Intracerebral hemorrhage risk in Alzheimer's disease patients on anticoagulants: A nationwide cohort study

Abstract

Background

The risk of intracerebral hemorrhage (ICH) in Alzheimer's disease (AD) patients undergoing anticoagulation (AC) remains unclear.

Objective

This nationwide cohort study assessed ICH risk in AC-treated AD patients.

Methods

Using Taiwan's National Health Insurance Research Database, we identified 1638 AC-treated AD patients and established four matched cohorts: AD with AC, AD without AC, non-AD with AC, and non-AD without AC. We applied inverse probability of treatment weighting (IPTW) and competing risks regression (CRR) to adjust for confounders and mortality risk. Cox proportional hazards regression estimated ICH risk.

Results

The ICH incidence per 100 person-years was 2.21 in AD patients with AC, 1.03 without AC, 1.71 in non-AD with AC, and 0.28 in non-AD without AC. After IPTW adjustment, compared to non-AD without AC, AD patients with AC had the highest ICH risk (aHR 1.94, 95% CI: 1.44–2.61), followed by non-AD with AC (aHR 1.84, 95% CI: 1.51–2.23) and AD without AC (aHR 1.74, 95% CI: 1.42–2.11). ICH risk in AC-treated AD patients was comparable to non-AD with AC. Subgroup analysis showed higher ICH risk in females and those with hyperlipidemia, diabetes, heart failure, chronic kidney disease, or cancer.

Conclusions

AD is associated with increased ICH risk, which is further elevated by AC use. These findings highlight the need for individualized risk-benefit evaluation, particularly in high-risk populations.

Keywords

Alzheimer's disease amyloid targeting therapy anticoagulant intracerebral hemorrhage risk

Introduction

Alzheimer's disease (AD) is the most prevalent neurodegenerative dementia, affecting 57.4 million people worldwide in 2019, with projections reaching 152.8 million by 2050.¹ The primary pathological hallmark of AD is amyloid-β accumulation,² and emerging treatments targeting amyloid-β are rapidly advancing.^3–5 In addition to amyloid pathology, other evolving pathologies, including amyloid-facilitated tau pathology, synaptic dysfunction, and neuroinflammation, have been widely recognized.⁶ Furthermore, co-pathologies are frequently observed in AD patients, among which overlapping cerebrovascular pathology is common.⁷

Given the cerebrovascular involvement, patients with AD face an increased risk of clinical stroke, and the use of antithrombotic regimens may become necessary in this population.⁸ Among these therapies, anticoagulation (AC) therapy is commonly prescribed for the prevention of cardioembolic stroke, particularly in patients with atrial fibrillation (AF) or other cardiac conditions that predispose to thromboembolic events, such as valvular heart disease.⁹ Notably, both AD and AF share advanced age as a major risk factor, and their prevalence increases substantially in the aging population.^8,10 As a result, the likelihood of individuals concurrently diagnosed with AD and requiring stroke prevention for AF is expected to rise. However, due to the overlap of cerebrovascular pathology, such as cerebral amyloid angiopathy (CAA), patients with AD face an increased risk of intracerebral hemorrhage (ICH),¹¹ with a pooled effect size of 1.41 as demonstrated in a meta-analysis.¹² Moreover, a recent study also found a causal relationship between AD and ICH through Mendelian randomization, with amyloid-β playing an important role in the process.¹³ These findings raise important safety concerns when considering AC therapy in this vulnerable population.

Therefore, understanding the background incidence of ICH in AD patients receiving AC therapy may provide critical insights into safety considerations when prescribing these medications. This study aims to characterize the baseline ICH risk in AD patients receiving AC and try to clarify whether the ICH risk is primarily driven by AC therapy or AD itself.

Methods

Study design and dataset

This study is an observational, retrospective, population-based cohort study. The AD and non-AD cohorts were selected from Taiwan's National Health Insurance Research Database (NHIRD). Taiwan's National Health Insurance (NHI) is a universal healthcare program covering 23 million people nationwide, providing comprehensive clinical data. The NHIRD includes demographic information, diagnoses, treatment procedures, and prescription records, offering a robust dataset for analysis.

Standard protocol approvals, registrations, and patient consents

The study was approved by the institutional review board (IRB) of National Cheng Kung University Hospital (NCKUH) (IRB Approval No. A-ER-110-097). Participant consent was waived under IRB approval.

