Effectiveness and Safety of Direct Oral Anticoagulants versus Aspirin in Patients with Non-Valvular Atrial Fibrillation and Intermediate Stroke Risk

Abstract

Aims

This study compared the effectiveness and safety of direct oral anticoagulants (DOACs) versus aspirin for stroke prevention in patients with non-valvular atrial fibrillation (NVAF) and intermediate stroke risk, defined by a CHA₂DS₂-VASc score of 1 in men and 2 in women.

Method

This retrospective cohort study used Korean Health Insurance Review and Assessment Service claims data. Patients diagnosed with NVAF between January 1, 2017, and December 31, 2019, who initiated DOACs or aspirin, were included. Eligibility was restricted by CHA₂DS₂-VASc score, and patients were matched by age and sex to reduce confounding. The observation period extended until November 30, 2021. The primary outcomes were the incidence of ischemic stroke (effectiveness) and major bleeding events (safety).

Results

Of the 2234 patients included, 977 (43.7%) were treated with DOACs, and 1257 (56.3%) received aspirin. After matching, no statistically significant differences were observed between the DOAC and aspirin groups in ischemic stroke incidence (1.46 vs 0.92 per 100 person-years, P = .060) or major bleeding events (2.26 vs 2.10 per 100 person-years, P = .100). Notably, rheumatic disease was associated with an increased risk of ischemic stroke, while liver disease was linked to a higher risk of major bleeding.

Conclusion

In NVAF patients with intermediate stroke risk, no statistically significant differences were observed in ischemic stroke or major bleeding between DOACs and aspirin. These results should not be interpreted as therapeutic equivalence but highlight the need for further real-world research, particularly in Asian populations, to inform optimal antithrombotic strategies in this subgroup.

Introduction

Atrial fibrillation (AF) increases the risk of stroke, which can be significantly reduced by anticoagulation therapy.^1–4 Therefore, current treatment guidelines recommend anticoagulation therapy for AF patients with stroke risk factors.^5–10 These guidelines strongly advise anticoagulation therapy in patients with a CHA₂DS₂-VASc (congestive heart failure, hypertension, age ≥75 years [doubled], diabetes mellitus, prior stroke/transient ischemic attack [TIA]/thromboembolism [doubled], vascular disease, age 65–74 years, and female sex) score of ≥2 in men and ≥3 in women.^5,6,8–11 They prefer the use of direct oral anticoagulants (DOACs) over warfarin for anticoagulation, except in patients with moderate-to-severe rheumatic mitral stenosis or a mechanical heart valve.^5–10 For patients with a CHA₂DS₂-VASc score of 1 in men and 2 in women, the strength of recommendation has been comparatively weaker than for higher-risk patients and has been changed over time. In the 2014 AHA/ACC/HRS guideline, either anticoagulation or aspirin was recommended. ¹² Subsequently, the 2018–2020 guidelines stated that anticoagulation “can be considered”,^6,8–10 whereas the 2023–2024 guidelines further strengthened this recommendation to “reasonable” or “should be considered”,^5,11 reflecting a gradual shift toward favoring anticoagulation therapy.

A systematic review comparing anticoagulants and aspirin in AF patients without prior stroke or TIA found that anticoagulants significantly reduced the risk of stroke compared to aspirin, although they increased the risk of intracranial hemorrhage. ¹³ However, these results pertained to studies that used warfarin, not DOACs.^14–20 Regarding direct comparisons between DOACs and aspirin for stroke prevention, a study comparing apixaban with aspirin in AF patients with stroke risk factors showed that apixaban reduced the risk of stroke or systemic embolism without an increased risk of major bleeding compared to aspirin. ²¹ However, this study primarily included non-Asian populations, with less than 20% Asians participants, and the average CHADS₂ (congestive heart failure, hypertension, age ≥75 years, diabetes mellitus, prior stroke/TIA/thromboembolism [doubled]) score was approximately 2, leaving a lack of research focusing on Asian populations with intermediate risk.

