Long-Term Clinical Outcomes Following the WATCHMAN Device Use in Medicare Beneficiaries

Abstract

BACKGROUND:

Long-term outcomes following left atrial appendage occlusion outside clinical trials and small registries are largely unknown. Collecting these data was a condition of US market authorization of the WATCHMAN device. The aim of this analysis was to evaluate the rates of stroke, bleeding, and death among Medicare beneficiaries following left atrial appendage occlusion implantation during initial commercial availability of the WATCHMAN left atrial appendage occlusion device overall and in important subgroups.

METHODS:

All Medicare fee-for-service beneficiaries ≥65 years of age who underwent left atrial appendage occlusion from April 1, 2016, to August 31, 2020, were included based on the International Classification of Diseases, Tenth Revision, and Current Procedural Terminology codes. Over a 5-year follow-up period, the cumulative incidence over time of mortality, ischemic stroke, and major bleeding were calculated using the International Classification of Diseases, Tenth Revision, diagnosis codes for the full study cohort and within important prespecified subgroups.

RESULTS:

WATCHMAN recipients (n=48 763) were a median of 77 (interquartile range, 72–82) years of age, 42% female, and mostly White (93%). The median CHA₂DS₂VASc score was 4 (interquartile range, 3–5) with prior major bleeding in 42% and prior stroke in 12%. At 5 years, death occurred in 44%, bleeding in 15% (with higher risk early following implantation), and ischemic stroke in 7%. Each of these end points was more common with greater baseline age. Male patients had greater 5-year mortality than female patients (46.9% versus 40.6%), but there was no difference between sexes in the rates of ischemic stroke (6.6% versus 7.5%) or major bleeding (14.9% for both). WATCHMAN recipients with prior ischemic stroke or a major bleeding event were older and frailer; these groups had higher rates of ischemic stroke, major bleeding, and death.

CONCLUSIONS:

Compared with patients enrolled in the pivotal clinical trials, Medicare beneficiaries undergoing WATCHMAN implantation were older, more female, and had more comorbid conditions. Substantial long-term mortality and major bleeding following WATCHMAN reflect the high-risk nature of the patient population, while the ischemic stroke rate was relatively low (<1.5% per year).

WHAT IS KNOWN

• For patients with at least moderate risk of atrial fibrillation–related stroke, left atrial appendage occlusion with the WATCHMAN device is an alternative to oral anticoagulation for stroke risk reduction.

• Due to the relative novelty of the device, there are limited long-term data regarding important atrial fibrillation–related outcomes, and these data are especially limited outside of clinical trials.

WHAT THE STUDY ADDS

• Medicare beneficiaries who underwent first-generation WATCHMAN left atrial appendage occlusion in clinical practice were more reflective of the general population with atrial fibrillation than those patients included in clinical trials, but Black patients remained underrepresented compared with atrial fibrillation prevalence.

• Long-term mortality is high among the Medicare population treated with left atrial appendage occlusion, in clinical practice, but the rate of stroke is less than would be predicted.

Atrial fibrillation (AF) is a highly prevalent cardiovascular disease, and it is estimated that it will afflict over 12 million Americans by 2030.¹ Patients with AF are at an increased risk for stroke, due primarily to stagnant blood in the fibrillating atrium forming a thrombus, which can then embolize to the cerebral circulation. Numerous studies have demonstrated that the vast majority of these thrombi develop in the left atrial appendage.^2–4 Historically, patients with AF have been stratified for ischemic stroke risk based on demographic and clinical characteristics, and treatment with oral anticoagulation was recommended for all but the lowest-risk patients. Oral anticoagulation, however, presents its own risks, primarily of bleeding. First approved by the US Food and Drug Administration in 2015, the WATCHMAN device was developed to prevent thromboembolism from the left atrial appendage for patients with AF and at least moderate risk of AF-related thromboembolism.

