Pharmacology

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.

Anticholinergic medications block the activity of the neurotransmitter acetylcholine by binding to either muscarinic or nicotinic receptors in both the peripheral and central nervous system. Anticholinergic medications typically refer to antimuscarinic medications and have been prescribed to treat a variety of conditions common in older adults, including overactive bladder, allergies, muscle spasms, and sleep disorders.^1,2 Since muscarinic receptors are present throughout the body, anticholinergic medications are associated with many adverse effects (AEs), including constipation, urinary retention, xerostomia, and delirium. Older adults are more sensitive to these AEs due to physiological changes associated with aging.¹

The American Geriatric Society Beers Criteria for Potentially Inappropriate Medications Use in Older Adults identifies drugs with strong anticholinergic properties. The Beers Criteria strongly recommends avoiding these medications in patients with dementia or cognitive impairment due to the risk of central nervous system AEs. In the updated 2023 Beers Criteria, the rationale was expanded to recognize the risks of the cumulative anticholinergic burden associated with concurrent anticholinergic use.^3,4

Given the prevalent use of anticholinergic medications in older adults, there has been significant research demonstrating their AEs, specifically delirium and cognitive impairment in geriatric patients. A systematic review of 14 articles conducted in 7 different countries of patients with median age of 76.4 to 86.1 years reviewed clinical outcomes of anticholinergic use in patients with dementia. Five studies found anticholinergics were associated with increased all-cause mortality in patients with dementia, and 3 studies found anticholinergics were associated with longer hospital stays. Other studies found that anticholinergics were associated with delirium and reduced health-related quality of life.⁵

About 35% of veterans with dementia have been prescribed a medication regimen with a high anticholinergic burden.⁶ In 2018, the US Department of Veterans Affairs (VA) Pharmacy Benfits Management Center for Medical Safety completed a centrally aggregated medication use evaluation (CAMUE) to assess the appropriateness of anticholinergic medication use in patients with dementia. The retrospective chart review included 1094 veterans from 19 sites. Overall, about 15% of the veterans experienced new falls, delirium, or worsening dementia within 30 days of starting an anticholinergic medication. Furthermore, < 40% had documentation of a nonanticholinergic alternative medication trial, and < 20% had documented nonpharmacologic therapy. The documentation of risk-benefit assessment acknowledging the risks of anticholinergic medication use in veterans with dementia occurred only about 13% of the time. The CAMUE concluded that the risks of initiating an anticholinergic medication in veterans with dementia are likely underdocumented and possibly under considered by prescribers.⁷

Developed within the Veterans Health Administration (VHA), VIONE (Vital, Important, Optional, Not Indicated, Every medication has an indication) is a medication management methodology that aims to reduce polypharmacy and improve patient safety consistent with high-reliability organizations. Since it launched in 2016, VIONE has gradually been implemented at many VHA facilities. The VIONE deprescribing dashboard had not been used at the VA Louisville Healthcare System prior to this quality improvement project.

This dashboard uses the Beers Criteria to identify potentially inappropriate anticholinergic medications. It uses the Anticholinergic Cognitive Burden (ACB) scale to calculate the cumulative anticholinergic risk for each patient. Medications with an ACB score of 2 or 3 have clinically relevant cognitive effects such as delirium and dementia (Table 1). For each point increase in total ACB score, a decline in mini-mental state examination score of 0.33 points over 2 years has been shown. Each point increase has also been correlated with a 26% increase in risk of death.^8-10

Methods

The purpose of this quality improvement project was to determine the impact of pharmacist-driven deprescribing on the anticholinergic burden in veterans with dementia at VA Louisville Healthcare System. Data were obtained through the Computerized Patient Record System (CPRS) and VIONE deprescribing dashboard and entered in a secure Microsoft Excel spreadsheet. Pharmacist deprescribing steps were entered as CPRS progress notes. A deprescribing note template was created, and 11 templates with indication-specific recommendations were created for each anticholinergic indication identified (contact authors for deprescribing note template examples). Usage of anticholinergic medications was reexamined 3 months after the deprescribing note was entered.

Eligible patients identified in the VIONE deprescribing dashboard had an outpatient order for a medication with strong anticholinergic properties as identified using the Beers Criteria and were aged ≥ 65 years. Patients also had to be diagnosed with dementia or cognitive impairment. Patients were excluded if they were receiving hospice care or if the anticholinergic medication was from a non-VA prescriber or filled at a non-VA pharmacy. The VIONE deprescribing dashboard also excluded skeletal muscle relaxants if the patient had a spinal cord-related visit in the previous 2 years, first-generation antihistamines if the patient had a vertigo diagnosis, hydroxyzine if the indication was for anxiety, trospium if the indication was for overactive bladder, and antipsychotics if the patient had been diagnosed with schizophrenia or bipolar disorder. The following were included in the deprescribing recommendations if the dashboard identified the patient due to receiving a second strongly anticholinergic medication: first generation antihistamines if the patient was diagnosed with vertigo and hydroxyzine if the indication is for anxiety.

Each eligible patient received a focused medication review by a pharmacist via electronic chart review and a templated CPRS progress note with patient-specific recommendations. The prescriber and the patient’s primary care practitioner were recommended to perform a patient-specific risk-benefit assessment, deprescribe potentially inappropriate anticholinergic medications, and consider nonanticholinergic alternatives (both pharmacologic and nonpharmacologic). Data collected included baseline age, sex, prespecified comorbidities (type of dementia, cognitive impairment, delirium, benign prostatic hyperplasia/lower urinary tract symptoms), duration of prescribed anticholinergic medication, indication and deprescribing rate for each anticholinergic agent, and concurrent dementia medications (acetylcholinesterase inhibitors, memantine, or both).

The primary outcome was the number of patients that had = 1 medication with strong anticholinergic properties deprescribed. Deprescribing was defined as medication discontinuation or reduction of total daily dose. Secondary outcomes were the mean change in ACB scale, the number of patients with dose tapering, documented patient-specific risk-benefit assessment, and initiated nonanticholinergic alternative per pharmacist recommendation.

Results

The VIONE deprescribing dashboard identified 121 patients; 45 were excluded for non-VA prescriber or pharmacy, and 8 patients were excluded for other reasons. Sixty-eight patients were included in the deprescribing initiative. The mean age was 73.4 years (range, 67-93), 65 (96%) were male, and 34 (50%) had unspecified dementia (Table 2). Thirty-one patients (46%) had concurrent cholinesterase inhibitor prescriptions for dementia. The median duration of use of a strong anticholinergic medication was 11 months.

Twenty-nine patients (43%) had ≥ 1 medication with strong anticholinergic properties deprescribed. Anticholinergic medication was discontinued for 26 patients, and the dose was decreased for 3 patients. ACB score fell by a mean of 1.1 per patient. There was an increase in the documented risk-benefit assessment for anticholinergic medications from a baseline of 4 (6%) to 19 (28%) 3 months after the deprescribing note. Cyclobenzaprine, paroxetine, and oxybutynin were deprescribed the most, and amitriptyline had the lowest rate of deprescribing (Table 3). Thirty patients (44%) had a pharmacologic, nonanticholinergic alternative initiated per pharmacist recommendation, and 6 patients (9%) had a nonpharmacologic alternative initiated per pharmacist recommendation.

Discussion

This quality improvement project suggests that with the use of population health management tools such as the VIONE deprescribing dashboard, pharmacists can help identify and deprescribe strong anticholinergic medications in patients with cognitive impairment or dementia. Pharmacists can also aid in deprescribing through evidence-based recommendations to guide risk-benefit discussion and consider safer, nonanticholinergic alternatives. The authors were able to help reduce anticholinergic cognitive burden in 43% of patients in this sample. The mean 1.1 ACB score reduction was considered clinically significant based on prior studies that found that each 1-point increase in ACB score correlated with declined cognition and increased mortality.^8,10 The VIONE deprescribing dashboard provided real-time patient data and helped target patients at the highest risk of anticholinergic AEs. The creation of the note templates based on the indication helped streamline recommendations. Typically, the prescriber addressed the recommendations at a routine follow-up appointment. The deprescribing method used in this project was time-efficient and could be easily replicated once the CPRS note templates were created. Future deprescribing projects could consider more direct pharmacist intervention and medication management.

Limitations

There was no direct assessment of clinical outcomes such as change in cognition using cognitive function tests. However, multiple studies have demonstrated AEs associated with strong anticholinergic medication use and additive anticholinergic burden in patients with dementia or cognitive impairment.1,5 Also, the 3-month follow-up period was relatively short. The pharmacist’s deprescribing recommendations may have been accepted after 3 months, or patients could have restarted their anticholinergic medications. Longer follow-up time could provide more robust results and conclusions. Thirdly, there was no formal definition of what constituted a risk-benefit assessment of anticholinergic medications. The risk-benefit assessment was determined at the discretion of the authors, which was subjective and allowed for bias. Finally, 6 patients died during the 3-month follow-up. The data for these patients were included in the baseline characteristics but not in the study outcomes. If these patients had been excluded from the results, a higher percentage of patients (47%) would have had ≥ 1 anticholinergic medication deprescribed.

Conclusions

In collaboration with the interdisciplinary team, pharmacist recommendations resulted in deprescribing of anticholinergic medications in veterans with dementia or cognitive impairment. The VIONE deprescribing dashboard, an easily accessible population health management tool, can identify patients prescribed potentially inappropriate medications and help target patients at the highest risk of anticholinergic AEs. To prevent worsening cognitive impairment, delirium, falls, and other AEs, this deprescribing initiative can be replicated at other VHA facilities. Future projects could have a longer follow-up period, incorporate more direct pharmacist intervention, and assess clinical outcomes of deprescribing.

Anticholinergic medications block the activity of the neurotransmitter acetylcholine by binding to either muscarinic or nicotinic receptors in both the peripheral and central nervous system. Anticholinergic medications typically refer to antimuscarinic medications and have been prescribed to treat a variety of conditions common in older adults, including overactive bladder, allergies, muscle spasms, and sleep disorders.^1,2 Since muscarinic receptors are present throughout the body, anticholinergic medications are associated with many adverse effects (AEs), including constipation, urinary retention, xerostomia, and delirium. Older adults are more sensitive to these AEs due to physiological changes associated with aging.¹

The American Geriatric Society Beers Criteria for Potentially Inappropriate Medications Use in Older Adults identifies drugs with strong anticholinergic properties. The Beers Criteria strongly recommends avoiding these medications in patients with dementia or cognitive impairment due to the risk of central nervous system AEs. In the updated 2023 Beers Criteria, the rationale was expanded to recognize the risks of the cumulative anticholinergic burden associated with concurrent anticholinergic use.^3,4

Given the prevalent use of anticholinergic medications in older adults, there has been significant research demonstrating their AEs, specifically delirium and cognitive impairment in geriatric patients. A systematic review of 14 articles conducted in 7 different countries of patients with median age of 76.4 to 86.1 years reviewed clinical outcomes of anticholinergic use in patients with dementia. Five studies found anticholinergics were associated with increased all-cause mortality in patients with dementia, and 3 studies found anticholinergics were associated with longer hospital stays. Other studies found that anticholinergics were associated with delirium and reduced health-related quality of life.⁵

About 35% of veterans with dementia have been prescribed a medication regimen with a high anticholinergic burden.⁶ In 2018, the US Department of Veterans Affairs (VA) Pharmacy Benfits Management Center for Medical Safety completed a centrally aggregated medication use evaluation (CAMUE) to assess the appropriateness of anticholinergic medication use in patients with dementia. The retrospective chart review included 1094 veterans from 19 sites. Overall, about 15% of the veterans experienced new falls, delirium, or worsening dementia within 30 days of starting an anticholinergic medication. Furthermore, < 40% had documentation of a nonanticholinergic alternative medication trial, and < 20% had documented nonpharmacologic therapy. The documentation of risk-benefit assessment acknowledging the risks of anticholinergic medication use in veterans with dementia occurred only about 13% of the time. The CAMUE concluded that the risks of initiating an anticholinergic medication in veterans with dementia are likely underdocumented and possibly under considered by prescribers.⁷

Developed within the Veterans Health Administration (VHA), VIONE (Vital, Important, Optional, Not Indicated, Every medication has an indication) is a medication management methodology that aims to reduce polypharmacy and improve patient safety consistent with high-reliability organizations. Since it launched in 2016, VIONE has gradually been implemented at many VHA facilities. The VIONE deprescribing dashboard had not been used at the VA Louisville Healthcare System prior to this quality improvement project.

