Small Vessel Disease and Ischemic Stroke Risk During Anticoagulation for Atrial Fibrillation After Cerebral Ischemia

Introduction

Although oral anticoagulation (OAC) is highly effective for secondary prevention after ischemic stroke or transient ischemic attack (TIA) associated with atrial fibrillation (AF), some studies indicate a substantial risk (4.7%–7.7%/year)^1,2 of recurrent cerebrovascular events despite OAC. Recurrent ischemic stroke despite anticoagulation for AF may reflect either inadequate anticoagulation, or alternative stroke mechanisms, since AF is commonly present concurrently with other relevant pathologies, including large artery atherosclerosis or cerebral small vessel disease (SVD). A recent case-control study reported a prevalence of 32.7% of other stroke etiologies than cardioembolism in patients with AF on anticoagulant therapy for stroke prevention, with SVD contributing 10.4%.³ Cerebral SVD—associated with age, hypertension,^4,5 and a major contributor to stroke and cognitive dysfunction⁴—can be detected and quantified by magnetic resonance imaging (MRI) biomarkers including perivascular spaces, cerebral microbleeds (CMBs), lacunes, and white matter hyperintensities (WMHs).⁶ Previous studies in unselected ischemic stroke populations demonstrated basal ganglia perivascular spaces (BGPVS), CMB burden, severe WMH, and multiple lacunes to be associated with recurrent ischemic stroke.^7,8 Conversely, in patients with intracranial arterial stenosis, SVD was not associated with ischemic stroke risk.⁵

See related article, p 100

Our aim was to examine the association between SVD presence and the risk of ischemic stroke during follow-up in a prospective multicenter observational inception cohort of patients anticoagulated for AF after ischemic stroke or TIA; we hypothesized that the risk of ischemic stroke during follow-up despite OAC is associated with baseline SVD presence and severity.

Methods

Data Availability Statement

Data of this study are available from UCL Institute of Neurology on reasonable request and after consideration by the CROMIS-2 (Clinical Relevance of Microbleeds in Stroke Study) Steering Committee. Persons interested in obtaining access to the data should contact the corresponding author at: d.werring@ucl.ac.uk.

Clinical Data

We included participants recruited to the CROMIS-2 study, a multicenter prospective inception cohort study of patients anticoagulated for AF after ischemic stroke or TIA, between August 3, 2011 and July 31, 2015. The details of study design, inclusion and exclusion criteria have previously been published.^9,10 We followed all adult patients (ie, ≥18 years of age) using standard structured postal questionnaires or telephone interviews to patients and their general practitioners or hospital visit at 6, 12, and 24 months following their index ischemic event. We defined ischemic stroke during follow-up as a new clinical event with neurological deficit consistent with ischemic stroke.¹¹ Clinicians at different participating centers assessed the recurrent event subtype using the TOAST (Trial of ORG 10172 in Acute Stroke Treatment) classification. We excluded patients without appropriate brain MRI at baseline or follow-up data.

Imaging Analysis

We included standardized parameters for the MRI sequences, including axial T2-weighted, T2-weighted gradient-recalled echo (T2*-GRE), coronal T1-weighted, fluid-attenuated inversion recovery and diffusion-weighted imaging at each site.^9,10 All structural markers of SVD were rated in accordance with consensus criteria.⁶ BGPVS and centrum semiovale perivascular spaces were defined and rated by 2 raters (G. Banerjee, H. Du) on T2 and fluid-attenuated inversion recovery sequences using a validated 4-point visual rating scale on a single predefined slice.^6,12 CMBs were evaluated by 2 raters (Dr Wilson, G. Banerjee) using T2*-GRE sequences and the validated Microbleed Anatomic Rating Scale.^6,10 Lacunes of presumed vascular origin were rated by 2 raters (G. Banerjee, H. Du) as ovoid or round, subcortical, cerebrospinal fluid-signal like cavities between 3 and 15 mm in diameter on T2 and fluid-attenuated inversion recovery sequences.⁶ WMH were rated by a single rater (Dr Wilson) on T2 and fluid-attenuated inversion recovery sequences using the Fazekas scale.^6,13 All raters were well trained as documented previously.^10,12 We quantified intrarater and interrater reliability for structural markers of SVD using Cohen κ coefficient. SVD presence was defined as any presence of the following: ≥11 BGPVS, ≥1 CMBs, ≥1 lacunes, or moderate to severe WMH (Fazekas grade 3 in periventricular or grade 2 to 3 in deep)^4,5 as defined in a composite SVD scale.⁴ We also calculated the total SVD score, which allocates 1 point for the presence of each of the following: ≥11 BGPVS, ≥1 CMBs, ≥1 lacunes, or moderate to severe WMH.⁴