There are limited long-term data available on outcomes in patients who undergo left atrial appendage occlusion (LAAO) with WATCHMAN in routine clinical practice. Claims data from the Centers for Medicare and Medicaid Services (CMS) are a useful source of long-term follow-up data, since most patients eligible for and treated with WATCHMAN are Medicare beneficiaries, and during the early period following market authorization of the WATCHMAN device, there were no other commercially available LAAO devices that shared similar coding. Furthermore, many of the outcomes of interest following LAAO including death and stroke are measurable in claims data. Thus, Medicare claims data offer an opportunity to understand how this novel technology has been applied in clinical practice, identify evidence gaps, and generate hypotheses for future analyses and clinical trials. In this study, we used Medicare claims data to evaluate the rates of stroke, bleeding, and death among the first cohort of WATCHMAN recipients in clinical practice.

METHODS

Investigators had access to the data. Because of the sensitive nature of the data collected for this study, a separate data use agreement with CMS is required to access the data. This study was reviewed and approved by the WCG Institutional Review Board with a waiver of authorization for use and disclosure of protected health information. The data set was received deidentified from CMS.

Data Sources

The data source for this analysis was the 100% Medicare Standard Analytical Files (Medicare). These data capture fee for service (FFS) beneficiaries’ encounters with the health care system and receipt of therapeutic interventions, including procedures and services. Because the majority of US citizens ≥65 years of age have Medicare insurance and >90% of WATCHMAN recipients are of Medicare age,⁵ Medicare data are generally representative of the characteristics and long-term outcomes of most patients undergoing these procedures.

Study Cohort

The study cohort comprised all Medicare beneficiaries who underwent WATCHMAN implantation between April 1, 2016, and August 31, 2020 (the index identification period). This time frame was chosen to ensure a minimum follow-up of 12 months for all surviving patients. In addition, restriction of the analysis to this time period ensured that only the International Classification of Diseases, Tenth Revision, Clinical Modification codes would be used to identify key outcomes. The chosen time period ensured that all WATCHMAN device implants were the WATCHMAN 2.5 device since the subsequent generation device (WATCHMAN FLX) was not commercially available in the United States until September 2020. Patients were eligible for inclusion if they had 12 months of continuous enrollment in Medicare FFS before the index date and underwent WATCHMAN implantation during the index identification period at the age of ≥65 years. WATCHMAN implantation was defined based on the relevant Current Procedural Terminology codes (33340 and 0281T) or the International Classification of Diseases, Tenth Revision, procedure code (02L73DK).

Outcomes of Interest

The outcomes of interest included ischemic stroke, hemorrhagic stroke, all-cause death, and major bleeding defined by the International Classification of Diseases, Tenth Revision, diagnoses codes listed as primary diagnoses (Tables S1 and S2). The follow-up period for each patient began on the index date and concluded at the patient’s death or the censor date defined as the end of the study period (August 31, 2021) or the end of Medicare FFS enrollment, whichever came first.

Statistical Analyses

All cohort construction was performed using the Instant Health Data software (Panalgo, Boston, MA) and R, version 3.2.1 (R Foundation for Statistical Computing, Vienna, Austria). Analyses were conducted using SAS, version 9.4 (SAS Institute, Cary, NC).

For demographic data and baseline comorbidities, continuous variables are summarized as mean, SD, and range, as well as median and interquartile range. Categorical variables are summarized as frequencies and percentages. Cumulative incidence was reported for the first occurrence of each event. All results comply with the CMS Cell Suppression Policy, meaning that no data cells (eg, admittances, discharges, patients) with fewer than 11 beneficiaries are reported.

Stratified analyses were planned based on race/ethnicity, sex, and age at the time of the index procedure. In addition, LAAO is often considered in patients who have not tolerated or have failed first-line stroke prevention therapy (ie, oral anticoagulation). We, therefore, performed outcome analyses in patients with prior stroke or major bleeding.

