Original Research

Low-dose aspirin commonly is used for the prevention of cardiovascular disease (CVD) but is associated with an increased risk of major bleeding.¹ The use of aspirin for primary prevention is largely extrapolated from clinical trials showing benefit in the secondary prevention of myocardial infarction and ischemic stroke. However, results from the Aspirin in Reducing Events in the Elderly (ASPREE) trial challenged this practice.² The ASPREE trial, conducted in the United States and Australia from 2010 to 2014, sought to determine whether daily 100 mg aspirin, was superior to placebo in promoting disability-free survival among older adults. Participants were aged ≥ 70 years (≥ 65 years for Hispanic and Black US participants), living in the community, and were free from preexisting CVD, cerebrovascular disease, or any chronic condition likely to limit survival to < 5 years. The study found no significant difference in the primary endpoints of death, dementia, or persistent physical disability, but there was a significantly higher risk of major hemorrhage in the aspirin group (3.8% vs 2.8%; hazard ratio, 1.38; 95% CI, 1.18-1.62; P < .001).

Several medical societies have updated their guideline recommendations for aspirin for primary prevention of CVD. The 2022 United States Public Service Task Force (USPSTF) provides a grade C recommendation (at least moderate certainty that the net benefit is small) to consider low-dose aspirin for the primary prevention of CVD on an individual patient basis for adults aged 40 to 59 years who have a ≥ 10% 10-year CVD risk. For adults aged ≥ 60 years, the USPSTF recommendation is grade D (moderate or high certainty that the practice has no net benefit or that harms outweigh the benefits) for low-dose aspirin use.^1,3 The American College of Cardiology and American Heart Association (ACC/AHA) recommend considering low-dose aspirin for primary prevention of atherosclerotic cardiovascular disease (ASCVD) among select adults aged 40 to 70 years at higher CVD risk but not at increased risk of bleeding.⁴ The American Diabetes Association (ADA) recommends low-dose aspirin for primary prevention of CVD in patients with diabetes and additional risk factors such as family history of premature ASCVD, hypertension, dyslipidemia, smoking, or chronic kidney disease, and who are not at higher risk of bleeding.⁵ The ADA standards also caution against the use of aspirin as primary prevention in patients aged > 70 years. Low-dose aspirin use is not recommended for the primary prevention of CVD in older adults or adults of any age who are at increased risk of bleeding.

Recent literature using the US Department of Veterans Affairs (VA) Corporate Data Warehouse database confirms 86,555 of 1.8 million veterans aged > 70 years (5%) were taking low-dose aspirin for primary prevention of ASCVD despite guideline recommendations.⁶ Higher risk of gastrointestinal and other major bleeding from low-dose aspirin has been reported in the literature.¹ Major bleeds represent a significant burden to the health care system with an estimated mean $13,093 cost for gastrointestinal bleed hospitalization.⁷

Considering the large scale aspirin use without appropriate indication within the veteran population, the risk of adverse effects, and the significant cost to patients and the health care system, it is imperative to determine the best approach to efficiently deprescribe aspirin for primary prevention among geriatric patients. Deprescribing refers to the systematic and supervised process of dose reduction or drug discontinuation with the goal of improving health and/or reducing the risk of adverse effects.⁸ During patient visits, primary care practitioners (PCPs) have opportunities to discontinue aspirin, but these encounters are time-limited and deprescribing might be secondary to more acute primary care needs. The shortage of PCPs is expected to worsen in coming years, which could further reduce their availability to assess inappropriate aspirin use.⁹

VA clinical pharmacist practitioners (CPPs) serve as medication experts and work autonomously under a broad scope of practice as part of the patient aligned care team.^10-12 CPPs can free up time for PCPs and facilitate deprescribing efforts, especially for older adults. One retrospective cohort study conducted at a VA medical center found that CPPs deprescribed more potentially inappropriate medications among individuals aged ≥ 80 years compared with usual care with PCPs (26.8% vs 16.1%; P < .001).^12,13 An aspirin deprescribing protocol conducted in 2022 resulted in nearly half of veterans aged ≥ 70 years contacted by phone agreeing to stop aspirin. Although this study supports the role pharmacists can play in reducing aspirin use in accordance with guidelines, the authors acknowledge that their interventions had a mean time of 12 minutes per patient and would require workflow changes.¹⁴ The purpose of this study is to evaluate the efficiency of aspirin deprescribing through 2 approaches: direct deprescribing by pharmacists using populationlevel review compared with clinicians following a pharmacist-led education.

Methods

This was a single-center quality improvement cohort study at the Durham VA Health Care System (DVAHCS) in North Carolina. Patients included were aged ≥ 70 years without known ASCVD who received care at any of 3 DVAHCS community-based outpatient clinics and prescribed aspirin. Patient data was obtained using the VIONE (Deprescribing Dashboard called Vital, Important, Optional, Not indicated, and Every medication has a specific indication or diagnosis) dashboard.¹⁵ VIONE was developed to identify potentially inappropriate medications (PIMs) that are eligible to deprescribe based on Beers Criteria, Screening Tool of Older Personsf Prescriptions criteria, and common clinical scenarios when clinicians determine the risk outweighs the benefit to continue a specific medication. ^16,17 VIONE is used to reduce polypharmacy and improve patient safety, comfort, and medication adherence. Aspirin for patients aged ≥ 70 years without a history of ASCVD is a PIM identified by VIONE. Patients aged ≥ 70 years were chosen as an inclusion criteria in this study to match the ASPREE trial inclusion criteria and age inclusion criteria in the VIONE dashboard for aspirin deprescribing.² Patient lists were generated for these potentially inappropriate aspirin prescriptions for 3 months before clinician staff education presentations, the day of the presentations, and 3 months after.

The primary endpoint was the number of veterans with aspirin deprescribed directly by 2 pharmacists over 12 weeks, divided by total patient care time spent, compared with the change in number of veterans with aspirin deprescribed by any DVAHCS physician, nurse practitioner, physician assistant, or CPP over 12 weeks, divided by the total pharmacist time spent on PCP education. Secondary endpoints were the number of aspirin orders discontinued by pharmacists and CPPs, the number of aspirin orders discontinued 12 weeks before pharmacist-led education compared with the number of aspirin orders discontinued 12 weeks after CPP-led education, average and median pharmacist time spent per patient encounter, and time of direct patient encounters vs time spent on PCP education.

Pharmacists reviewed each patient who met the inclusion criteria from the list generated by VIONE on December 1, 2022, for aspirin appropriateness according to the ACC/AHA and USPSTF guidelines, with the goal to discontinue aspirin for primary prevention of ASCVD and no other indications.^1,4 Pharmacists documented their visits using VIONE methodology in the Computerized Patient Record System (CPRS) using a polypharmacy review note. CPPs contacted patients who were taking aspirin for primary prevention by unscheduled telephone call to assess for aspirin adherence, undocumented history of ASCVD, cardiovascular risk factors, and history of bleeding. Aspirin was discontinued if patients met guideline criteria recommendations and agreed to discontinuation. Risk-benefit discussions were completed when patients without known ASCVD were considered high risk because the ACC/AHA guidelines mention there is insufficient evidence of safety and efficacy of aspirin for primary prevention for patients with other known ASCVD risk factors (eg, strong family history of premature myocardial infarction, inability to achieve lipid, blood pressure, or glucose targets, or significant elevation in coronary artery calcium score).

High risk was defined as family history of premature ASCVD (in a male first-degree relative aged < 55 years or a female first-degree relative aged < 65 years), most recent blood pressure or 2 blood pressure results in the last 12 months > 160/100 mm Hg, recent hemoglobin A_1c > 9%, and/or low-density lipoprotein > 190 mg/dL or not prescribed an indicated statin.³ Aspirin was continued or discontinued according to patient preference after the personalized risk-benefit discussion.

For patients with a clinical indication for aspirin use other than ASCVD (eg, atrial fibrillation not on anticoagulation, venous thromboembolism prophylaxis, carotid artery disease), CPPs documented their assessment and when appropriate deferred to the PCP for consideration of stopping aspirin. For patients with undocumented ASCVD, CPPs added their ASCVD history to their problem list and aspirin was continued. PCPs were notified by alert when aspirin was discontinued and when patients could not be reached by telephone.

presented a review of recent guideline updates and supporting literature at 2 online staff meetings. The education sessions lasted about 10 minutes and were presented to PCPs across 3 community-based outpatient clinics. An estimated 40 minutes were spent creating the PowerPoint education materials, seeking feedback, making edits, and answering questions or emails from PCPs after the presentation. During the presentation, pharmacists encouraged PCPs to discontinue aspirin (active VA prescriptions and reported over-the-counter use) for primary prevention of ASCVD in patients aged ≥ 70 years during their upcoming appointments and consider risk factors recommended by the ACC/AHA guidelines when applicable. PCPs were notified that CPPs planned to start a population review for discontinuing active VA aspirin prescriptions on December 1, 2022. The primary endpoint and secondary endpoints were analyzed using descriptive statistics. All data were analyzed using Microsoft Excel.

Results

A total of 868 patients aged ≥ 70 years with active prescriptions for aspirin were identified on December 1, 2022. After applying inclusion and exclusion criteria for the pharmacist population review, 224 patients were included for cohort final analysis (Figure). All 868 patients were eligible for the CPP intervention. Primary reasons for exclusion from the CPP population included over-thecounter aspirin and a history of ASCVD in the patient’s problem list. All patients were male, with a mean (SD) age of 75 (4.4) years (Table 1). Most patients were prescribed aspirin, 81 mg daily (n = 220; 98%).

The direct CPP deprescribing intervention resulted in 2 aspirin prescriptions discontinued per hour of pharmacist time and 67 aspirin prescriptions discontinued per hour of pharmacist time via the PCP education intervention. CPPs discontinued 66 aspirin orders in the 12 weeks before the PCP education sessions. A total of 230 aspirin prescriptions were discontinued in the 12 weeks following the PCP education sessions, with 97 discontinued directly by CPPs and 133 discontinued by PCPs. The PCP education session yielded an additional 67 discontinued aspirin orders compared with the 12 weeks before the education sessions (Table 2).

The CPP direct deprescribing intervention took about 48.3 hours, accounting for health record review and time interacting with patients. The PCP education intervention took about 60 minutes, which included time for preparing and delivering education materials (Table 3). CPP deprescribing encounter types, interventions, and related subcategories, and other identified indications to continue aspirin are listed in Table 4.

Discussion

Compared with direct deprescribing by pharmacists, the PCP education intervention was more efficient based on number of aspirin orders discontinued by pharmacist time. PCPs discontinued twice as many aspirin prescriptions in the 12 weeks after pharmacist-led education compared with the 12 weeks before.

Patients were primarily contacted by telephone (73%) for deprescribing. Among the 163 patients reached by phone and encouraged to discontinue aspirin, 97 patients (60%) accepted the recommendation, which was similar to the acceptance rates found in the literature (48% to 55%).^14,18 Although many veterans continued taking aspirin (78%), most had indications for its continued use, such as a history of ASCVD, atrial fibrillation without anticoagulation, and carotid artery stenosis, and complex comorbidities that required further discussion with their PCP. Less common uses for aspirin were identified through CPRS review or patient reports included cerebral small vessel disease without history of ASCVD, subclavian artery stenosis, thrombocytosis, bioprosthetic valve replacement, giant cell arteritis, rheumatoid arthritis, and prevention of second eye involvement of ischemic optic neuropathy.

to describe the benefit of clinical pharmacy services for deprescribing aspirin for primary prevention of ASCVD through PCP education. Previously published literature has assessed alternative ways to identify or discontinue PIMs—including aspirin—among geriatric patients. One study evaluated the use of marking inappropriate aspirin prescriptions in the electronic health database, leading to a significant reduction in incidence of inappropriate aspirin prescribing; however, it did not assess changes in discontinuation rates of existing aspirin prescriptions.¹⁹ The previous VA pharmacist aspirin deprescribing protocol demonstrated pharmacists’ aptitude at discontinuing aspirin for primary prevention but only used direct patient contact and did not compare efficiency with other methods, including PCP education.¹⁴

This quality improvement project contributes new data to the existing literature to support the use of clinical pharmacists to discontinue aspirin for primary prevention and suggests a strong role for pharmacists as educators on clinical guidelines, in addition to their roles directly deprescribing PIMs in clinical practice. This study is further strengthened by its use of VIONE, which previously has demonstrated effectiveness in deprescribing a variety of PIMs in primary care settings.²⁰

Despite using VIONE for generating a list of patients eligible for deprescription, our CPRS review found that this list was frequently inaccurate. For example, a small portion of patients were on the VIONE generated list indicating they had no ASCVD history, but had transient ischemic attack listed in their problem lists. Patient problem lists often were missing documented ASCVD history that was revealed by patient interview or CPRS review. It is possible that patients interviewed might have omitted relevant ASCVD history because of low health literacy, conditions affecting memory, or use of health care services outside the VA system.

There were several instances of aspirin used for other non-ASCVD indications, such as primary stroke prevention in atrial fibrillation. The ACC/AHA atrial fibrillation guidelines previously provided a Class IIb recommendation (benefit is greater than risk but additional studies are needed) for considering no antithrombic therapy or treatment with oral anticoagulant or aspirin for nonvalvular atrial fibrillation with CHA₂DS₂-VASc (Congestive heart failure, Hypertension, Age [> 65 y, 1 point; > 75 y, 2 points], Diabetes, previous Stroke/transient ischemic attack [2 points]) score of 1.²¹ The ACC/ AHA guidelines were updated in 2023 to recommend against antiplatelet therapy as an alternative to anticoagulation for reducing cardioembolic stroke risk among patients with atrial fibrillation with no indication for antiplatelet therapy because of risk of harm.²² If a patient has no risk factors for stroke, aspirin is not recommended to prevent thromboembolic events because of a lack of benefit. Interventions from this quality improvement study were completed before the 2023 atrial fibrillation guideline was published and therefore in this study aspirin was not discontinued when used for atrial fibrillation. Aspirin use for atrial fibrillation might benefit from similar discontinuation efforts analyzed within this study. Beyond atrial fibrillation, major guidelines do not comment on the use of aspirin for any other indications in the absence of clinical ASCVD.

Limitations

This study is limited by the lack of clinical consensus for complex patients and demonstrates the importance of individualized patient assessment when considering discontinuing aspirin. Because of the project’s relatively short intervention period, aspirin deprescribing rates could decrease over time and repeated education efforts might be necessary to see lasting impact. Health care professionals from services outside of primary care also might have discontinued aspirin during the study period unrelated to the education and these discontinued aspirin prescriptions could contribute to the higher rate observed among PCPs. This study included a specific population cohort of male, US veterans and might not reflect other populations where these interventions could be implemented.

The measurement of time spent by pharmacists and PCPs is an additional limitation. Although it is expected that PCPs attempt to discontinue aspirin during their existing patient care appointments, the time spent during visits was not measured or documented. Direct deprescribing by pharmacist CPRS review required a significant amount of time and could be a barrier to successful intervention by CPPs in patient aligned care teams.

To reduce the time pharmacists spent completing CPRS reviews, an aspirin deprescribing clinical reminder tool could be used to assess use and appropriate indication quickly during any primary care visit led by a PCP or CPP. In addition, it is recommended that pharmacists regularly educate health care professionals on guideline recommendations for aspirin use among geriatric patients. Future studies of the incidence of major cardiovascular events after aspirin deprescribing among geriatric patients and a longitudinal cost/benefit analysis could support these initiatives.

Conclusions

In this study, pharmacists successfully deprescribed inappropriate medications, such as aspirin. However, pharmacist-led PCP education is more efficient compared with direct deprescribing using a population-level review. PCP education requires less time and could allow ambulatory care pharmacists to spend more time on other direct patient care interventions to improve quality and access to care in primary care clinics. This study’s results further support the role of pharmacists in deprescribing PIMs for older adults and the use of a deprescribing tool, such as VIONE, in a primary care setting.

Low-dose aspirin commonly is used for the prevention of cardiovascular disease (CVD) but is associated with an increased risk of major bleeding.¹ The use of aspirin for primary prevention is largely extrapolated from clinical trials showing benefit in the secondary prevention of myocardial infarction and ischemic stroke. However, results from the Aspirin in Reducing Events in the Elderly (ASPREE) trial challenged this practice.² The ASPREE trial, conducted in the United States and Australia from 2010 to 2014, sought to determine whether daily 100 mg aspirin, was superior to placebo in promoting disability-free survival among older adults. Participants were aged ≥ 70 years (≥ 65 years for Hispanic and Black US participants), living in the community, and were free from preexisting CVD, cerebrovascular disease, or any chronic condition likely to limit survival to < 5 years. The study found no significant difference in the primary endpoints of death, dementia, or persistent physical disability, but there was a significantly higher risk of major hemorrhage in the aspirin group (3.8% vs 2.8%; hazard ratio, 1.38; 95% CI, 1.18-1.62; P < .001).

Several medical societies have updated their guideline recommendations for aspirin for primary prevention of CVD. The 2022 United States Public Service Task Force (USPSTF) provides a grade C recommendation (at least moderate certainty that the net benefit is small) to consider low-dose aspirin for the primary prevention of CVD on an individual patient basis for adults aged 40 to 59 years who have a ≥ 10% 10-year CVD risk. For adults aged ≥ 60 years, the USPSTF recommendation is grade D (moderate or high certainty that the practice has no net benefit or that harms outweigh the benefits) for low-dose aspirin use.^1,3 The American College of Cardiology and American Heart Association (ACC/AHA) recommend considering low-dose aspirin for primary prevention of atherosclerotic cardiovascular disease (ASCVD) among select adults aged 40 to 70 years at higher CVD risk but not at increased risk of bleeding.⁴ The American Diabetes Association (ADA) recommends low-dose aspirin for primary prevention of CVD in patients with diabetes and additional risk factors such as family history of premature ASCVD, hypertension, dyslipidemia, smoking, or chronic kidney disease, and who are not at higher risk of bleeding.⁵ The ADA standards also caution against the use of aspirin as primary prevention in patients aged > 70 years. Low-dose aspirin use is not recommended for the primary prevention of CVD in older adults or adults of any age who are at increased risk of bleeding.

Recent literature using the US Department of Veterans Affairs (VA) Corporate Data Warehouse database confirms 86,555 of 1.8 million veterans aged > 70 years (5%) were taking low-dose aspirin for primary prevention of ASCVD despite guideline recommendations.⁶ Higher risk of gastrointestinal and other major bleeding from low-dose aspirin has been reported in the literature.¹ Major bleeds represent a significant burden to the health care system with an estimated mean $13,093 cost for gastrointestinal bleed hospitalization.⁷

Considering the large scale aspirin use without appropriate indication within the veteran population, the risk of adverse effects, and the significant cost to patients and the health care system, it is imperative to determine the best approach to efficiently deprescribe aspirin for primary prevention among geriatric patients. Deprescribing refers to the systematic and supervised process of dose reduction or drug discontinuation with the goal of improving health and/or reducing the risk of adverse effects.⁸ During patient visits, primary care practitioners (PCPs) have opportunities to discontinue aspirin, but these encounters are time-limited and deprescribing might be secondary to more acute primary care needs. The shortage of PCPs is expected to worsen in coming years, which could further reduce their availability to assess inappropriate aspirin use.⁹

VA clinical pharmacist practitioners (CPPs) serve as medication experts and work autonomously under a broad scope of practice as part of the patient aligned care team.^10-12 CPPs can free up time for PCPs and facilitate deprescribing efforts, especially for older adults. One retrospective cohort study conducted at a VA medical center found that CPPs deprescribed more potentially inappropriate medications among individuals aged ≥ 80 years compared with usual care with PCPs (26.8% vs 16.1%; P < .001).^12,13 An aspirin deprescribing protocol conducted in 2022 resulted in nearly half of veterans aged ≥ 70 years contacted by phone agreeing to stop aspirin. Although this study supports the role pharmacists can play in reducing aspirin use in accordance with guidelines, the authors acknowledge that their interventions had a mean time of 12 minutes per patient and would require workflow changes.¹⁴ The purpose of this study is to evaluate the efficiency of aspirin deprescribing through 2 approaches: direct deprescribing by pharmacists using populationlevel review compared with clinicians following a pharmacist-led education.

Methods

This was a single-center quality improvement cohort study at the Durham VA Health Care System (DVAHCS) in North Carolina. Patients included were aged ≥ 70 years without known ASCVD who received care at any of 3 DVAHCS community-based outpatient clinics and prescribed aspirin. Patient data was obtained using the VIONE (Deprescribing Dashboard called Vital, Important, Optional, Not indicated, and Every medication has a specific indication or diagnosis) dashboard.¹⁵ VIONE was developed to identify potentially inappropriate medications (PIMs) that are eligible to deprescribe based on Beers Criteria, Screening Tool of Older Personsf Prescriptions criteria, and common clinical scenarios when clinicians determine the risk outweighs the benefit to continue a specific medication. ^16,17 VIONE is used to reduce polypharmacy and improve patient safety, comfort, and medication adherence. Aspirin for patients aged ≥ 70 years without a history of ASCVD is a PIM identified by VIONE. Patients aged ≥ 70 years were chosen as an inclusion criteria in this study to match the ASPREE trial inclusion criteria and age inclusion criteria in the VIONE dashboard for aspirin deprescribing.² Patient lists were generated for these potentially inappropriate aspirin prescriptions for 3 months before clinician staff education presentations, the day of the presentations, and 3 months after.

The primary endpoint was the number of veterans with aspirin deprescribed directly by 2 pharmacists over 12 weeks, divided by total patient care time spent, compared with the change in number of veterans with aspirin deprescribed by any DVAHCS physician, nurse practitioner, physician assistant, or CPP over 12 weeks, divided by the total pharmacist time spent on PCP education. Secondary endpoints were the number of aspirin orders discontinued by pharmacists and CPPs, the number of aspirin orders discontinued 12 weeks before pharmacist-led education compared with the number of aspirin orders discontinued 12 weeks after CPP-led education, average and median pharmacist time spent per patient encounter, and time of direct patient encounters vs time spent on PCP education.

Pharmacists reviewed each patient who met the inclusion criteria from the list generated by VIONE on December 1, 2022, for aspirin appropriateness according to the ACC/AHA and USPSTF guidelines, with the goal to discontinue aspirin for primary prevention of ASCVD and no other indications.^1,4 Pharmacists documented their visits using VIONE methodology in the Computerized Patient Record System (CPRS) using a polypharmacy review note. CPPs contacted patients who were taking aspirin for primary prevention by unscheduled telephone call to assess for aspirin adherence, undocumented history of ASCVD, cardiovascular risk factors, and history of bleeding. Aspirin was discontinued if patients met guideline criteria recommendations and agreed to discontinuation. Risk-benefit discussions were completed when patients without known ASCVD were considered high risk because the ACC/AHA guidelines mention there is insufficient evidence of safety and efficacy of aspirin for primary prevention for patients with other known ASCVD risk factors (eg, strong family history of premature myocardial infarction, inability to achieve lipid, blood pressure, or glucose targets, or significant elevation in coronary artery calcium score).

High risk was defined as family history of premature ASCVD (in a male first-degree relative aged < 55 years or a female first-degree relative aged < 65 years), most recent blood pressure or 2 blood pressure results in the last 12 months > 160/100 mm Hg, recent hemoglobin A_1c > 9%, and/or low-density lipoprotein > 190 mg/dL or not prescribed an indicated statin.³ Aspirin was continued or discontinued according to patient preference after the personalized risk-benefit discussion.

For patients with a clinical indication for aspirin use other than ASCVD (eg, atrial fibrillation not on anticoagulation, venous thromboembolism prophylaxis, carotid artery disease), CPPs documented their assessment and when appropriate deferred to the PCP for consideration of stopping aspirin. For patients with undocumented ASCVD, CPPs added their ASCVD history to their problem list and aspirin was continued. PCPs were notified by alert when aspirin was discontinued and when patients could not be reached by telephone.

presented a review of recent guideline updates and supporting literature at 2 online staff meetings. The education sessions lasted about 10 minutes and were presented to PCPs across 3 community-based outpatient clinics. An estimated 40 minutes were spent creating the PowerPoint education materials, seeking feedback, making edits, and answering questions or emails from PCPs after the presentation. During the presentation, pharmacists encouraged PCPs to discontinue aspirin (active VA prescriptions and reported over-the-counter use) for primary prevention of ASCVD in patients aged ≥ 70 years during their upcoming appointments and consider risk factors recommended by the ACC/AHA guidelines when applicable. PCPs were notified that CPPs planned to start a population review for discontinuing active VA aspirin prescriptions on December 1, 2022. The primary endpoint and secondary endpoints were analyzed using descriptive statistics. All data were analyzed using Microsoft Excel.

Results

A total of 868 patients aged ≥ 70 years with active prescriptions for aspirin were identified on December 1, 2022. After applying inclusion and exclusion criteria for the pharmacist population review, 224 patients were included for cohort final analysis (Figure). All 868 patients were eligible for the CPP intervention. Primary reasons for exclusion from the CPP population included over-thecounter aspirin and a history of ASCVD in the patient’s problem list. All patients were male, with a mean (SD) age of 75 (4.4) years (Table 1). Most patients were prescribed aspirin, 81 mg daily (n = 220; 98%).

The direct CPP deprescribing intervention resulted in 2 aspirin prescriptions discontinued per hour of pharmacist time and 67 aspirin prescriptions discontinued per hour of pharmacist time via the PCP education intervention. CPPs discontinued 66 aspirin orders in the 12 weeks before the PCP education sessions. A total of 230 aspirin prescriptions were discontinued in the 12 weeks following the PCP education sessions, with 97 discontinued directly by CPPs and 133 discontinued by PCPs. The PCP education session yielded an additional 67 discontinued aspirin orders compared with the 12 weeks before the education sessions (Table 2).

The CPP direct deprescribing intervention took about 48.3 hours, accounting for health record review and time interacting with patients. The PCP education intervention took about 60 minutes, which included time for preparing and delivering education materials (Table 3). CPP deprescribing encounter types, interventions, and related subcategories, and other identified indications to continue aspirin are listed in Table 4.

Discussion

Compared with direct deprescribing by pharmacists, the PCP education intervention was more efficient based on number of aspirin orders discontinued by pharmacist time. PCPs discontinued twice as many aspirin prescriptions in the 12 weeks after pharmacist-led education compared with the 12 weeks before.

Patients were primarily contacted by telephone (73%) for deprescribing. Among the 163 patients reached by phone and encouraged to discontinue aspirin, 97 patients (60%) accepted the recommendation, which was similar to the acceptance rates found in the literature (48% to 55%).^14,18 Although many veterans continued taking aspirin (78%), most had indications for its continued use, such as a history of ASCVD, atrial fibrillation without anticoagulation, and carotid artery stenosis, and complex comorbidities that required further discussion with their PCP. Less common uses for aspirin were identified through CPRS review or patient reports included cerebral small vessel disease without history of ASCVD, subclavian artery stenosis, thrombocytosis, bioprosthetic valve replacement, giant cell arteritis, rheumatoid arthritis, and prevention of second eye involvement of ischemic optic neuropathy.

to describe the benefit of clinical pharmacy services for deprescribing aspirin for primary prevention of ASCVD through PCP education. Previously published literature has assessed alternative ways to identify or discontinue PIMs—including aspirin—among geriatric patients. One study evaluated the use of marking inappropriate aspirin prescriptions in the electronic health database, leading to a significant reduction in incidence of inappropriate aspirin prescribing; however, it did not assess changes in discontinuation rates of existing aspirin prescriptions.¹⁹ The previous VA pharmacist aspirin deprescribing protocol demonstrated pharmacists’ aptitude at discontinuing aspirin for primary prevention but only used direct patient contact and did not compare efficiency with other methods, including PCP education.¹⁴

This quality improvement project contributes new data to the existing literature to support the use of clinical pharmacists to discontinue aspirin for primary prevention and suggests a strong role for pharmacists as educators on clinical guidelines, in addition to their roles directly deprescribing PIMs in clinical practice. This study is further strengthened by its use of VIONE, which previously has demonstrated effectiveness in deprescribing a variety of PIMs in primary care settings.²⁰

Despite using VIONE for generating a list of patients eligible for deprescription, our CPRS review found that this list was frequently inaccurate. For example, a small portion of patients were on the VIONE generated list indicating they had no ASCVD history, but had transient ischemic attack listed in their problem lists. Patient problem lists often were missing documented ASCVD history that was revealed by patient interview or CPRS review. It is possible that patients interviewed might have omitted relevant ASCVD history because of low health literacy, conditions affecting memory, or use of health care services outside the VA system.

