Association of Antithrombotic Drug Use With Incident Intracerebral Hemorrhage Location

Abstract

Background and Objectives

Few population-based studies have assessed associations between the use of antithrombotic (platelet antiaggregant or anticoagulant) drugs and location-specific risks of spontaneous intracerebral hemorrhage (s-ICH). In this study, we estimated associations between antithrombotic drug use and the risk of lobar vs nonlobar incident s-ICH.

Methods

Using Danish nationwide registries, we identified cases in the Southern Denmark Region of first-ever s-ICH in patients aged 50 years or older between 2009 and 2018. Each verified case was classified as lobar or nonlobar s-ICH and matched to controls in the general population by age, sex, and calendar year. Prior antithrombotic use was ascertained from a nationwide prescription registry. We calculated odds ratios (aORs) for associations between the use of clopidogrel, aspirin, direct oral anticoagulants (DOACs) or vitamin K antagonists (VKA), and lobar and nonlobar ICH in conditional logistic regression analyses that were adjusted for potential confounders.

Results

A total of 1,040 cases of lobar (47.9% men, mean age [SD] 75.2 [10.7] years) and 1,263 cases of nonlobar s-ICH (54.2% men, mean age 73.6 [11.4] years) were matched to 41,651 and 50,574 controls, respectively. A stronger association with lobar s-ICH was found for clopidogrel (cases: 7.6%, controls: 3.5%; aOR 3.46 [95% CI 2.45–4.89]) vs aspirin (cases: 22.9%, controls: 20.4%; aOR 2.14 [1.74–2.63; p = 0.019). Corresponding estimates for nonlobar s-ICH were not different between clopidogrel (cases: 5.4%, controls: 3.4%; aOR 2.44 [1.71–3.49]) and aspirin (cases: 20.7%, controls: 19.2%; aOR 1.77 [1.47–2.15]; p = 0.12). VKA use was associated with higher odds of both lobar (cases: 14.3%, controls: 6.1%; aOR 3.66 [2.78–4.80]) and nonlobar (cases: 15.4%, controls: 5.5%; aOR 4.62 [3.67–5.82]) s-ICH. The association of DOAC use with lobar s-ICH (cases: 3.5%, controls: 2.7%; aOR 1.66 [1.02–2.70]) was weaker than that of VKA use (p = 0.006). Corresponding estimates for nonlobar s-ICH were not different between DOACs (cases: 5.1%, controls: 2.4%; aOR 3.44 [2.33–5.08]) and VKAs (p = 0.20).

Discussion

Antithrombotics were associated with higher risks of s-ICH, but the strength of the associations varied by s-ICH location and drug, which may reflect differences in the cerebral microangiopathies associated with lobar vs nonlobar hemorrhages and the mechanisms of drug action.

Introduction

Antithrombotic drugs are effective in preventing thromboembolic events but carry a risk of intracerebral hemorrhage (ICH).^1-4 Differences in the pathology underlying ICH based on location may affect the association between antithrombotic use and incident ICH. Supratentorial deep (nonlobar) ICH is commonly due to perforator vessel arteriolosclerosis, often attributed to hypertension.⁵ Although moderate-severe arteriolosclerosis can also underlie lobar ICH,^6,7 in 42% of cases, it coexists with cerebral amyloid angiopathy (CAA), which is the only underlying microangiopathy in 15% of lobar ICH cases.⁶ A meta-analysis of neuroimaging findings in patients with oral anticoagulant (OAC)-associated ICH vs non–OAC-associated ICH reported pooled risk ratios of 1.02 [95% confidence interval (CI) 0.89–1.17] for lobar ICH vs 0.94 [95% CI 0.88–1.00] for deep hemorrhage.⁸ Subsequent studies not included in the meta-analysis reported higher odds of incident ICH with use of anticoagulants for both lobar and nonlobar locations.^9,10 No population-based study has yet reported risk estimates for the association between direct oral anticoagulant (DOAC) use and incident ICH by location. Studies investigating an association between platelet antiaggregant use and risk of lobar vs nonlobar ICH have reported neutral results,^10-12 or a higher risk of lobar ICH,¹³ and none have provided risk estimates for clopidogrel. We performed this population-based study to further explore the association of OAC and platelet antiaggregant use with the risk of incident ICH by location.

Methods

Design

This population-based case-control study was nested within the Region of Southern Denmark (RSD) using data from nationwide registries (see eMethods).^14-17 We used the unique 10-digit personal identifier assigned to all Danish residents at birth or on immigration to link data across registries. The Strengthening and Reporting of Observational Studies in Epidemiology reporting guidelines were followed.

Setting

Health care is paid by taxes and free of charge at the time of use for all residents in Denmark. When stroke is suspected, patients are evaluated using established pathways with rapid access to neuroimaging. Patients are also subsequently transferred or admitted to stroke units for specialized care. One of the 5 regions in Denmark, the RSD has a population of 1.2 million and in regard to sociodemographic and health related characteristics, is representative of the population in Denmark.¹⁸ The RSD has 5 stroke units and a single neurosurgical department at a university hospital.

Standard Protocol Approvals, Registration, and Patient Consents

Authorities in the RSD approved this study. Data were pseudonymized,¹⁹ and informed consent was waived.

Data Availability

Danish law prohibits the sharing of or the authors granting access to the data used for this study.

Source Population

For this study, the source population was all residents of the RSD aged 50 years or older in 2009–2018. We chose this age threshold because it is a component of the Boston criteria for CAA²⁰ and due to the higher likelihood of secondary causes of ICH in younger populations (i.e., trauma, hemorrhagic transformation of an ischemic stroke, intracranial venous sinus thrombosis, aneurysmal subarachnoid hemorrhage-related ICH, vascular malformations, or neoplasms).