RESULTS

Patient Cohort

The total patient cohort with at least 1 year of follow-up consisted of 48 763 Medicare beneficiaries (Table). The cohort was elderly (median age, 77 years) and had a high level of comorbidity, with a median CHA₂DS₂VASC score of 4 (interquartile range, 4–5). The cohort was 42% female, and >93% of the participants were White. Approximately 15% had a history of stroke or transient ischemic attack and >40% had a prior bleeding event.

Table.

Baseline Characteristics

Outcomes

Overall

At 5 years, 44.2% of WATCHMAN recipients had died. The rate of death was reasonably stable with 8% to 10% of patients dying per year (Figure 1A). Ischemic stroke was infrequent with a rate of ≈1.0% to 1.5% per year for the first 5 years post-implantation leading to an overall rate of 6.6% at 5 years (Figure 2A). Hemorrhagic stroke was rare with a rate of 0.29% in the first year, most of which occurred in the first 6 months (0.18%). At 5 years, the cumulative frequency of hemorrhagic stroke was 0.97%. Bleeding was most frequent in the first year (7.3%), of which most occurred in the first 6 months after the WATCHMAN procedure with a 5-year total of 14.7% (Figure 3A).

Figure 1.

All-cause death among Medicare beneficiaries undergoing the WATCHMAN implantation procedure 2016-2020. Overall (A), by age group (B), and by sex (C).

Figure 2.

Ischemic stroke among Medicare beneficiaries undergoing the WATCHMAN implantation procedure 2016-2020. Overall (A), by age group (B), and by sex (C).

Figure 3.

Major bleeding among Medicare beneficiaries undergoing the WATCHMAN implantation procedure 2016-2020. Overall (A), by age group (B), and by sex (C).

Analyses of Subgroups

Demographic Subgroups

Variation in patient characteristics after stratifying the population by 5-year age increments is shown in the Table. The proportion of women increased among older strata, ranging from 40% of the 65- to 69- and 70- to 74-year age groups to 44% of the 85+-year age group. In addition, while the South region was the dominant region in all age groups, there was increasing implantation among patients in the Northeast and West with increasing age. Some comorbidities increased with increasing age including prior stroke/transient ischemic attack and prior major bleeding. Other comorbidities decreased with age including diabetes and end-stage renal disease, likely reflecting the competing risks of death associated with these conditions. All outcomes increased with increasing age: the oldest groups had the highest rates of death, stroke, and bleeding, whereas the youngest groups had the lowest (Figures 1B, 2B, and 3B).

Women were, on average, 0.6 years older than men with a CHA₂DS₂VASc score about 1 point higher (Table). Women were also more likely to have had a prior stroke but were less likely to have peripheral artery disease, prior myocardial infarction, or a prior major bleeding event than men. Over 5 years of follow-up, there were no significant differences in bleeding or ischemic stroke rates by sex. The rate of all-cause death was higher among men than women with the curves appearing to separate after 6 months post-index (P<0.0001; Figure 1C).

Only a small percentage of patients in this cohort belonged to racial/ethnic subgroups. As such, drawing conclusions about racial differences was challenging, especially among Asian patients, Hispanic patients, and patients from other, unknown, or multiple race groups. Black patients who underwent WATCHMAN implantation had generally greater comorbidity burden than White patients including higher CHA₂DS₂VASc scores, Charlson comorbidity indices, and Hospital Frailty Risk Scores. Black patients were also more likely to be female and had a greater burden of nearly all measured comorbidities with large differences for several factors including end-stage renal disease, which was present among 13.6% of Black WATCHMAN recipients compared with 2.6% of White WATCHMAN recipients. Black patients had higher rates of death and major bleeding than White patients, but there was no difference in the rate of stroke among racial/ethnic groups.

Subgroups Based on Prior Events

Approximately half of all Medicare beneficiaries who underwent WATCHMAN implantation during this period had a prior bleeding event, a prior ischemic stroke, or both, so separate analyses of these groups were performed. Medicare beneficiaries with a prior stroke who underwent WATCHMAN implantation differed from those without prior stroke in several ways. In particular, they were frailer and were more likely to have serious comorbidities including congestive heart failure and diabetes (Table S3). Patients with prior stroke were also more likely to have had a prior major bleeding event (64% versus 40%). At 5 years of follow-up after WATCHMAN implantation, patients with prior ischemic stroke had higher rates of death (48% versus 44%), ischemic stroke (12% versus 6%), and major bleeding (19% versus 14%; Figure S1).