Although the overall use of guideline-recommended anticoagulation therapy has gradually increased over time, ²² a substantial number of AF patients continue to use antiplatelet agents instead of anticoagulants.^23,24 Previous research using claims data in South Korea found that approximately one-third of AF patients eligible for anticoagulation therapy are not prescribed anticoagulants, and the use of aspirin was identified as a significant risk factor influencing this underutilization. ²³ Previous studies have suggested that vascular disease is associated with the use of aspirin in patients with atrial fibrillation, particularly in those with lower CHA₂DS₂-VASc scores.^23,25 Despite guideline updates discouraging aspirin monotherapy in AF, its use persists in intermediate-risk patients, possibly due to the lack of direct comparative evidence with DOACs in this population.

Recent international guidelines highlight this clinical uncertainty. The 2024 European Society of Cardiology (ESC) guideline states that DOACs should be considered in patients with CHA₂DS₂-VASc score of 1 in men and 2 in women, but acknowledges limited evidence in this group. ¹¹ Meanwhile, the 2023 ACC/AHA/ACCP/HRS guideline also notes that decisions regarding anticoagulation in this population should be individualized, given the lack of robust comparative outcome data. ⁵ These recommendations underscore the need for real-world evidence comparing DOACs and aspirin in intermediate-risk patients, especially in Asian populations where stroke and bleeding risk profiles differ.

This study aims to address this evidence gap in Asian NVAF patients. We compare the effectiveness and safety of DOACs and aspirin as antithrombotic agents for stroke prevention in AF patients with a CHA₂DS₂-VASc score of 1 in men and 2 in women. By evaluating stroke prevention effects and bleeding risks between these treatments, this study aims to contribute additional real-world evidence for this specific population, where direct comparative research remains limited.

Methods

Study Design and Database

This study was a retrospective observational cohort study using customized research data from the Korean Health Insurance Review and Assessment Service (HIRA) claims database (No. M20220704002). The data encompassed the entire medical history between January 1, 2015, and November 30, 2021, of patients with an initial diagnosis of AF as the primary diagnosis between January 1, 2017, and December 31, 2019. Patients were followed from the index date, defined as the first AF diagnosis, until the occurrence of an outcome event or the last recorded medical visit before November 30, 2021. Medical records for at least two years prior to the index date were reviewed to confirm the absence of prior AF diagnoses.

The extracted data included basic information such as age, gender, and insurance type, as well as disease codes, prescribed medication codes, and treatment durations. In Korea, the health insurance system is a social security system operated by a single-payer system, and all citizens are required to subscribe. ²⁶ It consists of three insurance types: National Health Insurance (NHI), Medical Aid (MedAid), and Patriots & Veterans Insurance (PVI). Since all medical claims are submitted to HIRA, the HIRA data covers about 98% of the total population. ²⁶

This study was approved by the Institutional Review Board (IRB) of Pusan National University (PNU IRB/2022_111_HR). The requirement for informed consent was waived by the IRB because this study used anonymized secondary data without personally identifiable information.

Study Population

The study included adults aged 20 years or older who were newly diagnosed with non-valvular atrial fibrillation (NVAF) between 2017 and 2019, initiated treatment with either DOACs (dabigatran, rivaroxaban, apixaban, or edoxaban) or aspirin, and had an intermediate risk of stroke defined as a CHA₂DS₂-VASc score of 1 in men and 2 in women. NVAF was operationally defined as AF without diagnostic codes for moderate-to-severe rheumatic mitral stenosis or mechanical heart valves (Supplementary Table 1). ⁵

Exclusion criteria included: (1) prior use of any oral anticoagulants or aspirin; (2) any prescription of warfarin during the observation period; (3) combination therapy with or switching between DOACs and aspirin; and (4) failure to initiate DOAC or aspirin treatment on the index date to address immortal time bias.

