Associations of obesity measures with subtypes of ischemic stroke in the ARIC Study

J Epidemiol. Author manuscript; available in PMC 2011 September 1.

Published in final edited form as:

J Epidemiol. 2010 September 16; 20(5): 347–354.

Published online 2010 June 26.

PMCID: PMC2944008

NIHMSID: NIHMS188381

Associations of obesity measures with subtypes of ischemic stroke in the ARIC Study

Hiroshi Yatsuya,^1,² Kazumasa Yamagishi,³ Kari E. North,⁴ Frederick L. Brancati,⁵ June Stevens,⁶ Aaron R. Folsom,¹ and the ARIC Study Investigators

¹Division of Epidemiology and Community Health, School of Public Health, University of Minnesota, Minneapolis, MN, USA

²Department of Public Health, Graduate School of Medicine, Nagoya University, Nagoya, Japan

³Department of Public Health Medicine, Graduate School of Comprehensive Human Sciences, and Institute of Community Medicine, University of Tsukuba, Tsukuba, Japan

⁴Department of Epidemiology and Carolina Center for Genome Sciences, Gillings School of Global Public Health, University of North Carolina, Chapel Hill, NC USA

⁵Departments of Medicine and Epidemiology, Johns Hopkins University, Baltimore, MD, USA

⁶Department of Nutrition, Gillings School of Global Public Health, University of North Carolina, Chapel Hill, NC USA

Correspondence to: Hiroshi Yatsuya, MD, Division of Epidemiology & Community Health, School of Public Health, University of Minnesota, 1300 South 2nd Street, Suite 300, Minneapolis, MN 55454 USA. Tel: +1-612-625-6730; +1-612-624-0315; Email: yatsu002/at/umn.edu or ; Email: yatsuya/at/gmail.com

The publisher's final edited version of this article is available free at J Epidemiol

See other articles in PMC that cite the published article.

Abstract

Background

Associations between obesity, and lacunar, nonlacunar thrombotic and cardioembolic stroke are not firmly established.

Methods

Body mass index (BMI), waist circumference, and waist-to-hip ratio (WHR) were obtained at baseline between 1987 and 1989 in the Atherosclerosis Risk in Communities (ARIC) Study for 13,549 black and white subjects aged 45 and 64 years without a history of cardiovascular disease or cancer. The incidence of ischemic stroke subtypes was ascertained from surveillance of hospital records, over a median follow-up of 16.9 years. Cox proportional hazard regression analyses adjusted for age, sex, race, education, smoking status, pack years, usual ethanol consumption, and physical activity, were used to estimate hazard ratios (HRs).

Results

The ARIC sample at baseline was 43.8% men, 27.3% blacks, and had a mean age of 53.9 years. Mean BMI, waist circumference, and WHR were 27.7 kg/m², 96.8 cm, and 0.92, respectively. The associations of lacunar (n=138), nonlacunar (n=338), and cardioembolic (n=122) ischemic stroke incidence with obesity measures were all generally positive and linear. The HRs for the highest versus lowest quintile of the three obesity measures ranged from 1.43–2.21 for lacunar stroke, 1.90–2.16 for nonlacunar stroke, and 2.37–2.91 for cardioembolic stroke.

Conclusion

Although different pathophysiological mechanisms may exist, incidence of lacunar, nonlacunar, and cardioembolic stroke were all significantly positively associated with the degree of obesity, regardless of the measure used.

