The burden of undertreatment and non-treatment among patients with non-valvular atrial fibrillation and elevated stroke risk: a systematic review

Abstract

Objective

Global treatment guidelines recommend treatment with oral anticoagulants (OACs) for patients with non-valvular atrial fibrillation (NVAF) and an elevated stroke risk. However, not all patients with NVAF and an elevated stroke risk receive guideline-recommended therapy. A literature review and synthesis of observational studies were undertaken to identify the body of evidence on untreated and undertreated NVAF and the association with clinical and economic outcomes.

Methods

An extensive search (1/2010–4/2020) of MEDLINE, the Cochrane Library, conference proceedings, and health technology assessments (HTAs) was conducted. Studies must have evaluated rates of nontreatment or undertreatment in NVAF. Nontreatment was defined as absence of OACs (but with possible antiplatelet treatment), while undertreatment was defined as treatment with only antiplatelet agents.

Results

Sixteen studies met our inclusion criteria. Rates of nontreatment for patients with elevated stroke risk ranged from 2.0–51.1%, while rates of undertreatment ranged from 10.0–45.1%. The clinical benefits of anticoagulation were reported in the evaluated studies with reductions in stroke and mortality outcomes observed among patients treated with anticoagulants compared to untreated or undertreated patients. Adverse events associated with all bleeding types (i.e. hemorrhagic stroke, major bleeding or gastrointestinal hemorrhaging) were found to be higher for warfarin patients compared to untreated patients in real-world practice. Healthcare resource utilization was found to be lower among patients highly-adherent to warfarin compared to untreated patients.

Conclusions

Rates of nontreatment and undertreatment among NVAF patients remain high and are associated with preventable cardiovascular events and death. Strategies to increase rates of treatment may improve clinical outcomes.

Keywords:

• Non-valvular atrial fibrillation
• stroke
• bleeding
• warfarin
• direct oral anticoagulant

Introduction

Atrial fibrillation (AF), the most common form of arrhythmia with 5.2 million prevalent cases in the United States (US) in 2010, is predicted to affect more than 12 million people in the US by the year 2030¹. The prevalence of AF among the general population is approximately 1%, but it increases with age to approximately 9% for persons 80 years or older². AF leads to more than 450,000 hospitalizations annually in the US³, and contributes to more than 150,000 deaths each year⁴. Non-valvular AF (NVAF), defined as AF in the absence of moderate-to-severe mitral stenosis or a mechanical heart valve⁵, accounts for 95% of all AF patients².

AF has been associated with a four- to five-fold increased risk of ischemic stroke due to the high risk of clotting leading to thromboembolic events⁶. An individual’s risk of stroke is measured by CHA₂DS₂-VASc score, which is a composite measure of the presence of age, history of cardiovascular comorbidities and acute events, non-cardiovascular comorbidities, and sex, with weights for each criterion. The score is commonly used in clinical practice to estimate the risk of stroke and, by extension, to determine the need for anticoagulation⁵.

Anticoagulant guidelines issued by the American College of Cardiology/American Heart Association Task Force on Clinical Practice Guidelines and the Heart Rhythm Society (AHA/ACC/HRS) in 2014 and 2019 recommended OACs, including vitamin K antagonists (VKAs) and direct oral anticoagulants (DOACs), for patients with an elevated risk of stroke (CHA₂DS₂-VASc score of ≥2 in men or ≥3 in women)^5,7–11. Moreover, the 2014 AHA/ACC/HRS guidelines considered aspirin an option for patients with NVAF and a moderate risk of stroke (CHA₂DS₂-VASc score of 1) but not an option for patients with an elevated risk of stroke; patients with an elevated risk of stroke who were treated with aspirin only could therefore be considered an undertreated patient population.

Warfarin, a VKA, has historically been the first-line stroke prevention option in patients with NVAF¹²; trials have shown that warfarin can reduce the relative risk of stroke by up to two-thirds^6,13. DOACs, a novel class of OACs including factor Xa and direct thrombin inhibitors, were first approved in 2010^14,15–17 based on their noninferiority of efficacy and safety compared with warfarin in randomized controlled trials^18–21. The 2019 AHA/ACC/HRS guidelines recommend DOACs as the first-line OAC for stroke prevention ahead of VKAs, including warfarin⁵.