Similarly, when compared with those without a prior major bleeding event, WATCHMAN recipients with a prior major bleeding event were older, frailer, and were more likely to have important comorbidities including prior stroke (18% versus 8%; Table S3). Over the 5-year follow-up period, patients with prior bleeding were more likely to die (49% versus 41%) and had higher rates of ischemic stroke (8% versus 6%) and major bleeding (22% versus 10%) than those without a prior major bleeding event (Figure S2).

Only a small fraction of the cohort had prior hemorrhagic stroke (1797/48 763; 3.7%). These WATCHMAN recipients were similar to those without prior hemorrhagic stroke with regard to age and sex but were dramatically more likely to have had prior ischemic stroke (42% versus 9%). Given the small sample size, outcomes were not examined in this subgroup.

DISCUSSION

There are few reports of long-term outcomes following WATCHMAN implantation in general clinical practice.^6–8 To address this evidence gap, our claims-based analysis complements data collected from long-term follow-up of clinical trials and registries and offers several important and clinically relevant observations.

The first important observation from this analysis is that Medicare beneficiaries undergoing WATCHMAN implantation tended to be of more advanced age with a higher burden of comorbidity than patients included in clinical trials of WATCHMAN. The distribution of this cohort was skewed toward older age with >35% being older than 80 years of age. This cohort had more female representation than the clinical trial cohorts (40% versus ≈30%), which has consequential implications for clinical outcomes. On the one hand, female sex has been associated with more frequent adverse events following LAAO⁹ in clinical practice. On the other hand, these unadjusted data suggest a lower risk of mortality for the female cohort compared with the male cohort (Figure 1C) following LAAO, although no formal comparisons were performed based on this nonmodifiable risk factor. This all-comer Medicare cohort had more women and a greater burden of comorbidity^10,11 than would be typical in a clinical trial cohort. For example, the mean CHA₂DS₂VASc score in this cohort was 4.3, which was slightly lower than reported from the National Cardiovascular Data Registry- Left Atrial Appendage Occlusion Registry⁹ but higher than that seen in premarket randomized trials (3.5–4.0), which reflects, in part, a greater proportion of women who are granted an additional point for female sex.¹² These differences in demographics and comorbidity in general practice, as informally compared with the premarket trials, likely relate to the CMS national coverage decision stipulating that WATCHMAN candidates must be deemed unable to take long-term oral anticoagulation and should have CHADS₂ score ≥2 or CHA₂DS₂-VASc score ≥3.

This journey from clinical trial to clinical practice is emblematic of the typical path for a novel cardiovascular device. Premarket clinical trials must answer questions of feasibility first and then safety and effectiveness under controlled and ideal circumstances, and this approach has been shown to exclude patients with multiple comorbidities and advanced age, women, and other minoritized groups.^13,14 Nonetheless, market authorization and coverage decisions¹⁵ must be made in the face of this lingering uncertainty, so subsequent application in clinical practice populations may be associated with unanticipated outcomes. Thus, questions of clinical or real-world performance must be addressed to identify factors that meaningfully change risk-benefit considerations.