There were several instances of aspirin used for other non-ASCVD indications, such as primary stroke prevention in atrial fibrillation. The ACC/AHA atrial fibrillation guidelines previously provided a Class IIb recommendation (benefit is greater than risk but additional studies are needed) for considering no antithrombic therapy or treatment with oral anticoagulant or aspirin for nonvalvular atrial fibrillation with CHA₂DS₂-VASc (Congestive heart failure, Hypertension, Age [> 65 y, 1 point; > 75 y, 2 points], Diabetes, previous Stroke/transient ischemic attack [2 points]) score of 1.²¹ The ACC/ AHA guidelines were updated in 2023 to recommend against antiplatelet therapy as an alternative to anticoagulation for reducing cardioembolic stroke risk among patients with atrial fibrillation with no indication for antiplatelet therapy because of risk of harm.²² If a patient has no risk factors for stroke, aspirin is not recommended to prevent thromboembolic events because of a lack of benefit. Interventions from this quality improvement study were completed before the 2023 atrial fibrillation guideline was published and therefore in this study aspirin was not discontinued when used for atrial fibrillation. Aspirin use for atrial fibrillation might benefit from similar discontinuation efforts analyzed within this study. Beyond atrial fibrillation, major guidelines do not comment on the use of aspirin for any other indications in the absence of clinical ASCVD.

Limitations

This study is limited by the lack of clinical consensus for complex patients and demonstrates the importance of individualized patient assessment when considering discontinuing aspirin. Because of the project’s relatively short intervention period, aspirin deprescribing rates could decrease over time and repeated education efforts might be necessary to see lasting impact. Health care professionals from services outside of primary care also might have discontinued aspirin during the study period unrelated to the education and these discontinued aspirin prescriptions could contribute to the higher rate observed among PCPs. This study included a specific population cohort of male, US veterans and might not reflect other populations where these interventions could be implemented.

The measurement of time spent by pharmacists and PCPs is an additional limitation. Although it is expected that PCPs attempt to discontinue aspirin during their existing patient care appointments, the time spent during visits was not measured or documented. Direct deprescribing by pharmacist CPRS review required a significant amount of time and could be a barrier to successful intervention by CPPs in patient aligned care teams.

To reduce the time pharmacists spent completing CPRS reviews, an aspirin deprescribing clinical reminder tool could be used to assess use and appropriate indication quickly during any primary care visit led by a PCP or CPP. In addition, it is recommended that pharmacists regularly educate health care professionals on guideline recommendations for aspirin use among geriatric patients. Future studies of the incidence of major cardiovascular events after aspirin deprescribing among geriatric patients and a longitudinal cost/benefit analysis could support these initiatives.

Conclusions

In this study, pharmacists successfully deprescribed inappropriate medications, such as aspirin. However, pharmacist-led PCP education is more efficient compared with direct deprescribing using a population-level review. PCP education requires less time and could allow ambulatory care pharmacists to spend more time on other direct patient care interventions to improve quality and access to care in primary care clinics. This study’s results further support the role of pharmacists in deprescribing PIMs for older adults and the use of a deprescribing tool, such as VIONE, in a primary care setting.

Low-dose aspirin commonly is used for the prevention of cardiovascular disease (CVD) but is associated with an increased risk of major bleeding.¹ The use of aspirin for primary prevention is largely extrapolated from clinical trials showing benefit in the secondary prevention of myocardial infarction and ischemic stroke. However, results from the Aspirin in Reducing Events in the Elderly (ASPREE) trial challenged this practice.² The ASPREE trial, conducted in the United States and Australia from 2010 to 2014, sought to determine whether daily 100 mg aspirin, was superior to placebo in promoting disability-free survival among older adults. Participants were aged ≥ 70 years (≥ 65 years for Hispanic and Black US participants), living in the community, and were free from preexisting CVD, cerebrovascular disease, or any chronic condition likely to limit survival to < 5 years. The study found no significant difference in the primary endpoints of death, dementia, or persistent physical disability, but there was a significantly higher risk of major hemorrhage in the aspirin group (3.8% vs 2.8%; hazard ratio, 1.38; 95% CI, 1.18-1.62; P < .001).

Several medical societies have updated their guideline recommendations for aspirin for primary prevention of CVD. The 2022 United States Public Service Task Force (USPSTF) provides a grade C recommendation (at least moderate certainty that the net benefit is small) to consider low-dose aspirin for the primary prevention of CVD on an individual patient basis for adults aged 40 to 59 years who have a ≥ 10% 10-year CVD risk. For adults aged ≥ 60 years, the USPSTF recommendation is grade D (moderate or high certainty that the practice has no net benefit or that harms outweigh the benefits) for low-dose aspirin use.^1,3 The American College of Cardiology and American Heart Association (ACC/AHA) recommend considering low-dose aspirin for primary prevention of atherosclerotic cardiovascular disease (ASCVD) among select adults aged 40 to 70 years at higher CVD risk but not at increased risk of bleeding.⁴ The American Diabetes Association (ADA) recommends low-dose aspirin for primary prevention of CVD in patients with diabetes and additional risk factors such as family history of premature ASCVD, hypertension, dyslipidemia, smoking, or chronic kidney disease, and who are not at higher risk of bleeding.⁵ The ADA standards also caution against the use of aspirin as primary prevention in patients aged > 70 years. Low-dose aspirin use is not recommended for the primary prevention of CVD in older adults or adults of any age who are at increased risk of bleeding.

Recent literature using the US Department of Veterans Affairs (VA) Corporate Data Warehouse database confirms 86,555 of 1.8 million veterans aged > 70 years (5%) were taking low-dose aspirin for primary prevention of ASCVD despite guideline recommendations.⁶ Higher risk of gastrointestinal and other major bleeding from low-dose aspirin has been reported in the literature.¹ Major bleeds represent a significant burden to the health care system with an estimated mean $13,093 cost for gastrointestinal bleed hospitalization.⁷

Considering the large scale aspirin use without appropriate indication within the veteran population, the risk of adverse effects, and the significant cost to patients and the health care system, it is imperative to determine the best approach to efficiently deprescribe aspirin for primary prevention among geriatric patients. Deprescribing refers to the systematic and supervised process of dose reduction or drug discontinuation with the goal of improving health and/or reducing the risk of adverse effects.⁸ During patient visits, primary care practitioners (PCPs) have opportunities to discontinue aspirin, but these encounters are time-limited and deprescribing might be secondary to more acute primary care needs. The shortage of PCPs is expected to worsen in coming years, which could further reduce their availability to assess inappropriate aspirin use.⁹

VA clinical pharmacist practitioners (CPPs) serve as medication experts and work autonomously under a broad scope of practice as part of the patient aligned care team.^10-12 CPPs can free up time for PCPs and facilitate deprescribing efforts, especially for older adults. One retrospective cohort study conducted at a VA medical center found that CPPs deprescribed more potentially inappropriate medications among individuals aged ≥ 80 years compared with usual care with PCPs (26.8% vs 16.1%; P < .001).^12,13 An aspirin deprescribing protocol conducted in 2022 resulted in nearly half of veterans aged ≥ 70 years contacted by phone agreeing to stop aspirin. Although this study supports the role pharmacists can play in reducing aspirin use in accordance with guidelines, the authors acknowledge that their interventions had a mean time of 12 minutes per patient and would require workflow changes.¹⁴ The purpose of this study is to evaluate the efficiency of aspirin deprescribing through 2 approaches: direct deprescribing by pharmacists using populationlevel review compared with clinicians following a pharmacist-led education.

Methods

This was a single-center quality improvement cohort study at the Durham VA Health Care System (DVAHCS) in North Carolina. Patients included were aged ≥ 70 years without known ASCVD who received care at any of 3 DVAHCS community-based outpatient clinics and prescribed aspirin. Patient data was obtained using the VIONE (Deprescribing Dashboard called Vital, Important, Optional, Not indicated, and Every medication has a specific indication or diagnosis) dashboard.¹⁵ VIONE was developed to identify potentially inappropriate medications (PIMs) that are eligible to deprescribe based on Beers Criteria, Screening Tool of Older Personsf Prescriptions criteria, and common clinical scenarios when clinicians determine the risk outweighs the benefit to continue a specific medication. ^16,17 VIONE is used to reduce polypharmacy and improve patient safety, comfort, and medication adherence. Aspirin for patients aged ≥ 70 years without a history of ASCVD is a PIM identified by VIONE. Patients aged ≥ 70 years were chosen as an inclusion criteria in this study to match the ASPREE trial inclusion criteria and age inclusion criteria in the VIONE dashboard for aspirin deprescribing.² Patient lists were generated for these potentially inappropriate aspirin prescriptions for 3 months before clinician staff education presentations, the day of the presentations, and 3 months after.

The primary endpoint was the number of veterans with aspirin deprescribed directly by 2 pharmacists over 12 weeks, divided by total patient care time spent, compared with the change in number of veterans with aspirin deprescribed by any DVAHCS physician, nurse practitioner, physician assistant, or CPP over 12 weeks, divided by the total pharmacist time spent on PCP education. Secondary endpoints were the number of aspirin orders discontinued by pharmacists and CPPs, the number of aspirin orders discontinued 12 weeks before pharmacist-led education compared with the number of aspirin orders discontinued 12 weeks after CPP-led education, average and median pharmacist time spent per patient encounter, and time of direct patient encounters vs time spent on PCP education.

Pharmacists reviewed each patient who met the inclusion criteria from the list generated by VIONE on December 1, 2022, for aspirin appropriateness according to the ACC/AHA and USPSTF guidelines, with the goal to discontinue aspirin for primary prevention of ASCVD and no other indications.^1,4 Pharmacists documented their visits using VIONE methodology in the Computerized Patient Record System (CPRS) using a polypharmacy review note. CPPs contacted patients who were taking aspirin for primary prevention by unscheduled telephone call to assess for aspirin adherence, undocumented history of ASCVD, cardiovascular risk factors, and history of bleeding. Aspirin was discontinued if patients met guideline criteria recommendations and agreed to discontinuation. Risk-benefit discussions were completed when patients without known ASCVD were considered high risk because the ACC/AHA guidelines mention there is insufficient evidence of safety and efficacy of aspirin for primary prevention for patients with other known ASCVD risk factors (eg, strong family history of premature myocardial infarction, inability to achieve lipid, blood pressure, or glucose targets, or significant elevation in coronary artery calcium score).

High risk was defined as family history of premature ASCVD (in a male first-degree relative aged < 55 years or a female first-degree relative aged < 65 years), most recent blood pressure or 2 blood pressure results in the last 12 months > 160/100 mm Hg, recent hemoglobin A_1c > 9%, and/or low-density lipoprotein > 190 mg/dL or not prescribed an indicated statin.³ Aspirin was continued or discontinued according to patient preference after the personalized risk-benefit discussion.

For patients with a clinical indication for aspirin use other than ASCVD (eg, atrial fibrillation not on anticoagulation, venous thromboembolism prophylaxis, carotid artery disease), CPPs documented their assessment and when appropriate deferred to the PCP for consideration of stopping aspirin. For patients with undocumented ASCVD, CPPs added their ASCVD history to their problem list and aspirin was continued. PCPs were notified by alert when aspirin was discontinued and when patients could not be reached by telephone.

presented a review of recent guideline updates and supporting literature at 2 online staff meetings. The education sessions lasted about 10 minutes and were presented to PCPs across 3 community-based outpatient clinics. An estimated 40 minutes were spent creating the PowerPoint education materials, seeking feedback, making edits, and answering questions or emails from PCPs after the presentation. During the presentation, pharmacists encouraged PCPs to discontinue aspirin (active VA prescriptions and reported over-the-counter use) for primary prevention of ASCVD in patients aged ≥ 70 years during their upcoming appointments and consider risk factors recommended by the ACC/AHA guidelines when applicable. PCPs were notified that CPPs planned to start a population review for discontinuing active VA aspirin prescriptions on December 1, 2022. The primary endpoint and secondary endpoints were analyzed using descriptive statistics. All data were analyzed using Microsoft Excel.

Results

A total of 868 patients aged ≥ 70 years with active prescriptions for aspirin were identified on December 1, 2022. After applying inclusion and exclusion criteria for the pharmacist population review, 224 patients were included for cohort final analysis (Figure). All 868 patients were eligible for the CPP intervention. Primary reasons for exclusion from the CPP population included over-thecounter aspirin and a history of ASCVD in the patient’s problem list. All patients were male, with a mean (SD) age of 75 (4.4) years (Table 1). Most patients were prescribed aspirin, 81 mg daily (n = 220; 98%).

The direct CPP deprescribing intervention resulted in 2 aspirin prescriptions discontinued per hour of pharmacist time and 67 aspirin prescriptions discontinued per hour of pharmacist time via the PCP education intervention. CPPs discontinued 66 aspirin orders in the 12 weeks before the PCP education sessions. A total of 230 aspirin prescriptions were discontinued in the 12 weeks following the PCP education sessions, with 97 discontinued directly by CPPs and 133 discontinued by PCPs. The PCP education session yielded an additional 67 discontinued aspirin orders compared with the 12 weeks before the education sessions (Table 2).

The CPP direct deprescribing intervention took about 48.3 hours, accounting for health record review and time interacting with patients. The PCP education intervention took about 60 minutes, which included time for preparing and delivering education materials (Table 3). CPP deprescribing encounter types, interventions, and related subcategories, and other identified indications to continue aspirin are listed in Table 4.

Discussion

Compared with direct deprescribing by pharmacists, the PCP education intervention was more efficient based on number of aspirin orders discontinued by pharmacist time. PCPs discontinued twice as many aspirin prescriptions in the 12 weeks after pharmacist-led education compared with the 12 weeks before.

Patients were primarily contacted by telephone (73%) for deprescribing. Among the 163 patients reached by phone and encouraged to discontinue aspirin, 97 patients (60%) accepted the recommendation, which was similar to the acceptance rates found in the literature (48% to 55%).^14,18 Although many veterans continued taking aspirin (78%), most had indications for its continued use, such as a history of ASCVD, atrial fibrillation without anticoagulation, and carotid artery stenosis, and complex comorbidities that required further discussion with their PCP. Less common uses for aspirin were identified through CPRS review or patient reports included cerebral small vessel disease without history of ASCVD, subclavian artery stenosis, thrombocytosis, bioprosthetic valve replacement, giant cell arteritis, rheumatoid arthritis, and prevention of second eye involvement of ischemic optic neuropathy.

to describe the benefit of clinical pharmacy services for deprescribing aspirin for primary prevention of ASCVD through PCP education. Previously published literature has assessed alternative ways to identify or discontinue PIMs—including aspirin—among geriatric patients. One study evaluated the use of marking inappropriate aspirin prescriptions in the electronic health database, leading to a significant reduction in incidence of inappropriate aspirin prescribing; however, it did not assess changes in discontinuation rates of existing aspirin prescriptions.¹⁹ The previous VA pharmacist aspirin deprescribing protocol demonstrated pharmacists’ aptitude at discontinuing aspirin for primary prevention but only used direct patient contact and did not compare efficiency with other methods, including PCP education.¹⁴

This quality improvement project contributes new data to the existing literature to support the use of clinical pharmacists to discontinue aspirin for primary prevention and suggests a strong role for pharmacists as educators on clinical guidelines, in addition to their roles directly deprescribing PIMs in clinical practice. This study is further strengthened by its use of VIONE, which previously has demonstrated effectiveness in deprescribing a variety of PIMs in primary care settings.²⁰

Despite using VIONE for generating a list of patients eligible for deprescription, our CPRS review found that this list was frequently inaccurate. For example, a small portion of patients were on the VIONE generated list indicating they had no ASCVD history, but had transient ischemic attack listed in their problem lists. Patient problem lists often were missing documented ASCVD history that was revealed by patient interview or CPRS review. It is possible that patients interviewed might have omitted relevant ASCVD history because of low health literacy, conditions affecting memory, or use of health care services outside the VA system.

There were several instances of aspirin used for other non-ASCVD indications, such as primary stroke prevention in atrial fibrillation. The ACC/AHA atrial fibrillation guidelines previously provided a Class IIb recommendation (benefit is greater than risk but additional studies are needed) for considering no antithrombic therapy or treatment with oral anticoagulant or aspirin for nonvalvular atrial fibrillation with CHA₂DS₂-VASc (Congestive heart failure, Hypertension, Age [> 65 y, 1 point; > 75 y, 2 points], Diabetes, previous Stroke/transient ischemic attack [2 points]) score of 1.²¹ The ACC/ AHA guidelines were updated in 2023 to recommend against antiplatelet therapy as an alternative to anticoagulation for reducing cardioembolic stroke risk among patients with atrial fibrillation with no indication for antiplatelet therapy because of risk of harm.²² If a patient has no risk factors for stroke, aspirin is not recommended to prevent thromboembolic events because of a lack of benefit. Interventions from this quality improvement study were completed before the 2023 atrial fibrillation guideline was published and therefore in this study aspirin was not discontinued when used for atrial fibrillation. Aspirin use for atrial fibrillation might benefit from similar discontinuation efforts analyzed within this study. Beyond atrial fibrillation, major guidelines do not comment on the use of aspirin for any other indications in the absence of clinical ASCVD.

Limitations

This study is limited by the lack of clinical consensus for complex patients and demonstrates the importance of individualized patient assessment when considering discontinuing aspirin. Because of the project’s relatively short intervention period, aspirin deprescribing rates could decrease over time and repeated education efforts might be necessary to see lasting impact. Health care professionals from services outside of primary care also might have discontinued aspirin during the study period unrelated to the education and these discontinued aspirin prescriptions could contribute to the higher rate observed among PCPs. This study included a specific population cohort of male, US veterans and might not reflect other populations where these interventions could be implemented.

The measurement of time spent by pharmacists and PCPs is an additional limitation. Although it is expected that PCPs attempt to discontinue aspirin during their existing patient care appointments, the time spent during visits was not measured or documented. Direct deprescribing by pharmacist CPRS review required a significant amount of time and could be a barrier to successful intervention by CPPs in patient aligned care teams.

To reduce the time pharmacists spent completing CPRS reviews, an aspirin deprescribing clinical reminder tool could be used to assess use and appropriate indication quickly during any primary care visit led by a PCP or CPP. In addition, it is recommended that pharmacists regularly educate health care professionals on guideline recommendations for aspirin use among geriatric patients. Future studies of the incidence of major cardiovascular events after aspirin deprescribing among geriatric patients and a longitudinal cost/benefit analysis could support these initiatives.

Conclusions

In this study, pharmacists successfully deprescribed inappropriate medications, such as aspirin. However, pharmacist-led PCP education is more efficient compared with direct deprescribing using a population-level review. PCP education requires less time and could allow ambulatory care pharmacists to spend more time on other direct patient care interventions to improve quality and access to care in primary care clinics. This study’s results further support the role of pharmacists in deprescribing PIMs for older adults and the use of a deprescribing tool, such as VIONE, in a primary care setting.

Clinical practice has shifted from vitamin K antagonists to direct oral anticoagulants (DOACs) for atrial fibrillation treatment due to their more favorable risk-benefit profile and less lifestyle modification required.^1,2 However, the advantage of a lower bleeding risk with DOACs could be compromised by potentially problematic pharmacokinetic interactions like those conferred by antiplatelets or nonsteroidal anti-inflammatory drugs (NSAIDs).^3,4 Treating a patient needing anticoagulation with a DOAC who has comorbidities may introduce unavoidable drug-drug interactions. This particularly happens with over-the-counter and prescription NSAIDs used for the management of pain and inflammatory conditions.⁵

NSAIDs primarily affect 2 cyclooxygenase (COX) enzyme isomers, COX-1 and COX-2.⁶ COX-1 helps maintain gastrointestinal (GI) mucosa integrity and platelet aggregation processes, whereas COX-2 is engaged in pain signaling and inflammation mediation. COX-1 inhibition is associated with more bleeding-related adverse events (AEs), especially in the GI tract. COX-2 inhibition is thought to provide analgesia and anti-inflammatory properties without elevating bleeding risk. This premise is responsible for the preferential use of celecoxib, a COX-2 selective NSAID, which should confer a lower bleeding risk compared to nonselective NSAIDs such as ibuprofen and naproxen.⁷ NSAIDs have been documented as independent risk factors for bleeding. NSAID users are about 3 times as likely to develop GI AEs compared to nonNSAID users.⁸

Many clinicians aim to further mitigate NSAID-associated bleeding risk by coprescribing a proton pump inhibitor (PPI). PPIs provide gastroprotection against NSAID-induced mucosal injury and sequential complication of GI bleeding. In a multicenter randomized control trial, patients who received concomitant PPI therapy while undergoing chronic NSAID therapy—including nonselective and COX-2 selective NSAIDs—had a significantly lower risk of GI ulcer development (placebo, 17.0%; 20 mg esomeprazole, 5.2%; 40 mg esomeprazole, 4.6%).⁹ Current clinical guidelines for preventing NSAIDassociated bleeding complications recommend using a COX-2 selective NSAID in combination with PPI therapy for patients at high risk for GI-related bleeding, including the concomitant use of anticoagulants.¹⁰

There is evidence suggesting an increased bleeding risk with NSAIDs when used in combination with vitamin K antagonists such as warfarin.^11,12 A systematic review of warfarin and concomitant NSAID use found an increased risk of overall bleeding with NSAID use in combination with warfarin (odds ratio 1.58; 95% CI, 1.18-2.12), compared to warfarin alone.¹²

Posthoc analyses of randomized clinical trials have also demonstrated an increased bleeding risk with oral anticoagulation and concomitant NSAID use.^13,14 In the RE-LY trial, NSAID users on warfarin or dabigatran had a statistically significant increased risk of major bleeding compared to non-NSAID users (hazard ratio [HR] 1.68; 95% CI, 1.40- 2.02; P < .001).¹³ In the ARISTOTLE trial, patients on warfarin or apixaban who were incident NSAID users were found to have an increased risk of major bleeding (HR 1.61; 95% CI, 1.11-2.33) and clinically relevant nonmajor bleeding (HR 1.70; 95% CI, 1.16- 2.48).¹⁴ These trials found a statistically significant increased bleeding risk associated with NSAID use, though the populations evaluated included patients taking warfarin and patients taking DOACs. These trials did not evaluate the bleeding risk of concomitant NSAID use among DOACs alone.

Evidence on NSAID-associated bleeding risk with DOACs is lacking in settings where the patient population, prescribing practices, and monitoring levels are variable. Within the Veterans Health Administration, clinical pharmacist practitioners (CPPs) in anticoagulation clinics oversee DOAC therapy management. CPPs monitor safety and efficacy of DOAC therapies through a population health management tool, the DOAC Dashboard.¹⁵ The DOAC Dashboard creates alerts for patients who may require an intervention based on certain clinical parameters, such as drug-drug interactions.¹⁶ Whenever a patient on a DOAC is prescribed an NSAID, an alert is generated on the DOAC Dashboard to flag the CPPs for the potential need for an intervention. If NSAID therapy remains clinically indicated, CPPs may recommend risk reduction strategies such as a COX-2 selective NSAID or coprescribing a PPI.¹⁰

The DOAC Dashboard provides an ideal setting for investigating the effects of NSAID use, NSAID selectivity, and PPI coprescribing on DOAC bleeding rates. With an increasing population of patients receiving anticoagulation therapy with a DOAC, more guidance regarding the bleeding risk of concomitant NSAID use with DOACs is needed. Studies evaluating the bleeding risk with concomitant NSAID use in patients on a DOAC alone are limited. This is the first study to date to compare bleeding risk with concomitant NSAID use between DOACs. This study provides information on bleeding risk with NSAID use among commonly prescribed DOACs, rivaroxaban and apixaban, and the potential impacts of current risk reduction strategies.

METHODS

This single-center retrospective cohort review was performed using the electronic health records (EHRs) of patients enrolled in the US Department of Veterans Affairs (VA) Mountain Home Healthcare System who received rivaroxaban or apixaban from December 2020 to December 2022. This study received approval from the East Tennessee State University/VA Institutional Review Board committee.

Patients were identified through the DOAC Dashboard, aged 21 to 100 years, and received rivaroxaban or apixaban at a therapeutic dose: rivaroxaban 10 to 20 mg daily or apixaban 2.5 to 5 mg twice daily. Patients were excluded if they were prescribed dual antiplatelet therapy, received rivaroxaban at dosing indicated for peripheral vascular disease, were undergoing dialysis, had evidence of moderate to severe hepatic impairment or any hepatic disease with coagulopathy, were undergoing chemotherapy or radiation, or had hematological conditions with predisposed bleeding risk. These patients were excluded to mitigate the potential confounding impact from nontherapeutic DOAC dosing strategies and conditions associated with an increased bleeding risk.

Eligible patients were stratified based on NSAID use. NSAID users were defined as patients prescribed an oral NSAID, including both acute and chronic courses, at any point during the study time frame while actively on a DOAC. Bleeding events were reviewed to evaluate rates between rivaroxaban and apixaban among NSAID and nonNSAID users. Identified NSAID users were further assessed for NSAID selectivity and PPI coprescribing as a subgroup analysis for the secondary assessment.

Data Collection

Baseline data were collected, including age, body mass index, anticoagulation indication, DOAC agent, DOAC dose, and DOAC total daily dose. Baseline serum creatinine levels, liver function tests, hemoglobin levels, and platelet counts were collected from the most recent data available immediately prior to the bleeding event, if applicable.

The DOAC Dashboard was reviewed for active and dismissed drug interaction alerts to identify patients taking rivaroxaban or apixaban who were prescribed an NSAID. Patients were categorized in the NSAID group if an interacting drug alert with an NSAID was reported during the study time frame. Data available through the interacting drug alerts on NSAID use were limited to the interacting drug name and date of the reported flag. Manual EHR review was required to confirm dates of NSAID therapy initiation and NSAID discontinuation, if applicable.

Data regarding concomitant antiplatelet use were obtained through review of the active and dismissed drug interaction alerts on the DOAC Dashboard. Concomitant antiplatelet use was defined as the prescribing of a single antiplatelet agent at any point while receiving DOAC therapy. Data on concomitant antiplatelets were collected regardless of NSAID status.

Data on coprescribed PPI therapy were obtained through manual EHR review of identified NSAID users. Coprescribed PPI therapy was defined as the prescribing of a PPI at any point during NSAID therapy. Data regarding PPI use among non-NSAID users were not collected because the secondary endpoint was designed to assess PPI use only among patients coprescribed a DOAC and NSAID.

