Association of Timing of Aortic Valve Replacement Surgery After Stroke With Risk of Recurrent Stroke and Mortality

Key Points

Question  Is there a time-dependent association of previous stroke and surgical aortic valve replacement with the risk of adverse outcomes among patients with aortic valve stenosis?

Findings  This cohort study using Danish administrative registry data included 14 030 surgical aortic valve replacements. Patients with prior stroke had 14.7-, 4.0-, and 2.3-fold higher risk of recurrent stroke compared with patients without prior stroke if surgery was performed within 3 months, 3 to less than 12 months, and 12 months or more, respectively, after the incident stroke.

Meaning  Surgical aortic valve replacement within the first 3 months after a stroke may be associated with a particularly pronounced risk of recurrent ischemic stroke.

Importance  Timing of surgical aortic valve replacement (SAVR) in patients with aortic valve stenosis and previous stroke for the risk of recurrent stroke is insufficiently investigated.

Objective  To evaluate the association of time elapsed between previous stroke and SAVR with the risk of recurrent perioperative stroke, major adverse cardiovascular events (MACE), and mortality among patients with aortic valve stenosis.

Design, Setting, and Participants  This cohort study using data from Danish administrative registries included all patients with aortic valve stenosis older than 18 years who underwent SAVR between 1996 and 2014 (n = 14 030). Patients who received simultaneous mitral, tricuspid, or pulmonary valve surgery and patients with endocarditis 1 year prior to surgery were excluded. Data were analyzed from March 2017 to January 2018.

Exposures  Time elapsed between prior stroke and SAVR (<3 months, 3-<12 months, ≥12 months, and no prior stroke).

Main Outcomes and Measures  Thirty-day risks of MACE, ischemic stroke, and all-cause mortality reported as absolute events and multivariable adjusted odds ratios with 95% confidence intervals. Restricted cubic spline regression models were additionally applied on the subgroup with prior stroke.

Results  Of the 14 030 included patients, 616 patients (190 [30.8%] women; mean [SD] age, 72.0 [9.1] years) with prior stroke underwent surgery, and 13 414 (4837 [36.1%] women; mean [SD] age, 69.8 [10.8] years) without prior stroke underwent surgery. The absolute risk of ischemic stroke was significantly increased in patients with stroke less than 3 months prior to surgery compared with patients with no prior stroke (18.4% [37 of 201] vs 1.2% [160 of 13 219]; odds ratio, 14.69; 95% CI, 9.69-22.27). Likewise, compared with patients without stroke, patients with stroke less than 3 months prior surgery were at significantly increased risk of MACE (23.3% [53 of 227] vs 5.7% [768 of 13 414]; odds ratio, 4.57; 95% CI, 3.24-6.44) but not all-cause mortality (6.8% [50 of 730] vs 3.6% [374 of 10 370]; odds ratio, 1.45; 95% CI, 0.83-2.54). Spline analyses supported a declining risk over time, reaching nadir after 2 to 4 months.

Conclusions and Relevance  Previous stroke is a major risk factor of recurrent ischemic stroke and MACE in patients undergoing SAVR, especially if time elapsed between previous stroke and surgery is less than 3 months.

Progression of aortic valve stenosis is medically unpreventable. Likewise, the mortality rate among severely symptomatic patients is markedly increased unless aortic valve replacement is performed. Ischemic stroke is a feared complication of surgical aortic valve replacement (SAVR),^1-3 and prior stroke is a well-established risk factor of periprocedural stroke.^4-6 For noncardiac surgery, the risk of recurrent stroke was shown in 2014 to be time dependent⁷ and appears to be particularly pronounced if time between stroke and surgery is less than 9 months. Cerebral autoregulation is impaired immediately after a stroke, making penumbral tissue vulnerable to alterations in blood pressure. Thus, perioperative manipulation of blood pressure and interruption of antithrombotic medication may pose a particular risk in patients with a recent stroke.^8,9 As the prevalence of aortic valve stenosis increases with age^10-12 and predictions forecast a substantial increase in the proportion of elderly individuals (older than 65 years) in the Western world by 2080,^13,14 the socioeconomic burden of aortic valve stenosis and SAVR are expected to grow notably. Recurrent stroke is associated with more severe functional disability, greater financial burden, and increased mortality compared with an incident stroke.^15,16 Consequently, it is of particular relevance to investigate the association of time elapsed between previous stroke and SAVR with the risk of adverse outcomes in patients with aortic valve stenosis.

