Association Between Left Atrial Appendage Occlusion and Readmission for Thromboembolism Among Patients With Atrial Fibrillation Undergoing Concomitant Cardiac Surgery

Key Points

Question  Is surgical left atrial appendage occlusion associated with a reduction in long-term thromboembolic events?

Findings  In this retrospective cohort study of 10 524 Medicare recipients with atrial fibrillation undergoing cardiac surgery, surgical left atrial appendage occlusion, compared with no surgical left atrial appendage occlusion, was significantly associated with lower risk of readmission for thromboembolism at 3 years (unadjusted, 4.2% vs 6.2%; adjusted hazard ratio, 0.67).

Meaning  Surgical left atrial appendage occlusion may be of benefit in preventing thromboembolic events in older patients with atrial fibrillation undergoing cardiac surgery, although randomized trials would be necessary to provide definitive evidence.

Importance  The left atrial appendage is a key site of thrombus formation in atrial fibrillation (AF) and can be occluded or removed at the time of cardiac surgery. There is limited evidence regarding the effectiveness of surgical left atrial appendage occlusion (S-LAAO) for reducing the risk of thromboembolism.

Objective  To evaluate the association of S-LAAO vs no receipt of S-LAAO with the risk of thromboembolism among older patients undergoing cardiac surgery.

Design, Setting, and Participants  Retrospective cohort study of a nationally representative Medicare-linked cohort from the Society of Thoracic Surgeons Adult Cardiac Surgery Database (2011-2012). Patients aged 65 years and older with AF undergoing cardiac surgery (coronary artery bypass grafting [CABG], mitral valve surgery with or without CABG, or aortic valve surgery with or without CABG) with and without concomitant S-LAAO were followed up until December 31, 2014.

Exposures  S-LAAO vs no S-LAAO.

Main Outcomes and Measures  The primary outcome was readmission for thromboembolism (stroke, transient ischemic attack, or systemic embolism) at up to 3 years of follow-up, as defined by Medicare claims data. Secondary end points included hemorrhagic stroke, all-cause mortality, and a composite end point (thromboembolism, hemorrhagic stroke, or all-cause mortality).

Results  Among 10 524 patients undergoing surgery (median age, 76 years; 39% female; median CHA₂DS₂-VASc score, 4), 3892 (37%) underwent S-LAAO. Overall, at a mean follow-up of 2.6 years, thromboembolism occurred in 5.4%, hemorrhagic stroke in 0.9%, all-cause mortality in 21.5%, and the composite end point in 25.7%. S-LAAO, compared with no S-LAAO, was associated with lower unadjusted rates of thromboembolism (4.2% vs 6.2%), all-cause mortality (17.3% vs 23.9%), and the composite end point (20.5% vs 28.7%) but no significant difference in rates of hemorrhagic stroke (0.9% vs 0.9%). After inverse probability–weighted adjustment, S-LAAO was associated with a significantly lower rate of thromboembolism (subdistribution hazard ratio [HR], 0.67; 95% CI, 0.56-0.81; P < .001), all-cause mortality (HR, 0.88; 95% CI, 0.79-0.97; P = .001), and the composite end point (HR, 0.83; 95% CI, 0.76-0.91; P < .001) but not hemorrhagic stroke (subdistribution HR, 0.84; 95% CI, 0.53-1.32; P = .44). S-LAAO, compared with no S-LAAO, was associated with a lower risk of thromboembolism among patients discharged without anticoagulation (unadjusted rate, 4.2% vs 6.0%; adjusted subdistribution HR, 0.26; 95% CI, 0.17-0.40; P < .001), but not among patients discharged with anticoagulation (unadjusted rate, 4.1% vs 6.3%; adjusted subdistribution HR, 0.88; 95% CI, 0.56-1.39; P = .59).

Conclusions and Relevance  Among older patients with AF undergoing concomitant cardiac surgery, S-LAAO, compared with no S-LAAO, was associated with a lower risk of readmission for thromboembolism over 3 years. These findings support the use of S-LAAO, but randomized trials are necessary to provide definitive evidence.

