Body Mass Index and Etiology of Intracerebral Hemorrhage
Background and Purpose—Extremes of body mass index (BMI) are associated with increased incidence of intracerebral hemorrhage (ICH). Because ICH can result from different vessel pathologies, we investigated whether the effect of BMI depends on ICH etiology.
Methods—We analyzed 384 consecutive ICH cases (188 lobar ICH and 196 deep ICH) and 388 control subjects enrolled between 2004 and 2009 in an ongoing single-center prospective study of primary ICH. ICH was categorized as lobar or deep based on CT imaging at admission. BMI was calculated based on subjects' height and weight at enrollment. We constructed multivariate logistic regression models including BMI and known predictors of ICH occurrence. Analyses were performed separately for lobar and deep ICH subjects versus control subjects.
Results—Low BMI (<18.5 kg/m2) and very high BMI (>30.0 kg/m2) were associated with deep ICH risk (OR, 1.76; P=0.011 and OR, 1.75; P=0.013, respectively), whereas no effect was found for lobar ICH. Furthermore, sex-stratified analyses uncovered that among low BMI individuals, males were at higher ICH risk (OR, 2.85; P=0.041) but females were not (OR, 0.89; P=0.54, respectively).
Conclusions—Extremes of BMI are associated with increased risk of deep ICH, but not lobar ICH, suggesting a role for BMI in the vascular pathologies underlying deep ICH, but not in pathologies such as cerebral amyloid angiopathy that cause ICH in the lobar brain regions.
- amyloid angiopathy
- body mass index
- intracerebral hemorrhage
- intracerebral hypertension
Intracerebral hemorrhage (ICH) most commonly occurs when a chronically diseased vessel ruptures with extravasation of blood into the brain parenchyma. Although chronic small vessel disorders such as hypertensive vasculopathy and cerebral amyloid angiopathy are well-recognized causative mechanisms for ICH, effective preventive strategies are limited. As a result, ICH continues to account for approximately 15% of all strokes, although it is responsible for >30% of mortality due to cerebrovascular disorders.1 Further insight into ICH etiology is needed to identify novel targets for preventive and therapeutic interventions.
Body mass index (BMI) has been associated with ICH risk in several large cohorts2–4 with extremes of BMI (both low and high) predicting increased risk for ICH. The biological mechanisms underlying these associations, however, remain unclear. We sought to determine whether the association between BMI and ICH risk is dependent on the underlying vessel pathology. Taking advantage of the observation that hemorrhage location (deep hemispheric versus lobar) is an important clue to etiology, we conducted a case–control study separately comparing lobar and deep ICH cases drawn from a longitudinal prospective cohort to ICH-free control subjects recruited from the same geographical area.
Subject Enrollment and Data Collection
Subjects were drawn from an ongoing longitudinal cohort study of primary ICH as previously described.5 Briefly, enrolled cases were consecutive patients aged ≥55 years presenting to the Massachusetts General Hospital emergency department between 2004 and 2009 with a diagnosis of spontaneous primary ICH (Table 1). Eligibility for study participation required CT scan confirmation of hemorrhagic stroke. Exclusion criteria included presence of trauma, brain tumor, hemorrhagic transformation of a cerebral infarction, vascular malformation, or any other cause of secondary ICH. Control subjects were enrolled from primary care practices at Massachusetts General Hospital among individuals presenting for an annual health check-up and included individuals aged >55 years at the time of enrollment. Control subjects were confirmed to have no medical history of ICH through interview and review of medical records. The Institutional Review Board of Massachusetts General Hospital approved all aspects of this study, and informed consent was obtained from all subjects or their legal guardians.
Clinical data were recorded within 48 hours of index ICH hospital admission by study-specific research staff. Collected data included demographics and medical history, which included hypertension (physician diagnosis), atrial fibrillation (medical history or electrocardiographic findings at admission), diabetes (physician diagnosis), coronary artery disease (medical history of angina, myocardial infarction, coronary artery bypass graft, or percutaneous transluminal coronary angioplasty), hyperlipidemia (physician diagnosis), history of previous ICH/ischemic stroke/transient ischemic attack (based on history and review of neuroimaging), and cognitive impairment (medical history). We also collected information on alcohol and tobacco use and pre-ICH medications (warfarin, aspirin or other antiplatelet agents, antihypertensive agents, oral hypoglycemic agents or insulin, and statins). Clinical data for control subjects were recorded by study staff at the time of enrollment using identical definitions and recording procedures. Height and weight were recorded within 48 hours of admission (cases) or enrollment (control subjects) and used to calculate BMI.
