Ischemic stroke and systemic embolism among patients with non-valvular atrial fibrillation who abandon oral anticoagulant therapy

Abstract

Objectives

To compare the risk of stroke and systemic embolism (SE) among patients with nonvalvular atrial fibrillation (NVAF) who abandoned their first direct oral anticoagulant (DOAC) fill (“abandoners”) relative to patients who continued DOACs beyond the first fill (“continuers”).

Methods

In this retrospective longitudinal study, adults with NVAF prescribed DOACs were selected from Symphony Health, an ICON plc Company, PatientSource, 1 April 2017 to 31 October 2020. A 90-day landmark period following the first DOAC fill was used to classify patients as abandoners or continuers. Inverse probability of treatment weighting was used to balance baseline characteristics between cohorts. Time to ischemic stroke/SE was described and compared between cohorts using weighted Kaplan-Meier and Cox proportional hazard models from the end of the landmark period until end of clinical activity or data.

Results

After weighting, 200,398 and 211,352 patients comprised the abandoner and continuer cohorts, respectively. The mean duration of follow-up was 14.9 and 15.7 months, respectively. At 12 months of follow-up, the probability of ischemic stroke/SE was 1.34% in the abandoner cohort and 1.00% in the continuer cohort; the risk of ischemic stroke/SE was 35% higher in the abandoner versus continuer cohort (hazard ratio [95% confidence interval] = 1.35 [1.20, 1.51]; p < 0.0001).

Conclusions

Patients with NVAF who abandoned the first DOAC fill had significantly higher risk of ischemic stroke/SE compared to patients who continued therapy beyond the first fill. There is an unmet need for better access to DOACs so that the long-term risk of poor outcomes may be minimized.

Keywords:

• Direct oral anticoagulants
• ischemic stroke
• nonvalvular atrial fibrillation
• systemic embolism
• treatment abandonment

Introduction

Atrial fibrillation (AF) is associated with a 5-fold higher risk of stroke and an 8-fold higher risk of experiencing multiple cardiovascular hospitalizations^1,2. In turn, stroke accounts for more than 900,000 inpatient stays per year, in addition to 1 in every 19 deaths in the United States (US)³, highlighting the large clinical burden associated with AF and its complications.

To reduce the risk of stroke and systemic embolism (SE), guidelines recommend long-term treatment with oral anticoagulants (OACs)⁴, which have been shown to significantly lower the odds of moderate or severe stroke and in-hospital mortality compared to no treatment in patients with AF⁵. While the initiation and continuous use of anticoagulant therapy is key to adequate disease control among patients with AF, suboptimal compliance to direct oral anticoagulants (DOACs) is common (1 in 3 patients with AF adhere to DOACs <80% of the time) and is associated with poor clinical outcomes, including increased risk of stroke, transient ischemic attack, and mortality⁶. For instance, one retrospective claims study estimated that more than one-quarter of patients had suboptimal adherence to DOACs, with each 10% reduction in adherence conferring a 7% increase in the risk of stroke⁷.

Anticoagulant nonadherence and its clinical effects have previously been characterized in AF populations^6,8,9, but prior literature has largely focused on secondary nonadherence (i.e. patient fills first prescription[s] and then discontinues treatment) rather than primary nonadherence or abandonment (i.e. patient does not fill, or abandons a first prescription for a new medication and does not obtain an alternative medication within a reasonable time)¹⁰. In one US claims-based study, the rate of abandonment was 26.7% among DOAC users, but the study included any patients receiving OACs, regardless of diagnosis¹¹. Additionally, the analysis did not evaluate the clinical repurcussions of DOAC abandonment.

Therefore, the current study was conducted to compare the risk of stroke and SE among patients with nonvalvular AF (NVAF) who abandoned their first DOAC fill relative to patients who continued DOAC therapy beyond the first fill.

Methods

Data source

This study used data from Symphony Health, an ICON plc Company, PatientSource, 1 April 2017 to 31 October 2020, which links health care data for the US population from the following three basic sources: pharmacy point-of-service, switch/network transactions, and additional direct prescription, medical, and hospital claims data. The database includes patients’ demographics, medical and procedure claims, and prescription drug claims, including status of prescription drug claims (approved, rejected, abandoned). The database represents over three-quarters of the US population of patients annually across multiple payer channels (i.e. commercial, Medicare, Medicaid), capturing >75% of all US retail prescription claims.