This dashboard uses the Beers Criteria to identify potentially inappropriate anticholinergic medications. It uses the Anticholinergic Cognitive Burden (ACB) scale to calculate the cumulative anticholinergic risk for each patient. Medications with an ACB score of 2 or 3 have clinically relevant cognitive effects such as delirium and dementia (Table 1). For each point increase in total ACB score, a decline in mini-mental state examination score of 0.33 points over 2 years has been shown. Each point increase has also been correlated with a 26% increase in risk of death.^8-10

Methods

The purpose of this quality improvement project was to determine the impact of pharmacist-driven deprescribing on the anticholinergic burden in veterans with dementia at VA Louisville Healthcare System. Data were obtained through the Computerized Patient Record System (CPRS) and VIONE deprescribing dashboard and entered in a secure Microsoft Excel spreadsheet. Pharmacist deprescribing steps were entered as CPRS progress notes. A deprescribing note template was created, and 11 templates with indication-specific recommendations were created for each anticholinergic indication identified (contact authors for deprescribing note template examples). Usage of anticholinergic medications was reexamined 3 months after the deprescribing note was entered.

Eligible patients identified in the VIONE deprescribing dashboard had an outpatient order for a medication with strong anticholinergic properties as identified using the Beers Criteria and were aged ≥ 65 years. Patients also had to be diagnosed with dementia or cognitive impairment. Patients were excluded if they were receiving hospice care or if the anticholinergic medication was from a non-VA prescriber or filled at a non-VA pharmacy. The VIONE deprescribing dashboard also excluded skeletal muscle relaxants if the patient had a spinal cord-related visit in the previous 2 years, first-generation antihistamines if the patient had a vertigo diagnosis, hydroxyzine if the indication was for anxiety, trospium if the indication was for overactive bladder, and antipsychotics if the patient had been diagnosed with schizophrenia or bipolar disorder. The following were included in the deprescribing recommendations if the dashboard identified the patient due to receiving a second strongly anticholinergic medication: first generation antihistamines if the patient was diagnosed with vertigo and hydroxyzine if the indication is for anxiety.

Each eligible patient received a focused medication review by a pharmacist via electronic chart review and a templated CPRS progress note with patient-specific recommendations. The prescriber and the patient’s primary care practitioner were recommended to perform a patient-specific risk-benefit assessment, deprescribe potentially inappropriate anticholinergic medications, and consider nonanticholinergic alternatives (both pharmacologic and nonpharmacologic). Data collected included baseline age, sex, prespecified comorbidities (type of dementia, cognitive impairment, delirium, benign prostatic hyperplasia/lower urinary tract symptoms), duration of prescribed anticholinergic medication, indication and deprescribing rate for each anticholinergic agent, and concurrent dementia medications (acetylcholinesterase inhibitors, memantine, or both).

The primary outcome was the number of patients that had = 1 medication with strong anticholinergic properties deprescribed. Deprescribing was defined as medication discontinuation or reduction of total daily dose. Secondary outcomes were the mean change in ACB scale, the number of patients with dose tapering, documented patient-specific risk-benefit assessment, and initiated nonanticholinergic alternative per pharmacist recommendation.

Results

The VIONE deprescribing dashboard identified 121 patients; 45 were excluded for non-VA prescriber or pharmacy, and 8 patients were excluded for other reasons. Sixty-eight patients were included in the deprescribing initiative. The mean age was 73.4 years (range, 67-93), 65 (96%) were male, and 34 (50%) had unspecified dementia (Table 2). Thirty-one patients (46%) had concurrent cholinesterase inhibitor prescriptions for dementia. The median duration of use of a strong anticholinergic medication was 11 months.

Twenty-nine patients (43%) had ≥ 1 medication with strong anticholinergic properties deprescribed. Anticholinergic medication was discontinued for 26 patients, and the dose was decreased for 3 patients. ACB score fell by a mean of 1.1 per patient. There was an increase in the documented risk-benefit assessment for anticholinergic medications from a baseline of 4 (6%) to 19 (28%) 3 months after the deprescribing note. Cyclobenzaprine, paroxetine, and oxybutynin were deprescribed the most, and amitriptyline had the lowest rate of deprescribing (Table 3). Thirty patients (44%) had a pharmacologic, nonanticholinergic alternative initiated per pharmacist recommendation, and 6 patients (9%) had a nonpharmacologic alternative initiated per pharmacist recommendation.

Discussion

This quality improvement project suggests that with the use of population health management tools such as the VIONE deprescribing dashboard, pharmacists can help identify and deprescribe strong anticholinergic medications in patients with cognitive impairment or dementia. Pharmacists can also aid in deprescribing through evidence-based recommendations to guide risk-benefit discussion and consider safer, nonanticholinergic alternatives. The authors were able to help reduce anticholinergic cognitive burden in 43% of patients in this sample. The mean 1.1 ACB score reduction was considered clinically significant based on prior studies that found that each 1-point increase in ACB score correlated with declined cognition and increased mortality.^8,10 The VIONE deprescribing dashboard provided real-time patient data and helped target patients at the highest risk of anticholinergic AEs. The creation of the note templates based on the indication helped streamline recommendations. Typically, the prescriber addressed the recommendations at a routine follow-up appointment. The deprescribing method used in this project was time-efficient and could be easily replicated once the CPRS note templates were created. Future deprescribing projects could consider more direct pharmacist intervention and medication management.

Limitations

There was no direct assessment of clinical outcomes such as change in cognition using cognitive function tests. However, multiple studies have demonstrated AEs associated with strong anticholinergic medication use and additive anticholinergic burden in patients with dementia or cognitive impairment.1,5 Also, the 3-month follow-up period was relatively short. The pharmacist’s deprescribing recommendations may have been accepted after 3 months, or patients could have restarted their anticholinergic medications. Longer follow-up time could provide more robust results and conclusions. Thirdly, there was no formal definition of what constituted a risk-benefit assessment of anticholinergic medications. The risk-benefit assessment was determined at the discretion of the authors, which was subjective and allowed for bias. Finally, 6 patients died during the 3-month follow-up. The data for these patients were included in the baseline characteristics but not in the study outcomes. If these patients had been excluded from the results, a higher percentage of patients (47%) would have had ≥ 1 anticholinergic medication deprescribed.

Conclusions

In collaboration with the interdisciplinary team, pharmacist recommendations resulted in deprescribing of anticholinergic medications in veterans with dementia or cognitive impairment. The VIONE deprescribing dashboard, an easily accessible population health management tool, can identify patients prescribed potentially inappropriate medications and help target patients at the highest risk of anticholinergic AEs. To prevent worsening cognitive impairment, delirium, falls, and other AEs, this deprescribing initiative can be replicated at other VHA facilities. Future projects could have a longer follow-up period, incorporate more direct pharmacist intervention, and assess clinical outcomes of deprescribing.

Anticholinergic medications block the activity of the neurotransmitter acetylcholine by binding to either muscarinic or nicotinic receptors in both the peripheral and central nervous system. Anticholinergic medications typically refer to antimuscarinic medications and have been prescribed to treat a variety of conditions common in older adults, including overactive bladder, allergies, muscle spasms, and sleep disorders.^1,2 Since muscarinic receptors are present throughout the body, anticholinergic medications are associated with many adverse effects (AEs), including constipation, urinary retention, xerostomia, and delirium. Older adults are more sensitive to these AEs due to physiological changes associated with aging.¹

The American Geriatric Society Beers Criteria for Potentially Inappropriate Medications Use in Older Adults identifies drugs with strong anticholinergic properties. The Beers Criteria strongly recommends avoiding these medications in patients with dementia or cognitive impairment due to the risk of central nervous system AEs. In the updated 2023 Beers Criteria, the rationale was expanded to recognize the risks of the cumulative anticholinergic burden associated with concurrent anticholinergic use.^3,4

Given the prevalent use of anticholinergic medications in older adults, there has been significant research demonstrating their AEs, specifically delirium and cognitive impairment in geriatric patients. A systematic review of 14 articles conducted in 7 different countries of patients with median age of 76.4 to 86.1 years reviewed clinical outcomes of anticholinergic use in patients with dementia. Five studies found anticholinergics were associated with increased all-cause mortality in patients with dementia, and 3 studies found anticholinergics were associated with longer hospital stays. Other studies found that anticholinergics were associated with delirium and reduced health-related quality of life.⁵

About 35% of veterans with dementia have been prescribed a medication regimen with a high anticholinergic burden.⁶ In 2018, the US Department of Veterans Affairs (VA) Pharmacy Benfits Management Center for Medical Safety completed a centrally aggregated medication use evaluation (CAMUE) to assess the appropriateness of anticholinergic medication use in patients with dementia. The retrospective chart review included 1094 veterans from 19 sites. Overall, about 15% of the veterans experienced new falls, delirium, or worsening dementia within 30 days of starting an anticholinergic medication. Furthermore, < 40% had documentation of a nonanticholinergic alternative medication trial, and < 20% had documented nonpharmacologic therapy. The documentation of risk-benefit assessment acknowledging the risks of anticholinergic medication use in veterans with dementia occurred only about 13% of the time. The CAMUE concluded that the risks of initiating an anticholinergic medication in veterans with dementia are likely underdocumented and possibly under considered by prescribers.⁷

Developed within the Veterans Health Administration (VHA), VIONE (Vital, Important, Optional, Not Indicated, Every medication has an indication) is a medication management methodology that aims to reduce polypharmacy and improve patient safety consistent with high-reliability organizations. Since it launched in 2016, VIONE has gradually been implemented at many VHA facilities. The VIONE deprescribing dashboard had not been used at the VA Louisville Healthcare System prior to this quality improvement project.

This dashboard uses the Beers Criteria to identify potentially inappropriate anticholinergic medications. It uses the Anticholinergic Cognitive Burden (ACB) scale to calculate the cumulative anticholinergic risk for each patient. Medications with an ACB score of 2 or 3 have clinically relevant cognitive effects such as delirium and dementia (Table 1). For each point increase in total ACB score, a decline in mini-mental state examination score of 0.33 points over 2 years has been shown. Each point increase has also been correlated with a 26% increase in risk of death.^8-10

Methods

The purpose of this quality improvement project was to determine the impact of pharmacist-driven deprescribing on the anticholinergic burden in veterans with dementia at VA Louisville Healthcare System. Data were obtained through the Computerized Patient Record System (CPRS) and VIONE deprescribing dashboard and entered in a secure Microsoft Excel spreadsheet. Pharmacist deprescribing steps were entered as CPRS progress notes. A deprescribing note template was created, and 11 templates with indication-specific recommendations were created for each anticholinergic indication identified (contact authors for deprescribing note template examples). Usage of anticholinergic medications was reexamined 3 months after the deprescribing note was entered.

Eligible patients identified in the VIONE deprescribing dashboard had an outpatient order for a medication with strong anticholinergic properties as identified using the Beers Criteria and were aged ≥ 65 years. Patients also had to be diagnosed with dementia or cognitive impairment. Patients were excluded if they were receiving hospice care or if the anticholinergic medication was from a non-VA prescriber or filled at a non-VA pharmacy. The VIONE deprescribing dashboard also excluded skeletal muscle relaxants if the patient had a spinal cord-related visit in the previous 2 years, first-generation antihistamines if the patient had a vertigo diagnosis, hydroxyzine if the indication was for anxiety, trospium if the indication was for overactive bladder, and antipsychotics if the patient had been diagnosed with schizophrenia or bipolar disorder. The following were included in the deprescribing recommendations if the dashboard identified the patient due to receiving a second strongly anticholinergic medication: first generation antihistamines if the patient was diagnosed with vertigo and hydroxyzine if the indication is for anxiety.

Each eligible patient received a focused medication review by a pharmacist via electronic chart review and a templated CPRS progress note with patient-specific recommendations. The prescriber and the patient’s primary care practitioner were recommended to perform a patient-specific risk-benefit assessment, deprescribe potentially inappropriate anticholinergic medications, and consider nonanticholinergic alternatives (both pharmacologic and nonpharmacologic). Data collected included baseline age, sex, prespecified comorbidities (type of dementia, cognitive impairment, delirium, benign prostatic hyperplasia/lower urinary tract symptoms), duration of prescribed anticholinergic medication, indication and deprescribing rate for each anticholinergic agent, and concurrent dementia medications (acetylcholinesterase inhibitors, memantine, or both).

The primary outcome was the number of patients that had = 1 medication with strong anticholinergic properties deprescribed. Deprescribing was defined as medication discontinuation or reduction of total daily dose. Secondary outcomes were the mean change in ACB scale, the number of patients with dose tapering, documented patient-specific risk-benefit assessment, and initiated nonanticholinergic alternative per pharmacist recommendation.

Results

The VIONE deprescribing dashboard identified 121 patients; 45 were excluded for non-VA prescriber or pharmacy, and 8 patients were excluded for other reasons. Sixty-eight patients were included in the deprescribing initiative. The mean age was 73.4 years (range, 67-93), 65 (96%) were male, and 34 (50%) had unspecified dementia (Table 2). Thirty-one patients (46%) had concurrent cholinesterase inhibitor prescriptions for dementia. The median duration of use of a strong anticholinergic medication was 11 months.