Every individual living in Denmark receives a personal identification number either at birth or immigration, permitting cross-linkage of several health care–related registries at an individual level. Data were collected from 5 nationwide Danish registries. Register-based studies in which individuals cannot be identified do not require ethical approval in Denmark, and informed consent was not needed because deidentified, administrative databases were used.

The Central Population Registry contains information on sex, date of birth, and vital status. The Danish National Patient Registry contains information on all in-hospital admissions since 1977; diagnoses are coded according to the International Classification of Disease, Eight Revision (ICD-8) until 1993 and the International Statistical Classification of Diseases and Related Health Problems, Tenth Revision (ICD-10) from 1994 onwards.¹⁷ The Danish National Prescription Registry contains information on all dispensed medical prescriptions since 1995, and prescribed drugs have been registered by the Anatomical Therapeutic Chemical Classification System. Cause of death was obtained from the Danish Causes of Death Registry (since 1970), which contains information on primary and contributing causes of death. The Danish Heart Registry contains information on European System for Cardiac Operative Risk Evaluation (EuroSCORE), which is a commonly used risk stratification score for prediction of 30-day mortality,¹⁸ body mass index, smoking status, and surgery priority for cardiac surgeries since 2003.

We included patients with aortic valve stenosis older than 18 years who underwent SAVR from January 1996 to December 2014. Aortic valve stenosis was defined as a discharge diagnosis any time prior surgery. Coding details for all diagnoses and procedures are available in eTable 1 in the Supplement.

Surgical aortic valve replacement was identified using KFM codes (Nordic Medico-Statistical Committee Classification of Surgical Procedures). Patients who received simultaneous mitral, tricuspid, or pulmonary valve surgery were excluded. Patients with an endocarditis diagnosis within 1 year prior to surgery were also excluded because endocarditis is associated with increased risk of stroke 5 to 6 months after initiation of antibiotic therapy.¹⁹ However, we chose a more conservative cutoff value of 1 year prior to surgery to ensure exclusion of any strokes caused by endocarditis.

Ischemic stroke was identified using both ICD-8 and ICD-10 diagnostic codes. Ischemic stroke (ICD code I63) has been validated to have a positive predictive value between 88% and 97%. Most strokes classified as unspecified strokes (ICD code I64) have been shown to be of ischemic origin (positive predictive value between 72% and 80%).^20,21 The stroke diagnosis was considered outdated if more than 5 years had passed between stroke and surgery.⁷ Validation of the ICD-10 diagnostic coding system in the Danish National Patient Registry has been performed for several cardiovascular conditions, including aortic valve stenosis, myocardial infarction, congestive heart failure, peripheral artery disease, and cerebrovascular disease, with positive predictive values between 82% and 100%.^17,22,23

Comorbidities were identified using both ICD-8 and ICD-10 codes for acute myocardial infarction, ischemic heart disease, chronic heart failure, atrial fibrillation, peripheral artery disease (composite of atherosclerosis in the peripheral arteries or aorta, arterial embolism or thrombosis, or other peripheral vascular disease), carotid stenosis, chronic obstructive pulmonary disease, cancer, renal disease, rheumatic disease, and liver disease. Diabetes was defined as at least 1 claimed prescription of glucose-lowering drugs.⁷ Use of cardiovascular medication within 120 days prior surgery was obtained using Anatomical Therapeutic Chemical codes (eTable 1 in the Supplement).

We divided our study population into 4 groups: individuals with no prior stroke, those with stroke less than 3 months prior to surgery, those with stroke within 3 to less than 12 months prior to surgery, and those with stroke 12 months or more prior to surgery. Subanalyses were performed for patients with and without atrial fibrillation (preoperative atrial fibrillation), with and without carotid stenosis at baseline, with available EuroSCORE only, and undergoing transcatheter aortic valve implantation (TAVI), identified using KFM codes.

We investigated the 30-day risks of major adverse cardiovascular events (MACE; composite of nonfatal myocardial infarction, nonfatal ischemic stroke, or death from cardiovascular causes [ICD codes I00-I99]), ischemic stroke (ICD codes I63 and I64), and all-cause mortality. Secondary outcomes included nonfatal myocardial infarction and death from cardiovascular causes.

