroke. We observed low rates of myocardial infarctions in all of the patient subgroups (1% or less). Ischemic strokes were especially frequent in patients with stroke less than 3 months before surgery (9.9%) compared to patients with more distant stroke (2.3 to 2.8%). Cardiovascular death was the main contributor to the MACE endpoint for no-stroke patients (1.8%), as well as in patients with recent (9.9%) or more distant stroke (5.9%). All-cause mortality was significantly higher after surgery in stroke patients (11.7 to 16.4%) compared with no-stroke patients (4.8%). P value for the difference between the stroke groups was less than 0.001 for all of the endpoints.
Results from adjusted logistic regression models stratified by vascular and nonvascular surgery are presented in figure 1. Patients with stroke less than 3 months were at high risk of MACE for both nonvascular surgery (OR = 4.71 [95% CI, 4.18 to 5.32]) and vascular surgery (OR = 3.42 [95% CI, 2.02 to 5.78]) compared with no-stroke patients. Similar findings were seen for all-cause mortality. Patients with stroke more than 9 months before surgery undergoing nonvascular surgery remained at a significantly increased risk of MACE (OR = 1.62 [95% CI, 1.43 to 1.84]) and mortality (OR = 1.20 [95% CI, 1.08 to 1.34]) compared with no-stroke patients. We observed very high risks of repeat ischemic stroke in patients with stroke less than 3 months before surgery undergoing nonvascular (OR = 23.36 [95% CI, 19.24 to 28.37]) and vascular surgery (OR = 25.93 [95% CI, 12.55 to 53.55]).
Time between stroke and surgery was analyzed as a continuous variable for nonvascular surgery in spline analyses (fig. 2A–C). As time between stroke and surgery increased, risks of MACE, death, and ischemic stroke declined. Patients undergoing surgery approximately 5 months after their index stroke were no longer at a significantly increased risk of MACE (OR = 1.21 [95% CI, 0.98 to 1.49]) or death (OR = 1.20 [95% CI, 0.98 to 1.45]) compared with patients undergoing surgery approximately 12 months after initial stroke (reference; fig. 2A–C).
Long-term Interval between Stroke and Surgery in All Patients: Absolute Risk by Comorbidity, Age, and Sex
The absolute risks of MACE in all patients undergoing nonvascular surgery are presented for a total of 36 clinically relevant patient subgroups, stratified by comorbidities, sex, and age (fig. 3). Risks of MACE in any male stroke patient more than 70 yr of age (16.2%) were lower than those of men more than 70 yr of age with concomitant stroke and chronic obstructive pulmonary disease (23.7%), stroke and myocardial infarction (27.7%), and stroke and kidney disease (22.7%). On the contrary, the impact of comorbidities in addition to previous stroke was less pronounced in women less than 70 yr of age where the absolute risks of MACE included any stroke patient (9.1%), stroke and chronic obstructive pulmonary disease (7.3%), stroke and myocardial infarction (12.1%), and stroke and kidney disease (12.5%). The sensitivity analysis of absolute risks of MACE stratified by the revised cardiac risk index is presented in table 4, Supplemental Digital Content (http://links.lww.com/ALN/B461).
Immediate and Early Surgery: Baseline
A large proportion of patients with previous stroke underwent surgery within 14 days of index stroke (see fig. 2, A and B, Supplemental Digital Content, http://links.lww.com/ALN/B461, which displays time between previous stroke and surgery). Using propensity-score matching, we matched 323 patients undergoing early surgery (4 to 14 days after stroke) with the same number of patients undergoing immediate surgery (1 to 3 days after stroke). Baseline characteristics before and after matching are shown in table 5, Supplemental Digital Content (http://links.lww.com/ALN/B461). After successful matching, standardized mean differences were less than 0.10, indicating minimal imbalance between the groups.
Immediate and Early Surgery: Crude Events and Adjusted Models
In the propensity score–matched population, risk of 30-day MACEs were significantly lower after immediate surgery (69 events) compared with early surgery (93 events; P = 0.029). No difference was observed for 30-day all-cause mortality (P = 0.678; table 3). In spline analyses, time between stroke and surgery up to 14 days was assessed as a continuous variable (fig. 2A–C). The point estimates indicated that risks of MACE and death after surgery increased within the first 7 and 3 days, respectively, after a stroke; however, 95% CIs were wide and the day-by-day differences not statistically significant.