Twenty-nine patients (43%) had ≥ 1 medication with strong anticholinergic properties deprescribed. Anticholinergic medication was discontinued for 26 patients, and the dose was decreased for 3 patients. ACB score fell by a mean of 1.1 per patient. There was an increase in the documented risk-benefit assessment for anticholinergic medications from a baseline of 4 (6%) to 19 (28%) 3 months after the deprescribing note. Cyclobenzaprine, paroxetine, and oxybutynin were deprescribed the most, and amitriptyline had the lowest rate of deprescribing (Table 3). Thirty patients (44%) had a pharmacologic, nonanticholinergic alternative initiated per pharmacist recommendation, and 6 patients (9%) had a nonpharmacologic alternative initiated per pharmacist recommendation.

Discussion

This quality improvement project suggests that with the use of population health management tools such as the VIONE deprescribing dashboard, pharmacists can help identify and deprescribe strong anticholinergic medications in patients with cognitive impairment or dementia. Pharmacists can also aid in deprescribing through evidence-based recommendations to guide risk-benefit discussion and consider safer, nonanticholinergic alternatives. The authors were able to help reduce anticholinergic cognitive burden in 43% of patients in this sample. The mean 1.1 ACB score reduction was considered clinically significant based on prior studies that found that each 1-point increase in ACB score correlated with declined cognition and increased mortality.^8,10 The VIONE deprescribing dashboard provided real-time patient data and helped target patients at the highest risk of anticholinergic AEs. The creation of the note templates based on the indication helped streamline recommendations. Typically, the prescriber addressed the recommendations at a routine follow-up appointment. The deprescribing method used in this project was time-efficient and could be easily replicated once the CPRS note templates were created. Future deprescribing projects could consider more direct pharmacist intervention and medication management.

Limitations

There was no direct assessment of clinical outcomes such as change in cognition using cognitive function tests. However, multiple studies have demonstrated AEs associated with strong anticholinergic medication use and additive anticholinergic burden in patients with dementia or cognitive impairment.1,5 Also, the 3-month follow-up period was relatively short. The pharmacist’s deprescribing recommendations may have been accepted after 3 months, or patients could have restarted their anticholinergic medications. Longer follow-up time could provide more robust results and conclusions. Thirdly, there was no formal definition of what constituted a risk-benefit assessment of anticholinergic medications. The risk-benefit assessment was determined at the discretion of the authors, which was subjective and allowed for bias. Finally, 6 patients died during the 3-month follow-up. The data for these patients were included in the baseline characteristics but not in the study outcomes. If these patients had been excluded from the results, a higher percentage of patients (47%) would have had ≥ 1 anticholinergic medication deprescribed.

Conclusions

In collaboration with the interdisciplinary team, pharmacist recommendations resulted in deprescribing of anticholinergic medications in veterans with dementia or cognitive impairment. The VIONE deprescribing dashboard, an easily accessible population health management tool, can identify patients prescribed potentially inappropriate medications and help target patients at the highest risk of anticholinergic AEs. To prevent worsening cognitive impairment, delirium, falls, and other AEs, this deprescribing initiative can be replicated at other VHA facilities. Future projects could have a longer follow-up period, incorporate more direct pharmacist intervention, and assess clinical outcomes of deprescribing.

More than 37 million Americans have diabetes mellitus (DM), and approximately 90% have type 2 DM (T2DM), including about 25% of veterans.^1,2 The current guidelines suggest that therapy depends on a patient's comorbidities, management needs, and patient-centered treatment factors.³ About 1 in 3 adults with DM have chronic kidney disease (CKD), defined as the presence of kidney damage or an estimated glomerular filtration rate (eGFR) < 60 mL/min per 1.73 m², persisting for ≥ 3 months.⁴

Sodium-glucose cotransporter-2 (SGLT-2) inhibitors are a class of antihyperglycemic agents acting on the SGLT-2 proteins expressed in the renal proximal convoluted tubules. They exert their effects by preventing the reabsorption of filtered glucose from the tubular lumen. There are 4 SGLT-2 inhibitors approved by the US Food and Drug Administration: canagliflozin, dapagliflozin, empagliflozin, and ertugliflozin. Empagliflozin is currently the preferred SGLT-2 inhibitor on the US Department of Veterans Affairs (VA) formulary.

According to the American Diabetes Association guidelines, empagliflozin is considered when an individual has or is at risk for atherosclerotic cardiovascular disease, heart failure, and CKD.³ SGLT-2 inhibitors are a favorable option due to their low risk for hypoglycemia while also promoting weight loss. The EMPEROR-Reduced trial demonstrated that, in addition to benefits for patients with heart failure, empagliflozin also slowed the progressive decline in kidney function in those with and without DM.⁵ The purpose of this study was to evaluate the effectiveness of empagliflozin on hemoglobin A_1c (HbA_1c) levels in patients with CKD at the Hershel “Woody” Williams VA Medical Center (HWWVAMC) in Huntington, West Virginia, along with other laboratory test markers.

Methods

The Marshall University Institutional Review Board #1 (Medical) and the HWWVAMC institutional review board and research and development committee each reviewed and approved this study. A retrospective chart review was conducted on patients diagnosed with T2DM and stage 3 CKD who were prescribed empagliflozin for DM management between January 1, 2015, and October 1, 2022, yielding 1771 patients. Data were obtained through the VHA Corporate Data Warehouse (CDW) and stored on the VA Informatics and Computing Infrastructure (VINCI) research server.

Patients were included if they were aged 18 to 89 years, prescribed empagliflozin by a VA clinician for the treatment of T2DM, had an eGFR between 30 and 59 mL/min/1.73 m², and had an initial HbA_1c between 7% and 10%. Using further random sampling, patients were either excluded or divided into, those with stage 3a CKD and those with stage 3b CKD. The primary endpoint of this study was the change in HbA_1c levels in patients with stage 3b CKD (eGFR 30-44 mL/min/1.73 m²) compared with stage 3a (eGFR 45-59 mL/min/1.73 m²) after 12 months. The secondary endpoints included effects on renal function, weight, blood pressure, incidence of adverse drug events, and cardiovascular events. Of the excluded, 38 had HbA_1c < 7%, 30 had HbA_1c ≥ 10%, 21 did not have data at 1-year mark, 15 had the medication discontinued due to decline in renal function, 14 discontinued their medication without documented reason, 10 discontinued their medication due to adverse drug reactions (ADRs), 12 had eGFR > 60 mL/ min/1.73 m², 9 died within 1 year of initiation, 4 had eGFR < 30 mL/min/1.73 m², 1 had no baseline eGFR, and 1 was the spouse of a veteran.

Statistical Analysis

All statistical analyses were performed using STATA v.15. We used t tests to examine changes within each group, along with paired t tests to compare the 2 groups. Two-sample t tests were used to analyze the continuous data at both the primary and secondary endpoints.

Results

Of the 1771 patients included in the initial data set, a randomized sample of 255 charts were reviewed, 155 were excluded, and 100 were included. Fifty patients, had stage 3a CKD and 50 had stage 3b CKD. Baseline demographics were similar between the stage 3a and 3b groups (Table 1). Both groups were predominantly White and male, with mean age > 70 years.

The primary endpoint was the differences in HbA_1c levels over time and between groups for patients with stage 3a and stage 3b CKD 1 year after initiation of empagliflozin. The starting doses of empagliflozin were either 12.5 mg or 25.0 mg. For both groups, the changes in HbA_1c levels were statistically significant (Table 2). HbA_1c levels dropped 0.65% for the stage 3a group and 0.48% for the 3b group. When compared to one another, the results were not statistically significant (P = .51).

Secondary Endpoint

There was no statistically significant difference in serum creatinine levels within each group between baselines and 1 year later for the stage 3a (P = .21) and stage 3b (P = .22) groups, or when compared to each other (P = .67). There were statistically significant changes in weight for patients in the stage 3a group (P < .05), but not for stage 3b group (P = .06) or when compared to each other (P = .41). A statistically significant change in systolic blood pressure was observed for the stage 3a group (P = .003), but not the stage 3b group (P = .16) or when compared to each other (P = .27). There were statistically significant changes in diastolic blood pressure within the stage 3a group (P = .04), but not within the stage 3b group (P = .61) or when compared to each other (P = .31).

Ten patients discontinued empagliflozin before the 1-year mark due to ADRs, including dizziness, increased incidence of urinary tract infections, rash, and tachycardia (Table 3). Additionally, 3 ADRs resulted in the empagliflozin discontinuation after 1 year (Table 3).

Discussion

This study showed a statistically significant change in HbA_1c levels for patients with stage 3a and stage 3b CKD. With eGFR levels in these 2 groups > 30 mL/min/1.73 m², patients were able to achieve glycemic benefits. There were no significant changes to the serum creatinine levels. Both groups saw statistically significant changes in weight loss within their own group; however, there were no statistically significant changes when compared to each other. With both systolic and diastolic blood pressure, the stage 3a group had statistically significant changes.

The EMPA-REG BP study demonstrated that empagliflozin was associated with significant and clinically meaningful reductions in blood pressure and HbA_1c levels compared with placebo and was well tolerated in patients with T2DM and hypertension.^6,7,8

Limitations

This study had a retrospective study design, which resulted in missing information for many patients and higher rates of exclusion. The population was predominantly older, White, and male and may not reflect other populations. The starting doses of empagliflozin varied between the groups. The VA employs tablet splitting for some patients, and the available doses were either 10.0 mg, 12.5 mg, or 25.0 mg. Some prescribers start veterans at lower doses and gradually increase to the higher dose of 25.0 mg, adding to the variability in starting doses.

Patients with eGFR < 30 mL/min/1.73 m² make it difficult to determine any potential benefit in this population. The EMPA-KIDNEY trial demonstrated that the benefits of empagliflozin treatment were consistent among patients with or without DM and regardless of eGFR at randomization.⁹ Furthermore, many veterans had an initial HbA_1c levels outside the inclusion criteria range, which was a factor in the smaller sample size.

Conclusions

While the reduction in HbA_1c levels was less in patients with stage 3b CKD compared to patients stage 3a CKD, all patients experienced a benefit. The overall incidence of ADRs was low in the study population, showing empagliflozin as a favorable choice for those with T2DM and CKD. Based on the findings of this study, empagliflozin is a potentially beneficial option for reducing HbA_1c levels in patients with CKD.

More than 37 million Americans have diabetes mellitus (DM), and approximately 90% have type 2 DM (T2DM), including about 25% of veterans.^1,2 The current guidelines suggest that therapy depends on a patient's comorbidities, management needs, and patient-centered treatment factors.³ About 1 in 3 adults with DM have chronic kidney disease (CKD), defined as the presence of kidney damage or an estimated glomerular filtration rate (eGFR) < 60 mL/min per 1.73 m², persisting for ≥ 3 months.⁴

Sodium-glucose cotransporter-2 (SGLT-2) inhibitors are a class of antihyperglycemic agents acting on the SGLT-2 proteins expressed in the renal proximal convoluted tubules. They exert their effects by preventing the reabsorption of filtered glucose from the tubular lumen. There are 4 SGLT-2 inhibitors approved by the US Food and Drug Administration: canagliflozin, dapagliflozin, empagliflozin, and ertugliflozin. Empagliflozin is currently the preferred SGLT-2 inhibitor on the US Department of Veterans Affairs (VA) formulary.

According to the American Diabetes Association guidelines, empagliflozin is considered when an individual has or is at risk for atherosclerotic cardiovascular disease, heart failure, and CKD.³ SGLT-2 inhibitors are a favorable option due to their low risk for hypoglycemia while also promoting weight loss. The EMPEROR-Reduced trial demonstrated that, in addition to benefits for patients with heart failure, empagliflozin also slowed the progressive decline in kidney function in those with and without DM.⁵ The purpose of this study was to evaluate the effectiveness of empagliflozin on hemoglobin A_1c (HbA_1c) levels in patients with CKD at the Hershel “Woody” Williams VA Medical Center (HWWVAMC) in Huntington, West Virginia, along with other laboratory test markers.

Methods

The Marshall University Institutional Review Board #1 (Medical) and the HWWVAMC institutional review board and research and development committee each reviewed and approved this study. A retrospective chart review was conducted on patients diagnosed with T2DM and stage 3 CKD who were prescribed empagliflozin for DM management between January 1, 2015, and October 1, 2022, yielding 1771 patients. Data were obtained through the VHA Corporate Data Warehouse (CDW) and stored on the VA Informatics and Computing Infrastructure (VINCI) research server.