Cases

Cases were identified as described previously.^21-23 We used both the Danish Stroke Registry and the National Danish Patient Registry to identify all individuals with first-ever spontaneous ICH (s-ICH) in the RSD during the study period.²¹ We excluded patients with secondary causes for ICH. The physicians who classified the hemorrhage location verified the diagnosis of s-ICH based on both brain imaging reports and discharge summaries using a validated method described elsewhere.²⁴ The index date was the date of symptom onset of s-ICH according to medical records, or if unavailable, the date of hospital admission. Data that were collected for this project on the patients with s-ICH were linked with information collected prospectively on all RSD residents from 4 nationwide registries^14-17 using the 10-digit personal identifier assigned to all Danish residents at birth or on immigration.¹⁷

Based on criteria modified from a previous population-based study,²⁵ s-ICH location was classified as nonlobar if a single supratentorial deep ICH (i.e., located in the basal ganglia, internal or external capsule, or thalamus), single infratentorial ICH (i.e., located in the brain stem or cerebellum), or multiple nonlobar ICHs (supratentorial deep or infratentorial) were present; lobar classification included all other ICHs. The nonlobar hemorrhages were subclassified as either “deep” (i.e., subcortical or pontine) or “cerebellar” (a location occasionally associated with CAA²⁶). Isolated intraventricular hemorrhages (IVH) were included in overall analyses of ICH but excluded from location-specific analyses. IVH can be related to small vessel disease (SVD).²⁷ Patients with ICH with mortality within 30 days of the index date (fatal ICH) were identified based on data in the vital status data from the Danish Civil Registry.

Controls

We used a high control:case sampling ratio (40:1) to attain maximum statistical precision with the available cases.²⁸ For each case, we identified controls without ICH from the source population in the Danish Civil Registration System.¹⁷ We used risk-set sampling to match controls to an index case (i.e., the date of ICH onset of their corresponding case) by birth year and sex. Nationwide data on controls were retrieved from the same registry as data on the cases. Cases were eligible as controls before the occurrence of first-ever s-ICH. Thus, the odds ratio (OR) was ensured to be an unbiased estimate of the incidence rate ratio.

Exposure to Antithrombotic Drugs

Low-dose aspirin (only available as ≤150 mg doses in Denmark and ≈90% of total sales prescribed²⁹), clopidogrel, prasugrel, ticagrelor, dipyridamole, apixaban, rivaroxaban, edoxaban, dabigatran, and warfarin were available for clinical use in Denmark during the study period. We determined the exposure to any of these drugs up to the index date, for both cases and controls, based on data in the prescription registry (codes in eTable 1). Owing to very limited use of prasugrel, ticagrelor, and dipyridamole, we did not include data on these drugs in the analyses. We classified exposure to the prescribed antithrombotic drugs on the index date as current use (i.e., prescription supply ended 0–30 days before the index date), past use (supply ended 31–365 days before the index date), and nonuse (no supply in the 365 days before the index date). Drug exposure definitions can be found in the eMethods.

Supplemental Analyses

We performed separate analyses for ICH located in deep and cerebellar locations. Large hemorrhages at a location that could not be classified (n = 282, 10.6%) were assessed separately and not included in the primary analyses. We repeated the primary analyses among s-ICH cases only (i.e., direct comparison of lobar and nonlobar cases) and compared the antithrombotic drug exposure for nonlobar vs lobar s-ICH. For this analysis, we included age, sex, and calendar time in addition to the confounders from the primary analysis. This approach was chosen to directly ascertain the variability in the association between antithrombotic drug use and s-ICH location (lobar vs nonlobar). Further analyses are outlined in eMethods.

Brain CT Substudy

In a substudy, we included patients aged 55 years or older in 2015–2018 with an available index brain CT (compared with ≥50 years in the main analyses) because this part of the study was completed before the revised Boston criteria version 2.0 for CAA were published.²⁰ We excluded patients with isolated intraventricular ICH or large unclassifiable ICH (Figure 1). CT scans were re-evaluated, for hematoma location, masked to clinical data (using CHARTS³⁰). We used the above classifications for hematoma location. We further classified lobar cases into low/intermediate probability of CAA and high probability of CAA using the simplified Edinburgh criteria (i.e., brain CT features alone, without the ApoE genotype).⁶ Lobar cases were matched to controls from the general population as in the main analysis. We also repeated the primary analysis with a direct comparison of lobar vs nonlobar cases based on the results of this substudy. An additional analysis is outlined in eMethods.

Figure. Study Flowchart.

Covariates

We obtained data from the Danish Stroke Registry¹⁴ and the Danish National Prescription Registry¹⁵ (Prescription Registry) and the Danish National Patient Registry¹⁶ for both patients with incident s-ICH and controls in regard to risk factors, potential confounders, and proxies for confounders (eMethods, eTable 1).

Statistical Analysis

We performed descriptive analyses in which categorical data were presented as frequency counts and percentages. Continuous data were presented as means and standard deviations (SDs). ORs were calculated by conditional logistic regression as a measure of the overall risk of s-ICH and the risk of s-ICH by location in nested case-control analyses of the association with antithrombotic drug use (vs nonuse). All analyses were adjusted for the following covariates: hypertension, diabetes, history of ischemic stroke, chronic kidney disease, chronic hepatic disease, ischemic heart disease, peripheral artery disease, heart failure, cancer, dementia, disorders indicative of high alcohol consumption, history of chronic obstructive pulmonary disease as a proxy for smoking, and current use of nonsteroidal anti-inflammatory drugs, selective serotonin reuptake inhibitors, proton pump inhibitors), hormone replacement therapy, oral corticosteroid drugs, or drugs with antihypertensive effects (one covariate for current use of any of the following: loop-diuretics, nonloop diuretics, beta-blockers, calcium channel blockers, ACE inhibitors, and angiotensin II receptor blockers). In all VKA/DOAC analyses, we also adjusted for platelet antiaggregant use. We chose the covariates based on currently known risk factors and potential confounders for ICH. The minimum strength of association that an unmeasured confounder requires for both exposure and outcome to fully explain the observed association is represented by the E-value. Thus, as a measure of the potential effect of unmeasured confounding, E-values were calculated for the main results.³¹ Higher E-values were indicative of the observed association being less likely to be explained by unmeasured confounding. Statistical analyses are further outlined in eMethods. The two-sampled Wald test was used to evaluate differences in the strength of the association between subgroups. A two-sided limit of 0.05 was used as the significance threshold.

Literature Search

We conducted a systematic literature search to identify studies providing risk estimates for any antithrombotic agent and location-specific ICH as outlined in eMethods, eFigure 1, and eTable 2. Abstracted data from the articles that met eligibility criteria are presented in eTable 3.

