Predicted log-mean hematoma volume applying the fitted models. Results are presented stratified by INR category and by ICH location.
eTable.Comparison Between Groups With Available and Missing Data
Falcone GJ, Biffi A, Brouwers HB, Anderson CD, Battey TWK, Ayres AM, Vashkevich A, Schwab K, Rost NS, Goldstein JN, Viswanathan A, Greenberg SM, Rosand J. Predictors of Hematoma Volume in Deep and Lobar Supratentorial Intracerebral
Hemorrhage. JAMA Neurol. 2013;70(8):988-994. doi:10.1001/jamaneurol.2013.98
Copyright 2013 American Medical Association. All Rights Reserved. Applicable
FARS/DFARS Restrictions Apply to Government Use.
Hematoma volume is the strongest predictor of outcome in intracerebral hemorrhage (ICH). Despite
known differences in the underlying biology between deep and lobar ICHs, limited data are available
on location specificity of factors reported to affect hematoma volume.
To evaluate whether determinants of ICH volume differ by topography, we sought to estimate
location-specific effects for potential predictors of this radiological outcome.
Prospective cohort study.
Academic medical center.
A total of 744 supratentorial primary ICH patients (388 deep and 356 lobar) aged older than 18
years admitted between January 1, 2000, and December 31, 2010.
Main Outcomes and Measures
Intracerebral hemorrhage volume measured from the computed tomography scan obtained on
presentation to the emergency department. Linear regression analysis, stratified by ICH location,
was implemented to identify determinants of log-transformed ICH volume.
Median ICH volume was larger in lobar hemorrhages (39 mL; interquartile range, 16-75 mL) than in
deep hemorrhages (13 mL; interquartile range, 5-40 mL; P < .001). In
multivariable linear regression, independent predictors of deep ICH volume were intensity of
anticoagulation (β = 0.32; standard error [SE] = 0.08;
P < .001; test for trend across 4 categories of the international
normalized ratio), history of coronary artery disease (β = 0.33;
SE = 0.17; P = .05), male sex (β = 0.28;
SE = 0.14; P = .05), and age
(β = −0.02; SE = 0.01; P = .001).
Independent predictors of lobar ICH volume were intensity of anticoagulation
(β = 0.14; SE = 0.06; P = .02) and
antiplatelet treatment (β = 0.27; SE = 0.13;
P = .03).
Conclusions and Relevance
Predictors of hematoma volume only partially overlap between deep and lobar ICHs. These findings
suggest that the mechanisms that determine the extent of bleeding differ for deep and lobar ICHs.
Further studies are needed to characterize the specific biological pathways that underlie the
The biological pathways that lead to intracerebral hemorrhage (ICH) differ depending on the location of the hemorrhage.1 Previous genetic studies indicate that these differences in underlying mechanisms also affect the size of the hematoma.2 The volume of intraparenchymal bleeding measured on hospital admission is the strongest predictor of clinical outcome in ICH.3 Hematoma volume at presentation is, in turn, associated with hematoma expansion,4 another important determinant of poor outcome in this condition. Owing to their importance as prognostic factors, ICH volume and expansion have been targeted by 2 different therapeutic interventions assessed in clinical trials: administration of activated recombinant factor VII5,6 and aggressive blood pressure (BP) lowering.7,8
Previous studies identified bleeding location as a potent influencing factor of ICH volume. Lobar hemorrhages were consistently found to be larger than those located in deep structures of the brain.9 A number of additional predictors of supratentorial ICH volume have been described: warfarin treatment,9 statin treatment,10 leukoaraiosis,11 and asymptomatic microbleeds.12 However, whether these factors influence bleeding, regardless of ICH location, remains unknown.
We hypothesized that biological pathways determining the volume of the hematoma differ between deep and lobar ICHs. To test this hypothesis, we undertook the largest single-center study to date of ICH volume to establish whether predictors of deep and lobar ICH volume overlap completely, partially, or not at all. To achieve this goal, we sought to estimate location-specific effects for each potential predictor considered in the study.
We performed a retrospective analysis of data drawn from an ongoing prospective cohort study of ICH performed at Massachusetts General Hospital in Boston. The study was approved by the hospital’s institutional review board, and written informed consent was obtained from all participants or their surrogates when patients were unable to provide consent.