Although guidelines recommend treatment for NVAF with anticoagulants, a large proportion of patients in the US remain untreated. A study conducted by the National Cardiovascular Data Registry (NCDR) Practice Innovation and Clinical Excellence (PINNACLE) program found that only 55% of patients deemed eligible for warfarin were actually treated with that agent²². Additionally, almost 35% of NVAF patients eligible for warfarin were not treated with either warfarin or antithrombotic or antiplatelet agents, which was discordant with treatment guidelines during the time of the study’s analysis²². There are several reasons why patients with NVAF remain untreated, including physician underestimation of patient stroke risk, physician overestimation of bleeding risk, and patient reluctance to maintain adherence to OACs due to pharmacokinetics and interactions with food and other medications²³. Furthermore, educating physicians and patients regarding the benefits of anticoagulant treatment are crucial to increasing rates and compliance associated with anticoagulation²³.

When left untreated, NVAF poses a substantial burden to society, as stroke events are estimated to cost the US healthcare system about $34 billion each year in both direct medical costs and indirect productivity losses (e.g. absenteeism, presenteeism, disability)²⁴. From both clinical and economic perspectives, there remains a high unmet need despite current viable treatment options. Understanding the rates of anticoagulant treatment, as well as the burden posed by nonadherence to treatment guidelines, may further inform clinicians and patients about the benefits of anticoagulant treatment. The goals of this research project were three-fold: (1) to conduct a systematic literature review to identify rates of undertreatment or nontreatment with OACs in NVAF, and assess them over time, (2) evaluate the clinical and economic burden of undertreated or untreated NVAF, and (3) identify gaps in the literature corresponding to the clinical and economic consequences of undertreatment or nontreatment.

Methods

Inclusion and exclusion criteria

A global systematic literature search was conducted in accordance with guidelines developed by PRISMA²⁵, and the study protocol was registered with PROSPERO (ID: CRD42020187381). An extensive search of the literature was conducted over a 10-year period, from 1 January 2010 to 15 April 2020, for observational studies evaluating the burden of untreated or undertreated NVAF. Sources included MEDLINE (PubMed, including Medline in-process), GoogleScholar, the Cochrane Library, and relevant cardiovascular or pharmacoeconomic/pharmacoepidemiologic conference proceedings. The following search terms were explored in the systematic search process: (“atrial fibrillation”) AND (“epidem*” OR “prevalence” OR “incidence” Or “burden” OR “cost” OR “economic”). The key search terms included “atrial fibrillation” in an effort to capture the full body of evidence relating to the study objectives. Studies were limited to those written in the English language. Only full-text publications of research studies including database analyses, systematic literature reviews, meta-analyses, and registry analyses were included in our review. Studies were excluded if they focused exclusively on valvular AF, did not have full-length publications or did not provide an abstract.

The literature search and study selection were conducted in accordance with relevant best practices²⁵. Three reviewers (NA, CT, MS) applied criteria to search results to identify relevant studies. See the supplemental materials for more details.

Study populations must have included patients with an elevated stroke risk based on the risk score described in the publication, either CHADS₂ (score ≥2 based on 2011 ACCF/AHA/HRS guidelines²⁶) or CHA₂DS₂-VASc (score ≥2 based on 2014 AHA/ACC/HRS guidelines²). Patients with elevated stroke risk receive the strongest recommendations from global treatment guidelines with regard to initiating and continuing anticoagulation treatment, and thus were the focus of this review.

Treatment definitions

AHA/ACC/HRS treatment guidelines have evolved over time as DOACs have become more commonly used. The 2014 AHA/ACC/HRS treatment guidelines considered aspirin an option for patients with moderate risk of stroke and suggested that patients with high risk would be considered undertreated if treated only with aspirin. In contrast, the 2019 AHA/ACC/HRS treatment guidelines indicate that patients with NVAF, regardless of stroke risk, should not receive monotherapy with antiplatelets unless contraindicated to an OAC.

For the purpose of this analysis, the 2014 AHA/ACC/HRS treatment guidelines were used to define rates of treatment for two reasons. First, despite the change in the guidelines from 2014 to 2019, aspirin monotherapy for stroke prevention is still observed among those patients without an OAC contraindication^5,11. Second, the time horizon of this literature review dated back to 1 January 2010 and therefore included multiple iterations of guidelines. Using the current 2019 AHA/ACC/HRS guidelines for this analysis would preclude populations treated with antiplatelets. Applying this standard to older studies would exclude an important body of evidence and would not accurately reflect real-world prescribing behavior in prior years. Thus, inclusion of rates and outcomes related to aspirin use only is warranted.

Undertreated patients were defined as those high-risk patients treated with antiplatelet agents only. Untreated patients were defined as those who did not receive anticoagulant therapy; depending on the specificity of the publication under review, untreated patients may or may not have been treated with antiplatelets.