Results

In the main analyses, we included 1,040 patients with lobar s-ICH (47.9% men, 52.1% women; mean age, 75.2 [SD 10.7] years) and 1,263 patients with nonlobar s-ICH (54.2% men, 45.8% women; mean age, 73.6 [SD 11.4] years), who were matched to 41,651 and 50,574 controls, respectively. Table 1 summarizes the clinical characteristics of the cases and controls.

Table 1.

Characteristics of Cases With Intracerebral Hemorrhage by Location and Their Respective Age and Sex-Matched General Population Controls

Characteristics, No. (%)	Lobar s-ICH location			Nonlobar s-ICH location
	Cases (n = 1,040)	Controls (n = 41,651)	ORa (95% CI)	Cases (n = 1,263)	Controls (n = 50,574)	ORa (95% CI)
Sex
Male	498 (47.9)	19,955 (47.9)	N/Ab	684 (54.2)	27,343 (54.1)	N/Ab
Female	542 (52.1)	21,696 (52.1)	N/Ab	579 (45.8)	23,231 (45.9)	N/Ab
Age, mean ± SD, y	75.2 ± 10.7	75.1 ± 10.7	N/Ab	73.6 ± 11.4	73.5 ± 11.4	N/Ab
Age, categories, y
50–74	485 (46.6)	19,538 (46.9)	N/Ab	673 (53.3)	27,112 (53.6)	N/Ab
75–84	356 (34.2)	14,166 (34.0)	N/Ab	368 (29.1)	14,544 (28.8)	N/Ab
85+	199 (19.1)	7,947 (19.1)	N/Ab	222 (17.6)	8,918 (17.6)	N/Ab
Index year
2009–2013	486 (46.7)	19,476 (46.8)	N/Ab	560 (44.3)	22,450 (44.4)	N/Ab
2014–2018	554 (53.3)	22,175 (53.2)	N/Ab	703 (55.7)	28,124 (55.6)	N/Ab
30-d fatality	328 (31.5)	N/Ac	N/Ac	389 (30.8)	N/Ac	N/Ac
Medication—current used
Antihypertensive drugs	558 (53.7)	24,102 (57.9)	0.74 (0.61–0.90)	692 (54.8)	27,809 (55.0)	0.70 (0.58–0.83)
Statins	312 (30.0)	12,114 (29.1)	1.06 (0.92–1.22)	370 (29.3)	14,305 (28.3)	1.08 (0.95–1.23)
Nonsteroidal anti-inflammatory drugs	74 (7.1)	2,478 (5.9)	1.19 (0.94–1.52)	113 (8.9)	3,063 (6.1)	1.50 (1.23–1.83)
Selective serotonin reuptake inhibitors	147 (14.1)	3,282 (7.9)	1.99 (1.66–2.39)	136 (10.8)	3,717 (7.3)	1.54 (1.28–1.85)
Proton pump inhibitors	191 (18.4)	6,868 (16.5)	1.14 (0.97–1.34)	230 (18.2)	7,807 (15.4)	1.21 (1.04–1.40)
Oral corticosteroids	31 (3.0)	1,391 (3.3)	0.89 (0.62–1.28)	36 (2.9)	1,554 (3.1)	0.94 (0.67–1.32)
Hormone replacement therapy (women only)	44 (4.2)	1843 (4.4)	0.97 (0.71–1.32)	48 (3.8)	2017 (4.0)	0.95 (0.70–1.28)
Comorbidity
Hypertension	678 (65.2)	26,599 (63.9)	1.07 (0.93–1.22)	870 (68.9)	30,731 (60.8)	1.50 (1.32–1.71)
Previous ischemic stroke	131 (12.6)	2063 (5.0)	2.81 (2.32–3.41)	181 (14.3)	2,411 (4.8)	3.43 (2.90–4.05)
Diabetes	129 (12.4)	5,243 (12.6)	0.98 (0.81–1.19)	174 (13.8)	6,305 (12.5)	1.12 (0.95–1.32)
Chronic kidney disease	33 (3.2)	1,020 (2.4)	1.31 (0.92–1.87)	45 (3.6)	1,069 (2.1)	1.72 (1.27–2.33)
Chronic hepatic disease	35 (3.4)	547 (1.3)	2.63 (1.85–3.72)	38 (3.0)	608 (1.2)	2.55 (1.83–3.56)
High alcohol consumption	68 (6.5)	1,463 (3.5)	1.96 (1.52–2.53)	106 (8.4)	2000 (4.0)	2.29 (1.86–2.82)
Chronic obstructive pulmonary disease	318 (30.6)	12,451 (29.9)	1.03 (0.90–1.18)	409 (32.4)	14,730 (29.1)	1.17 (1.04–1.32)
Ischemic heart disease	183 (17.6)	6,915 (16.6)	1.07 (0.91–1.27)	193 (15.3)	8,009 (15.8)	0.95 (0.81–1.12)
Peripheral artery disease	61 (5.9)	2074 (5.0)	1.19 (0.91–1.55)	82 (6.5)	2,352 (4.7)	1.43 (1.14–1.80)
Congestive heart failure	90 (8.7)	2,486 (6.0)	1.51 (1.21–1.89)	83 (6.6)	2,840 (5.6)	1.18 (0.94–1.49)
Cancer	188 (18.1)	6,239 (15.0)	1.26 (1.07–1.48)	159 (12.6)	6,884 (13.6)	0.91 (0.77–1.08)
Dementia	81 (7.8)	1,496 (3.6)	2.38 (1.87–3.04)	63 (5.0)	1,580 (3.1)	1.67 (1.28–2.18)
Atrial fibrillatione	203 (19.5)	4,675 (11.2)	2.00 (1.70–2.35)	259 (20.5)	5,145 (10.2)	2.41 (2.08–2.79)
Venous thromboembolisme	56 (5.4)	1,509 (3.6)	1.52 (1.15–2.00)	68 (5.4)	1767 (3.5)	1.57 (1.22–2.02)
Mechanical heart valvee	6 (0.6)	19 (0.0)	12.88 (5.10–32.52)	5 (0.4)	52 (0.1)	3.86 (1.54–9.66)
Gastrointestinal bleede	100 (9.6)	2,695 (6.5)	1.54 (1.25–1.90)	105 (8.3)	3,074 (6.1)	1.40 (1.14–1.72)
Coagulopathye	5 (0.5)	97 (0.2)	2.07 (0.84–5.08)	7 (0.6)	123 (0.2)	2.28 (1.06–4.89)

Abbreviations: OR = odds ratio; s-ICH = spontaneous intracerebral hemorrhage.