Outcomes

Bleeding events were identified through an outcomes report generated by the DOAC Dashboard based on International Classification of Diseases, Tenth Revision diagnosis codes associated with a bleeding event. The outcomes report captures diagnoses from the outpatient and inpatient care settings. Reported bleeding events were limited to patients who received a DOAC at any point in the 6 months prior to the event and excluded patients with recent DOAC initiation within 7 days of the event, as these patients are not captured on the DOAC Dashboard.

All reported bleeding events were manually reviewed in the EHR and categorized as a major or clinically relevant nonmajor bleed, according to International Society of Thrombosis and Haemostasis criteria. Validated bleeding events were then crossreferenced with the interacting drug alerts report to identify events with potentially overlapping NSAID therapy at the time of the event. Overlapping NSAID therapy was defined as the prescribing of an NSAID at any point in the 6 months prior to the event. All events with potential overlapping NSAID therapies were manually reviewed for confirmation of NSAID status at the time of the event.

The primary endpoint was a composite of any bleeding event per International Society of Thrombosis and Haemostasis criteria. The secondary endpoint evaluated the potential impact of NSAID selectivity or PPI coprescribing on the bleeding rate among the NSAID user groups.

Statistical Analysis

Analyses were performed consistent with the methods used in the ARISTOTLE and RE-LY trials. It was determined that a sample size of 504 patients, with ≥ 168 patients in each group, would provide 80% power using a 2-sided a of 0.05. HRs with 95% CIs and respective P values were calculated using a SPSS-adapted online calculator.

RESULTS

The DOAC Dashboard identified 681 patients on rivaroxaban and 3225 patients on apixaban; 72 patients on rivaroxaban (10.6%) and 300 patients on apixaban (9.3%) were NSAID users. The mean age of NSAID users was 66.9 years in the rivaroxaban group and 72.4 years in the apixaban group. The mean age of non-NSAID users was 71.5 years in the rivaroxaban group and 75.6 years in the apixaban group. No appreciable differences were observed among subgroups in body mass index, renal function, hepatic function, hemoglobin, or platelet counts, and no statistically significant differences were identified (Table 1). Antiplatelet agents identified included aspirin, clopidogrel, prasugrel, and ticagrelor. Fifteen patients (20.3%) in the rivaroxaban group and 87 patients (28.7%) in the apixaban group had concomitant antiplatelet and NSAID use. Forty-five patients on rivaroxaban (60.8%) and 170 (55.9%) on apixaban were prescribed concomitant PPI and NSAID at baseline. Among non-NSAID users, there was concomitant antiplatelet use for 265 patients (43.6%) in the rivaroxaban group and 1401 patients (47.9%) in the apixaban group. Concomitant PPI use was identified among 63 patients (60.0%) taking selective NSAIDs and 182 (57.2%) taking nonselective NSAIDs.

A total of 423 courses of NSAIDs were identified: 85 NSAID courses in the rivaroxaban group and 338 NSAID courses in the apixaban group. Most NSAID courses involved a nonselective NSAID in the rivaroxaban and apixaban NSAID user groups: 75.2% (n = 318) aggregately compared to 71.8% (n = 61) and 76.0% (n = 257) in the rivaroxaban and apixaban groups, respectively. The most frequent NSAID courses identified were meloxicam (26.7%; n = 113), celecoxib (24.8%; n = 105), ibuprofen (19.1%; n = 81), and naproxen (13.5%; n = 57). Data regarding NSAID therapy initiation and discontinuation dates were not readily available. As a result, the duration of NSAID courses was not captured.

There was no statistically significant difference in bleeding rates between rivaroxaban and apixaban among NSAID users (HR 1.04; 95% CI, 0.98-1.12) or non-NSAID users (HR 1.15; 95% CI, 0.80-1.66) (Table 2). Apixaban non-NSAID users had a higher rate of major bleeds (HR 0.32; 95% CI, 0.17-0.61) while rivaroxaban non-NSAID users had a higher rate of clinically relevant nonmajor bleeds (HR 1.63; 95% CI, 1.10-2.54).

The sample size for the secondary endpoint consisted of bleeding events that were confirmed to have had an overlapping NSAID prescribed at the time of the event. For this secondary assessment, there was 1 rivaroxaban NSAID user bleeding event and 4 apixaban NSAID user bleeding events. For the rivaroxaban NSAID user bleeding event, the NSAID was nonselective and a PPI was not coprescribed. For the apixaban NSAID user bleeding events, 2 NSAIDs were nonselective and 2 were selective. All patients with apixaban and NSAID bleeding events had a coprescribed PPI. There was no clinically significant difference in the bleeding rates observed for NSAID selectivity or PPI coprescribing among the NSAID user subgroups.

DISCUSSION

This study found that there was no statistically significant difference for bleeding rates of major and nonmajor bleeding events between rivaroxaban and apixaban among NSAID users and non-NSAID users. This study did not identify a clinically significant impact on bleeding rates from NSAID selectivity or PPI coprescribing among the NSAID users.

There were notable but not statistically significant differences in baseline characteristics observed between the NSAID and non-NSAID user groups. On average, the rivaroxaban and apixaban NSAID users were younger compared with those not taking NSAIDs. NSAIDs, specifically nonselective NSAIDs, are recognized as potentially inappropriate medications for older adults given that this population is at an increased risk for GI ulcer development and/or GI bleeding.¹⁷ The non-NSAID user group likely consisted of older patients compared to the NSAID user group as clinicians may avoid prescribing NSAIDs to older adults regardless of concomitant DOAC therapy.

In addition to having an older patient population, non-NSAID users were more frequently prescribed a concomitant antiplatelet when compared with NSAID users. This prescribing pattern may be due to clinicians avoiding the use of NSAIDs in patients receiving DOAC therapy in combination with antiplatelet therapy, as these patients have been found to have an increased bleeding rate compared to DOAC therapy alone.¹⁸

Non-NSAID users had an overall higher bleeding rate for both major and nonmajor bleeding events. Based on this observation, it could be hypothesized that antiplatelet agents have a higher risk of bleeding in comparison to NSAIDs. In a subanalysis of the EXPAND study evaluating risk factors of major bleeding in patients receiving rivaroxaban, concomitant use of antiplatelet agents demonstrated a statistically significant increased risk of bleeding (HR 1.6; 95% CI, 1.2-2.3; P = .003) while concomitant use of NSAIDs did not (HR 0.8; 95% CI, 0.3-2.2; P = .67).¹⁹

In assessing PPI status at baseline, a majority of both rivaroxaban and apixaban NSAID users were coprescribed a PPI. This trend aligns with current clinical guideline recommendations for the prescribing of PPI therapy for GI protection in high-risk patients, such as those on DOAC therapy and concomitant NSAID therapy.¹⁰ Given the high proportion of NSAID users coprescribed a PPI at baseline, it may be possible that the true incidence of NSAID-associated bleeding events was higher than what this study found. This observation may reflect the impact from timely implementation of risk mitigation strategies by CPPs in the anticoagulation clinic. However, this study was not constructed to assess the efficacy of PPI use in this manner.

It is important to note the patients included in this study were followed by a pharmacist in an anticoagulation clinic using the DOAC Dashboard.¹⁵ This population management tool allows CPPs to make proactive interventions when a patient taking a DOAC receives an NSAID prescription, such as recommending the coprescribing of a PPI or use of a selective NSAID.^10,16 These standards of care may have contributed to an overall reduced bleeding rate among the NSAID user group and may not be reflective of private practice.

The planned analysis of this study was modeled after the posthoc analysis of the RE-LY and ARISTOTLE trials. Both trials demonstrated an increased risk of bleeding with oral anticoagulation, including DOAC and warfarin, in combination with NSAID use. However, both trials found that NSAID use in patients treated with a DOAC was not independently associated with increased bleeding events compared with warfarin.^13,14 The results of this study are comparable to the RE-LY and ARISTOTLE findings that NSAID use among patients treated with rivaroxaban or apixaban did not demonstrate a statistically significant increased bleeding risk.

Studies of NSAID use in combination with DOAC therapy have been limited to patient populations consisting of both DOAC and warfarin. Evidence from these trials outlines the increased bleeding risk associated with NSAID use in combination with oral anticoagulation; however, these patient populations include those on a DOAC and warfarin.^13,14,19,20 Given the limited evidence on NSAID use among DOACs alone, it is assumed NSAID use in combination with DOACs has a similar risk of bleeding as warfarin use. This may cause clinicians to automatically exclude NSAID therapy as a treatment option for patients on a DOAC who are otherwise clinically appropriate candidates, such as those with underlying inflammatory conditions. Avoiding NSAID therapy in this patient population may lead to suboptimal pain management and increase the risk of patient harm from methods such as inappropriate opioid therapy prescribing.

DOAC therapy should not be a universal limitation to the use of NSAIDs. Although the risk of bleeding with NSAID therapy is always present, deliberate NSAID prescribing in addition to the timely implementation of risk mitigation strategies may provide an avenue for safe NSAID prescribing in patients receiving a DOAC. A population health-based approach to DOAC management, such as the DOAC Dashboard, appears to be effective at preventing patient harm when NSAIDs are prescribed in conjunction with DOACs.

Limitations

The DOAC Dashboard has been shown to be effective and efficient at monitoring DOAC therapy from a population-based approach.¹⁶ Reports generated through the DOAC Dashboard provide convenient access to patient data which allows for timely interventions; however, there are limits to its use for data collection. All the data elements necessary to properly assess bleeding risk with validated tools, such as HAS-BLED (hypertension, abnormal renal/liver function, stroke, bleeding history or predisposition, labile international normalized ratio, elderly, drugs/ alcohol concomitantly), are not available on DOAC Dashboard reports. Due to this constraint, bleeding risk assessments were not conducted at baseline and this study was unable to include risk modeling. Additionally, data elements like initiation and discontinuation dates and duration of therapies were not readily available. As a result, this study was unable to incorporate time as a data point.

This was a retrospective study that relied on manual review of chart documentation to verify bleeding events, but data obtained through the DOAC Dashboard were transferred directly from the EHR.¹⁵ Bleeding events available for evaluation were restricted to those that occurred at a VA facility. Additionally, the sample size within the rivaroxaban NSAID user group did not reach the predefined sample size required to reach power and may have been too small to detect a difference if one did exist. The secondary assessment had a low sample size of NSAID user bleeding events, making it difficult to fully assess its impact on NSAID selectivity and PPI coprescribing on bleeding rates. All courses of NSAIDs were equally valued regardless of the dose or therapy duration; however, this is consistent with how NSAID use was defined in the RE-LY and ARISTOTLE trials.

CONCLUSIONS

This retrospective cohort review found no statistically significant difference in the composite bleeding rates between rivaroxaban and apixaban among NSAID users and non-NSAID users. Moreover, there was no clinically significant impact observed for bleeding rates in regard to NSAID selectivity and PPI coprescribing among NSAID users. However, coprescribing of PPI therapy to patients on a DOAC who are clinically indicated for an NSAID may reduce the risk of bleeding. Population health management tools, such as the DOAC Dashboard, may also allow clinicians to safely prescribe NSAIDs to patients on a DOAC. Further large-scale observational studies are needed to quantify the real-world risk of bleeding with concomitant NSAID use among DOACs alone and to evaluate the impact from NSAID selectivity or PPI coprescribing.

Clinical practice has shifted from vitamin K antagonists to direct oral anticoagulants (DOACs) for atrial fibrillation treatment due to their more favorable risk-benefit profile and less lifestyle modification required.^1,2 However, the advantage of a lower bleeding risk with DOACs could be compromised by potentially problematic pharmacokinetic interactions like those conferred by antiplatelets or nonsteroidal anti-inflammatory drugs (NSAIDs).^3,4 Treating a patient needing anticoagulation with a DOAC who has comorbidities may introduce unavoidable drug-drug interactions. This particularly happens with over-the-counter and prescription NSAIDs used for the management of pain and inflammatory conditions.⁵

NSAIDs primarily affect 2 cyclooxygenase (COX) enzyme isomers, COX-1 and COX-2.⁶ COX-1 helps maintain gastrointestinal (GI) mucosa integrity and platelet aggregation processes, whereas COX-2 is engaged in pain signaling and inflammation mediation. COX-1 inhibition is associated with more bleeding-related adverse events (AEs), especially in the GI tract. COX-2 inhibition is thought to provide analgesia and anti-inflammatory properties without elevating bleeding risk. This premise is responsible for the preferential use of celecoxib, a COX-2 selective NSAID, which should confer a lower bleeding risk compared to nonselective NSAIDs such as ibuprofen and naproxen.⁷ NSAIDs have been documented as independent risk factors for bleeding. NSAID users are about 3 times as likely to develop GI AEs compared to nonNSAID users.⁸

Many clinicians aim to further mitigate NSAID-associated bleeding risk by coprescribing a proton pump inhibitor (PPI). PPIs provide gastroprotection against NSAID-induced mucosal injury and sequential complication of GI bleeding. In a multicenter randomized control trial, patients who received concomitant PPI therapy while undergoing chronic NSAID therapy—including nonselective and COX-2 selective NSAIDs—had a significantly lower risk of GI ulcer development (placebo, 17.0%; 20 mg esomeprazole, 5.2%; 40 mg esomeprazole, 4.6%).⁹ Current clinical guidelines for preventing NSAIDassociated bleeding complications recommend using a COX-2 selective NSAID in combination with PPI therapy for patients at high risk for GI-related bleeding, including the concomitant use of anticoagulants.¹⁰

There is evidence suggesting an increased bleeding risk with NSAIDs when used in combination with vitamin K antagonists such as warfarin.^11,12 A systematic review of warfarin and concomitant NSAID use found an increased risk of overall bleeding with NSAID use in combination with warfarin (odds ratio 1.58; 95% CI, 1.18-2.12), compared to warfarin alone.¹²

Posthoc analyses of randomized clinical trials have also demonstrated an increased bleeding risk with oral anticoagulation and concomitant NSAID use.^13,14 In the RE-LY trial, NSAID users on warfarin or dabigatran had a statistically significant increased risk of major bleeding compared to non-NSAID users (hazard ratio [HR] 1.68; 95% CI, 1.40- 2.02; P < .001).¹³ In the ARISTOTLE trial, patients on warfarin or apixaban who were incident NSAID users were found to have an increased risk of major bleeding (HR 1.61; 95% CI, 1.11-2.33) and clinically relevant nonmajor bleeding (HR 1.70; 95% CI, 1.16- 2.48).¹⁴ These trials found a statistically significant increased bleeding risk associated with NSAID use, though the populations evaluated included patients taking warfarin and patients taking DOACs. These trials did not evaluate the bleeding risk of concomitant NSAID use among DOACs alone.

Evidence on NSAID-associated bleeding risk with DOACs is lacking in settings where the patient population, prescribing practices, and monitoring levels are variable. Within the Veterans Health Administration, clinical pharmacist practitioners (CPPs) in anticoagulation clinics oversee DOAC therapy management. CPPs monitor safety and efficacy of DOAC therapies through a population health management tool, the DOAC Dashboard.¹⁵ The DOAC Dashboard creates alerts for patients who may require an intervention based on certain clinical parameters, such as drug-drug interactions.¹⁶ Whenever a patient on a DOAC is prescribed an NSAID, an alert is generated on the DOAC Dashboard to flag the CPPs for the potential need for an intervention. If NSAID therapy remains clinically indicated, CPPs may recommend risk reduction strategies such as a COX-2 selective NSAID or coprescribing a PPI.¹⁰

The DOAC Dashboard provides an ideal setting for investigating the effects of NSAID use, NSAID selectivity, and PPI coprescribing on DOAC bleeding rates. With an increasing population of patients receiving anticoagulation therapy with a DOAC, more guidance regarding the bleeding risk of concomitant NSAID use with DOACs is needed. Studies evaluating the bleeding risk with concomitant NSAID use in patients on a DOAC alone are limited. This is the first study to date to compare bleeding risk with concomitant NSAID use between DOACs. This study provides information on bleeding risk with NSAID use among commonly prescribed DOACs, rivaroxaban and apixaban, and the potential impacts of current risk reduction strategies.

METHODS

This single-center retrospective cohort review was performed using the electronic health records (EHRs) of patients enrolled in the US Department of Veterans Affairs (VA) Mountain Home Healthcare System who received rivaroxaban or apixaban from December 2020 to December 2022. This study received approval from the East Tennessee State University/VA Institutional Review Board committee.

Patients were identified through the DOAC Dashboard, aged 21 to 100 years, and received rivaroxaban or apixaban at a therapeutic dose: rivaroxaban 10 to 20 mg daily or apixaban 2.5 to 5 mg twice daily. Patients were excluded if they were prescribed dual antiplatelet therapy, received rivaroxaban at dosing indicated for peripheral vascular disease, were undergoing dialysis, had evidence of moderate to severe hepatic impairment or any hepatic disease with coagulopathy, were undergoing chemotherapy or radiation, or had hematological conditions with predisposed bleeding risk. These patients were excluded to mitigate the potential confounding impact from nontherapeutic DOAC dosing strategies and conditions associated with an increased bleeding risk.

Eligible patients were stratified based on NSAID use. NSAID users were defined as patients prescribed an oral NSAID, including both acute and chronic courses, at any point during the study time frame while actively on a DOAC. Bleeding events were reviewed to evaluate rates between rivaroxaban and apixaban among NSAID and nonNSAID users. Identified NSAID users were further assessed for NSAID selectivity and PPI coprescribing as a subgroup analysis for the secondary assessment.

Data Collection

Baseline data were collected, including age, body mass index, anticoagulation indication, DOAC agent, DOAC dose, and DOAC total daily dose. Baseline serum creatinine levels, liver function tests, hemoglobin levels, and platelet counts were collected from the most recent data available immediately prior to the bleeding event, if applicable.

The DOAC Dashboard was reviewed for active and dismissed drug interaction alerts to identify patients taking rivaroxaban or apixaban who were prescribed an NSAID. Patients were categorized in the NSAID group if an interacting drug alert with an NSAID was reported during the study time frame. Data available through the interacting drug alerts on NSAID use were limited to the interacting drug name and date of the reported flag. Manual EHR review was required to confirm dates of NSAID therapy initiation and NSAID discontinuation, if applicable.

Data regarding concomitant antiplatelet use were obtained through review of the active and dismissed drug interaction alerts on the DOAC Dashboard. Concomitant antiplatelet use was defined as the prescribing of a single antiplatelet agent at any point while receiving DOAC therapy. Data on concomitant antiplatelets were collected regardless of NSAID status.

Data on coprescribed PPI therapy were obtained through manual EHR review of identified NSAID users. Coprescribed PPI therapy was defined as the prescribing of a PPI at any point during NSAID therapy. Data regarding PPI use among non-NSAID users were not collected because the secondary endpoint was designed to assess PPI use only among patients coprescribed a DOAC and NSAID.

Outcomes

Bleeding events were identified through an outcomes report generated by the DOAC Dashboard based on International Classification of Diseases, Tenth Revision diagnosis codes associated with a bleeding event. The outcomes report captures diagnoses from the outpatient and inpatient care settings. Reported bleeding events were limited to patients who received a DOAC at any point in the 6 months prior to the event and excluded patients with recent DOAC initiation within 7 days of the event, as these patients are not captured on the DOAC Dashboard.

All reported bleeding events were manually reviewed in the EHR and categorized as a major or clinically relevant nonmajor bleed, according to International Society of Thrombosis and Haemostasis criteria. Validated bleeding events were then crossreferenced with the interacting drug alerts report to identify events with potentially overlapping NSAID therapy at the time of the event. Overlapping NSAID therapy was defined as the prescribing of an NSAID at any point in the 6 months prior to the event. All events with potential overlapping NSAID therapies were manually reviewed for confirmation of NSAID status at the time of the event.

The primary endpoint was a composite of any bleeding event per International Society of Thrombosis and Haemostasis criteria. The secondary endpoint evaluated the potential impact of NSAID selectivity or PPI coprescribing on the bleeding rate among the NSAID user groups.

Statistical Analysis

Analyses were performed consistent with the methods used in the ARISTOTLE and RE-LY trials. It was determined that a sample size of 504 patients, with ≥ 168 patients in each group, would provide 80% power using a 2-sided a of 0.05. HRs with 95% CIs and respective P values were calculated using a SPSS-adapted online calculator.

RESULTS

The DOAC Dashboard identified 681 patients on rivaroxaban and 3225 patients on apixaban; 72 patients on rivaroxaban (10.6%) and 300 patients on apixaban (9.3%) were NSAID users. The mean age of NSAID users was 66.9 years in the rivaroxaban group and 72.4 years in the apixaban group. The mean age of non-NSAID users was 71.5 years in the rivaroxaban group and 75.6 years in the apixaban group. No appreciable differences were observed among subgroups in body mass index, renal function, hepatic function, hemoglobin, or platelet counts, and no statistically significant differences were identified (Table 1). Antiplatelet agents identified included aspirin, clopidogrel, prasugrel, and ticagrelor. Fifteen patients (20.3%) in the rivaroxaban group and 87 patients (28.7%) in the apixaban group had concomitant antiplatelet and NSAID use. Forty-five patients on rivaroxaban (60.8%) and 170 (55.9%) on apixaban were prescribed concomitant PPI and NSAID at baseline. Among non-NSAID users, there was concomitant antiplatelet use for 265 patients (43.6%) in the rivaroxaban group and 1401 patients (47.9%) in the apixaban group. Concomitant PPI use was identified among 63 patients (60.0%) taking selective NSAIDs and 182 (57.2%) taking nonselective NSAIDs.

A total of 423 courses of NSAIDs were identified: 85 NSAID courses in the rivaroxaban group and 338 NSAID courses in the apixaban group. Most NSAID courses involved a nonselective NSAID in the rivaroxaban and apixaban NSAID user groups: 75.2% (n = 318) aggregately compared to 71.8% (n = 61) and 76.0% (n = 257) in the rivaroxaban and apixaban groups, respectively. The most frequent NSAID courses identified were meloxicam (26.7%; n = 113), celecoxib (24.8%; n = 105), ibuprofen (19.1%; n = 81), and naproxen (13.5%; n = 57). Data regarding NSAID therapy initiation and discontinuation dates were not readily available. As a result, the duration of NSAID courses was not captured.

There was no statistically significant difference in bleeding rates between rivaroxaban and apixaban among NSAID users (HR 1.04; 95% CI, 0.98-1.12) or non-NSAID users (HR 1.15; 95% CI, 0.80-1.66) (Table 2). Apixaban non-NSAID users had a higher rate of major bleeds (HR 0.32; 95% CI, 0.17-0.61) while rivaroxaban non-NSAID users had a higher rate of clinically relevant nonmajor bleeds (HR 1.63; 95% CI, 1.10-2.54).

The sample size for the secondary endpoint consisted of bleeding events that were confirmed to have had an overlapping NSAID prescribed at the time of the event. For this secondary assessment, there was 1 rivaroxaban NSAID user bleeding event and 4 apixaban NSAID user bleeding events. For the rivaroxaban NSAID user bleeding event, the NSAID was nonselective and a PPI was not coprescribed. For the apixaban NSAID user bleeding events, 2 NSAIDs were nonselective and 2 were selective. All patients with apixaban and NSAID bleeding events had a coprescribed PPI. There was no clinically significant difference in the bleeding rates observed for NSAID selectivity or PPI coprescribing among the NSAID user subgroups.

DISCUSSION

This study found that there was no statistically significant difference for bleeding rates of major and nonmajor bleeding events between rivaroxaban and apixaban among NSAID users and non-NSAID users. This study did not identify a clinically significant impact on bleeding rates from NSAID selectivity or PPI coprescribing among the NSAID users.

There were notable but not statistically significant differences in baseline characteristics observed between the NSAID and non-NSAID user groups. On average, the rivaroxaban and apixaban NSAID users were younger compared with those not taking NSAIDs. NSAIDs, specifically nonselective NSAIDs, are recognized as potentially inappropriate medications for older adults given that this population is at an increased risk for GI ulcer development and/or GI bleeding.¹⁷ The non-NSAID user group likely consisted of older patients compared to the NSAID user group as clinicians may avoid prescribing NSAIDs to older adults regardless of concomitant DOAC therapy.

In addition to having an older patient population, non-NSAID users were more frequently prescribed a concomitant antiplatelet when compared with NSAID users. This prescribing pattern may be due to clinicians avoiding the use of NSAIDs in patients receiving DOAC therapy in combination with antiplatelet therapy, as these patients have been found to have an increased bleeding rate compared to DOAC therapy alone.¹⁸

Non-NSAID users had an overall higher bleeding rate for both major and nonmajor bleeding events. Based on this observation, it could be hypothesized that antiplatelet agents have a higher risk of bleeding in comparison to NSAIDs. In a subanalysis of the EXPAND study evaluating risk factors of major bleeding in patients receiving rivaroxaban, concomitant use of antiplatelet agents demonstrated a statistically significant increased risk of bleeding (HR 1.6; 95% CI, 1.2-2.3; P = .003) while concomitant use of NSAIDs did not (HR 0.8; 95% CI, 0.3-2.2; P = .67).¹⁹

In assessing PPI status at baseline, a majority of both rivaroxaban and apixaban NSAID users were coprescribed a PPI. This trend aligns with current clinical guideline recommendations for the prescribing of PPI therapy for GI protection in high-risk patients, such as those on DOAC therapy and concomitant NSAID therapy.¹⁰ Given the high proportion of NSAID users coprescribed a PPI at baseline, it may be possible that the true incidence of NSAID-associated bleeding events was higher than what this study found. This observation may reflect the impact from timely implementation of risk mitigation strategies by CPPs in the anticoagulation clinic. However, this study was not constructed to assess the efficacy of PPI use in this manner.

It is important to note the patients included in this study were followed by a pharmacist in an anticoagulation clinic using the DOAC Dashboard.¹⁵ This population management tool allows CPPs to make proactive interventions when a patient taking a DOAC receives an NSAID prescription, such as recommending the coprescribing of a PPI or use of a selective NSAID.^10,16 These standards of care may have contributed to an overall reduced bleeding rate among the NSAID user group and may not be reflective of private practice.

The planned analysis of this study was modeled after the posthoc analysis of the RE-LY and ARISTOTLE trials. Both trials demonstrated an increased risk of bleeding with oral anticoagulation, including DOAC and warfarin, in combination with NSAID use. However, both trials found that NSAID use in patients treated with a DOAC was not independently associated with increased bleeding events compared with warfarin.^13,14 The results of this study are comparable to the RE-LY and ARISTOTLE findings that NSAID use among patients treated with rivaroxaban or apixaban did not demonstrate a statistically significant increased bleeding risk.

Studies of NSAID use in combination with DOAC therapy have been limited to patient populations consisting of both DOAC and warfarin. Evidence from these trials outlines the increased bleeding risk associated with NSAID use in combination with oral anticoagulation; however, these patient populations include those on a DOAC and warfarin.^13,14,19,20 Given the limited evidence on NSAID use among DOACs alone, it is assumed NSAID use in combination with DOACs has a similar risk of bleeding as warfarin use. This may cause clinicians to automatically exclude NSAID therapy as a treatment option for patients on a DOAC who are otherwise clinically appropriate candidates, such as those with underlying inflammatory conditions. Avoiding NSAID therapy in this patient population may lead to suboptimal pain management and increase the risk of patient harm from methods such as inappropriate opioid therapy prescribing.