Study participants, validation of AD diagnosis, and cohort formation

Data from the NHIRD were collected from 2010 to 2021. The AD cohort was defined as patients with at least two NHI ambulatory claims or one inpatient record with an AD-related diagnosis, based on ICD-9-CM code 331.0 or ICD-10-CM code G30. Additionally, patients diagnosed with dementia (ICD-9-CM codes 290.0–290.3, 294.1, 331.2; ICD-10-CM codes F03.9, F02.80, F02.81, G31.1) were included if they concurrently received prescriptions for acetylcholinesterase inhibitors (AChEIs) or memantine, ensuring diagnostic validation. In Taiwan, strict regulations govern the prescription of NHI-covered AChEIs and memantine for dementia patients.^14,15 Before 2017, prescriptions required approval from an expert committee composed of neurologists or psychiatrists, who confirmed the clinical diagnosis of AD on a case-by-case basis. After 2017, clinicians were still required to exclude other potential causes of cognitive decline, such as vascular dementia or metabolic disorders, before prescribing these medications. Furthermore, only board-certified neurologists or psychiatrists were authorized to prescribe NHI-covered AChEIs or memantine. Given these stringent regulations, the diagnostic validity of AD in the NHIRD is high, as patients diagnosed with dementia and receiving NHI-covered AChEIs or memantine are likely to have undergone rigorous diagnostic evaluation.

The validation of ICD-9-CM and ICD-10-CM codes for identifying AD was conducted by reviewing the medical records of 200 patients at NCKUH, a 1198-bed tertiary medical center in southern Taiwan. We randomly selected 100 patients diagnosed with AD (ICD-9-CM code 331.0; ICD-10-CM code G30) from inpatient and outpatient databases, ensuring each had at least two outpatient records or one inpatient record within a year. Medical charts and neuroimaging were reviewed to confirm the clinical diagnosis of AD based on the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-V). The positive predictive value (PPV) for AD diagnosis using these criteria was estimated at 93% (95% CI: 86.3%–96.6%),¹⁴ indicating a high level of diagnostic accuracy.

The AD cohort was divided into two groups based on AC use. The AD with AC group (AD ⁺ AC⁺) included patients who received oral AC, either vitamin K antagonists (VKAs) such as warfarin or direct oral anticoagulants (DOACs), including dabigatran (a direct thrombin inhibitor) and factor Xa inhibitors (rivaroxaban, apixaban, edoxaban). Only oral AC users were included, while those receiving intravenous or intramuscular anticoagulants, such as unfractionated heparin or low-molecular-weight heparin (LMWH), were excluded. Patients with continuous AC use for at least three months were included, and the index date was defined as the date of first AC prescription. Patients who received AC therapy less than three continuous months were excluded from the final cohort to avoid misclassification and potential bias in bleeding risk estimation. The AD without AC group (AD ⁺ AC⁻) consisted of patients with AD who did not receive any AC therapy. Patients were excluded if they had a history of ICH or subarachnoid hemorrhage (SAH) (ICD-9-CM codes 430–432; ICD-10-CM codes I60–I62) before the index date (either before AC initiation in the AC group or before AD diagnosis in the non-AC group) or had overlapping use of antiplatelets and AC for at least one month in the AC group. After applying these criteria, the AD ⁺ AC⁻ and AD ⁺ AC⁺ included 137,913 and 3596 patients, respectively. Figure 1 illustrates the study flowchart.

Figure 1.

Study flowchart.

To establish the non-AD cohort, we extracted patients from the NHIRD who had no history of AChEIs or memantine use and no diagnosis of AD or any other forms of dementia (ICD-9-CM codes 290.0–290.4, 294.1, 331.0–331.2; ICD-10-CM codes G30, F03.9, F02.80, F02.81, G31.1). Patients diagnosed with ICH or SAH before the index date of AC use (for the AC group) and those with overlapping antiplatelet and AC use for at least one month were also excluded. This resulted in a non-AD cohort of 27,205,764 cases, of which 151,844 were non-AD patients with AC use (AD⁻AC⁺), and 27,053,920 were non-AD patients without AC use (AD⁻AC⁻). The criteria for classifying AC + and AC− status in the non-AD cohort were consistent with those applied in the AD cohort.

We then performed exact matching at a 1:5:5:5 ratio, matching each patient in the AD ⁺ AC⁺ group with five patients from each of the AD ⁺ AC⁻, AD⁻AC⁺, and AD⁻AC⁻ groups based on age, sex, and index year. After exact matching, the final sample sizes were 1638 patients in the AD + AC + group and 8190 patients in each of the other three groups.

Potential confounders

Several potential confounders associated with an increased risk of AD, stroke, and ICH were identified, including age, sex, and pre-existing or coexisting medical conditions. These conditions comprised hypertension, diabetes mellitus, hyperlipidemia, ischemic stroke, coronary artery disease, heart failure, AF, chronic renal disease, and peripheral artery occlusive disease. Additionally, other factors contributing to bleeding tendencies, such as malignancy, chronic alcohol use, and liver cirrhosis, were also considered.

We also recognize that urbanization level, monthly income, and living area can influence various risk factors. The urbanization level was classified based on the method proposed by Liu et al. (2006),¹⁶ categorizing areas into three levels, where level 1 represents the most urbanized regions and level 3 the least urbanized.