^aAdjusted for sex, age, and calendar year by design (cases matched to controls on these variables).

^bNot applicable because cases and controls were matched on age, sex, and index year by design.

^cNot applicable—data only apply to cases.

^dCurrent use of drug defined as most recent prescription supply covering index date or ending up to 30 d before index data.

^eOnly included for descriptive purposes, not included in models as covariates.

Comparison of Antithrombotic Drug Associations Within the Same s-ICH Location

The association between single use of clopidogrel (i.e., with no concurrent use of another platelet antiaggregant) and risk of lobar s-ICH (aOR 3.46 [95% CI 2.45–4.89]) was stronger than for aspirin (aOR 2.14 [1.74–2.63]) (p = 0.019). Corresponding estimates for nonlobar s-ICH were not different for single use of clopidogrel (aOR 2.44 [1.71–3.49]) vs aspirin (aOR 1.77 [1.47–2.15]).

The association with risk of lobar s-ICH was weaker for DOAC use (aOR 1.66 [1.02–2.70]) than for VKA (aOR 3.66 [2.78–4.80]) (p = 0.006). Corresponding estimates for nonlobar s-ICH were not different for DOAC (aOR 3.44 [2.33–5.08]) vs VKA (aOR 4.62 [3.67–5.82]). Comparison of the associations in patients who were prescribed these anticoagulants, but with no concurrent platelet antiaggregant use, produced similar results (Table 2).

Table 2.

Association Between Use of Oral Antithrombotic Drugs and the Risk of Intracerebral Hemorrhage by Location

Use of antithrombotic drugs	Lobar s-ICH location					Nonlobar s-ICH location
	Cases, no. (%) (n = 1,040)	Controls, no. (%) (n = 41,651)	ORa (95% CI)	Adjusted ORb (95% CI)	E-valuec	Cases, no. (%) (n = 1,263)	Controls, no. (%) (n = 50,574)	ORa (95% CI)	Adjusted ORb (95% CI)	E-valuec
Nonuse of any antithromboticd	468 (45.0)	26,194 (62.9)	1 (reference)	1 (reference)		612 (48.5)	32,956 (65.2)	1 (reference)	1 (reference)
Current usee,f
Platelet antiaggregant single-agent or dual-agent use	337 (32.4)	10,333 (24.8)	1.98 (1.70–2.31)	2.47 (2.05–2.97)	4.37	343 (27.2)	11,855 (23.4)	1.76 (1.52–2.03)	1.88 (1.57–2.24)	3.16
Platelet antiaggregant single-agent use	317 (30.5)	9,941 (23.9)	1.93 (1.65–2.25)	2.39 (1.98–2.89)	4.21	330 (26.1)	11,419 (22.6)	1.75 (1.51–2.02)	1.84 (1.54–2.19)	3.07
Low-dose aspirin	238 (22.9)	8,482 (20.4)	1.70 (1.44–2.01)	2.14 (1.74–2.63)	3.70	262 (20.7)	9,722 (19.2)	1.60 (1.37–1.87)	1.77 (1.47–2.15)	2.95
Clopidogrel	79 (7.6)	1,459 (3.5)	3.34 (2.52–4.42)	3.46 (2.45–4.89)	6.38	68 (5.4)	1,697 (3.4)	2.56 (1.91–3.44)	2.44 (1.71–3.49)	4.32
Dual-platelet antiaggregant agent use (aspirin and clopidogrel)	20 (1.9)	392 (0.9)	2.98 (1.85–4.82)	3.32 (1.83–6.03)	6.09	13 (1.0)	436 (0.9)	1.82 (1.02–3.23)	1.98 (1.00–3.94)	3.38
Oral anticoagulant use	185 (17.8)	3,670 (8.8)	2.78 (2.22–3.47)	3.16 (2.46–4.05)	5.77	260 (20.6)	4,027 (8.0)	3.91 (3.25–4.71)	4.23 (3.42–5.21)	7.92
Vitamin K antagonist	149 (14.3)	2,530 (6.1)	3.10 (2.42–3.95)	3.66 (2.78–4.80)	6.77	195 (15.4)	2,806 (5.5)	4.18 (3.41–5.13)	4.62 (3.67–5.82)	8.72
Direct oral anticoagulantg	36 (3.5)	1,140 (2.7)	1.69 (1.09–2.64)	1.66 (1.02–2.70)	2.70	65 (5.1)	1,221 (2.4)	3.34 (2.33–4.79)	3.44 (2.33–5.08)	6.34
Oral anticoagulants and no concurrent platelet antiaggregant use	133 (12.8)	3,064 (7.4)	2.61 (2.08–3.28)	2.98 (2.29–3.87)	5.41	204 (16.2)	3,311 (6.5)	3.68 (3.04–4.44)	3.92 (3.16–4.86)	7.30
Vitamin K antagonist and no concurrent platelet antiaggregant	NRh	2043 (4.9)	2.96 (2.30–3.81)	3.54 (2.66–4.70)	6.53	150 (11.9)	2,225 (4.4)	3.94 (3.20–4.86)	4.28 (3.37–5.42)	8.02
Direct oral anticoagulants and no concurrent platelet antiaggregant	NRh	1,021 (2.5)	1.61 (1.02–2.55)	1.54 (0.93–2.55)	2.45	54 (4.3)	1,086 (2.1)	3.19 (2.20–4.64)	3.31 (2.22–4.95)	6.08
Oral anticoagulant and concurrent platelet antiaggregant	52 (5.0)	606 (1.5)	6.50 (1.74–24.31)	6.65 (1.52–29.21)	12.79	56 (4.4)	716 (1.4)	7.91 (2.85–21.94)	5.45 (1.76–16.91)	10.37
Vitamin K antagonist and concurrent platelet antiaggregant	NRh	487 (1.2)	6.06 (1.08–34.07)	9.36 (1.56–56.15)	18.21	45 (3.6)	581 (1.1)	4.94 (1.06–23.01)	5.27 (1.04–26.62)	10.01
Direct oral anticoagulant and concurrent platelet antiaggregant	NRh	119 (0.3)	NEi	NEi		11 (0.9)	135 (0.3)	10.44 (2.68–40.64)	5.57 (1.18–26.31)	10.62

Abbreviations: NE = not estimated due to sparse data; NR = not reported to preserve anonymity in view of small cell counts in this or related cells of these tables; OR = odds ratio; s-ICH = spontaneous intracerebral hemorrhage.