Outcome Definition and Stroke Risk Prediction Tool

Among the study subjects, those who received dabigatran, rivaroxaban, apixaban, or edoxaban were categorized as the DOACs group, and those who received aspirin were categorized as the aspirin group. The two groups were matched by age and sex to adjust for confounding factors.

The primary outcome for effectiveness was the occurrence of ischemic stroke, while the primary safety outcome was the occurrence of major bleeding, including intracranial hemorrhage and gastrointestinal bleeding.

We used CHA₂DS₂-VASc as the stroke risk prediction tool, and the score was calculated based on the comorbidities using the disease information of the research data.

Since lifestyle factors such as smoking or alcohol consumption were not available in the claims database, alcohol use was identified indirectly through diagnostic codes related to alcohol use disorders. A detailed list of diagnostic codes used to define disease, comorbidities, and clinical outcomes is provided in Supplementary Table 1.

Statistical Analysis

Baseline characteristics were summarized as mean and standard deviation (SD) for the continuous variable age, and as frequencies and percentages for categorical variables including gender, insurance type, and comorbidities. Differences in age between groups were assessed using the independent t-test, while categorical variables were compared using the chi-square test or Fisher's exact test, as appropriate. Matching by age and sex was conducted to ensure balance between the two groups. Standardized mean differences (SMDs) were used to assess the balance of baseline characteristics after matching, with an SMD <0.1 considered indicative of a negligible difference. ²⁷ The matched data were analyzed to compare the effectiveness and safety between the two groups, and variables showing differences between the two groups after matching were included in the Cox proportional hazard regression to determine whether their impact on the results.

Survival analyses were used to evaluate outcomes, with individual time at risk calculated from the index date to the first occurrence of an outcome or a censoring event. Because the date of death could not be confirmed in the HIRA data used, censoring was done using the date of the last medical record of each patient. The log-rank test was used to compare survival curves between the groups, and Cox proportional hazards models were applied to estimate hazard ratios (HRs) and 95% confidence intervals (CIs) for risk factors associated with outcomes.

Statistical analyses were conducted using R software (version 3.5.1), and statistical significance was set at P < .05. All analyses were performed using the remote statistical analysis service site provided by HIRA (https://ras.hira.or.kr), and analytical results were exported through a secure, approved process, as HIRA prohibits direct data transfer.

Results

Characteristics of Study Population

A total of 144,688 patients were newly diagnosed with NVAF between January 1, 2017 and December 31, 2019. Among them, 2234 patients with a CHA₂DS₂-VASc score of 1 in men and 2 in women were included in the analysis. Of the included patients, 977 (43.7%) received DOACs, and 1257 (56.3%) were prescribed aspirin (Figure 1). After age and sex matching, a total of 1600 matched patients (800 per group) were included in the final analysis. Baseline characteristics of the unmatched and propensity score-matched cohorts are shown in Table 1. In the unmatched cohort, DOAC users were older (59.1 ± 8.9 vs 54.7 ± 9.8 years, P < .001) and included a higher proportion of females (42.3% vs 20.2%, P < .001). Regarding comorbidities, they showed higher rates of dementia (1.2% vs 0.4%, P = .046) and cancer (9.2% vs 6.8%, P = .047), while a lower rate of vascular disease (4.0% vs 12.9%, P < .001). After matching by age and sex, no significant differences were observed for these characteristics, except for a higher prevalence of congestive heart failure (16.9% vs 13.1%, P = .036) and a lower prevalence of vascular disease (4.5% vs 13.3%, P < .001) and liver disease (37.9% vs 43.5%, P = .022) in the DOAC group.

Figure 1.

Flowchart of study population selection.