All admission CT scans were reviewed by stroke neurologists blinded to all clinical data to determine ICH location. ICH isolated to the cortex (with or without involvement of subcortical white matter) was defined as lobar, whereas ICH selectively involving the internal capsule, thalamus, basal ganglia, or brain stem was defined as deep (nonlobar) ICH. Because we sought to investigate the association between BMI and ICH risk based on bleeding etiology, single and/or multiple concurrent bleeds involving both deep and lobar territories were defined as mixed ICH and represented an exclusion criterion. Cerebellar hemorrhages were also not analyzed in the present study, because existing evidence suggests that both cerebral amyloid angiopathy and hypertension can be responsible for intracranial bleeding in this location.6 Individuals with CT scans of insufficient quality for location determination (n=3) were excluded from all analyses. When ICH location assignment was not clear (n=7), the scan was reviewed by a group of study neurologists and neuroradiologists for consensus. Scans lacking a consensus location (n=3) were excluded from analysis.
Age at index ICH was analyzed as a continuous variable. BMI was categorized according to World Health Organization guidelines: low BMI (underweight): <18.5 kg/m2, normal BMI: 18.5 to 24.9 kg/m2, high BMI (overweight): 25.0 to 29.9 kg/m2, and very high BMI (obese): >30.0 kg/m2. Different BMI categories were separately compared with the normal BMI group, which was assumed as a reference for analytic purposes. Demographics, medical history, and drug use data were all analyzed as binary variables.
Continuous numeric variables were expressed as mean±SD. Cases and control subjects were compared in univariate analyses using t test, Wilcoxon rank sum, χ2, or Fisher exact test as appropriate. Differences in cohort characteristics between lobar and deep ICH cases were assessed in univariate analyses using identical methods.
Multivariate analyses were performed using logistic regression models, including BMI categories as ICH risk predictors. Candidate covariates for all multivariate models included all variables showing a trend in association with ICH risk in univariate analysis (P<0.20) as well as variables showing trends toward association with BMI (P<0.20). Backward elimination of nonsignificant variables (P>0.05) was subsequently used to generate a minimal model. Potential confounders and predictors were then reincorporated into the resulting minimal model using change-in-effect criteria; all remaining variables were individually added and retained if they improved the overall model fit (as assessed by the likelihood ratio method) by >10%.7
Analyses were performed separately for deep and lobar ICH subjects versus control subjects, and effect sizes for BMI categories were compared using the Breslow-Day (B-D) test. Based on observed results, we further compared effect sizes for BMI categories by sex using the B-D tests and performed a multivariate sex-stratified logistic regression using aforementioned methods. Statistical power for discovery of association between BMI and ICH risk assumed existing sample sizes for both lobar and deep ICH.
All statistical analyses were performed using STATA 10.0. All statistical tests were considered significant at P<0.05 (2-tailed).
A total of 188 lobar ICH cases and 196 deep (nonlobar) ICH cases enrolled during the recruitment period were eligible for inclusion in the present study. A total of 388 control subjects were available for case–control analysis (Table 1). Compared with lobar ICH cases, subjects diagnosed with deep ICH were younger, more likely to have a pre-ICH history of hypertension and coronary artery disease, and to be prescribed antihypertensive agents and statins (all P<0.05). Patients with lobar ICH were more likely to have a history of pre-ICH cognitive impairment and to be in the normal BMI range (all P<0.05).
BMI values were not associated with lobar ICH risk in either univariate and multivariate analysis (Table 2). Hypertension, hyperlipidemia, and history of cognitive impairment were more frequent in lobar ICH cases compared with control subjects in both univariate and multivariate analysis. Consequently, predicted lobar ICH risk showed no association with BMI categories in either female or male study participants (Figure, top panel). Based on available data, we estimated our study to have statistical power of 90% to detect an effect size (OR) of 1.70 for low (<18.5 kg/m2), high (25.0 to 30.0 kg/m2), and very high BMI (>30.0 kg/m2). Statistical power for an effect size of 1.50 for the same BMI categories was estimated to be >80%.
Both low BMI (<18.5 kg/m2) and very high BMI (>30.0 kg/m2) were associated with deep hemorrhage. In univariate analyses, age, sex, hypertension, diabetes, coronary artery disease, hyperlipidemia, and alcohol consumption (>3 oz/day) were associated with deep ICH risk (Table 3). Multivariate logistic regression confirmed the association between low BMI/very high BMI and increased risk of deep ICH (both P<0.05).
Male sex was associated with deep ICH risk in both univariate and multivariate (OR, 1.94; 95% CI, 1.28 to 2.96; P=0.002) analyses but did not influence lobar ICH risk (P>0.2). Given that there are established sex-based differences in adipose tissue and lipid metabolism,8,9 we hypothesized the existence of heterogeneity of effect size(s) for the association between BMI categories and deep ICH when comparing men and women. Difference in effect size for the association between low BMI and deep ICH was significant (B-D P=0.024). Specifically, low BMI was associated with increased deep ICH risk in men but not women (Figure, bottom panel) in multivariate logistic regression (Table 4) stratified by sex and adjusted for age, hypertension, diabetes, coronary artery disease, hyperlipidemia, and alcohol consumption. For high and very high BMI, the B-D test identified a consistent trend toward stronger association with ICH risk in men (BMI 25.0 to 30.0 kg/m2: B-D P=0.17 and BMI >30.0 kg/m2: B-D P=0.12).