Main outcome measures

The main outcome was the diagnosis of ICH within the observational period, identified using diagnostic codes from Taiwan's NHIRD. In Taiwan, the clinical diagnosis of ICH typically adheres to the 2013 American Heart Association criteria for stroke,¹⁷ which are based on clinical presentation combined with neuroimaging findings. To ensure diagnostic validity, only patients with an inpatient ICH diagnosis code—whether diagnosed in the emergency room or during hospitalization—and with a corresponding procedure code indicating that neuroimaging (computed tomography [CT] or magnetic resonance imaging [MRI]) was performed during the same admission were included in the analysis.

Statistical analysis

Each group was followed from the index diagnosis until December 31, 2021, with follow-up periods ranging from 1 to 11 years. To assess differences in demographic and clinical characteristics, we used the Wald chi-square test for categorical variables and a generalized linear model for continuous variables, analyzing factors such as age, sex, living area, urbanization level, enrollee category, monthly income, and comorbidities. To account for baseline imbalances and adjust for potential confounders, we applied inverse probability of treatment weighting (IPTW) based on propensity scores, ensuring adequate sample size while improving comparability between groups.¹⁸

To assess the difference in ICH risk between the AD and non-AD cohorts, with and without AC treatment, we applied weighted Cox proportional hazard regression analysis to estimate the hazard ratio. Given that death may serve as a competing risk for both AD and ICH, we utilized the Fine and Gray competing risk regression (CRR) model to account for this factor and adjust for the risk of death. All statistical analyses were conducted using SAS for Windows (version 9.2; SAS Institute Inc., Cary, NC, USA), with statistical significance defined as p < 0.05.

Results

Demographic analysis of patient data after matching for age, sex, and index year showed that the majority of patients were men (55.37%) and those aged over 65 years (99.82%). While the main covariates of age, sex, and index year were balanced between the AD cases and non-AD controls, differences were observed among the four cohorts in factors such as living area, urbanization level, enrollee category, and prevalence of comorbidities (Table 1). Additional demographic characteristics are provided in the Supplemental Material (Supplemental Table 1). To adjust for unbalanced covariates, we then applied the IPTW method before estimating the ICH risk and using a CRR model to adjust for mortality among the four groups.

Table 1.

Demographic information among the four cohorts (before IPTW).

Variables	AD⁻AC⁻ (N = 8190)	AD⁻AC⁺ (N = 8190)	AD ⁺ AC⁻ (N = 8190)	AD ⁺ AC⁺ (N = 1638)	p
Variables	n (%)	n (%)	n (%)	n (%)	p
Age, Mean ± SD	81.8 ± 7.4	81.8 ± 7.4	81.8 ± 7.4	81.8 ± 7.4	1.000
<65	15 (0.18)	15 (0.18)	15 (0.18)	3 (0.18)	1.000
≥65	8175 (99.82)	8175 (99.82)	8175 (99.82)	1635 (99.82)
Male	4535 (55.37)	4535 (55.37)	4535 (55.37)	907 (55.37)	1.000
Comorbidities
IS	54 (0.66)	1427 (17.42)	207 (2.53)	316 (19.29)	<0.001
HL	990 (12.09)	2637 (32.20)	1781 (21.75)	451 (27.53)	<0.001
DM	932 (11.38)	2934 (35.82)	1901 (23.21)	592 (36.14)	<0.001
Hypertension	1998 (24.40)	6530 (79.73)	3951 (48.24)	1240 (75.70)	<0.001
CAD	220 (2.69)	3239 (39.55)	483 (5.90)	614 (37.48)	<0.001
Heart failure	124 (1.51)	3579 (43.70)	332 (4.05)	637 (38.89)	<0.001
CKD	324 (3.96)	2166 (26.45)	804 (9.82)	418 (25.52)	<0.001
PAOD	16 (0.20)	142 (1.73)	42 (0.51)	32 (1.95)	<0.001
AF	31 (0.38)	5410 (66.06)	127 (1.55)	1002 (61.17)	<0.001
Smoking	12 (0.15)	65 (0.79)	27 (0.33)	12 (0.73)	<0.001
Cancer	643 (7.85)	1525 (18.62)	1031 (12.59)	300 (18.32)	<0.001
Alcohol use	7 (0.09)	12 (0.15)	20 (0.24)	5 (0.31)	0.042
Cirrhosis	176 (2.15)	345 (4.21)	275 (3.36)	67 (4.09)	<0.001

IPTW: inverse probability of treatment weighting; IS: ischemic stroke; HL: Hyperlipidemia; DM: Diabetes mellitus; CAD: Coronary arterial disease; CKD: Chronic kidney disease; PAOD: peripheral arterial occlusion disease; AF: Atrial fibrillation.