Standard Protocol Approvals, Registrations, and Patient Consents

CROMIS-2 was approved by the National Institute for Health Clinical Research Network (Research Ethics Committee reference: 10/H0716/64).^9,10

Statistics

Continuous data were summarized as mean values with SDs (if normally distributed) or median value with interquartile range (if not normally distributed) and categorical data as counts with proportions. We calculated absolute event rates per 100 patient-years for ischemic stroke during follow-up. We compared the univariate Kaplan-Meier probabilities of ischemic stroke during follow-up for patients with and without SVD presence and individual SVD markers. We developed 2 Cox proportional hazards models to evaluate the unadjusted and adjusted risks of ischemic stroke during follow-up in patients with the presence and absence of SVD and individual SVD markers: first including CHA₂DS₂-VASc as a single confounder; and second, including variables which were associated with recurrent ischemic stroke in univariable analysis (P<0.1). The risk of ischemic stroke during follow-up with increasing SVD score was also investigated using Cox proportional hazards models. We used the Fisher exact test to compare the baseline SVD presence among different recurrent mechanisms. All the statistical analyses were done using SPSS 25.0 (IBM; SPSS, Inc, Chicago, IL).

Results

We included 1419 eligible patients (75.8 years [SD, 10.4]; 42.1% female) in the final analysis (Figure 1). The 43 patients without follow-up were similar to those with follow-up in age (73.8 years [SD 11.4] versus 75.8 years [10.3]; P=0.173), sex (female 46.5% versus 42.1%; P=0.561), hypertension (57.1% versus 63.4%; P=0.411), and SVD presence (46.5% versus 54.1%; P=0.324). The median follow-up duration was 2.3 years (SD, 1.0), providing 3265 patient-years of follow-up data. Of 1419 participants, 53 patients had ischemic stroke during follow-up, an event rate of 1.62 per 100-patient years (95% CI, 1.24–2.12).

• Download figure
• Download PowerPoint

SVD was present in 768 (54.1%) patients. The prevalence of individual SVD markers were as follows: ≥11 BGPVS presence 375 (26.4%) patients; ≥11 centrum semiovale perivascular spaces presence 681 (48.0%) patients; CMBs presence 298 (21.0%) patients (strictly lobar in 109 patients, strictly deep in 117 patients, and mixed in 72 patients); lacunes 295 (20.8%) patients; and moderate to severe WMH 283 (19.9%) patients. Among those with ≥11 BGPVS, 261(18.4%) patients had 11 to 20 BGPVS and 114 (8.0%) patients had >20 BGPVS. Intrarater and interrater reliability were as follows: for the presence of BGPVS 0.86 (95% CI, 0.69–1.00) and 0.82 (95% CI, 0.60–0.96); for the presence of centrum semiovale perivascular spaces 0.90 (95% CI, 0.78–1.00) and 0.80 (95% CI, 0.66–0.95), for presence of CMBs 0.93 (95% CI, 0.86–1.00) and 0.85 (95% CI, 0.74–0.96); and for presence of lacunes 0.87 (95% CI, 0.70–1.00) and 0.81 (95% CI, 0.63–1.00).