DOAC therapy should not be a universal limitation to the use of NSAIDs. Although the risk of bleeding with NSAID therapy is always present, deliberate NSAID prescribing in addition to the timely implementation of risk mitigation strategies may provide an avenue for safe NSAID prescribing in patients receiving a DOAC. A population health-based approach to DOAC management, such as the DOAC Dashboard, appears to be effective at preventing patient harm when NSAIDs are prescribed in conjunction with DOACs.

Limitations

The DOAC Dashboard has been shown to be effective and efficient at monitoring DOAC therapy from a population-based approach.¹⁶ Reports generated through the DOAC Dashboard provide convenient access to patient data which allows for timely interventions; however, there are limits to its use for data collection. All the data elements necessary to properly assess bleeding risk with validated tools, such as HAS-BLED (hypertension, abnormal renal/liver function, stroke, bleeding history or predisposition, labile international normalized ratio, elderly, drugs/ alcohol concomitantly), are not available on DOAC Dashboard reports. Due to this constraint, bleeding risk assessments were not conducted at baseline and this study was unable to include risk modeling. Additionally, data elements like initiation and discontinuation dates and duration of therapies were not readily available. As a result, this study was unable to incorporate time as a data point.

This was a retrospective study that relied on manual review of chart documentation to verify bleeding events, but data obtained through the DOAC Dashboard were transferred directly from the EHR.¹⁵ Bleeding events available for evaluation were restricted to those that occurred at a VA facility. Additionally, the sample size within the rivaroxaban NSAID user group did not reach the predefined sample size required to reach power and may have been too small to detect a difference if one did exist. The secondary assessment had a low sample size of NSAID user bleeding events, making it difficult to fully assess its impact on NSAID selectivity and PPI coprescribing on bleeding rates. All courses of NSAIDs were equally valued regardless of the dose or therapy duration; however, this is consistent with how NSAID use was defined in the RE-LY and ARISTOTLE trials.

CONCLUSIONS

This retrospective cohort review found no statistically significant difference in the composite bleeding rates between rivaroxaban and apixaban among NSAID users and non-NSAID users. Moreover, there was no clinically significant impact observed for bleeding rates in regard to NSAID selectivity and PPI coprescribing among NSAID users. However, coprescribing of PPI therapy to patients on a DOAC who are clinically indicated for an NSAID may reduce the risk of bleeding. Population health management tools, such as the DOAC Dashboard, may also allow clinicians to safely prescribe NSAIDs to patients on a DOAC. Further large-scale observational studies are needed to quantify the real-world risk of bleeding with concomitant NSAID use among DOACs alone and to evaluate the impact from NSAID selectivity or PPI coprescribing.

Clinical practice has shifted from vitamin K antagonists to direct oral anticoagulants (DOACs) for atrial fibrillation treatment due to their more favorable risk-benefit profile and less lifestyle modification required.^1,2 However, the advantage of a lower bleeding risk with DOACs could be compromised by potentially problematic pharmacokinetic interactions like those conferred by antiplatelets or nonsteroidal anti-inflammatory drugs (NSAIDs).^3,4 Treating a patient needing anticoagulation with a DOAC who has comorbidities may introduce unavoidable drug-drug interactions. This particularly happens with over-the-counter and prescription NSAIDs used for the management of pain and inflammatory conditions.⁵

NSAIDs primarily affect 2 cyclooxygenase (COX) enzyme isomers, COX-1 and COX-2.⁶ COX-1 helps maintain gastrointestinal (GI) mucosa integrity and platelet aggregation processes, whereas COX-2 is engaged in pain signaling and inflammation mediation. COX-1 inhibition is associated with more bleeding-related adverse events (AEs), especially in the GI tract. COX-2 inhibition is thought to provide analgesia and anti-inflammatory properties without elevating bleeding risk. This premise is responsible for the preferential use of celecoxib, a COX-2 selective NSAID, which should confer a lower bleeding risk compared to nonselective NSAIDs such as ibuprofen and naproxen.⁷ NSAIDs have been documented as independent risk factors for bleeding. NSAID users are about 3 times as likely to develop GI AEs compared to nonNSAID users.⁸

Many clinicians aim to further mitigate NSAID-associated bleeding risk by coprescribing a proton pump inhibitor (PPI). PPIs provide gastroprotection against NSAID-induced mucosal injury and sequential complication of GI bleeding. In a multicenter randomized control trial, patients who received concomitant PPI therapy while undergoing chronic NSAID therapy—including nonselective and COX-2 selective NSAIDs—had a significantly lower risk of GI ulcer development (placebo, 17.0%; 20 mg esomeprazole, 5.2%; 40 mg esomeprazole, 4.6%).⁹ Current clinical guidelines for preventing NSAIDassociated bleeding complications recommend using a COX-2 selective NSAID in combination with PPI therapy for patients at high risk for GI-related bleeding, including the concomitant use of anticoagulants.¹⁰

There is evidence suggesting an increased bleeding risk with NSAIDs when used in combination with vitamin K antagonists such as warfarin.^11,12 A systematic review of warfarin and concomitant NSAID use found an increased risk of overall bleeding with NSAID use in combination with warfarin (odds ratio 1.58; 95% CI, 1.18-2.12), compared to warfarin alone.¹²

Posthoc analyses of randomized clinical trials have also demonstrated an increased bleeding risk with oral anticoagulation and concomitant NSAID use.^13,14 In the RE-LY trial, NSAID users on warfarin or dabigatran had a statistically significant increased risk of major bleeding compared to non-NSAID users (hazard ratio [HR] 1.68; 95% CI, 1.40- 2.02; P < .001).¹³ In the ARISTOTLE trial, patients on warfarin or apixaban who were incident NSAID users were found to have an increased risk of major bleeding (HR 1.61; 95% CI, 1.11-2.33) and clinically relevant nonmajor bleeding (HR 1.70; 95% CI, 1.16- 2.48).¹⁴ These trials found a statistically significant increased bleeding risk associated with NSAID use, though the populations evaluated included patients taking warfarin and patients taking DOACs. These trials did not evaluate the bleeding risk of concomitant NSAID use among DOACs alone.

Evidence on NSAID-associated bleeding risk with DOACs is lacking in settings where the patient population, prescribing practices, and monitoring levels are variable. Within the Veterans Health Administration, clinical pharmacist practitioners (CPPs) in anticoagulation clinics oversee DOAC therapy management. CPPs monitor safety and efficacy of DOAC therapies through a population health management tool, the DOAC Dashboard.¹⁵ The DOAC Dashboard creates alerts for patients who may require an intervention based on certain clinical parameters, such as drug-drug interactions.¹⁶ Whenever a patient on a DOAC is prescribed an NSAID, an alert is generated on the DOAC Dashboard to flag the CPPs for the potential need for an intervention. If NSAID therapy remains clinically indicated, CPPs may recommend risk reduction strategies such as a COX-2 selective NSAID or coprescribing a PPI.¹⁰

The DOAC Dashboard provides an ideal setting for investigating the effects of NSAID use, NSAID selectivity, and PPI coprescribing on DOAC bleeding rates. With an increasing population of patients receiving anticoagulation therapy with a DOAC, more guidance regarding the bleeding risk of concomitant NSAID use with DOACs is needed. Studies evaluating the bleeding risk with concomitant NSAID use in patients on a DOAC alone are limited. This is the first study to date to compare bleeding risk with concomitant NSAID use between DOACs. This study provides information on bleeding risk with NSAID use among commonly prescribed DOACs, rivaroxaban and apixaban, and the potential impacts of current risk reduction strategies.

METHODS

This single-center retrospective cohort review was performed using the electronic health records (EHRs) of patients enrolled in the US Department of Veterans Affairs (VA) Mountain Home Healthcare System who received rivaroxaban or apixaban from December 2020 to December 2022. This study received approval from the East Tennessee State University/VA Institutional Review Board committee.

Patients were identified through the DOAC Dashboard, aged 21 to 100 years, and received rivaroxaban or apixaban at a therapeutic dose: rivaroxaban 10 to 20 mg daily or apixaban 2.5 to 5 mg twice daily. Patients were excluded if they were prescribed dual antiplatelet therapy, received rivaroxaban at dosing indicated for peripheral vascular disease, were undergoing dialysis, had evidence of moderate to severe hepatic impairment or any hepatic disease with coagulopathy, were undergoing chemotherapy or radiation, or had hematological conditions with predisposed bleeding risk. These patients were excluded to mitigate the potential confounding impact from nontherapeutic DOAC dosing strategies and conditions associated with an increased bleeding risk.

Eligible patients were stratified based on NSAID use. NSAID users were defined as patients prescribed an oral NSAID, including both acute and chronic courses, at any point during the study time frame while actively on a DOAC. Bleeding events were reviewed to evaluate rates between rivaroxaban and apixaban among NSAID and nonNSAID users. Identified NSAID users were further assessed for NSAID selectivity and PPI coprescribing as a subgroup analysis for the secondary assessment.

Data Collection

Baseline data were collected, including age, body mass index, anticoagulation indication, DOAC agent, DOAC dose, and DOAC total daily dose. Baseline serum creatinine levels, liver function tests, hemoglobin levels, and platelet counts were collected from the most recent data available immediately prior to the bleeding event, if applicable.

The DOAC Dashboard was reviewed for active and dismissed drug interaction alerts to identify patients taking rivaroxaban or apixaban who were prescribed an NSAID. Patients were categorized in the NSAID group if an interacting drug alert with an NSAID was reported during the study time frame. Data available through the interacting drug alerts on NSAID use were limited to the interacting drug name and date of the reported flag. Manual EHR review was required to confirm dates of NSAID therapy initiation and NSAID discontinuation, if applicable.

Data regarding concomitant antiplatelet use were obtained through review of the active and dismissed drug interaction alerts on the DOAC Dashboard. Concomitant antiplatelet use was defined as the prescribing of a single antiplatelet agent at any point while receiving DOAC therapy. Data on concomitant antiplatelets were collected regardless of NSAID status.

Data on coprescribed PPI therapy were obtained through manual EHR review of identified NSAID users. Coprescribed PPI therapy was defined as the prescribing of a PPI at any point during NSAID therapy. Data regarding PPI use among non-NSAID users were not collected because the secondary endpoint was designed to assess PPI use only among patients coprescribed a DOAC and NSAID.

Outcomes

Bleeding events were identified through an outcomes report generated by the DOAC Dashboard based on International Classification of Diseases, Tenth Revision diagnosis codes associated with a bleeding event. The outcomes report captures diagnoses from the outpatient and inpatient care settings. Reported bleeding events were limited to patients who received a DOAC at any point in the 6 months prior to the event and excluded patients with recent DOAC initiation within 7 days of the event, as these patients are not captured on the DOAC Dashboard.

All reported bleeding events were manually reviewed in the EHR and categorized as a major or clinically relevant nonmajor bleed, according to International Society of Thrombosis and Haemostasis criteria. Validated bleeding events were then crossreferenced with the interacting drug alerts report to identify events with potentially overlapping NSAID therapy at the time of the event. Overlapping NSAID therapy was defined as the prescribing of an NSAID at any point in the 6 months prior to the event. All events with potential overlapping NSAID therapies were manually reviewed for confirmation of NSAID status at the time of the event.

The primary endpoint was a composite of any bleeding event per International Society of Thrombosis and Haemostasis criteria. The secondary endpoint evaluated the potential impact of NSAID selectivity or PPI coprescribing on the bleeding rate among the NSAID user groups.

Statistical Analysis

Analyses were performed consistent with the methods used in the ARISTOTLE and RE-LY trials. It was determined that a sample size of 504 patients, with ≥ 168 patients in each group, would provide 80% power using a 2-sided a of 0.05. HRs with 95% CIs and respective P values were calculated using a SPSS-adapted online calculator.

RESULTS

The DOAC Dashboard identified 681 patients on rivaroxaban and 3225 patients on apixaban; 72 patients on rivaroxaban (10.6%) and 300 patients on apixaban (9.3%) were NSAID users. The mean age of NSAID users was 66.9 years in the rivaroxaban group and 72.4 years in the apixaban group. The mean age of non-NSAID users was 71.5 years in the rivaroxaban group and 75.6 years in the apixaban group. No appreciable differences were observed among subgroups in body mass index, renal function, hepatic function, hemoglobin, or platelet counts, and no statistically significant differences were identified (Table 1). Antiplatelet agents identified included aspirin, clopidogrel, prasugrel, and ticagrelor. Fifteen patients (20.3%) in the rivaroxaban group and 87 patients (28.7%) in the apixaban group had concomitant antiplatelet and NSAID use. Forty-five patients on rivaroxaban (60.8%) and 170 (55.9%) on apixaban were prescribed concomitant PPI and NSAID at baseline. Among non-NSAID users, there was concomitant antiplatelet use for 265 patients (43.6%) in the rivaroxaban group and 1401 patients (47.9%) in the apixaban group. Concomitant PPI use was identified among 63 patients (60.0%) taking selective NSAIDs and 182 (57.2%) taking nonselective NSAIDs.

A total of 423 courses of NSAIDs were identified: 85 NSAID courses in the rivaroxaban group and 338 NSAID courses in the apixaban group. Most NSAID courses involved a nonselective NSAID in the rivaroxaban and apixaban NSAID user groups: 75.2% (n = 318) aggregately compared to 71.8% (n = 61) and 76.0% (n = 257) in the rivaroxaban and apixaban groups, respectively. The most frequent NSAID courses identified were meloxicam (26.7%; n = 113), celecoxib (24.8%; n = 105), ibuprofen (19.1%; n = 81), and naproxen (13.5%; n = 57). Data regarding NSAID therapy initiation and discontinuation dates were not readily available. As a result, the duration of NSAID courses was not captured.

There was no statistically significant difference in bleeding rates between rivaroxaban and apixaban among NSAID users (HR 1.04; 95% CI, 0.98-1.12) or non-NSAID users (HR 1.15; 95% CI, 0.80-1.66) (Table 2). Apixaban non-NSAID users had a higher rate of major bleeds (HR 0.32; 95% CI, 0.17-0.61) while rivaroxaban non-NSAID users had a higher rate of clinically relevant nonmajor bleeds (HR 1.63; 95% CI, 1.10-2.54).

The sample size for the secondary endpoint consisted of bleeding events that were confirmed to have had an overlapping NSAID prescribed at the time of the event. For this secondary assessment, there was 1 rivaroxaban NSAID user bleeding event and 4 apixaban NSAID user bleeding events. For the rivaroxaban NSAID user bleeding event, the NSAID was nonselective and a PPI was not coprescribed. For the apixaban NSAID user bleeding events, 2 NSAIDs were nonselective and 2 were selective. All patients with apixaban and NSAID bleeding events had a coprescribed PPI. There was no clinically significant difference in the bleeding rates observed for NSAID selectivity or PPI coprescribing among the NSAID user subgroups.

DISCUSSION

This study found that there was no statistically significant difference for bleeding rates of major and nonmajor bleeding events between rivaroxaban and apixaban among NSAID users and non-NSAID users. This study did not identify a clinically significant impact on bleeding rates from NSAID selectivity or PPI coprescribing among the NSAID users.

There were notable but not statistically significant differences in baseline characteristics observed between the NSAID and non-NSAID user groups. On average, the rivaroxaban and apixaban NSAID users were younger compared with those not taking NSAIDs. NSAIDs, specifically nonselective NSAIDs, are recognized as potentially inappropriate medications for older adults given that this population is at an increased risk for GI ulcer development and/or GI bleeding.¹⁷ The non-NSAID user group likely consisted of older patients compared to the NSAID user group as clinicians may avoid prescribing NSAIDs to older adults regardless of concomitant DOAC therapy.

In addition to having an older patient population, non-NSAID users were more frequently prescribed a concomitant antiplatelet when compared with NSAID users. This prescribing pattern may be due to clinicians avoiding the use of NSAIDs in patients receiving DOAC therapy in combination with antiplatelet therapy, as these patients have been found to have an increased bleeding rate compared to DOAC therapy alone.¹⁸

Non-NSAID users had an overall higher bleeding rate for both major and nonmajor bleeding events. Based on this observation, it could be hypothesized that antiplatelet agents have a higher risk of bleeding in comparison to NSAIDs. In a subanalysis of the EXPAND study evaluating risk factors of major bleeding in patients receiving rivaroxaban, concomitant use of antiplatelet agents demonstrated a statistically significant increased risk of bleeding (HR 1.6; 95% CI, 1.2-2.3; P = .003) while concomitant use of NSAIDs did not (HR 0.8; 95% CI, 0.3-2.2; P = .67).¹⁹

In assessing PPI status at baseline, a majority of both rivaroxaban and apixaban NSAID users were coprescribed a PPI. This trend aligns with current clinical guideline recommendations for the prescribing of PPI therapy for GI protection in high-risk patients, such as those on DOAC therapy and concomitant NSAID therapy.¹⁰ Given the high proportion of NSAID users coprescribed a PPI at baseline, it may be possible that the true incidence of NSAID-associated bleeding events was higher than what this study found. This observation may reflect the impact from timely implementation of risk mitigation strategies by CPPs in the anticoagulation clinic. However, this study was not constructed to assess the efficacy of PPI use in this manner.

It is important to note the patients included in this study were followed by a pharmacist in an anticoagulation clinic using the DOAC Dashboard.¹⁵ This population management tool allows CPPs to make proactive interventions when a patient taking a DOAC receives an NSAID prescription, such as recommending the coprescribing of a PPI or use of a selective NSAID.^10,16 These standards of care may have contributed to an overall reduced bleeding rate among the NSAID user group and may not be reflective of private practice.

The planned analysis of this study was modeled after the posthoc analysis of the RE-LY and ARISTOTLE trials. Both trials demonstrated an increased risk of bleeding with oral anticoagulation, including DOAC and warfarin, in combination with NSAID use. However, both trials found that NSAID use in patients treated with a DOAC was not independently associated with increased bleeding events compared with warfarin.^13,14 The results of this study are comparable to the RE-LY and ARISTOTLE findings that NSAID use among patients treated with rivaroxaban or apixaban did not demonstrate a statistically significant increased bleeding risk.

Studies of NSAID use in combination with DOAC therapy have been limited to patient populations consisting of both DOAC and warfarin. Evidence from these trials outlines the increased bleeding risk associated with NSAID use in combination with oral anticoagulation; however, these patient populations include those on a DOAC and warfarin.^13,14,19,20 Given the limited evidence on NSAID use among DOACs alone, it is assumed NSAID use in combination with DOACs has a similar risk of bleeding as warfarin use. This may cause clinicians to automatically exclude NSAID therapy as a treatment option for patients on a DOAC who are otherwise clinically appropriate candidates, such as those with underlying inflammatory conditions. Avoiding NSAID therapy in this patient population may lead to suboptimal pain management and increase the risk of patient harm from methods such as inappropriate opioid therapy prescribing.

DOAC therapy should not be a universal limitation to the use of NSAIDs. Although the risk of bleeding with NSAID therapy is always present, deliberate NSAID prescribing in addition to the timely implementation of risk mitigation strategies may provide an avenue for safe NSAID prescribing in patients receiving a DOAC. A population health-based approach to DOAC management, such as the DOAC Dashboard, appears to be effective at preventing patient harm when NSAIDs are prescribed in conjunction with DOACs.

Limitations

The DOAC Dashboard has been shown to be effective and efficient at monitoring DOAC therapy from a population-based approach.¹⁶ Reports generated through the DOAC Dashboard provide convenient access to patient data which allows for timely interventions; however, there are limits to its use for data collection. All the data elements necessary to properly assess bleeding risk with validated tools, such as HAS-BLED (hypertension, abnormal renal/liver function, stroke, bleeding history or predisposition, labile international normalized ratio, elderly, drugs/ alcohol concomitantly), are not available on DOAC Dashboard reports. Due to this constraint, bleeding risk assessments were not conducted at baseline and this study was unable to include risk modeling. Additionally, data elements like initiation and discontinuation dates and duration of therapies were not readily available. As a result, this study was unable to incorporate time as a data point.

This was a retrospective study that relied on manual review of chart documentation to verify bleeding events, but data obtained through the DOAC Dashboard were transferred directly from the EHR.¹⁵ Bleeding events available for evaluation were restricted to those that occurred at a VA facility. Additionally, the sample size within the rivaroxaban NSAID user group did not reach the predefined sample size required to reach power and may have been too small to detect a difference if one did exist. The secondary assessment had a low sample size of NSAID user bleeding events, making it difficult to fully assess its impact on NSAID selectivity and PPI coprescribing on bleeding rates. All courses of NSAIDs were equally valued regardless of the dose or therapy duration; however, this is consistent with how NSAID use was defined in the RE-LY and ARISTOTLE trials.

CONCLUSIONS

This retrospective cohort review found no statistically significant difference in the composite bleeding rates between rivaroxaban and apixaban among NSAID users and non-NSAID users. Moreover, there was no clinically significant impact observed for bleeding rates in regard to NSAID selectivity and PPI coprescribing among NSAID users. However, coprescribing of PPI therapy to patients on a DOAC who are clinically indicated for an NSAID may reduce the risk of bleeding. Population health management tools, such as the DOAC Dashboard, may also allow clinicians to safely prescribe NSAIDs to patients on a DOAC. Further large-scale observational studies are needed to quantify the real-world risk of bleeding with concomitant NSAID use among DOACs alone and to evaluate the impact from NSAID selectivity or PPI coprescribing.

Limb loss is a significant and growing concern in the United States. Nearly 2 million Americans are living with limb loss, and up to 185,000 people undergo amputations annually.^1-4 Of these patients, about 35% are women.⁵ The Veterans Health Administration (VHA) provides about 10% of US amputations.^6-8 Between 2015 and 2019, the number of prosthetic devices provided to female veterans increased from 3.3 million to 4.6 million.^5,9,10

Previous research identified disparities in prosthetic care between men and women, both within and outside the VHA. These disparities include slower prosthesis prescription and receipt among women, in addition to differences in self-reported mobility, satisfaction, rates of prosthesis rejection, and challenges related to prosthesis appearance and fit.^5,10,11 Recent studies suggest women tend to have worse outcomes following amputation, and are underrepresented in amputation research.^12,13 However, these disparities are poorly described in a large, national sample. Because women represent a growing portion of patients with limb loss in the VHA, understanding their needs is critical.¹⁴

The Johnny Isakson and David P. Roe, MD Veterans Health Care and Benefits Improvement Act of 2020 was enacted, in part, to improve the care provided to women veterans.¹⁵ The law required the VHA to conduct a survey of ≥ 50,000 veterans to assess the satisfaction of women veterans with prostheses provided by the VHA. To comply with this legislation and understand how women veterans rate their prostheses and related care in the VHA, the US Department of Veterans Affairs (VA) Center for Collaborative Evaluation (VACE) conducted a large national survey of veterans with limb loss that oversampled women veterans. This article describes the survey results, including characteristics of female veterans with limb loss receiving care from the VHA, assesses their satisfaction with prostheses and prosthetic care, and highlights where their responses differ from those of male veterans.

Methods

We conducted a cross-sectional, mixedmode survey of eligible amputees in the VHA Support Service Capital Assets Amputee Data Cube. We identified a cohort of veterans with any major amputation (above the ankle or wrist) or partial hand or foot amputation who received VHA care between October 1, 2019, and September 30, 2020. The final cohort yielded 46,646 potentially eligible veterans. Thirty-three had invalid contact information, leaving 46,613 veterans who were asked to participate, including 1356 women.

Survey

We created a survey instrument de novo that included questions from validated instruments, including the Trinity Amputation Prosthesis and Experience Scales to assess prosthetic device satisfaction, the Prosthesis Evaluation Questionnaire to assess quality of life (QOL) satisfaction, and the Orthotics Prosthetics Users Survey to assess prosthesis-related care satisfaction. ^16-18 Additional questions were incorporated from a survey of veterans with upper limb amputation to assess the importance of cosmetic considerations related to the prosthesis and comfort with prosthesis use in intimate relationships.¹⁹ Questions were also included to assess amputation type, year of amputation, if a prosthesis was currently used, reasons for ceasing use of a prosthesis, reasons for never using a prosthesis, the types of prostheses used, intensity of prosthesis use, satisfaction with time required to receive a prosthetic limb, and if the prosthesis reflected the veteran’s selfidentified gender. Veterans were asked to answer questions based on their most recent amputation.

We tested the survey using cognitive interviews with 6 veterans to refine the survey and better understand how veterans interpreted the questions. Pilot testers completed the survey and participated in individual interviews with experienced interviewers (CL and RRK) to describe how they selected their responses.²⁰ This feedback was used to refine the survey. The online survey was programmed using Qualtrics Software and manually translated into Spanish.

Given the multimodal design, surveys were distributed by email, text message, and US Postal Service (USPS). Surveys were emailed to all veterans for whom a valid email address was available. If emails were undeliverable, veterans were contacted via text message or the USPS. Surveys were distributed by text message to all veterans without an email address but with a cellphone number. We were unable to consistently identify invalid numbers among all text message recipients. Invitations with a survey URL and QR code were sent via USPS to veterans who had no valid email address or cellphone number. Targeted efforts were made to increase the response rate for women. A random sample of 200 women who had not completed the survey 2 weeks prior to the closing date (15% of women in sample) was selected to receive personal phone calls. Another random sample of 400 women was selected to receive personalized outreach emails. The survey data were confidential, and responses could not be traced to identifying information.

Data Analyses

We conducted a descriptive analysis, including percentages and means for responses to variables focused on describing amputation characteristics, prosthesis characteristics, and QOL. All data, including missing values, were used to document the percentage of respondents for each question. Removing missing data from the denominator when calculating percentages could introduce bias to the analysis because we cannot be certain data are missing at random. Missing variables were removed to avoid underinflation of mean scores.

We compared responses across 2 groups: individuals who self-identified as men and individuals who self-identified as women. For each question, we assessed whether each of these groups differed significantly from the remaining sample. For example, we examined whether the percentage of men who answered affirmatively to a question was significantly higher or lower than that of individuals not identifying as male, and whether the percentage of women who answered affirmatively was significantly higher or lower than that of individuals not identifying as female. We utilized x² tests to determine significant differences for percentage calculations and t tests to determine significant differences in means across gender.

Since conducting multiple comparisons within a dataset may result in inflating statistical significance (type 1 errors), we used a more conservative estimate of statistical significance (α = 0.01) and high significance (α = 0.001). This study was deemed quality improvement by the VHA Rehabilitation and Prosthetic Services (12RPS) and acknowledged by the VA Research Office at Eastern Colorado Health Care System and was not subject to institutional review board review.

Results

Surveys were distributed to 46,613 veterans and were completed by 4981 respondents for a 10.7% overall response rate. Survey respondents were generally similar to the eligible population invited to participate, but the proportion of women who completed the survey was higher than the proportion of women eligible to participate (2.0% of eligible population vs 16.7% of respondents), likely due to specific efforts to target women. Survey respondents were slightly younger than the general population (67.3 years vs 68.7 years), less likely to be male (97.1% vs 83.3%), showed similar representation of Operation Enduring Freedom/Operation Iraqi Freedom (OEF/OIF) veterans (4.4% vs 4.1%), and were less likely to have diabetes (58.0% vs 52.7% had diabetes) (Table 1).

The mean age of male respondents was 67.3 years, while the mean age of female respondents was 58.3 years. The majority of respondents were male (83.3%) and White (77.2%). Female respondents were less likely to have diabetes (35.4% of women vs 53.5% of men) and less likely to report that their most recent amputation resulted from diabetes (10.1% of women vs 22.2% of men). Women respondents were more likely to report an amputation due to other causes, such as adverse results of surgery, neurologic disease, suicide attempt, blood clots, tumors, rheumatoid arthritis, and revisions of previous amputations. Most women respondents did not serve during the OEF or OIF eras. The most common amputation site for women respondents was lower limb, either below the knee and above the ankle or above the knee.