Abbreviations: AF, atrial fibrillation; CHA₂DS₂-VA, congestive heart failure, hypertension, age ≥75 years [doubled], diabetes mellitus, prior stroke/transient ischemic attack/thromboembolism [doubled], vascular disease, age 65–74 years; DOAC, direct oral anticoagulant; NVAF, non-valvular atrial fibrillation; OAC, oral anticoagulant.

Table 1.

Baseline Characteristics of the Unmatched and Propensity Score-Matched Cohorts.

	Unmatched (n = 2234)			Matched (n = 1600)
Characteristics	DOAC (n = 977)	Aspirin (n = 1257)	P-value	DOAC (n = 800)	Aspirin (n = 800)	P-value	SMD
Age, y a	59.1 ± 8.9	54.7 ± 9.8	<.001	57.8 ± 9.0	57.3 ± 8.9	.250	0.058
Sex			<.001			1.000	<0.001
Male	564 (57.7)	1003 (79.8)		563 (70.4)	563 (70.4)
Female a	413 (42.3)	254 (20.2)		237 (29.6)	237 (29.6)
Insurance type			.277			.075	0.089
NHI	961 (98.4)	1227 (97.6)		786 (98.3)	775 (96.9)
MedAid/PVI	16 (1.6)	30 (2.4)		14 (1.8)	25 (3.1)
Comorbidities
Congestive heart failure a	153 (15.7)	198 (15.8)	1.000	135 (16.9)	105 (13.1)	.036	0.105
Hypertension a	427 (43.7)	595 (47.3)	.096	377 (47.1)	347 (43.4)	.132	0.075
Diabetes mellitus a	98 (10.0)	140 (11.1)	.440	88 (11.0)	92 (11.5)	.752	0.016
Vascular disease a	39 (4.0)	162 (12.9)	<.001	36 (4.5)	106 (13.3)	<.001	0.311
Severe renal disease	4 (0.4)	2 (0.2)	.414 b	3 (0.4)	2 (0.3)	1.000 b	0.022
Liver disease	369 (37.8)	521 (41.4)	.086	303 (37.9)	348 (43.5)	.022	0.115
Anemia	94 (9.6)	124 (9.9)	.904	72 (9.0)	82 (10.3)	.397	0.042
Dyslipidemia	546 (55.9)	685 (54.5)	.540	440 (55.0)	442 (55.3)	.920	0.005
Dementia	12 (1.2)	5(0.4)	.046	9 (1.1)	5 (0.6)	.283	0.054
Chronic pulmonary disease	468 (47.9)	552 (43.9)	.067	373 (46.6)	362 (45.3)	.581	0.028
Rheumatic disease	56 (5.7)	72 (5.7)	1.000	40 (5.0)	52 (6.5)	.198	0.064
Previous hemorrhage	192 (19.7)	242 (19.3)	.855	149 (18.6)	154 (19.3)	.750	0.016
Alcohol	39 (4.0)	60 (4.8)	.431	37 (4.6)	35 (4.4)	.809	0.012
Intracranial hemorrhage	6 (0.6)	7 (0.6)	1.000	4 (0.5)	4 (0.5)	1.000 b	<0.001
Gastrointestinal bleeding	65 (6.7)	73 (5.8)	.462	52 (6.5)	50 (6.3)	.838	0.010
Peripheral vascular disease	6 (0.6)	2 (0.2)	.087 b	5 (0.6)	2 (0.3)	.452 b	0.057
Cerebrovascular disease	50 (5.1)	52 (4.1)	.318	38 (4.8)	39 (4.9)	.907	0.006
Peptic ulcer	360 (36.8)	457 (36.4)	.846	292 (36.5)	318 (39.8)	.181	0.067
Cancer	90 (9.2)	86 (6.8)	.047	71 (8.9)	67 (8.4)	.722	0.018

Note: Data are presented as mean ± SD or n (%).

^aComponents of the CHA₂DS₂-VASc score. Age 65–74 years, female sex, and the other components contribute 1 point each. Patients aged ≥75 years or with prior stroke, TIA, or systemic embolism (each assigned 2 points) were excluded from the study subjects by design.