Data were de-identified and comply with the patient requirements of the Health Insurance Portability and Accountability Act (HIPAA). Studies with de-identified retrospective claims data do not require a waiver of an ethics committee, as per the Health and Human Services Office for Human Research Protections Guidance on Research Involving Coded Private Information or Biological Specimens (Title 45 CFR 46.102(f))¹², they do not meet the definition of human subject research. Throughout the study, researchers complied strictly with all applicable HIPAA data management rules and the 1964 Helsinki Declaration and its later amendments or comparable ethical standards.

Study design

This study used a retrospective longitudinal intent-to-treat design. The index date was the date of the first claim for a DOAC indicated for NVAF (i.e. apixaban, dabigatran, or rivaroxaban). The study focused on these 3 DOACs because they are the most used in the US. Only patients with the final status of the first claim being “approved” or “abandoned” were retained. Approved claims were those submitted by a pharmacy and approved for payment by health plans after claims adjudication, while abandoned claims were adjudicated claims for prescriptions that were approved by health plans but abandoned by patients, usually due to cost or non-compliance. The baseline period comprised the 6-month period of clinical activity before the index date (also the washout period for oral anticoagulants). Clinical activity was identified based on the first and last patient-level activity flags (i.e. either a pharmacy or a medical claim). The 90-day landmark period starting on the index date was used to classify patients into the “abandoner” and “continuer” cohorts. The follow-up period started on day 91 post-index and spanned until the earliest of the end of clinical activity or data availability.

Sample selection

Patients meeting the following criteria were included in the study: (1) the final approved or abandoned status of the first claim (index date) for apixaban, dabigatran, rivaroxaban; (2) ≥6 months of clinical activity before the index date; (3) no claims for oral anticoagulants (i.e. apixaban, dabigatran, rivaroxaban, betrixaban, edoxaban, warfarin) before the index date; (4) ≥1 claim with a diagnosis for AF (International Classification of Diseases, 10th Revision, Clinical Modification [ICD-10-CM]: I48) during the baseline period or on the index date; and (5) ≥18 years old on the index date.

Patients were excluded if they had any of the following exclusion criteria: (1) ≥1 claim at baseline for mitral-stenosis, mechanical heart-valve, venous thromboembolism (VTE), hip or knee replacement surgery, or organ or tissue replaced by transplant; (2) pregnancy during or after the baseline period; (3) ≥1 claim for stroke (ischemic or hemorrhagic) or SE during an inpatient admission on the index date or within 30 days prior to the index date; (4) >1 DOAC (i.e. apixaban, betrixaban, dabigatran, edoxaban, rivaroxaban) claimed on the index date; or (5) >1 final claim status (e.g. approved and abandoned) for the index DOAC on the index date. The rationale for exclusion of patients with clinical events on the index date or shortly before it is as follows: (1) it is impossible to identify if the event on the index date happened before or after the DOAC claim, and (2) the event on the index date or shortly before might impact the decision to obtain treatment with DOACs.

Additionally, the following exclusion criteria for the identification of cohorts during the landmark period were applied: (1) ≤90 days of clinical activity post-index (i.e. incomplete landmark period); (2) ≥1 stroke or SE diagnosis in an inpatient setting; (3) ≥1 approved DOAC claim (abandoner cohort only); (4) >30 days of supply on the index DOAC claim (continuer cohort only).

Patients were classified into the abandoner cohort if they abandoned their first DOAC claim and did not have an approved claim for a DOAC within the next 90 days (i.e. primary nonadherence). Patients were classified into the continuer cohort if they had an approved first DOAC claim and continued DOAC therapy beyond the first claim (i.e. had ≥2 approved claims for a DOAC during the landmark period).

Outcome measures

Outcomes measured in each cohort during the follow-up period included the first instance of ischemic stroke (ICD-10-CM: I63) or SE (ICD-10-CM: I74) in an inpatient setting, reported as a composite outcome (i.e. stroke/SE) and separately. The diagnosis codes used to identify stroke have been validated in a previous study¹³ and other claims-based studies have used these ICD-10-CM diagnosis codes to identify stroke or systemic embolism^14–16.