Baseline characteristics, crude event rates, and adjusted odds ratios (ORs) were reported for each of the 4 a priori selected stroke groups. Only the first surgery for each individual over the study period was included. Using patients with no prior stroke as a reference, we estimated ORs with 95% CIs by multivariable logistic regression models adjusted for the following variables: age, sex, concomitant coronary artery bypass grafting (CABG) surgery, renal disease, chronic obstructive lung disease, diabetes, previous myocardial infarction, peripheral artery disease, atrial fibrillation, chronic heart failure, calendar year, and preoperative antithrombotic therapy (eg, aspirin, clopidogrel, dipyridamole, and vitamin K antagonists). Because of insufficient power, sensitivity analyses were adjusted for fewer variables, as specified in the Results section. The proportions of variance in outcomes explained by the logistic regression models were computed using Tjur R-squared²⁴ and are presented in eTable 2 in the Supplement.

Tests for interactions were chosen a priori based on clinical relevance and included atrial fibrillation, concomitant CABG surgery, preoperative antithrombotic therapy, age, and sex. Time between stroke and surgery was evaluated by restricted cubic spline regression models to determine continuous changes in risk over time among the subgroup with prior stroke.²⁵ Five knots were placed at the p5, p25, p50, p75, and p95 percentiles. The median value was used as reference.

All analyses were performed using SAS version 9.4 (SAS Institute), and RStudio version 1.1.383 (RStudio Inc) was used for illustrations. Two-sided P values less than .05 were considered statistically significant for all tests.

We identified 14 030 patients who underwent SAVR for the first time during the study period (eFigure in the Supplement). Of these, 616 patients (190 [30.8%] women; mean [SD] age, 72.0 [9.1] years) with prior stroke underwent surgery, and 13 414 patients (4837 [36.1%] women; mean [SD] age, 69.8 [10.8] years) without prior stroke underwent surgery. Baseline characteristics for the study population stratified by stroke history and time from stroke to surgery are presented in Table 1.

Most patients in each stroke group were men (range, 63.9%-69.8%) and received bioprosthetic valves (range, 68.8%-81.3%). Concomitant CABG surgery was performed in 35.3% to 45.3% of patients in the stroke groups. Patients with prior stroke had a higher prevalence of preoperative atrial fibrillation, ischemic heart disease, peripheral artery disease, and carotid stenosis and more commonly used antithrombotic medication, statins, and renin-angiotensin system inhibitors compared with patients with no prior stroke. Furthermore, diabetes and insulin treatment was more common in patients with stroke 12 months or more prior to surgery compared with patients without stroke.

Sample sizes, crude events, and adjusted ORs for 30-day risks of MACE, ischemic stroke, and all-cause mortality are presented in Figure 1. The absolute percentage of MACE was 5.7% (768 of 13 414) in patients without prior stroke compared with 23.3% (53 of 227) in patients with stroke less than 3 months prior to surgery. In adjusted analyses, stroke less than 3 months prior to surgery remained associated with markedly increased risk of MACE (OR, 4.57; 95% CI, 3.24-6.44) and ischemic stroke (OR, 14.69; 95% CI, 9.69-22.27) compared with patients with no prior stroke. All-cause mortality did not differ significantly between the groups after adjustment for various confounders (Figure 1).

Risk estimates of the secondary outcomes, ie, perioperative myocardial infarction and cardiovascular death, are available in eTable 3 in the Supplement. Event rates were low for these outcomes, and adjusted analyses did not show any differences in risk between the groups.

The association of MACE, recurrent stroke, and all-cause mortality with time since prior stroke and surgery among the subgroup with prior stroke is illustrated in Figure 2 (splines). For all outcomes, the ORs declined with time and appeared stable after approximately 2 to 4 months.

Stroke any time prior to surgery was associated with a higher risk of 30-day MACE among patients with atrial fibrillation (OR, 5.26; 95% CI, 3.49-7.92) compared with patients without atrial fibrillation (OR, 1.81; 95% CI, 1.27-2.59) (P < .001). Prior stroke was also associated with a higher risk of 30-day MACE in patients younger than 75 years (OR, 3.19; 95% CI, 2.24-4.54) compared with patients older than 75 years (OR, 2.22; 95% CI, 1.48-3.32) (P = .03). No statistically significant difference was found in the association between prior stroke and our outcomes for the remaining variables (including similar risks observed for patients with and without concomitant CABG) (eTable 4 in the Supplement).

We identified 3092 patients (22.0%) with and 10 938 (78.0%) without preoperative atrial fibrillation at baseline. Of the 10 938 patients without known preoperative atrial fibrillation, 1025 patients (9.4%) developed postoperative atrial fibrillation (defined as a diagnosis of atrial fibrillation within 30 days after surgery) and were equally distributed among the groups (eTable 5 in the Supplement). Crude events, sample size, and adjusted ORs (adjusted for age, sex, and antithrombotic therapy) for patients with and without preoperative atrial fibrillation are presented in eTables 6 and 7, respectively, in the Supplement.