Study measures

For each study meeting our selection criteria, extracted study measures included rates of treatment by class (i.e. DOACs, warfarin, antiplatelets/undertreatment, nontreatment) and associated clinical and economic outcomes. Clinical outcomes included rates of ischemic stroke events, death, and bleeding events (i.e. hemorrhagic stroke events, major bleeding events, gastrointestinal hemorrhaging). Economic outcomes included healthcare resource utilization and direct medical costs corresponding to the clinical outcomes as well as indirect costs (i.e. lost productivity costs, caregiver costs). Treatment rates were stratified by geographic setting (i.e. US vs. ex-US) and separately by time period. The time period under evaluation was segmented into two periods, 1 January 2010–December 2012 (termed the “pre-DOAC era”) and January 2013–15 April 2020 (termed the “DOAC era”). The end of 2012 was used as the threshold time point since the three most prescribed DOACs were approved and on the market²⁷.

Risk of bias assessment

The modified Newcastle-Ottawa Scale for cohort studies was used to assess the quality of studies reporting an association between treatment rates for NVAF and clinical and/or economic outcomes²⁸. Studies were graded on a scale of 0 (lowest quality) to 9 (highest quality) by a single reviewer and validated by a second reviewer.

Results

Search results

A total of 11,231 publications were reviewed for inclusion in this analysis, including peer-reviewed publications, non-peer-reviewed HTA documents, and conference abstracts and posters. After applying inclusion and exclusion criteria, our initial full-text review identified 43 manuscripts^22,29–69, published between 2011 and 2020 that evaluated the burden of NVAF in undertreated, untreated, and treated populations. Figure 1 provides a complete flowchart of the study selection methodology. Further review clarified that a subset of sixteen manuscripts stratified findings by stroke risk score which are the focus of this analysis^29–44. Table 1 provides a detailed summary of the sixteen manuscripts.

Figure 1. Attrition chart using PRISMA criteria. Abbreviations. AF, atrial fibrillation; CADTH, Canadian Agency for Drugs and Technologies in Health; HTA, health technology assessment; ICER, Institute for Clinical and Economic Review; NICE, National Institute for Health and Care Excellence; PRISMA, Preferred Reporting Items for Systematic Review and Meta-Analyses.

Table 1. Summary table of studies of the burden of nontreatment and undertreatment in NVAF.