A second key observation is that mortality was high among WATCHMAN recipients over the 5 years after implantation. Nearly 45% of WATCHMAN recipients were dead at 5 years with consistent yearly mortality rates of 8% to 9%. This mortality rate is higher than that seen in the clinical practice registry of WATCHMAN recipients in Europe (15% at 2 years¹⁶) and nearly double that seen in long-term follow-up of patients enrolled in the continued access registries, which demonstrated a 5-year all-cause mortality rate of ≈20%.¹⁷ In the 5-year follow-up of subjects randomized to LAAO with WATCHMAN in Watchman Left Atrial Appendage System for Embolic Protection in Patients With Atrial Fibrillation and Prospective Randomized Evaluation of the WATCHMAN LAA Closure Device in Patients With Atrial Fibrillation, combined all-cause mortality was 3.6% per 100 patient-years.¹⁸ However, the mortality seen in this population is slightly better than that seen in an all-comer population of Medicare beneficiaries with AF. Piccini et al¹⁹ demonstrated that following incident AF in 2007, 3-year mortality among Medicare beneficiaries was >40% and twice that seen in the general Medicare population standardized to age and sex. Of note, the 5-year mortality rate in the Medicare population far exceeds the 5-year rates of stroke (≈7%) and major bleeding (≈15%), which are like those seen in clinical trials.¹⁸ These findings highlight the high competing risks of death in this older and highly comorbid Medicare population.

The observed ischemic stroke rate in this population (≈1.4% per year) was similar to the predicted rate with warfarin treatment and less than would be predicted without any treatment based on the CHA₂DS₂VASc score alone (median, 4), which would estimate an annual stroke risk of 4.8%.^17,18 This finding mirrors findings from long-term follow-up of patients enrolled in the WATCHMAN clinical trials,¹⁸ a European real-world registry,¹⁶ and continued access registries.¹⁷ For example, in the latter case, the expected ischemic stroke rate based on the CHA₂DS₂VASc score was 5.8% to 7.1% per 100 patient-years compared with the observed rate of 1.3% to 2.2%.¹⁷ While comparisons of stroke risk between LAAO and no treatment or warfarin were historically relevant, contemporary practice includes the widespread availability of direct-acting oral anticoagulants for which randomized comparison data are lacking but forthcoming.^20,21 No such comparative analyses were performed in this analysis, but prior observational investigations have attempted to address this gap in knowledge.^7,8

Third, there were substantial demographic, clinical, and outcome similarities between subgroups with prior major bleeding and prior stroke. As such, it is not surprising that both of these key subgroups experienced particularly high rates of post-WATCHMAN bleeding, stroke, and death. These relationships were particularly pronounced among beneficiaries with prior history of bleeding among whom 22% had a subsequent major bleeding event compared with only 10% among WATCHMAN recipients without a history of bleeding. It is noteworthy that >40% of our study cohort had a history of major bleeding—a finding that is consistent with our personal observations that a history of bleeding is a common reason for referral for WATCHMAN among patients with AF. In this analysis, no comparison was made between WATCHMAN and long-term oral anticoagulation, but comparative outcomes between these 2 strategies for AF-related stroke risk reduction are needed to better inform decision-making for a population that may have complex and competing relative contraindications to both strategies. Future randomized trials will be critical to untangle comparative outcomes in this complex space.

Finally, there was a striking imbalance noted by race with >93% of the cohort being White. The cohort of Black patients included just 1583 beneficiaries (3.25%) despite this group making up about 10% of the total Medicare population. While Black Americans have a lower lifetime likelihood of AF diagnosis compared with other groups,²² the prescription of oral anticoagulation after AF diagnosis is less likely compared with other groups,²³ and the associated risk of stroke and bleeding is higher^24,25 regardless of anticoagulation prescription status.²⁴ Whether this latter finding reflects differences in pathophysiology, medication persistence, or other factors remains uncertain and is deserving of future investigation to identify optimal strategies to balance risks of stroke and bleeding. In the meantime, however, patients with AF with lower socioeconomic status and those in racially and ethnically minoritized groups have less access to LAAO even when controlling for geographic factors.²⁶ These differences mirror those seen for other cardiovascular procedures, suggesting that the differences are not related to medical decision-making per se but, rather, structural differences in access to care.