Sensitivity Analyses: All Patients and Orthopedic Surgery Patients
Risks of MACE associated with each stroke subgroup were estimated in analyses stratified by time of surgery, anesthesia type, and duration of anesthesia, including all patients undergoing nonvascular surgery (see table 6, Supplemental Digital Content [http://links.lww.com/ALN/B461], which display the sensitivity analysis for the full study cohort). Patients with stroke less than 3 months were at significantly higher risks of MACE with duration of anesthesia less than 120 min (OR = 6.69 [95% CI, 5.44 to 8.23]) compared with duration of anesthesia at 120 min or more (OR = 3.93 [95% CI, 3.39 to 4.56]), with an overall P for interaction of less than 0.001. Risks of MACE varied with time of surgery at regular and odd hours, although the estimates for patients with stroke less than 3 months were very similar (surgery performed in regular hours, OR = 5.51 [95% CI, 4.58 to 6.63], and odd hours, OR = 4.25 [95% CI, 3.62 to 4.99]; P for interaction = 0.012). Risks of MACE did not differ by type of anesthesia stratified as general anesthesia and other types (P for interaction = 0.175). We repeated our main analyses in the subgroup of patients undergoing minor or major orthopedic surgery (46% of the study cohort). Unadjusted absolute risks and ORs for patients with stroke less than 3 months were largely comparable with our main findings, including risks of MACE (OR = 4.25 [95% CI, 3.62 to 5.00]), all-cause mortality (OR = 1.54 [95% CI, 1.29 to 1.84]), and ischemic stroke (OR = 22.28 [95% CI, 17.43 to 28.49]). (See table 7, Supplemental Digital Content [http://links.lww.com/ALN/B461], which shows the main analysis only for the cohort of orthopedic surgery). In the subgroup of orthopedic surgery, we also performed analyses stratified by time of surgery, anesthesia type, and anesthesia duration (see table 8, Supplemental Digital Content [http://links.lww.com/ALN/B461], for variable characteristics and full results). Risks of MACEs did not differ by anesthesia type (P for interaction = 0.827) or regular versus odd hours of operation (P for interaction = 0.101). Risks of MACE were significantly higher in orthopedic surgery patients if duration of anesthesia was less than 120 min versus duration of anesthesia more than 120 min (P for interaction < 0.001). In the group of patients undergoing orthopedic surgery, 818 patients underwent surgery during the first 14 days after stroke. The majority of these early surgeries (635 surgeries) were on the hip or femoral bone, and, of these, 58% were fracture repairs and only 1% were due to infection of a prosthesis.
This nationwide study of patients with and without previous stroke undergoing emergency noncardiac, nonintracranial surgery demonstrated a time-dependent increased risk of 30-day MACE and all-cause mortality associated with previous stroke. Noticeably, among patients with stroke 14 days or less before surgery, we found that undergoing surgery 4 to 14 days after stroke was associated with significantly increased risks of a MACE compared with undergoing surgery within 1 to 3 days of stroke. In a long-term perspective, patients with a stroke less than 3 months before surgery were at elevated risk, which plateaued as time between stroke and surgery exceeded 4 to 5 months. Elderly patients with additional comorbidities, such as kidney disease or previous myocardial infarction in addition to stroke were at an especially high risk of a perioperative MACE. In addition, we found that the risk of a MACE was largely dependent on the duration of anesthesia, whereas the type of anesthesia and surgery at regular versus odd hours did not significantly affect the risk of adverse events.
The observed long-term time dependency between stroke and perioperative risks was similar to that observed in a previous study from our group examining previous stroke in patients having elective noncardiac and nonintracranial surgery.7 The associations in time between stroke and surgery and risks of perioperative adverse events have not previously been investigated in a setting of emergency noncardiac surgery, and only a few studies have previously addressed these issues in an elective surgery setting.22–25 In addition, this study is the first to look at perioperative risks in patients undergoing surgery 14 days or less after stroke.
The lack of guidance for patients with previous stroke undergoing surgery was addressed in a recent consensus statement from the Society for Neuroscience in Anesthesiology and Critical Care; they suggested that elective surgery should be delayed 1 to 3 months but did not include any guidance for emergency surgery.26,27 They concluded that the ultimate decision should always be based on the balance between risks of perioperative stroke and the risks of delaying surgery, which must be assessed thoroughly for each patient in need of emergency surgery. The observed increasing risk of adverse events within the first days after stroke and the long-term increased risk within the first 4 to 5 months may provide guidance for clinicians preparing patients with previous stroke for surgery.
