Abstract
Background: Cerebral microbleeds are known to be indicative of bleeding-prone microangiopathy and may predict incident intracerebral hemorrhage (ICH). In this study, we investigated whether microbleeds are associated with the incidence of warfarin-related ICH.
Methods: Twenty-four patients with ICH while on outpatient treatment with warfarin were selected from a consecutive cohort. Control, warfarin-using subjects with no history of ICH were randomly selected during the same time period (n = 48). We compared demographic factors, vascular risk factors, laboratory findings, and radiologic findings including microbleeds between the groups.
Result: There were more cases of patients with microbleeds in the ICH than control group (79.2% vs 22.9%: p < 0.001), and the number of microbleeds was much higher for the ICH group (9.0 ± 26.8 vs 0.5 ± 1.03: p < 0.001). Moreover, the number of microbleeds was significantly correlated with the presence of warfarin-related ICH (r = 0.299; p < 0.001). Conditional logistic regression analysis showed that increased prothrombin time and the presence of microbleeds were independently related to the incidence of warfarin-related ICH (microbleeds: adjusted OR, 83.12).
Conclusion: This study suggests that underlying microbleeds are independently associated with an incidence of warfarin-related intracerebral hemorrhage. Future research should focus on elucidating the risks and benefits of warfarin medication in patients with microbleeds.
Glossary
- CI=
- confidence interval;
- GRE=
- gradient-echo;
- ICH=
- intracerebral hemorrhage;
- INR=
- international normalized ratio;
- NS=
- not significant;
- OR=
- odds ratio;
- PT=
- prothrombin time;
- WMH=
- white matter hyperintensity.
Spontaneous intracerebral hemorrhage (ICH) results from a rupture of the small penetrating cerebral arteriole. ICH accounts for 10 to 15% of acute first-ever strokes, and it reaches an incidence rate of 30% in Asian countries like Japan and Korea.1 Moreover, ICH is associated with the highest mortality of all cerebrovascular events, and most survivors never regain functional independence.2 The incidence of ICH is increasing in elderly people in some developed countries, in part because of the increased use of warfarin medication.3,4 Warfarin remains a highly effective therapy for prevention of thromboembolic strokes in common clinical situations like atrial fibrillation, but it increases both the risk of developing ICH and mortality. More severe outcomes in warfarin-related ICH are associated with an increased risk of hematoma expansion.5 Thus, the prevention of ICH in patients with warfarin medication is critical, and identification of risk factors for bleeding remains an important issue.
Cerebral microbleeds are seen as small round hypointense lesions on T2*-weighted gradient-echo (GRE) MRI, and they are pathologically tiny extravasations of blood from lipohyalinized cerebral arterioles.6 Hypertension,7 old age,8 low serum cholesterol,9 cerebral amyloid angiopathy,10 and glycated hemoglobin11 have been demonstrated as risk factors for these lesions. It is of potential importance that the lesions are closely associated with spontaneous ICH,12 and they may predict future occurrence or recurrence of ICH.13,14 These lesions are also associated with the severity of ICH.15 Microbleeds may be indicative of a bleeding-prone state in the brain, and they may be regarded as risk lesions of spontaneous ICH. However, it has been understood that the lesions are not associated with hemorrhagic transformation after acute ischemic stroke in patients treated with thrombolysis.16,17 This difference on predictability of subsequent hemorrhagic events may be related to differences of bleeding mechanism-reperfusion to the ischemic tissue (hemorrhagic transformation) vs rupture of arteriosclerotic arterioles (ICH).
To our knowledge, an association between microbleeds and incident warfarin-related ICH has not been studied. In this case-control study, we investigated whether microbleeds are associated with the incidence of ICH in patients taking warfarin medication.
METHODS
Study population.
All patients with ICH while on outpatient treatment with warfarin were selected from consecutive patients aged ≥45 years admitted to the Seoul National University Hospital with ICH. The electric medical record system in our hospital was used to identify patients between January 2002 and July 2007. We reviewed medical records to confirm whether patients were eligible for our study. Patients were excluded if their ICH was due to head trauma or acute ischemic stroke with hemorrhagic transformation. Patients with brain tumor, vascular malformation, vasculitis of the CNS, or invalid medical records were also excluded. We excluded patients who were not taking warfarin at the time of hemorrhage. From the database, 54 patients were matches for the diagnosis of symptomatic warfarin-related ICH. The initial provisional diagnosis of ICH was not confirmed by imaging studies in eight subjects. Twenty-two patients did not undergo brain MRI before or at the time of ICH incidence (economic problem, n = 13; metallic materials in the body [e.g., pacemaker], n = 3; poor cooperation, n = 6). Thus, 24 patients were finally included as cases for analysis (table e-1 on the Neurology® Web site at www.neurology.org). Characteristics of selected patients were not different from those of unselected patients (table e-2).
