Background and purpose Cerebral microbleeds (CMBs) are radiological markers which predict future intracerebral haemorrhage. Researchers are exploring how CMBs can guide anticoagulation decisions in atrial fibrillation (AF). The purpose of this study is to evaluate the correlation of non-vitamin K antagonist oral anticoagulants (NOAC) exposure and prevalence of CMBs in Chinese patients with AF.
Methods We prospectively recruited Chinese patients with AF on NOAC therapy of ≥30 days for 3T MRI brain for evaluation of CMBs and white matter hyperintensities. Patients with AF without prior exposure to oral anticoagulation were recruited as control group.
Results A total of 282 patients were recruited, including 124 patients in NOAC group and 158 patients in control group. Mean duration of NOAC exposure was 723.8±500.3 days. CMBs were observed in 103 (36.5%) patients. No significant correlation was observed between duration of NOAC exposure and quantity of CMBs. After adjusting for confounding factors (ie, age, hypertension, labile hypertension, stroke history and white matter scores), previous intracerebral haemorrhage was predictive of CMBs (OR 15.28, 95% CI 1.81 to 129.16), particularly lobar CMBs (OR 5.37, 95% CI 1.27 to 22.6). While white matter score was predictive of mixed lobar CMBs (OR 1.65, 95% CI 1.1 to 2.5), both exposure and duration of NOAC use were not predictive of presence of CMBs.
Conclusions In Chinese patients with AF, duration of NOAC exposure did not correlate with prevalence and burden of CMBs. Further studies with follow-up MRI are needed to determine if long-term NOAC therapy can lead to development of new CMBs.
Statistics from Altmetric.com
Atrial fibrillation (AF) is the most common cardiac arrhythmia worldwide, with a fivefold increase in risk of ischaemic stroke with high morbidity and mortality.1 With ageing population, the number of AF is projected to double by 2060.2 Over the past 15 years, the number of ischaemic stroke due to AF has increased by almost threefold in Chinese population.3 Currently, oral anticoagulants are still underused in around 40% of the eligible patients globally due to their potential risk of devastating intracerebral haemorrhage (ICH).4 5 The dilemma of coexisting risk of ischaemic and haemorrhagic stroke in patients with AF necessitates better risk stratification tools to help select appropriate stroke prevention strategies for high-risk individuals.
Cerebral microbleeds (CMBs) are small, round areas of signal loss on blood-sensitive MR sequences. They are perivascular hemosiderin deposits, indicative of previous asymptomatic leakage from bleeding-prone microangiopathy.6 It is commonly observed in patients with risk factors for ICH (eg, ageing, hypertension, white matter change and cerebral amyloid angiopathy, etc), and is an independent predictor for future ICH.7 Data from retrospective studies have shown that CMBs were observed in 30.5% of patients with AF,8 which has raised concerns about safety of anticoagulation in these patients, who may have higher risk of ICH that tilts the balance away from treatment benefit. Recently, there are growing interests in exploring how anticoagulant decisions may be influenced by CMBs.9 10
One of the areas which remains unclear is the influence of anticoagulation on prevalence of CMBs. Under impaired hemostasis by anticoagulation, there may be higher tendency for red blood cells to extravasate through bleeding-prone vessels affected by small vessel diseases. The number of CMBs may therefore theoretically increase with the duration or intensity of anticoagulant therapy, leading to cumulative risk of ICH in the long run. Hence, understanding the dynamic interactions between anticoagulation, microangiopathy and development of CMBs is important to determine the best approach to prognosticate and monitor the long-term risk of ICH in patients with AF under anticoagulation therapy. Currently, there are limited data addressing this issue.
According to data from the Rotterdam scan study, previous warfarin use and a higher maximum international normalised ratio (INR) were associated with more CMBs in deep or infratentorial areas (OR 1.7, 95% CI 1.24 to 2.34).11 Concerning non-vitamin K antagonist oral anticoagulants (NOAC), the only study available up-to-date was a small prospective study in Japan, which evaluated CMBs in 69 patients with AF (including 21 patients on warfarin, 9 patients on warfarin plus antiplatelets and 23 patients on NOAC) at baseline and 1 year.12 In this study, warfarin, but not NOAC, was independently associated with new development of CMBs (HR 10.75, P=0.03). Unfortunately, the study is limited by its small sample size and heterogeneous sample with patients on different types of antithrombotic agents. Hence, more prospective data with larger sample size, more homogeneous sample are urgently needed for better understanding of impact of NOAC on CMBs.
In this study, we aimed to evaluate the correlation of NOAC exposure and the number of CMBs in Chinese patients with AF.