Study	Country	Study type and data source	Study population size	Study follow-up (months)	Baseline patient characteristics		Anticoagulation treatment rates	Impact of treatment on clinical outcomes	Impact of treatment on healthcare resource utilization	Impact of treatment on healthcare costs
Apenteng 2018	United Kingdom	Study of GARFIELD-AF registry, 2011–2016	3482	Mean = 24	CHF	7.2%	High-risk population: 47.9% treated with warfarin, 20.3% treated with DOAC, 20.7% undertreated, 11.1% untreated	Not reported	Not reported	Not reported
					Hypertension	69.6%
					Diabetes	17.6%
					Vascular disease	13.6%
					Stroke or TIA	12.3%
Casciano 2013	United States	Retrospective cohort study of Thomson Reuters MarketScan, 2003–2007	13,289	Mean = 15.6	CHF	23.8%	High-risk population: 52.9% treated with anticoagulants, 47.1% untreated	Incidence of ischemic stroke events per 100 person-years: Untreated: 4.41 Low adherence to warfarin: 1.87 High adherence to warfarin: 0.72 Adherence to  warfarin was defined by  proportion of days covered (PDC) by warfarin: >0.8 (high) and ≤0.8 (low)	High adherence warfarin treatment was associated with 7% fewer inpatient encounters (p < .001), 16% fewer ER visits (p = .019), and nearly a 40% reduction in the number of hospital days (p < .001) compared with patients who did not receive warfarin Inpatient and ER services utilization were not significantly different between patients with low warfarin adherence and patients who did not receive warfarin	Total costs by warfarin adherence (costs presented for inpatient, outpatient, medical cost, etc.): No warfarin = $1344 Low adherence = $1384 High adherence = $1165 Adherence to  warfarin was defined by proportion of days covered (PDC) by warfarin: > 0.8 (high) and ≤ 0.8 (low)
					Hypertension	75.1%
					Diabetes	34.8%
					Vascular disease (PAD)	11.1%
					Stroke or TIA	10.9%
de Andrés-Nogales 2015	Spain	Retrospective cohort study of the Real Evidence of Anti Coagulation Treatment in AF (REACT-AF) cohort, 2010–2012	12,514	Not reported	CHF	20.0%	Overall population: 60.1% treated with warfarin, 1.3% treated with DOACs, 30.9% undertreated, 7.7% untreated.	Bleeding incidence rates per 100 person-years: DOAC: 0.6 (95% CI: 0.0–3.4) Antiplatelets: 2.4 (95% CI: 2.1–2.7) VKAs: 2.4 (95% CI: 2.0–2.8) Untreated: 1.7 (95% CI: 1.3–2.3) Ischemic stroke  incidence rates  per 100  person-years: Untreated : 3.1; 95% CI: 2.67–3.55 Antiplatelets: 0.6; 95% CI: 0.49–0.77 DOACs: 0.0; 95% CI: 0.0–10.4 VKAs: 0.3; 95%CI: 0.22–0.35	Not reported	Not reported
					Hypertension	Not reported
					Diabetes	26.1%
					Vascular disease	7.5%
					Stroke or TIA	6.9%
Frewen 2013	Ireland	Prospective cohort study of Irish Longitudinal Study on Ageing, 2009–2011	118	Not reported	Not reported		High-risk population: 40.7% treated with anticoagulants, 34.6% undertreated, 24.7% untreated	Not reported	Not reported	Not reported
Hess 2017	United States	Study of ORBIT-AF registry, 2010–2011	9553	Median = 30	CHF	35.6%	High-risk population: 78.7% treated with warfarin, 21.3% untreated	- Patients not receiving OAC with CHADS-VASc ≥ 2 were associated with higher adjusted rates of death (OR = 1.22), stroke/non-central nervous system embolism (OR = 1.14), stroke/non-central nervous system embolism/TIA (OR = 1.18), and composite outcome of stroke and non-central nervous system embolism (OR = 1.19), and lower bleeding rates (OR = 0.35) [ORs adjusted for propensity to receive OAC]	Not reported	Not reported
					Hypertension	87.3%
					Diabetes	32.2%
					Vascular disease (MI + CAD)	56.9%
					Stroke or TIA	16.4%
Holt 2013	United Kingdom	Retrospective survey study of QResearch database, 2007–2010	59,804	N/A	Not reported		High-risk population: 50.7% treated with warfarin, 35.9% undertreated, 13.4% untreated	Not reported	Not reported	Not reported
Huisman 2017	Multiple	Study of GLORIA-AF registry, 2011–2014	15,641	Mean = 24	CHF	24.2%	High-risk population: 33.3% treated with warfarin, 48.9% treated with DOAC, 10.0% undertreated, 7.8% untreated	Not reported	Not reported	Not reported
					Hypertension	74.6%
					Diabetes	23.1%
					Vascular disease	20.3%
					Stroke or TIA	10.5%
Inoue 2014	Japan	Study of J-RHYTHM registry, 2009–2011	7516	Mean = 24	CHF	32.2%	Overall population: 86% treated with warfarin, 8% undertreated, 6% untreated	Warfarin and AP use were associated with lower rates of thromboembolism (ORs = 0.36 and 0.92) and higher rates of major hemorrhaging (ORs = 2.35 and 1.34) and all-cause mortality (ORs = 0.50 and 0.94) as compared to untreated patients	Not reported	Not reported
					Hypertension	60.4%
					Diabetes	18.3%
					Vascular disease	10.5%
					Stroke or TIA	13.8%
Jolliffe 2016	New Zealand	Retrospective cohort study of Wellington Regional Hospital, 2012–2012	751	Not reported	CHF	40.3%	High-risk population: 31.6% treated with warfarin, 45.1% undertreated, 23.3% untreated	Not reported	Not reported	Not reported
					Hypertension	68.4%
					Diabetes	22.3%
					Vascular disease	32.3%
					Stroke or TIA	24.9%
LaMori 2013	United States	Retrospective survey study of National Health and Wellness Survey, 2009–2009	1297	Not reported	CHF	19.0%	High-risk population: 48.9% treated with warfarin, 51.1% untreated	Not reported	Not reported	Not reported
					Hypertension	66.0%
					Diabetes	29.0%
					Vascular disease	Not reported
					Stroke or TIA	13.0%
Piccini 2013	United States	Retrospective cohort study of Medicare 5% database, 2006–2007	45,711	Not reported	CHF	52.4%	High-risk population: 57.5% treated with warfarin, 42.5% untreated	Not reported	Not reported	Not reported
					Hypertension	87.9%
					Diabetes	63.5%
					Vascular disease	Not reported
					Stroke or TIA	Not reported
Proietti 2018	Europe	Study of EURObservational Research Programme-Atrial Fibrillation (EORP-AF) Pilot registry, 2012–2014	1990	Mean = 24	Not reported		High-risk population: 69.4% treated with warfarin, 13.2% treated with DOAC, 16.4% untreated	Not reported	-Proportion of patients experiencing readmissions and for what reason are stratified by AF subtype -AF (12.0%), other CV events (8.2%) and non-CV events (9.7%) were the most common reasons for readmission	Not reported
Son 2017	South Korea	Retrospective cohort study of National Health Insurance Services database, 2007–2013	10,654	Mean = 38	CHF	25.5%	High-risk population: 13.6% treated with warfarin, 43.5% undertreated, 43.0% untreated	Not reported	Not reported	Not reported
					Hypertension	69.9%
					Diabetes	36.9%
					Vascular disease	10.1%
					Stroke or TIA	Not reported
Hsu 2016	United States	Study of PINNACLE registry, 2008–2012	294,642	12	CHF	28.6%	High-risk population: 54.2% treated with warfarin, 5.6% treated with DOAC, 40.2% untreated	Not reported	Not reported	Not reported
					Hypertension	84.4%
					Diabetes	26.1%
					Vascular disease (PAD + CAD)	66.2%
					Stroke or TIA	15.8%
Hsu 2016	United States	Study of PINNACLE registry, 2008–2012	429,417	Not reported	CHF	24.9%	High-risk population: 42.3% treated with warfarin, 4.3% treated with DOAC, 31.2% undertreated, 22.2% untreated	Not reported	Not reported	Not reported
					Hypertension	76.8%
					Diabetes	22.9%
					Vascular disease (PAD + MI)	26.4%
					Stroke or TIA	14.3%
Marzec 2017	United States	Study of PINNACLE registry, 2008–2014	655,000	Not reported	CHF	32.4%	High-risk population: 38.1% treated with warfarin, 16.1% treated with DOAC, 6.9% undertreated, 39.0% untreated	Not reported	Not reported	Not reported
					Hypertension	86.6%
					Diabetes	24.8%
					Vascular disease	28.8%
					Stroke or TIA	8.5%