Atrial fibrillation (AF) is the most common sustained arrhythmia and has been projected to affect 3.3 million US adults by 2020.¹ AF is associated with an increased risk of thromboembolic stroke.² In the setting of nonrheumatic AF, approximately 90% of strokes originate from the left atrial appendage (LAA),³ which exhibits poor contractile function during AF, permitting blood stasis and thrombus formation. Although oral anticoagulation^4,5 is effective at reducing the risk of thromboembolic stroke, as few as half of all eligible patients use anticoagulation,⁶ frequently citing high perceived hemorrhage risk, cost, and patient preference. Quiz Ref IDThe low rates of anticoagulant use and the understanding that AF-related thrombus formation is most likely to occur in the LAA has led to increasing interest in occluding the LAA as a potential alternative to anticoagulation, particularly among those with difficulty tolerating anticoagulation.

Quiz Ref IDLAA occlusion can be performed surgically (S-LAAO) at the time of cardiac surgery, yet use of S-LAAO varies widely among physicians. Although in aggregate, data⁷ from 2 randomized trials^8,9 demonstrated percutaneous LAAO was noninferior to warfarin, no similar data exist for S-LAAO. Furthermore, there have been reports of increased thromboembolism risk due to incomplete S-LAAO with persistent LAA to left atrium communication.¹⁰ These limited data on the effectiveness of S-LAAO have led to its class IIb recommendation in US¹¹ and European¹² guidelines. As a result, information on the safety and effectiveness of S-LAAO is needed.

The Society of Thoracic Surgeons (STS) Adult Cardiac Surgery Database (ACSD) was used to perform an analysis of S-LAAO vs no S-LAAO in a contemporary, nationally representative cohort of Medicare beneficiaries with AF who underwent cardiac surgery. The main study objective was to evaluate whether S-LAAO was associated with a lower risk of readmission for thromboembolism in a broad population including patients who were discharged with and without anticoagulation.

The STS ACSD is a national US registry that collects detailed in-hospital data on all adults undergoing cardiac surgery at more than 1000 participating institutions (approximately 90% of cardiac surgical programs in the United States).¹³ Through the use of a validated deterministic linkage with fee-for-service Medicare claims data and the Medicare Denominator File,¹⁴ longitudinal data on morbidity and mortality can be determined for patients aged 65 years and older with fee-for-service Medicare insurance coverage. Documentation of concomitant S-LAAO began on January 14, 2011 (data collection form version 2.73). This project was approved by the STS Task Force on Longitudinal Follow-up and Linked Registries Committee and the Duke University institutional review board. The Duke University institutional review board waived the need for informed consent given the nature of this study.

This retrospective cohort study included older patients (age ≥65 years) with a history of AF or atrial flutter undergoing first-time cardiac surgery (coronary artery bypass grafting [CABG], mitral valve surgery ± CABG, or aortic valve surgery ± CABG). We excluded patients with planned off-pump operations, endocarditis, double valve procedures (both aortic and mitral operations), congenital heart disease or cardiac transplant, left ventricular assist device, cardiogenic shock, missing data on S-LAAO, inability to link to Medicare claims, missing anticoagulation data, those without information on the primary surgical procedure, and those with a duplicate Medicare record number. Paroxysmal AF was defined by the STS ACSD as AF that terminates within 7 days of initiation. All other AF subtypes were combined for the purpose of this analysis and labeled nonparoxysmal AF because the data collection form does not differentiate between persistent and permanent AF. The STS Registry collects data on the use of oral anticoagulation within 24 hours of surgery and oral anticoagulation prescription at discharge. Because oral anticoagulation is typically discontinued more than 24 hours prior to surgery, the preoperative anticoagulation variable likely reflects patients who underwent an urgent procedure (and did not have anticoagulation held) or mistakenly received oral anticoagulation. The discharge anticoagulation variable reflects patients who were discharged with a prescription for oral anticoagulation, but it does not capture information regarding adherence or duration of therapy.

Geographic regions were defined based on the STS ACSD convention.¹⁵ Baseline patient characteristics were defined by the STS ACSD data specifications.¹⁶ Race and ethnicity were defined by patient (or family) report using fixed categories and were included in this analysis because race and ethnicity have been associated with perioperative risk and could be associated with use of S-LAAO. The glomerular filtration rate was calculated based on the Modification of Diet in Renal Disease equation.¹⁷

The treatment of interest was S-LAAO (by any method) vs no S-LAAO, as defined by the STS Registry.