Cases of ICH were ascertained by stroke neurologists and enrolled according to methods previously described.13 Study subjects were consecutive white patients admitted to the Massachusetts General Hospital from January 1, 2000, to December 31, 2010, with primary and warfarin-related supratentorial ICH. Intracerebral hemorrhage was defined as a new and acute (<24 hours) neurological deficit with compatible brain imaging showing the presence of intraparenchymal bleeding. Inclusion criteria were age older than 18 years and confirmation of ICH through computed tomography (CT). Exclusion criteria included trauma, brain tumor, hemorrhagic transformation of a cerebral infarction, vascular malformation, or any other suspected cause of secondary ICH.
Patients (or their families or surrogates) were interviewed to determine age; sex; time of ictus; medical history; family history; pre-ICH treatment with warfarin, antiplatelets, antihypertensives, and statins; and alcohol and tobacco use. Hospital records were reviewed for Glasgow Coma Scale score on arrival, admission arterial BP, time to baseline imaging, routine laboratory measurements, and in-hospital mortality. For patients who survived and were discharged, 90-day mortality was assessed by telephone by trained study staff and supplemented by regular surveillance of the Social Security Death Index.14
For all study subjects, the first available CT was evaluated. Intracerebral hemorrhage location was assigned based on admission CT by study neurologists blinded to clinical data. Intracerebral hemorrhage exclusively involving the thalamus, basal ganglia, internal capsule, and deep periventricular white matter was defined as deep ICH, whereas ICH originating at the cortex and cortical-subcortical junction was defined as lobar ICH. Hemorrhages involving more than 1 territory were defined as mixed ICH. Differences in ICH location were adjudicated by consensus. Intracerebral hemorrhage volume was measured using Alice (Parexel International) and Analyze version 9.0 (Mayo Clinic) software, using previously described methods.15 Intraventricular bleeding was not included in volume calculations. Subjects with mixed (n = 10) and primary intraventricular (n = 52) hemorrhages, as well as those initially imaged beyond 72 hours of symptom onset (n = 35), were excluded from the analysis. Unless contraindicated, imaging of the intracranial vasculature (CT angiography, conventional angiography, or magnetic resonance angiography) was performed to rule out secondary causes of ICH.
Discrete variables are expressed as count (percentage) and continuous variables as mean (standard deviation) or median (interquartile range [IQR]), as appropriate. Intracerebral hemorrhage volume was natural-log transformed to approximate normality. Following previous reports,9 subjects not taking warfarin with missing international normalized ratio (INR) values were assigned a value of 1, and categories were created based on the following values of the admission INR: less than or equal to 1.2, greater than 1.2 to less than 2.0, greater than or equal to 2.0 to less than or equal to 3.0, and greater than 3.
Linear regression was used to model mean log-ICH volume. Univariable linear regression was initially implemented to evaluate unadjusted associations between log-ICH volume and covariates. Subsequently, multivariable linear regression was applied to identify independent associations after accounting for potential confounders. Multivariable models were constructed separately for deep and lobar ICHs. Model building proceeded as follows: age and sex were included in all models; covariates with P values less than .20 in univariable analysis were entered into the model and backward eliminated to a significance level of .20; finally, collinear factors (as measured through the variance inflation factor) were removed when appropriate. Time from symptom onset to first CT scan (time to scan) was modeled as a continuous variable. For INR, the bin of less than or equal to 1.2 was set as the reference category.
Mortality at 90 days was evaluated by fitting logistic regression models that incorporated the same covariates used in linear regression analyses. Subsequently, ICH volume was included in the model as a predictor to assess its role as a mediator of identified associations.
To assess the possibility of missing data bias, we compared baseline characteristics between the initial cohort that met the inclusion criteria and the final cohort that had available data for both ICH volume and time to scan. Differences in ICH volume between subjects treated with warfarin and those treated with warfarin plus antiplatelets were evaluated in both unadjusted (t test between these 2 groups) and adjusted (interaction term for warfarin × antiplatelets) analyses.
All statistical analyses were performed using SAS version 9.3. The results of the statistical tests were considered significant at P < .05 (2-tailed).