Most participants use an everyday prosthesis, but women were more likely to report using a sports-specific prosthesis (Table 2). Overall, most respondents report using a prosthesis (87.7%); however, women were more likely to report not using a prosthesis (19.4% of women vs 11.1% of men; P ≤ .01). Additionally, a lower proportion of women report using a prosthesis for < 12 hours per day (30.6% of women vs 46.4% of men; P ≤ .01) or using a prosthesis every day (54.8% of women vs 74.6% of men; P ≤ .001).

In the overall sample, the mean satisfaction score with a prosthesis was 2.7 on a 5-point scale, and women had slightly lower overall satisfaction scores (2.6 for women vs 2.7 for men; P ≤ .001) (Table 3). Women also had lower satisfaction scores related to appearance, usefulness, reliability, and comfort. Women were more likely to indicate that it was very important to be able to wear jewelry and accessories (20.2% of women vs 11.6% of men; P ≤ .01), while men were less likely to indicate that it was somewhat or very important that the prosthesis not restrict clothing or shoes (95.2% of women vs 82.9% of men; P ≤ .001). Men were more likely than women to report being comfortable or very comfortable using their prosthesis in intimate contact: 40.5% vs 29.0%, respectively (P ≤ .001).

Overall, participants reported high satisfaction with appointment times, wait times, courteous treatment, opportunities to express concerns, and staff responsiveness. Men were slightly more likely than women to be satisfied with training (P ≤ 0.001) and problem discussion (P ≤ 0.01) (Table 4). There were no statistically significant differences in satisfaction or QOL ratings between women and men. The overall sample rated both QOL and satisfaction with QOL 6.7 on a 10-point scale.

Discussion

The goal of this study was to characterize the experience of veterans with limb loss receiving care in the VHA and assess their satisfaction with prostheses and prosthetic care. We received responses from nearly 5000 veterans, 158 of whom were women. Women veteran respondents were slightly younger and less likely to have an amputation due to diabetes. We did not observe significant differences in amputation level between men and women but women were less likely to use a prosthesis, reported lower intensity of prosthesis use, and were less satisfied with certain aspects of their prostheses. Women may also be less satisfied with prosthesis training and problem discussion. However, we found no differences in QOL ratings between men and women.

Findings indicating women were more likely to report not using a prosthesis and that a lower proportion of women report using a prosthesis for > 12 hours a day or every day are consistent with previous research. 21,22 Interestingly, women were more likely to report using a sports-specific prosthesis. This is notable because prior research suggests that individuals with amputations may avoid participating in sports and exercise, and a lack of access to sports-specific prostheses may inhibit physical activity.^23,24 Women in this sample were slightly less satisfied with their prostheses overall and reported lower satisfaction scores regarding appearance, usefulness, reliability, and comfort, consistent with previous findings.²⁵

A lower percentage of women in this sample reported being comfortable or very comfortable using their prosthesis during intimate contact. Previous research on prosthesis satisfaction suggests individuals who rate prosthesis satisfaction lower also report lower body image across genders. ²⁶ While women in this sample did not rate their prosthesis satisfaction lower than men, they did report lower intensity of prosthesis use, suggesting potential issues with their prostheses this survey did not evaluate. Women indicated the importance of prostheses not restricting jewelry, accessories, clothing, or shoes. These results have significant clinical and social implications. A recent qualitative study emphasizes that women veterans feel prostheses are primarily designed for men and may not work well with their physiological needs.⁹ Research focused on limbs better suited to women’s bodies could result in better fitting sockets, lightweight limbs, or less bulky designs. Additional research has also explored the difficulties in accommodating a range of footwear for patients with lower limb amputation. One study found that varying footwear heights affect the function of adjustable prosthetic feet in ways that may not be optimal.²⁷

Ratings of satisfaction with prosthesisrelated services between men and women in this sample are consistent with a recent study showing that women veterans do not have significant differences in satisfaction with prosthesis-related services.²⁸ However, this study focused specifically on lower limb amputations, while the respondents of this study include those with both upper and lower limb amputations. Importantly, our findings that women are less likely to be satisfied with prosthesis training and problem discussions support recent qualitative findings in which women expressed a desire to work with prosthetists who listen to them, take their concerns seriously, and seek solutions that fit their needs. We did not observe a difference in QOL ratings between men and women in the sample despite lower satisfaction among women with some elements of prosthesis-related services. Previous research suggests many factors impact QOL after amputation, most notably time since amputation.^16,29

Limitations

This survey was deployed in a short timeline that did not allow for careful sample selection or implementing strategies to increase response rate. Additionally, the study was conducted among veterans receiving care in the VHA, and findings may not be generalizable to limb loss in other settings. Finally, the discrepancy in number of respondents who identified as men vs women made it difficult to compare differences between the 2 groups.

Conclusions

This is the largest sample of survey respondents of veterans with limb loss to date. While the findings suggest veterans are generally satisfied with prosthetic-related services overall, they also highlight several areas for improvement with services or prostheses. Given that most veterans with limb loss are men, there is a significant discrepancy between the number of women and men respondents. Additional studies with more comparable numbers of men and women have found similar ratings of satisfaction with prostheses and services.²⁸ Further research specifically focused on improving the experiences of women should focus on better characterizing their experiences and identifying how they differ from those of male veterans. For example, understanding how to engage female veterans with limb loss in prosthesis training and problem discussions may improve their experience with their care teams and improve their use of prostheses. Understanding experiences and needs that are specific to women could lead to the development of processes, resources, or devices that are tailored to the unique requirements of women with limb loss.

Limb loss is a significant and growing concern in the United States. Nearly 2 million Americans are living with limb loss, and up to 185,000 people undergo amputations annually.^1-4 Of these patients, about 35% are women.⁵ The Veterans Health Administration (VHA) provides about 10% of US amputations.^6-8 Between 2015 and 2019, the number of prosthetic devices provided to female veterans increased from 3.3 million to 4.6 million.^5,9,10

Previous research identified disparities in prosthetic care between men and women, both within and outside the VHA. These disparities include slower prosthesis prescription and receipt among women, in addition to differences in self-reported mobility, satisfaction, rates of prosthesis rejection, and challenges related to prosthesis appearance and fit.^5,10,11 Recent studies suggest women tend to have worse outcomes following amputation, and are underrepresented in amputation research.^12,13 However, these disparities are poorly described in a large, national sample. Because women represent a growing portion of patients with limb loss in the VHA, understanding their needs is critical.¹⁴

The Johnny Isakson and David P. Roe, MD Veterans Health Care and Benefits Improvement Act of 2020 was enacted, in part, to improve the care provided to women veterans.¹⁵ The law required the VHA to conduct a survey of ≥ 50,000 veterans to assess the satisfaction of women veterans with prostheses provided by the VHA. To comply with this legislation and understand how women veterans rate their prostheses and related care in the VHA, the US Department of Veterans Affairs (VA) Center for Collaborative Evaluation (VACE) conducted a large national survey of veterans with limb loss that oversampled women veterans. This article describes the survey results, including characteristics of female veterans with limb loss receiving care from the VHA, assesses their satisfaction with prostheses and prosthetic care, and highlights where their responses differ from those of male veterans.

Methods

We conducted a cross-sectional, mixedmode survey of eligible amputees in the VHA Support Service Capital Assets Amputee Data Cube. We identified a cohort of veterans with any major amputation (above the ankle or wrist) or partial hand or foot amputation who received VHA care between October 1, 2019, and September 30, 2020. The final cohort yielded 46,646 potentially eligible veterans. Thirty-three had invalid contact information, leaving 46,613 veterans who were asked to participate, including 1356 women.

Survey

We created a survey instrument de novo that included questions from validated instruments, including the Trinity Amputation Prosthesis and Experience Scales to assess prosthetic device satisfaction, the Prosthesis Evaluation Questionnaire to assess quality of life (QOL) satisfaction, and the Orthotics Prosthetics Users Survey to assess prosthesis-related care satisfaction. ^16-18 Additional questions were incorporated from a survey of veterans with upper limb amputation to assess the importance of cosmetic considerations related to the prosthesis and comfort with prosthesis use in intimate relationships.¹⁹ Questions were also included to assess amputation type, year of amputation, if a prosthesis was currently used, reasons for ceasing use of a prosthesis, reasons for never using a prosthesis, the types of prostheses used, intensity of prosthesis use, satisfaction with time required to receive a prosthetic limb, and if the prosthesis reflected the veteran’s selfidentified gender. Veterans were asked to answer questions based on their most recent amputation.

We tested the survey using cognitive interviews with 6 veterans to refine the survey and better understand how veterans interpreted the questions. Pilot testers completed the survey and participated in individual interviews with experienced interviewers (CL and RRK) to describe how they selected their responses.²⁰ This feedback was used to refine the survey. The online survey was programmed using Qualtrics Software and manually translated into Spanish.

Given the multimodal design, surveys were distributed by email, text message, and US Postal Service (USPS). Surveys were emailed to all veterans for whom a valid email address was available. If emails were undeliverable, veterans were contacted via text message or the USPS. Surveys were distributed by text message to all veterans without an email address but with a cellphone number. We were unable to consistently identify invalid numbers among all text message recipients. Invitations with a survey URL and QR code were sent via USPS to veterans who had no valid email address or cellphone number. Targeted efforts were made to increase the response rate for women. A random sample of 200 women who had not completed the survey 2 weeks prior to the closing date (15% of women in sample) was selected to receive personal phone calls. Another random sample of 400 women was selected to receive personalized outreach emails. The survey data were confidential, and responses could not be traced to identifying information.

Data Analyses

We conducted a descriptive analysis, including percentages and means for responses to variables focused on describing amputation characteristics, prosthesis characteristics, and QOL. All data, including missing values, were used to document the percentage of respondents for each question. Removing missing data from the denominator when calculating percentages could introduce bias to the analysis because we cannot be certain data are missing at random. Missing variables were removed to avoid underinflation of mean scores.

We compared responses across 2 groups: individuals who self-identified as men and individuals who self-identified as women. For each question, we assessed whether each of these groups differed significantly from the remaining sample. For example, we examined whether the percentage of men who answered affirmatively to a question was significantly higher or lower than that of individuals not identifying as male, and whether the percentage of women who answered affirmatively was significantly higher or lower than that of individuals not identifying as female. We utilized x² tests to determine significant differences for percentage calculations and t tests to determine significant differences in means across gender.

Since conducting multiple comparisons within a dataset may result in inflating statistical significance (type 1 errors), we used a more conservative estimate of statistical significance (α = 0.01) and high significance (α = 0.001). This study was deemed quality improvement by the VHA Rehabilitation and Prosthetic Services (12RPS) and acknowledged by the VA Research Office at Eastern Colorado Health Care System and was not subject to institutional review board review.

Results

Surveys were distributed to 46,613 veterans and were completed by 4981 respondents for a 10.7% overall response rate. Survey respondents were generally similar to the eligible population invited to participate, but the proportion of women who completed the survey was higher than the proportion of women eligible to participate (2.0% of eligible population vs 16.7% of respondents), likely due to specific efforts to target women. Survey respondents were slightly younger than the general population (67.3 years vs 68.7 years), less likely to be male (97.1% vs 83.3%), showed similar representation of Operation Enduring Freedom/Operation Iraqi Freedom (OEF/OIF) veterans (4.4% vs 4.1%), and were less likely to have diabetes (58.0% vs 52.7% had diabetes) (Table 1).

The mean age of male respondents was 67.3 years, while the mean age of female respondents was 58.3 years. The majority of respondents were male (83.3%) and White (77.2%). Female respondents were less likely to have diabetes (35.4% of women vs 53.5% of men) and less likely to report that their most recent amputation resulted from diabetes (10.1% of women vs 22.2% of men). Women respondents were more likely to report an amputation due to other causes, such as adverse results of surgery, neurologic disease, suicide attempt, blood clots, tumors, rheumatoid arthritis, and revisions of previous amputations. Most women respondents did not serve during the OEF or OIF eras. The most common amputation site for women respondents was lower limb, either below the knee and above the ankle or above the knee.

Most participants use an everyday prosthesis, but women were more likely to report using a sports-specific prosthesis (Table 2). Overall, most respondents report using a prosthesis (87.7%); however, women were more likely to report not using a prosthesis (19.4% of women vs 11.1% of men; P ≤ .01). Additionally, a lower proportion of women report using a prosthesis for < 12 hours per day (30.6% of women vs 46.4% of men; P ≤ .01) or using a prosthesis every day (54.8% of women vs 74.6% of men; P ≤ .001).

In the overall sample, the mean satisfaction score with a prosthesis was 2.7 on a 5-point scale, and women had slightly lower overall satisfaction scores (2.6 for women vs 2.7 for men; P ≤ .001) (Table 3). Women also had lower satisfaction scores related to appearance, usefulness, reliability, and comfort. Women were more likely to indicate that it was very important to be able to wear jewelry and accessories (20.2% of women vs 11.6% of men; P ≤ .01), while men were less likely to indicate that it was somewhat or very important that the prosthesis not restrict clothing or shoes (95.2% of women vs 82.9% of men; P ≤ .001). Men were more likely than women to report being comfortable or very comfortable using their prosthesis in intimate contact: 40.5% vs 29.0%, respectively (P ≤ .001).

Overall, participants reported high satisfaction with appointment times, wait times, courteous treatment, opportunities to express concerns, and staff responsiveness. Men were slightly more likely than women to be satisfied with training (P ≤ 0.001) and problem discussion (P ≤ 0.01) (Table 4). There were no statistically significant differences in satisfaction or QOL ratings between women and men. The overall sample rated both QOL and satisfaction with QOL 6.7 on a 10-point scale.

Discussion

The goal of this study was to characterize the experience of veterans with limb loss receiving care in the VHA and assess their satisfaction with prostheses and prosthetic care. We received responses from nearly 5000 veterans, 158 of whom were women. Women veteran respondents were slightly younger and less likely to have an amputation due to diabetes. We did not observe significant differences in amputation level between men and women but women were less likely to use a prosthesis, reported lower intensity of prosthesis use, and were less satisfied with certain aspects of their prostheses. Women may also be less satisfied with prosthesis training and problem discussion. However, we found no differences in QOL ratings between men and women.

Findings indicating women were more likely to report not using a prosthesis and that a lower proportion of women report using a prosthesis for > 12 hours a day or every day are consistent with previous research. 21,22 Interestingly, women were more likely to report using a sports-specific prosthesis. This is notable because prior research suggests that individuals with amputations may avoid participating in sports and exercise, and a lack of access to sports-specific prostheses may inhibit physical activity.^23,24 Women in this sample were slightly less satisfied with their prostheses overall and reported lower satisfaction scores regarding appearance, usefulness, reliability, and comfort, consistent with previous findings.²⁵

A lower percentage of women in this sample reported being comfortable or very comfortable using their prosthesis during intimate contact. Previous research on prosthesis satisfaction suggests individuals who rate prosthesis satisfaction lower also report lower body image across genders. ²⁶ While women in this sample did not rate their prosthesis satisfaction lower than men, they did report lower intensity of prosthesis use, suggesting potential issues with their prostheses this survey did not evaluate. Women indicated the importance of prostheses not restricting jewelry, accessories, clothing, or shoes. These results have significant clinical and social implications. A recent qualitative study emphasizes that women veterans feel prostheses are primarily designed for men and may not work well with their physiological needs.⁹ Research focused on limbs better suited to women’s bodies could result in better fitting sockets, lightweight limbs, or less bulky designs. Additional research has also explored the difficulties in accommodating a range of footwear for patients with lower limb amputation. One study found that varying footwear heights affect the function of adjustable prosthetic feet in ways that may not be optimal.²⁷

Ratings of satisfaction with prosthesisrelated services between men and women in this sample are consistent with a recent study showing that women veterans do not have significant differences in satisfaction with prosthesis-related services.²⁸ However, this study focused specifically on lower limb amputations, while the respondents of this study include those with both upper and lower limb amputations. Importantly, our findings that women are less likely to be satisfied with prosthesis training and problem discussions support recent qualitative findings in which women expressed a desire to work with prosthetists who listen to them, take their concerns seriously, and seek solutions that fit their needs. We did not observe a difference in QOL ratings between men and women in the sample despite lower satisfaction among women with some elements of prosthesis-related services. Previous research suggests many factors impact QOL after amputation, most notably time since amputation.^16,29

Limitations

This survey was deployed in a short timeline that did not allow for careful sample selection or implementing strategies to increase response rate. Additionally, the study was conducted among veterans receiving care in the VHA, and findings may not be generalizable to limb loss in other settings. Finally, the discrepancy in number of respondents who identified as men vs women made it difficult to compare differences between the 2 groups.

Conclusions

This is the largest sample of survey respondents of veterans with limb loss to date. While the findings suggest veterans are generally satisfied with prosthetic-related services overall, they also highlight several areas for improvement with services or prostheses. Given that most veterans with limb loss are men, there is a significant discrepancy between the number of women and men respondents. Additional studies with more comparable numbers of men and women have found similar ratings of satisfaction with prostheses and services.²⁸ Further research specifically focused on improving the experiences of women should focus on better characterizing their experiences and identifying how they differ from those of male veterans. For example, understanding how to engage female veterans with limb loss in prosthesis training and problem discussions may improve their experience with their care teams and improve their use of prostheses. Understanding experiences and needs that are specific to women could lead to the development of processes, resources, or devices that are tailored to the unique requirements of women with limb loss.

Limb loss is a significant and growing concern in the United States. Nearly 2 million Americans are living with limb loss, and up to 185,000 people undergo amputations annually.^1-4 Of these patients, about 35% are women.⁵ The Veterans Health Administration (VHA) provides about 10% of US amputations.^6-8 Between 2015 and 2019, the number of prosthetic devices provided to female veterans increased from 3.3 million to 4.6 million.^5,9,10

Previous research identified disparities in prosthetic care between men and women, both within and outside the VHA. These disparities include slower prosthesis prescription and receipt among women, in addition to differences in self-reported mobility, satisfaction, rates of prosthesis rejection, and challenges related to prosthesis appearance and fit.^5,10,11 Recent studies suggest women tend to have worse outcomes following amputation, and are underrepresented in amputation research.^12,13 However, these disparities are poorly described in a large, national sample. Because women represent a growing portion of patients with limb loss in the VHA, understanding their needs is critical.¹⁴

The Johnny Isakson and David P. Roe, MD Veterans Health Care and Benefits Improvement Act of 2020 was enacted, in part, to improve the care provided to women veterans.¹⁵ The law required the VHA to conduct a survey of ≥ 50,000 veterans to assess the satisfaction of women veterans with prostheses provided by the VHA. To comply with this legislation and understand how women veterans rate their prostheses and related care in the VHA, the US Department of Veterans Affairs (VA) Center for Collaborative Evaluation (VACE) conducted a large national survey of veterans with limb loss that oversampled women veterans. This article describes the survey results, including characteristics of female veterans with limb loss receiving care from the VHA, assesses their satisfaction with prostheses and prosthetic care, and highlights where their responses differ from those of male veterans.

Methods

We conducted a cross-sectional, mixedmode survey of eligible amputees in the VHA Support Service Capital Assets Amputee Data Cube. We identified a cohort of veterans with any major amputation (above the ankle or wrist) or partial hand or foot amputation who received VHA care between October 1, 2019, and September 30, 2020. The final cohort yielded 46,646 potentially eligible veterans. Thirty-three had invalid contact information, leaving 46,613 veterans who were asked to participate, including 1356 women.

Survey

We created a survey instrument de novo that included questions from validated instruments, including the Trinity Amputation Prosthesis and Experience Scales to assess prosthetic device satisfaction, the Prosthesis Evaluation Questionnaire to assess quality of life (QOL) satisfaction, and the Orthotics Prosthetics Users Survey to assess prosthesis-related care satisfaction. ^16-18 Additional questions were incorporated from a survey of veterans with upper limb amputation to assess the importance of cosmetic considerations related to the prosthesis and comfort with prosthesis use in intimate relationships.¹⁹ Questions were also included to assess amputation type, year of amputation, if a prosthesis was currently used, reasons for ceasing use of a prosthesis, reasons for never using a prosthesis, the types of prostheses used, intensity of prosthesis use, satisfaction with time required to receive a prosthetic limb, and if the prosthesis reflected the veteran’s selfidentified gender. Veterans were asked to answer questions based on their most recent amputation.

We tested the survey using cognitive interviews with 6 veterans to refine the survey and better understand how veterans interpreted the questions. Pilot testers completed the survey and participated in individual interviews with experienced interviewers (CL and RRK) to describe how they selected their responses.²⁰ This feedback was used to refine the survey. The online survey was programmed using Qualtrics Software and manually translated into Spanish.

Given the multimodal design, surveys were distributed by email, text message, and US Postal Service (USPS). Surveys were emailed to all veterans for whom a valid email address was available. If emails were undeliverable, veterans were contacted via text message or the USPS. Surveys were distributed by text message to all veterans without an email address but with a cellphone number. We were unable to consistently identify invalid numbers among all text message recipients. Invitations with a survey URL and QR code were sent via USPS to veterans who had no valid email address or cellphone number. Targeted efforts were made to increase the response rate for women. A random sample of 200 women who had not completed the survey 2 weeks prior to the closing date (15% of women in sample) was selected to receive personal phone calls. Another random sample of 400 women was selected to receive personalized outreach emails. The survey data were confidential, and responses could not be traced to identifying information.

Data Analyses

We conducted a descriptive analysis, including percentages and means for responses to variables focused on describing amputation characteristics, prosthesis characteristics, and QOL. All data, including missing values, were used to document the percentage of respondents for each question. Removing missing data from the denominator when calculating percentages could introduce bias to the analysis because we cannot be certain data are missing at random. Missing variables were removed to avoid underinflation of mean scores.

We compared responses across 2 groups: individuals who self-identified as men and individuals who self-identified as women. For each question, we assessed whether each of these groups differed significantly from the remaining sample. For example, we examined whether the percentage of men who answered affirmatively to a question was significantly higher or lower than that of individuals not identifying as male, and whether the percentage of women who answered affirmatively was significantly higher or lower than that of individuals not identifying as female. We utilized x² tests to determine significant differences for percentage calculations and t tests to determine significant differences in means across gender.

Since conducting multiple comparisons within a dataset may result in inflating statistical significance (type 1 errors), we used a more conservative estimate of statistical significance (α = 0.01) and high significance (α = 0.001). This study was deemed quality improvement by the VHA Rehabilitation and Prosthetic Services (12RPS) and acknowledged by the VA Research Office at Eastern Colorado Health Care System and was not subject to institutional review board review.

Results

Surveys were distributed to 46,613 veterans and were completed by 4981 respondents for a 10.7% overall response rate. Survey respondents were generally similar to the eligible population invited to participate, but the proportion of women who completed the survey was higher than the proportion of women eligible to participate (2.0% of eligible population vs 16.7% of respondents), likely due to specific efforts to target women. Survey respondents were slightly younger than the general population (67.3 years vs 68.7 years), less likely to be male (97.1% vs 83.3%), showed similar representation of Operation Enduring Freedom/Operation Iraqi Freedom (OEF/OIF) veterans (4.4% vs 4.1%), and were less likely to have diabetes (58.0% vs 52.7% had diabetes) (Table 1).

The mean age of male respondents was 67.3 years, while the mean age of female respondents was 58.3 years. The majority of respondents were male (83.3%) and White (77.2%). Female respondents were less likely to have diabetes (35.4% of women vs 53.5% of men) and less likely to report that their most recent amputation resulted from diabetes (10.1% of women vs 22.2% of men). Women respondents were more likely to report an amputation due to other causes, such as adverse results of surgery, neurologic disease, suicide attempt, blood clots, tumors, rheumatoid arthritis, and revisions of previous amputations. Most women respondents did not serve during the OEF or OIF eras. The most common amputation site for women respondents was lower limb, either below the knee and above the ankle or above the knee.

Most participants use an everyday prosthesis, but women were more likely to report using a sports-specific prosthesis (Table 2). Overall, most respondents report using a prosthesis (87.7%); however, women were more likely to report not using a prosthesis (19.4% of women vs 11.1% of men; P ≤ .01). Additionally, a lower proportion of women report using a prosthesis for < 12 hours per day (30.6% of women vs 46.4% of men; P ≤ .01) or using a prosthesis every day (54.8% of women vs 74.6% of men; P ≤ .001).

In the overall sample, the mean satisfaction score with a prosthesis was 2.7 on a 5-point scale, and women had slightly lower overall satisfaction scores (2.6 for women vs 2.7 for men; P ≤ .001) (Table 3). Women also had lower satisfaction scores related to appearance, usefulness, reliability, and comfort. Women were more likely to indicate that it was very important to be able to wear jewelry and accessories (20.2% of women vs 11.6% of men; P ≤ .01), while men were less likely to indicate that it was somewhat or very important that the prosthesis not restrict clothing or shoes (95.2% of women vs 82.9% of men; P ≤ .001). Men were more likely than women to report being comfortable or very comfortable using their prosthesis in intimate contact: 40.5% vs 29.0%, respectively (P ≤ .001).

Overall, participants reported high satisfaction with appointment times, wait times, courteous treatment, opportunities to express concerns, and staff responsiveness. Men were slightly more likely than women to be satisfied with training (P ≤ 0.001) and problem discussion (P ≤ 0.01) (Table 4). There were no statistically significant differences in satisfaction or QOL ratings between women and men. The overall sample rated both QOL and satisfaction with QOL 6.7 on a 10-point scale.

Discussion

The goal of this study was to characterize the experience of veterans with limb loss receiving care in the VHA and assess their satisfaction with prostheses and prosthetic care. We received responses from nearly 5000 veterans, 158 of whom were women. Women veteran respondents were slightly younger and less likely to have an amputation due to diabetes. We did not observe significant differences in amputation level between men and women but women were less likely to use a prosthesis, reported lower intensity of prosthesis use, and were less satisfied with certain aspects of their prostheses. Women may also be less satisfied with prosthesis training and problem discussion. However, we found no differences in QOL ratings between men and women.

Findings indicating women were more likely to report not using a prosthesis and that a lower proportion of women report using a prosthesis for > 12 hours a day or every day are consistent with previous research. 21,22 Interestingly, women were more likely to report using a sports-specific prosthesis. This is notable because prior research suggests that individuals with amputations may avoid participating in sports and exercise, and a lack of access to sports-specific prostheses may inhibit physical activity.^23,24 Women in this sample were slightly less satisfied with their prostheses overall and reported lower satisfaction scores regarding appearance, usefulness, reliability, and comfort, consistent with previous findings.²⁵

A lower percentage of women in this sample reported being comfortable or very comfortable using their prosthesis during intimate contact. Previous research on prosthesis satisfaction suggests individuals who rate prosthesis satisfaction lower also report lower body image across genders. ²⁶ While women in this sample did not rate their prosthesis satisfaction lower than men, they did report lower intensity of prosthesis use, suggesting potential issues with their prostheses this survey did not evaluate. Women indicated the importance of prostheses not restricting jewelry, accessories, clothing, or shoes. These results have significant clinical and social implications. A recent qualitative study emphasizes that women veterans feel prostheses are primarily designed for men and may not work well with their physiological needs.⁹ Research focused on limbs better suited to women’s bodies could result in better fitting sockets, lightweight limbs, or less bulky designs. Additional research has also explored the difficulties in accommodating a range of footwear for patients with lower limb amputation. One study found that varying footwear heights affect the function of adjustable prosthetic feet in ways that may not be optimal.²⁷

Ratings of satisfaction with prosthesisrelated services between men and women in this sample are consistent with a recent study showing that women veterans do not have significant differences in satisfaction with prosthesis-related services.²⁸ However, this study focused specifically on lower limb amputations, while the respondents of this study include those with both upper and lower limb amputations. Importantly, our findings that women are less likely to be satisfied with prosthesis training and problem discussions support recent qualitative findings in which women expressed a desire to work with prosthetists who listen to them, take their concerns seriously, and seek solutions that fit their needs. We did not observe a difference in QOL ratings between men and women in the sample despite lower satisfaction among women with some elements of prosthesis-related services. Previous research suggests many factors impact QOL after amputation, most notably time since amputation.^16,29

Limitations

This survey was deployed in a short timeline that did not allow for careful sample selection or implementing strategies to increase response rate. Additionally, the study was conducted among veterans receiving care in the VHA, and findings may not be generalizable to limb loss in other settings. Finally, the discrepancy in number of respondents who identified as men vs women made it difficult to compare differences between the 2 groups.