Quiz Ref IDThe primary outcome was rehospitalization (ie, after the index hospitalization) for thromboembolism (International Classification of Diseases, Ninth Revision [ICD-9] codes 434.x or 444.x [thromboembolic stroke or systemic embolism] or 435.x [transient ischemic attack])¹⁸ to 3 years. Secondary outcomes included hemorrhagic stroke (ICD-9 codes 430-432),¹⁸ all-cause mortality, and a composite end point comprised of thromboembolism, hemorrhagic stroke, or all-cause mortality to 3 years. Vital status was determined by the Medicare denominator file. Follow-up was censored on December 31, 2014, or on the date at which the patient’s data were no longer available (owing to death or transition to a managed care plan).

We also considered falsification end points,¹⁹ which are outcomes that would not be expected to be affected by S-LAAO but could differ if treatment selection bias remained after adjustment. These prospectively designed falsification end points included pneumonia (ICD-9 codes 480.x, 481, 482.xx, 483.xx, 485, 486, 487.0, and 488.11)²⁰ and lower extremity fracture (ICD-9 codes 820.00-820.9, 821.00-821.11, 821.20-821.39, 808.0-808.9, 823.02-823.92, 827.0-827.1, and 828.0-828.1).²¹

Baseline characteristics of the study population by treatment group (S-LAAO vs no S-LAAO) were described using proportions for categorical variables and means with SDs for continuous variables. Differences between groups were tested using the χ² test for categorical variables and t test or Wilcoxon test for continuous variables, depending on the nature of the distribution.

We compared observed event rates by treatment group. For the all-cause mortality and composite end points, we used the Kaplan-Meier methods to calculate event rates. For thromboembolism, hemorrhagic stroke, and the falsification end points, we used the cumulative incidence function to calculate event rates. Unadjusted Cox proportional hazards models and Fine-Gray models were used to test between-group differences. The Markov Chain Monte Carlo method was used for imputation. Model results represent a pooled estimate generated by combing the results obtained by running each model in 10 independent imputed data sets.

To estimate the risk-adjusted association between S-LAAO vs no S-LAAO and outcomes, we used inverse probability–weighted (IPW) Fine-Gray or Cox proportional hazard models. Fine-Gray models were used for nonfatal study end points (thromboembolism, hemorrhagic stroke, pneumonia, and lower extremity fracture) to account for the competing risk of death, which is high in this population. Cox proportional hazards models were used for all-cause mortality and the composite end point, which included death. We used nonparsimonious logistic regression models to predict the use of S-LAAO to generate a propensity score and, subsequently, an inverse propensity weight. Adjusted models were derived using IPW and did not additionally include individual variables that were used to generate the propensity score.

For the primary analysis, we used the following variables to derive the propensity score: age; sex; race/ethnicity; AF subtype; current smoking; body mass index; ejection fraction; congestive heart failure; prior stroke; hypertension; hyperlipidemia; diabetes; coronary artery disease; acute coronary syndrome prior to operation; glomerular filtration rate; lung disease; obstructive sleep apnea; congestive heart failure, hypertension, age 75 years or older (doubled), diabetes, stroke, transient ischemic attack, or thromboembolism (doubled), vascular disease, age 65 to 74 years, sex category (female) (CHA₂DS₂-VASc score); STS Risk Score; warfarin use within the 24 hours prior to surgery; academic hospital status; CABG volume; valve surgery volume; geographic region; operation type; mechanical valve; operation status; nonsternotomy surgical approach; number of diseased vessels; and presence of left main coronary artery disease. For secondary analyses, all variables (including surgical ablation) were included in the logistic regression model to derive a propensity score. The IPW cohorts were stabilized to prevent individuals with extreme weights from having excess influence.²² The differences between observed characteristics in the weighted S-LAAO and no S-LAAO groups were examined by calculating the Cramer φ statistic. Small differences (<10%) suggest balance across the observed baseline covariates among patients in the S-LAAO and no S-LAAO groups.

The primary analysis assessed the association between S-LAAO vs no S-LAAO and outcomes without adjustment for discharge anticoagulation strategy. We performed a series of prespecified secondary analyses to explore whether concomitant surgical ablation of AF or discharge anticoagulation strategies may affect the association between S-LAAO and risk of thromboembolism and all-cause mortality. We assessed the interaction between S-LAAO, discharge anticoagulation, and the study outcomes in the primary IPW models. Owing to the theoretical (but unproven) association between surgical AF ablation and stroke, we included concomitant surgical AF ablation in these secondary IPW models. Balance across the IPW populations for those discharged with and without oral anticoagulation was assessed by the Cramer φ statistics.

