Abstract
Background: Intracerebral hemorrhage (ICH) is the most feared complication of warfarin therapy. The pathogenesis of this often-fatal complication remains obscure. Cerebral amyloid angiopathy (CAA) is a major cause of spontaneous lobar hemorrhage in the elderly and is associated with specific alleles of the APOE gene.
Objective: To assess the role of CAA in warfarin-associated ICH.
Methods: Clinical characteristics and APOE genotype were compared between 41 patients with warfarin-related ICH (from a cohort of 59 consecutive patients aged ≥65 years with supratentorial ICH on warfarin) and 66 randomly selected individuals aged ≥65 years without ICH taking warfarin. In addition, all neuropathologic specimens from ICH patients were reviewed for the presence and severity of CAA.
Results: Hemorrhages tended to be in the lobar regions of the brain, and most (76%) occurred with an international normalized ratio of ≤3.0. The APOE ε2 allele was overrepresented among patients with warfarin-associated lobar hemorrhage (allele frequency 0.13 versus 0.04 in control subjects; p = 0.031). After controlling for other variables associated with ICH, carriers of the ε2 allele had an OR of 3.8 (95% CI, 1.0 to 14.6) for lobar ICH. CAA was pathologically diagnosed as the cause of lobar hemorrhage in 7 of 11 patients with available tissue samples.
Conclusions: CAA is an important cause of warfarin-associated lobar ICH in the elderly. Although diagnosis of CAA before hemorrhage is not yet possible, these data offer hope that future patients at high risk for hemorrhage may be identified before initiation of warfarin therapy.
Intracerebral hemorrhage (ICH) is the most feared complication of warfarin therapy, with a mortality of ∼60%.1 Because of its high morbidity and mortality, the risk of warfarin-associated ICH can tip the balance in favor of withholding warfarin from patients who might otherwise be candidates for anticoagulant therapy. A risk for ICH of 2.0% per year, for example, would be enough to offset the benefit of warfarin for stroke prevention in patients with atrial fibrillation.2
Despite accumulating knowledge of the risk factors for warfarin-associated ICH, our understanding of its cause and whether it can be predicted or prevented remains limited. The annual risk of ICH among patients receiving warfarin is estimated to be between 0.2 and 0.6%, with higher rates noted in clinical trials using international normalized ratio (INR) target ranges exceeding 4.0.3-7⇓⇓⇓⇓ Case-control studies have pointed to several clinical risk factors
for warfarin-associated ICH, including age of ≥65 years, past history of cerebrovascular disease, and intensity of anticoagulation.5,8-11⇓⇓⇓⇓ The existence of specific risk factors for warfarin-related hemorrhage suggests that this event might result from particular pathologic processes in cerebral vessels rather than as a random occurrence.
The current study concentrates on one potential underlying vascular mechanism for warfarin-related hemorrhage: cerebral amyloid angiopathy (CAA). CAA causes hemorrhages located primarily in the cortical/subcortical (or lobar) regions of the cerebral hemispheres. This predilection for a lobar location is likely the consequence of the fact that CAA most commonly and severely involves the superficial vessels of the cortex (as opposed to the penetrating vessels of the basal ganglia, thalamus, and brainstem typically involved by hypertensive vasculopathy in the brain). There are several features of CAA that suggest it may be one of the pathologic processes underlying warfarin-associated ICH. It is associated with multiple, often asymptomatic microhemorrhages and, like warfarin-related ICH, increases in prevalence with age.12,13⇓ Furthermore, CAA has been identified among individual patients who have developed ICH as a complication of anticoagulation or thrombolysis,14,15⇓ although no systematic search for its presence in patients with warfarin-related ICH has been undertaken.
We report a prospective case-control study using two independent approaches of assessing the role of CAA in warfarin-associated ICH: correlation of ICH with the genetic risk factor APOE and neuropathologic analysis of available specimens from hemorrhage patients. APOE is a widely distributed protein involved in the mobilization and redistribution of cholesterol and phospholipid.16 The frequencies of the three human alleles of APOE (ε2, ε3, ε4) have been studied in populations with AD and CAA. CAA appears to associate with elevated frequencies of the APOE ε4 allele (as in AD) as well as the ε2 allele (differing from AD).17-22⇓⇓⇓⇓⇓ Based on these associations, we used the distribution of APOE genotypes in our study group as a surrogate marker for the aggregate presence of CAA. We hypothesized that if CAA were overrepresented among patients with warfarin-associated ICH, then the frequencies of APOE ε2 or ε4 should be elevated relative to age-matched control subjects taking warfarin without ICH.