The baseline characteristics of clinical and MRI factors with and without ischemic stroke during follow-up are shown in Table 1. Patients with ischemic stroke during follow-up had a higher median CHA₂DS₂-VASc score of 6 (interquartile range, 5–7) versus 5 (4–6). Age (unadjusted hazard ratio [HR], 1.04 [95% CI, 1.01–1.07]), female sex (unadjusted HR, 2.01 [95% CI, 1.17–3.48]), diabetes (unadjusted HR, 2.11 [95% CI, 1.16–3.84]), and vascular disease (unadjusted HR, 2.24 [95% CI, 1.26–3.99]) were associated with ischemic stroke during follow-up. Of the individual baseline SVD radiological features, ≥11 BGPVS (unadjusted HR, 2.44 [95% CI, 1.42–4.19]) were associated with ischemic stroke during follow-up. We found weak evidence of an association between CMBs presence (unadjusted HR, 1.62 [95% CI, 0.90–2.91]), moderate to severe WMH presence (unadjusted HR, 1.72 [95% CI, 0.95–3.12]) and ischemic stroke during follow-up.

The ischemic stroke event rate during follow-up in patients with SVD presence was 2.20 per 100-patient years (95% CI, 1.60–3.02), compared with 0.98 per 100 patient-years (95% CI, 0.59–1.62) in those without SVD presence. The absolute rate increase associated with SVD presence was 1.22 per 100 patient-years (95% CI, 1.01–1.40). In Kaplan-Meier analysis, the ischemic stroke event during follow-up was more frequent in patients with SVD presence compared with those without SVD presence (log-rank test, P=0.008, Figure 2A). In univariate Cox regression analysis, patients with SVD had a 2.2-times higher risk of ischemic stroke during follow-up (95% CI, 1.21–4.01, P=0.009). We found an association between higher SVD score with increasing risk of ischemic stroke during follow-up (HR per point increase, 1.45 [95% CI, 1.14–1.83]; P=0.002). After adjusting for CHA₂DS₂-VASc, patients with SVD presence still had a 1.9× higher risk of ischemic stroke during follow-up ([95% CI, 1.01–3.53]; P=0.046); the risk of ischemic stroke during follow-up remained significantly increased with increasing SVD score (HR per point increase, 1.33 [95% CI, 1.04–1.07]; P=0.023). To better appreciate the effects of the different components in the CHA₂DS₂-VASc score, we adjusted for age, sex, diabetes, and vascular disease in another multivariate model. The adjusted HRs for the presence of SVD, SVD per point increase, and individual SVD markers were very similar to those adjusted for CHA₂DS₂-VASc as a single confounder (Table 2).

• Download figure
• Download PowerPoint

We performed 3 sensitivity analyses to investigate whether the associations of SVD presence and SVD score with ischemic stroke during follow-up were robust to other potential confounders; (1) adjusted for therapeutic time in range in Vitamin K antagonist-treated patients (available in 861 patients); (2) adjusted for carotid-artery stenosis (available in 682 patients, of which 581 were <50% stenosis, 60 were 50% to 70% stenosis, 41 were >70% stenosis); and (3) adjusted for antihypertensive and statin treatment (data available in 1383 patients; Table 3).