We performed a series of sensitivity analyses in which the primary and secondary analyses were repeated using fully adjusted regression models to ensure consistency of results. Cox proportional hazards models or Fine-Gray models were used, as appropriate. Hazard ratios (HRs) or subdistribution hazard ratios (sHRs) and their 95% CIs were reported. HRs and sHRs reflect results from Cox proportional hazards and Fine-Gray models, respectively. A P value less than .05 was considered statistically significant for all tests. All tests were 2-sided. Analyses were performed using SAS version 9.4 (SAS Institute Inc). The eAppendix in the Supplement contains additional information on the analytic approach.

Between January 14, 2011, and June 1, 2012, a total of 13 578 older patients with AF underwent first-time cardiac surgery. After excluding patients with planned off-pump operations (n = 841), endocarditis (n = 205), double valve procedures (both aortic and mitral operations; n = 807), congenital heart disease or cardiac transplant (n = 92), left ventricular assist device (n = 42), cardiogenic shock (n = 123), missing data on S-LAAO (n = 64), inability to link to Medicare claims (n = 337), missing anticoagulation data (n = 429), those without information on the primary surgical procedure (n = 95), and those with a duplicate Medicare record number (n = 19), a total of 10 524 patients met the study criteria.

The overall study cohort was older (median age, 76 years; interquartile range, 71-81 years), predominantly male (61%), and at high risk for stroke as demonstrated by a median CHA₂DS₂-VASc score of 4 (interquartile range, 3-5). Thirty percent of patients (n = 3163) underwent a mitral operation with or without CABG, 35% of patients (n = 3635) underwent an aortic procedure with or without CABG, and 35% of patients (n = 3726) underwent isolated CABG. Thirty-seven percent of patients (n = 3892) underwent S-LAAO. A comparison of the baseline characteristics of patients who did and did not receive S-LAAO at the time of cardiac surgery is presented in Table 1. Compared with patients who did not undergo S-LAAO, those who received S-LAAO more commonly had nonparoxysmal AF, a higher ejection fraction, a lower STS Predicted Risk of Mortality score, and lower rates of common stroke risk factors (diabetes, hypertension, and prior stroke). S-LAAO was performed more commonly at the time of certain types of cardiac surgery (eg, mitral valve operations, surgical AF ablation), in certain geographic regions (eg, Great Lakes region), and in academic centers (Table 2).

In the overall cohort, at a mean follow-up of 2.6 years, thromboembolism occurred in 5.4% of patients, hemorrhagic stroke in 0.9%, all-cause mortality in 21.5%, and the composite end point in 25.7% by 3 years. S-LAAO was associated with lower absolute rates of thromboembolism (4.2% vs 6.2%), all-cause mortality (17.3% vs 23.9%), and the composite end point (20.5% vs 28.7%), but no difference in rates of hemorrhagic stroke (0.9% vs 0.9%).

In unadjusted analyses, use of S-LAAO was associated with a significantly lower risk of thromboembolism (sHR, 0.66; 95% CI, 0.56-0.79; P < .001), all-cause mortality (HR, 0.70; 95% CI, 0.64-0.77; P < .001), and the composite end point (HR, 0.69; 95% CI, 0.63-0.75; P < .001) (Table 3; Figure).

The IPW cohort for the primary analysis was well-balanced based on assessment of the Cramer φ (eTable 1 in the Supplement) and the falsification end points that were not associated with S-LAAO use in the IPW models: readmission for lower extremity fracture (HR, 0.95; 95% CI, 0.72-1.26; P = .72) or pneumonia (HR, 1.10; 95% CI, 0.93-1.30; P = .29) (eTable 2 in the Supplement).

The results of the IPW-adjusted analyses were similar to the unadjusted analyses, and S-LAAO was associated with a significantly lower risk of thromboembolism (sHR, 0.67; 95% CI, 0.56-0.81; P < .001), all-cause mortality (HR, 0.88; 95% CI, 0.79-0.97; P = .001), and the composite end point (HR, 0.83; 95% CI, 0.76-0.91; P < .001) (Table 3). Results were unchanged in sensitivity analyses using regression analyses (eTable 3 in the Supplement). The adjusted rates of the study end points by exposure group are depicted in eTable 4 in the Supplement.