Methods.
Patient recruitment and genetic analysis.
Cases.
Between July 1, 1994, and December 31, 1998, we prospectively identified 57 patients with warfarin-associated ICH among all patients aged ≥65 years with ICH admitted to Massachusetts General Hospital (MGH). We also performed a retrospective review of discharge diagnoses for the same period and identified two additional patients. To be eligible for the study, patients had to be taking warfarin at the time of the hemorrhage and the hemorrhage had to be located in the cerebral hemispheres, basal ganglia, or thalamus. Patients with antecedent head trauma, acute ischemic stroke with hemorrhagic transformation, brain tumor, vascular malformation, or vasculitis of the CNS were excluded.
All patients received CT scans of the brain upon admission. Hemorrhages were prospectively classified as either lobar or deep on the basis of the CT scan. Hemorrhages centered in the subcortical white matter of the frontal, parietal, temporal, or occipital lobes were defined as lobar hemorrhages. Hemorrhages of the thalamus and basal ganglia were identified as deep hemorrhages.
Among the 59 consecutive patients with primary supratentorial ICH, 41 donated blood for APOE genotype analysis. Of the 18 patients in whom APOE genotype was unavailable, 10 died before blood could be obtained (six lobar, four deep), five declined to donate a blood sample (two lobar, three deep), and three were discharged before donating blood and could not subsequently be located (one lobar, two deep). Compared with the 41 patients with available blood samples, the 18 patients without samples were more likely to be hypertensive and to have deep hemorrhages but did not differ significantly in age, gender, history of stroke, or INR (data not shown).
Each patient or family member was interviewed by a neurologist regarding medications, indication for warfarin, and features of the medical history. Medical records were also reviewed to determine history of hypertension or prior stroke. The INR recorded was the value determined in the emergency room upon presentation.
Determination of APOE genotype was performed without knowledge of the patient’s clinical history by restriction enzyme digestion of an APOE PCR product prepared from blood samples, as described previously.19 In brief, DNA was extracted from blood, amplified with APOE primers, and digested with HhaI to produce DNA fragments characteristic of the ε2, ε3, or ε4 alleles.
Control subjects.
Control patients were randomly selected from among the patients enrolled in the MGH Anticoagulation Therapy Unit (MGH-ATU) who were ≥65 years old. No control patient had a history of intracerebral hemorrhage. Study recruitment letters were sent to 129 patients, and 66 agreed to provide a blood sample for APOE genotyping. History of hypertension and previous ischemic stroke was obtained from review of patient medical records. The INR value for each control patient was chosen from among all INR determinations in the MGH-ATU database by matching calendar date to that of a randomly selected case patient’s admission date. The INR in the database closest to the month and year of the admission date of the matched case patient was selected for each control.
The primary indication for warfarin was atrial fibrillation in a majority of both the ICH cases (35/59, 59%) and control cases (40/66, 61%). Indications for the remaining patients were distributed among cerebrovascular, cardiovascular, and peripheral vascular disorders.
Pathology.
Tissue specimens were available from 13 patients in the 59-patient cohort who: 1) died and underwent autopsy; 2) had clinical deterioration leading to urgent craniotomy and hematoma evacuation; or 3) were referred by their physician for diagnostic brain biopsy. We reviewed all 13 available specimens. Eight samples were obtained during life (six from hematoma evacuation, two from cortical biopsy) and five at postmortem examination. All procedures except one (a patient readmitted for recurrent hemorrhage) were performed during the index hospitalization. Two of the hematoma specimens did not contain blood vessels and were dropped from the analysis, leaving 11 pathologic samples for evaluation, all from patients with lobar hemorrhage. Ten of the 11 patients also donated blood for APOE genotype determination.
Samples were examined under light microscopy by routine histopathologic stains, including Congo red. The presence of β-amyloid was confirmed by immunocytochemistry.23 Among patients who underwent postmortem brain examination, hemorrhages were judged to be caused by CAA in the presence of severe CAA pathology, defined by at least one focus (among all examined tissue sections) of complete replacement of the vessel wall by amyloid, cracking of the media from the intima, and paravascular blood leakage.19 Because biopsy or hematoma evacuation yields only a single tissue specimen with a limited number of vessels, hemorrhages examined by these methods were judged to be related to CAA in the presence of any vascular amyloid staining positively with Congo red. This approach has been shown to correlate to CAA-related hemorrhage at autopsy with high sensitivity and specificity.23
Statistical analysis.