We identified 257 patients with and 13 773 patients without preoperative carotid stenosis (eTables 8 and 9 in the Supplement). For patients without known preoperative carotid stenosis and stroke less than 3 months prior to surgery, the absolute rate of recurrent ischemic stroke was 18.4% (37 of 201) compared with 1.2% (160 of 13 219) in patients with no stroke (eTable 9 in the Supplement).

In the Danish Heart Registry, information on EuroSCORE was available for 11 100 patients who underwent nonacute SAVR. Mean EuroSCORE did not differ markedly among the 4 groups (eTable 10 in the Supplement), and most patients were at intermediate risk of 30-day mortality. The absolute rate of ischemic stroke for patients with stroke less than 3 months prior to surgery was 14.8% (48 of 325) compared with 1.2% (121 of 10 370) for patients with no prior stroke. Full results (crude events and total sample size as well as adjusted ORs for sex, age, and EuroSCORE) stratified by the 4 stroke groups for 30-day risks of MACE, ischemic stroke, and all-cause mortality are shown in Table 2. Additional adjustment for body mass index and smoking status yielded similar results (eTable 11 in the Supplement).

A total of 1508 patients with aortic valve stenosis underwent TAVI surgery. Prior stroke was identified in 132 patients. Crude events and sample size as well as adjusted ORs for sex and age of 30-day MACE, ischemic stroke, and all-cause mortality are available in eTable 12 in the Supplement. In line with the main results, among patients undergoing TAVI, the absolute rates of 30-day ischemic stroke and MACE were 26.5% (9 of 34) and 34.2% (12 of 34), respectively, in patients with stroke less than 3 months prior to surgery compared with 1.1% (15 of 1376) and 4.5% (62 of 1376) in patients with no prior stroke.

This nationwide study examined the 30-day risk of adverse outcomes associated with SAVR in patients with and without prior stroke, with particular focus on the importance of time elapsed between stroke and surgery. Patients with recent stroke (within 3 months) had markedly heightened risk of recurrent stroke during surgery, and the risk declined with time. We found no association of time from stroke to surgery with the risk of all-cause mortality. The splines of the subgroup with prior stroke supports that the risk of recurrent stroke declines with time and reaches a nadir after approximately 4 months. Thus, data suggest that postponement of SAVR for at least 3 to 4 months after a stroke, if possible, may reduce the risk of recurrent stroke during surgery. However, because of the observational nature of this study, the results are only hypothesis generating, and more studies are encouraged before drawing such a conclusion.

Consistent with previously reported rates, we found an overall perioperative stroke rate of 2.0% for SAVR with concomitant CABG and 1.3% for SAVR without concomitant CABG.^18,26-28 For both surgical and nonsurgical patients, the risk of recurrent stroke has been reported to be particularly pronounced during the first months after an incident stroke.^29-31 Importantly, the event rate of recurrent stroke that our sample would have experienced if they did not undergo surgery is unknown. However, our estimates were substantially higher than those reported in prior cohort studies of nonsurgical patients (approximately 3% at 30 days),³² and our regression models were adjusted for key clinical factors that predict the risk of recurrent stroke, supporting the assumption that surgery may heighten the risk of recurrent stroke.¹⁶ Similarly, in a noncardiac surgery setting, we observed in 2014 that the relative risk of recurrent stroke was strikingly higher in patients undergoing surgery shortly after a recent stroke compared with patients who did not undergo surgery shortly after a stroke.^7,33 Nevertheless, for the present study, we were unable to find a suitable comparison group because the groups would have been inherently different—the controls would not have an indication for surgery. Therefore, ideally, a randomized clinical trial would be needed to show whether postponement of surgery reduces the risk of recurrent stroke and improves neurological outcomes.

Numerous studies have indicated that previous stroke is an independent predictor for postoperative stroke and mortality.^6,34-40 In 2008, Gulbins et al⁴¹ presented a hazard ratio of 2.6 for postoperative stroke among patients with a history of stroke receiving a stentless aortic valve; however, the follow-up began 6 months after surgery. In 2013, Bottle et al³⁹ performed a study on preoperative stroke and outcomes after CABG surgery. Similar to our study, they observed that preoperative stroke increased the risk of postoperative recurrent stroke. However, in contrast to our study, they found no evidence of any vulnerable period in the first 3 to 6 months after a stroke. Notably, the number of patients with recent stroke was low, with only 28 patients with stroke less than 3 months prior to surgery (<0.1%) and 64 patients with stroke less than 6 months prior to surgery (0.1%). Somewhat surprisingly, they found a small increase in risk of postoperative stroke with longer duration between surgery and preoperative stroke; the reason for discrepancy of our study results and those by Bottle et al³⁹ is unknown.