To assess the extent to which the observed association between S-LAAO and outcomes could be mediated by S-LAAO being a marker for higher overall quality of care, the proportion of eligible patients receiving standard of care treatments were compared among hospitals that performed S-LAAO 50% or more of the time vs less than 50% of the time. When comparing hospitals that performed more vs less S-LAAO, there were no differences in the rates of discharge oral anticoagulation use, prescription of β-blocker or lipid-lowering medications among patients undergoing CABG, or use of an internal mammary artery among patients with CABG and left main or proximal left anterior descending artery disease (eTable 5 in the Supplement). Subsequent multivariate adjusted Fine-Gray models demonstrated that the association between S-LAAO and thromboembolism did not vary based on concomitant surgical ablation (P for interaction = .89) or AF subtype (paroxysmal vs nonparoxysmal; P for interaction = .64).

Anticoagulation was prescribed to 68.9% of patients (n = 2680) who received S-LAAO and 60.3% (n = 3996) who did not receive S-LAAO (P < .001). There was a significant interaction between S-LAAO, discharge anticoagulation, and all-cause mortality (P = .004). In the IPW cohort of patients discharged without anticoagulation (37%, n = 3848), S-LAAO was associated with a significantly lower rate of thromboembolism (sHR, 0.26; 95% CI, 0.17-0.40; P < .001), but not all-cause mortality, hemorrhagic stroke, or the composite end point (Table 3). In the IPW cohort of patients discharged with anticoagulation (63%, n = 6676), there was no association between S-LAAO, thromboembolism, and all-cause mortality, although S-LAAO was associated with a lower risk for hemorrhagic stroke (sHR, 0.32; 95% CI, 0.17-0.57; P < .001) (Table 3). Results were unchanged in sensitivity analyses using regression analyses (eTable 3 in the Supplement). The Cramer φ statistics assessing balance among the cohorts stratified by discharge anticoagulation status are displayed in eTable 6 in the Supplement. Among patients who received S-LAAO, there was no adjusted association between discharge anticoagulation and thromboembolism (P = .79).

Quiz Ref IDIn a nationally representative cohort of older patients with AF undergoing cardiac surgery, S-LAAO (compared with no S-LAAO) was associated with a significantly lower risk of readmission for thromboembolism and all-cause mortality over the subsequent 3 years. Furthermore, the observed association between S-LAAO and lower rates of thromboembolism may have been primarily related to lower observed rates of thromboembolism in the substantial group of patients discharged without anticoagulation. While observational in nature, this analysis supports the use of S-LAAO in patients with AF at the time of cardiac surgery.

Prior reports on surgical S-LAAO, including a recent meta-analysis (which included 171 patients from 3 randomized clinical trials and 3482 patients from 4 observational studies)²³ have generally supported the notion that S-LAAO is associated with a reduction in stroke and mortality. However, the individual reports did not account for concomitant surgical ablation and anticoagulation strategies, both of which could have a meaningful effect on AF-related morbidity and mortality. Furthermore, the individual observational studies included in the meta-analysis were generally small, single center, and included only 30-day follow-up; notably, the largest study²⁴ (N = 1777) included in the meta-analysis did not use any statistical adjustment.

A recently published propensity-matched analysis of S-LAAO from the Mayo Clinic²⁵ (which was not included in the aforementioned meta-analysis²³) did not demonstrate an association between S-LAAO and long-term thromboembolic events. Most patients in the Mayo Clinic analysis did not have preoperative AF, those with surgical AF ablation were excluded, and the study size (461 patients per arm) was limited. The Left Atrial Appendage Closure by Surgery Study, which was recently presented at the European Society of Cardiology meeting but has not yet been published, was a small (n = 187) randomized trial demonstrating that S-LAAO reduced symptomatic and magnetic resonance image–defined asymptomatic thromboembolic events.²⁶ The current study was based on data from a representative national registry, exclusively included patients with AF, accounted for key AF management strategies, and used propensity score–based techniques with falsification end points to minimize the potential for residual confounding. To our knowledge, this study represents the largest study assessing long-term outcomes of LAAO by any method.