Abbreviations. CHF, congestive heart failure; CI, confidence interval; DOAC, direct oral anticoagulant; MI, myocardial infarction; NVAF, non-valvular atrial fibrillation; OAC, oral anticoagulant; OR, odds ratio; PAD, peripheral arterial disease; TIA, transient ischemic attack; VKA, vitamin K antagonist. Percentages of treatment shares may not add up to 100% due to rounding of values.

Study characteristics

The majority of publications (15) were retrospective studies of claims databases, electronic health records, and preexisting registries^30–44; the remaining study was a prospective evaluation²⁹.Seven studies focused on US populations^30–36, two studies focused on UK populations^37,38, one study evaluated a multi-country population³⁹, and the remaining six studies^29,40–44, covered other individual countries (e.g. Ireland, Japan, South Korea). All studies evaluated rates of treatment stratified by treatment class, including untreated patients. Several studies evaluated clinical outcomes pertaining to treatment, undertreatment, and nontreatment of AF, while one study also evaluated economic outcomes including resource utilization and direct costs of treatment³⁰. Of these studies, 13^{29,31,33–43} stratified study measures by CHA₂DS₂-VASc score while the remaining 3 studies stratified measures by CHADS₂ score^30,32,44. All publications described high risk score as ≥2 on their respective scales.

Rates of undertreatment and nontreatment

Fourteen papers^{29–39,41–43} reported treatment rates for the NVAF subpopulation with elevated stroke risk. In patients with elevated stroke risk, the median rate of undertreatment was 35.3% (range: 12.1%³⁹ to 45.8%³⁶), while the median rate of nontreatment was 23.3% (range: 7.9%³⁹ to 51.1%³²). These studies used data from Ireland, New Zealand, South Korea, the United Kingdom, and the US, as well as combined global and pan-European data. A comparison between US and ex-US rates suggests higher rates of nontreatment in the US setting; the median rate of nontreatment in the US was 40% (range: 21% to 51%) while the median rate in ex-US settings was 20% (range: 11–43%).