Although atrial fibrillation (preoperative or postoperative) has been associated with worse postoperative outcomes, including increased risks of ischemic stroke, a few factors suggest that it may not drive the observed increased risks of recurrent stroke in our cohort of patients with prior stroke.^42-44 First, the association of recent stroke (less than 3 months) with increased risk of ischemic stroke was apparent both in patients with and without atrial fibrillation. Second, postoperative atrial fibrillation was not more commonly observed in patients with recent stroke. Potential mechanisms underlying our observations remain unknown and warrant more studies.

The main strength of our study was the unselected cohort of 14 030 patients undergoing aortic valve replacement surgery. However, the number of patients with prior stroke within 5 years prior to surgery was rather small, potentially caused by pre-exclusion because of, for example, a high EuroSCORE. The small sample size may have influenced the estimates to become more uncertain, and it cannot be excluded that some of the interaction analyses had insufficient power to reveal a true difference (ie, a statistical type 2 error). Moreover, the number of patients with prior stroke undergoing TAVI (n = 132) was also small, hindering any conclusions to be drawn on this population.

Stratification of the patients in the stroke groups revealed that surgeries were more frequently performed less than 3 months or 12 months or more after a stroke. Although administrative data may not enable us to provide full granularity of our data and, thus, we cannot exclude that the patients undergoing surgery less than 3 months after a stroke might constitute a specific group of critically ill patients where postponement may not be an option (and thus, the indication rather than the history of stroke may heighten the risk of recurrent stroke), the stroke groups were fairly similar with regards to comorbidity burden. Moreover, mean EuroSCORE values were not markedly different among the stroke groups, and exclusion of all acute surgery did not alter the results.

The Danish population is predominantly white, which may limit generalizability of our findings to other races/ethnicities. Although we adjusted our analyses for several clinical variables, we lacked information on symptoms and severity of aortic valve stenosis, neurological status, general frailty, and circulating biomarkers, including creatinine values, troponins concentrations, and natriuretic peptide levels. Moreover, we lacked data on the perioperative handling of antiplatelet and anticoagulation therapy (continuation vs interruption), which may have influenced the outcomes. Finally, adjustment for carotid stenosis was not performed because of unknown validity of this diagnosis in our registries. However, excluding patients with preoperative carotid stenosis did not alter the results.

Previous stroke is a major risk factor for recurrent ischemic stroke and MACE in patients undergoing SAVR, especially if time elapsed since previous stroke and surgery is less than 3 months.

Accepted for Publication: February 14, 2018.

Corresponding Author: Charlotte Andreasen, MD, Department of Cardiology, Copenhagen University Hospital Herlev and Gentofte, Kildegaardsvej 28, 2900 Hellerup, Denmark (andreasencharlotte7@gmail.com).

Published Online: April 25, 2018. doi:10.1001/jamacardio.2018.0899

Author Contributions: Drs Andreasen and Andersson had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: Andreasen, Gislason, Sanders, Køber, Andersson.

Acquisition, analysis, or interpretation of data: Andreasen, Jørgensen, Gislason, Martinsson, Sanders, Abdulla, Jensen, Torp-Pedersen, Andersson.

Drafting of the manuscript: Andreasen.

Critical revision of the manuscript for important intellectual content: All authors.

Statistical analysis: Andreasen, Jørgensen, Gislason, Andersson.

Obtained funding: Andreasen, Gislason, Køber, Andersson.

Administrative, technical, or material support: Gislason, Torp-Pedersen, Køber.

Study supervision: Jørgensen, Gislason, Martinsson, Sanders, Abdulla, Jensen, Torp-Pedersen, Køber, Andersson.

Conflict of Interest Disclosures: All authors have completed and submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Dr Gislason has received grants from AstraZeneca, Bayer, Boehringer Ingelheim, Bristol-Myers Squibb, and Pfizer. Dr Torp-Pedersen has received grants from Bayer during the conduct of the study. Dr Køber has received personal fees from Novartis as a speaker. No other disclosures were reported.

Funding/Support: Dr Andreasen was supported by an unrestricted grant by the Danish Heart Foundation and Gerda & Hans Hansens Fund.

Role of the Funder/Sponsor: The funders had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.