Conclusions

This is the largest sample of survey respondents of veterans with limb loss to date. While the findings suggest veterans are generally satisfied with prosthetic-related services overall, they also highlight several areas for improvement with services or prostheses. Given that most veterans with limb loss are men, there is a significant discrepancy between the number of women and men respondents. Additional studies with more comparable numbers of men and women have found similar ratings of satisfaction with prostheses and services.²⁸ Further research specifically focused on improving the experiences of women should focus on better characterizing their experiences and identifying how they differ from those of male veterans. For example, understanding how to engage female veterans with limb loss in prosthesis training and problem discussions may improve their experience with their care teams and improve their use of prostheses. Understanding experiences and needs that are specific to women could lead to the development of processes, resources, or devices that are tailored to the unique requirements of women with limb loss.

Atherosclerotic cardiovascular disease (ASCVD) is a significant cause of morbidity and mortality in the United States. ASCVD involves the buildup of cholesterol plaque in arteries and includes acute coronary syndrome, peripheral arterial disease, and events such as myocardial infarction and stroke.¹ Cardiovascular disease (CVD) risk factors include high cholesterol levels, elevated blood pressure, insulin resistance, elevated blood glucose levels, smoking, poor dietary habits, and a sedentary lifestyle.²

According to the Centers for Disease Control and Prevention, about 86 million adults aged ≥ 20 years have total cholesterol levels > 200 mg/dL. More than half (54.5%) who could benefit are currently taking cholesterol-lowering medications.³ Controlling high cholesterol in American adults, especially veterans, is essential for reducing CVD morbidity and mortality.

The 2018 American College of Cardiology/American Heart Association (ACC/AHA) guideline recommends a low-density lipoprotein cholesterol (LDL-C) target goal of < 70 mg/dL for patients at high risk for ASCVD. Very high-risk ASCVD includes a history of multiple major ASCVD events or 1 major ASCVD event and multiple high-risk conditions (eg, age ≥ 65 years, smoking, or diabetes).⁴ Major ASCVD events include recent acute coronary syndrome (within the past 12 months), a history of myocardial infarction or ischemic stroke, and symptomatic peripheral artery disease. 

The ACC/AHA guideline suggests that if the LDL-C level remains ≥ 70 mg/dL, adding ezetimibe (a dietary cholesterol absorption inhibitor) to maximally tolerated statin therapy is reasonable. If LDL-C levels remain ≥ 70 mg/dL, adding a proprotein convertase subtilisin/kexin type 9 (PCSK9) inhibitor, such as alirocumab, is reasonable.⁴ The US Departments of Veterans Affairs/US Department of Defense guidelines recommend using maximally tolerated statins and ezetimibe before PCSK9 inhibitors due to established long-term safety and reduction in CVD events. 

Generic statins and ezetimibe are administered orally and widely available. In contrast, PCSK9 inhibitors have unknown long-term safety profiles, require subcutaneous injection once or twice monthly, and are significantly more expensive. They also require patient education on proper use while providing comparable or lesser relative risk reductions.²

These 3 classes of medication vary in their mechanisms of action to reduce LDL.^5,6 Ezetimibe and several statin medications are included on the Veterans Affairs Sioux Falls Health Care System (VASFHCS) formulary and do not require review prior to prescribing. Alirocumab is available at VASFHCS but is restricted to patients with a history of ASCVD or a diagnosis of familial hypercholesterolemia, and who are receiving maximally tolerated statin and ezetimibe therapy but require further LDL-C lowering to reduce their ASCVD risk. 

Studies have found ezetimibe monotherapy reduces LDL-C in patients with dyslipidemia by 18% after 12 weeks.⁷ One found that the percentage reduction in LDL-C was significantly greater (P < .001) with all doses of ezetimibe plus simvastatin (46% to 59%) compared with either atorvastatin 10 mg (37%) or simvastatin 20 mg (38%) monotherapy after 6 weeks.⁸

Although alirocumab can be added to other lipid therapies, most VASFHCS patients are prescribed alirocumab monotherapy. In the ODYSSEY CHOICE II study, patients were randomly assigned to receive either a placebo or alirocumab 150 mg every 4 weeks or alirocumab 75 mg every 2 weeks. The primary efficacy endpoint was LDL-C percentage change from baseline to week 24. In the alirocumab 150 mg every 4 weeks and 75 mg every 2 weeks groups, the least-squares mean LDL-C changes from baseline to week 24 were 51.7% and 53.5%, respectively, compared to a 4.7% increase in the placebo group (both groups P < .001 vs placebo). The authors also reported that alirocumab 150 mg every 4 weeks as monotherapy demonstrated a 47.4% reduction in LDL-C levels from baseline in a phase 1 study.⁹Although alirocumab monotherapy and ezetimibe plus statin therapy have been shown to effectively decrease LDL-C independently, a direct comparison of alirocumab monotherapy vs ezetimibe plus statin therapy has not been assessed, to our knowledge. Understanding the differences in effectiveness and safety between these 2 regimens will be valuable for clinicians when selecting a medication regimen for veterans with a history of ASCVD.

METHODS

This retrospective, single-center chart review used VASFHCS Computerized Patient Record System (CPRS) and Joint Longitudinal Viewer (JLV) records to compare patients with a history of ASCVD events who were treated with alirocumab monotherapy or ezetimibe plus statin. The 2 groups were randomized in a 1:3 ratio. The primary endpoint was achieving LDL-C < 70 mg/dL after 4 to 12 weeks, 13 to 24 weeks, and 25 to 52 weeks. Secondary endpoints included the mean percentage change from baseline in total cholesterol (TC), high-density lipoprotein cholesterol (HDL-C), LDL-C, and triglycerides (TG) over 52 weeks. The incidence of ASCVD events during this period was also assessed. If LDL-C < 70 mg/dL was achieved > 1 time during each time frame, only 1 incident was counted for analysis. Safety was assessed based on the incidence of any adverse event (AE) that led to treatment discontinuation.

Patients were identified by screening the prescription fill history between October 1, 2019, and December 31, 2022. The 52-week data collection period was counted from the first available fill date. Additionally, the prior authorization drug request file from January 1, 2017, to December 31, 2022, was used to obtain a list of patients prescribed alirocumab. Patients were included if they were veterans aged ≥ 18 years and had a history of an ASCVD event, had a alirocumab monotherapy or ezetimibe plus statin prescription between October 1, 2019, and December 31, 2022, or had an approved prior authorization drug request for alirocumab between January 1, 2017, and December 31, 2022. Patients missing a baseline or follow-up lipid panel and those with concurrent use of alirocumab and ezetimibe and/or statin were excluded. 

Baseline characteristics collected for patients included age, sex, race, weight, body mass index, lipid parameters (LDL-C, TC, HDL-C, and TG), dosing of each type of statin before adding ezetimibe, and use of any other antihyperlipidemic medication. We also collected histories of hypertension, hyperlipidemia, diabetes, chronic kidney disease, congestive heart failure, and smoking or tobacco use status. The baseline lipid panel was the most recent lipid panel documented before starting alirocumab or ezetimibe plus statin therapy. Follow-up lipid panel values were gathered at 4 to 12 weeks, 13 to 24 weeks, and 25 to 52 weeks following initiation of either therapy.

High-, moderate-, and low-intensity dosing of statin therapy and alirocumab dosing (75 mg every 2 weeks, 150 mg every 2 weeks, or 300 mg every 4 weeks) were recorded at the specified intervals. However, no patients in this study received the latter dosing regimen. ASCVD events and safety endpoints were recorded based on a review of clinical notes over the 52 weeks following the first available start date.

Statistical Analysis

The primary endpoint of achieving the LDL-C < 70 mg/dL goal from baseline to 4 to 12 weeks, 13 to 24 weeks, and 25 to 52 weeks after initiation was compared between alirocumab monotherapy and ezetimibe plus statin therapy using the χ² test. Mean percentage change from baseline in TC, HDL-C, LDL-C, and TG were compared using the independent t test. P < .05 was considered statistically significant. Incidence of ASCVD events and the safety endpoint (incidence of AEs leading to treatment discontinuation) were also compared using the χ² test. Continuous baseline characteristics were reported mean (SD) and nominal baseline characteristics were reported as a percentage.

RESULTS

There were 80 participants in this study: 20 in the alirocumab monotherapy group and 60 in the ezetimibe plus statin therapy group. More than 100 patients did not meet the prespecified inclusion criteria and were excluded. Mean (SD) age was 75 (8) years in the alirocumab group and 74 (8) years in the ezetimibe plus statin group. There was no significant differences in mean (SD) weight or mean (SD) body mass index. All study participants identified as White and male except for 2 patients in the ezetimibe plus statin therapy group whose race was not documented. Differences in lipid parameters were observed between groups, with mean baseline LDL-C, HDL-C, and TC higher in the alirocumab monotherapy group than in the ezetimibe plus statin therapy group, with significant differences in LDL-C and TC (Table 1).

Fourteen patients (70%) in the alirocumab monotherapy group had hypertension, compared with 31 (52%) in the ezetimibe plus statin therapy group. In both groups, most patients had previously been diagnosed with hyperlipidemia. More patients (60%) in the alirocumab group had diabetes than in the ezetimibe plus statin therapy group (37%). The alirocumab monotherapy group also had a higher percentage of patients with diagnoses of congestive heart failure and used other antihyperlipidemic medications than in the ezetimibe plus statin therapy group. Five patients (25%) in the alirocumab monotherapy group and 12 patients (20%) in the ezetimibe plus statin therapy group took fish oil. In the ezetimibe plus statin therapy group, 2 patients (3%) took gemfibrozil, and 2 patients (3%) took fenofibrate. Six (30%) patients in the alirocumab monotherapy group and 12 (20%) patients in the ezetimibe plus statin therapy group had chronic kidney disease. Although the majority of patients in each group did not use tobacco products, there were more tobacco users in the ezetimibe plus statin therapy group.

In the alirocumab monotherapy group, 15 patients (75%) were prescribed 75 mg every 2 weeks and 5 patients (25%) were prescribed 150 mg every 2 weeks. In the ezetimibe plus statin therapy group, 59 patients (98%) were prescribed ezetimibe 10 mg/d (Table 2). Forty-three patients (72%) were prescribed a high-intensity statin 10 received moderate-intensity (17%) and 7 received low-intensity statin (12%). Most patients were prescribed rosuvastatin (45%), followed by atorvastatin (42%), pravastatin (10%), and simvastatin (3%).

Primary Endpoint

During the 52-week study, more patients met the LDL-C goal of < 70 mg/dL in the alirocumab monotherapy group (70%) than in the ezetimibe plus statin therapy group (57%); however, the difference was not significant (P = .29). Of the patients prescribed alirocumab monotherapy who achieved LDL-C < 70 mg/dL, 15% achieved this goal in 4 to 12 weeks, 40% in 13 to 24 weeks, and 45% in 25 to 52 weeks. In the ezetimibe plus statin therapy group, 28% of patients achieved LDL-C < 70 mg/dL in 4 to 12 weeks, 31% in 13 to 24 weeks, and 41% in 25 to 52 weeks (Table 3).

Secondary Endpoints

During weeks 4 to 52 of treatment, the mean percentage change decreased in LDL-C (37.7% vs 21.4%; P = .01), TC (24.7% vs 12.5%; P = .01), and TG (0.9% vs 7.0%; P = .28) in the alirocumab monotherapy group and the ezetimibe plus statin therapy group, respectively (Table 4). The mean percentage change increased in HDL-C by 3.6% in the alirocumab monotherapy group and 1.8% in the ezetimibe plus statin therapy group (P = .36). During the study, ASCVD events occurred in 1 patient (5%) in the alirocumab monotherapy group and 3 patients (5%) in the ezetimibe plus statin therapy group (P = .99). The patient in the alirocumab monotherapy group had unstable angina 1 month after taking alirocumab. One patient in the ezetimibe plus statin therapy group had coronary artery disease and 2 patients had coronary heart disease that required stents during the 52-week period. There was 1 patient in each group who reported an AE that led to treatment discontinuation (P = .41). One patient stopped alirocumab after a trial of 2 months due to intolerance, but no specific AE was reported in the CPRS. In the ezetimibe plus statin therapy group, 1 patient requested to discontinue ezetimibe after a trial of 3 months without a specific reason noted in the medical record.

DISCUSSION

This study found no statistically significant difference in the incidence of reaching an LDL-C goal of < 70 mg/dL after alirocumab monotherapy initiation compared with ezetimibe plus statin therapy. This occurred despite baseline LDL-C being lower in the ezetimibe plus statin therapy group, which required a smaller reduction in LDL-C to reach the primary goal. Most patients on alirocumab monotherapy were prescribed a lower initial dose of 75 mg every 2 weeks. Of those patients, 30% did not achieve the LDL-C goal < 70 mg/dL. Thus, a higher dose may have led to more patients achieving the LDL-C goal.

Secondary endpoints, including mean percentage change in HDL-C and TG and incidence of ASCVD events during 52 weeks of treatment, were not statistically significant. The mean percentage increase in HDL-C was negligible in both groups, while the mean percentage reduction in TG favored the ezetimibe plus statin therapy group. In the ezetimibe plus statin therapy group, patients who also took fenofibrate experienced a significant reduction in TG while none of the patients in the alirocumab group were prescribed fenofibrate. Although the alirocumab monotherapy group had a statistically significant greater reduction in LDL-C and TC compared with those prescribed ezetimibe plus statin, the mean baseline LDL-C and TC were significantly greater in the alirocumab monotherapy group, which could contribute to higher reductions in LDL-C and TC after alirocumab monotherapy.Based on the available literature, we expected greater reductions in LDL-C in both study groups compared with statin therapy alone.^8,9 However, it was unclear whether the LDL-C and TC reductions were clinically significant.

Limitations

The study design did not permit randomization prior to the treatments, restricting our ability to account for some confounding factors, such as diet, exercise, other antihyperlipidemic medication, and medication adherence, which may have affected LDL-C, HDL-C, TG, and TC levels. Differences in baseline characteristics—particularly major risk factors, such as hypertension, diabetes, and tobacco use—also could have confounding affect on lipid levels and ASCVD events. Additionally, patients prescribed alirocumab monotherapy may have switched from statin or ezetimibe therapy, and the washout period was not reviewed or recorded, which could have affected the lipid panel results.

The small sample size of this study also may have limited the ability to detect significant differences between groups. A direct comparison of alirocumab monotherapy vs ezetimibe plus statin therapy has not been performed, making it difficult to prospectively evaluate what sample size would be needed to power this study. A posthoc analysis was used to calculate power, which was found to be only 17%. Many patients were excluded due to a lack of laboratory results within the study period, contributing to the small sample size. 

Another limitation was the reliance on documentation in CPRS and JLV. For example, having documentation of the specific AEs for the 2 patients who discontinued alirocumab or ezetimibe could have helped determine the severity of the AEs. Several patients were followed by non-VA clinicians, which could have contributed to limited documentation in the CPRS and JLV. It is difficult to draw any conclusions regarding ASCVD events and AEs that led to treatment discontinuation between alirocumab monotherapy and ezetimibe plus statin therapy based on the results of this retrospective study due to the limited number of events within the 52-week period.

CONCLUSIONS

This study found that there was no statistically significant difference in LDL-C reduction to < 70 mg/dL between alirocumab monotherapy and ezetimibe plus statin therapy in a small population of veterans with ASCVD, with a higher percentage of participants in both groups achieving that goal in 25 to 52 weeks. There also was no significant difference in percentage change in HDL-C or TG or in incidence of ASCVD events and AEs leading to treatment discontinuation. However, there was a statistically significant difference in percentage reduction for LDL-C and TC during 52 weeks of alirocumab monotherapy vs ezetimibe plus statin therapy.

Although there was no significant difference in LDL-C reduction to < 70 mg/dL, targeting this goal in patients with ASCVD is still clinically warranted. This study does not support a change in current VA criteria for use of alirocumab or a change in current guidelines for secondary prevention of ASCVD. Still, this study does indicate that the efficacy of alirocumab monotherapy is similar to that of ezetimibe plus statin therapy in patients with a history of ASCVD and may be useful in clinical settings when an alternative to ezetimibe plus statin therapy is needed. Alirocumab also may be more effective in lowering LDL-C and TC than ezetimibe plus statin therapy in veterans with ASCVD and could be added to statin therapy or ezetimibe when additional LDL-C or TC reduction is needed.

Atherosclerotic cardiovascular disease (ASCVD) is a significant cause of morbidity and mortality in the United States. ASCVD involves the buildup of cholesterol plaque in arteries and includes acute coronary syndrome, peripheral arterial disease, and events such as myocardial infarction and stroke.¹ Cardiovascular disease (CVD) risk factors include high cholesterol levels, elevated blood pressure, insulin resistance, elevated blood glucose levels, smoking, poor dietary habits, and a sedentary lifestyle.²

According to the Centers for Disease Control and Prevention, about 86 million adults aged ≥ 20 years have total cholesterol levels > 200 mg/dL. More than half (54.5%) who could benefit are currently taking cholesterol-lowering medications.³ Controlling high cholesterol in American adults, especially veterans, is essential for reducing CVD morbidity and mortality.

The 2018 American College of Cardiology/American Heart Association (ACC/AHA) guideline recommends a low-density lipoprotein cholesterol (LDL-C) target goal of < 70 mg/dL for patients at high risk for ASCVD. Very high-risk ASCVD includes a history of multiple major ASCVD events or 1 major ASCVD event and multiple high-risk conditions (eg, age ≥ 65 years, smoking, or diabetes).⁴ Major ASCVD events include recent acute coronary syndrome (within the past 12 months), a history of myocardial infarction or ischemic stroke, and symptomatic peripheral artery disease. 

The ACC/AHA guideline suggests that if the LDL-C level remains ≥ 70 mg/dL, adding ezetimibe (a dietary cholesterol absorption inhibitor) to maximally tolerated statin therapy is reasonable. If LDL-C levels remain ≥ 70 mg/dL, adding a proprotein convertase subtilisin/kexin type 9 (PCSK9) inhibitor, such as alirocumab, is reasonable.⁴ The US Departments of Veterans Affairs/US Department of Defense guidelines recommend using maximally tolerated statins and ezetimibe before PCSK9 inhibitors due to established long-term safety and reduction in CVD events. 

Generic statins and ezetimibe are administered orally and widely available. In contrast, PCSK9 inhibitors have unknown long-term safety profiles, require subcutaneous injection once or twice monthly, and are significantly more expensive. They also require patient education on proper use while providing comparable or lesser relative risk reductions.²

These 3 classes of medication vary in their mechanisms of action to reduce LDL.^5,6 Ezetimibe and several statin medications are included on the Veterans Affairs Sioux Falls Health Care System (VASFHCS) formulary and do not require review prior to prescribing. Alirocumab is available at VASFHCS but is restricted to patients with a history of ASCVD or a diagnosis of familial hypercholesterolemia, and who are receiving maximally tolerated statin and ezetimibe therapy but require further LDL-C lowering to reduce their ASCVD risk. 

Studies have found ezetimibe monotherapy reduces LDL-C in patients with dyslipidemia by 18% after 12 weeks.⁷ One found that the percentage reduction in LDL-C was significantly greater (P < .001) with all doses of ezetimibe plus simvastatin (46% to 59%) compared with either atorvastatin 10 mg (37%) or simvastatin 20 mg (38%) monotherapy after 6 weeks.⁸

Although alirocumab can be added to other lipid therapies, most VASFHCS patients are prescribed alirocumab monotherapy. In the ODYSSEY CHOICE II study, patients were randomly assigned to receive either a placebo or alirocumab 150 mg every 4 weeks or alirocumab 75 mg every 2 weeks. The primary efficacy endpoint was LDL-C percentage change from baseline to week 24. In the alirocumab 150 mg every 4 weeks and 75 mg every 2 weeks groups, the least-squares mean LDL-C changes from baseline to week 24 were 51.7% and 53.5%, respectively, compared to a 4.7% increase in the placebo group (both groups P < .001 vs placebo). The authors also reported that alirocumab 150 mg every 4 weeks as monotherapy demonstrated a 47.4% reduction in LDL-C levels from baseline in a phase 1 study.⁹Although alirocumab monotherapy and ezetimibe plus statin therapy have been shown to effectively decrease LDL-C independently, a direct comparison of alirocumab monotherapy vs ezetimibe plus statin therapy has not been assessed, to our knowledge. Understanding the differences in effectiveness and safety between these 2 regimens will be valuable for clinicians when selecting a medication regimen for veterans with a history of ASCVD.

METHODS

This retrospective, single-center chart review used VASFHCS Computerized Patient Record System (CPRS) and Joint Longitudinal Viewer (JLV) records to compare patients with a history of ASCVD events who were treated with alirocumab monotherapy or ezetimibe plus statin. The 2 groups were randomized in a 1:3 ratio. The primary endpoint was achieving LDL-C < 70 mg/dL after 4 to 12 weeks, 13 to 24 weeks, and 25 to 52 weeks. Secondary endpoints included the mean percentage change from baseline in total cholesterol (TC), high-density lipoprotein cholesterol (HDL-C), LDL-C, and triglycerides (TG) over 52 weeks. The incidence of ASCVD events during this period was also assessed. If LDL-C < 70 mg/dL was achieved > 1 time during each time frame, only 1 incident was counted for analysis. Safety was assessed based on the incidence of any adverse event (AE) that led to treatment discontinuation.

Patients were identified by screening the prescription fill history between October 1, 2019, and December 31, 2022. The 52-week data collection period was counted from the first available fill date. Additionally, the prior authorization drug request file from January 1, 2017, to December 31, 2022, was used to obtain a list of patients prescribed alirocumab. Patients were included if they were veterans aged ≥ 18 years and had a history of an ASCVD event, had a alirocumab monotherapy or ezetimibe plus statin prescription between October 1, 2019, and December 31, 2022, or had an approved prior authorization drug request for alirocumab between January 1, 2017, and December 31, 2022. Patients missing a baseline or follow-up lipid panel and those with concurrent use of alirocumab and ezetimibe and/or statin were excluded. 

Baseline characteristics collected for patients included age, sex, race, weight, body mass index, lipid parameters (LDL-C, TC, HDL-C, and TG), dosing of each type of statin before adding ezetimibe, and use of any other antihyperlipidemic medication. We also collected histories of hypertension, hyperlipidemia, diabetes, chronic kidney disease, congestive heart failure, and smoking or tobacco use status. The baseline lipid panel was the most recent lipid panel documented before starting alirocumab or ezetimibe plus statin therapy. Follow-up lipid panel values were gathered at 4 to 12 weeks, 13 to 24 weeks, and 25 to 52 weeks following initiation of either therapy.

High-, moderate-, and low-intensity dosing of statin therapy and alirocumab dosing (75 mg every 2 weeks, 150 mg every 2 weeks, or 300 mg every 4 weeks) were recorded at the specified intervals. However, no patients in this study received the latter dosing regimen. ASCVD events and safety endpoints were recorded based on a review of clinical notes over the 52 weeks following the first available start date.

Statistical Analysis

The primary endpoint of achieving the LDL-C < 70 mg/dL goal from baseline to 4 to 12 weeks, 13 to 24 weeks, and 25 to 52 weeks after initiation was compared between alirocumab monotherapy and ezetimibe plus statin therapy using the χ² test. Mean percentage change from baseline in TC, HDL-C, LDL-C, and TG were compared using the independent t test. P < .05 was considered statistically significant. Incidence of ASCVD events and the safety endpoint (incidence of AEs leading to treatment discontinuation) were also compared using the χ² test. Continuous baseline characteristics were reported mean (SD) and nominal baseline characteristics were reported as a percentage.

RESULTS

There were 80 participants in this study: 20 in the alirocumab monotherapy group and 60 in the ezetimibe plus statin therapy group. More than 100 patients did not meet the prespecified inclusion criteria and were excluded. Mean (SD) age was 75 (8) years in the alirocumab group and 74 (8) years in the ezetimibe plus statin group. There was no significant differences in mean (SD) weight or mean (SD) body mass index. All study participants identified as White and male except for 2 patients in the ezetimibe plus statin therapy group whose race was not documented. Differences in lipid parameters were observed between groups, with mean baseline LDL-C, HDL-C, and TC higher in the alirocumab monotherapy group than in the ezetimibe plus statin therapy group, with significant differences in LDL-C and TC (Table 1).

Fourteen patients (70%) in the alirocumab monotherapy group had hypertension, compared with 31 (52%) in the ezetimibe plus statin therapy group. In both groups, most patients had previously been diagnosed with hyperlipidemia. More patients (60%) in the alirocumab group had diabetes than in the ezetimibe plus statin therapy group (37%). The alirocumab monotherapy group also had a higher percentage of patients with diagnoses of congestive heart failure and used other antihyperlipidemic medications than in the ezetimibe plus statin therapy group. Five patients (25%) in the alirocumab monotherapy group and 12 patients (20%) in the ezetimibe plus statin therapy group took fish oil. In the ezetimibe plus statin therapy group, 2 patients (3%) took gemfibrozil, and 2 patients (3%) took fenofibrate. Six (30%) patients in the alirocumab monotherapy group and 12 (20%) patients in the ezetimibe plus statin therapy group had chronic kidney disease. Although the majority of patients in each group did not use tobacco products, there were more tobacco users in the ezetimibe plus statin therapy group.

In the alirocumab monotherapy group, 15 patients (75%) were prescribed 75 mg every 2 weeks and 5 patients (25%) were prescribed 150 mg every 2 weeks. In the ezetimibe plus statin therapy group, 59 patients (98%) were prescribed ezetimibe 10 mg/d (Table 2). Forty-three patients (72%) were prescribed a high-intensity statin 10 received moderate-intensity (17%) and 7 received low-intensity statin (12%). Most patients were prescribed rosuvastatin (45%), followed by atorvastatin (42%), pravastatin (10%), and simvastatin (3%).

Primary Endpoint

During the 52-week study, more patients met the LDL-C goal of < 70 mg/dL in the alirocumab monotherapy group (70%) than in the ezetimibe plus statin therapy group (57%); however, the difference was not significant (P = .29). Of the patients prescribed alirocumab monotherapy who achieved LDL-C < 70 mg/dL, 15% achieved this goal in 4 to 12 weeks, 40% in 13 to 24 weeks, and 45% in 25 to 52 weeks. In the ezetimibe plus statin therapy group, 28% of patients achieved LDL-C < 70 mg/dL in 4 to 12 weeks, 31% in 13 to 24 weeks, and 41% in 25 to 52 weeks (Table 3).

