Warfarin, hematoma expansion, and outcome of intracerebral hemorrhage

Methods.

Study subjects consisted of prospectively identified consecutive ICH patients age ≥ 18 years recruited as participants in an ongoing cohort study of ICH outcome.4,13⇓ For this cohort study, consecutive patients with ICH are prospectively characterized and mortality is assessed at 3 months. For patients who die after hospital discharge, vital status is determined through follow-up phone calls.4 Patients with ICH secondary to antecedent head trauma, acute ischemic stroke with hemorrhagic transformation, brain tumor, vascular malformation, or vasculitis of the CNS are not enrolled.

The present study was restricted solely to patients within the cohort admitted directly to the Massachusetts General Hospital (MGH) Emergency Department (ED). Patients were therefore excluded if they presented to a community hospital prior to being transferred to the MGH ED. In addition, we excluded patients with symptoms > 72 hours in duration at presentation to the ED. Because of their relative rarity, their differing clinical courses, and the difficulty of reliably using CT to measure ICH volumes in the posterior fossa, patients with cerebellar and brainstem ICH were also excluded. The Institutional Review Board approved all aspects of this study.

To determine the effect of warfarin on baseline ICH volume, we analyzed consecutive patients age ≥ 18 years presenting to the MGH ED with primary supratentorial ICH (hemorrhage originating in the lobes or deep structures of the cerebral hemispheres) between January 1, 1998, and December 31, 2002 (baseline cohort, figure 1). Patient demographics and medical history were characterized as previously described.4 Time of ICH onset was defined as the time at which a subject or companion reported acute onset of a neurologic deficit. When time of onset could not be established, or when individuals awoke with a deficit, the time a subject was last known to be normal was considered the time of onset. Hemorrhages were categorized as warfarin-related if warfarin was reported as a regularly ingested medication. All recorded laboratory values were measured in the ED upon presentation. Glasgow Coma Score (GCS) was determined at the time of neurologic evaluation in the ED. All patients on warfarin received repeated doses of vitamin K and fresh-frozen plasma until the international normalized ratio (INR) of the prothrombin time was in the normal range, according to routine practices of the MGH Stroke Service. CT scans were repeated based on clinical judgment of the team caring for the patient.4

• Download figure
• Open in new tab
• Download powerpoint

Figure 1. Study subjects and design.

The effect of warfarin on hematoma expansion was determined in the subgroup of patients admitted between January 1, 2000, and December 31, 2002 (expansion cohort, see figure 1) because full sets of serial CT scans were not reliably available prior to 2000. For each patient we analyzed all CT scans obtained up to 7 days after hospital admission. Expansion was defined as any increase in volume ≥33% of baseline ICH volume determined from the admission CT scan.10 In order to assess the possibility of drift in the measure, we also identified patients with ICH volume reduction of ≥33% of baseline ICH volume. For warfarin patients, we recorded all serial INR values measured up to 7 days from admission. Patients who underwent craniotomy for hematoma evacuation during their hospital course were excluded from the analysis of hematoma expansion.

Results.

We examined the correlates of baseline ICH volume. Baseline CT scans obtained in the ED were available in 183/186 (98%) patients. Of these 23% (42/183) were taking warfarin at the time of ICH (table 1). Median time from symptom onset to baseline CT scan was 10.3 hours (range 25 minutes to 62 hours). Lobar hemorrhages were significantly larger than those centered in the thalamus and basal ganglia (table 2). Warfarin-related hemorrhages were not significantly larger than those unassociated with warfarin (p = 0.65). Among baseline characteristics, only serum glucose on admission correlated with hemorrhage volume (Spearman ρ = 0.3, p < 0.0001). This correlation was restricted to patients without diabetes (Spearman ρ = 0.5, p < 0.0001) and was not present in diabetic patients (Spearman ρ = 0.1, p = 0.51). There was no correlation between INR at presentation and baseline ICH volume in the entire baseline cohort (p = 1.0) or among the 42/183 (23%) patients on warfarin (p = 0.4). There was no effect of antiplatelet agent use, or admission systolic (p = 0.79), diastolic (p = 0.79), or pulse pressure (p = 0.75) on baseline hematoma volume.