Patients were included if they were aged 18 to 89 years, prescribed empagliflozin by a VA clinician for the treatment of T2DM, had an eGFR between 30 and 59 mL/min/1.73 m², and had an initial HbA_1c between 7% and 10%. Using further random sampling, patients were either excluded or divided into, those with stage 3a CKD and those with stage 3b CKD. The primary endpoint of this study was the change in HbA_1c levels in patients with stage 3b CKD (eGFR 30-44 mL/min/1.73 m²) compared with stage 3a (eGFR 45-59 mL/min/1.73 m²) after 12 months. The secondary endpoints included effects on renal function, weight, blood pressure, incidence of adverse drug events, and cardiovascular events. Of the excluded, 38 had HbA_1c < 7%, 30 had HbA_1c ≥ 10%, 21 did not have data at 1-year mark, 15 had the medication discontinued due to decline in renal function, 14 discontinued their medication without documented reason, 10 discontinued their medication due to adverse drug reactions (ADRs), 12 had eGFR > 60 mL/ min/1.73 m², 9 died within 1 year of initiation, 4 had eGFR < 30 mL/min/1.73 m², 1 had no baseline eGFR, and 1 was the spouse of a veteran.

Statistical Analysis

All statistical analyses were performed using STATA v.15. We used t tests to examine changes within each group, along with paired t tests to compare the 2 groups. Two-sample t tests were used to analyze the continuous data at both the primary and secondary endpoints.

Results

Of the 1771 patients included in the initial data set, a randomized sample of 255 charts were reviewed, 155 were excluded, and 100 were included. Fifty patients, had stage 3a CKD and 50 had stage 3b CKD. Baseline demographics were similar between the stage 3a and 3b groups (Table 1). Both groups were predominantly White and male, with mean age > 70 years.

The primary endpoint was the differences in HbA_1c levels over time and between groups for patients with stage 3a and stage 3b CKD 1 year after initiation of empagliflozin. The starting doses of empagliflozin were either 12.5 mg or 25.0 mg. For both groups, the changes in HbA_1c levels were statistically significant (Table 2). HbA_1c levels dropped 0.65% for the stage 3a group and 0.48% for the 3b group. When compared to one another, the results were not statistically significant (P = .51).

Secondary Endpoint

There was no statistically significant difference in serum creatinine levels within each group between baselines and 1 year later for the stage 3a (P = .21) and stage 3b (P = .22) groups, or when compared to each other (P = .67). There were statistically significant changes in weight for patients in the stage 3a group (P < .05), but not for stage 3b group (P = .06) or when compared to each other (P = .41). A statistically significant change in systolic blood pressure was observed for the stage 3a group (P = .003), but not the stage 3b group (P = .16) or when compared to each other (P = .27). There were statistically significant changes in diastolic blood pressure within the stage 3a group (P = .04), but not within the stage 3b group (P = .61) or when compared to each other (P = .31).

Ten patients discontinued empagliflozin before the 1-year mark due to ADRs, including dizziness, increased incidence of urinary tract infections, rash, and tachycardia (Table 3). Additionally, 3 ADRs resulted in the empagliflozin discontinuation after 1 year (Table 3).

Discussion

This study showed a statistically significant change in HbA_1c levels for patients with stage 3a and stage 3b CKD. With eGFR levels in these 2 groups > 30 mL/min/1.73 m², patients were able to achieve glycemic benefits. There were no significant changes to the serum creatinine levels. Both groups saw statistically significant changes in weight loss within their own group; however, there were no statistically significant changes when compared to each other. With both systolic and diastolic blood pressure, the stage 3a group had statistically significant changes.

The EMPA-REG BP study demonstrated that empagliflozin was associated with significant and clinically meaningful reductions in blood pressure and HbA_1c levels compared with placebo and was well tolerated in patients with T2DM and hypertension.^6,7,8

Limitations

This study had a retrospective study design, which resulted in missing information for many patients and higher rates of exclusion. The population was predominantly older, White, and male and may not reflect other populations. The starting doses of empagliflozin varied between the groups. The VA employs tablet splitting for some patients, and the available doses were either 10.0 mg, 12.5 mg, or 25.0 mg. Some prescribers start veterans at lower doses and gradually increase to the higher dose of 25.0 mg, adding to the variability in starting doses.

Patients with eGFR < 30 mL/min/1.73 m² make it difficult to determine any potential benefit in this population. The EMPA-KIDNEY trial demonstrated that the benefits of empagliflozin treatment were consistent among patients with or without DM and regardless of eGFR at randomization.⁹ Furthermore, many veterans had an initial HbA_1c levels outside the inclusion criteria range, which was a factor in the smaller sample size.

Conclusions

While the reduction in HbA_1c levels was less in patients with stage 3b CKD compared to patients stage 3a CKD, all patients experienced a benefit. The overall incidence of ADRs was low in the study population, showing empagliflozin as a favorable choice for those with T2DM and CKD. Based on the findings of this study, empagliflozin is a potentially beneficial option for reducing HbA_1c levels in patients with CKD.

More than 37 million Americans have diabetes mellitus (DM), and approximately 90% have type 2 DM (T2DM), including about 25% of veterans.^1,2 The current guidelines suggest that therapy depends on a patient's comorbidities, management needs, and patient-centered treatment factors.³ About 1 in 3 adults with DM have chronic kidney disease (CKD), defined as the presence of kidney damage or an estimated glomerular filtration rate (eGFR) < 60 mL/min per 1.73 m², persisting for ≥ 3 months.⁴

Sodium-glucose cotransporter-2 (SGLT-2) inhibitors are a class of antihyperglycemic agents acting on the SGLT-2 proteins expressed in the renal proximal convoluted tubules. They exert their effects by preventing the reabsorption of filtered glucose from the tubular lumen. There are 4 SGLT-2 inhibitors approved by the US Food and Drug Administration: canagliflozin, dapagliflozin, empagliflozin, and ertugliflozin. Empagliflozin is currently the preferred SGLT-2 inhibitor on the US Department of Veterans Affairs (VA) formulary.

According to the American Diabetes Association guidelines, empagliflozin is considered when an individual has or is at risk for atherosclerotic cardiovascular disease, heart failure, and CKD.³ SGLT-2 inhibitors are a favorable option due to their low risk for hypoglycemia while also promoting weight loss. The EMPEROR-Reduced trial demonstrated that, in addition to benefits for patients with heart failure, empagliflozin also slowed the progressive decline in kidney function in those with and without DM.⁵ The purpose of this study was to evaluate the effectiveness of empagliflozin on hemoglobin A_1c (HbA_1c) levels in patients with CKD at the Hershel “Woody” Williams VA Medical Center (HWWVAMC) in Huntington, West Virginia, along with other laboratory test markers.

Methods

The Marshall University Institutional Review Board #1 (Medical) and the HWWVAMC institutional review board and research and development committee each reviewed and approved this study. A retrospective chart review was conducted on patients diagnosed with T2DM and stage 3 CKD who were prescribed empagliflozin for DM management between January 1, 2015, and October 1, 2022, yielding 1771 patients. Data were obtained through the VHA Corporate Data Warehouse (CDW) and stored on the VA Informatics and Computing Infrastructure (VINCI) research server.

Patients were included if they were aged 18 to 89 years, prescribed empagliflozin by a VA clinician for the treatment of T2DM, had an eGFR between 30 and 59 mL/min/1.73 m², and had an initial HbA_1c between 7% and 10%. Using further random sampling, patients were either excluded or divided into, those with stage 3a CKD and those with stage 3b CKD. The primary endpoint of this study was the change in HbA_1c levels in patients with stage 3b CKD (eGFR 30-44 mL/min/1.73 m²) compared with stage 3a (eGFR 45-59 mL/min/1.73 m²) after 12 months. The secondary endpoints included effects on renal function, weight, blood pressure, incidence of adverse drug events, and cardiovascular events. Of the excluded, 38 had HbA_1c < 7%, 30 had HbA_1c ≥ 10%, 21 did not have data at 1-year mark, 15 had the medication discontinued due to decline in renal function, 14 discontinued their medication without documented reason, 10 discontinued their medication due to adverse drug reactions (ADRs), 12 had eGFR > 60 mL/ min/1.73 m², 9 died within 1 year of initiation, 4 had eGFR < 30 mL/min/1.73 m², 1 had no baseline eGFR, and 1 was the spouse of a veteran.

Statistical Analysis

All statistical analyses were performed using STATA v.15. We used t tests to examine changes within each group, along with paired t tests to compare the 2 groups. Two-sample t tests were used to analyze the continuous data at both the primary and secondary endpoints.

Results

Of the 1771 patients included in the initial data set, a randomized sample of 255 charts were reviewed, 155 were excluded, and 100 were included. Fifty patients, had stage 3a CKD and 50 had stage 3b CKD. Baseline demographics were similar between the stage 3a and 3b groups (Table 1). Both groups were predominantly White and male, with mean age > 70 years.

The primary endpoint was the differences in HbA_1c levels over time and between groups for patients with stage 3a and stage 3b CKD 1 year after initiation of empagliflozin. The starting doses of empagliflozin were either 12.5 mg or 25.0 mg. For both groups, the changes in HbA_1c levels were statistically significant (Table 2). HbA_1c levels dropped 0.65% for the stage 3a group and 0.48% for the 3b group. When compared to one another, the results were not statistically significant (P = .51).

Secondary Endpoint

There was no statistically significant difference in serum creatinine levels within each group between baselines and 1 year later for the stage 3a (P = .21) and stage 3b (P = .22) groups, or when compared to each other (P = .67). There were statistically significant changes in weight for patients in the stage 3a group (P < .05), but not for stage 3b group (P = .06) or when compared to each other (P = .41). A statistically significant change in systolic blood pressure was observed for the stage 3a group (P = .003), but not the stage 3b group (P = .16) or when compared to each other (P = .27). There were statistically significant changes in diastolic blood pressure within the stage 3a group (P = .04), but not within the stage 3b group (P = .61) or when compared to each other (P = .31).

Ten patients discontinued empagliflozin before the 1-year mark due to ADRs, including dizziness, increased incidence of urinary tract infections, rash, and tachycardia (Table 3). Additionally, 3 ADRs resulted in the empagliflozin discontinuation after 1 year (Table 3).

Discussion

This study showed a statistically significant change in HbA_1c levels for patients with stage 3a and stage 3b CKD. With eGFR levels in these 2 groups > 30 mL/min/1.73 m², patients were able to achieve glycemic benefits. There were no significant changes to the serum creatinine levels. Both groups saw statistically significant changes in weight loss within their own group; however, there were no statistically significant changes when compared to each other. With both systolic and diastolic blood pressure, the stage 3a group had statistically significant changes.

The EMPA-REG BP study demonstrated that empagliflozin was associated with significant and clinically meaningful reductions in blood pressure and HbA_1c levels compared with placebo and was well tolerated in patients with T2DM and hypertension.^6,7,8

Limitations

This study had a retrospective study design, which resulted in missing information for many patients and higher rates of exclusion. The population was predominantly older, White, and male and may not reflect other populations. The starting doses of empagliflozin varied between the groups. The VA employs tablet splitting for some patients, and the available doses were either 10.0 mg, 12.5 mg, or 25.0 mg. Some prescribers start veterans at lower doses and gradually increase to the higher dose of 25.0 mg, adding to the variability in starting doses.

Patients with eGFR < 30 mL/min/1.73 m² make it difficult to determine any potential benefit in this population. The EMPA-KIDNEY trial demonstrated that the benefits of empagliflozin treatment were consistent among patients with or without DM and regardless of eGFR at randomization.⁹ Furthermore, many veterans had an initial HbA_1c levels outside the inclusion criteria range, which was a factor in the smaller sample size.

Conclusions

While the reduction in HbA_1c levels was less in patients with stage 3b CKD compared to patients stage 3a CKD, all patients experienced a benefit. The overall incidence of ADRs was low in the study population, showing empagliflozin as a favorable choice for those with T2DM and CKD. Based on the findings of this study, empagliflozin is a potentially beneficial option for reducing HbA_1c levels in patients with CKD.

Antimicrobial resistance is a global threat and burden to health care, with > 2.8 million antibiotic-resistant infections occurring annually in the United States.¹ To combat this issue and improve patient care, the US Department of Veterans Affairs (VA) has implemented antimicrobial stewardship programs (ASPs) across its health care systems. ASPs are multidisciplinary teams that promote evidence-based use of antimicrobials through activities supporting appropriate selection, dosing, route, and duration of antimicrobial therapy. ASP best practices are also included in the Joint Commission and Centers for Medicare and Medicaid Services accreditation standards.²

The foundational charge for VA facilities to develop and maintain ASPs was outlined in 2014 and updated in 2023 in the Veterans Health Administration (VHA) Directive 1031 on antimicrobial stewardship programs.2 This directive outlines specific requirements for all VA ASPs, including personnel, staffing levels, and the roles and responsibilities of all team members. VHA now requires that Veterans Integrated Services Networks (VISNs) establish robust ASP collaboratives. A VISN ASP collaborative consists of stewardship champions from each VA medical center in the VISN and is designed to support, develop, and enhance ASP programs across all facilities within that VISN.² Some VISNs may lack an ASP collaborative altogether, and others with existing groups may seek ways to expand their collaboratives in line with the updated directive. Prior to VHA Directive 1031, the VA Sunshine Healthcare Network (VISN 8) established an ASP collaborative. This article describes the structure and activities of the VISN 8 ASP collaborative and highlights a recent VISN 8 quality assurance initiative related to vancomycin area under the curve (AUC) dosing that illustrates how ASP collaboratives can enhance stewardship and clinical care across broad geographic areas.