Strengths and Limitations

First, this report relies on Medicare claims data, which lack the rigor, specificity, and nuance of data collected prospectively through clinical trials or registries. Nonetheless, these data offer a broad view of the current clinical application of a novel medical device about which there remain unanswered questions. Furthermore, it has long been recognized that clinical practice populations differ in important ways from clinical trial populations, with each offering important and complementary insights. Indeed, in recognition of these limitations of clinical trials, LAAO devices are covered under the Coverage With Evidence Development program,²⁷ which compels ongoing postmarket data collection as a condition of Medicare coverage, which is achieved primarily through the LAAO registry. The claims-based analyses presented here complement the rigorous but more selected data from clinical trials, as well as the more detailed but shorter-term data available through the LAAO registry. Analyses based on claims-based diagnoses are imprecise, and outcomes cannot be independently verified; there are no clear standards for what defines major bleeding in administrative claims, so comparisons between analyses are not straightforward. No inference is made with regard to cause and effect between interventions and outcomes, and no comparisons have been made between WATCHMAN and any other treatments for the reduction in AF-related stroke risk. Nonetheless, we believe the observed patterns can help to inform future hypothesis generation, clinical decision-making, and regulatory and policy considerations.

Second, despite the lack of inclusion or exclusion criteria, our study of Medicare beneficiaries still has limitations to generalizability. Although most patients eligible for WATCHMAN fall into this population, there are other populations of patients for whom patterns seen here may not apply, including individuals insured by Medicare Advantage, which now comprise ≈50% of Medicare beneficiaries²⁸ without substantial differences in beneficiary characteristics compared with fee-for-service Medicare,^29,30 the Department of Defense (Veteran’s Affairs), or private insurance. Also, during the period of study, WATCHMAN 2.5 was the only commercially available LAAO device, so the findings here may not generalize to other LAAO devices including the contemporary version of the WATCHMAN device (WATCHMAN FLX). Finally, medical therapy has changed in meaningful ways since these data were collected. This includes recommended periprocedural anticoagulation and antiplatelet therapy, as well as therapy for the management of AF and related cardiovascular conditions like heart failure. These evolutions may have a meaningful impact on the outcomes reported.

Conclusions

This report represents an unfiltered view of long-term outcomes following WATCHMAN implantation among Medicare beneficiaries. These descriptive data help define how this novel technology has been applied after it became commercially available. Consistent with previous studies, Medicare beneficiaries undergoing WATCHMAN implantation tended to be older and more comorbid than the typical clinical trial participant, and the proportion of women receiving WATCHMAN in clinical practice was numerically greater than that seen in clinical trials. Patients undergoing WATCHMAN placement in clinical practice have high mortality at 5 years that cannot be directly attributed to AF (eg, stroke) or treatment thereof (eg, major bleeding), suggesting significant competing risks of death. Future investigations should seek to better untangle comparative risks and benefits of WATCHMAN and alternative treatment strategies, as well as to better understand and address racial differences in WATCHMAN utilization and general AF-related outcomes.

ARTICLE INFORMATION

Acknowledgments

The authors acknowledge Jianbo Na, PhD, who provided additional statistical support for this analysis. Dr Na is a full-time employee of Boston Scientific Corporation.

Sources of Funding

This study was supported by Boston Scientific with analyses performed by analysts employed by Boston Scientific, but the design, interpretation, and writing were the responsibility of all authors.

Disclosures

Dr Zeitler reports consulting and speaking fees for Abbott, Biosense Webster, Medtronic, and Sanofi; travel support from Abbott, Biosense Webster, and Medtronic; and research support from Biosense Webster, Sanofi, and the National Institutes of Health. B. Bian, B.G., Dr Allocco, Dr Christen, and Dr Roy were full-time employees of Boston Scientific Corporation at the time of this work. Dr Cohen reports research grant support (to institution) from Boston Scientific, Abbott, Edwards Lifesciences, Philips, CathWorks, I-Rhythm, and Zoll Medical and consulting income from Boston Scientific, Abbott, Edwards Lifesciences, Medtronic, Elixir Medical, and HeartBeam. The other authors report no conflicts.

Supplemental Material

Tables S1–S3

Figures S1 and S2