We next examined whether the effect of warfarin on 3-month ICH mortality was dependent on baseline ICH volume. Three-month mortality was 35%. Warfarin, along with increasing age, baseline ICH and IVH volume, GCS < 9, and possession of the APOE ε2 allele predicted death in univariate analysis (table 3). There was no association between the APOE ε4 allele and mortality. In multivariable analysis warfarin significantly predicted fatal outcome even when controlling for baseline ICH and IVH volume. (ICH volume, IVH volume, and GCS also remained significant.)

We used the subgroup of patients with more than one CT scan to identify predictors of hematoma expansion (expansion cohort, see figure 1). Of 117 potentially eligible patients, 11 ultimately underwent craniotomy. Of the remaining 106 patients, 70 (66%) received at least two CT scans, 44 (42%) at least three scans, and 23 (22%) at least four scans (maximum of seven) during the first 7 days of hospitalization. Three-month mortality (53% vs 29%, p = 0.02), baseline ICH volume (50.6 ± 56.8 mL vs 21.6 ± 20.0 mL, p = 0.18), and proportion of patients with GCS < 9 (44% vs 11%, p < 0.001) were greater for subjects with single scans compared to those with two or more scans. Fifty percent of warfarin patients received a single scan as compared to 27% of non-warfarin patients (p = 0.05).

Hematoma expansion was detected in 16 (23%) of 70 patients with ≥ two CT scans. Median percentage increase in volume was 100% (range 33% to 867%). Three patients (4%) experienced ≥ 33% reduction in their hematoma volume (3 mL, 5 mL, 7 mL). Median time to detection of expansion was 17.0 hours after onset (range 2.4 to 60.8 hours). The median time to presentation to the ED did not differ between expanders and nonexpanders (4.7 vs 2.6 hours, p = 0.93). Although expansion was associated with increased 3-month mortality, the association was not significant (OR = 3.5, 95% CI = 0.7 to 18.9, p = 0.14).

Warfarin users were significantly more likely to undergo expansion than those not on warfarin at the time of ICH (7/13 [54%] vs 9/57 [16%], p = 0.007) (table 4). Use of an antiplatelet agent at the time of ICH was not associated with an increased risk of developing expansion. Systolic, diastolic, or pulse pressure on admission were not associated with expansion. There was no effect of glucose in either diabetic or non-diabetic patients. After controlling for time from symptom onset to initial CT scan, the effect of warfarin remained significant (OR = 10.4, 95% CI = 2.2 to 49.7, p = 0.003). Among warfarin patients with serial CT scans, there was no effect of baseline INR or time to INR normalization (INR ≤ 1.2) on hematoma growth.

Hematoma expansion was not only more frequent among warfarin users, but was also detected later in the hospital course. The median time to detection of expansion was 21.4 hours (range 4.6 to 60.8) for warfarin patients who developed ICH expansion compared to 8.4 hours (range 2.4 to 31.3) for those with ICH expansion unrelated to warfarin (figure 2).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2. Kaplan-Meier estimate of rate of hemorrhage expansion among subjects with hematoma expansion. Data are stratified according to whether patients were on warfarin at the time of ICH and include only those 16 patients with expansion. Testing for significance performed by log-rank method.

We performed a multivariable analysis of mortality in the 70 patients analyzed for hematoma expansion, incorporating the independent predictors identified in the baseline cohort of 183 patients. Warfarin predicted 3-month mortality (OR = 4.6, 95% CI = 1.0 to 21.8) when controlling for baseline ICH and IVH volume and GCS. When hematoma expansion was added to the model, the marginal effect of warfarin was reduced (OR = 3.0, 95% CI = 0.6 to 15.7).

Discussion.

The results of this study demonstrate that patients on warfarin at the time of their ICH are at increased risk for hematoma expansion. Expansion not only occurs more commonly in warfarin patients but is also found later in the hospital course, suggesting that warfarin patients experience prolonged bleeding. In fact, our data imply that warfarin patients are at risk for bleeding well beyond 24 hours, offering a generous theoretical time window for intervention with hemostatic therapy. Although it remains possible that unmeasured factors could also, in part, explain the increase in ICH mortality seen in warfarin patients, this ICH expansion is likely to be a crucial mechanism through which warfarin exerts its fatal effect. Indeed the trend toward increased fatality among expanders in our cohort is consistent with the results of larger retrospective studies that have found associations between hematoma expansion and clinical deterioration and mortality.9,11⇓ While larger studies will be necessary to evaluate whether rapid correction of warfarin effect with vitamin K, blood products, or targeted therapies such as recombinant factor VIIa can alter outcome, the connection between expansion and increased mortality underscores the opportunity for effective hemostatic therapies in all ICH patients at risk for expansion.18

Use of an antiplatelet agent at the time of ICH conversely was not associated with hematoma expansion in our cohort. This is consistent with the observed lack of effect of antiplatelet agents on baseline ICH volume as well ICH outcome.4 These data therefore do not support the use of platelet transfusions in patients who develop ICH while taking antiplatelet agents.