The p value calculated using the Wald test was statistically significant for the following comparisons: (a) same agent/agent class across locations (i.e., lobar vs nonlobar): platelet antiaggregants single or dual drugs (p = 0.038), platelet antiaggregants agent single use (p = 0.047), DOAC use (p = 0.022), DOAC use and no concurrent platelet antiaggregant use (p = 0.020) and (b) different agent/agent class in same location (i.e., lobar or nonlobar): lobar location low-dose aspirin vs clopidogrel (p = 0.019), lobar VKA vs DOAC (p = 0.006), lobar location VKA and no concurrent platelet antiaggregant agent vs DOAC and no concurrent platelet antiaggregant (p = 0.005).

^aAdjusted for age, sex, and calendar year by design.

^bAdjusted for age, sex, and calendar period and for the following, based on register data: hypertension, previous ischemic stroke, diabetes, chronic renal insufficiency, chronic hepatic disease, heart failure, ischemic heart disease, peripheral artery disease, cancer, dementia, disorders indicative of high alcohol consumption, chronic obstructive pulmonary disease (as a marker of smoking), and the use of antihypertensive drugs, statins, nonsteroidal anti-inflammatory drugs, selective serotonin reuptake inhibitors, proton pump inhibitors, hormone replacement therapy (women only), and oral corticosteroid drugs. In analyses of oral anticoagulants, individual variables for use of aspirin, and clopidogrel, were adjusted for. In analyses of aspirin, individual variables for use of clopidogrel were adjusted for, and vice versa. Analyses of platelet antiaggregants, aspirin, and clopidogrel did not include patients with use of oral anticoagulants within the past 12 mo before index date, see footnote f below.

^cMetric for the potential effect of unmeasured confounding.

^dNo use of any antithrombotic agent in the 12 mo before index date was the reference group for all analyses in the table.

^eBased on the most recent treatment episode before the index date (date of diagnosis for cases and date of selection for controls), exposure categorized as current use if treatment episode ended 0–30 d before index date.

^fCurrent users of both an oral anticoagulant and one or more platelet antiaggregants were classified exclusively as oral anticoagulant users (and as vitamin K antagonist/direct oral anticoagulant users in analyses of these drug classes).

^gRivaroxan 2.5 mg was used by no cases and less than 5 controls (exact number not reported to preserve anonymity) and rivaroxaban 10 mg was used by less than 5 cases (exact number not reported to preserve anonymity) and 15 controls.

^hNot reported to preserve anonymity.

ⁱNot estimated due to small numbers.

Brain CT Substudy Results

In the brain CT substudy, platelet antiaggregant use was more strongly associated with high probability of CAA (aOR 3.90 [95% CI 2.41–6.30]) than low/medium probability of CAA (aOR 1.88 [1.33–2.65]; p = 0.018). In corresponding analyses by type of platelet antiaggregant, there were differences in the strength of the associations for low-dose aspirin (high probability of CAA: aOR 3.61 [2.11–6.18]; vs low/intermediate probability of CAA: aOR 1.44 [0.94–2.20] p = 0.010), but not clopidogrel (Table 3). Of the antithrombotic drugs assessed, DOACs had the weakest association with both high probability of CAA (aOR 2.18 [0.84–5.62]) and low/intermediate probability of CAA (aOR 1.86 [1.02–3.38]) bleeding. For corresponding estimates for the association of VKA, see Table 3.

Table 3.

Association Between Use of Oral Antithrombotic Drugs and the Risk of Cerebral Amyloid Angiopathy-Related Intracerebral Hemorrhage According to the Simplified Edinburgh Criteria

Use of antithrombotic drugs	Lobar ICH			Lobar s-ICH—low/intermediate CAA probability (only FLP, only cSAH, or no FLP or cSAH)			Lobar s-ICH—high CAA probability (FLP and cSAH)			p Valueb
	Cases, no. (%) (n = 467)	Controls, no. (%) (n = 18,679)	Adjusted ORa (95% CI)	Cases, no. (%) (n = 319)	Controls, no. (%) (n = 12,760)	Adjusted ORa (95% CI)	Cases, no. (%) (n = 148)	Controls, no. (%) (n = 5,919)	Adjusted ORa (95% CI)
No use of any antithromboticc	186 (39.8)	10,992 (58.8)	1 (reference)	134 (42.0)	7,480 (58.6)	1 (reference)	52 (35.1)	3,512 (59.3)	1 (reference)
Current used,e
Platelet antiaggregantsf	142 (30.4)	4,515 (24.2)	2.43 (1.84–3.22)	87 (27.3)	3,086 (24.2)	1.88 (1.33–2.65)	55 (37.2)	1,429 (24.1)	3.90 (2.41–6.30)	0.018
Low-dose aspirinf	83 (17.8)	3,373 (18.1)	2.04 (1.47–2.83)	44 (13.8)	2,308 (18.1)	1.44 (0.94–2.20)	39 (26.4)	1,065 (18.0)	3.61 (2.11–6.18)	0.010
Clopidogrelf	59 (12.6)	1,142 (6.1)	3.74 (2.51–5.58)	43 (13.5)	778 (6.1)	3.45 (2.14–5.57)	16 (10.8)	364 (6.1)	4.27 (1.97–9.26)	0.65
Oral anticoagulants	99 (21.2)	2,290 (12.3)	3.31 (2.38–4.61)	73 (22.9)	1,596 (12.5)	3.41 (2.30–5.03)	26 (17.6)	694 (11.7)	3.26 (1.74–6.09)	0.90
VKA	63 (13.5)	1,262 (6.8)	4.17 (2.82–6.19)	47 (14.7)	894 (7.0)	4.83 (3.05–7.66)	16 (10.8)	368 (6.2)	3.17 (1.44–6.96)	0.37
DOAC	36 (7.7)	1,028 (5.5)	1.95 (1.18–3.22)	26 (8.2)	702 (5.5)	1.86 (1.02–3.38)	10 (6.8)	326 (5.5)	2.18 (0.84–5.62)	0.78

Abbreviations: CAA = cerebral amyloid angiopathy; DOAC = direct oral anticoagulant; OAC = oral anticoagulant; OR = odds ratio; s-ICH = spontaneous intracerebral hemorrhage; VKA = vitamin K antagonist.