Statistical analysis

Inverse probability of treatment weighting (IPTW) was used to balance baseline characteristics between the abandoner and continuer cohorts. The propensity score was computed from a logistic regression model adjusting for age; sex; region of residence; insurance plan type; index year (to account for temporal factors); Quan Charlson Comorbidity Index (Quan-CCI) ¹⁷; CHA₂DS₂-VASc score; HAS-BLED score; patient out-of-pocket paid amount for the first DOAC prescription; polypharmacy (concurrent use of ≥5 different medications); use of antihypertensive, antihyperlipidemic, and antiplatelet agents; the index DOAC medication; and all-cause pharmacy costs. The balance of baseline characteristics was evaluated using the standardized difference, where <10% indicated balance¹⁸.

Probability of an ischemic stroke or SE event was described in each cohort during the follow-up period using weighted Kaplan Meier survival analysis and was compared between cohorts using log-rank tests. Weighted Cox proportional hazards models were used to compare the risk of ischemic stroke or SE between the cohorts, reported as hazard ratios (HRs) with their 95% confidence intervals (CIs) and p-values. Weighted Cox proportional hazards models were further adjusted for the patients’ out-of-pocket paid amounts for the first DOAC prescription that remained slightly imbalanced between cohorts after IPTW. The time to the first event was defined as the time from day 91 post-index (i.e. first day after the end of the landmark period) to the first ischemic stroke or SE event during the follow-up period. Patients without an observed event during the follow-up period were censored at the end of the follow-up period.

Sensitivity analysis

A sensitivity analysis was conducted by shortening the landmark period from 90 to 75 days to minimize the loss of patients with stroke and SE events occurring during the landmark period. This sensitivity also allowed to assess how inclusion of abandoners, who initiated DOACs between days 75 and 90 post-index, impacted the results.

Results

Among 779,468 identified patients with NVAF who were prescribed a study DOAC, 86,338 (11.1%) had the final status of their fist DOAC claim as “abandoned” and 667,417 (85.6%) as “approved”. A total of 45,038 (52.2%) of patients with the final status of their fist DOAC claim as “abandoned” met additional criteria during the landmark period to be classified into the abandoner cohort. A total of 366,712 (54.9%) of patients with the final status of their fist DOAC claim as “approved” met additional criteria during the landmark period to be classified in the continuer cohort (Figure 1). Overall, 5.8% (45,038/779,468) abandoned their first DOAC prescription fill in this retrospective, real-world study. This estimate should be interpreted bearing in mind that half of patients with the final status of their first DOAC claim as “abandoned” did not have enough information to confirm the prescription abandonment, thus the true proportion of abandoners could be between 5.8% and 11.1%.

Figure 1. Identification of DOAC abandoners and continuers among patients with NVAF. AF, atrial fibrillation; DOAC, direct oral anticoagulant; NVAF, non-valvular atrial fibrillation; SE, systemic embolism. Notes: ^aPatients were excluded from the study population if they had (1) ≥1 claim with any of the following diagnoses or procedures at baseline: mitral-stenosis, mechanical heart-valve, venous thromboembolism, hip or knee replacement surgery, organ or tissue replaced by transplant; (2) pregnancy during or after the baseline period; (3) ≥1 claim for stroke (ischemic or hemorrhagic) or systemic embolism during an inpatient admission on the index date or within 30 days prior to the index date; (4) >1 DOAC (i.e. apixaban, betrixaban, dabigatran, edoxaban, rivaroxaban) claimed on the index date; or (5) >1 final claim status (e.g., approved and abandoned) for the index DOAC on the index date. ^bPatients were excluded during the landmark period if they had (1) ≤90 days of clinical activity post-index (i.e. incomplete landmark period); (2) ≥1 stroke or SE diagnosis in an inpatient setting within the first 90 days post-index (i.e. during the landmark period); or (3) ≥1 approved DOAC claim within the first 90 days post-index (i.e. during the landmark period). ^cPatients were excluded during the landmark period if they had (1) ≤90 days of clinical activity post-index (i.e. incomplete landmark period); (2) ≥1 stroke or SE diagnosis in an inpatient setting within the first 90 days post-index (i.e. during the landmark period); or (3) >30 days of supply on the index DOAC claim.