We selected control subjects from warfarin users with no history of ICH during the same time period. From the database, we found 1,970 patients who meet these conditions. We reviewed medical records to identify eligible controls. We excluded patients who had a previous ICH or who did not receive brain MRI. Two controls per a case were randomly selected, and subjects were matched for sex and age. Based on the patients’ medical records, most of the brain MR imaging studies were performed to evaluate suspicious neurologic symptoms (headache, dizziness, or subjective weakness). Finally, 48 patients were included as controls. All study procedures were approved by our institutional review board.
Clinical information.
We reviewed medical records for cases and controls to obtain demographic data, concurrent antiplatelet medication, previous medical history, and an indication for warfarin. Hypertension and coronary heart disease were defined as present if they had been diagnosed and treated by a physician. Diabetes was defined as present if the patient’s fasting glucose level was at least 7.0 mmol/L or if the patient had taken a hypoglycemic agent. A diagnosis of hypercholesterolemia was based on a history of hypercholesterolemia with medication or a fasting serum cholesterol level >6.2 mmol/L. A history of smoking was defined as present if the subject was a current smoker or an ex-smoker who had quit smoking within 5 years of admission. Indications of warfarin included atrial fibrillation, valvular heart disease, undetermined causes, and other-determined causes. The prothrombin time (PT)–international normalized ratio (INR) values for patients with ICH were recorded during presentation to the emergency room. For the matched control subjects, the PT-INR values were determined by selecting the date from our database that was closest to the date of emergency room presentation.
Image analysis.
MRI was performed on a 1.5 T superconducting magnet system (GE Medical System, Milwaukee, WI). GRE MRI was performed as a part of the routine protocol in our hospital, and the images were obtained in the axial plane with the following parameters: repetition time/echo time, 500/15 msec; flip angle, 26°; matrix size, 256 × 192; slice thickness, 6 mm; and gap width, 2 mm. Standard T2-weighted and fluid-attenuated inversion recovery (FLAIR) sequences were also obtained. Cerebral microbleeds were defined as well-defined focal areas of low signal on the GRE MRI of less than 5 mm in diameter, and they were counted throughout the whole brain by study neurologists (S.-H.L., W.-S.R.) blinded to the clinical characteristics. We did not include the microbleeds around the symptomatic ICH lesion because the lesions might be caused by the ICH itself. Interobserver agreement on the presence of microbleeds was found to be excellent (κ = 0.88). The presence and numbers of microbleeds were finally determined by consensus between the readers. Because some cases exhibited microbleeds and symptomatic ICH in the same GRE images, total blinding of the readers of the MRI for the presence of ICH was difficult. The GRE images were re-reviewed by other investigators (Drs. E.K. Bae and J.I. Cha) who did not know the hypothesis of this study and did not participate in this study as authors. Because interobserver agreement between the original data and the re-reviewed data were excellent (κ = 0.83), we do not believe that our reading was biased by the lack of blinding to the presence of ICH.
White matter hyperintensities (WMHs) were diagnosed according to the criteria of Fazekas et al.18 based on whether the subject had one of the following: 1) multiple periventricular hyperintense punctate lesions and early confluence or 2) multiple areas reaching confluence seen on T2-weighted or FLAIR images.
Statistical analysis.
For baseline comparisons, dichotomous variables were compared between groups using the χ2 test or Fisher exact analysis, and continuous variables (age, PT-INR, and number of microbleeds) were compared by the Wilcoxon rank sum test. To find a correlation between the number of microbleeds and the presence of warfarin-related ICH, we used Spearman correlation analysis. Finally, we used conditional logistic regression models with the matched sets to determine which variable was associated with an occurrence of ICH. In this analysis, WMHs and microbleeds were analyzed as dichotomous variables. A two-tailed p value of <0.05 was considered to be significant. Data analysis was performed with SPSS (version 13.0, Chicago, IL).
RESULTS
Characteristics of the patients included in this study are described in table 1. There were higher numbers of women (62.5% in both groups) with no significant difference. Hypertension was the most frequent vascular risk factor in both ICH case and control subjects, and smoking was more frequent in control subjects. However, none of the demographic variables were significantly different between groups. Warfarin was largely used for the prevention of stroke in the patients with atrial fibrillation or valvular replacement treatment. The durations of warfarin treatment were not different between the ICH cases and controls and, when we set the target range of PT-INR to 2.0–3.0 according to the general recommendation, the successful time of anticoagulation was not also different between the groups. PT-INR was significantly higher in the ICH cases than in the control subjects (3.22 ± 1.88 vs 2.21 ± 0.62; p = 0.001). In the patients with ICH, the location of ICH was quite typical: 8 patients (33.3%) had ICH in the lobar area, 7 (29.2%) in the basal ganglia, 2 (8.3%) in the thalamus, 1 (4.2%) in the brainstem, and 6 (25.0%) in the cerebellum. In addition, ratio of patients under adequate blood pressure control in the hypertensive subjects (<140/90 mm Hg) was 84.6% in the ICH cases and 82.6% in the controls. See table e-2 for comparisons with unselected cases (n = 22) and controls (n = 837).