Categorical variables were compared between groups using the Fisher exact test. APOE allele frequencies, calculated as the proportion of chromosomes (two per patient) composed of a given APOE allele, were compared among groups by the same method. Age (displayed as mean ± SD) was analyzed as a continuous variable by t-test. Multivariate analysis for odds of ICH was performed by logistic regression, treating APOE ε2 as a categorical variable (present or absent) and controlling for age, INR (as a continuous variable), and history of hypertension or stroke. A two-tailed p value of 0.05 was required for significance. All analyses were performed with Stata software (Stata Corp., College Station, TX).
This study was performed with the approval of and in accord with the guidelines of the MGH Institutional Review Board and with informed consent of all participants or family members.
Results.
Fifty-nine consecutive patients aged ≥65 years (mean 77.3 ± 7.0 years) with warfarin-related supratentorial hemorrhages were evaluated. Forty-one (69%) of the hemorrhages were centered in lobar rather than deep brain territories. Twenty patients (34%) died during their hospitalization, with similar mortalities noted among lobar (34%) and deep (33%) ICH.
Clinical characteristics and APOE genotype.
Forty-one ICH patients donated blood samples for determination of APOE genotype. The clinical features and APOE data from this group were compared with those from 66 control patients taking warfarin without ICH (table) . Among the clinical features examined, a higher proportion of patients with lobar ICH had a history of prior stroke, and a higher proportion with deep ICH had hypertension, than control subjects. ⇓
Clinical characteristics of genotyped patients: hemorrhages versus controls
Comparison of allele frequencies demonstrated an association between ICH and the APOE ε2 allele. The frequency of the ε2 allele was 0.11 in cases compared with 0.04 for control subjects (p = 0.048). The elevation in ε2 was largely owing to the subgroup with lobar hemorrhage. When cases were separated into those with lobar and deep ICH, the frequency of ε2 among patients with deep hemorrhage was 0.06, approximating that of the control group, whereas the frequency of ε2 in patients with lobar hemorrhage was 0.13 (p = 0.031 versus control subjects). In multivariate analysis controlling for age, hypertension, history of stroke, and INR, carriers of the ε2 allele had an OR for lobar hemorrhage of 3.8 (95% CI 1.0 to 14.6). The frequency of ε4 among all patients with ICH as well as those with lobar hemorrhage did not differ significantly from that among control subjects.
Neuropathologic analysis.
Tissue specimens containing cerebral blood vessels were available from 11 patients with lobar hemorrhage and none with deep hemorrhage. Pathology from seven of the 11 specimens (three of five autopsies, one of two biopsies, and three of four evacuations) indicated hemorrhage related to CAA. The four tissue specimens without CAA did not demonstrate an alternative cause for ICH when examined by standard techniques. The frequency of APOE ε2 was 0.29 (4 of 14 chromosomes) in the seven patients with pathologically manifest CAA and 0 (0 of 6 chromosomes) in the three patients with negative pathology and available DNA samples.
Discussion.
The major conclusion from our data is that CAA is an important contributor to warfarin-associated lobar hemorrhage. This is supported by two lines of evidence: genetic data showing significantly elevated frequency of the APOE ε2 allele and histopathologic demonstration of CAA in 7 of 11 patients with available tissue samples. We interpret the genetic data as indicating a high prevalence of CAA in the cohort, as the ε2 allele appears to associate specifically with hemorrhages related to CAA.21,22,24⇓⇓ Our conclusion is consistent with anecdotal reports of CAA as the underlying cause of hemorrhage in two patients, one taking heparin and the other (who is included in the current series) warfarin,14,25⇓ although it will be important to confirm it in other larger series of patients with warfarin-associated ICH.