VISN 8 ASP

The VHA, the largest integrated US health care system, is divided into 18 VISNs that provide regional systems of care to enhance access and meet the local health care needs of veterans.³ VISN 8 serves > 1.5 million veterans across 165,759 km² in Florida, South Georgia, Puerto Rico, and the US Virgin Islands.⁴ The network is composed of 7 health systems with 8 medical centers and > 60 outpatient clinics. These facilities provide comprehensive acute, primary, and specialty care, as well as mental health and extended care services in inpatient, outpatient, nursing home, and home care settings.⁴

The 2023 VHA Directive 1031 update recognizes the importance of VISN-level coordination of ASP activities to enhance the standardization of care and build partnerships in stewardship across all levels of care. The VISN 8 ASP collaborative workgroup (ASPWG) was established in 2015. Consistent with Directive 1031, the ASPWG is guided by clinician and pharmacist VISN leads. These leads serve as subject matter experts, facilitate access to resources, establish VISN-level consensus, and enhance communication among local ASP champions at medical centers within the VISN. All 7 health systems include = 1 ASP champion (clinician or pharmacist) in the ASPWG. Ad hoc members, whose routine duties are not solely focused on antimicrobial stewardship, contribute to specific stewardship projects as needed. For example, the ASPWG has included internal medicine, emergency department, community living center pharmacists, representatives from pharmacy administration, and trainees (pharmacy students and residents, and infectious diseases fellows) in antimicrobial stewardship initiatives. The inclusion of non-ASP champions is not discussed in VHA Directive 1031. However, these members have made valuable contributions to the ASPWG.

The ASPWG meets monthly. Agendas and priorities are developed by the VISN pharmacist and health care practitioner (HCP) leads. Monthly discussions may include but are not limited to a review of national formulary decisions, VISN goals and metrics, infectious diseases hot topics, pharmacoeconomic initiatives, strong practice presentations, regulatory and accreditation preparation, preparation of tracking reports, as well as the development of both patient-level and HCPlevel tools, resources, and education materials. This forum facilitates collaborative learning: members process and synthesize information, share and reframe ideas, and listen to other viewpoints to gain a complete understanding as a group.⁵ For example, ASPWG members have leaned on each other to prepare for Joint Commission accreditation surveys and strengthen the VISN 8 COVID-19 program through the rollout of vaccines and treatments. Other collaborative projects completed over the past few years included a penicillin allergy testing initiative and anti-methicillin-resistant Staphylococcus aureus (MRSA) and pseudomonal medication use evaluations. This team-centric problem-solving approach is highly effective while also fostering professional and social relationships. However, collaboratives could be perceived to have drawbacks. There may be opportunity costs if ASP time is allocated for issues that have already been addressed locally or concerns that standardization might hinder rapid adoption of practices at individual sites. Therefore, participation in each distinct group initiative is optional. This allows sites to choose projects related to their high priority areas and maintain bandwidth to implement practices not yet adopted by the larger group.

The ASPWG tracks metrics related to antimicrobial use with quarterly data presented by the VISN pharmacist lead. Both inpatient and outpatient metrics are evaluated, such as days of therapy per 1000 days and outpatient antibiotic prescriptions per 1000 unique patients. Facilities are benchmarked against their own historical data and other VISN sites, as well as other VISNs across the country. When outliers are identified, facilities are encouraged to conduct local projects to identify reasons for different antimicrobial use patterns and subsequent initiatives to optimize antimicrobial use. Benchmarking against VISN facilities can be useful since VISN facilities may be more similar than facilities in different geographic regions. Each year, the ASPWG reviews the current metrics, makes adjustments to address VISN priorities, and votes for approval of the metrics that will be tracked in the coming year.

Participation in an ASP collaborative streamlines the rollout of ASP and quality improvement initiatives across multiple sites, allowing ASPs to impact a greater number of veterans and evaluate initiatives on a larger scale. In 2019, with the anticipation of revised vancomycin dosing and monitoring guidelines, our ASPWG began to strategize the transition to AUC-based vancomycin monitoring.⁶ This multisite initiative showcases the strengths of implementing and evaluating practice changes as part of an ASP collaborative.

Vancomycin Dosing

The antibiotic vancomycin is used primarily for the treatment of MRSA infections.⁶ The 2020 consensus guidelines for vancomycin therapeutic monitoring recommend using the AUC to minimum inhibitory concentration (MIC) ratio as the pharmacodynamic target for serious MRSA infections, with an AUC/MIC goal of 400 to 600 mcg*h/mL.⁶ Prior guidelines recommended using vancomycin trough concentrations of 15 to 20 mcg/mL as a surrogate for this AUC target. However, subsequent studies have shown that trough-based dosing is associated with higher vancomycin exposures, supratherapeutic AUCs, and increased risk of vancomycin-associated acute kidney injury (AKI).^7,8 Therefore, more direct AUC estimation is now recommended.⁶ The preferred approach for AUC calculations is through Bayesian modeling. Due to limited resources and software availability, many facilities use an alternative method involving 2 postdistributive serum vancomycin concentrations and first-order pharmacokinetic equations. This approach can optimize vancomycin dosing but is more mathematically and logistically challenging. Transitioning from troughto AUC-based vancomycin monitoring requires careful planning and comprehensive staff education.

In 2019, the VISN 8 ASPWG created a comprehensive vancomycin AUC toolkit to facilitate implementation. Components included a pharmacokinetic management policy and procedure, a vancomycin dosing guide, a progress note template, educational materials specific to pharmacy, nursing, laboratory, and medical services, a pharmacist competency examination, and a vancomycin AUC calculator (eAppendix). Each component was developed by a subgroup with the understanding that sites could incorporate variations based on local practices and needs.

The vancomycin AUC calculator was developed to be user-friendly and included safety validation protocols to prevent the entry of erroneous data (eg, unrealistic patient weight or laboratory values). The calculator allowed users to copy data into the electronic health record to avoid manual transcription errors and improve operational efficiency. It offered suggested volume of distribution estimates and 2 methods to estimate elimination constant (Ke ) depending on the patient’s weight.^9,10 Creatinine clearance could be estimated using serum creatinine or cystatin C and considered amputation history. The default AUC goal in the calculator was 400 to 550 mcg*h/mL. This range was chosen based on consensus guidelines, data suggesting increased risk of AKI with AUCs > 515 mcg*h/mL, and the preference for conservative empiric dosing in the generally older VA population.¹¹ The calculator suggested loading doses of about 25 mg/kg with a 2500 mg limit. VHA facilities could make limited modifications to the calculator based on local policies and procedures (eg, adjusting default infusion times or a dosing intervals).

The VISN 8 Pharmacy Pharmacokinetic Dosing Manual was developed as a comprehensive document to guide pharmacy staff with dosing vancomycin across diverse patient populations. This document included recommendations for renal function assessment, patient-specific considerations when choosing an empiric vancomycin dose, methods of ordering vancomycin peak, trough, and surveillance levels, dose determination based on 2 levels, and other clinical insights or frequently asked questions.

ASPWG members presented an accredited continuing education webinar for pharmacists, which reviewed the rationale for AUC-targeted dosing, changes to the current pharmacokinetic dosing program, case-based scenarios across various patient populations, and potential challenges associated with vancomycin AUC-based dosing. A recording of the live training was also made available. A vancomycin AUC dosing competency test was developed with 11 basic pharmacokinetic and case-based questions and comprehensive explanations provided for each answer.

VHA facilities implemented AUC dosing in a staggered manner, allowing for lessons learned at earlier adopters to be addressed proactively at later sites. The dosing calculator and education documents were updated iteratively as opportunities for improvement were discovered. ASPWG members held local office hours to address questions or concerns from staff at their facilities. Sharing standardized materials across the VISN reduced individual site workload and complications in rolling out this complex new process.

VISN-WIDE QUALITY ASSURANCE

At the time of project conception, 4 of 7 VISN 8 health systems had transitioned to AUC-based dosing. A quality assurance protocol to compare patient outcomes before and after changing to AUC dosing was developed. Each site followed local protocols for project approval and data were deidentified, collected, and aggregated for analysis.

The primary objectives were to compare the incidence of AKI and persistent bacteremia and assess rates of AUC target attainment (400-600 mcg*h/mL) in the AUC-based and trough-based dosing groups.⁶ Data for both groups included anthropomorphic measurements, serum creatinine, amputation status, vancomycin dosing, and infection characteristics. The X² test was used for categorical data and the t test was used for continuous data. A 2-tailed α of 0.05 was used to determine significance. Each site sequentially reviewed all patients receiving ≥ 48 hours of intravenous vancomycin over a 3-month period and contributed up to 50 patients for each group. Due to staggered implementation, the study periods for sites spanned 2018 to 2023. A minimum 6-month washout period was observed between the trough and AUC groups at each site. Patients were excluded if pregnant, receiving renal replacement therapy, or presenting with AKI at the time of vancomycin initiation.

There were 168 patients in the AUC group and 172 patients in the trough group (Table 1). The rate of AUC target attainment with the initial dosing regimen varied across sites from 18% to 69% (mean, 48%). Total daily vancomycin exposure was lower in the AUC group compared with the trough group (2402 mg vs 2605 mg, respectively), with AUC-dosed patients being less likely to experience troughs level ≥ 15 or 20 mcg/mL (Table 2). There was a statistically significant lower rate of AKI in the AUC group: 2.4% in the AUC group (range, 2%-3%) vs 10.4% (range 7%-12%) in the trough group (P = .002). Rates of AKI were comparable to those observed in previous interventions.⁶ There was no statistical difference in length of stay, time to blood culture clearance, or rate of persistent bacteremia in the 2 groups, but these assessments were limited by sample size.

We did not anticipate such variability in initial target attainment across sites. The multisite quality assurance design allowed for qualitative evaluation of variability in dosing practices, which likely arose from sites and individual pharmacists having some flexibility in adjusting dosing tool parameters. Further analysis revealed that the facility with low initial target attainment was not routinely utilizing vancomycin loading doses. Sites routinely use robust loading doses achieved earlier and more consistent target attainment. Some sites used a narrower AUC target range in certain clinical scenarios (eg, > 500 mcg*h/mL for septic patients and < 500 mcg*h/mL for patients with less severe infections) rather than the 400 to 550 mcg*h/mL range for all patients. Sites targeting broader AUC ranges for all patients had higher rates of target attainment. Reviewing differences among sites allowed the ASPWG to identify best practices to optimize future care.

CONCLUSIONS

VHA ASPs must meet the standards outlined in VHA Directive 1031, including the new requirement for each VISN to develop an ASP collaborative. The VISN 8 ASPWG demonstrates how ASP champions can collaborate to solve common issues, complete tasks, explore new infectious diseases concepts, and impact large veteran populations. Furthermore, ASP collaboratives can harness their collective size to complete robust quality assurance evaluations that might otherwise be underpowered if completed at a single center. A limitation of the collaborative model is that a site with a robust ASP may already have specific practices in place. Expanding the ASP collaborative model further highlights the VHA role as a nationwide leader in ASP best practices.

Antimicrobial resistance is a global threat and burden to health care, with > 2.8 million antibiotic-resistant infections occurring annually in the United States.¹ To combat this issue and improve patient care, the US Department of Veterans Affairs (VA) has implemented antimicrobial stewardship programs (ASPs) across its health care systems. ASPs are multidisciplinary teams that promote evidence-based use of antimicrobials through activities supporting appropriate selection, dosing, route, and duration of antimicrobial therapy. ASP best practices are also included in the Joint Commission and Centers for Medicare and Medicaid Services accreditation standards.²

The foundational charge for VA facilities to develop and maintain ASPs was outlined in 2014 and updated in 2023 in the Veterans Health Administration (VHA) Directive 1031 on antimicrobial stewardship programs.2 This directive outlines specific requirements for all VA ASPs, including personnel, staffing levels, and the roles and responsibilities of all team members. VHA now requires that Veterans Integrated Services Networks (VISNs) establish robust ASP collaboratives. A VISN ASP collaborative consists of stewardship champions from each VA medical center in the VISN and is designed to support, develop, and enhance ASP programs across all facilities within that VISN.² Some VISNs may lack an ASP collaborative altogether, and others with existing groups may seek ways to expand their collaboratives in line with the updated directive. Prior to VHA Directive 1031, the VA Sunshine Healthcare Network (VISN 8) established an ASP collaborative. This article describes the structure and activities of the VISN 8 ASP collaborative and highlights a recent VISN 8 quality assurance initiative related to vancomycin area under the curve (AUC) dosing that illustrates how ASP collaboratives can enhance stewardship and clinical care across broad geographic areas.