A total of 1030 subjects with ICH fulfilled the inclusion criteria and were enrolled during the study. Of these, 45 and 241 had missing data for ICH volume and time to scan, respectively. After these exclusions, 744 (mean [SD] age, 74  years; 47% female ) were included in the present analysis, comprising 388 deep and 356 lobar ICHs (Table 1). Subjects with complete and missing data differed only on history of coronary artery disease (eTable in Supplement).
Hemorrhage volumes in the lobar regions were larger than those in deep hemispheric locations, while intraventricular extension of the hemorrhage, when it occurred, was larger in volume for deep hemispheric ICH. The median hematoma volumes were 13 mL (IQR, 5-40 mL) and 39 mL (IQR, 16-75 mL) for deep and lobar ICHs, respectively (P < .001). The median intraventricular hemorrhage volumes were 15 mL (IQR, 5-42 mL) and 6 mL (IQR, 2-18 mL) for deep and lobar ICHs, respectively (P < .001). Large hemorrhages (30-60 mL) occurred in 313 subjects (42%) and massive bleeding (>60 mL) occurred in 246 cases (33%). Median time to scan was 5 hours (IQR, 2-8 hours) and 6 hours (IQR, 4-12 hours) for deep and lobar ICHs, respectively. A total of 143 patients (19%) were being treated with warfarin when the hemorrhage took place, and 115 (15%) had an INR greater than 2.0. Overall, 290 patients (39%) did not survive to 90 days. There was no statistically significant difference in 90-day mortality between deep (141 deaths, 36%) and lobar (149 deaths, 42%) locations.
Predictors of hematoma volume in deep ICH identified through univariable analysis were age, sex, coronary artery disease, atrial fibrillation, warfarin treatment, intensity of anticoagulation (expressed by the INR), admission blood glucose, admission systolic BP, admission diastolic BP, and time to scan (Table 2). As in previous reports,9 admission measures of systolic BP, diastolic BP, and blood glucose were not included in multivariable models because measurements were ascertained after the ICH had occurred, rendering it impossible to determine the causal relationship between them. Warfarin treatment and atrial fibrillation were also excluded owing to collinearity with INR, the latter associated with the largest effect on ICH volume. Of the predictors previously described, age, sex, intensity of anticoagulation, and time to scan remained significant in multivariable linear regression analysis (Table 3). Percentage changes in mean ICH volume estimated through multivariable linear regression were a 2% decrease per additional year of age (β = −0.02; standard error [SE] = 0.01; P = .001); a 28% increase by male sex (β = 0.28; SE = 0.14; P = .05); a 33% increase by prior diagnosis of coronary artery disease (β = 0.33; SE = 0.17; P = .05); 58%, 71%, and 85% increases by the INR categories of greater than 1.2 to less than 2.0, greater than or equal to 2.0 to less than or equal to 3.0, and greater than 3.0, respectively (reference INR, ≤1.2; test for trend across INR categories; P = 5 × 105); and a 3.6% decrease per additional hour in time to scan (β = −0.04; SE = 0.008; P = 3 × 106). The dose-response curve describing the effect of the INR on ICH volume followed a linear pattern (Figure). No interaction was found between warfarin and antiplatelet medication (P = .45).
Predictors of lobar ICH volume identified in univariable analysis were previous ICH, intensity of anticoagulation (as measured by the INR), antiplatelet treatment, and time to scan (Table 2). Of these, intensity of anticoagulation, antiplatelet treatment, and time to scan remained significant in multivariable analysis (Table 4). Percentage changes in mean ICH volume estimated through multivariable linear regression included a 27% increase by antiplatelet treatment (β = 0.27; SE = 0.13; P = .03), an 85% increase by intensity of anticoagulation (when comparing INR ≤ 1.2 to INR > 3; β = 0.85; SE = 0.26; P = .001), and a 2.8% decrease per additional hour to scan (β = −0.028; SE = 0.006; P = 34 × 106). In contrast to deep ICH, for lobar hemorrhages, there was a threshold effect at INR greater than 3 for the effect of anticoagulation on hematoma volume (Figure). No interaction was found between warfarin and antiplatelet medication (P = .30).