When broken down by era, the median rate of undertreatment decreased from 39.7% in the pre-DOAC era (range: 34.6%²⁹ to 45.1%⁴¹) to 25.9% in the DOAC era (range: 12.1%³⁹ to 45.8%³⁶) (Figure 2), while the median rate of nontreatment decreased from 33.6% in the pre-DOAC era (range: 13.4%³⁸ to 51.1%³²) to 17.4% in the DOAC era (range: 7.9%³⁹ to 40.2%³⁴) (Figure 3). These median rates of undertreatment and nontreatment suggest declining trends in undertreated or nontreated NVAF patients with elevated stroke risk once DOACs became available.

Figure 2. Treatment rates by class among high-risk NVAF patients, pre-DOAC era. Abbreviations. DOAC, direct oral anticoagulants; NVAF, non-valvular atrial fibrillation.

Figure 3. Treatment rates by class among high-risk NVAF patients, DOAC era. Abbreviations. DOAC, direct oral anticoagulants; NVAF, non-valvular atrial fibrillation.

Only Jolliffe 2016 reported reasons for nontreatment with OACs among patients with elevated stroke risk. The most frequently documented reason included decisions deferred to the primary care physician (15.6%), followed by the risk of falls or frailty (7.2%) and high risk of bleeding (6.6%); “no reason” was reported in 56.9% of these patients⁴¹.

Clinical consequences of undertreatment and nontreatment

Compared to treated patients with NVAF receiving anticoagulation, untreated patients had a significantly higher risk of adverse clinical outcomes such as strokes, mortality, and various composite outcomes, although patients on anticoagulants did have higher rates of bleeding compared to untreated patients.

Strokes and other ischemia were the most commonly reported clinical consequences of undertreatment or nontreatment of AF. Casciano 2013 found that high-risk patients without warfarin exposure had significantly higher rates of ischemic stroke than patients with low adherence (4.41 vs. 1.87 per 100 person-years, p < .001) or high adherence (4.41 vs. 0.72 per 100 person-years, p < .001) to warfarin³⁰. Similarly, patients without warfarin exposure had higher rates of transient ischemic attack than patients with low adherence (1.77 vs. 0.72 per 100 person-years, p < .001) or high adherence (1.77 vs. 0.46 per 100 person-years, p < .001) to warfarin³⁰. Additionally, in a mostly high-risk population comprised of 81% of patients with elevated risk, de Andrés-Nogales 2015 reported the incidence of ischemic stroke per 100 person-years to be highest for untreated patients (incidence: 3.1; 95% CI: 2.67–3.55) followed by patients treated with antiplatelets (incidence: 0.6; 95% CI: 0.49–0.77) and patients treated with DOACs (0.0; 95% CI: 0.0–10.4))⁴⁰.

Two studies reported mortality associated with nontreatment of AF. Hess 2017 reported that untreated high-risk patients were associated with higher adjusted annual rates of death (OR: 1.22; 95% CI: 1.05–1.41, p = .006) compared to patients on warfarin³¹. In a mixed-risk population comprised of 50% of patients with elevated risk, Inoue 2014 also found that warfarin-treated patients had reduced all-cause mortality compared to untreated patients (OR: 0.50; 95% CI: 0.33–0.75; p < .001); antiplatelet use, however, did not significantly reduce risks of mortality (OR: 0.94; 95% CI: 0.66–1.34; p = .739)⁴⁴.

Three studies reported differences in bleeding risk for untreated patients compared to treated patients, with results indicating higher risks of bleeding among patients treated with antiplatelets or VKAs. de Andrés-Nogales 2015 reported incidence of bleeding events (i.e. hemorrhagic stroke, major bleeding, gastrointestinal hemorrhaging) per 100 person-years to be highest for patients treated with antiplatelets (incidence: 2.4; 95% CI: 2.0–2.8) or VKAs (incidence: 2.4; 95% CI: 2.1–2.7), followed by untreated patients (incidence: 1.7; 95% CI: 1.3–2.3) and patients treated with DOACs (incidence: 0.6; 95% CI: 0.0–3.6)⁴⁰. Inoue 2014 found that warfarin-treated patients had significantly higher rates of major hemorrhaging compared to untreated patients (OR: 2.35; 95% CI 1.12–4.93; p = .024)⁴⁴. Hess 2017 reported that untreated high-risk patients had significantly lower bleeding rates (OR: 0.35; 95% CI: 0.15–0.81, p = .0147) compared to patients on warfarin³¹.