In the AD cohort, the incidence of ICH was 2.21 per 100 person-years with AC use and 1.03 per 100 person-years without AC use. In the non-AD cohort, the incidence was 1.71 per 100 person-years with AC use and 0.28 per 100 person-years without AC use. Overall, AD patients had a higher risk of ICH than non-AD controls (Table 2 and Supplemental Table 2). After IPTW adjustment, compared to the AD⁻AC⁻ cohort, the AD ⁺ AC⁺ cohort (adjusted hazard ratio [AHR], 1.94; 95% CI: 1.44–2.61; p < 0.001), the AD ⁺ AC⁻ cohort (aHR, 1.74; 95% CI: 1.42–2.11; p < 0.001), and the AD⁻AC⁺ cohort (aHR, 1.84; 95% CI: 1.51–2.23; p < 0.001) all exhibited a significantly increased risk of ICH. After adjusting for mortality, the CRR model continued to show elevated risks of ICH in the AD ⁺ AC⁺, AD ⁺ AC⁻, and AD⁻AC⁺ cohorts compared to the AD⁻AC⁻ cohort, with subdistribution hazard ratios (SHR) of 2.22 (95% CI: 1.61–3.07), 1.78 (95% CI:1.45–2.19), and 2.16 (95% CI: 1.61–3.07), respectively (Table 2).

Table 2.

The risk of ICH among the four cohorts.

	AD⁻AC⁻^(N ⁼ ⁸¹⁹⁰⁾	AD⁻AC⁺^(N ⁼ ⁸¹⁹⁰⁾	AD ⁺ AC⁻^(N ⁼ ⁸¹⁹⁰⁾	AD ⁺ AC⁺^(N ⁼ ¹⁶³⁸⁾
Before IPTW
Incidence of ICH (per 100 person-year)	0.28	1.71	1.03	2.21
Incidence of ICH, n (%)	76 (0.93)	382 (4.66)	232 (2.83)	99 (6.04)
After IPTW
Crude HR	Ref.	2.72 (2.26–3.29)	2.21 (1.83–2.68)	2.95 (2.20–3.94)
Adjusted HR^a	Ref.	1.84 (1.51–2.23)	1.74 (1.42–2.11)	1.94 (1.44–2.61)
p	-	<0.001	<0.001	<0.001
Adjusted CRR^a	Ref.	2.16 (1.61–3.07)	1.78 (1.45–2.19)	2.22 (1.61–3.07)
p	-	<0.001	<0.001	<0.001
Using the AD⁻AC⁺group as the reference
Crude HR	0.37 (0.30–0.44)	Ref.	0.81 (0.68–0.97)	1.08 (0.82–1.43)
Adjusted HR^a	0.55 (0.45–0.66)	Ref.	0.95 (0.79–1.13)	1.06 (0.80–1.39)
p	<0.001	-	0.530	0.699
Adjusted CRR^a	0.46 (0.37–0.58)	Ref.	0.82 (0.69–0.99)	1.03 (0.78–1.36)
p	<0.001	-	0.035	0.841

Adjustment was made for variables’ SMD > 0.1.

In both the AD⁻ and AD⁺ groups, AC treatment was associated with an increased risk of developing ICH. To evaluate the relative contributions of AD and AC to ICH risk, we conducted additional analyses using the AD⁻AC⁺ group as reference (Table 2 and Supplemental Table 2). In this analysis, the risk of ICH in the AD ⁺ AC⁺ group was comparable to the reference group (aHR 1.06, 95% CI: 0.80–1.39, p = 0.699). The CRR model showed similar results (SHR 1.03, 95% CI: 0.78–1.36, p = 0.841). We also used the AD ⁺ AC⁻ group as the reference group and observed a slightly elevated risk of ICH in the AD ⁺ AC⁺ group, although this difference did not reach statistical significance. After adjusting for mortality, an elevated risk of ICH was noted in the AD⁻AC⁺ group compared to the AD ⁺ AC⁻ group (SHR 1.22, 95% CI:1.01–1.46, p = 0.035) (Supplemental Table 3). These findings may suggest that the increased ICH risk in AD patients receiving AC treatment is primarily attributable to AC rather than AD itself.

In the subgroup analysis, we aimed to identify covariate factors that interact with the risk of ICH in AD patients receiving AC treatment by assessing relative excess risk due to interaction (RERI).¹⁹ To minimize the potential influence of IPTW weighting on the subgroup analysis, we conducted the analyses using the original, unweighted cohort prior to IPTW adjustment. We defined the AHR for the subgroups as follows: AD⁻AC⁺, AD ⁺ AC⁻, and AD ⁺ AC⁺ as AHR_AD⁻_AC⁺, AHR_AD⁺ _AC⁻, and AHR_AD ⁺ _AC⁺, respectively. The RERI was then calculated with the formula listed below:

RERI = {AHR}_{A D^{+} A C^{+}} - ({AHR}_{A D^{-} A C^{+}} + {AHR}_{A D^{+} A C^{-}} - 1)

A synergistic interaction was defined as RERI > 0, an attenuated (less-than-additive) interaction as RERI < 0, and no interaction as RERI = 0. For example, among female AD patients, the AHRs were 3.48 for AD⁻AC⁺, 3.26 for AD ⁺ AC⁻, and 5.94 for AD ⁺ AC ⁺. Since RERI value in the female subgroup was greater than 0, this finding suggests the presence of a synergistic interaction between AC therapy and AD on ICH risk. Based on this framework, we found that AD patients with certain covariates—including female sex, hyperlipidemia, statin use, diabetes mellitus, heart failure, chronic kidney disease, and cancer—exhibited synergistic interaction in ICH risk when receiving AC treatment (Figure 2).