The ischemic stroke event rate during follow-up in patients with ≥11 BGPVS was 2.98 per 100-patient years (95% CI, 1.99–4.44), compared with 1.18 per 100 patient-years (95% CI, 0.82–1.70) in those with 0 to 10 BGPVS. In Kaplan-Meier analysis, ischemic stroke during follow-up was more frequent in patients with ≥11 BGPVS compared with those with 0–10 BGPVS (log-rank test, P=0.001, Figure 2B). We found weak evidence that the rate of ischemic stroke during follow-up was higher with the presence of moderate to severe WMH (log-rank test, P=0.073), CMBs (log-rank test, P=0.104) but not with the presence of centrum semiovale perivascular spaces (log-rank test, P=0.196), and lacunes (log-rank test, P=0.494). However, after adjusting for CHA₂DS₂-VASc as a single confounder in the multivariate Cox regression analysis, only ≥11 BGPVS (adjusted HR, 2.11 [95% CI, 1.21–3.67], P=0.008) remained statistically significant. The presence of moderate to severe WMH (adjusted HR, 1.33 [95% CI, 0.72–2.45]; P=0.366) and CMBs (adjusted HR, 1.47 [95% CI, 0.81–2.65]; P=0.206) were not significantly associated with ischemic stroke during follow-up.

Data on the likely mechanism of recurrent ischemic stroke are shown in Table 4. There were no significant differences in baseline SVD presence among different recurrent mechanisms. However, the absolute rate of recurrent small artery occlusion in patients with baseline SVD was higher than those without SVD (7.9% versus 0%).

Discussion

In our prospective multicenter inception cohort study of patients anticoagulated for AF after ischemic stroke or TIA, MRI-defined SVD is independently associated with an increased risk of ischemic stroke during follow-up; moreover, a higher composite SVD score is associated with increasing risk. Thus, these findings suggest that the risk of ischemic stroke during follow-up despite OAC in some patients might be related to SVD presence and severity (and, in some cases, because of small vessel occlusion), suggesting a need for better stroke prevention strategies in this patient group.

Having AF does not necessarily mean that this is the exclusive cause of an index or recurrent stroke,^3,14 since AF commonly coexists with other etiologies, including SVD. In line with our findings, data from the OXVASC study (The Oxford Vascular Study) also showed that patients with cardioembolic stroke who had increasing SVD score were at increasing risk of ischemic stroke recurrence (unadjusted HR, 1.34 [95% CI, 1.00–1.78]).⁷ Therefore, preventing recurrent ischemic stroke in stroke or TIA patients with AF and SVD requires better management options. Previous randomized controlled trials (ORBIT-AF [The Outcomes Registry for Better Informed Treatment of AF],¹⁵ ARISTOTLE [Apixaban for Reduction in Stroke and Other Thromboembolic Events in AF],¹⁶ and ROCKET-AF [Rivaroxaban Once Daily Oral Direct Factor Xa Inhibition Compared with Vitamin K Antagonism for Prevention of Stroke and Embolism Trial in AF]¹⁷) do not support combining antiplatelet and standard dose anticoagulation therapy for stroke prevention in patients with AF. Findings of recent studies (COMPASS [Cardiovascular Outcomes for People Using Anticoagulation Strategies]¹⁸ and PIONEER AF-PCI [An Open-Label, Randomized, Controlled, Multicenter Study Exploring Two Treatment Strategies of Rivaroxaban and a Dose-Adjusted Oral Vitamin K Antagonist Treatment Strategy in Subjects with AF who Undergo Percutaneous Coronary Intervention]¹⁹) suggested that low-dose rivaroxaban 2.5 mg twice daily plus single antiplatelet therapy might be a reasonable treatment option for patients with AF and atherosclerotic (including carotid) cardiovascular disease. However, these trials did not measure SVD or include participants known to have SVD.

The use of antiplatelet or anticoagulant medication might usefully be guided by stroke subtype and mechanism. For example, a prospective intervention study showed the recurrent stroke event rate in warfarin-treated patients with AF and lacunar stroke was similar to that in those treated with aspirin alone (8.8% [3.1%–14.6%] versus 8.9% [1.8%–16.1%]).¹⁴ This finding raises the uncertainty whether antiplatelet therapy (eg, using cilostazol, which is a promising agent in SVD) might be an alternative approach to anticoagulation for secondary prevention in mild stroke or patients with TIA with AF and SVD. Nonpharmacological approaches, such as left atrial appendage closure also reduce stroke occurrence in patients with AF.²⁰ Whether long-term antiplatelet therapy following left atrial appendage closure might improve secondary prevention for patients with AF and coexistent SVD has not been investigated.