Economic consequences of undertreatment and nontreatment

Only one study compared healthcare resource utilization between high-risk patients treated with warfarin and untreated patients. Compared to untreated patients, Casciano 2013 found that patients highly-adherent to warfarin experienced significantly lower risks of inpatient hospitalizations (OR: 0.73; 95% CI: 0.61–0.86) and outpatient visits (OR: 0.84; 95% CI: 0.73–0.97) and shorter lengths of hospital stay (OR: 0.60; 95% CI: 0.55–0.66); adherence was measured according to the proportion of days covered³⁰. No studies evaluated the resource utilization consequences of nontreatment compared to DOAC use. To the best of our knowledge, currently-published studies have not evaluated the burden of indirect costs (e.g. caregiver burden, loss of productivity) due to undertreatment or nontreatment of NVAF⁴⁵.

Risk of bias assessment

The risk of bias rating for each evaluated study is reported in the supplemental material (Supplemental Appendix, Table 1). Scores ranged from 6 to 8 (out of a total of 9 points), with most studies representing “good quality.”

Discussion

Overview

The results from this analysis indicate that median rates of nontreatment were higher in the US setting compared to ex-US settings (40% vs. 20%). Additionally, global rates of undertreatment and nontreatment in the high-risk population were lower in the DOAC era compared to the pre-DOAC era. More specifically, median rates of undertreatment and nontreatment were 16.4% and 11.1% in the DOAC era compared to 39.7% and 33.6% in the pre-DOAC era, respectively.

This review of observational studies showed that the clinical benefits of anticoagulation led to reductions in stroke and mortality outcomes among treated patients. Additionally, the risk of bleeding events (i.e. hemorrhagic stroke, major bleeding or gastrointestinal hemorrhaging) was found to be higher for patients treated with warfarin compared to untreated patients in real-world practice. Consistent with the reductions observed with stroke events, healthcare resource utilization was found to be lower among patients highly-adherent to warfarin compared to untreated patients.

Gaps and future opportunities

While the literature search identified important findings related to rates of treatment and corresponding clinical and economic consequences of undertreatment or nontreatment, the body of evidence was generally limited. Only three publications evaluated rates of treatment in the DOAC era, and few studies reported the clinical (four) and economic (one) outcomes associated with undertreatment or nontreatment. Therefore, there are notable gaps in the evidence base that warrant future study.

For instance, among the studies evaluating treatment rates in the DOAC era, either ex-US data or older US data was used for analysis. Proietti 2017 evaluated European AF patients in current cardiology practice in Belgium, Denmark, Netherlands, Norway, Poland, Romania, Greece, Italy, and Portugal⁴²; Huisman 2017 globally evaluated patients in Europe, Asia, North America (without further segmentation by country), Latin America, and Africa/Middle East³⁹; and Apenteng 2018 studied UK patients³⁷. Meanwhile, the three US studies used data from the PINNACLE database. Two of these studies used data through 2012, and one used data through 2014, which accounts only for the beginning of the DOAC period. Given the predicted size of the US population with AF by 2030¹, it would be important to evaluate treatment rates using newer US-specific data sources especially with regards to potential increases in DOAC market uptake in more recent years.

Of the few studies evaluating the impact of undertreatment and nontreatment on the clinical burden of NVAF, most analyses focus on warfarin as the sole OAC comparator. Only one study, de Andres Nogales 2015, included DOACs in its evaluation; however, only 1.3% of studied patients were prescribed DOACs which was likely a result of the enrollment period (May 2010-April 2012) coinciding with the very start of the DOAC era⁴⁰. As a result, opportunities exist to compare outcomes in populations treated with DOACs with undertreated or untreated populations. This would provide an opportunity to evaluate real-world aspects of DOAC use that have been demonstrated within the context of clinical trials, such as superior control of bleeding as compared to warfarin, and in the realm of intracranial hemorrhage and hemorrhagic stroke specifically.

Only one study of the economic burden of NVAF undertreatment and nontreatment was identified in this literature search. Casciano 2013 was based on an evaluation of US administrative claims data from 2003 to 2007 and, as the authors note, the data set may have underrepresented the US Medicare population³⁰. Additionally, the analysis reported economic outcomes based on levels of adherence to warfarin and did not consider the effects of DOAC treatment. Many of the same opportunities for evaluating the clinical effects of DOAC treatment would apply to future opportunities for economic studies of OAC. In addition, published studies stratifying the indirect cost burden of NVAF, such as lost productivity and the impact on caregivers, by treatment status (i.e. undertreatment, nontreatment, treatment) were not identified. An evaluation of the indirect cost burden could reveal additional significant burden of illness beyond direct medical costs.