As expected, location-specific determinants of ICH volume also influenced clinical outcome. Independent predictors of mortality in deep ICH were age, history of coronary artery disease, intensity of anticoagulation, and time to scan (Table 3). Age, intensity of anticoagulation, and time to scan were predictive of mortality in lobar ICH (Table 4). For both deep and lobar ICH volume, only age remained significant after including ICH volume in the model (data not shown).
The results of the current study demonstrate that the determinants of hematoma volume differ by the location in ICH. These findings indicate that the different biological pathways underlying the development of deep and lobar ICHs not only determine ICH risk, but also the volume of bleeding once the hemorrhage has taken place. The different effect of antithrombotic medications on ICH volume is probably the most striking finding of the study. The intensity of anticoagulation with warfarin, as measured by the INR, was independently associated with ICH volume in lobar and deep locations, although the effect size and shape of the dose-response curve were different at each location (Figure). On the other hand, use of an antiplatelet agent had a substantial association with lobar ICH volume and had little, if any, relationship to deep ICH volume.
Our findings corroborate those of previous studies. Hemorrhages in lobar locations have been consistently found to be larger than those in the deep brain regions.9,16 Additionally, warfarin treatment and the intensity of anticoagulation as measured by the INR were previously shown to be associated with ICH volume.16 However, none of these studies investigated whether hemorrhages in different locations were affected by different factors. We confirmed the association between INR and ICH volume in a large ICH cohort and contribute to refine our understanding of this relationship by describing differences in the strength of this association, identified through location-specific analyses. First, we showed that the effect of warfarin treatment on ICH volume was markedly stronger for deep hemorrhages. Second, we demonstrated that the shape of the dose-response curve that describes the INR-volume relationship also varies with location: while a clear linear pattern was observed in deep ICH, a threshold effect at INR of 3.0 was found in lobar ICH. Interestingly, Flaherty et al9 found a similar threshold effect for anticoagulation when considering ICH in all locations, with only INR greater than 3.0 being associated with larger hematomas.
We were also able to establish that previously reported associations between ICH volume and time to scan, admission BP, and admission blood glucose held true for both lobar and deep ICHs. In accordance with previous reports, time to scan was inversely associated with ICH volume in our study.9 Of note, time to scan played an equally important role in both deep and lobar hemorrhages in our study, likely a reflection of the fact that patients with larger hematomas, and hence more severe symptoms, will present to the emergency department earlier. We chose to exclude admission BP and admission blood glucose (ascertained after the onset of symptoms) from multivariable analyses owing to uncertainty surrounding the causal meaning of these findings. In subjects with larger hematomas, a higher admission BP could be the consequence of increased intracranial pressure, signaling the occurrence of a more intense Cushing response.17 Likewise, in patients with more severe bleeding, elevated admission blood glucose levels could be the consequence of a stronger metabolic response to stress. The role of these factors should be clarified further, especially in light of the evidence indicating that aggressive BP management reduces hematoma growth.7,8
The present study identifies novel location-specific associations that merit further evaluation. In deep ICH, male sex and history of coronary artery disease were associated with increases of 23% and 38%, respectively, in mean ICH volume. These findings could reflect the larger burden of the vascular risk factors, and cardiovascular disease in general, known to affect men compared with women.18 In this regard, previous studies reported a positive association between the burden of leukoaraiosis and ICH volume,11 suggesting that the degree of chronic vascular damage sustained before the ICH takes place could play a role in defining the size of the hematoma. Also for deep ICH, age was inversely correlated with ICH volume. A previous report that combined deep and lobar hemorrhages found a nonsignificant (P = .10) effect of age on volume in the same direction.19 Nonetheless, this finding should be treated with caution because this inverse effect of age on ICH volume did not translate into a similar effect on clinical outcome; to the contrary, increasing age was strongly associated with a worse clinical outcome.
In lobar hemorrhages, pre-ICH treatment with antiplatelets was associated with a 23% increase in mean ICH volume. In agreement with this result, a previous study found a trend toward significance between antiplatelet treatment and ICH volume when assessing hemorrhages in deep and lobar locations jointly.19 A previous report from our group described an association between treatment with aspirin and the risk for recurrent ICH; importantly, this relation was observed only in lobar hemorrhages.20 If corroborated in future studies, this finding could have important implications for clinical practice. Given that pre-ICH treatment with both warfarin and antiplatelets influences hematoma volume in lobar ICH, caution should be exercised when prescribing either, and especially both, types of medications to individuals at high risk for sustaining lobar hemorrhages.