^bFisher's exact test.

Abbreviations: DOAC, direct oral anticoagulant; MedAid, Medical Aid; NHI, National Health Insurance; PVI, Patriots and Veterans Insurance; SMD, standardized mean difference.

Clinical Outcomes of Matched Cohort

In the matched cohort, the incidence rate of ischemic stroke was higher in the DOAC group (1.46 per 100 person-years; 95% CI, 1.00-1.97) compared to the aspirin group (0.92 per 100 person-years; 95% CI, 0.57-1.30), although the difference was not statistically significant (P = .060). Similarly, the incidence rate of major bleeding was 2.26 per 100 person-years (95% CI, 1.69-2.94) in the DOAC group and 2.10 per 100 person-years (95% CI, 1.57-2.73) in the aspirin group, with no significant difference (P = .100). (Figure 2 and Table 2).

Figure 2.

Kaplan-Meier survival curves for clinical outcomes. (A) Ischemic stroke, (B) Major bleeding.

Abbreviation: DOAC, direct oral anticoagulant.

Table 2.

Event Rates of Clinical Outcomes in the Propensity Score-Matched Cohort.

Clinical Outcome	DOAC	Aspirin	P-value
Ischemic stroke	1.46 (1.00–1.97)	0.92 (0.57–1.30)	.060
Major bleeding	2.26 (1.69–2.94)	2.10 (1.57–2.73)	.100

Note: Event rates (95% confidence interval) are presented as per 100 person-years.

Abbreviations: DOAC, direct oral anticoagulant.

In the adjusted model, no statistically significant association was found between the treatment group (DOAC vs aspirin) and the risk of ischemic stroke (Supplementary Table 2) or major bleeding (Supplementary Table 3). Instead, we identified several clinical predictors for these outcomes: rheumatic disease showed a significant association with ischemic stroke, whereas liver disease was significantly associated with major bleeding.

Discussion

Recommendations for anticoagulation therapy for high-risk AF patients, as well as the preferential use of DOACs over warfarin, are consistent across various guidelines.^{5–10,28–30} For NVAF patients with a CHA₂DS₂-VASc score of 1 in men and 2 in women, recommendations have gradually shifted from allowing no therapy or aspirin (2014) to suggesting anticoagulation as “reasonable” in the 2023 update.^5,12

In Korea, the 2014 guidelines recommended anticoagulation therapy for AF patients with a CHA₂DS₂-VASc score of ≥1 in men and ≥2 in women. ³¹ However, the 2018 update adjusted this to “consider” anticoagulation for patients with a CHA₂DS₂-VASc score of 1 in men and 2 in women. ⁹ Despite these evolving recommendations, underutilization of anticoagulants remains high in clinical practice. Studies have shown that even high-risk patients often fail to receive anticoagulation therapy. In a cross-sectional study using national claims data from 2018, only 65.5% of elderly NVAF patients with a CHA₂DS₂-VASc score of ≥2 were receiving anticoagulation. ²³ Therefore, our findings should be interpreted within the context of these evolving recommendations and applied cautiously in higher risk patients.

The 2020 Canadian Cardiovascular Society/Canadian Heart Rhythm Society (CCS/CHRS) guidelines used the CHADS-65 (congestive heart failure, hypertension, age ≥65 years, diabetes mellitus, prior stroke/TIA) score, excluding vascular disease and gender, and recommended anticoagulants for patients with a CHADS-65 ≥ 1, while antiplatelet therapy was reserved for those with a CHADS-65 = 0 and vascular disease. ⁷ This approach highlights the emphasis on age (≥65 years) as a decisive factor for anticoagulation, which differs from the CHA₂DS₂-VASc scoring system. In this context, our study contributes additional evidence for patients younger than 65 years with a CHA₂DS₂-VASc score of 1 in men or 2 in women, who are not directly addressed by the CHADS-65 scheme but remain clinically challenging with respect to treatment decisions.