^aAdjusted for age, sex, and calendar period and for the following, based on register data: hypertension, previous ischemic stroke, diabetes, chronic renal insufficiency, chronic hepatic disease, heart failure, ischemic heart disease, peripheral artery disease, cancer, dementia, disorders indicative of high alcohol consumption, chronic obstructive pulmonary disease (as a marker of smoking), and the use of antihypertensive drugs, statins, nonsteroidal anti-inflammatory drugs, selective serotonin reuptake inhibitors, proton pump inhibitors, hormone replacement therapy (women only), and oral corticosteroid drugs. In analyses of OACs, individual variables for use of aspirin, and clopidogrel, were adjusted for. Analyses of platelet antiaggregants, aspirin, and clopidogrel did not include patients with use of OACs within the past 12 mo before index date, see footnote e below.

^bWald test comparing same agent for high CAA probability vs low/intermediate CAA probability.

^cNo use of any antithrombotic drug in the 12 mo before index date.

^dBased on the most recent treatment episode before the index date (date of diagnosis for cases and date of selection for controls), exposure categorized as current use if treatment episode ended 0–30 d before index date.

^eCurrent users of both an OAC and one or more platelet antiaggregants were classified exclusively as OAC users (and as VKA/DOAC users in analyses of these drug classes).

^fSingle-platelet antiaggregant use only, that is, concurrent dual platelet antiaggregant use not included.

Comparison of Antithrombotic Drug Associations by s-ICH Locations

In comparison with the use of the same antithrombotic drug across the 2 s-ICH locations, platelet antiaggregant use overall (single or dual regimen; p = 0.038) and single-platelet antiaggregant use (p = 0.047) were more strongly associated with lobar s-ICH than for nonlobar s-ICH (Table 2). Although the associations did not differ for VKA use, the association of DOAC use was stronger for nonlobar s-ICH than for lobar s-ICH (p = 0.022).

In the direct comparison of cases in the brain CT substudy, we also found that DOAC use was more strongly associated with a nonlobar location (aOR 0.48 [95% CI 0.25–0.91]; Table 4).

Table 4.

Association Between Use of Oral Antithrombotic Drugs and the Risk of Intracerebral Hemorrhage by Location—Direct Comparison of Cases, Lobar vs Nonlobar (Reference Group)

Use of antithrombotic drugs	Main study
	Lobar, no. (%) (n = 1,040)	Nonlobar, no. (%) (n = 1,263)	ORa (95% CI)	Adjusted ORb (95% CI)
Nonuse of any antithromboticc	468 (45.0)	612 (48.5)	1 (reference)	1 (reference)
Current used,e
Platelet antiaggregantsf	317 (30.5)	330 (26.1)	1.17 (0.95–1.43)	1.31 (1.02–1.69)
Low-dose aspirinf	238 (22.9)	262 (20.7)	1.11 (0.89–1.39)	1.21 (0.92–1.59)
Clopidogrelf	79 (7.6)	68 (5.4)	1.41 (0.98–2.02)	1.67 (1.05–2.65)
Oral anticoagulants	185 (17.8)	260 (20.6)	0.89 (0.70–1.13)	0.82 (0.60–1.13)
Vitamin K antagonist	149 (14.3)	195 (15.4)	0.95 (0.74–1.24)	0.82 (0.58–1.16)
Direct oral anticoagulants	36 (3.5)	65 (5.1)	0.64 (0.41–1.01)	0.75 (0.42–1.36)

Use of antithrombotic drugs	Brain CT substudyg
	Lobar, no. (%) (n = 467)	Nonlobar, no. (%) (n = 572)	ORa (95% CI)	Adjusted ORb (95% CI)
Nonuse of any antithromboticc	186 (39.8)	236 (41.3)	1 (reference)	1 (reference)
Current used,e
Platelet antiaggregantsf	142 (30.4)	142 (24.8)	1.06 (0.77–1.45)	1.07 (0.73–1.59)
Low-dose aspirinf	83 (17.8)	91 (15.9)	0.94 (0.65–1.37)	0.99 (0.63–1.56)
Clopidogrelf	59 (12.6)	51 (8.9)	1.18 (0.76–1.83)	1.22 (0.68–2.19)
Oral anticoagulants	99 (21.2)	164 (28.7)	0.60 (0.43–0.85)	0.71 (0.46–1.10)
Vitamin K antagonist	63 (13.5)	95 (16.6)	0.69 (0.46–1.03)	0.72 (0.43–1.22)
Direct oral anticoagulants	36 (7.7)	69 (12.1)	0.47 (0.29–0.77)	0.48 (0.25–0.91)

Abbreviations: CT = computerized tomography; OR = odds ratio.

^aAdjusted for age, sex, and calendar period.

^bAdjusted for age, sex, and calendar period and for the following, based on register data: hypertension, previous ischemic stroke, diabetes, chronic renal insufficiency, chronic hepatic disease, heart failure, ischemic heart disease, peripheral artery disease, cancer, dementia, disorders indicative of high alcohol consumption, chronic obstructive pulmonary disease (as a marker of smoking), and the use of antihypertensive drugs, statins, nonsteroidal anti-inflammatory drugs, selective serotonin reuptake inhibitors, proton pump inhibitors, hormone replacement therapy (women only), and oral corticosteroid drugs. In analyses of oral anticoagulants, individual variables for use of aspirin, and clopidogrel, were adjusted for. In analyses of direct oral anticoagulants, individual variables for use of vitamin K antagonist were adjusted for, and vice versa. In analyses of aspirin, individual variables for use of clopidogrel were adjusted for, and vice versa. Analyses of platelet antiaggregants, aspirin, and clopidogrel did not include patients with use of oral anticoagulants within the past 12 mo before index date, see footnote e below.