Patient population and weighted baseline characteristics

After weighting the cohorts to balance baseline characteristics, 200,398 patients comprised the abandoner cohort and 211,352 patients comprised the continuer cohort. Most characteristics between cohorts were well balanced. Mean [standard deviation (SD)] age was 69.2 [9.8] years in the abandoner cohort and 69.4 [9.3] years in the continuer cohort, and 46.3 and 45.8% were female, respectively (Table 1). Most patients were covered by Medicare (57.9% in the abandoner cohort and 57.0% in the continuer cohort), followed by commercial insurance (34.3 and 34.7%, respectively), and Medicaid (6.1 and 7.0%), or another type of insurance (1.7 and 1.3%). Both cohorts had the same mean [SD] CHA₂DS₂-VASc score of 3.1 [1.6] and similar prevalence of risk factors for stroke, with the most common being hypertension (67.5% in the abandoner and 67.8% in the continuer cohort), hyperlipidemia (45.1 and 45.9%), diabetes (27.4 and 26.4%), and congestive heart failure (26.6 and 26.4%).

Table 1. Baseline characteristics in the weighted abandoner and continuer cohorts^a,b.

Mean ± SD [median] or n (%)	Abandoner cohort	Continuer cohort	Standardized difference (%)
	N = 200,398	N = 211,352
Age (years)	69.2 ± 9.8 [72.0]	69.4 ± 9.3 [72.0]	2.0
Female	92,867 (46.3)	96,782 (45.8)	1.1
Region of residence
South	82,475 (41.2)	87,114 (41.2)	0.1
Midwest	49,027 (24.5)	50,103 (23.7)	1.8
Northeast	38,954 (19.4)	42,830 (20.3)	2.1
West	29,498 (14.7)	30,816 (14.6)	0.4
Unknown	444 (0.2)	488 (0.2)	0.2
Insurance plan
Medicare	116,035 (57.9)	120,383 (57.0)	1.9
Commercial	68,683 (34.3)	73,409 (34.7)	1.0
Medicaid	12,237 (6.1)	14,878 (7.0)	3.8
Other	3,442 (1.7)	2,682 (1.3)	3.7
Year of index date
2017	13,736 (6.9)	14,477 (6.8)	0.0
2018	73,473 (36.7)	78,093 (36.9)	0.6
2019	75,791 (37.8)	79,973 (37.8)	0.0
2020	37,398 (18.7)	38,809 (18.4)	0.8
Patient out-of-pocket paid amount for the index drug claim (US $2021)	93.8 ± 160.9 [22.6]	77.2 ± 152.3 [20.2]	10.6
Quan-CCI	1.8 ± 2.1 [1.0]	1.8 ± 2.1 [1.0]	0.1
CHA2DS2-VASc score	3.1 ± 1.6 [3.0]	3.1 ± 1.6 [3.0]	1.1
HAS-BLED score	1.9 ± 1.1 [2.0]	1.9 ± 1.1 [2.0]	1.4
Baseline diagnosis for stroke and SE in an inpatient setting
Ischemic stroke	1,734 (0.9)	2,244 (1.1)	2.0
Hemorrhagic stroke	210 (0.1)	287 (0.1)	0.9
SE	84 (0.0)	118 (0.1)	0.6
Major bleeding	2,963 (1.5)	2,254 (1.1)	3.7
Number of unique prescription drugs used	7.2 ± 5.9 [6.0]	7.6 ± 5.8 [7.0]	6.2
Polypharmacy (use of ≥5 different medications concurrently)	107,848 (53.8)	118,525 (56.1)	4.5
Use of non-oral anticoagulants	13,605 (6.8)	15,976 (7.6)	3.0
Use of cardiovascular-related medications	139,128 (69.4)	154,296 (73.0)	7.9
Antihypertensive agents	124,257 (62.0)	135,176 (64.0)	4.0
Antihyperlipidemic agents	85,347 (42.6)	94,663 (44.8)	4.4
Antiplatelet agents	18,337 (9.2)	20,195 (9.6)	1.4
Cardiovascular procedures	5,044 (2.5)	6,554 (3.1)	3.5
Cancer diagnoses and/or treatments	23,364 (11.7)	26,795 (12.7)	3.1
Index DOAC medication
Apixaban	150,209 (75.0)	158,867 (75.2)	0.5
Rivaroxaban	45,536 (22.7)	47,854 (22.6)	0.2
Dabigatran	4,653 (2.3)	4,632 (2.2)	0.9
Risk factors for stroke
Hypertension	135,192 (67.5)	143,395 (67.8)	0.8
Hyperlipidemia	90,369 (45.1)	96,940 (45.9)	1.5
Diabetes	54,974 (27.4)	55,703 (26.4)	2.4
Congestive heart failure	53,380 (26.6)	55,899 (26.4)	0.4
COPD	34,542 (17.2)	32,153 (15.2)	5.5
Obesity	29,469 (14.7)	32,617 (15.4)	2.0
Morbid obesity	14,214 (48.2)	15,248 (46.7)	3.0
Depression	21,075 (10.5)	19,015 (9.0)	5.1
Prior stroke	7,274 (3.6)	9,508 (4.5)	4.4
Family history of CVD	7,565 (3.8)	8,232 (3.9)	0.6
Transient ischemic attack	3,720 (1.9)	5,029 (2.4)	3.6
All-cause monthly pharmacy costs^c (US $2021)	215 ± 897 [29]	228 ± 748 [47]	1.6