This study found no statistically significant difference between DOACs and aspirin in preventing ischemic stroke or major bleeding in patients with a CHA₂DS₂-VASc score of 1 in men and 2 in women. However, this should not be interpreted as evidence of therapeutic equivalence. The results likely reflect real-world treatment patterns in an intermediate-risk population where treatment decisions are nuanced. Further subgroup analyses based on individual CHA₂DS₂-VASc components, particularly vascular disease, are warranted.

Current guidelines do not recommend aspirin monotherapy for stroke prevention in AF patients, even those at intermediate risk. Therefore, our findings should not be misinterpreted to support the use of aspirin over anticoagulants in such patients. Rather, these findings reflect a clinical reality where aspirin continues to be used despite guideline recommendations, highlighting the gap between evidence-based guidelines and real-world practice. This study contributes to bridging that gap by providing real-world data, especially in Asian populations, and underscores the need for continued efforts to improve anticoagulation rates where indicated.

In this study, rheumatic diseases, including rheumatoid arthritis and systemic connective tissue disorders, were significantly associated with increased stroke incidence. This finding aligns with previous literature suggesting a prothrombotic state and increased vascular risk in patients with systemic inflammatory conditions.^32,33

Several limitations of this study should be noted. First, residual confounding may exist despite propensity score matching, and the marginally higher stroke incidence in the DOAC group (P = .060) may reflect unmeasured or imbalanced clinical factors. In particular, aspirin-treated patients were more often classified as intermediate risk due to vascular disease, whereas DOAC-treated patients were more often classified based on congestive heart failure. Second, reliance on claims data limits the accuracy of diagnoses and outcomes, and precludes access to laboratory results or imaging findings. Third, our study design was based on the CHA₂DS₂-VASc score calculated at the index date. We could not account for potential changes in a patient's stroke risk profile over the follow-up period (ie, time-varying covariates), such as the new onset of hypertension or diabetes, which may have led to some misclassification. Fourth, our analysis grouped dabigatran, rivaroxaban, apixaban, and edoxaban into a single ‘DOAC’ class. Due to the modest sample size of this specific intermediate-risk subgroup, we lacked the statistical power to compare individual DOACs, which may have distinct efficacy and safety profiles. Fifth, this study did not include systemic embolism, death, or cardiovascular hospitalizations as outcomes, which may limit the generalizability of the findings. Sixth, these results are derived from an Asian population and may not apply to other ethnic groups or healthcare systems. Lastly, lifestyle factors such as smoking and medication adherence were not accounted for, which may have influenced clinical outcomes.

Nevertheless, this study offers several important strengths. It utilized a large-scale, nationally representative claims database, allowing for generalizability within the Korean population and reflecting real-world clinical practice. Furthermore, it focused on an intermediate-risk subgroup (CHA₂DS₂-VASc score of 1 in men and 2 in women) that is often underrepresented in randomized clinical trials and guideline development. Importantly, it provides Asian-specific real-world evidence, addressing a critical gap in stroke prevention data for this population. These findings may help inform region-specific clinical decisions and future updates to stroke prevention strategies.

Conclusion

This study found no statistically significant differences in the risk of ischemic stroke or major bleeding between DOACs and aspirin in NVAF patients with a CHA₂DS₂-VASc score of 1 in men and 2 in women. However, the borderline difference observed in ischemic stroke incidence (P = .060) should be interpreted with caution, as it may partly reflect differences in underlying vascular disease burden and stroke mechanisms between treatment groups rather than a true difference in antithrombotic efficacy. While these findings may support the consideration of both treatments in selected cases, individualized treatment decisions based on patient characteristics and strict adherence to clinical guidelines remain crucial. Further real-world and prospective studies are warranted to confirm these findings and clarify their generalizability.