Our data suggest that emergency surgery is associated with better outcomes if operations occur within 72 h of the stroke; thereafter, a higher risk period ensues that coincides with dysregulated cerebral autoregulation (see following paragraph). Hence, we hypothesize that anesthesia and surgery may constitute a larger second hit in the poststroke phase when conducted between 4 to 14 days after the initial insult. Nonetheless, the results of this study are important for guiding perioperative decision-making and to inform the patient and relatives about realistic expectations of perioperative outcomes. For example, patients with stroke less than 3 months had a 16% absolute risk of 30-day mortality and a 21% risk of MACE, which is substantial. In addition, estimating absolute risks in clinically relevant subgroups of patients, we found that patient risks were also highly dependent on other comorbidities, sex, and age, as outlined in figure 3. Informing clinicians about expected outcomes in these high-risk patients according to comorbidities and demographics may improve individual preoperative evaluation, rigorous monitoring, and patient care. Medical diseases such as atrial fibrillation and renal disease may be stabilized even in the setting of emergency surgery, and antithrombotic strategies should be carefully considered. Because this study showed a high incidence of repeat ischemic strokes (10% in patients with stroke less than 3 months), increased attention toward perioperative stroke screening seems relevant. The straightforward Face Arm Speech test for identifying neurologic deficits has shown promise, and the routine use of this test in the postoperative setting may be warranted.28
Surgery, Anesthesia, and Cerebral Autoregulation
Interestingly, in propensity score–matched analyses, we found that immediate surgery within 1 to 3 days of the index stroke was associated with significantly fewer MACEs than early surgery within 4 to 14 days. Similarly, findings from spline analyses suggested that risks of adverse events and death increased over the first 3 to 7 days, followed by a decline in risks. A systematic review of 23 studies on autoregulation of cerebral blood flow has provided a possible explanation for the increasing risk of adverse events within the first days after stroke and the continuous decrease in risks for the following several months.29 The review suggests that autoregulation deteriorates during the first 5 days after an ischemic stroke, followed by a recovery period of an estimated three months. These time intervals coincide with our periods of higher risk and hence support biologic plausibility. However, because we did not have information on important perioperative parameters, clinical variables, or the indication for surgery, we urge caution with this interpretation. Future prospective studies with more detailed perioperative data should seek to address these issues in detail.
In our study, subgroup analysis stratified by duration of anesthesia showed an unexpected increased risk of MACE with shorter duration of surgery. One explanation might be that longer anesthesia time is due to more thorough preparation and thus reduced perioperative risks. We estimated preparation time as the difference in anesthesia time and total procedure time, which showed that mean preparation time was in fact longer in patients with previous stroke compared with no-stroke patients (data not shown). We were, however, unable to investigate these associations in depth because surgery- and anesthesia-specific variables, beyond the type of anesthesia used, were not available in our registries.
We hypothesized that general anesthesia might exert more profound effects on the cardiovascular system and thus further aggravate the impaired autoregulation, increasing the risks of adverse outcomes. The analyses stratified by type of anesthesia showed only a nonsignificant association toward increased risks with general anesthesia (70 vs. 80% had general anesthesia for patients with and without stroke, respectively), whereas no difference was observed in the subgroup of patients undergoing orthopedic surgery. However, because we lack details of intraoperative care, we cannot be sure that the patients allocated regional anesthesia had nonanesthetic depth sedation30 and different hemodynamics compared with the general anesthetic group.
Strengths and Limitations
Our study included a contemporary unselected nationwide cohort of patients undergoing emergency noncardiac surgery. The Danish universal healthcare system ensures equal access to healthcare services irrespective of socioeconomic status. The comprehensive and validated administrative registers and the possibility for linking individual records using a unique personal identifier made possible the adjustment for important confounders, such as demographics, comorbidities, drug use, and surgery-related variables. Despite these efforts, residual confounding cannot be ruled out. The register-based method reduced the risk of report and selection bias, as well as other imprecisions related to data collection. We found that 19% of all previous stroke patients underwent emergency surgery within 14 days of the index stroke. This suggests that the stroke episode and the need for surgery are somehow correlated, but whether the ischemic stroke caused the emergency surgical condition (i.e., the stroke facilitated an injurious fall resulting in a hip fracture) or whether a surgical condition caused a subsequent stroke (i.e., a severe trauma resulting in disseminated intravascular coagulation causing an ischemic stroke) cannot be made out from the data available. In addition, information on in-hospital distributed medications and perioperative care was not available, and changes in medication during hospital admission, including preoperative antithrombotic and anticoagulation therapy, could not be accounted for in our study. No information on periprocedural blood loss or blood transfusions was available. It was not possible to distinguish between thromboembolic and atherothrombotic strokes, which are known to be different in pathophysiology. We have shown previously that stroke patients not undergoing surgery experience a similar reduction in risks of repeat stroke over time but that these risks and reductions are much more pronounced in patients undergoing surgery, which suggests that surgery constitutes a significant risk factor for repeat stroke.16 Previous literature from our Danish registers suggests an underestimation of myocardial infarction due to the lack of a systematic assessment of the myocardial infarction by troponins after surgery.31 The use of continuous time variables, such as anesthesia time, may be problematic and dependent on the specific surgery.32
In this study, we demonstrated that patients with previous stroke are at high risk of perioperative death and MACE for several months after emergency noncardiac and nonintracranial surgery. Interestingly, for the first 3 to 7 days after stroke, risks seemed to increase as cerebral autoregulation may have deteriorated, suggesting that immediate surgery may be associated with lower risks. We observed that patients were at particularly high risk within the first 3 months of an ischemic stroke, whereas the rapid decrease in risks of adverse events leveled off after approximately 4 to 5 months. Stroke-associated risks were dependent on patient sex, age, and additional comorbidities. Although emergency surgeries might not be feasible to postpone until the risk period subsides, high risk of adverse events and death should be taken into account during the perioperative risk assessment and conveyed when informing the patient and relatives about realistic expectations.
Support was provided solely from institutional and/or departmental sources.
The authors declare no competing interests.