Figure 2.

ICH risk stratified by demographics and comorbidities.

Discussion

Our study provided real-world data on the background incidence rate of ICH in AD patients, with or without AC therapy, using a nationwide database. Previous studies have indicated that both AD patients and those receiving AC therapy are at an increased risk of developing ICH.^12,13,20,21 Our study suggests that AD patients have a higher ICH risk, which increases further with AC use. However, in our study, AD patients on AC have a similar ICH risk to non-AD patients on AC, suggesting that AC use may play a more prominent role in ICH risk, while AD may contribute to a lesser extent.

Our study demonstrated that the incidence of ICH in AD⁻AC⁺ patients was 1.71 per 100 person-years, which is consistent with prior studies on ICH rates in patients receiving oral anticoagulants in Taiwan using the NHIRD, reporting an ICH incidence of 0.5–2.4 per 100 person-years.^22,23 In contrast, the incidence of ICH in AD ⁺ AC⁺, AD ⁺ AC⁻, and AD⁻AC⁻ patients were 2.21, 1.03, and 0.28 per 100 persons-year, respectively. The higher ICH incidence in AD ⁺ AC⁻ patients compared to the AD⁻AC⁻ group suggests that AD patients have an inherently higher risk of ICH than non-AD patients. These findings align with previous studies.^11,12,24 The potential mechanism may involve the co-pathology of AD and CAA—a microangiopathy associated with cerebral microbleeds and lobar ICH—recognized as an additional risk factor for increased ICH risk in AD patients,²⁴ along with the co-pathology of AD and cerebrovascular pathology. Our findings also provide important background information ahead of the global implementation of anti-amyloid treatments for AD, as data on ICH risk in AD patients receiving anticoagulation therapy remain limited.²⁵

Our study found that AD patients had a 1.74-fold increased risk of ICH compared to non-AD patients not taking AC, while AC use mildly elevated the risk to 1.94-fold. In contrast, non-AD patients receiving AC faced a 1.84-fold risk of ICH. A direct comparison between the AC users, regardless of AD status, showed no significant difference in ICH risk. These results may support the appropriate use of AC in AD patients when clinically indicated. However, given the known background risk of ICH in AD, appropriate and careful dose selection may lead to safer outcomes, as demonstrated in the ELDERCARE-AF trial,²⁶ where a very low dose selection was safe in very elderly patients with comorbidities or higher bleeding risk. Nonetheless, given the high risk of ARIA-H and the underlying pathophysiology, careful evaluation is recommended when considering the simultaneous use of amyloid targeting therapies and AC to mitigate the potential risk of ICH.

In accordance with previous studies,^27–31 our study found that AD patients with certain comorbidities, including diabetes mellitus, heart failure, CKD, and cancer, had a higher risk of ICH when receiving AC. The underlying mechanisms are complex and require further investigation. Diabetes mellitus can lead to vasculopathy, affecting small to large blood vessels through macrovascular atherosclerosis, microvascular endothelial dysfunction, basement membrane thickening, and microthrombosis,^27,32 while β-amyloidopathy in AD may further exacerbate endothelial dysfunction, increasing ICH risk in patients with both conditions. Heart failure, with its diverse etiologies—including ischemic, valvular, hypertensive, and cardiomyopathic origins—is often associated with multiple comorbidities,^28,33 and its interaction with β-amyloidopathy may contribute to an elevated risk of ICH. Similarly, CKD is known to cause endothelial dysfunction,³⁴ uremic bleeding,³⁵ reduced platelet retention and aggregation,³⁶ and impaired cerebral autoregulation,³⁷ all of which increase susceptibility to ICH due to endothelial injury, as supported by a recent three-stage analysis combining observational and genetic data.²⁹ The interaction between β-amyloidopathy and these pathophysiological changes may further heighten ICH risk in AD patients with CKD. Additionally, cancer patients exhibit an increased risk of ICH due to cancer-related coagulopathy,³⁰ cytokine-induced vascular inflammation,³¹ and the potential for brain metastasis,³⁸ all of which contribute to hemorrhagic vulnerability. Furthermore, cancer treatments such as chemotherapy and radiation therapy may further compromise vascular integrity, increasing the likelihood of ICH.^39,40

We found that AD patients with hyperlipidemia had an additive increase in ICH risk when receiving AC therapy. Several global epidemiological studies have reported an inverse association between cholesterol and LDL levels and ICH risk.^41–44 A recent study identified hyperlipidemia as an independent risk factor for ICH in patients receiving AC.⁴⁵ The underlying pathophysiology remains unclear, but one possible explanation is the use of LDL-lowering therapy in patients with hyperlipidemia.^44,45 In our study, subgroup analysis further demonstrated that among statin users, a synergistic effect between AD and AC use on ICH risk was observed, suggesting that statin therapy may play a role in modulating the interaction between hyperlipidemia, AC, and ICH risk in AD patients. However, due to the ongoing controversy surrounding LDL-lowering therapy and ICH risk,⁴⁶ as well as the absence of LDL level data in our study, the interaction between hyperlipidemia, LDL-lowering therapy, AC use, and ICH risk in AD patients warrants further investigation.