^cNo use of any antithrombotic drug in the 12 mo before index date.

^dBased on the most recent treatment episode before the index date (date of diagnosis for cases and date of selection for controls), exposure categorized as current use if treatment episode ended 0–30 d before index date.

^eCurrent users of both an oral anticoagulant and one or more platelet antiaggregants were classified exclusively as oral anticoagulant users (and as vitamin K antagonist/direct oral anticoagulant users in analyses of these drug-classes).

^fSingle-platelet antiaggregant use only, that is, concurrent dual platelet antiaggregant use not included.

^gIncludes patients aged 55 years or older at onset of ICH admitted to hospital in 2015–2018 with first-ever spontaneous intracerebral hemorrhage. Index brain CTs of these patients were re-evaluated masked to clinical data with hematoma location classified according to the Cerebral Haemorrhage Anatomical RaTing Instrument (CHARTS).

E-values for all comparisons in the main analysis were consistently >2.4 (Table 2).

Supplemental Analyses

In case-control analyses of the associations between antithrombotic drug use and the risk of all s-ICHs, regardless of location (including large unclassifiable, isolated IVH, or unclassifiable due to missing information), the association was weakest with current aspirin use (aOR 2.04 [95% CI 1.79–2.32]) and strongest with VKA use (aOR 4.68 [3.98–5.49]) (eTable 4).

In analyses of nonlobar s-ICH subclassified into deep and cerebellar, the strongest associations were for OACs and cerebellar location (VKA aOR 8.25 [4.95–13.75], DOAC 4.58 [1.91–11.00]). Similar results were found for large unclassifiable s-ICH (eTable 5).

Other analyses were consistent with the main results (eTables 6–10).

Discussion

In this population-based study, VKA and DOAC use was each associated with both lobar and nonlobar s-ICH. For VKAs, the strength of the association was similar in both locations. For DOACs, the s-ICH risk was higher in nonlobar locations than lobar locations, a pattern that was also present among platelet antiaggregant users. Clopidogrel was associated with a higher risk of s-ICH than aspirin, regardless of location. Of all antithrombotic drugs investigated, VKA use was associated with the highest risk of s-ICH, both lobar and nonlobar, whereas DOAC use was associated with the smallest increase in risk of lobar s-ICH, including in an analysis of lobar s-ICH by likelihood of underlying CAA (according to simplified Edinburgh criteria). In the analyses by CAA likelihood, platelet antiaggregants were more strongly associated with high probability of CAA compared with low/intermediate probability of CAA bleeding, an association that was mainly driven by a differential association with aspirin use.

Our results are consistent with 2 studies reporting an association between OAC use and higher risk of nonlobar than lobar ICH.^10,11 In a cohort study, OAC use was associated with a higher risk of nonlobar compared with lobar ICH (adjusted hazard ratio 3.49 [95% CI 1.20–10.2]) vs 1.46 [0.31–6.79], respectively).¹¹ A case-control study from the USA also reported a higher risk of nonlobar than lobar ICH (aOR 3.06 [95% CI 2.29–4.09]) vs 2.81 [2.04–3.89]) with OAC use.¹⁰ Other studies reported an association between OAC use and a higher risk of lobar ICH.^9,13,32,33 The majority of previous studies had considerably smaller cohorts than our study and lacked the power to test these differences robustly (eTable 2). In addition, the definitions of location varied across studies, which may partly explain the discrepant results. Studies with results similar to ours classified cerebellar hemorrhages as nonlobar as in our analysis.^10,11 Studies reporting OAC use associated with higher odds of lobar s-ICH than nonlobar s-ICH classified cerebellar s-ICH as lobar,^13,32,33 with the exception of one study that excluded cerebellar hemorrhages.⁹ This may have influenced the results because OAC use is reported strongly associated with cerebellar s-ICH location,⁸ a finding we also observed in our study. OAC-associated ICH results in larger hematomas,³⁴ a phenomenon that probably explains the strong association we found between OAC use and “large unclassifiable” ICH. In the CT substudy, we measured larger s-ICH volumes assessed by ABC/2³⁵ in both lobar and nonlobar locations in patients on OACs compared with antithrombotic drug nonusers (data not shown). Our novel results underscore the safety advantages of DOACs over warfarin regarding ICH risk. We speculated to what degree our findings related to DOAC use are due to vessels in nonlobar areas of the brain being more susceptible to the effects of untreated or inadequately controlled hypertension, but we lacked data to adequately address this question.

A case series reported a higher risk of lobar vs nonlobar s-ICH in patients using platelet antiaggregants,¹³ whereas most other studies^8,10-12 did not find location-specific differences (eTable 2). Our ICH cases were unselected, and we used a register-based method of establishing medication use (90% of all low-dose aspirin sale in Denmark are recorded in the Prescription Registry).²⁹ Of interest, overall platelet antiaggregant use (i.e., regardless of type of antiaggregant) and aspirin use had higher risks of lobar ICH, and these associations were particularly strong in the subset of ICH patients with high probability of CAA. A possible explanation is that these findings reflect differences in the microangiopathies underlying lobar vs nonlobar s-ICH, with CAA being more susceptible to aspirin. This hypothesis requires further testing particularly because the pattern differed for clopidogrel (i.e., risks were similar across locations).

This study has several strengths. We used multiple sources to verify all hospitalized ICH cases in a defined population.^21,24 Although some ICH cases may have not been hospitalized, or may have died before neuroimaging, these numbers are likely small because the threshold for rapid in-hospital evaluation of stroke in Denmark is low and independent of income. As the frequency of autopsy is quite low in Denmark,³⁶ supplementing the data with additional data from the Cause-of-Death Registry would likely not improve the ascertainment of cases. In addition, previous population-based studies indicate a low number of ICH cases being diagnosed only at autopsy.^25,37 We eliminated recall bias by using registries that prospectively collect data on medication use and previous medical history. We used the Civil Registration System to identify all residents in the RSD and, therefore, selected true general population controls.