Secondary Endpoints

During weeks 4 to 52 of treatment, the mean percentage change decreased in LDL-C (37.7% vs 21.4%; P = .01), TC (24.7% vs 12.5%; P = .01), and TG (0.9% vs 7.0%; P = .28) in the alirocumab monotherapy group and the ezetimibe plus statin therapy group, respectively (Table 4). The mean percentage change increased in HDL-C by 3.6% in the alirocumab monotherapy group and 1.8% in the ezetimibe plus statin therapy group (P = .36). During the study, ASCVD events occurred in 1 patient (5%) in the alirocumab monotherapy group and 3 patients (5%) in the ezetimibe plus statin therapy group (P = .99). The patient in the alirocumab monotherapy group had unstable angina 1 month after taking alirocumab. One patient in the ezetimibe plus statin therapy group had coronary artery disease and 2 patients had coronary heart disease that required stents during the 52-week period. There was 1 patient in each group who reported an AE that led to treatment discontinuation (P = .41). One patient stopped alirocumab after a trial of 2 months due to intolerance, but no specific AE was reported in the CPRS. In the ezetimibe plus statin therapy group, 1 patient requested to discontinue ezetimibe after a trial of 3 months without a specific reason noted in the medical record.

DISCUSSION

This study found no statistically significant difference in the incidence of reaching an LDL-C goal of < 70 mg/dL after alirocumab monotherapy initiation compared with ezetimibe plus statin therapy. This occurred despite baseline LDL-C being lower in the ezetimibe plus statin therapy group, which required a smaller reduction in LDL-C to reach the primary goal. Most patients on alirocumab monotherapy were prescribed a lower initial dose of 75 mg every 2 weeks. Of those patients, 30% did not achieve the LDL-C goal < 70 mg/dL. Thus, a higher dose may have led to more patients achieving the LDL-C goal.

Secondary endpoints, including mean percentage change in HDL-C and TG and incidence of ASCVD events during 52 weeks of treatment, were not statistically significant. The mean percentage increase in HDL-C was negligible in both groups, while the mean percentage reduction in TG favored the ezetimibe plus statin therapy group. In the ezetimibe plus statin therapy group, patients who also took fenofibrate experienced a significant reduction in TG while none of the patients in the alirocumab group were prescribed fenofibrate. Although the alirocumab monotherapy group had a statistically significant greater reduction in LDL-C and TC compared with those prescribed ezetimibe plus statin, the mean baseline LDL-C and TC were significantly greater in the alirocumab monotherapy group, which could contribute to higher reductions in LDL-C and TC after alirocumab monotherapy.Based on the available literature, we expected greater reductions in LDL-C in both study groups compared with statin therapy alone.^8,9 However, it was unclear whether the LDL-C and TC reductions were clinically significant.

Limitations

The study design did not permit randomization prior to the treatments, restricting our ability to account for some confounding factors, such as diet, exercise, other antihyperlipidemic medication, and medication adherence, which may have affected LDL-C, HDL-C, TG, and TC levels. Differences in baseline characteristics—particularly major risk factors, such as hypertension, diabetes, and tobacco use—also could have confounding affect on lipid levels and ASCVD events. Additionally, patients prescribed alirocumab monotherapy may have switched from statin or ezetimibe therapy, and the washout period was not reviewed or recorded, which could have affected the lipid panel results.

The small sample size of this study also may have limited the ability to detect significant differences between groups. A direct comparison of alirocumab monotherapy vs ezetimibe plus statin therapy has not been performed, making it difficult to prospectively evaluate what sample size would be needed to power this study. A posthoc analysis was used to calculate power, which was found to be only 17%. Many patients were excluded due to a lack of laboratory results within the study period, contributing to the small sample size. 

Another limitation was the reliance on documentation in CPRS and JLV. For example, having documentation of the specific AEs for the 2 patients who discontinued alirocumab or ezetimibe could have helped determine the severity of the AEs. Several patients were followed by non-VA clinicians, which could have contributed to limited documentation in the CPRS and JLV. It is difficult to draw any conclusions regarding ASCVD events and AEs that led to treatment discontinuation between alirocumab monotherapy and ezetimibe plus statin therapy based on the results of this retrospective study due to the limited number of events within the 52-week period.

CONCLUSIONS

This study found that there was no statistically significant difference in LDL-C reduction to < 70 mg/dL between alirocumab monotherapy and ezetimibe plus statin therapy in a small population of veterans with ASCVD, with a higher percentage of participants in both groups achieving that goal in 25 to 52 weeks. There also was no significant difference in percentage change in HDL-C or TG or in incidence of ASCVD events and AEs leading to treatment discontinuation. However, there was a statistically significant difference in percentage reduction for LDL-C and TC during 52 weeks of alirocumab monotherapy vs ezetimibe plus statin therapy.

Although there was no significant difference in LDL-C reduction to < 70 mg/dL, targeting this goal in patients with ASCVD is still clinically warranted. This study does not support a change in current VA criteria for use of alirocumab or a change in current guidelines for secondary prevention of ASCVD. Still, this study does indicate that the efficacy of alirocumab monotherapy is similar to that of ezetimibe plus statin therapy in patients with a history of ASCVD and may be useful in clinical settings when an alternative to ezetimibe plus statin therapy is needed. Alirocumab also may be more effective in lowering LDL-C and TC than ezetimibe plus statin therapy in veterans with ASCVD and could be added to statin therapy or ezetimibe when additional LDL-C or TC reduction is needed.

Atherosclerotic cardiovascular disease (ASCVD) is a significant cause of morbidity and mortality in the United States. ASCVD involves the buildup of cholesterol plaque in arteries and includes acute coronary syndrome, peripheral arterial disease, and events such as myocardial infarction and stroke.¹ Cardiovascular disease (CVD) risk factors include high cholesterol levels, elevated blood pressure, insulin resistance, elevated blood glucose levels, smoking, poor dietary habits, and a sedentary lifestyle.²

According to the Centers for Disease Control and Prevention, about 86 million adults aged ≥ 20 years have total cholesterol levels > 200 mg/dL. More than half (54.5%) who could benefit are currently taking cholesterol-lowering medications.³ Controlling high cholesterol in American adults, especially veterans, is essential for reducing CVD morbidity and mortality.

The 2018 American College of Cardiology/American Heart Association (ACC/AHA) guideline recommends a low-density lipoprotein cholesterol (LDL-C) target goal of < 70 mg/dL for patients at high risk for ASCVD. Very high-risk ASCVD includes a history of multiple major ASCVD events or 1 major ASCVD event and multiple high-risk conditions (eg, age ≥ 65 years, smoking, or diabetes).⁴ Major ASCVD events include recent acute coronary syndrome (within the past 12 months), a history of myocardial infarction or ischemic stroke, and symptomatic peripheral artery disease. 

The ACC/AHA guideline suggests that if the LDL-C level remains ≥ 70 mg/dL, adding ezetimibe (a dietary cholesterol absorption inhibitor) to maximally tolerated statin therapy is reasonable. If LDL-C levels remain ≥ 70 mg/dL, adding a proprotein convertase subtilisin/kexin type 9 (PCSK9) inhibitor, such as alirocumab, is reasonable.⁴ The US Departments of Veterans Affairs/US Department of Defense guidelines recommend using maximally tolerated statins and ezetimibe before PCSK9 inhibitors due to established long-term safety and reduction in CVD events. 

Generic statins and ezetimibe are administered orally and widely available. In contrast, PCSK9 inhibitors have unknown long-term safety profiles, require subcutaneous injection once or twice monthly, and are significantly more expensive. They also require patient education on proper use while providing comparable or lesser relative risk reductions.²

These 3 classes of medication vary in their mechanisms of action to reduce LDL.^5,6 Ezetimibe and several statin medications are included on the Veterans Affairs Sioux Falls Health Care System (VASFHCS) formulary and do not require review prior to prescribing. Alirocumab is available at VASFHCS but is restricted to patients with a history of ASCVD or a diagnosis of familial hypercholesterolemia, and who are receiving maximally tolerated statin and ezetimibe therapy but require further LDL-C lowering to reduce their ASCVD risk. 

Studies have found ezetimibe monotherapy reduces LDL-C in patients with dyslipidemia by 18% after 12 weeks.⁷ One found that the percentage reduction in LDL-C was significantly greater (P < .001) with all doses of ezetimibe plus simvastatin (46% to 59%) compared with either atorvastatin 10 mg (37%) or simvastatin 20 mg (38%) monotherapy after 6 weeks.⁸

Although alirocumab can be added to other lipid therapies, most VASFHCS patients are prescribed alirocumab monotherapy. In the ODYSSEY CHOICE II study, patients were randomly assigned to receive either a placebo or alirocumab 150 mg every 4 weeks or alirocumab 75 mg every 2 weeks. The primary efficacy endpoint was LDL-C percentage change from baseline to week 24. In the alirocumab 150 mg every 4 weeks and 75 mg every 2 weeks groups, the least-squares mean LDL-C changes from baseline to week 24 were 51.7% and 53.5%, respectively, compared to a 4.7% increase in the placebo group (both groups P < .001 vs placebo). The authors also reported that alirocumab 150 mg every 4 weeks as monotherapy demonstrated a 47.4% reduction in LDL-C levels from baseline in a phase 1 study.⁹Although alirocumab monotherapy and ezetimibe plus statin therapy have been shown to effectively decrease LDL-C independently, a direct comparison of alirocumab monotherapy vs ezetimibe plus statin therapy has not been assessed, to our knowledge. Understanding the differences in effectiveness and safety between these 2 regimens will be valuable for clinicians when selecting a medication regimen for veterans with a history of ASCVD.

METHODS

This retrospective, single-center chart review used VASFHCS Computerized Patient Record System (CPRS) and Joint Longitudinal Viewer (JLV) records to compare patients with a history of ASCVD events who were treated with alirocumab monotherapy or ezetimibe plus statin. The 2 groups were randomized in a 1:3 ratio. The primary endpoint was achieving LDL-C < 70 mg/dL after 4 to 12 weeks, 13 to 24 weeks, and 25 to 52 weeks. Secondary endpoints included the mean percentage change from baseline in total cholesterol (TC), high-density lipoprotein cholesterol (HDL-C), LDL-C, and triglycerides (TG) over 52 weeks. The incidence of ASCVD events during this period was also assessed. If LDL-C < 70 mg/dL was achieved > 1 time during each time frame, only 1 incident was counted for analysis. Safety was assessed based on the incidence of any adverse event (AE) that led to treatment discontinuation.

Patients were identified by screening the prescription fill history between October 1, 2019, and December 31, 2022. The 52-week data collection period was counted from the first available fill date. Additionally, the prior authorization drug request file from January 1, 2017, to December 31, 2022, was used to obtain a list of patients prescribed alirocumab. Patients were included if they were veterans aged ≥ 18 years and had a history of an ASCVD event, had a alirocumab monotherapy or ezetimibe plus statin prescription between October 1, 2019, and December 31, 2022, or had an approved prior authorization drug request for alirocumab between January 1, 2017, and December 31, 2022. Patients missing a baseline or follow-up lipid panel and those with concurrent use of alirocumab and ezetimibe and/or statin were excluded. 

Baseline characteristics collected for patients included age, sex, race, weight, body mass index, lipid parameters (LDL-C, TC, HDL-C, and TG), dosing of each type of statin before adding ezetimibe, and use of any other antihyperlipidemic medication. We also collected histories of hypertension, hyperlipidemia, diabetes, chronic kidney disease, congestive heart failure, and smoking or tobacco use status. The baseline lipid panel was the most recent lipid panel documented before starting alirocumab or ezetimibe plus statin therapy. Follow-up lipid panel values were gathered at 4 to 12 weeks, 13 to 24 weeks, and 25 to 52 weeks following initiation of either therapy.

High-, moderate-, and low-intensity dosing of statin therapy and alirocumab dosing (75 mg every 2 weeks, 150 mg every 2 weeks, or 300 mg every 4 weeks) were recorded at the specified intervals. However, no patients in this study received the latter dosing regimen. ASCVD events and safety endpoints were recorded based on a review of clinical notes over the 52 weeks following the first available start date.

Statistical Analysis

The primary endpoint of achieving the LDL-C < 70 mg/dL goal from baseline to 4 to 12 weeks, 13 to 24 weeks, and 25 to 52 weeks after initiation was compared between alirocumab monotherapy and ezetimibe plus statin therapy using the χ² test. Mean percentage change from baseline in TC, HDL-C, LDL-C, and TG were compared using the independent t test. P < .05 was considered statistically significant. Incidence of ASCVD events and the safety endpoint (incidence of AEs leading to treatment discontinuation) were also compared using the χ² test. Continuous baseline characteristics were reported mean (SD) and nominal baseline characteristics were reported as a percentage.

RESULTS

There were 80 participants in this study: 20 in the alirocumab monotherapy group and 60 in the ezetimibe plus statin therapy group. More than 100 patients did not meet the prespecified inclusion criteria and were excluded. Mean (SD) age was 75 (8) years in the alirocumab group and 74 (8) years in the ezetimibe plus statin group. There was no significant differences in mean (SD) weight or mean (SD) body mass index. All study participants identified as White and male except for 2 patients in the ezetimibe plus statin therapy group whose race was not documented. Differences in lipid parameters were observed between groups, with mean baseline LDL-C, HDL-C, and TC higher in the alirocumab monotherapy group than in the ezetimibe plus statin therapy group, with significant differences in LDL-C and TC (Table 1).

Fourteen patients (70%) in the alirocumab monotherapy group had hypertension, compared with 31 (52%) in the ezetimibe plus statin therapy group. In both groups, most patients had previously been diagnosed with hyperlipidemia. More patients (60%) in the alirocumab group had diabetes than in the ezetimibe plus statin therapy group (37%). The alirocumab monotherapy group also had a higher percentage of patients with diagnoses of congestive heart failure and used other antihyperlipidemic medications than in the ezetimibe plus statin therapy group. Five patients (25%) in the alirocumab monotherapy group and 12 patients (20%) in the ezetimibe plus statin therapy group took fish oil. In the ezetimibe plus statin therapy group, 2 patients (3%) took gemfibrozil, and 2 patients (3%) took fenofibrate. Six (30%) patients in the alirocumab monotherapy group and 12 (20%) patients in the ezetimibe plus statin therapy group had chronic kidney disease. Although the majority of patients in each group did not use tobacco products, there were more tobacco users in the ezetimibe plus statin therapy group.

In the alirocumab monotherapy group, 15 patients (75%) were prescribed 75 mg every 2 weeks and 5 patients (25%) were prescribed 150 mg every 2 weeks. In the ezetimibe plus statin therapy group, 59 patients (98%) were prescribed ezetimibe 10 mg/d (Table 2). Forty-three patients (72%) were prescribed a high-intensity statin 10 received moderate-intensity (17%) and 7 received low-intensity statin (12%). Most patients were prescribed rosuvastatin (45%), followed by atorvastatin (42%), pravastatin (10%), and simvastatin (3%).

Primary Endpoint

During the 52-week study, more patients met the LDL-C goal of < 70 mg/dL in the alirocumab monotherapy group (70%) than in the ezetimibe plus statin therapy group (57%); however, the difference was not significant (P = .29). Of the patients prescribed alirocumab monotherapy who achieved LDL-C < 70 mg/dL, 15% achieved this goal in 4 to 12 weeks, 40% in 13 to 24 weeks, and 45% in 25 to 52 weeks. In the ezetimibe plus statin therapy group, 28% of patients achieved LDL-C < 70 mg/dL in 4 to 12 weeks, 31% in 13 to 24 weeks, and 41% in 25 to 52 weeks (Table 3).

Secondary Endpoints

During weeks 4 to 52 of treatment, the mean percentage change decreased in LDL-C (37.7% vs 21.4%; P = .01), TC (24.7% vs 12.5%; P = .01), and TG (0.9% vs 7.0%; P = .28) in the alirocumab monotherapy group and the ezetimibe plus statin therapy group, respectively (Table 4). The mean percentage change increased in HDL-C by 3.6% in the alirocumab monotherapy group and 1.8% in the ezetimibe plus statin therapy group (P = .36). During the study, ASCVD events occurred in 1 patient (5%) in the alirocumab monotherapy group and 3 patients (5%) in the ezetimibe plus statin therapy group (P = .99). The patient in the alirocumab monotherapy group had unstable angina 1 month after taking alirocumab. One patient in the ezetimibe plus statin therapy group had coronary artery disease and 2 patients had coronary heart disease that required stents during the 52-week period. There was 1 patient in each group who reported an AE that led to treatment discontinuation (P = .41). One patient stopped alirocumab after a trial of 2 months due to intolerance, but no specific AE was reported in the CPRS. In the ezetimibe plus statin therapy group, 1 patient requested to discontinue ezetimibe after a trial of 3 months without a specific reason noted in the medical record.

DISCUSSION

This study found no statistically significant difference in the incidence of reaching an LDL-C goal of < 70 mg/dL after alirocumab monotherapy initiation compared with ezetimibe plus statin therapy. This occurred despite baseline LDL-C being lower in the ezetimibe plus statin therapy group, which required a smaller reduction in LDL-C to reach the primary goal. Most patients on alirocumab monotherapy were prescribed a lower initial dose of 75 mg every 2 weeks. Of those patients, 30% did not achieve the LDL-C goal < 70 mg/dL. Thus, a higher dose may have led to more patients achieving the LDL-C goal.

Secondary endpoints, including mean percentage change in HDL-C and TG and incidence of ASCVD events during 52 weeks of treatment, were not statistically significant. The mean percentage increase in HDL-C was negligible in both groups, while the mean percentage reduction in TG favored the ezetimibe plus statin therapy group. In the ezetimibe plus statin therapy group, patients who also took fenofibrate experienced a significant reduction in TG while none of the patients in the alirocumab group were prescribed fenofibrate. Although the alirocumab monotherapy group had a statistically significant greater reduction in LDL-C and TC compared with those prescribed ezetimibe plus statin, the mean baseline LDL-C and TC were significantly greater in the alirocumab monotherapy group, which could contribute to higher reductions in LDL-C and TC after alirocumab monotherapy.Based on the available literature, we expected greater reductions in LDL-C in both study groups compared with statin therapy alone.^8,9 However, it was unclear whether the LDL-C and TC reductions were clinically significant.

Limitations

The study design did not permit randomization prior to the treatments, restricting our ability to account for some confounding factors, such as diet, exercise, other antihyperlipidemic medication, and medication adherence, which may have affected LDL-C, HDL-C, TG, and TC levels. Differences in baseline characteristics—particularly major risk factors, such as hypertension, diabetes, and tobacco use—also could have confounding affect on lipid levels and ASCVD events. Additionally, patients prescribed alirocumab monotherapy may have switched from statin or ezetimibe therapy, and the washout period was not reviewed or recorded, which could have affected the lipid panel results.

The small sample size of this study also may have limited the ability to detect significant differences between groups. A direct comparison of alirocumab monotherapy vs ezetimibe plus statin therapy has not been performed, making it difficult to prospectively evaluate what sample size would be needed to power this study. A posthoc analysis was used to calculate power, which was found to be only 17%. Many patients were excluded due to a lack of laboratory results within the study period, contributing to the small sample size. 

Another limitation was the reliance on documentation in CPRS and JLV. For example, having documentation of the specific AEs for the 2 patients who discontinued alirocumab or ezetimibe could have helped determine the severity of the AEs. Several patients were followed by non-VA clinicians, which could have contributed to limited documentation in the CPRS and JLV. It is difficult to draw any conclusions regarding ASCVD events and AEs that led to treatment discontinuation between alirocumab monotherapy and ezetimibe plus statin therapy based on the results of this retrospective study due to the limited number of events within the 52-week period.

CONCLUSIONS

This study found that there was no statistically significant difference in LDL-C reduction to < 70 mg/dL between alirocumab monotherapy and ezetimibe plus statin therapy in a small population of veterans with ASCVD, with a higher percentage of participants in both groups achieving that goal in 25 to 52 weeks. There also was no significant difference in percentage change in HDL-C or TG or in incidence of ASCVD events and AEs leading to treatment discontinuation. However, there was a statistically significant difference in percentage reduction for LDL-C and TC during 52 weeks of alirocumab monotherapy vs ezetimibe plus statin therapy.

Although there was no significant difference in LDL-C reduction to < 70 mg/dL, targeting this goal in patients with ASCVD is still clinically warranted. This study does not support a change in current VA criteria for use of alirocumab or a change in current guidelines for secondary prevention of ASCVD. Still, this study does indicate that the efficacy of alirocumab monotherapy is similar to that of ezetimibe plus statin therapy in patients with a history of ASCVD and may be useful in clinical settings when an alternative to ezetimibe plus statin therapy is needed. Alirocumab also may be more effective in lowering LDL-C and TC than ezetimibe plus statin therapy in veterans with ASCVD and could be added to statin therapy or ezetimibe when additional LDL-C or TC reduction is needed.

Individuals with diabetes are at risk for developing foot ulcers or full-thickness defects in the epithelium of the foot. These defects can lead to bacterial invasion and foot infection, potentially resulting in leg amputation (Figure 1). Effective treatment to prevent leg amputation, known as limb salvage, requires management across multiple medical specialties including podiatry, vascular surgery, and infectious diseases. The multidisciplinary team approach to limb salvage was introduced in Boston in 1928 and has been the prevailing approach to this cross-specialty medical problem for at least a decade.^1,2

The Michael E. DeBakey Veterans Affairs Medical Center (MEDVAMC) has established an inpatient limb salvage program—a group of dedicated clinicians working collaboratively to provide evidence-guided management of patients hospitalized with foot ulcers, foot gangrene or any superimposed infection with the goal of avoiding leg amputations. We have seen a significant and durable reduction in the incidence of leg amputations among veterans at MEDVAMC.

This article describes the evolution and outcomes of the MEDVAMC limb salvage program over more than a decade. It includes changes to team structure and workflow, as well as past and present successes and challenges. The eAppendix provides a narrative summary with examples of how our clinical practice and research efforts have informed one another and how these findings are applied to clinical management. This process is part of the larger efforts of the Veterans Health Administration (VHA) to create a learning health system in which “internal data and experience are systematically integrated with external evidence, and that knowledge is put into practice.”³

Methods

Data from the VHA Support Service Center were used to obtain monthly major (leg) and minor (toe and partial foot) amputation records at MEDVAMC from October 2000 through May 2023. Yearly totals for the number of persons with diabetes and foot ulcers at MEDVAMC were also obtained from the support service center. Annual patient population sizes and number of persons with foot ulcers were converted to monthly estimates using cubic spline interpolation. Rates were calculated as 12-month rolling averages. Trend lines were created with locally weighted running line smoothing that used a span α of 0.1.

We characterized the patient population using data from cohorts of veterans treated for foot ulcers and foot infections at MEDVAMC. To compare the contemporary veteran population with nonveteran inpatients treated for foot ulcers and foot infections at other hospitals, we created a 2:1 nonveteran to veteran cohort matched by sex and zip code, using publicly available hospital admission data from the Texas Department of Health and State Health Services. Veterans used for this cohort comparison are consistent with the 100 consecutive patients who underwent angiography for limb salvage in 2022.

This research was approved by the Baylor College of Medicine Institutional Review Board (protocol H-34858) and the MEDVAMC Research Committee (IRBNet protocol 15A12. HB). All analyses used deidentified data in the R programming language version 4.2.2 using RStudio version 2022.06.0 Build 421.

Program Description

MEDVAMC is a 350-bed teaching hospital located in central Houston. Its hospital system includes 11 outpatient clinics, ranging from 28 to 126 miles (eAppendix, Supplemental Figure A) from MEDVAMC. MEDVAMC provides vascular, orthopedic, and podiatric surgery services, as well as many other highly specialized services such as liver and heart transplants. The hospital’s risk-adjusted rates of operative morbidity and mortality (observed-to-expected ratios) are significantly lower than expected.

Despite this, the incidence rate of leg amputations at MEDVAMC in early 2011 was nearly 3-times higher than the VHA average. The inpatient management of veterans with infected foot ulcers was fragmented, with the general, orthopedic, and vascular surgery teams separately providing siloed care. Delays in treatment were common. There was much service- and practitioner-level practice heterogeneity. No diagnostic or treatment protocols were used, and standard treatment components were sporadically provided.

Patient Population

Compared to the matched non-VHA patient cohort (Supplemental Table 1), veterans treated at MEDVAMC for limb salvage are older. Nearly half (46%) identify as Black, which is associated with a 2-fold higher riskadjusted rate of leg amputations.⁴ MEDVAMC patients also have significantly higher rates of diabetes, chronic kidney disease, and systolic heart failure. About 22% travel > 40 miles for treatment at MEDVAMC, double that of the matched cohort (10.7%). Additionally, 35% currently smoke and 37% have moderate to severe peripheral artery disease (PAD).⁵

Program Design

In late 2011, the MEDVAMC vascular surgery team led limb salvage efforts by implementing a single team model, which involved assuming the primary role of managing foot ulcers for all veterans, both infected and uninfected (eAppendix, Supplemental Figure B). Consultations were directed to a dedicated limb salvage pager. The vascular team provided interdisciplinary limb salvage management across the spectrum of disease, including the surgical treatment of infection, assessment for PAD, open surgical operations and endovascular interventions to treat PAD, and foot reconstruction (debridement, minor or partial foot amputations, and skin grafting). This care was complemented by frequent consultation with the infectious disease, vascular medicine, podiatry, and geriatric wound care teams. This approach streamlined the delivery of consistent multidisciplinary care.

This collaborative effort aimed to develop ideal multidisciplinary care plans through research spanning the spectrum of the diabetic foot infection disease process (eAppendix, Supplemental Table 1). Some of the most impactful practices were: (1) a proclivity towards surgical treatment of foot infections, especially osteomyelitis⁵; (2) improved identification of PAD^6,7; (3) early surgical closure of foot wounds following revascularization^8,9; and (4) palliative wound care as an alternative to leg amputation in veterans who are not candidates for revascularization and limb salvage.¹⁰ Initally, the vascular surgery team held monthly multidisciplinary limb salvage meetings to coordinate patient management, identify ways to streamline care and avoid waste, discuss research findings, and review the 12-month rolling average of the MEDVAMC leg amputation incidence rate.

During the study period, the MEDVAMC vascular surgery team consisted of 2 to 5 board certified vascular or general surgeons, 2 or 3 nurse practitioners, and 3 vascular ultrasound technologists. Associated specialists included 2 podiatrists, 3 geriatricians with wound care certification, as well as additional infectious diseases, vascular medicine, orthopedics, and general surgery specialists.

Program Assessment

We noted a significant and sustained decrease in the MEDVAMC leg amputation rate after implementing multidisciplinary meetings and a single- team model from early 2012 through 2017 (Figure 2). The amputation incidence rate decreased steadily over the period from a maximum of 160 per 100,000 per year in February 2012 to a nadir of 66 per 100,000 per year in April 2017, an overall 60% decrease. Increases were noted in early 2018 after ceasing the single- team model, and in the summer of 2022, following periods of bed shortages after the onset of the COVID-19 pandemic. Tracking this metric allowed clinicians to make course corrections.