The large sample size and detailed ascertainment of ICH cases constitute the fundamental strengths of the present study. Among its limitations is the hypothesis-generating nature of the results. Our primary goal was to identify location-specific predictors of ICH volume, estimate the size of their effect, and evaluate their role in clinical outcome. However, our analysis cannot establish the precise biological mechanisms that mediate these associations. A second limitation is that evaluated predictors could have been subject to misclassification. In particular, subjects were considered to receive pre-ICH medications (antiplatelets, warfarin, statins, and antihypertensives) based on self-reporting by the patient or their caregivers. This would not introduce information bias for warfarin treatment (as INR was the real measure of interest), but it could generate misclassification for other medications because precise information on dosing, compliance, and duration of treatment was lacking. Third, missing data bias could have been introduced by the absence of data on time to scan for some of the study subjects. We addressed this limitation by comparing the initial study population with that obtained after exclusion of cases that had incomplete data. Fourth, association results involving mortality should be interpreted with caution, given the multitude of factors that influence this particular outcome measure. Finally, data on hematoma expansion, a fundamental determinant of outcome in ICH, were not included in this study. Given the known positive relation between warfarin and the risk for hematoma expansion, it is possible that the present analysis underestimated the role of warfarin on outcome. Follow-up scans, and consequently assessment of hematoma expansion, were available in a significant proportion of the subjects included in this study. These results have been described separately.21
In conclusion, predictors of ICH volume differ between deep and lobar ICHs. Treatment with warfarin influences volume of deep and lobar hemorrhages, with a stronger effect in deep ICH. Antiplatelet use is associated with an increase in the size of lobar hemorrhages when they occur, but have little or no effect on the size of deep hemorrhages. These differences in the predictors of deep and lobar ICH volume indicate that different biological mechanisms determine the extent of bleeding in each location. Further studies are needed to replicate these preliminary findings and characterize the biological mechanisms that could account for these observations.
Accepted for Publication: January 10, 2012.
Corresponding Author: Jonathan Rosand, MD, MSc, Center for
Human Genetic Research, Massachusetts General Hospital, 185 Cambridge St, CPZN-6818, Boston, MA
Published Online: June 3, 2013. doi:10.1001/jamaneurol.2013.98
Author Contributions:Study concept and design: Falcone, Biffi, Anderson, Rosand.
Acquisition of data: Falcone, Brouwers, Battey, Ayres, Vashkevich, Schwab, Rost,
Analysis and interpretation of data: Falcone, Brouwers, Goldstein, Viswanathan,
Drafting of the manuscript: Falcone, Rosand.
Critical revision of the manuscript for important intellectual content: Biffi,
Brouwers, Anderson, Battey, Ayres, Vashkevich, Schwab, Rost, Goldstein, Viswanathan, Greenberg,
Statistical analysis: Falcone, Biffi, Brouwers.
Obtained funding: Schwab, Rosand.
Administrative, technical, and material support: Schwab, Goldstein.
Study supervision: Biffi, Anderson, Rost, Viswanathan, Rosand.
Conflict of Interest Disclosure: Dr
Brouwers’ work was supported by National Institutes of Health/National Institute of
Neurological Disorders and Stroke SPOTRIAS fellowship grant P50NS061343. Dr Anderson received a
research grant from the American Brain Foundation. Dr Goldstein has received a research grant from
the National Institutes of Health/National Institute of Neurological Disorders and Stroke, and he
serves as a consultant/advisory board member for CSL Behring. Dr Greenberg has received a research
grant from the National Institutes of Health. Dr Rosand has received a research grant from the
National Institutes of Health and support from the American Heart Association.
Funding: This study was supported by grants
R01NS073344, R01NS059727, and 5K23NS059774 from the National Institutes of
Health/National Institute of Neurological Disorders and Stroke and grant 0755984T
from the American Heart Association.
Role of the Sponsors: None of the funding entities had any involvement in study
design; data collection, analysis, and interpretation; writing of the manuscript; or decision to
submit the study for publication.