COPD, chronic obstructive pulmonary disease; CVD, cardiovascular disease; DOAC, direct oral anticoagulant; IPTW, inverse probability of treatment weighting; Quan-CCI, Quan Charlson Comorbidity Index; SD, standard deviation; SE, systemic embolism; US, United States.

Notes: ^aCohorts were weighted on baseline characteristics using IPTW; characteristics were considered well balanced if standardized difference was <10%. ^bThe number of patients reported in this weighted population represents the sum of weights for the corresponding patients, rounded to the nearest integer. The sum of patients in each category may therefore not add up to the total number of patients in each cohort. ^cPharmacy costs were reported from a payer’s perspective and comprised the plan paid amounts and patient copay amounts.

Mean patient out-of-pocket paid amount for the first DOAC prescription remained slightly unbalanced between cohorts: $93.80 in the abandoner and $77.20 in the continuer cohort (standardized difference = 10.6%); this variable was controlled for in Cox proportional hazard models.

Ischemic stroke and SE

The mean [SD] duration of follow-up was 14.9 [9.5] months in the abandoner cohort and 15.7 [9.5] months in the continuer cohort. A total of 3,203 stroke/SE events were observed in the weighted abandoner cohort (incidence rate of 1.29 per 100 patient-years) and 2,538 stroke/SE events were observed in the weighted continuer cohort (incidence rate of 0.92 per 100 patient-years). The probability of ischemic stroke or SE as a composite outcome at 3 months of follow-up was 0.39% in the abandoner cohort compared to 0.28% in the continuer cohort (log-rank p-value = 0.0027; Figure 2). It remained higher in the abandoner cohort compared to the continuer cohort at later points of follow-up (e.g. at 12 months, 1.34 and 1.00%, respectively [log-rank p-value <0.0001]). Probability of the composite outcome was mainly driven by ischemic stroke. Indeed, the probability of ischemic stroke alone at 3 months of follow-up was 0.37% in the abandoner cohort compared to 0.26% in the continuer cohort and at 12 months 1.29% compared to 0.92%, respectively (all log-rank p-values <0.01; data not shown).

Figure 2. Ischemic stroke or SE probability after first DOAC fill in weighted abandoner versus continuer cohorts^a,b. *Log-rank p-value <0.05; CI, confidence interval; DOAC, direct oral anticoagulant; IPTW, inverse probability of treatment weighting; Quan-CCI, Quan Charlson Comorbidity Index; SE, systemic embolism. Note: ^aIschemic stroke and SE events were identified in an inpatient setting. The time to the first event was measured from day 91 post-index date (i.e. first day after the end of the landmark period); patients without an observed event during the follow-up period were censored at the end of the follow-up period. ^bCompeting risk of death, e.g. from cancer, could lead to an overestimation of the incidence of stroke/SE in both cohorts. Information on death was not available in the data to account for this risk. Since cohorts were balanced on Quan-CCI and other comorbidities using IPTW, risk of death from competing events should be similar in both cohorts.

Based on weighted adjusted Cox proportional hazards models, the risk of ischemic stroke or SE as a composite outcome was 38% higher (HR = 1.38) in the abandoner cohort than in the continuer cohort at 3 months, 31% higher at 6 months, 35% higher at 12 months, and 40% higher at 24 months of follow-up (all p-values <0.01; HRs and 95% CIs reported in Figure 3). The risk of ischemic stroke alone was 33–42% higher in the abandoner cohort than in the continuer cohort at 3, 6, 12, and 24 months of follow-up (all p-values < 0.001), but the risk of SE alone was not significantly different between the cohorts.