Complementary treatments targeting common risk factors for SVD and nonvalvular AF, in addition to anticoagulation, also need to be considered. A previous study showed higher statin adherence predicted reduced recurrent ischemic stroke risk in patients with stroke because of AF after controlling for warfarin therapeutic time in range (HR, 0.61 [95% CI, 0.41–0.90]; P=0.012).²¹ A subgroup analysis of the SPS3 (Secondary Prevention of Small Subcortical Strokes) study showed aggressive blood pressure lowering was significantly associated with reduced risk of stroke recurrence in patients who have SVD with CMBs (HR, 0.5 [95% CI, 0.3–0.9]).²² In our cohort, neither antihypertensives (HR, 1.63 [95% CI, 0.69–3.89]) nor statin treatment (HR, 0.91 [95% CI, 0.48–1.92]) were associated with an increased ischemic stroke risk during follow-up after adjusting for CHA₂DS₂-VASc. However, our protocol did not measure statin adherence or strict blood pressure control.

Our data did not demonstrate that baseline SVD is associated with the subtype of recurrent ischemic stroke. However, the absolute rate of recurrent small artery occlusion in patients with baseline SVD was higher than in those without SVD. Our findings were in line with a prospective study that showed patients with stroke with large artery atherosclerosis and baseline SVD had higher risk of recurrent stroke regardless of its recurrence mechanism.²³ One possible explanation is that SVD is associated with risk factors for atherosclerosis, which are known predictors of stroke recurrence. It is always challenging to identify the exact cause of a stroke if there are 2 or more potential causes. The proportion of recurrent small vessel occlusion in our cohort is much lower than those in previous studies^14,24 but might be underestimated. First, only a minority of patients underwent MRI for their recurrent stroke, reducing our ability to accurately classify small vessel occlusions. Second, in patients with known AF clinicians are more likely to regard the recurrent event as cardioembolic, reflecting the TOAST criteria.²⁵

Another important finding of our study is that ≥11 BGPVS are independently associated with the risk of ischemic stroke during follow-up in patients anticoagulated for AF. BGPVS are defined as cerebrospinal fluid-filled spaces surrounding penetrating arteries in basal ganglia⁶ and are recognized as a marker of hypertensive arteriopathy because of their association with age and hypertension.^7,26 Previous studies demonstrated an independent association between BGPVS and severity of intracranial and extracranial atherosclerosis.^26–28 Possible mechanisms for this link include chronic cerebral hypoperfusion²⁸ and increased arterial stiffness.²⁹

We acknowledge limitations: first, there is inevitable selection bias at baseline since we only recruited patients able to undergo MRI. Second, despite standardization, minor local variations in the magnetic resonance sequences are inevitable. Third, we did not adjust for the time between index stroke/TIA events to receiving anticoagulation treatment; this is because clinicians decided on timing of OAC based on best clinical judgment. However, we found no evidence that OAC timing affected a composite of end point of stroke, TIA or death at 90 days follow-up³⁰; the optimal timing of OAC after stroke because of AF remains uncertain.^31,32

Our study has important strengths. We prospectively studied a large prospective inception cohort of patients at multiple hospital stroke units using standardized MRI sequences, rated for imaging markers of SVD using validated scales by trained observers. Our follow-up rate was 97%, and experienced observers adjudicated all primary events blinded to baseline neuroimaging findings. We undertook survival analysis to take into account baseline confounding factors.

Conclusions

Our findings suggest that OAC is not as effective in patients with AF who have SVD as it is in those without; this patient group should be a focus for trials of new prevention strategies.

Acknowledgments

We thank the principal investigators, research practitioners, and participants involved in CROMIS-2 (Clinical Relevance of Microbleeds in Stroke Study). Author contributions are listed in the Data Supplement.

Footnotes