Materials and methods
This is a substudy of the ongoing prospective IPAAC study (Risk of Intracerebral haemorrhage in Patients taking oral Anticoagulants for Atrial fibrillation with Cerebral microbleeds). Patients were recruited from medical outpatient clinics and acute stroke unit at Prince of Wales Hospital. Inclusion criteria for IPAAC study are Chinese patients aged 18 years or above, presence of AF or atrial flutter who require oral anticoagulation for stroke prevention. Exclusion criteria include patients who are not physically fit for MRI brain (eg, pacemakers, aphasic patients who are not cooperative for examination, etc), concurrent treatment with antiplatelet therapy, platelet count <100×109/L, known bleeding tendency and life expectancy <1 year due to comorbid medical conditions. Informed consents were obtained. After enrolment into the IPAAC study, 3T MRI brain were performed for evaluation of CMBs and white matter hyperintensities. For the current substudy, patients were divided into NOAC group and control group based on their prior exposure to anticoagulants before MRI. Patients with prior NOAC exposure for 30 days or more were assigned to NOAC group. Patients without prior exposure to any oral anticoagulants (including both warfarin and NOAC) before MRI were assigned to control group.
MR examinations were performed on a 3T scanner (Achieva TX; Philips Medical Systems, Best, The Netherlands) using an 8-channel head coil. All patients underwent our standard stroke scanning protocol. Venous Blood Oxygen Level Dependent parameters were repetition time/echo time=18/25 ms, flip angle=15° and reconstructed at 0.45×0.45×1.0 mm. Fluid attenuation inversion recovery (FLAIR) parameters were repetition time/inversion time/echo time=11 000 ms/2800 ms/125 ms, flip angle=90°, reconstructed at 0.33×0.33×5.0 mm.
A CMB was defined as an old, silent focus of signal loss in the gradient-echo sequence, measuring 2–10 mm in diameter. Symmetric signal loss or hypointensities in the globus pallidum, which might represent calcification, were excluded. Flow void artefacts of the pial blood vessels were distinguished from CMBs by their morphology and correlation with T1-weighted and T2-weighted images. CMBs located within territory of acute or old infarct were classified as possible haemorrhagic transformation and were excluded from analysis. Distribution of CMBs were classified as lobar, deep and infratentorial regions based on their anatomical location. White matter changes on FLAIR sequence were scored using a visual rating scale proposed by Fazekas et al.13
Continuous clinical characteristics were compared by 2-sample t-test or Mann-Whitney U test based on the result of normality test. Shapiro-Wilk test were used to examine the validity of normality assumption. Categorical clinical characteristics were compared by χ2 test.
Univariate logistic regressions were fitted for presence of CMBs overall and in each of these subgroups—pure lobar, pure deep, mixed lobar and infratentorial CMBs. Multiple logistic regression with stepwise approach for variable selection was then used to investigate the underlying relationship between presence of CMBs and clinical and radiological characteristics. The base model consisted of age only, while the full model included other clinical and radiological characteristics. The best model between base model and full model was chosen based on Akaike information criterion. Correlation between exposure of anticoagulant and number of CMBs was examined by Spearman’s rank correlation coefficient.
In this study design, we anticipated that most of the patients in NOAC group would comprise patients with or without stroke history recruited from outpatient clinic, while patients in control group would mostly comprise patients recruited from acute stroke unit with known or newly diagnosed AF. In order to explore if the study results could be affected by the unbalanced stroke history due to this potential selection bias, we adjusted for age, stroke history and other confounding factors identified at baseline comparison in multiple logistic regression for prediction of CMBs. We also repeated the logistic regressions in the NOAC group alone to see if the results would be significantly different after excluding patients in the control group.
All analyses were performed in R V.3.4.0. P values <0.05 were considered as statistically significant.
A total of 282 patients were recruited into this study, with 124 patients into NOAC group and 158 patients in control group. The mean duration of NOAC exposure was 723.8±500.3 days (median 617±477 days, ranged from 37 to 2500 days). Table 1 summarises the demographic, clinical and radiological characteristics of patients in the two groups. Majority of patients in both groups had hypertension. Patients in the NOAC group were more likely to have labile hypertension, history of ICH and a slightly but significantly higher white matter score. Compared with NOAC group, most of the patients in the control group had history of stroke or transient ischaemic attack with marginally higher CHA2DS2-VASc14 and HAS-BLED scores. Proportion of patients with prior use of antiplatelet agents and newly diagnosed AF were also higher in this group. There was no significant difference in the prevalence, quantity and distribution of CMBs between those with and without prior NOAC exposure.