Figure 3. Risk of ischemic stroke and SE in weighted abandoner versus continuer cohorts^a. *p-Value <0.05; HR, hazard ratio; CI, confidence interval; SE, systemic embolism. Note: ^aIschemic stroke and SE were identified in an inpatient setting. HRs were generated using univariate weighted Cox proportional hazard models.

Sensitivity analysis using a 75-day landmark period

In the sensitivity analysis using a 75-day landmark period, 201,678 patients were included in the weighted abandoner cohort and 212,367 patients were included in the weighted continuer cohort. The probabilities of ischemic stroke or SE in the two cohorts were comparable to the findings of the main analysis (Supplementary Figure 1). The risk of ischemic stroke or SE was 36–45% higher in the abandoner cohort compared to the continuer cohort at 3, 6, 12, and 24 months of follow-up (all p-values <0.0001; HRs and 95% CIs reported in Supplementary Figure 2).

Discussion

Among patients with NVAF prescribed DOAC therapy, at least 5% could be abandoning their first DOAC prescription fill, based on the results of this retrospective, real-world study. These patients compared to those who continued DOAC therapy beyond the first claim had a significantly higher risk of ischemic stroke or SE both in the short and long term, predominantly due to the higher risk of ischemic stroke. These results remained robust in the sensitivity analysis, which minimized the impact of the loss of stroke and SE events in the first three months after the first DOAC prescription date on study findings.

The rates of primary abandonment of DOAC therapy have not been well-characterized in US populations. One retrospective, claims-based study by Gupta et al. reported a primary abandonment rate of 26.7% for DOACs in a conference abstract, but this is difficult to compare with the current analysis given the limited details provided regarding the methods of the unpublished study¹¹.

Several factors could be associated with primary abandonment or poor DOAC compliance in general. Previous research suggests that higher copayment for DOACs is a factor associated with lower adherence and higher discontinuation rates among patients with AF in the US¹⁴. Restricted access to DOACs (i.e. prior authorization or step therapy requirements) could also lower the probability of DOAC use and adherence, as well as increase delays in filling the initial prescription¹⁹. Although factors associated with primary non-adherence were not explored in this study, before adjustment for differences in baseline characteristics, patients who abandoned their first prescription had over $100 higher out-of-pocket paid amounts for the first DOAC prescription on average compared to patients who continued DOACs beyond the first claim. This difference could be due to varying coverage provided by different insurance plans. Overall, policies and programs to facilitate access to DOACs may help to improve patient compliance; additional research into such options is warranted.

In the current study, DOAC abandonment was associated with a significantly higher risk of ischemic stroke or SE than continued DOAC therapy. This finding is consistent with the literature on outcomes of secondary DOAC non-adherence and non-persistence in general, though different definitions have been used in different studies^6,9,19. For instance, secondary DOAC non-adherence was associated with an increased risk of stroke while DOAC non-persistence was associated with an increased risk of stroke/transient ischemic attack in a meta-analysis of real-world AF studies⁶. Further, one claims-based study found that the association between DOAC non-persistence and risk of stroke was most evident among higher-risk patients (i.e. CHA₂DS₂-VASc score ≥2), potentially because they had a more easily modifiable stroke risk⁹. Finally, patients with AF in plans with more restricted access to DOACs, who were shown to have lower adherence, also displayed a higher risk of stroke and mortality¹⁹. Combined with previous evidence, results of the current study suggest that along with secondary non-adherence and non-persistence, primary DOAC non-adherence could have detrimental clinical consequences for the downstream risk of stroke.

Although not measured in this study, poor compliance with DOACs and the resulting potential for stroke can also have economic consequences. In a previous analysis, adults with AF and secondary non-adherence to DOACs had lower pharmacy costs compared to adherent patients, but this was offset by the higher inpatient and outpatient costs, resulting in significantly higher total healthcare costs for nonadherent patients²⁰. This argument further supports the need for better compliance with DOACs so that the long-term risk of stroke and SE may be minimized. Alternatively, other management options to reduce stroke risk may be considered, including surgical left atrial appendage closure²¹.