While increasing admission glucose correlated with baseline ICH volume in our cohort, it remains unclear whether hyperglycemia is simply a result of the hemorrhage or instead contributes to increased hemorrhage volume. The relationship between hyperglycemia and increasing ICH volume was restricted to patients without diabetes, although the small number of diabetic patients in the cohort may account for our inability to see a relationship. Both diabetes and hyperglycemia have been identified as independent risk factors for fatal outcome in ICH,4,19–21⇓⇓⇓ suggesting that they may act independently. Although it is plausible that large hemorrhage volumes trigger hyperglycemia, several observations suggest that hyperglycemia may itself predispose, or mark a predisposition, to increased bleeding.22–28⇓⇓⇓⇓⇓⇓ Clinical trials of aggressive glycemic control have demonstrated improved outcomes in critically ill patients.29 Such a clinical trial in acute ICH would clarify the importance of hyperglycemia in the clinical course of ICH.

Although included in our study because of its reported association with fatal outcome from ICH,16,17⇓ APOE ε4 did not predict ICH mortality. Instead, we found a possible association with the less common ε2 allele. Although this result is limited by the relatively small numbers of patients who donated blood for genotyping, the trends observed in multivariable analysis of mortality and analysis of ICH volume are all consistent with an effect of ε2 and not ε4. Our study, however, was not designed to assess the role of genetics in ICH outcome or hematoma volume. Rather, APOE was included solely as a potential covariate in the analysis of mortality. Clarification of the role of APOE in the clinical course of ICH will require further investigation.

An important limitation to the present study is the potential for unmeasured confounding that may result when care is not standardized throughout the cohort. Repeat CT scans were ordered at the discretion of the clinical teams caring for the patients, and we did find systematic differences between those patients receiving a single scan and those receiving more than one. Because those who received a single scan had an increased odds of fatal outcome, it is likely that the decision not to repeat a CT scan reflected early mortality or a conclusion on the part of the clinical team regarding the patient’s prognosis, extending to, perhaps, withdrawal of supportive care.30 ICH expansion that occurred in this group of patients would therefore go undetected. The most probable effect of this reduction in detection of expansion is underestimation of the amount of hematoma expansion in the cohort. This underestimate is likely to have been more severe for patients on warfarin, who were not only at greater risk for expansion, but also less likely to receive repeated CT scans. As a result, the bias introduced by inconsistencies in obtaining serial CT scans is likely to have obscured an even stronger association between warfarin and hematoma expansion rather than led to a false positive one. The potential for unmeasured confounding extends beyond the ordering of CT scans as there probably was other treatment variation among the cohort. Nonetheless, although the management of patients with ICH was not explicitly standardized, all patients received treatment within a single tertiary care center with an active neurologic intensive care unit and were managed by the same team of vascular neurosurgeons and vascular/critical care neurologists, suggesting such variation is likely to have been limited.

Although we did not find a relationship between the rapidity of INR correction and occurrence of hematoma expansion in the warfarin group, several additional limitations of the present study must be considered in interpreting this finding. As was the case with CT scans, serial INR measures were ordered by the clinical teams caring for the patients and not according to a standard protocol. In addition, they were not obtained simultaneously with repeat CT scans. Finally, the small number of warfarin patients in the expansion cohort left us with limited power to detect anything but large effects of INR correction on risk of ICH expansion in this subgroup. A prospective clinical trial will be required to determine the relationship between degree of INR correction and risk of hematoma expansion in warfarin-related ICH.

Hematoma expansion occurs commonly in patients with warfarin-related ICH and is likely to be an important contributor to the increased mortality of ICH in this setting. The presence of ICH expansion in warfarin patients, a subgroup currently excluded from the only ongoing clinical trial of hemostatic therapy in ICH,18 suggests a crucial opportunity for clinical trials targeting continued bleeding in patients with anticoagulant-related ICH. Simply reducing mortality to that of ICH unrelated to anticoagulation would be a major improvement.