The decreased leg amputation rate at MEDVAMC does not seem to be mirroring national or regional trends. During this 10-year period, the VHA annualized amputation rate decreased minimally, from 58 to 54 per 100,000 (eAppendix Supplemental Figure C). Leg amputation incidence at non-VHA hospitals in Texas slightly increased over the same period.¹¹

Value was also reflected in other metrics. MEDVAMC improved safety through a bundled strategy that reduced the risk-adjusted rate of surgical wound infections by 95%.¹² MEDVAMC prioritized limb salvage when selecting patients for angiography and nearly eliminated using stent-grafts, cryopreserved allogeneic saphenous vein grafts, and expensive surgical and endovascular implants, which were identified as more expensive and less effective than other options (Figure 3).^13-15 The MEDVAMC team achieved a > 90% patient trust rating on the Veterans Signals survey in fiscal years 2021 and 2022.

Challenges

A significant increase in the patient-physician ratio occurred 5 years into the program. In 2016, 2 vascular surgeons left MEDVAMC and a planned renovation of 1 of the 2 vascular surgery-assigned hybrid working facilities began even as the number of MEDVAMC patients with diabetes grew 120% (from 89,400 to 107,746 between 2010 and 2016), and the incidence rate of foot ulcers grew 300% (from 392 in 2010 to 1183 in 2016 per 100,000). The net result was a higher clinical workload among the remaining vascular surgeons with less operating room availability.

To stabilize surgeon retention, MEDVAMC reverted from the single team model back to inpatient care being distributed among general surgery, orthopedic surgery, and vascular surgery. After noting an increase in the leg amputation incidence rate, we adjusted the focus from multidisciplinary to interdisciplinary care (ie, majority of limb salvage clinical care can be provided by practitioners of any involved specialties). We worked to establish a local, written, interdisciplinary consensus on evaluating and managing veterans with nonhealing foot ulcers to mitigate the loss of a consolidated inpatient approach. Despite frequent staff turnover, ≥ 1 physician or surgeon from the core specialties of vascular surgery, podiatry, and infectious diseases remained throughout the study period.

The COVID-19 pandemic caused a shortage of hospital beds. This was followed by more bed shortages due to decreased nursing staff. Our health care system also had a period of restricted outpatient encounters early in the pandemic. During this time, we noted a delayed presentation of veterans with advanced infections and another increase in leg amputation incidence rate.

Like many health systems, MEDVAMC pivoted to telephone- and video-based outpatient encounters. Our team also used publicly available Texas hospitalization data to identify zip codes with particularly high leg amputation incidence rates, and > 3500 educational mailings to veterans categorized as moderate and high risk for leg amputation in these zip codes. These mailings provided information on recognizing foot ulcers and infections, emphasized timely evaluation, and named the MEDVAMC vascular surgery team as a point-of-contact. More recently, we have seen a further decrease in the MEDVAMC incidences of leg amputation to its lowest rate in > 20 years.

Discussion

A learning organization that directs its research based on clinical observations and informs its clinical care with research findings can produce palpable improvements in outcomes. Understanding the disease process and trying to better understand management across the entire range of this disease process has allowed our team to make consistent and systematic changes in care (Table). Consolidating inpatient care in a single team model seems to have been effective in reducing amputation rates among veterans with diabetes. The role the MEDVAMC vascular surgery team served for limb salvage patients may have been particularly beneficial because of the large impact untreated or unidentified PAD can have and because of the high prevalence of PAD among the limb salvage population seen at MEDVAMC. To be sustainable, though, a single-team model needs resources. A multiteam model can also be effective if the degree of multidisciplinary involvement for any given veteran is appropriate to the individual's clinical needs, teams are engaged and willing to contribute in a defined role within their specialty, and lines of communication remain open.

The primary challenge at MEDVAMC has been, and will continue to be, the retention of physicians and surgeons. MEDVAMC has excellent leadership and a collegial working environment, but better access to operating rooms for elective and time-sensitive operations, additional clinical staff support, and higher salary at non-VA positions have been the basis for many of physicians— especially surgeons—leaving MEDVAMC. Despite high staff turnover and a constant flow of resident and fellow trainees, MEDVAMC has been able to keep the clinical approach relatively consistent due to the use of written protocols and continuity of care as ≥ 1 physician or surgeon from each of the 4 main teams remained engaged with limb salvage throughout the entire period.

Going forward, we will work to ensure that all requirements of the 2022 Prevention of Amputation in Veterans Everywhere directive are incorporated into care.⁸ We plan to standardize MEDVAMC management algorithms further, both to streamline care and reduce the opportunity for disparities in treatment. More prophylactic podiatric procedures, surgical forms of offloading, and a shared multidisciplinary clinic space may also further help patients.

Conclusions

The introduction of multidisciplinary limb salvage at MEDVAMC has led to significant and sustained reductions in leg amputation incidence. These reductions do not seem dependent upon a specific team structure for inpatient care. To improve patient outcomes, efforts should focus on making improvements across the entire disease spectrum. For limb salvage, this includes primary prevention of foot ulcers, the treatment of foot infections, identification and management of PAD, surgical reconstruction/optimal wound healing, and care for patients who undergo leg amputation.

Individuals with diabetes are at risk for developing foot ulcers or full-thickness defects in the epithelium of the foot. These defects can lead to bacterial invasion and foot infection, potentially resulting in leg amputation (Figure 1). Effective treatment to prevent leg amputation, known as limb salvage, requires management across multiple medical specialties including podiatry, vascular surgery, and infectious diseases. The multidisciplinary team approach to limb salvage was introduced in Boston in 1928 and has been the prevailing approach to this cross-specialty medical problem for at least a decade.^1,2

The Michael E. DeBakey Veterans Affairs Medical Center (MEDVAMC) has established an inpatient limb salvage program—a group of dedicated clinicians working collaboratively to provide evidence-guided management of patients hospitalized with foot ulcers, foot gangrene or any superimposed infection with the goal of avoiding leg amputations. We have seen a significant and durable reduction in the incidence of leg amputations among veterans at MEDVAMC.

This article describes the evolution and outcomes of the MEDVAMC limb salvage program over more than a decade. It includes changes to team structure and workflow, as well as past and present successes and challenges. The eAppendix provides a narrative summary with examples of how our clinical practice and research efforts have informed one another and how these findings are applied to clinical management. This process is part of the larger efforts of the Veterans Health Administration (VHA) to create a learning health system in which “internal data and experience are systematically integrated with external evidence, and that knowledge is put into practice.”³

Methods

Data from the VHA Support Service Center were used to obtain monthly major (leg) and minor (toe and partial foot) amputation records at MEDVAMC from October 2000 through May 2023. Yearly totals for the number of persons with diabetes and foot ulcers at MEDVAMC were also obtained from the support service center. Annual patient population sizes and number of persons with foot ulcers were converted to monthly estimates using cubic spline interpolation. Rates were calculated as 12-month rolling averages. Trend lines were created with locally weighted running line smoothing that used a span α of 0.1.

We characterized the patient population using data from cohorts of veterans treated for foot ulcers and foot infections at MEDVAMC. To compare the contemporary veteran population with nonveteran inpatients treated for foot ulcers and foot infections at other hospitals, we created a 2:1 nonveteran to veteran cohort matched by sex and zip code, using publicly available hospital admission data from the Texas Department of Health and State Health Services. Veterans used for this cohort comparison are consistent with the 100 consecutive patients who underwent angiography for limb salvage in 2022.

This research was approved by the Baylor College of Medicine Institutional Review Board (protocol H-34858) and the MEDVAMC Research Committee (IRBNet protocol 15A12. HB). All analyses used deidentified data in the R programming language version 4.2.2 using RStudio version 2022.06.0 Build 421.

Program Description

MEDVAMC is a 350-bed teaching hospital located in central Houston. Its hospital system includes 11 outpatient clinics, ranging from 28 to 126 miles (eAppendix, Supplemental Figure A) from MEDVAMC. MEDVAMC provides vascular, orthopedic, and podiatric surgery services, as well as many other highly specialized services such as liver and heart transplants. The hospital’s risk-adjusted rates of operative morbidity and mortality (observed-to-expected ratios) are significantly lower than expected.

Despite this, the incidence rate of leg amputations at MEDVAMC in early 2011 was nearly 3-times higher than the VHA average. The inpatient management of veterans with infected foot ulcers was fragmented, with the general, orthopedic, and vascular surgery teams separately providing siloed care. Delays in treatment were common. There was much service- and practitioner-level practice heterogeneity. No diagnostic or treatment protocols were used, and standard treatment components were sporadically provided.

Patient Population

Compared to the matched non-VHA patient cohort (Supplemental Table 1), veterans treated at MEDVAMC for limb salvage are older. Nearly half (46%) identify as Black, which is associated with a 2-fold higher riskadjusted rate of leg amputations.⁴ MEDVAMC patients also have significantly higher rates of diabetes, chronic kidney disease, and systolic heart failure. About 22% travel > 40 miles for treatment at MEDVAMC, double that of the matched cohort (10.7%). Additionally, 35% currently smoke and 37% have moderate to severe peripheral artery disease (PAD).⁵

Program Design

In late 2011, the MEDVAMC vascular surgery team led limb salvage efforts by implementing a single team model, which involved assuming the primary role of managing foot ulcers for all veterans, both infected and uninfected (eAppendix, Supplemental Figure B). Consultations were directed to a dedicated limb salvage pager. The vascular team provided interdisciplinary limb salvage management across the spectrum of disease, including the surgical treatment of infection, assessment for PAD, open surgical operations and endovascular interventions to treat PAD, and foot reconstruction (debridement, minor or partial foot amputations, and skin grafting). This care was complemented by frequent consultation with the infectious disease, vascular medicine, podiatry, and geriatric wound care teams. This approach streamlined the delivery of consistent multidisciplinary care.

This collaborative effort aimed to develop ideal multidisciplinary care plans through research spanning the spectrum of the diabetic foot infection disease process (eAppendix, Supplemental Table 1). Some of the most impactful practices were: (1) a proclivity towards surgical treatment of foot infections, especially osteomyelitis⁵; (2) improved identification of PAD^6,7; (3) early surgical closure of foot wounds following revascularization^8,9; and (4) palliative wound care as an alternative to leg amputation in veterans who are not candidates for revascularization and limb salvage.¹⁰ Initally, the vascular surgery team held monthly multidisciplinary limb salvage meetings to coordinate patient management, identify ways to streamline care and avoid waste, discuss research findings, and review the 12-month rolling average of the MEDVAMC leg amputation incidence rate.

During the study period, the MEDVAMC vascular surgery team consisted of 2 to 5 board certified vascular or general surgeons, 2 or 3 nurse practitioners, and 3 vascular ultrasound technologists. Associated specialists included 2 podiatrists, 3 geriatricians with wound care certification, as well as additional infectious diseases, vascular medicine, orthopedics, and general surgery specialists.

Program Assessment

We noted a significant and sustained decrease in the MEDVAMC leg amputation rate after implementing multidisciplinary meetings and a single- team model from early 2012 through 2017 (Figure 2). The amputation incidence rate decreased steadily over the period from a maximum of 160 per 100,000 per year in February 2012 to a nadir of 66 per 100,000 per year in April 2017, an overall 60% decrease. Increases were noted in early 2018 after ceasing the single- team model, and in the summer of 2022, following periods of bed shortages after the onset of the COVID-19 pandemic. Tracking this metric allowed clinicians to make course corrections.

The decreased leg amputation rate at MEDVAMC does not seem to be mirroring national or regional trends. During this 10-year period, the VHA annualized amputation rate decreased minimally, from 58 to 54 per 100,000 (eAppendix Supplemental Figure C). Leg amputation incidence at non-VHA hospitals in Texas slightly increased over the same period.¹¹

Value was also reflected in other metrics. MEDVAMC improved safety through a bundled strategy that reduced the risk-adjusted rate of surgical wound infections by 95%.¹² MEDVAMC prioritized limb salvage when selecting patients for angiography and nearly eliminated using stent-grafts, cryopreserved allogeneic saphenous vein grafts, and expensive surgical and endovascular implants, which were identified as more expensive and less effective than other options (Figure 3).^13-15 The MEDVAMC team achieved a > 90% patient trust rating on the Veterans Signals survey in fiscal years 2021 and 2022.

Challenges

A significant increase in the patient-physician ratio occurred 5 years into the program. In 2016, 2 vascular surgeons left MEDVAMC and a planned renovation of 1 of the 2 vascular surgery-assigned hybrid working facilities began even as the number of MEDVAMC patients with diabetes grew 120% (from 89,400 to 107,746 between 2010 and 2016), and the incidence rate of foot ulcers grew 300% (from 392 in 2010 to 1183 in 2016 per 100,000). The net result was a higher clinical workload among the remaining vascular surgeons with less operating room availability.

To stabilize surgeon retention, MEDVAMC reverted from the single team model back to inpatient care being distributed among general surgery, orthopedic surgery, and vascular surgery. After noting an increase in the leg amputation incidence rate, we adjusted the focus from multidisciplinary to interdisciplinary care (ie, majority of limb salvage clinical care can be provided by practitioners of any involved specialties). We worked to establish a local, written, interdisciplinary consensus on evaluating and managing veterans with nonhealing foot ulcers to mitigate the loss of a consolidated inpatient approach. Despite frequent staff turnover, ≥ 1 physician or surgeon from the core specialties of vascular surgery, podiatry, and infectious diseases remained throughout the study period.

The COVID-19 pandemic caused a shortage of hospital beds. This was followed by more bed shortages due to decreased nursing staff. Our health care system also had a period of restricted outpatient encounters early in the pandemic. During this time, we noted a delayed presentation of veterans with advanced infections and another increase in leg amputation incidence rate.

Like many health systems, MEDVAMC pivoted to telephone- and video-based outpatient encounters. Our team also used publicly available Texas hospitalization data to identify zip codes with particularly high leg amputation incidence rates, and > 3500 educational mailings to veterans categorized as moderate and high risk for leg amputation in these zip codes. These mailings provided information on recognizing foot ulcers and infections, emphasized timely evaluation, and named the MEDVAMC vascular surgery team as a point-of-contact. More recently, we have seen a further decrease in the MEDVAMC incidences of leg amputation to its lowest rate in > 20 years.

Discussion

A learning organization that directs its research based on clinical observations and informs its clinical care with research findings can produce palpable improvements in outcomes. Understanding the disease process and trying to better understand management across the entire range of this disease process has allowed our team to make consistent and systematic changes in care (Table). Consolidating inpatient care in a single team model seems to have been effective in reducing amputation rates among veterans with diabetes. The role the MEDVAMC vascular surgery team served for limb salvage patients may have been particularly beneficial because of the large impact untreated or unidentified PAD can have and because of the high prevalence of PAD among the limb salvage population seen at MEDVAMC. To be sustainable, though, a single-team model needs resources. A multiteam model can also be effective if the degree of multidisciplinary involvement for any given veteran is appropriate to the individual's clinical needs, teams are engaged and willing to contribute in a defined role within their specialty, and lines of communication remain open.

The primary challenge at MEDVAMC has been, and will continue to be, the retention of physicians and surgeons. MEDVAMC has excellent leadership and a collegial working environment, but better access to operating rooms for elective and time-sensitive operations, additional clinical staff support, and higher salary at non-VA positions have been the basis for many of physicians— especially surgeons—leaving MEDVAMC. Despite high staff turnover and a constant flow of resident and fellow trainees, MEDVAMC has been able to keep the clinical approach relatively consistent due to the use of written protocols and continuity of care as ≥ 1 physician or surgeon from each of the 4 main teams remained engaged with limb salvage throughout the entire period.

Going forward, we will work to ensure that all requirements of the 2022 Prevention of Amputation in Veterans Everywhere directive are incorporated into care.⁸ We plan to standardize MEDVAMC management algorithms further, both to streamline care and reduce the opportunity for disparities in treatment. More prophylactic podiatric procedures, surgical forms of offloading, and a shared multidisciplinary clinic space may also further help patients.

Conclusions

The introduction of multidisciplinary limb salvage at MEDVAMC has led to significant and sustained reductions in leg amputation incidence. These reductions do not seem dependent upon a specific team structure for inpatient care. To improve patient outcomes, efforts should focus on making improvements across the entire disease spectrum. For limb salvage, this includes primary prevention of foot ulcers, the treatment of foot infections, identification and management of PAD, surgical reconstruction/optimal wound healing, and care for patients who undergo leg amputation.

Individuals with diabetes are at risk for developing foot ulcers or full-thickness defects in the epithelium of the foot. These defects can lead to bacterial invasion and foot infection, potentially resulting in leg amputation (Figure 1). Effective treatment to prevent leg amputation, known as limb salvage, requires management across multiple medical specialties including podiatry, vascular surgery, and infectious diseases. The multidisciplinary team approach to limb salvage was introduced in Boston in 1928 and has been the prevailing approach to this cross-specialty medical problem for at least a decade.^1,2

The Michael E. DeBakey Veterans Affairs Medical Center (MEDVAMC) has established an inpatient limb salvage program—a group of dedicated clinicians working collaboratively to provide evidence-guided management of patients hospitalized with foot ulcers, foot gangrene or any superimposed infection with the goal of avoiding leg amputations. We have seen a significant and durable reduction in the incidence of leg amputations among veterans at MEDVAMC.

This article describes the evolution and outcomes of the MEDVAMC limb salvage program over more than a decade. It includes changes to team structure and workflow, as well as past and present successes and challenges. The eAppendix provides a narrative summary with examples of how our clinical practice and research efforts have informed one another and how these findings are applied to clinical management. This process is part of the larger efforts of the Veterans Health Administration (VHA) to create a learning health system in which “internal data and experience are systematically integrated with external evidence, and that knowledge is put into practice.”³

Methods

Data from the VHA Support Service Center were used to obtain monthly major (leg) and minor (toe and partial foot) amputation records at MEDVAMC from October 2000 through May 2023. Yearly totals for the number of persons with diabetes and foot ulcers at MEDVAMC were also obtained from the support service center. Annual patient population sizes and number of persons with foot ulcers were converted to monthly estimates using cubic spline interpolation. Rates were calculated as 12-month rolling averages. Trend lines were created with locally weighted running line smoothing that used a span α of 0.1.

We characterized the patient population using data from cohorts of veterans treated for foot ulcers and foot infections at MEDVAMC. To compare the contemporary veteran population with nonveteran inpatients treated for foot ulcers and foot infections at other hospitals, we created a 2:1 nonveteran to veteran cohort matched by sex and zip code, using publicly available hospital admission data from the Texas Department of Health and State Health Services. Veterans used for this cohort comparison are consistent with the 100 consecutive patients who underwent angiography for limb salvage in 2022.

This research was approved by the Baylor College of Medicine Institutional Review Board (protocol H-34858) and the MEDVAMC Research Committee (IRBNet protocol 15A12. HB). All analyses used deidentified data in the R programming language version 4.2.2 using RStudio version 2022.06.0 Build 421.

Program Description

MEDVAMC is a 350-bed teaching hospital located in central Houston. Its hospital system includes 11 outpatient clinics, ranging from 28 to 126 miles (eAppendix, Supplemental Figure A) from MEDVAMC. MEDVAMC provides vascular, orthopedic, and podiatric surgery services, as well as many other highly specialized services such as liver and heart transplants. The hospital’s risk-adjusted rates of operative morbidity and mortality (observed-to-expected ratios) are significantly lower than expected.

Despite this, the incidence rate of leg amputations at MEDVAMC in early 2011 was nearly 3-times higher than the VHA average. The inpatient management of veterans with infected foot ulcers was fragmented, with the general, orthopedic, and vascular surgery teams separately providing siloed care. Delays in treatment were common. There was much service- and practitioner-level practice heterogeneity. No diagnostic or treatment protocols were used, and standard treatment components were sporadically provided.

Patient Population

Compared to the matched non-VHA patient cohort (Supplemental Table 1), veterans treated at MEDVAMC for limb salvage are older. Nearly half (46%) identify as Black, which is associated with a 2-fold higher riskadjusted rate of leg amputations.⁴ MEDVAMC patients also have significantly higher rates of diabetes, chronic kidney disease, and systolic heart failure. About 22% travel > 40 miles for treatment at MEDVAMC, double that of the matched cohort (10.7%). Additionally, 35% currently smoke and 37% have moderate to severe peripheral artery disease (PAD).⁵

Program Design

In late 2011, the MEDVAMC vascular surgery team led limb salvage efforts by implementing a single team model, which involved assuming the primary role of managing foot ulcers for all veterans, both infected and uninfected (eAppendix, Supplemental Figure B). Consultations were directed to a dedicated limb salvage pager. The vascular team provided interdisciplinary limb salvage management across the spectrum of disease, including the surgical treatment of infection, assessment for PAD, open surgical operations and endovascular interventions to treat PAD, and foot reconstruction (debridement, minor or partial foot amputations, and skin grafting). This care was complemented by frequent consultation with the infectious disease, vascular medicine, podiatry, and geriatric wound care teams. This approach streamlined the delivery of consistent multidisciplinary care.

This collaborative effort aimed to develop ideal multidisciplinary care plans through research spanning the spectrum of the diabetic foot infection disease process (eAppendix, Supplemental Table 1). Some of the most impactful practices were: (1) a proclivity towards surgical treatment of foot infections, especially osteomyelitis⁵; (2) improved identification of PAD^6,7; (3) early surgical closure of foot wounds following revascularization^8,9; and (4) palliative wound care as an alternative to leg amputation in veterans who are not candidates for revascularization and limb salvage.¹⁰ Initally, the vascular surgery team held monthly multidisciplinary limb salvage meetings to coordinate patient management, identify ways to streamline care and avoid waste, discuss research findings, and review the 12-month rolling average of the MEDVAMC leg amputation incidence rate.

During the study period, the MEDVAMC vascular surgery team consisted of 2 to 5 board certified vascular or general surgeons, 2 or 3 nurse practitioners, and 3 vascular ultrasound technologists. Associated specialists included 2 podiatrists, 3 geriatricians with wound care certification, as well as additional infectious diseases, vascular medicine, orthopedics, and general surgery specialists.

Program Assessment

We noted a significant and sustained decrease in the MEDVAMC leg amputation rate after implementing multidisciplinary meetings and a single- team model from early 2012 through 2017 (Figure 2). The amputation incidence rate decreased steadily over the period from a maximum of 160 per 100,000 per year in February 2012 to a nadir of 66 per 100,000 per year in April 2017, an overall 60% decrease. Increases were noted in early 2018 after ceasing the single- team model, and in the summer of 2022, following periods of bed shortages after the onset of the COVID-19 pandemic. Tracking this metric allowed clinicians to make course corrections.

The decreased leg amputation rate at MEDVAMC does not seem to be mirroring national or regional trends. During this 10-year period, the VHA annualized amputation rate decreased minimally, from 58 to 54 per 100,000 (eAppendix Supplemental Figure C). Leg amputation incidence at non-VHA hospitals in Texas slightly increased over the same period.¹¹

Value was also reflected in other metrics. MEDVAMC improved safety through a bundled strategy that reduced the risk-adjusted rate of surgical wound infections by 95%.¹² MEDVAMC prioritized limb salvage when selecting patients for angiography and nearly eliminated using stent-grafts, cryopreserved allogeneic saphenous vein grafts, and expensive surgical and endovascular implants, which were identified as more expensive and less effective than other options (Figure 3).^13-15 The MEDVAMC team achieved a > 90% patient trust rating on the Veterans Signals survey in fiscal years 2021 and 2022.

Challenges

A significant increase in the patient-physician ratio occurred 5 years into the program. In 2016, 2 vascular surgeons left MEDVAMC and a planned renovation of 1 of the 2 vascular surgery-assigned hybrid working facilities began even as the number of MEDVAMC patients with diabetes grew 120% (from 89,400 to 107,746 between 2010 and 2016), and the incidence rate of foot ulcers grew 300% (from 392 in 2010 to 1183 in 2016 per 100,000). The net result was a higher clinical workload among the remaining vascular surgeons with less operating room availability.

To stabilize surgeon retention, MEDVAMC reverted from the single team model back to inpatient care being distributed among general surgery, orthopedic surgery, and vascular surgery. After noting an increase in the leg amputation incidence rate, we adjusted the focus from multidisciplinary to interdisciplinary care (ie, majority of limb salvage clinical care can be provided by practitioners of any involved specialties). We worked to establish a local, written, interdisciplinary consensus on evaluating and managing veterans with nonhealing foot ulcers to mitigate the loss of a consolidated inpatient approach. Despite frequent staff turnover, ≥ 1 physician or surgeon from the core specialties of vascular surgery, podiatry, and infectious diseases remained throughout the study period.

The COVID-19 pandemic caused a shortage of hospital beds. This was followed by more bed shortages due to decreased nursing staff. Our health care system also had a period of restricted outpatient encounters early in the pandemic. During this time, we noted a delayed presentation of veterans with advanced infections and another increase in leg amputation incidence rate.

Like many health systems, MEDVAMC pivoted to telephone- and video-based outpatient encounters. Our team also used publicly available Texas hospitalization data to identify zip codes with particularly high leg amputation incidence rates, and > 3500 educational mailings to veterans categorized as moderate and high risk for leg amputation in these zip codes. These mailings provided information on recognizing foot ulcers and infections, emphasized timely evaluation, and named the MEDVAMC vascular surgery team as a point-of-contact. More recently, we have seen a further decrease in the MEDVAMC incidences of leg amputation to its lowest rate in > 20 years.

Discussion

A learning organization that directs its research based on clinical observations and informs its clinical care with research findings can produce palpable improvements in outcomes. Understanding the disease process and trying to better understand management across the entire range of this disease process has allowed our team to make consistent and systematic changes in care (Table). Consolidating inpatient care in a single team model seems to have been effective in reducing amputation rates among veterans with diabetes. The role the MEDVAMC vascular surgery team served for limb salvage patients may have been particularly beneficial because of the large impact untreated or unidentified PAD can have and because of the high prevalence of PAD among the limb salvage population seen at MEDVAMC. To be sustainable, though, a single-team model needs resources. A multiteam model can also be effective if the degree of multidisciplinary involvement for any given veteran is appropriate to the individual's clinical needs, teams are engaged and willing to contribute in a defined role within their specialty, and lines of communication remain open.

The primary challenge at MEDVAMC has been, and will continue to be, the retention of physicians and surgeons. MEDVAMC has excellent leadership and a collegial working environment, but better access to operating rooms for elective and time-sensitive operations, additional clinical staff support, and higher salary at non-VA positions have been the basis for many of physicians— especially surgeons—leaving MEDVAMC. Despite high staff turnover and a constant flow of resident and fellow trainees, MEDVAMC has been able to keep the clinical approach relatively consistent due to the use of written protocols and continuity of care as ≥ 1 physician or surgeon from each of the 4 main teams remained engaged with limb salvage throughout the entire period.

Going forward, we will work to ensure that all requirements of the 2022 Prevention of Amputation in Veterans Everywhere directive are incorporated into care.⁸ We plan to standardize MEDVAMC management algorithms further, both to streamline care and reduce the opportunity for disparities in treatment. More prophylactic podiatric procedures, surgical forms of offloading, and a shared multidisciplinary clinic space may also further help patients.

Conclusions

The introduction of multidisciplinary limb salvage at MEDVAMC has led to significant and sustained reductions in leg amputation incidence. These reductions do not seem dependent upon a specific team structure for inpatient care. To improve patient outcomes, efforts should focus on making improvements across the entire disease spectrum. For limb salvage, this includes primary prevention of foot ulcers, the treatment of foot infections, identification and management of PAD, surgical reconstruction/optimal wound healing, and care for patients who undergo leg amputation.