From the subgroup analysis and RERI assessment, an attenuated interaction effect between AD and AC on the risk of ICH was observed among certain clinical subgroups, including aged ≥65 years, males, those with or without ischemic stroke and coronary artery disease, and those without hyperlipidemia, diabetes mellitus, heart failure, chronic kidney disease, peripheral artery occlusive disease, AF, cancer, alcohol use, liver cirrhosis, or smoking history. These findings suggest that the combined impact of AD and AC therapy on ICH risk may be diminished in these subgroups, potentially due to underlying physiological factors, ceiling effects, or differential clinical management practices in these populations.

However, these findings should be interpreted with caution, as they are based on subgroup analyses, which may be subject to limited statistical power and potential issues related to multiple comparisons. Further research is warranted to elucidate the mechanisms underlying these subgroup-specific interactions and to determine whether tailored AC strategies could help optimize safety in AD patients with varying vascular risk profiles.

There are several limitations to our study. First, despite matching for age, sex, and index year, and applying the IPTW method, covariates could not be fully identical across the four cohorts, particularly with respect to comorbidities. Previous studies have identified certain comorbidities as predisposing factors for ICH risk.^{27–30,40,44} To address this, additional adjustments were made when performing the Cox proportional hazard regression and CRR models. Second, we were unable to identify the specific characteristics of ICH cases. Due to the constraints of the NHIRD, we could not provide detailed clinical or radiologic features of individual ICH patients. Third, our cohort included users of both DOACs and VKAs. Data on drug compliance among DOACs users and time in the therapeutic range for VKA users were not available in the NHIRD, both of which may influence ICH risk. In addition, information on appropriate DOAC dosing was lacking, which is important since DOAC dosage may also affect bleeding risk.^23,26,45 Furthermore, differences in bleeding risks between DOACs and VKAs among AD patients warrants future investigation. Fourth, the NHIRD does not provide information on individual lifestyle factors, such as smoking quantity, use of quit-aid medications, or levels of alcohol consumption, which are major risk factors for stroke and ICH.⁴⁷ Lastly, our study included only Taiwanese patients. Given that the risks, incidence, and pathogenesis of ICH vary across different racial and genetic backgrounds,^48,49 our findings may not be generalizable to global populations. Future prospective studies are needed to further explore these issues.

Conclusion

Our population-based study provided crucial background data on the ICH rate and risk in AD patients with and without AC therapy. The increased ICH risk observed in AD patients receiving AC may be more associated with AC use than with AD itself. When considering AC therapy for AD patients in clinical practice, individualized assessment of the indication for AC, careful adjustment of dosing strategies, and vigilant monitoring for bleeding risk are essential to optimizing outcomes in this vulnerable population, particularly among females and patients with comorbidities such as hyperlipidemia, diabetes mellitus, heart failure, chronic kidney disease, and cancer.

Supplemental Material

sj-docx-1-alz-10.1177_13872877251362362 - Supplemental material for Intracerebral hemorrhage risk in Alzheimer's disease patients on anticoagulants: A nationwide cohort study

Supplemental material, sj-docx-1-alz-10.1177_13872877251362362 for Intracerebral hemorrhage risk in Alzheimer's disease patients on anticoagulants: A nationwide cohort study by Che-Hui Hsiao, Cheng-En Wu, Chun-Min Wang, Tien-Yu Lin, Ren-Ying Wu, Meng-Ju Wu, Sheng-Hsiang Lin and Pi-Shan Sung in Journal of Alzheimer's Disease

Footnotes

Acknowledgements

We are grateful to Hsu Chih-Hui for providing the statistical consulting services from the Biostatistics Consulting Center, National Cheng Kung University Hospital.

ORCID iD

Pi-Shan Sung

Ethical considerations

This research was approved by the IRB of National Cheng Kung University Hospital (IRB Approval No. B-ER-109-168).

Consent to participate

Participant consent was waived under IRB approval.

Consent for publication

Not applicable

Author contributions

Che-Hui Hsiao: Conceptualization; Data curation; Writing - original draft.

Cheng-En Wu: Data curation; Investigation; Resources.

Chun-Min Wang: Data curation; Software.

Tien-Yu Lin: Data curation.

Ren-Ying Wu: Data curation.

Meng-Ju Wu: Conceptualization; Supervision.

Sheng-Hsiang Lin: Data curation; Formal analysis; Methodology; Software; Validation.

Pi-Shan Sung: Conceptualization; Formal analysis; Funding acquisition; Project administration; Supervision; Writing - review & editing.