Table 1 Baseline characteristics
As illustrated in table 2, there were more patients with microbleeds in the ICH than control group (79.2% vs 22.9%: p < 0.001), and the number of microbleeds was much higher in the ICH than control group (9.0 ± 26.8 vs 0.5 ± 1.03: p < 0.001). Moreover, Spearman correlation analysis revealed that a greater number of microbleeds was associated with a greater risk of ICH (r = 0.299; p < 0.001). Distribution of microbleeds appeared to be different between the groups. The most frequently involved area was the lobar area (58.3%) for ICH cases but thalamus (12.5%) for control subjects. In all the brain areas, the number of patients with microbleeds was larger for ICH cases than control subjects.
Table 2 Radiologic findings
Finally, we examined whether the effects of microbleeds on warfarin-related ICH were independent of other important clinical variables. Some variables were associated with warfarin-related ICH in the univariate analyses: PT-INR (odds ratio [OR], 2.51; 95% confidence interval [CI], 1.32–4.79), WMHs (OR, 3.00; 95% CI, 1.07–8.43), and the presence of microbleeds (OR, 12.78; 95% CI, 3.88–42.15). With regard to the location, microbleeds in the lobar area, basal ganglia, and cerebellum were significantly associated with warfarin-related ICH. Microbleeds in the brainstem and thalamus were not significantly associated (data not shown). In the conditional logistic regression analysis, the presence of microbleeds was independently related to warfarin-related ICH. Furthermore, the effect of microbleeds was the most powerful among the variables (table 3). In contrast, WMH did not remain significant after conditional logistic regression analysis (p = 0.093).
Table 3 Conditional logistic regression analysis
DISCUSSION
In this case-control study, we found that PT-INR and the presence of microbleeds on GRE MRI were independently associated with incident ICH in the patients undergoing warfarin treatment. In addition, the number of microbleeds was correlated with the incidence of warfarin-related ICH, and microbleeds in the lobar area and basal ganglia were more significantly associated. WMHs were associated with the incidence of warfarin-related ICH in a univariate analysis, but the significance did not remain after adjustment for PT-INR and the presence of microbleeds.
Cerebral microbleeds indicate previous extravasation of blood and signify bleeding-prone cerebral microangiopathy.6,15,19,20 It was first suggested that these lesions would predict all types of hemorrhages in the brain, including primary ICH and hemorrhagic transformation after acute ischemic stroke. With regard to hemorrhagic transformation after acute ischemic stroke, two observational studies19,21 indicated that presence of baseline microbleeds predicted incident hemorrhagic transformation in patients with or without thrombolytic treatment. However, recent prospective multicenter studies have failed to confirm these results in the patients who underwent IV thrombolysis.16,17,22 Furthermore, we recently found that the presence of baseline microbleeds is not associated with subsequent hemorrhagic transformation after acute atherothrombotic stroke regardless of thrombolytic therapy.23 Considering these results, microbleeds are not likely to be associated with the risk of hemorrhagic transformation in acute ischemic stroke. In contrast, microbleeds are closely associated with primary ICH. Several cross-sectional and prospective studies indicated that the presence or severity of microbleeds is strongly related to the incidence or recurrence of primary ICH.12–14 We believe that the differential effects of microbleeds on hemorrhagic transformation and ICH may be related to differences in the pathogenesis of hemorrhage. The underlying histopathologies of ICH include the rupture of penetrating arterioles damaged by hyaline degeneration and microaneurysm formation caused by longstanding hypertension or aging.24–26 This is quite the same in microbleeds, but the hemorrhagic transformation is caused by ischemic injury to the microvasculature in an extensive brain infarction.27 Warfarin-related ICH may share the mechanism of hemorrhage with primary ICH rather than hemorrhagic transformation.