Limitations

The study is subject to some limitations. The database does not capture services received from providers outside of the network, which may have led to an underestimation of the use of pharmacological treatments and medical services. For instance, some patients may have appeared as less compliant with DOAC therapy if they switched claims transaction networks for their prescriptions. The use of free drug samples was not captured in the data; however, their use among abandoners would have biased the results towards the null and led to more conservative risk estimates. Similarly, prescription fills did not account for whether the medication dispensed was taken as prescribed, which may have led to an overestimation of treatment continuation, further biasing the results towards the null. The requirement of the landmark period (i.e. ≥90 days post-index) introduced a selection bias, since patients with an abandoned DOAC claim on the index date may have been less likely to survive beyond 90 days than patients who continued DOAC therapy. Further, patients with early events (i.e. within the first 90 days) were excluded. However, the findings of the sensitivity analysis with a shorter landmark period (i.e. 75 days) were comparable to those of the main analysis. Additionally, we were unable to determine the reason for non-adherence due to the retrospective nature of the registry and data. A competing risk of death from comorbidities, e.g. cancer, could lead to an overestimation of the incidence of stroke/SE in both cohorts. Information on death was not available in the data to account for this risk. Nonetheless, given that cohorts were balanced on Quan-CCI and other comorbidities using IPTW, the risk of death from competing events should be similar in both cohorts. Lastly, there may have been residual confounding due to unmeasured factors, such as the use of over-the-counter medications (e.g. aspirin), which are not captured in claims data, as well as clinical characteristics of patient severity.

Conclusions

In this real-world study, patients who abandoned their first DOAC prescription fill had higher probabilities of ischemic stroke or SE than those who continued DOAC therapy beyond the first fill. Given the large and detrimental clinical burden associated with stroke, further research is warranted to identify factors associated with primary abandonment of DOACs so that management of NVAF may be optimized in clinical practice.

Transparency

Author contributions

Mark J. Alberts: conceptualization, methodology, resources, writing – original draft preparation, writing – review & editing; Maryia Zhdanava: conceptualization, data curation, formal analysis, investigation, methodology, project administration, resources, supervision, validation, visualization, writing – original draft preparation, writing – review & editing; Dominic Pilon: conceptualization, data curation, formal analysis, investigation, methodology, project administration, resources, supervision, validation, visualization, writing – original draft preparation, writing – review & editing; Gabrielle Caron-Lapointe: conceptualization, data curation, formal analysis, investigation, methodology, project administration, resources, supervision, validation, visualization, writing – original draft preparation, writing – review & editing; Patrick Lefebvre: conceptualization, data curation, formal analysis, investigation, methodology, project administration, resources, supervision, validation, visualization, writing – original draft preparation, writing – review & editing; Brahim Bookhart: conceptualization, methodology, resources, writing – original draft preparation, writing – review & editing; Akshay Kharat: conceptualization, funding acquisition, methodology, project administration, resources, writing – original draft preparation, writing – review & editing.

Acknowledgements

Isabelle Ghelerter contributed to the study design and the interpretation of the results for this study and was an employee of Analysis Group, Inc. at the time this study was conducted. Medical writing assistance was provided by Christine Tam, MSc, MWC, an employee of Analysis Group, Inc.

Ethics statement

Data were de-identified and comply with the patient requirements of the Health Insurance Portability and Accountability Act (HIPAA) of 1996. Studies with de-identified retrospective claims data do not require a waiver of an ethics committee, as per the Health and Human Services Office for Human Research Protections Guidance on Research Involving Coded Private Information or Biological Specimens (Title 45 CFR 46.102(f))¹², they do not meet the definition of human subject research. Throughout the study, researchers complied strictly with all applicable HIPAA data management rules and the 1964 Helsinki Declaration and its later amendments or comparable ethical standards.

Previous presentations

None.

Declaration of funding

This study was funded by Janssen Scientific Affairs, LLC., the producer of rivaroxaban.

Declaration of financial/other relationships

Akshay Kharat and Brahim Bookhart are employees of Janssen Scientific Affairs, LLC, and are stockholders of Johnson & Johnson. Maryia Zhdanava, Dominic Pilon, Gabrielle Caron-Lapointe, and Patrick Lefebvre are employees of Analysis Group, Inc., a consulting company that has provided paid consulting services to Janssen Scientific Affairs, LLC. Mark J. Alberts has not received compensation for this project.

A reviewer on this manuscript has disclosed that they have been or are a consultant/advisor for Daiichi-Sankyo; Bayer AG and Servier. Peer reviewers on this manuscript have received an honorarium from CMRO for their review work but have no other relevant financial relationships to disclose.

Data availability statement

The datasets generated and analyzed during the current study are not publicly available because they were used pursuant to a data use agreement. The data are available through requests made directly to Symphony Health, an ICON plc Company, via https://www.iconplc.com/contact/.