Funding

The study was supported by a research grant from the Ministry of Health and Welfare (MOHW 113-2314-B-006-108-) (MOHW 113-2321-B-006-018-) and National Cheng Kung University Hospital (NCKUH-11402007)

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Data availability statement

Data are available from the NHIRD published by Taiwan National Health Insurance Administration. Due to legal restrictions imposed by the government of Taiwan in relation to the “Personal Information Protection Act”, data cannot be made publicly available. Requests for data can be sent as a formal proposal to the NHIRD ().

Supplemental material

Supplemental material for this article is available online.

References

GBD 2019 Dementia Forecasting Collaborators. Estimation of the global prevalence of dementia in 2019 and forecasted prevalence in 2050: an analysis for the Global Burden of Disease Study 2019. Lancet Public Health 2022; 7: e105–e125.

Jack

Jr Andrews

Beach

, et al. Revised criteria for diagnosis and staging of Alzheimer's disease: alzheimer's Association Workgroup. Alzheimers Dement 2024; 20: 5143–5169.

Valenzuela

Pascual-Leone

. Lecanemab in early Alzheimer's disease. N Engl J Med 2023; 388: 1630.

Sims

Zimmer

Evans

, et al. Donanemab in early symptomatic Alzheimer disease: the TRAILBLAZER-ALZ 2 randomized clinical trial. JAMA 2023; 330: 512–527.

Sevigny

Chiao

Bussiere

, et al. The antibody aducanumab reduces Abeta plaques in Alzheimer's disease. Nature 2016; 537: 50–56.

Selkoe

Hardy

. The amyloid hypothesis of Alzheimer's disease at 25 years. EMBO Mol Med 2016; 8: 595–608.

Veitch

Weiner

Aisen

, et al. Using the Alzheimer's Disease Neuroimaging Initiative to improve early detection, diagnosis, and treatment of Alzheimer's disease. Alzheimers Dement 2022; 18: 824–857.

Behl

Kaur

Sehgal

, et al. The link between Alzheimer's disease and stroke: a detrimental synergism. Ageing Res Rev 2024; 99: 102388.

Font

Krupinski

Arboix

. Antithrombotic medication for cardioembolic stroke prevention. Stroke Res Treat 2011; 2011: 607852.

10.

Rivard

Friberg

Conen

, et al. Atrial fibrillation and dementia: a report from the AF-SCREEN International Collaboration. Circulation 2022; 145: 392–409.

11.

Chi

Chien

, et al. Alzheimer disease and risk of stroke: a population-based cohort study. Neurology 2013; 80: 705–711.

12.

Zhou

Wang

, et al. Association between stroke and Alzheimer's disease: systematic review and meta-analysis. J Alzheimers Dis 2015; 43: 479–489.

13.

Shen

Nie

Xiang

, et al. The relationship between Alzheimer's disease and intracerebral hemorrhage based on Mendelian randomization. J Alzheimers Dis 2025; 104: 1154–1166.

14.

Lee

Liu

Chang

, et al. The impact of antiplatelet use on the risk of intracerebral hemorrhage in patients with Alzheimer's disease: a nationwide cohort study. J Alzheimers Dis 2020; 73: 297–306.

15.

Yang

Chou

, et al. Longitudinal neuropsychological outcome in Taiwanese Alzheimer's disease patients treated with medication. Curr Alzheimer Res 2018; 15: 474–481.

16.

Liu

Hung

Chuang

, et al. Incorporating development stratification of Taiwan townships into sampling design of large scale health interview survey. J Health Manag 2006; 4: 1–22.

17.

Sacco

Kasner

Broderick

, et al. An updated definition of stroke for the 21st century: a statement for healthcare professionals from the American Heart Association/American Stroke Association. Stroke 2013; 44: 2064–2089.

18.

Austin

. An introduction to propensity score methods for reducing the effects of confounding in observational studies. Multivariate Behav Res 2011; 46: 399–424.

19.

Richardson

Kaufman

. Estimation of the relative excess risk due to interaction and associated confidence bounds. Am J Epidemiol 2009; 169: 756–760.

20.

Flaherty

Kissela

Woo

, et al. The increasing incidence of anticoagulant-associated intracerebral hemorrhage. Neurology 2007; 68: 116–121.

21.

, et al. Risk of intracranial hemorrhage with direct oral anticoagulants: a systematic review and meta-analysis of randomized controlled trials. J Neurol 2022; 269: 664–675.

22.

Chan

Lee

, et al. Effectiveness, safety, and major adverse limb events in atrial fibrillation patients with concomitant diabetes mellitus treated with non-vitamin K antagonist oral anticoagulants. Cardiovasc Diabetol 2020; 19: 63.

23.

Chang

Chou

Yeh

, et al. Association between use of non-vitamin K oral anticoagulants with and without concurrent medications and risk of major bleeding in nonvalvular atrial fibrillation. JAMA 2017; 318: 1250–1259.