The strongest data supporting the strategy of LAAO are from randomized clinical trials comparing warfarin therapy with percutaneous LAAO using the WATCHMAN device (Boston Scientific). A patient-level meta-analysis of the WATCHMAN Left Atrial Appendage System for Embolic Protection in Patients With Atrial Fibrillation (PROTECT AF)⁸ Study and the Evaluation of the WATCHMAN LAA Closure Device in Patients With Atrial Fibrillation vs Long Term Warfarin Therapy (PREVAIL)⁹ Study, as well as the associated continued access registries, demonstrated that percutaneous LAAO with the WATCHMAN device was noninferior to warfarin for the end point of stroke, cardiovascular death, or systemic embolism.⁷ There were notable differences in the individual components of the composite end point, and LAAO was associated with a reduction in hemorrhagic stroke and cardiovascular death but increased rates of ischemic stroke.⁹

The current study has a number of key differences compared with the randomized WATCHMAN trials. In the current study, S-LAAO and no S-LAAO were compared in the overall population and in subpopulations defined by discharge anticoagulation strategies. In contrast, the randomized WATCHMAN trials^8,9 compared LAAO vs warfarin; in these studies, patients randomized to LAAO used short-term warfarin, which was discontinued after device endothelialization if the appendage was deemed sufficiently closed after assessment with transesophageal echocardiography. In the WATCHMAN meta-analysis,⁷ the association between LAAO and a reduction in cardiovascular mortality appeared to be driven by a reduction in hemorrhagic stroke and occurred despite increased rates of ischemic stroke; therefore, the benefit appeared to be due to avoidance of anticoagulation-related morbidity and mortality rather than a reduction in thromboembolism. The current study demonstrated that S-LAAO was associated with a significantly lower rate of thromboembolism among patients without oral anticoagulation. In the cohort of patients discharged with oral anticoagulation, S-LAAO was not associated with thromboembolism but was associated with a lower risk for hemorrhagic stroke, presumably related to eventual discontinuation of oral anticoagulation among S-LAAO patients.

The results from the current study have implications for the ongoing Left Atrial Appendage Occlusion Study (LAAOS) III (NCT01561651),²⁷ which is randomizing 4700 cardiac surgery patients with AF to S-LAAO vs no S-LAAO with 4 years of planned follow-up and a primary end point of stroke or systemic embolism. In this study, patients and the research team (except for the surgeon) are blinded to treatment assignment, and oral anticoagulation is recommended in both groups. Based on the results from the current study, the outcome of LAAOS III may depend on the proportion of patients who are discharged without anticoagulation or have anticoagulation discontinued after S-LAAO. The anticipated increased use of direct oral anticoagulants in LAAOS III compared with the current study will be an important difference between the 2 studies. Additional randomized studies comparing S-LAAO without anticoagulation vs systemic anticoagulation alone will be needed to define the optimal use of S-LAAO.

Quiz Ref IDDifferences in unobserved covariates are of concern in nonrandomized observational analyses, particularly among older cohorts. To assess the likelihood for residual confounding in the IPW cohort, the associations between treatment (S-LAAO vs no S-LAAO) and 2 common admission diagnoses (pneumonia and lower extremity fracture) were tested. These admission diagnoses were selected because they are expected to be more common among frail individuals and those with increasing comorbidity burden. Although residual confounding can never be completely excluded in nonrandomized studies, the observed null association between treatment and the falsification end points supports that differences in unobserved covariates are likely minimal.

There are important limitations associated with this study. First, treatment (S-LAAO vs no S-LAAO) was not randomized and the rationale for varying treatment decisions is unknown; although robust statistical methods were used to account for differences between groups, the potential for residual confounding cannot be ruled out. Second, the study population included adults 65 years of age and older with fee-for-service Medicare; therefore, the results may not be generalizable to younger individuals or those with different insurance. Third, study outcomes were defined using claims data rather than adjudicated end points as is the standard in randomized trials. Fourth, the STS Registry Data Collection Form version 2.73 does not collect data on the method of S-LAAO, and it is possible that the method of S-LAAO may be associated with the procedure’s effectiveness. Similarly, the STS Registry does not contain information on the completeness of S-LAAO (presence or absence of a residual leak), which may also influence outcomes. Fifth, owing to differences in procedure characteristics, the results from this study may not be generalizable to patients who undergo LAAO via a percutaneous approach. Sixth, the determination of discharge anticoagulation was based on a variable that reflects patients who were discharged with a prescription for oral anticoagulation, but it does not capture information regarding adherence or duration of therapy. Therefore, the discharge anticoagulation variable may not reflect long-term anticoagulation; results of the exploratory analyses (with stratification by discharge anticoagulation) should be interpreted accordingly. Seventh, most patients discharged with anticoagulation were discharged with warfarin, so the results may not be generalizable to patients treated with direct oral anticoagulants.