CMBs were observed in 103 (36.5%) patients, including 41 (39.8%) patients with pure lobar CMBs, 11 (10.6%) with pure deep CMBs, 34 (33.0%) with mixed lobar CMBs and 41 (39.8%) with infratentorial CMBs (figure 1). Among 103 patients with CMBs, the mean number of CMBs was 4.8±10.1 (ranged from 1 to 75). Patients with CMBs were slightly older, more likely to have prior ICH and ischaemic heart disease, and had higher CHA2DS2-VASc and HAS-BLED scores (table 2). There was no significant difference in other baseline risk factors, prior use of antiplatelet agents and current use of individual NOAC between those with and without CMBs. No linear relationship was observed between creatinine clearance and number of CMBs (figure 2).
No significant correlation was observed between of duration of NOAC exposure and quantity of CMBs in all patients and in NOAC group alone (figure 3A and B). After adjusting for confounding factors in multiple logistic regression for all patients (ie, age, hypertension, labile hypertension, history of ischaemic and haemorrhagic stroke), history of ICH was found to be an independent predictor for presence of CMBs (OR 15.28, 95% CI 1.81 to 129.16), which was mostly driven by its association with pure lobar CMBs (OR 5.37, 95% CI 1.27 to 22.6) (table 3). In addition, white matter score was found to independently predict mixed lobar CMBs (OR 1.65, 95% CI 1.1 to 2.5). Neither NOAC exposure nor duration of NOAC use was predictive of presence of CMBs overall or CMBs in any regions (table 3).
Consistent results were observed when repeating multiple logistic regression analyses for NOAC group only. Prior ICH was an independent predictor for presence of CMBs in general (OR 11.57, 95% CI 1.28 to 105.02), and pure lobar CMBs (OR 7.24, 95% CI 1.21 to 43.5). On the other hand, white matter score was predictive of mixed lobar CMBs (OR 4.19, 95% CI 1.60 to 10.97) and infratentorial CMBs (OR 2.79, 95% CI 1.24 to 6.29) (table 3).
This is the first study evaluating correlation of NOAC exposure and prevalence of CMBs in Chinese patients with AF. In this study involving patients with NOAC use up to 6.8 years, both history of NOAC use and duration of NOAC exposure were not associated with higher prevalence and number of CMBs. On the other hand, history of ICH and underlying white matter scores were the most predictive factors for pure lobar and mixed lobar CMBs, respectively. The results of this study reinforce the safety of NOAC in Chinese patients with AF who have higher risk of ICH compared with Caucasians. Also, the data are helpful for future formulation of risk stratification strategies for patients with AF.
The first NOAC has been approved by the Food and Drug Administration for prevention of stroke in AF since October 2010. In this study, we were able to include 18.6% of patients with NOAC exposure for 3 or more years and evaluate the short-term and medium-term effect of NOAC on prevalence of CMBs. Compared with the study in Japanese patients,12 our study has much larger sample size, involves a more homogeneous sample and used 3T MRI brain, which is more sensitive in detecting CMBs than 1.5T MRI. Furthermore, including evaluation of white matter hyperintensities and adjustments for all confounding factors for CMBs further strengthen the power of the study in evaluating the influence of NOAC exposure on prevalence of CMBs.
Contradicting other CMBs studies, history of hypertension did not predict presence of CMBs in our study. This is likely due to the very high prevalence of hypertension (>80% in those with and without CMBs) among patients with AF, which masks its distinctive association with CMBs. Also, contrasting to our understanding that hypertension is usually associated with deep CMBs,16 we observed that over 70% of the CMBs affected the lobar regions (with or without CMBs elsewhere), while only one-third of the CMBs were found in the deep region. This suggests that other risk factors, such as history of ischaemic and haemorrhagic stroke, underlying white matter changes, cerebral amyloid angiopathy, etc, may be contributing to CMBs observed in patients with AF. Furthermore, we found that patients in the NOAC group are more likely to have previous ICH. As many of these ICH were secondary to head injuries due to syncope from AF, they were less likely to be present in patients in the control group who were mostly recruited from acute stroke unit with acute ischaemic stroke and newly diagnosed AF.