This study also has limitations. It was underpowered for the assessment of associations of combinations of antithrombotic drugs. Because this was an exploratory analysis, we tested many associations and did not correct for multiple comparisons, which may lead to false positive results. However, as most point estimates in our analyses showed consistent results regarding the direction, we think this is less likely. Although our method of classifying s-ICH location could raise concerns regarding diagnostic accuracy, similar results were reached in the CT substudy, which had masked reevaluation of brain scans according to CHARTS.³⁰ We lacked information on hematoma location within the cerebellum, which could allow subclassification into superficial s-ICH, which has been linked to CAA in recent studies, vs deep/midline cerebellar s-ICH (i.e., arteriolosclerosis-related). As an estimated 85% of cerebellar ICHs are likely due to non-CAA vasculopathy,³⁸ the extent of misclassification introduced by including cerebellar ICH in the nonlobar group in the main analysis is small. We lacked or had only partial data for several potential confounders, including blood pressure, alcohol use, smoking, genetic information (e.g., ApoE variants), and markers of socioeconomic status. As the E-value analyses suggest, strong unmeasured confounding is needed to account for the results of the primary analyses. In Denmark, health care is free of charge to the individual at the time of use and medication costs are covered to a varying degree depending on a patient's total health care expenditure over the previous year,³⁹ which are factors that reduce the effect of socioeconomic status.²³ We lacked data on SVD MRI biomarkers (e.g., cortical superficial siderosis and microbleeds) that could have improved the accuracy of the CAA classification.²⁰ Given that this is an observational study, we emphasize that the observed associations should not be interpreted as causal. Finally, the generalizability of our results is limited by the Danish population being mainly of European ancestry.

All investigated antithrombotic drugs were associated with a higher risk of incident s-ICH, but the strengths of these associations with drug type varied by hematoma location. The use of VKAs was associated with the highest risk, regardless of hematoma location. DOAC use was more strongly associated with nonlobar than lobar ICH, whereas the opposite was true for platelet antiaggregants. The immediate clinical/practice implications of our findings are unclear at this time. Our findings raise the intriguing possibility that the choice of antithrombotic agents may require balancing the risk of lobar vs nonlobar ICH in some patients. Further studies are needed to improve current understanding of the pathophysiology of antithrombotic-related ICH and appropriate patient selection for types of antithrombotic therapies.

Disclosure

R. Al-Shahi Salman reports receiving grants from the British Heart Foundation for RESTART (Restart or Stop Antithrombotics Randomised Trial) and ASPIRING (Antiplatelet Secondary Prevention International Randomized trial after INtracerebral hemorrhaGe) paid to the University of Edinburgh. J. Hallas has participated in regulatory mandated projects by Pfizer, Leo Pharma, Novo Nordisk, Roche, and Atellas with funding paid to his employer. L.A. García Rodríguez works for CEIFE, which received research grants from Bayer for research projects outside the submitted work. M. Selim receives support from the NIH (NINDS: U01NS102289, and NIA: UF1NS120871) and serves on the Advisory Board of MedRhythms Inc. D. Gaist has received speaker honoraria from Bristol-Myers Squibb and Pfizer outside the submitted work and participated in research outside the submitted work funded by Bayer with funds paid to the institution where he is employed. The other authors report no relevant disclosures. Go to Neurology.org/N for full disclosures.

Glossary

CAA

cerebral amyloid angiopathy

DOACs

direct oral anticoagulants

ICH

intracerebral hemorrhage

IVH

isolated intraventricular hemorrhages

OR

odds ratio

RSD

Region of Southern Denmark

s-ICH

spontaneous intracerebral hemorrhage

SVD

small vessel disease

VKA

vitamin K antagonists

Appendix. Authors

Name	Location	Contribution
Nils Jensen Boe, MD	Neurology Research Unit, Odense University Hospital; University of Southern Denmark	Drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data; study concept or design; analysis or interpretation of data
Stine Munk Hald, PhD	Neurology Research Unit, Odense University Hospital; University of Southern Denmark	Drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data; study concept or design; analysis or interpretation of data
Alexandra Redzkina Kristensen, MD	Neurology Research Unit, Odense University Hospital; University of Southern Denmark	Drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data; analysis or interpretation of data
Sören Möller, PhD	Open Patient Data Explorative Network, Odense University Hospital; Department Clinical Research, University of Southern Denmark, Odense, Denmark	Drafting/revision of the manuscript for content, including medical writing for content; study concept or design; analysis or interpretation of data
Jonas A. Bojsen, MD	Department of Radiology, Odense University Hospital, Odense, Denmark	Drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data; analysis or interpretation of data
Mohammad Talal Elhakim, MD	Department of Radiology, Odense University Hospital, Odense, Denmark	Drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data; analysis or interpretation of data
Mark A. Rodrigues, PhD, FRCR, MBChB, BSc	Centre for Clinical Brain Sciences, University of Edinburgh, United Kingdom	Drafting/revision of the manuscript for content, including medical writing for content; study concept or design; analysis or interpretation of data
Rustam Al-Shahi Salman, PhD	Centre for Clinical Brain Sciences, University of Edinburgh, United Kingdom	Drafting/revision of the manuscript for content, including medical writing for content; study concept or design; analysis or interpretation of data
Jesper Hallas, DrMedSci	Department of Clinical Pharmacology, Pharmacy and Environmental Medicine, University of Southern Denmark, Odense	Drafting/revision of the manuscript for content, including medical writing for content; study concept or design; analysis or interpretation of data
Luis A. García Rodríguez, MD	Centro Español Investigación Farmacoepidemiológica, Madrid, Spain	Drafting/revision of the manuscript for content, including medical writing for content; study concept or design; analysis or interpretation of data
Magdy Selim, MD, PhD	Beth Israel Deaconess Medical Center, Harvard Medical School, Boston, MA	Drafting/revision of the manuscript for content, including medical writing for content; study concept or design; analysis or interpretation of data
Larry B. Goldstein, MD	Department of Neurology and Kentucky Neuroscience Institute, University of Kentucky, Lexington	Drafting/revision of the manuscript for content, including medical writing for content; study concept or design; analysis or interpretation of data
David Gaist, PhD	Neurology Research Unit, Odense University Hospital; University of Southern Denmark	Drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data; study concept or design; analysis or interpretation of data

Study Funding

The project received funding from the Novo Nordisk Foundation (grant NNF20OC0064637; Dr. Gaist). The funding organization had no role in the design of the study or the collection, analysis, and the interpretation of the data, or the interpretation of data and manuscript writing.