Among older patients with AF undergoing concomitant cardiac surgery, S-LAAO compared with no S-LAAO was associated with a lower risk of readmission for thromboembolism over 3 years. These findings support the use of S-LAAO, but randomized trials are necessary to provide definitive evidence.

Corresponding Author: J. Matthew Brennan, MD, MPH, Division of Cardiology, Duke University School of Medicine, 2400 Pratt St, Room 0311 Terrace Level, Durham, NC 27705 (j.matthew.brennan@dm.duke.edu).

Accepted for Publication: December 1, 2017.

Author Contributions: Drs Brennan and Friedman 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.

Concept and design: Friedman, Wang, Holmes, Badhwar, Jacobs, Gaca, Chow, Peterson, Brennan.

Acquisition, analysis, or interpretation of data: Friedman, Piccini, Wang, Zheng, Malaisrie, Suri, Mack, Jacobs, Chow, Peterson, Brennan.

Drafting of the manuscript: Friedman, Wang, Brennan.

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

Statistical analysis: Friedman, Wang, Zheng, Chow, Peterson.

Obtained funding: Peterson, Brennan.

Administrative, technical, or material support: Friedman, Holmes, Suri, Brennan.

Supervision: Friedman, Holmes, Suri, Badhwar, Gaca, Peterson, Brennan.

Conflict of Interest Disclosures: All authors have completed and submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Dr Friedman reported receiving grants from Boston Scientific, Abbott, and the National Cardiovascular Data Registry and he is funded by T32 training grant HL069749-13 from the National Institutes of Health. Dr Piccini reported receiving grants from ARCA Biopharma, Boston Scientific, Gilead Sciences, Janssen Pharmaceuticals, and Abbott to Duke for conduct of clinical trials; consultancies for Allergan, Janssen Pharmaceuticals, Bayer, Sanofi, Spectranetics, and Medtronic; and grants/grants pending from Johnson & Johnson, Boston Scientific, Gilead, St Jude Medical, and Spectranetics. Ms Wang reported receiving grants from the Food and Drug Administration, Innovation in Regulatory Science Award from Burroughs Welcome Fund, and a T32 training grant from the National Institutes of Health. Dr Holmes and the Mayo Clinic reported a financial interest in technology related to this research; that technology has been licensed to Boston Scientific. Dr Suri reported receiving funding for the Sorin Perceval Trial and the Publications Committee Partner Trial. Dr Peterson reported running the analysis center for the ACC STS Adult Cardiac Surgery Database and receiving a research grant from Janssen Pharmaceuticals and Eli Lilly. Dr Peterson is also a consultant for Janssen Pharmaceuticals and Boehringer Ingelheim. Dr Brennan holds an Innovation in Regulatory Science Award from Burroughs Welcome Fund (1014158) and a Food and Drug Administration grant (1U01FD004591-01). No other disclosures were reported.

Funding/Support: Funding for this study was made possible, in part, by grant 1U01FD004591-01 from the US Food and Drug Administration, which was awarded to Dr Brennan. Additional funding was provided through an Innovation in Regulatory Science Award from Burroughs Welcome Fund (1014158) awarded to Dr Brennan. Dr Friedman is funded by T32 training grant HL069749-13 from the National Institutes of Health.

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

Disclaimer: The views expressed in written materials or publications and by speakers and moderators do not necessarily reflect the official policies of the Department of Health and Human Services, nor does any mention of trade names, commercial practices, or organization imply endorsement by the US government. Dr Peterson, JAMA associate editor, had no role in the review of the manuscript or the decision to accept the manuscript for publication.

Additional Contributions: We acknowledge Felicia Graham, MBA (Duke Clinical Research Institute), for her longitudinal project leadership and Erin Campbell, MS (Duke Clinical Research Institute), for her editorial support. Neither Ms Graham nor Ms Campbell received compensation for their contributions apart from their employment at the institution where this study was conducted.