VISN 8 ASP

The VHA, the largest integrated US health care system, is divided into 18 VISNs that provide regional systems of care to enhance access and meet the local health care needs of veterans.³ VISN 8 serves > 1.5 million veterans across 165,759 km² in Florida, South Georgia, Puerto Rico, and the US Virgin Islands.⁴ The network is composed of 7 health systems with 8 medical centers and > 60 outpatient clinics. These facilities provide comprehensive acute, primary, and specialty care, as well as mental health and extended care services in inpatient, outpatient, nursing home, and home care settings.⁴

The 2023 VHA Directive 1031 update recognizes the importance of VISN-level coordination of ASP activities to enhance the standardization of care and build partnerships in stewardship across all levels of care. The VISN 8 ASP collaborative workgroup (ASPWG) was established in 2015. Consistent with Directive 1031, the ASPWG is guided by clinician and pharmacist VISN leads. These leads serve as subject matter experts, facilitate access to resources, establish VISN-level consensus, and enhance communication among local ASP champions at medical centers within the VISN. All 7 health systems include = 1 ASP champion (clinician or pharmacist) in the ASPWG. Ad hoc members, whose routine duties are not solely focused on antimicrobial stewardship, contribute to specific stewardship projects as needed. For example, the ASPWG has included internal medicine, emergency department, community living center pharmacists, representatives from pharmacy administration, and trainees (pharmacy students and residents, and infectious diseases fellows) in antimicrobial stewardship initiatives. The inclusion of non-ASP champions is not discussed in VHA Directive 1031. However, these members have made valuable contributions to the ASPWG.

The ASPWG meets monthly. Agendas and priorities are developed by the VISN pharmacist and health care practitioner (HCP) leads. Monthly discussions may include but are not limited to a review of national formulary decisions, VISN goals and metrics, infectious diseases hot topics, pharmacoeconomic initiatives, strong practice presentations, regulatory and accreditation preparation, preparation of tracking reports, as well as the development of both patient-level and HCPlevel tools, resources, and education materials. This forum facilitates collaborative learning: members process and synthesize information, share and reframe ideas, and listen to other viewpoints to gain a complete understanding as a group.⁵ For example, ASPWG members have leaned on each other to prepare for Joint Commission accreditation surveys and strengthen the VISN 8 COVID-19 program through the rollout of vaccines and treatments. Other collaborative projects completed over the past few years included a penicillin allergy testing initiative and anti-methicillin-resistant Staphylococcus aureus (MRSA) and pseudomonal medication use evaluations. This team-centric problem-solving approach is highly effective while also fostering professional and social relationships. However, collaboratives could be perceived to have drawbacks. There may be opportunity costs if ASP time is allocated for issues that have already been addressed locally or concerns that standardization might hinder rapid adoption of practices at individual sites. Therefore, participation in each distinct group initiative is optional. This allows sites to choose projects related to their high priority areas and maintain bandwidth to implement practices not yet adopted by the larger group.

The ASPWG tracks metrics related to antimicrobial use with quarterly data presented by the VISN pharmacist lead. Both inpatient and outpatient metrics are evaluated, such as days of therapy per 1000 days and outpatient antibiotic prescriptions per 1000 unique patients. Facilities are benchmarked against their own historical data and other VISN sites, as well as other VISNs across the country. When outliers are identified, facilities are encouraged to conduct local projects to identify reasons for different antimicrobial use patterns and subsequent initiatives to optimize antimicrobial use. Benchmarking against VISN facilities can be useful since VISN facilities may be more similar than facilities in different geographic regions. Each year, the ASPWG reviews the current metrics, makes adjustments to address VISN priorities, and votes for approval of the metrics that will be tracked in the coming year.

Participation in an ASP collaborative streamlines the rollout of ASP and quality improvement initiatives across multiple sites, allowing ASPs to impact a greater number of veterans and evaluate initiatives on a larger scale. In 2019, with the anticipation of revised vancomycin dosing and monitoring guidelines, our ASPWG began to strategize the transition to AUC-based vancomycin monitoring.⁶ This multisite initiative showcases the strengths of implementing and evaluating practice changes as part of an ASP collaborative.

Vancomycin Dosing

The antibiotic vancomycin is used primarily for the treatment of MRSA infections.⁶ The 2020 consensus guidelines for vancomycin therapeutic monitoring recommend using the AUC to minimum inhibitory concentration (MIC) ratio as the pharmacodynamic target for serious MRSA infections, with an AUC/MIC goal of 400 to 600 mcg*h/mL.⁶ Prior guidelines recommended using vancomycin trough concentrations of 15 to 20 mcg/mL as a surrogate for this AUC target. However, subsequent studies have shown that trough-based dosing is associated with higher vancomycin exposures, supratherapeutic AUCs, and increased risk of vancomycin-associated acute kidney injury (AKI).^7,8 Therefore, more direct AUC estimation is now recommended.⁶ The preferred approach for AUC calculations is through Bayesian modeling. Due to limited resources and software availability, many facilities use an alternative method involving 2 postdistributive serum vancomycin concentrations and first-order pharmacokinetic equations. This approach can optimize vancomycin dosing but is more mathematically and logistically challenging. Transitioning from troughto AUC-based vancomycin monitoring requires careful planning and comprehensive staff education.

In 2019, the VISN 8 ASPWG created a comprehensive vancomycin AUC toolkit to facilitate implementation. Components included a pharmacokinetic management policy and procedure, a vancomycin dosing guide, a progress note template, educational materials specific to pharmacy, nursing, laboratory, and medical services, a pharmacist competency examination, and a vancomycin AUC calculator (eAppendix). Each component was developed by a subgroup with the understanding that sites could incorporate variations based on local practices and needs.

The vancomycin AUC calculator was developed to be user-friendly and included safety validation protocols to prevent the entry of erroneous data (eg, unrealistic patient weight or laboratory values). The calculator allowed users to copy data into the electronic health record to avoid manual transcription errors and improve operational efficiency. It offered suggested volume of distribution estimates and 2 methods to estimate elimination constant (Ke ) depending on the patient’s weight.^9,10 Creatinine clearance could be estimated using serum creatinine or cystatin C and considered amputation history. The default AUC goal in the calculator was 400 to 550 mcg*h/mL. This range was chosen based on consensus guidelines, data suggesting increased risk of AKI with AUCs > 515 mcg*h/mL, and the preference for conservative empiric dosing in the generally older VA population.¹¹ The calculator suggested loading doses of about 25 mg/kg with a 2500 mg limit. VHA facilities could make limited modifications to the calculator based on local policies and procedures (eg, adjusting default infusion times or a dosing intervals).

The VISN 8 Pharmacy Pharmacokinetic Dosing Manual was developed as a comprehensive document to guide pharmacy staff with dosing vancomycin across diverse patient populations. This document included recommendations for renal function assessment, patient-specific considerations when choosing an empiric vancomycin dose, methods of ordering vancomycin peak, trough, and surveillance levels, dose determination based on 2 levels, and other clinical insights or frequently asked questions.

ASPWG members presented an accredited continuing education webinar for pharmacists, which reviewed the rationale for AUC-targeted dosing, changes to the current pharmacokinetic dosing program, case-based scenarios across various patient populations, and potential challenges associated with vancomycin AUC-based dosing. A recording of the live training was also made available. A vancomycin AUC dosing competency test was developed with 11 basic pharmacokinetic and case-based questions and comprehensive explanations provided for each answer.

VHA facilities implemented AUC dosing in a staggered manner, allowing for lessons learned at earlier adopters to be addressed proactively at later sites. The dosing calculator and education documents were updated iteratively as opportunities for improvement were discovered. ASPWG members held local office hours to address questions or concerns from staff at their facilities. Sharing standardized materials across the VISN reduced individual site workload and complications in rolling out this complex new process.

VISN-WIDE QUALITY ASSURANCE

At the time of project conception, 4 of 7 VISN 8 health systems had transitioned to AUC-based dosing. A quality assurance protocol to compare patient outcomes before and after changing to AUC dosing was developed. Each site followed local protocols for project approval and data were deidentified, collected, and aggregated for analysis.

The primary objectives were to compare the incidence of AKI and persistent bacteremia and assess rates of AUC target attainment (400-600 mcg*h/mL) in the AUC-based and trough-based dosing groups.⁶ Data for both groups included anthropomorphic measurements, serum creatinine, amputation status, vancomycin dosing, and infection characteristics. The X² test was used for categorical data and the t test was used for continuous data. A 2-tailed α of 0.05 was used to determine significance. Each site sequentially reviewed all patients receiving ≥ 48 hours of intravenous vancomycin over a 3-month period and contributed up to 50 patients for each group. Due to staggered implementation, the study periods for sites spanned 2018 to 2023. A minimum 6-month washout period was observed between the trough and AUC groups at each site. Patients were excluded if pregnant, receiving renal replacement therapy, or presenting with AKI at the time of vancomycin initiation.

There were 168 patients in the AUC group and 172 patients in the trough group (Table 1). The rate of AUC target attainment with the initial dosing regimen varied across sites from 18% to 69% (mean, 48%). Total daily vancomycin exposure was lower in the AUC group compared with the trough group (2402 mg vs 2605 mg, respectively), with AUC-dosed patients being less likely to experience troughs level ≥ 15 or 20 mcg/mL (Table 2). There was a statistically significant lower rate of AKI in the AUC group: 2.4% in the AUC group (range, 2%-3%) vs 10.4% (range 7%-12%) in the trough group (P = .002). Rates of AKI were comparable to those observed in previous interventions.⁶ There was no statistical difference in length of stay, time to blood culture clearance, or rate of persistent bacteremia in the 2 groups, but these assessments were limited by sample size.

We did not anticipate such variability in initial target attainment across sites. The multisite quality assurance design allowed for qualitative evaluation of variability in dosing practices, which likely arose from sites and individual pharmacists having some flexibility in adjusting dosing tool parameters. Further analysis revealed that the facility with low initial target attainment was not routinely utilizing vancomycin loading doses. Sites routinely use robust loading doses achieved earlier and more consistent target attainment. Some sites used a narrower AUC target range in certain clinical scenarios (eg, > 500 mcg*h/mL for septic patients and < 500 mcg*h/mL for patients with less severe infections) rather than the 400 to 550 mcg*h/mL range for all patients. Sites targeting broader AUC ranges for all patients had higher rates of target attainment. Reviewing differences among sites allowed the ASPWG to identify best practices to optimize future care.

CONCLUSIONS

VHA ASPs must meet the standards outlined in VHA Directive 1031, including the new requirement for each VISN to develop an ASP collaborative. The VISN 8 ASPWG demonstrates how ASP champions can collaborate to solve common issues, complete tasks, explore new infectious diseases concepts, and impact large veteran populations. Furthermore, ASP collaboratives can harness their collective size to complete robust quality assurance evaluations that might otherwise be underpowered if completed at a single center. A limitation of the collaborative model is that a site with a robust ASP may already have specific practices in place. Expanding the ASP collaborative model further highlights the VHA role as a nationwide leader in ASP best practices.

Antimicrobial resistance is a global threat and burden to health care, with > 2.8 million antibiotic-resistant infections occurring annually in the United States.¹ To combat this issue and improve patient care, the US Department of Veterans Affairs (VA) has implemented antimicrobial stewardship programs (ASPs) across its health care systems. ASPs are multidisciplinary teams that promote evidence-based use of antimicrobials through activities supporting appropriate selection, dosing, route, and duration of antimicrobial therapy. ASP best practices are also included in the Joint Commission and Centers for Medicare and Medicaid Services accreditation standards.²

The foundational charge for VA facilities to develop and maintain ASPs was outlined in 2014 and updated in 2023 in the Veterans Health Administration (VHA) Directive 1031 on antimicrobial stewardship programs.2 This directive outlines specific requirements for all VA ASPs, including personnel, staffing levels, and the roles and responsibilities of all team members. VHA now requires that Veterans Integrated Services Networks (VISNs) establish robust ASP collaboratives. A VISN ASP collaborative consists of stewardship champions from each VA medical center in the VISN and is designed to support, develop, and enhance ASP programs across all facilities within that VISN.² Some VISNs may lack an ASP collaborative altogether, and others with existing groups may seek ways to expand their collaboratives in line with the updated directive. Prior to VHA Directive 1031, the VA Sunshine Healthcare Network (VISN 8) established an ASP collaborative. This article describes the structure and activities of the VISN 8 ASP collaborative and highlights a recent VISN 8 quality assurance initiative related to vancomycin area under the curve (AUC) dosing that illustrates how ASP collaboratives can enhance stewardship and clinical care across broad geographic areas.