Moreover, the study population evaluated in this review included a general population of NVAF patients. However, subpopulations such as those patients with NVAF who undergo percutaneous coronary intervention (PCI) are treated with OACs plus one or two antiplatelet agents which may impact bleeding risks and outcomes^70,71. Therefore, the burden of undertreatment and nontreatment in these subpopulations may be different than in the general NVAF population described in this analysis; future research evaluating rates of treatment and corresponding outcomes in this subpopulation, as well as other subpopulations of patients with NVAF, is warranted.

Limitations

The scope of this study was restricted to the review of analyses of real-world observational data, which presented a few limitations. First, retrospective cohort studies evaluating real-world data may be subject to certain limitations including selection biases, differential losses to follow-up, as well as the inability to control for residual or unmeasured confounders (i.e. sociodemographic and/or clinical variables unavailable in the data), all of which may restrict the interpretation of study results. Second, rates of treatment varied across studies partially due to differences in regional real-world practice and treatment guidelines captured in each regional data set. For example, the six DOAC-era studies of treatment rates were based on data from three US studies^34–36, two multinational studies^39,42, and one UK study³⁷. Third, studies based on claims data were also limited by the fact that aspirin treatment is not reimbursed by payers (e.g. private US payers); as a result, researchers are unable to distinguish between patients on aspirin and untreated patients when analyzing this data type.

Additionally, despite general trends in results across the four studies reporting clinical outcomes, the individual studies were difficult to compare based on differences in the analytic populations as well as the presentation of results. In the assessments of stroke events, one study presented incidence rates per 100 person-years based on adherence to warfarin³⁰ and a second study presented incidence rates per 100 person-years based on anticoagulant treatment status (i.e. undertreated, untreated, treated with DOACs); even among the treated anticoagulant cohort, adherence levels were not reported⁴⁰. In assessments of mortality, both studies presented adjusted odds of all-cause mortality; however, one study did so among patients with a high risk of stroke while the other paper did so among a mixed-risk population comprised of 50% of patients with a high risk of stroke. Those publications that evaluated mixed populations of patients based on stroke risk were included in our analysis due to the limited literature in this area. A meta-analysis of treatment rates and/or clinical outcomes would have provided an opportunity to statistically and succinctly combine results from the literature. However, it was deemed infeasible to do so, as described above, based on the heterogeneity of the geographic regions and data sources evaluated, population definitions (e.g. high vs. mixed risk of stroke), cohort definitions (e.g. based on warfarin adherence; based undertreatment/nontreatment), and measure definitions (e.g. presentation of incidence rates vs. odds ratios).

Finally, this literature review was guided by the 2014 AHA/ACC/HRS guidelines on treating AF, and readers should be aware of more recent updates to those guidelines when considering the findings. For instance, aspirin is no longer a recommended therapy for any stroke risk level per the 2019 AHA/ACC/HRS guidelines, and patients on aspirin who are described in this study as undertreated would now be considered untreated.

Conclusion

The limited evidence base suggests that the rate of treatment for NVAF has improved over time with the advent of DOACs. This is an important finding to consider given the increasing prevalence of AF as well as the association of undertreatment and nontreatment with adverse clinical and economic outcomes in this population. Ongoing efforts by clinicians, health insurers, and policy makers to develop programs to increase appropriate use of anticoagulation among NVAF patients with elevated stroke risk could lead to reductions in potentially preventable cardiovascular events and death, as well as the associated direct medical cost burden to healthcare systems. Contemporary research is warranted, however, to evaluate current clinical and economic consequences of undertreatment and nontreatment among NVAF patients with elevated stroke risk given the relatively limited evidence base, the evaluation of older data, and the absence of comparisons to patients treated with DOACs.

Transparency

Declaration of funding

This study was sponsored by the Bristol Myers Squibb-Pfizer Alliance.

Declaration of financial/other relationships

M Di Fusco, J Gillespie, I Shirkhorshidian, and M Cato are employees of Pfizer Inc. M Ferri and J Guo are employees of Bristol Myers Squibb. M Sussman, C Tao, and N Adair are employees of Panalgo who were paid consultants to the Bristol Myers Squibb-Pfizer Alliance in connection with the development of this manuscript. GD Barnes received consulting honoraria with regards to the development of the project design and interpretation. GD Barnes received no honoraria for his authorship activities of this manuscript.

Peer reviewers on this manuscript have no relevant financial or other relationships to disclose.

Author contributions

MS, GDB, JDG, CYT, JAG, NA, and MDF were materially involved in the conception, design, analysis, and interpretation of the data. MS, JDG, CYT, NA, and MDF were involved in the drafting of the paper and revising it critically for intellectual content. All authors approved the final version to be published and agree to be accountable for all aspects of the work.