In a pooled analysis of five primary stroke prevention trials, adjusted-dose warfarin reduced all strokes by approximately 70% in patients with atrial fibrillation.28 This preventive effect of warfarin is diminished by a significant 0.2% increase in the annual risk of ICH.29 Several risk factors for warfarin-related ICH have been reported. First, advancing age is one of the most important risk factors consistently reported by most studies.30,31 The relationship might be associated with an increased frequency of vascular rupture-related microangiopathic findings in cerebral amyloid angiopathy or hypertension in aged patients. In addition, racial background was also suggested as a new risk factor. A recent retrospective multiethnic study with a stroke-free atrial fibrillation cohort32 indicated that the risk of warfarin-related ICH was significantly higher in Hispanic (hazard ratio, 2.04) and Asian (hazard ratio, 4.06) than Caucasian white patients. Furthermore, an increased PT-INR value reaching 4.0 to 5.0 is another important risk factor for ICH.31,33,34 In many cases, however, warfarin-related ICHs occurred in the patients with the optimal range of PT-INR.35 In the present study, the PT-INR value in ICH cases was slightly increased (3.22 ± 1.88) compared to that of controls. Finally, underlying microangiopathy, such as cerebral amyloid angiopathy, may be associated with warfarin-related ICH; however, this causality has not yet been established. Aging is a global phenomenon, and hemorrhagic stroke caused by cerebral amyloid angiopathy would be more important than expected. In this context, the differential effects of cerebral amyloid angiopathy and hypertensive microangiopathy on warfarin-related ICH may be important in clinical practice. Thus, further studies on this issue should be conducted.
The Stroke Prevention In Reversible Ischemia Trial first suggested that white matter hypodensity lesions in CT scans might be associated with an increased risk of warfarin-related ICH, and this risk remained significant after adjustment for age, blood pressure, and PT-INR value.30 This suggestion was confirmed by a case-control study conducted in a single hospital.35 In the present study, we investigated an association between the presence of WMHs seen on T2-weighted or FLAIR images and the incidence of warfarin-related ICH. The effect of WMHs was significant in univariate analysis, but it did not remain significant after adjustment for possible confounders like the presence of microbleeds. These results signify that microbleeds are more specifically effective on incident warfarin-related ICH than WMHs. In fact, the presence and severity of microbleeds are strongly correlated with those of WMHs.36 However, microbleeds have a greater association with the incidence or recurrence of primary ICH than WMHs.37 This may arise because WMHs are closely associated with microangiopathic findings of ischemic nature whereas microbleeds represent a bleeding-prone microangiopathy.
There are important caveats in this study. First, in some patients, microbleeds and ICH were found on the same GRE images. The total blinding of presence of ICH was not possible in this context. To achieve sufficient validity, we invited two independent neurologists as reviewers who did not know the hypothesis of this study as indicated in Methods. As a result, we found that the re-reviewed data were nearly identical to the original review data. We do not believe that the lack of total blinding in this study seriously affected the study results. Second, because this study was conducted by a retrospective collection of data, selective brain MRI scanning possibly occurred in this situation. Moreover, the clinical practice guidelines have not recommended brain MRI scanning as a primary tool for diagnosis of acute hemorrhagic stroke. However, our hospital has performed brain MRI on all stroke patients regardless of stroke subtype to detect concomitant cerebrovascular lesions. Finally, because the cases and controls were only matched for age and sex, the groups were not fully matched for warfarin indications. Statistical methods were used to adjust for differences in warfarin indications, but we acknowledge that the adjustment may not be complete.
Microbleeds are a radiologic finding that is more often associated with ICH than other stroke subtypes. General application of brain MRI is not recommended for detecting the risk of ICH in the asymptomatic elderly population with vascular risk factors due to its high cost. If it is limited to the patients taking warfarin medication, however, brain MRI might be useful because warfarin-related ICH is a very critical complication in this population. Future research should focus on the risks and benefits of warfarin medication in patients with microbleeds.
ACKNOWLEDGMENT
The authors thank Drs. Eun-Gi Bae and Jung-In Cha for technical assistance.
Footnotes
-
*These authors contributed equally.
Supported by grants of the Korea Health 21 R&D Project, Ministry of Health and Welfare, Republic of Korea (A060143, A060171, and A060263).
Disclosure: The authors report no disclosures.
Supplemental data at www.neurology.org
Received April 21, 2008. Accepted in final form October 3, 2008.
REFERENCES
Disputes & Debates: Rapid online correspondence
REQUIREMENTS
If you are uploading a letter concerning an article:
You must have updated your disclosures within six months: http://submit.neurology.org
Your co-authors must send a completed Publishing Agreement Form to Neurology Staff (not necessary for the lead/corresponding author as the form below will suffice) before you upload your comment.
If you are responding to a comment that was written about an article you originally authored:
You (and co-authors) do not need to fill out forms or check disclosures as author forms are still valid
and apply to letter.
Submission specifications:
- Submissions must be < 200 words with < 5 references. Reference 1 must be the article on which you are commenting.
- Submissions should not have more than 5 authors. (Exception: original author replies can include all original authors of the article)
- Submit only on articles published within 6 months of issue date.
- Do not be redundant. Read any comments already posted on the article prior to submission.
- Submitted comments are subject to editing and editor review prior to posting.