VISN 8 ASP

The VHA, the largest integrated US health care system, is divided into 18 VISNs that provide regional systems of care to enhance access and meet the local health care needs of veterans.³ VISN 8 serves > 1.5 million veterans across 165,759 km² in Florida, South Georgia, Puerto Rico, and the US Virgin Islands.⁴ The network is composed of 7 health systems with 8 medical centers and > 60 outpatient clinics. These facilities provide comprehensive acute, primary, and specialty care, as well as mental health and extended care services in inpatient, outpatient, nursing home, and home care settings.⁴

The 2023 VHA Directive 1031 update recognizes the importance of VISN-level coordination of ASP activities to enhance the standardization of care and build partnerships in stewardship across all levels of care. The VISN 8 ASP collaborative workgroup (ASPWG) was established in 2015. Consistent with Directive 1031, the ASPWG is guided by clinician and pharmacist VISN leads. These leads serve as subject matter experts, facilitate access to resources, establish VISN-level consensus, and enhance communication among local ASP champions at medical centers within the VISN. All 7 health systems include = 1 ASP champion (clinician or pharmacist) in the ASPWG. Ad hoc members, whose routine duties are not solely focused on antimicrobial stewardship, contribute to specific stewardship projects as needed. For example, the ASPWG has included internal medicine, emergency department, community living center pharmacists, representatives from pharmacy administration, and trainees (pharmacy students and residents, and infectious diseases fellows) in antimicrobial stewardship initiatives. The inclusion of non-ASP champions is not discussed in VHA Directive 1031. However, these members have made valuable contributions to the ASPWG.

The ASPWG meets monthly. Agendas and priorities are developed by the VISN pharmacist and health care practitioner (HCP) leads. Monthly discussions may include but are not limited to a review of national formulary decisions, VISN goals and metrics, infectious diseases hot topics, pharmacoeconomic initiatives, strong practice presentations, regulatory and accreditation preparation, preparation of tracking reports, as well as the development of both patient-level and HCPlevel tools, resources, and education materials. This forum facilitates collaborative learning: members process and synthesize information, share and reframe ideas, and listen to other viewpoints to gain a complete understanding as a group.⁵ For example, ASPWG members have leaned on each other to prepare for Joint Commission accreditation surveys and strengthen the VISN 8 COVID-19 program through the rollout of vaccines and treatments. Other collaborative projects completed over the past few years included a penicillin allergy testing initiative and anti-methicillin-resistant Staphylococcus aureus (MRSA) and pseudomonal medication use evaluations. This team-centric problem-solving approach is highly effective while also fostering professional and social relationships. However, collaboratives could be perceived to have drawbacks. There may be opportunity costs if ASP time is allocated for issues that have already been addressed locally or concerns that standardization might hinder rapid adoption of practices at individual sites. Therefore, participation in each distinct group initiative is optional. This allows sites to choose projects related to their high priority areas and maintain bandwidth to implement practices not yet adopted by the larger group.

The ASPWG tracks metrics related to antimicrobial use with quarterly data presented by the VISN pharmacist lead. Both inpatient and outpatient metrics are evaluated, such as days of therapy per 1000 days and outpatient antibiotic prescriptions per 1000 unique patients. Facilities are benchmarked against their own historical data and other VISN sites, as well as other VISNs across the country. When outliers are identified, facilities are encouraged to conduct local projects to identify reasons for different antimicrobial use patterns and subsequent initiatives to optimize antimicrobial use. Benchmarking against VISN facilities can be useful since VISN facilities may be more similar than facilities in different geographic regions. Each year, the ASPWG reviews the current metrics, makes adjustments to address VISN priorities, and votes for approval of the metrics that will be tracked in the coming year.

Participation in an ASP collaborative streamlines the rollout of ASP and quality improvement initiatives across multiple sites, allowing ASPs to impact a greater number of veterans and evaluate initiatives on a larger scale. In 2019, with the anticipation of revised vancomycin dosing and monitoring guidelines, our ASPWG began to strategize the transition to AUC-based vancomycin monitoring.⁶ This multisite initiative showcases the strengths of implementing and evaluating practice changes as part of an ASP collaborative.

Vancomycin Dosing

The antibiotic vancomycin is used primarily for the treatment of MRSA infections.⁶ The 2020 consensus guidelines for vancomycin therapeutic monitoring recommend using the AUC to minimum inhibitory concentration (MIC) ratio as the pharmacodynamic target for serious MRSA infections, with an AUC/MIC goal of 400 to 600 mcg*h/mL.⁶ Prior guidelines recommended using vancomycin trough concentrations of 15 to 20 mcg/mL as a surrogate for this AUC target. However, subsequent studies have shown that trough-based dosing is associated with higher vancomycin exposures, supratherapeutic AUCs, and increased risk of vancomycin-associated acute kidney injury (AKI).^7,8 Therefore, more direct AUC estimation is now recommended.⁶ The preferred approach for AUC calculations is through Bayesian modeling. Due to limited resources and software availability, many facilities use an alternative method involving 2 postdistributive serum vancomycin concentrations and first-order pharmacokinetic equations. This approach can optimize vancomycin dosing but is more mathematically and logistically challenging. Transitioning from troughto AUC-based vancomycin monitoring requires careful planning and comprehensive staff education.

In 2019, the VISN 8 ASPWG created a comprehensive vancomycin AUC toolkit to facilitate implementation. Components included a pharmacokinetic management policy and procedure, a vancomycin dosing guide, a progress note template, educational materials specific to pharmacy, nursing, laboratory, and medical services, a pharmacist competency examination, and a vancomycin AUC calculator (eAppendix). Each component was developed by a subgroup with the understanding that sites could incorporate variations based on local practices and needs.

The vancomycin AUC calculator was developed to be user-friendly and included safety validation protocols to prevent the entry of erroneous data (eg, unrealistic patient weight or laboratory values). The calculator allowed users to copy data into the electronic health record to avoid manual transcription errors and improve operational efficiency. It offered suggested volume of distribution estimates and 2 methods to estimate elimination constant (Ke ) depending on the patient’s weight.^9,10 Creatinine clearance could be estimated using serum creatinine or cystatin C and considered amputation history. The default AUC goal in the calculator was 400 to 550 mcg*h/mL. This range was chosen based on consensus guidelines, data suggesting increased risk of AKI with AUCs > 515 mcg*h/mL, and the preference for conservative empiric dosing in the generally older VA population.¹¹ The calculator suggested loading doses of about 25 mg/kg with a 2500 mg limit. VHA facilities could make limited modifications to the calculator based on local policies and procedures (eg, adjusting default infusion times or a dosing intervals).

The VISN 8 Pharmacy Pharmacokinetic Dosing Manual was developed as a comprehensive document to guide pharmacy staff with dosing vancomycin across diverse patient populations. This document included recommendations for renal function assessment, patient-specific considerations when choosing an empiric vancomycin dose, methods of ordering vancomycin peak, trough, and surveillance levels, dose determination based on 2 levels, and other clinical insights or frequently asked questions.

ASPWG members presented an accredited continuing education webinar for pharmacists, which reviewed the rationale for AUC-targeted dosing, changes to the current pharmacokinetic dosing program, case-based scenarios across various patient populations, and potential challenges associated with vancomycin AUC-based dosing. A recording of the live training was also made available. A vancomycin AUC dosing competency test was developed with 11 basic pharmacokinetic and case-based questions and comprehensive explanations provided for each answer.

VHA facilities implemented AUC dosing in a staggered manner, allowing for lessons learned at earlier adopters to be addressed proactively at later sites. The dosing calculator and education documents were updated iteratively as opportunities for improvement were discovered. ASPWG members held local office hours to address questions or concerns from staff at their facilities. Sharing standardized materials across the VISN reduced individual site workload and complications in rolling out this complex new process.

VISN-WIDE QUALITY ASSURANCE

At the time of project conception, 4 of 7 VISN 8 health systems had transitioned to AUC-based dosing. A quality assurance protocol to compare patient outcomes before and after changing to AUC dosing was developed. Each site followed local protocols for project approval and data were deidentified, collected, and aggregated for analysis.

The primary objectives were to compare the incidence of AKI and persistent bacteremia and assess rates of AUC target attainment (400-600 mcg*h/mL) in the AUC-based and trough-based dosing groups.⁶ Data for both groups included anthropomorphic measurements, serum creatinine, amputation status, vancomycin dosing, and infection characteristics. The X² test was used for categorical data and the t test was used for continuous data. A 2-tailed α of 0.05 was used to determine significance. Each site sequentially reviewed all patients receiving ≥ 48 hours of intravenous vancomycin over a 3-month period and contributed up to 50 patients for each group. Due to staggered implementation, the study periods for sites spanned 2018 to 2023. A minimum 6-month washout period was observed between the trough and AUC groups at each site. Patients were excluded if pregnant, receiving renal replacement therapy, or presenting with AKI at the time of vancomycin initiation.

There were 168 patients in the AUC group and 172 patients in the trough group (Table 1). The rate of AUC target attainment with the initial dosing regimen varied across sites from 18% to 69% (mean, 48%). Total daily vancomycin exposure was lower in the AUC group compared with the trough group (2402 mg vs 2605 mg, respectively), with AUC-dosed patients being less likely to experience troughs level ≥ 15 or 20 mcg/mL (Table 2). There was a statistically significant lower rate of AKI in the AUC group: 2.4% in the AUC group (range, 2%-3%) vs 10.4% (range 7%-12%) in the trough group (P = .002). Rates of AKI were comparable to those observed in previous interventions.⁶ There was no statistical difference in length of stay, time to blood culture clearance, or rate of persistent bacteremia in the 2 groups, but these assessments were limited by sample size.

We did not anticipate such variability in initial target attainment across sites. The multisite quality assurance design allowed for qualitative evaluation of variability in dosing practices, which likely arose from sites and individual pharmacists having some flexibility in adjusting dosing tool parameters. Further analysis revealed that the facility with low initial target attainment was not routinely utilizing vancomycin loading doses. Sites routinely use robust loading doses achieved earlier and more consistent target attainment. Some sites used a narrower AUC target range in certain clinical scenarios (eg, > 500 mcg*h/mL for septic patients and < 500 mcg*h/mL for patients with less severe infections) rather than the 400 to 550 mcg*h/mL range for all patients. Sites targeting broader AUC ranges for all patients had higher rates of target attainment. Reviewing differences among sites allowed the ASPWG to identify best practices to optimize future care.

CONCLUSIONS

VHA ASPs must meet the standards outlined in VHA Directive 1031, including the new requirement for each VISN to develop an ASP collaborative. The VISN 8 ASPWG demonstrates how ASP champions can collaborate to solve common issues, complete tasks, explore new infectious diseases concepts, and impact large veteran populations. Furthermore, ASP collaboratives can harness their collective size to complete robust quality assurance evaluations that might otherwise be underpowered if completed at a single center. A limitation of the collaborative model is that a site with a robust ASP may already have specific practices in place. Expanding the ASP collaborative model further highlights the VHA role as a nationwide leader in ASP best practices.

The US Department of Veterans Affairs (VA) now has a permanent pharmacogenomics service that provides genetic tests to give clinicians insight into the best medication options for their patients.

The tests, which have no extra cost, are available to all veterans, said pharmacist Jill S. Bates, PharmD, MS, executive director of the VA National Pharmacogenomics Program, who spoke in an interview and a presentation at the annual meeting of the Association of VA Hematology/Oncology.

Genetic testing is “a tool that can help optimize care that we provide for veterans,” she said. “Pharmacogenomics is additional information to help the clinician make a decision. We know that most veterans—greater than 90%—carry a variant in a pharmacogenomics gene that is actionable.”

The genetic tests can provide insight into the optimal medication for multiple conditions such as mental illness, gastrointestinal disorders, cancer, pain, and heart disease. According to a 2019 analysis of over 6 years of data, more than half of the VA patient population used medications whose efficacy may have been affected by detectable genetic variants.

For instance, Bates said tests can let clinicians know whether patients are susceptible to statin-associated muscle adverse effects if they take simvastatin, the cholesterol medication. An estimated 25.6% of the VA population has this variant.

Elsewhere on the cardiac front, an estimated 58.3% of the VA population has a genetic variant that increases sensitivity to the blood thinner warfarin.

Testing could help psychiatrists determine whether certain medications should not be prescribed—or should be prescribed at lower doses—in patients who’ve had adverse reactions to antidepressants, Bates said.

In cancer, Bates said, genetic testing can identify patients who have a genetic variant that boosts toxicity from fluoropyrimidine chemotherapy treatments, which include capecitabine, floxuridine, and fluorouracil. Meanwhile, an estimated 0.9% will have no reaction or limited reaction to capecitabine and fluorouracil, and 4.8% will have hypersensitivity to carbamazepine and oxcarbazepine. 

Tests can also identify a genetic variant that can lead to poor metabolism of the chemotherapy drug irinotecan, which is used to treat colon cancer. “In those patients, you’d want to reduce the dose by 20%,” Bates said. In other cases, alternate drugs may be the best strategy to address genetic variations.