24.

Ellis

Olichney

Thal

, et al. Cerebral amyloid angiopathy in the brains of patients with Alzheimer's disease: the CERAD experience, Part XV. Neurology 1996; 46: 1592–1596.

25.

Honig

Sabbagh

van Dyck

, et al. Updated safety results from phase 3 lecanemab study in early Alzheimer's disease. Alzheimers Res Ther 2024; 16: 105.

26.

Okumura

Akao

Yoshida

, et al. Low-dose edoxaban in very elderly patients with atrial fibrillation. N Engl J Med 2020; 383: 1735–1745.

27.

Boulanger

Poon

Wild

, et al. Association between diabetes mellitus and the occurrence and outcome of intracerebral hemorrhage. Neurology 2016; 87: 870–878.

28.

Adelborg

Szepligeti

Sundboll

, et al. Risk of stroke in patients with heart failure: a population-based 30-year cohort study. Stroke 2017; 48: 1161–1168.

29.

Vanent

Leasure

Acosta

, et al. Association of chronic kidney disease with risk of intracerebral hemorrhage. JAMA Neurol 2022; 79: 911–918.

30.

Falanga

Russo

Milesi

. The coagulopathy of cancer. Curr Opin Hematol 2014; 21: 423–429.

31.

Park

Ranganathan

. Neoplastic and paraneoplastic vasculitis, vasculopathy, and hypercoagulability. Rheum Dis Clin North Am 2011; 37: 593–606.

32.

Liu

, et al. Diabetic vascular diseases: molecular mechanisms and therapeutic strategies. Signal Transduct Target Ther 2023; 8: 152.

33.

Ziaeian

Fonarow

. Epidemiology and aetiology of heart failure. Nat Rev Cardiol 2016; 13: 368–378.

34.

Six

Flissi

Lenglet

, et al. Uremic toxins and vascular dysfunction. Toxins (Basel) 2020; 12: 404.

35.

Duranton

Cohen

De Smet

, et al. Normal and pathologic concentrations of uremic toxins. J Am Soc Nephrol 2012; 23: 1258–1270.

36.

Boccardo

Remuzzi

Galbusera

. Platelet dysfunction in renal failure. Semin Thromb Hemost 2004; 30: 579–589.

37.

Sprick

Nocera

Hajjar

, et al. Cerebral blood flow regulation in end-stage kidney disease. Am J Physiol Renal Physiol 2020; 319: F782–F791.

38.

Lee

Jairam

, et al. Nationwide patterns of hemorrhagic stroke among patients hospitalized with brain metastases: influence of primary cancer diagnosis and anticoagulation. Sci Rep 2020; 10: 10084.

39.

Guo

Osouli

Shahripour

. Review of cerebral radiotherapy-induced vasculopathy in pediatric and adult patients. Adv Biol (Weinh) 2023; 7: e2300179.

40.

Mallikarjun

Eldaya

Miller-Thomas

, et al. Good gone bad: complications of chemotherapy, immunotherapy, and radiotherapy on the CNS. Curr Probl Diagn Radiol 2024; 53: 133–149.

41.

Kagan

Popper

Rhoads

. Factors related to stroke incidence in Hawaii Japanese men. The Honolulu Heart Study. Stroke 1980; 11: 14–21.

42.

Ueshima

Iida

Shimamoto

, et al. Multivariate analysis of risk factors for stroke. Eight-year follow-up study of farming villages in Akita, Japan. Prev Med 1980; 9: 722–740.

43.

Sun

Clarke

Bennett

, et al. Causal associations of blood lipids with risk of ischemic stroke and intracerebral hemorrhage in Chinese adults. Nat Med 2019; 25: 569–574.

44.

Zhang

Patel

Horibe

, et al. Cholesterol, coronary heart disease, and stroke in the Asia Pacific region. Int J Epidemiol 2003; 32: 563–572.

45.

Paciaroni

Agnelli

Giustozzi

, et al. Risk factors for intracerebral hemorrhage in patients with atrial fibrillation on non-vitamin K antagonist oral anticoagulants for stroke prevention. Stroke 2021; 52: 1450–1454.

46.

Goldstein

Toth

Dearborn-Tomazos

, et al. Aggressive LDL-C lowering and the brain: impact on risk for dementia and hemorrhagic stroke: a scientific statement from the American Heart Association. Arterioscler Thromb Vasc Biol 2023; 43: e404–e442.

47.

Boehme

Esenwa

Elkind

. Stroke risk factors, genetics, and prevention. Circ Res 2017; 120: 472–495.

48.

Kittner

Sekar

Comeau

, et al. Ethnic and racial variation in intracerebral hemorrhage risk factors and risk factor burden. JAMA Netw Open 2021; 4: e2121921.

49.

Larsson

Chen

Gill

, et al. Risk factors for intracerebral hemorrhage: genome-wide association study and Mendelian randomization analyses. Stroke 2024; 55: 1582–1591.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.03 MB