Hart et al.1 have proposed that anticoagulation with warfarin causes small hemorrhages that might otherwise remain asymptomatic to become large, symptomatic hemorrhagic stroke. This model is fully consistent with a role for CAA. Small, clinically silent hemorrhages are a recognized feature of advanced CAA and indeed appear to occur at a severalfold greater rate than symptomatic hemorrhages.26,27⇓
We note that the majority of hemorrhages in our cohort (as in a previously described cohort)9 occurred with levels of anticoagulation considered therapeutic (i.e., INR ≤3.0). This observation supports the possibility that most warfarin-associated hemorrhages occur from a particular vascular predisposition rather than from excessive anticoagulation. It is clear, however, that CAA is not the only mechanism involved in warfarin-related ICH. The four cases without definite vascular pathology (including two full-brain autopsies) indicate that even in the lobar regions, warfarin-related hemorrhage can occur in the absence of CAA. We also did not find genetic evidence for involvement of CAA in deep hemorrhages, consistent with the tendency of CAA to spare blood vessels in these deep brain regions.13
The APOE ε4 allele, though also associated with CAA-related hemorrhage,19,20⇓ was not increased in frequency in the warfarin-related hemorrhage group. One possible explanation for an association with APOE ε2 but not ε4 is that the ε2 allele might bear some specific relationship with hemorrhages caused by warfarin. This possibility, previously raised by another series of hemorrhage patients,28 could relate to the association between ε2 and particular vasculopathic processes in amyloid-laden vessels such as cracking and fibrinoid necrosis.22,29⇓ An alternative explanation for the absence of elevated ε4 frequency in this cohort is the selection bias favoring patients who survived their hemorrhage—10 of the 20 patients who died from their hemorrhage did not donate blood for genotype analysis before death. Several studies have suggested that APOE ε4 predisposes to high mortality following hemorrhagic stroke,30,31⇓ offering a potential mechanism by which this allele might have been selectively excluded from our analysis. We indeed observed a relatively high APOE ε4 allele frequency (0.25) in the remaining 10 patients who were able to donate blood samples before in-hospital death.
Despite the consecutive selection of hemorrhage patients in our study, there remains the possibility of selection bias. This issue is particularly relevant to the pathologic samples, which were drawn exclusively from lobar hemorrhage patients who came to surgical procedures or autopsy. It is thus possible that patients with less severe hemorrhages that were neither lethal nor life threatening might have lower rates of CAA than the group with available pathology. In this regard, it is reassuring that genetic data from a larger subset of the cohort also implicated CAA in lobar hemorrhage. Another indication of the role of CAA in the cohort as a whole is the high proportion of lobar hemorrhages (69%), a trend noted in some previous series.32
A role for CAA in warfarin-related ICH is also consistent with the observed increase in risk of this complication with advancing age.9-11⇓⇓ The prevalence of CAA is markedly age dependent. It can be identified pathologically in ∼35% of brains from individuals aged ≥85 years, one-third of whom are affected to the severe extent associated with hemorrhage.23 The elevated risk of CAA and ICH in the elderly is counterbalanced, however, by a parallel increase with age in risk of thromboembolic stroke. In atrial fibrillation, for example, pooled data from randomized trials demonstrated an annual risk of stroke in patients >75 years old of 3.5 to 8.1%. This was reduced to <2% with the administration of warfarin.33 The annual rate of ICH on warfarin is also relatively high in this age group, e.g., reaching 1.8% in the Stroke Prevention in Atrial Fibrillation II study.5 The benefits as well as the risks of warfarin thus appear to be greatest for the oldest patients.
The possibility that warfarin-associated ICH may arise from a specific vasculopathic process such as CAA offers the hope that patients at risk can be identified before initiating therapy. Improved identification of patients who will bleed on warfarin would almost certainly allow individuals at lower risk to use the drug with improved therapeutic index and for a wider range of indications. There is unfortunately no accurate diagnostic technique for CAA at its presymptomatic stage. APOE genotype, although predictive of recurrent lobar hemorrhage,27 is not diagnostic of CAA and should not, in our opinion, be used to screen for risk of anticoagulation. Future progress in the early diagnosis of CAA, however, is likely to have an important impact on the risk–benefit calculation for use of warfarin.
Acknowledgments
Supported by grants from the NIH (AG00725), American Heart Association, and Edward Mallinckrodt, Jr., Foundation.
Acknowledgment
The authors thank Dr. Jean-Paul Vonsattel for assistance in reviewing pathologic specimens, Dr. Yuchiao Chang for statistical advice, and Ms. Katherine Knudsen for help in gathering patient records.
Footnotes
-
See also page 907
-
J.R. received the 2000 American Academy of Neurology Founders’ Award for this work.
- Received February 29, 2000.
- Accepted June 5, 2000.
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.