Our results suggest that BMI differentially influences risk of lobar and deep ICH, consistent with the hypothesis that BMI influences specific vascular pathologies that lead to ICH.2–4 In particular, these findings are most likely related to biological differences between cerebral amyloid angiopathy-related lobar ICH and hypertensive deep ICH. While hypertension does play a role in lobar ICH risk, the mechanisms leading to cerebral amyloid deposition and associated vasculopathy appear to play a much larger role.10–12 Although we did have adequate statistical power to rule out effects of BMI in lobar ICH that were similar-sized to those observed in deep ICH, we are of course unable to discern more subtle effects. However, our findings strongly suggest that, even if present, effects of BMI on lobar ICH risk are likely to play only a minor role.
Results from analysis comparing deep ICH cases and control subjects suggest a role for both low and very high BMI in ICH. Low cholesterol levels are known to increase risk of deep ICH and other manifestations of cerebral small vessel disease such as leukoaraiosis13,14 and could be responsible for increased ICH risk in individuals with low BMI. Although we did adjust for hyperlipidemia in our analyses, cholesterol blood levels were not available. This limitation prevents us from further exploring the possible role of hypocholesterolemia in deep ICH in underweight subjects. We also uncovered a differential effect of low BMI on deep ICH; low BMI male subjects were at higher risk of deep ICH. Given the well-studied differences in lipid and adipose tissue metabolism in men and women, this observation may also point to low cholesterol levels as a potential mechanism.8,9 Specifically, a large analysis conducted as part of the World Health Organization Monitoring Trends and Determinants in Cardiovascular Disease (MONICA) project found that prevalence of hypercholesterolemia in males aged 50 to 64 years was lower than in females with similar BMI in the same age groups.13 Ultimately, however, larger studies will be required to clarify any sex-specific effects of BMI in ICH.
Hyperlipidemia is therefore unlikely to account for increased ICH risk in obese patients. Obesity is frequently associated with hyperlipidemia, which has been repeatedly found to be protective against ICH risk.14 Furthermore, although diabetes and hypertension are closely associated with general adiposity, adjustment for these risk factors in our analyses did not eliminate the association; this finding closely mirrors results from large population-based studies, which also failed to invalidate the association between very high BMI and hemorrhagic stroke after adjustment for metabolic covariates.2–4
Our study has limitations. Despite controlling for demographics and pre-ICH baseline conditions in our analyses, it is possible that our results are influenced by unmeasured confounders due to health and/or socioeconomic factors. Furthermore, although pre-existing medical conditions conferring risk for ICH and associated with BMI (eg, hypertension, hyperlipidemia)15,16 have been captured as binary variables in the present study, lack of detailed information on severity of these pre-ICH conditions may have limited our ability to adjust for cardiovascular risk factors in our analyses. Our analyses could also be affected by severity and survival bias, which may be more likely to disproportionately affect individuals with low and very high BMI. However, all the aforementioned confounding effects are unlikely to selectively influence only deep ICH cases and thus could not fully account for our findings. Finally, our cohort is largely comprised of white subjects, thus limiting generalizability to other ethnicities at higher risk for intracranial bleed. Dedicated studies in nonwhite populations will be needed to confirm our findings.
Extremes of BMI, both very low and very high, were associated with risk of deep ICH, whereas we found no association between BMI and risk of lobar ICH despite adequate statistical power. These findings suggest a role for metabolic factors in risk of ICH arising in the setting of hypertensive vasculopathy.
Sources of Funding
The authors' work on this study was supported by funding from the National Institute for Neurologic Disorders and Stroke (R01 NS063925, R01 NS059727, P50 NS051343, K23 NS064052, K23 NS059774). A.B. received support from the American Heart Association/Bugher Foundation Centers for Stroke Prevention Research (0775011N, 0755984T). All funding entities had no involvement in study design, data collection, analysis, interpretation, writing of the report, or in the decision to submit the article for publication.
We thank all clinical coordinators and research staff in the Hemorrhagic Stroke Research Group, Department of Neurology, Massachusetts General Hospital.
Natan M. Bornstein, MD, was the Consulting Editor for this paper.
- Received February 11, 2011.
- Revision received March 28, 2011.
- Accepted March 30, 2011.
- © 2011 American Heart Association, Inc.
- Song YM,
- Sung J,
- Davey Smith G,
- Ebrahim S
- Silventoinen K,
- Magnusson PK,
- Tynelius P,
- Batty GD,
- Rasmussen F
- Biffi A,
- Halpin A,
- Towfighi A,
- Gilson A,
- Busl K,
- Rost N,
- et al.
- Agresti A
- Agresti A
- Arima H,
- Tzourio C,
- Anderson C,
- Woodward M,
- Bousser MG,
- MacMahon S,
- et al.
- Greenberg SM,
- Eng JA,
- Ning M,
- Smith EE,
- Rosand J
- Smith EE,
- Gurol ME,
- Eng JA,
- Engel CR,
- Nguyen TN,
- Rosand J,
- et al.
- Hu G,
- Barengo NC,
- Tuomilehto J,
- Lakka TA,
- Nissinen A,
- Jousilahti P