Risk of anticoagulant-associated ICH is well known to be higher in Asians compared with Caucasians. This is partly due to the ethnic difference in genetic polymorphism of CYP2C9 and VKCORC1 that affect warfarin metabolism and sensitivity, respectively, as well as lower body weight in Asians.17HR of ICH with Asian patients taking warfarin was 4.06 compared with Caucasians.18 Data from meta-analysis have also shown that NOAC-associated ICH was numerically higher in Asians than non-Asians.19 This treatment-related complication is of particular concern in Asians, as Asians with CMBs have higher risk of ICH than Western populations (OR 3.87, 95% CI 0.91 to 16.40).20 Hence, it is important to develop better prognostic tools to estimate the risk of ICH in Asian patients with AF. MRI detection of CMBs provides direct evidence of asymptomatic microvascular leakage before the occurrence of clinical devastating ICH. Understanding the pathophysiology of CMBs and their correlation with anticoagulants exposure is important to determine how this emerging biomarker can help individualise stroke prevention strategies in patients with AF.
The interactive effects between CMBs, AF and anticoagulation is complex. On one hand, the presence of CMBs may indicate intrinsic bleeding-prone microangiopathy where small asymptomatic leaks could evolve into large haematoma when haemostasis is impaired under anticoagulation therapy. On the other hand, anticoagulation may enhance development of new CMBs by promoting red cell extravasation through damaged vessel walls. When these factors are added up together, the dynamic process can cause cumulative risk of ICH in patients under anticoagulation, where regular MRI brain may be helpful in monitoring for progression of CMBs, a surrogate marker for clinical ICH. In this study, there was no significant correlation observed between duration of NOAC exposure and prevalence of CMBs. As recent study has shown that CMBs could continue to exist for 9 years,21 prospective studies with follow-up MRI scans to compare for interval changes are needed to determine if patients on long-term NOAC are more likely to develop new CMBs with time. A number of observational studies including CROMIS-2 (The Clinical Relevance of Microbleeds in Stroke study), HERO (Intracerebral Haemorrhage due to Oral Anticoagulants), IPAAC, SAMURAI-NVAF (Stroke Acute Management With Urgent Risk-factor Assessment and Improvement Study on Anticoagulant Therapy in Nonvalvular Atrial Fibrillation) and CMB-NOW (Cerebral Microbleeds During NOACs or Warfarin Therapy in Nonvalvular Atrial Fibrillation Patients With Acute Ischaemic Stroke) etc are also underway to evaluate the risk of clinical ICH in patients with AF with CMBs.22 23
Strength of this study include prospective data collection and analysis, homogeneous sample including white matter evaluation in the analysis and a control group of patients with AF without prior oral anticoagulants use for comparison. Limitations of this study include, first, there were significantly more patients with prior stroke and antiplatelet use in the control group. As IPAAC study recruited patients with AF who were on or intended to start oral anticoagulants, patients in the control group were mainly those who were admitted with acute ischaemic stroke with known AF, many of them were on prior antiplatelet therapy, which accounts for the differences in risk factor profile. However, after adjusting for these confounding factors, and repeating the logistic regressions in NOAC group alone, the results remain the same, confirming the absence of any significant correlation between duration of NOAC exposure and prevalence of CMBs. Second, only 18.5% of patients were on NOAC for over 3 years in this study. Hence, it remains uncertain if longer duration of NOAC treatment may affect development of CMBs. Finally, we did not measure body weight of our patients in this study. Although dose of apixaban was adjusted partially by body weight, doses of dabigatran and rivaroxaban were determined mostly by renal function. Therefore, we were not able to address if body weight has any influence on the prevalence of CMBs among those treated by different NOAC.
In this study involving Chinese patients with AF who were treated with NOAC for up to 6.8 years, there was no significant correlation observed between duration of NOAC exposure and prevalence of CMBs. However, it remains uncertain if long-term treatment of NOAC can lead to development of new CMBs demanding need of regular MRI brain for monitoring of CMBs. Further studies involving patients with longer duration of treatment and follow-up MRI scans will be helpful in determining the dynamic interactions between NOAC use, CMBs and ICH risk.
Contributors YS conceptualised project, wrote protocol, performed literature review, reviewed radiological images, collected and analysed data and prepared manuscript. JA and WC provided radiological support. KTL designed data extraction form, collected data and performed all statistical analyses. LW rated white matter score. SFT rated white matter score, collected and analysed data. VI, KM, VM, SHM and FF assisted patient recruitment, collected data and reviewed manuscript. LW conceptualised project, reviewed protocol and manuscript. VM interpreted data and reviewed manuscript. TWL conceptualised project, reviewed protocol and manuscript.
Funding This study was supported by Health and Medical Research Fund and Kwok Tak Seng Centre for Stroke Research and Intervention.
Competing interests None declared.
Ethics approval New Territories East Cluster Clinical Research Ethics Committee (NTEC CREC).
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.