Prior to 2019, clinicians had to order pharmacogenomic tests outside of the VA system, according to Bates. That year, a donation from Sanford Health brought VA pharmacogenomics to 40 pilot sites. Since then, more than 88,000 tests have been performed.

The VA has now made its pharmacogenomic program permanent, Bates said. As of early September, testing was available at 139 VA sites and is coming soon to 4 more. It’s not available at another 23 sites that are scattered across the country.

A tool in the VA electronic health record now reminds clinicians about the availability of genetic testing and allows them to order tests. However, testing isn’t available for patients who have had liver transplants or certain bone marrow transplants.

The VA is working on developing decision-making tools to help clinicians determine when the tests are appropriate, Bates said. It typically takes 2 to 3 weeks to get results, she said, adding that external laboratories provide results. “We eventually would like to bring in all pharmacogenomics testing to be conducted within the VA enterprise.”

Bates reported that she had no disclosures.

The US Department of Veterans Affairs (VA) now has a permanent pharmacogenomics service that provides genetic tests to give clinicians insight into the best medication options for their patients.

The tests, which have no extra cost, are available to all veterans, said pharmacist Jill S. Bates, PharmD, MS, executive director of the VA National Pharmacogenomics Program, who spoke in an interview and a presentation at the annual meeting of the Association of VA Hematology/Oncology.

Genetic testing is “a tool that can help optimize care that we provide for veterans,” she said. “Pharmacogenomics is additional information to help the clinician make a decision. We know that most veterans—greater than 90%—carry a variant in a pharmacogenomics gene that is actionable.”

The genetic tests can provide insight into the optimal medication for multiple conditions such as mental illness, gastrointestinal disorders, cancer, pain, and heart disease. According to a 2019 analysis of over 6 years of data, more than half of the VA patient population used medications whose efficacy may have been affected by detectable genetic variants.

For instance, Bates said tests can let clinicians know whether patients are susceptible to statin-associated muscle adverse effects if they take simvastatin, the cholesterol medication. An estimated 25.6% of the VA population has this variant.

Elsewhere on the cardiac front, an estimated 58.3% of the VA population has a genetic variant that increases sensitivity to the blood thinner warfarin.

Testing could help psychiatrists determine whether certain medications should not be prescribed—or should be prescribed at lower doses—in patients who’ve had adverse reactions to antidepressants, Bates said.

In cancer, Bates said, genetic testing can identify patients who have a genetic variant that boosts toxicity from fluoropyrimidine chemotherapy treatments, which include capecitabine, floxuridine, and fluorouracil. Meanwhile, an estimated 0.9% will have no reaction or limited reaction to capecitabine and fluorouracil, and 4.8% will have hypersensitivity to carbamazepine and oxcarbazepine. 

Tests can also identify a genetic variant that can lead to poor metabolism of the chemotherapy drug irinotecan, which is used to treat colon cancer. “In those patients, you’d want to reduce the dose by 20%,” Bates said. In other cases, alternate drugs may be the best strategy to address genetic variations.

Prior to 2019, clinicians had to order pharmacogenomic tests outside of the VA system, according to Bates. That year, a donation from Sanford Health brought VA pharmacogenomics to 40 pilot sites. Since then, more than 88,000 tests have been performed.

The VA has now made its pharmacogenomic program permanent, Bates said. As of early September, testing was available at 139 VA sites and is coming soon to 4 more. It’s not available at another 23 sites that are scattered across the country.

A tool in the VA electronic health record now reminds clinicians about the availability of genetic testing and allows them to order tests. However, testing isn’t available for patients who have had liver transplants or certain bone marrow transplants.

The VA is working on developing decision-making tools to help clinicians determine when the tests are appropriate, Bates said. It typically takes 2 to 3 weeks to get results, she said, adding that external laboratories provide results. “We eventually would like to bring in all pharmacogenomics testing to be conducted within the VA enterprise.”

Bates reported that she had no disclosures.

The US Department of Veterans Affairs (VA) now has a permanent pharmacogenomics service that provides genetic tests to give clinicians insight into the best medication options for their patients.

The tests, which have no extra cost, are available to all veterans, said pharmacist Jill S. Bates, PharmD, MS, executive director of the VA National Pharmacogenomics Program, who spoke in an interview and a presentation at the annual meeting of the Association of VA Hematology/Oncology.

Genetic testing is “a tool that can help optimize care that we provide for veterans,” she said. “Pharmacogenomics is additional information to help the clinician make a decision. We know that most veterans—greater than 90%—carry a variant in a pharmacogenomics gene that is actionable.”

The genetic tests can provide insight into the optimal medication for multiple conditions such as mental illness, gastrointestinal disorders, cancer, pain, and heart disease. According to a 2019 analysis of over 6 years of data, more than half of the VA patient population used medications whose efficacy may have been affected by detectable genetic variants.

For instance, Bates said tests can let clinicians know whether patients are susceptible to statin-associated muscle adverse effects if they take simvastatin, the cholesterol medication. An estimated 25.6% of the VA population has this variant.

Elsewhere on the cardiac front, an estimated 58.3% of the VA population has a genetic variant that increases sensitivity to the blood thinner warfarin.

Testing could help psychiatrists determine whether certain medications should not be prescribed—or should be prescribed at lower doses—in patients who’ve had adverse reactions to antidepressants, Bates said.

In cancer, Bates said, genetic testing can identify patients who have a genetic variant that boosts toxicity from fluoropyrimidine chemotherapy treatments, which include capecitabine, floxuridine, and fluorouracil. Meanwhile, an estimated 0.9% will have no reaction or limited reaction to capecitabine and fluorouracil, and 4.8% will have hypersensitivity to carbamazepine and oxcarbazepine. 

Tests can also identify a genetic variant that can lead to poor metabolism of the chemotherapy drug irinotecan, which is used to treat colon cancer. “In those patients, you’d want to reduce the dose by 20%,” Bates said. In other cases, alternate drugs may be the best strategy to address genetic variations.

Prior to 2019, clinicians had to order pharmacogenomic tests outside of the VA system, according to Bates. That year, a donation from Sanford Health brought VA pharmacogenomics to 40 pilot sites. Since then, more than 88,000 tests have been performed.

The VA has now made its pharmacogenomic program permanent, Bates said. As of early September, testing was available at 139 VA sites and is coming soon to 4 more. It’s not available at another 23 sites that are scattered across the country.

A tool in the VA electronic health record now reminds clinicians about the availability of genetic testing and allows them to order tests. However, testing isn’t available for patients who have had liver transplants or certain bone marrow transplants.

The VA is working on developing decision-making tools to help clinicians determine when the tests are appropriate, Bates said. It typically takes 2 to 3 weeks to get results, she said, adding that external laboratories provide results. “We eventually would like to bring in all pharmacogenomics testing to be conducted within the VA enterprise.”

Bates reported that she had no disclosures.

Background

Veterans receiving care in the community (CITC) are prescribed oral oncology medications to be filled at VA pharmacies. Many of the outpatient prescriptions written for oncology medications require a prior authorization review by a pharmacist. A standardized workflow to obtain outside records to ensure patient safety, appropriate therapeutic selections, and maximize cost avoidance was established in March 2023. This quality improvement project evaluated the implementation of a clinical peer-to-peer prescription referral process between operational and oncology clinical pharmacists (CPS) to include a prior authorization drug request (PADR) review.

Methods

A retrospective chart review was completed to assess the effectiveness of the CITC Rx review process. Patients who had a CITC PADR consult entered between April 2023 and March 2024 were included. Metrics obtained included medication ordered, diagnosis, line of treatment, date prescription received, time to PADR completion, PADR outcome, FDA approval status, and conformity to VA National Oncology Program (NOP) disease pathway. Descriptive statistics were used to describe the data.

Results

Top reasons for referral for CITC included best medical interest and drive time. Fifty-one PADR requests were submitted for 41 patients. Forty-six PADR consults were completed. Approval rate was 85%. Consults involved 32 different oncolytics, 78% had VA Pharmacy Benefits Manager criteria for use. Thirty-seven percent of the PADR requests adhered to the NOP pathways. Approximately 30% of PADR requests did not have an associated NOP pathway. Seventy-four percent of drugs had an associated FDA approval. On average, two calls were made to CITC provider by the operational pharmacist to obtain necessary information for clinical review, resulting in a 5 day time to PADR entry. The average time to PADR consult completion was 9.5 hours. Four interventions addressed drug interactions or dosing adjustments.

Conclusions

This review demonstrated the feasibility and framework for implementing a standardized peer-to-peer PADR consult review process for CITC prescriptions requiring prior authorization. Having separate intake of CITC prescriptions by the operational pharmacist who is responsible for obtaining outside records, the CPS provided a timely clinical review of PADR consults, assuring appropriate therapeutic selections to maximize cost avoidance while maintaining patient safety.

Background

Veterans receiving care in the community (CITC) are prescribed oral oncology medications to be filled at VA pharmacies. Many of the outpatient prescriptions written for oncology medications require a prior authorization review by a pharmacist. A standardized workflow to obtain outside records to ensure patient safety, appropriate therapeutic selections, and maximize cost avoidance was established in March 2023. This quality improvement project evaluated the implementation of a clinical peer-to-peer prescription referral process between operational and oncology clinical pharmacists (CPS) to include a prior authorization drug request (PADR) review.

Methods

A retrospective chart review was completed to assess the effectiveness of the CITC Rx review process. Patients who had a CITC PADR consult entered between April 2023 and March 2024 were included. Metrics obtained included medication ordered, diagnosis, line of treatment, date prescription received, time to PADR completion, PADR outcome, FDA approval status, and conformity to VA National Oncology Program (NOP) disease pathway. Descriptive statistics were used to describe the data.

Results

Top reasons for referral for CITC included best medical interest and drive time. Fifty-one PADR requests were submitted for 41 patients. Forty-six PADR consults were completed. Approval rate was 85%. Consults involved 32 different oncolytics, 78% had VA Pharmacy Benefits Manager criteria for use. Thirty-seven percent of the PADR requests adhered to the NOP pathways. Approximately 30% of PADR requests did not have an associated NOP pathway. Seventy-four percent of drugs had an associated FDA approval. On average, two calls were made to CITC provider by the operational pharmacist to obtain necessary information for clinical review, resulting in a 5 day time to PADR entry. The average time to PADR consult completion was 9.5 hours. Four interventions addressed drug interactions or dosing adjustments.

Conclusions

This review demonstrated the feasibility and framework for implementing a standardized peer-to-peer PADR consult review process for CITC prescriptions requiring prior authorization. Having separate intake of CITC prescriptions by the operational pharmacist who is responsible for obtaining outside records, the CPS provided a timely clinical review of PADR consults, assuring appropriate therapeutic selections to maximize cost avoidance while maintaining patient safety.

Background

Veterans receiving care in the community (CITC) are prescribed oral oncology medications to be filled at VA pharmacies. Many of the outpatient prescriptions written for oncology medications require a prior authorization review by a pharmacist. A standardized workflow to obtain outside records to ensure patient safety, appropriate therapeutic selections, and maximize cost avoidance was established in March 2023. This quality improvement project evaluated the implementation of a clinical peer-to-peer prescription referral process between operational and oncology clinical pharmacists (CPS) to include a prior authorization drug request (PADR) review.

Methods

A retrospective chart review was completed to assess the effectiveness of the CITC Rx review process. Patients who had a CITC PADR consult entered between April 2023 and March 2024 were included. Metrics obtained included medication ordered, diagnosis, line of treatment, date prescription received, time to PADR completion, PADR outcome, FDA approval status, and conformity to VA National Oncology Program (NOP) disease pathway. Descriptive statistics were used to describe the data.

Results

Top reasons for referral for CITC included best medical interest and drive time. Fifty-one PADR requests were submitted for 41 patients. Forty-six PADR consults were completed. Approval rate was 85%. Consults involved 32 different oncolytics, 78% had VA Pharmacy Benefits Manager criteria for use. Thirty-seven percent of the PADR requests adhered to the NOP pathways. Approximately 30% of PADR requests did not have an associated NOP pathway. Seventy-four percent of drugs had an associated FDA approval. On average, two calls were made to CITC provider by the operational pharmacist to obtain necessary information for clinical review, resulting in a 5 day time to PADR entry. The average time to PADR consult completion was 9.5 hours. Four interventions addressed drug interactions or dosing adjustments.

Conclusions

This review demonstrated the feasibility and framework for implementing a standardized peer-to-peer PADR consult review process for CITC prescriptions requiring prior authorization. Having separate intake of CITC prescriptions by the operational pharmacist who is responsible for obtaining outside records, the CPS provided a timely clinical review of PADR consults, assuring appropriate therapeutic selections to maximize cost avoidance while maintaining patient safety.