Keywords: obesity, stroke, lacunar infarction, incidence, risk factors

Introduction

Ischemic stroke consists of three major subtypes with different etiologies: lacunar, nonlacunar thrombotic, and cardioembolic. Since differences exist in the underlying vascular or non-vascular pathophysiologies of each subtype, such as lipohyalinosis and microatheroma, or atherosclerosis, risk factors may differ accordingly. For example, serum cholesterol level, atherosclerosis score, and extent of artery stenosis were different between infarction in penetrating artery regions (corresponds to lacunar stroke), and infarction in cortical artery regions (presumably corresponds to nonlacunar thrombotic stroke).¹ Although overweight/obesity is generally considered as a risk factor for ischemic stroke,²^,³^,⁴^,⁵ differences in the associations of obesity with the subtypes of ischemic stroke are less thoroughly studied and not yet clear. A previous investigation in the ARIC Study found that body mass index and waist-to-hip ratio were positively associated only with nonlacunar and cardioembolic ischemic stroke not with lacunar stroke in minimally adjusted model.⁶ But there are reports that suggested a positive association of the degree of obesity with lacunar stroke incidence in women,⁷ or with prevalence of silent lacunar infarct identified by MRI.⁸ Thus, further studies on this topic are warranted, especially in an era of growing obesity in most populations.

Both BMI and waist circumference are used in research and clinical settings for defining the health risks of obesity. In the present study, we described incidence rates and hazard ratios of ischemic stroke subtypes in relation to BMI, waist circumference, and WHR. We tested the hypothesis that there are differences in the associations according to the subtypes, and that any associations of obesity measures with ischemic stroke subtypes would be mediated by known stroke predictors (particularly hypertension and diabetes).

Methods

Study population

The ARIC Study included a cohort of 15,792 persons between 45 and 64 years of age at recruitment in 1987 through 1989.⁹ Population samples were selected by probability sampling methods from Forsyth County, NC (n=4,035); Jackson, MS (blacks only, n=3,728); northwest suburbs of Minneapolis, MN (n=4,009); and Washington County, MD (n=4,020). Baseline response rates ranged from 46% in Jackson to 65–67% in the other three communities. Participants were subsequently contacted annually by telephone and three additional clinic visits. The retention rate was 93% through 2005, and the rate did not differ appreciably between races.

Baseline assessment

Body mass index (BMI: kg/m²) was calculated from measurements of weight to the nearest pound and height to the nearest centimeter, with the participants wearing a scrub suit and no shoes. The ratio of waist (umbilical level) and hip (maximum buttocks) circumference (waist-to-hip ratio: WHR) was calculated as a measure of fat distribution in addition to waist circumference alone. The inter-technician reliability coefficients for waist and hip circumference, and WHR were all r > 0.94.¹⁰

Questionnaires were used to assess educational level, cigarette smoking, alcohol drinking, leisure time sports index, use of antihypertensive or diabetic medications, and histories of physician-diagnosed diabetes, cancer, CHD or stroke. The sports index was derived from questionnaire items on hours per week spent in up to four sports and the months per year each sport was done. By assuming a sport intensity level (light, moderate or heavy), a sport score was calculated ranging from 1 (lowest) to 5 (highest).¹¹ Prevalent CHD at baseline was defined for exclusion as a reported history of a physician-diagnosed myocardial infarction, prior myocardial infarction detected by ECG, or prior cardiovascular surgery or coronary angioplasty.

Three sitting brachial blood pressure measurements were taken with a random-zero sphygmomanometer on the right arm after five minutes’ rest; the last two measurements were averaged. The manual for ARIC blood pressure measurement can be accessed online (http://www.cscc.unc.edu/aric/). Blood was drawn after an 8-hour fasting period with minimal trauma from an antecubital vein. The proportion of subjects that met the 8-hour fasting criterion was 97%. Blood levels of von Willebrand factor (vWF), glucose, HDL cholesterol, and albumin were measured centrally by standard methods. Prevalent diabetes was defined a history of, or treatment for, diabetes, a fasting glucose level of 126 mg/dl or greater, or a casual blood glucose level of 200 mg/dl or greater.

Ascertainment of incident stroke

Ischemic strokes that occurred by December 31, 2005 (median follow-up 16.9 years) were included in the present study. During annual telephone contacts, interviewers asked each ARIC participant to list all hospitalizations during the past year; hospital records were obtained. In addition, all local hospitals annually provided lists of stroke discharges (International Classification of Diseases, Ninth Revision, Clinical Modification codes 430–438), which were scrutinized for ARIC participants’ discharges. Details on quality assurance for ascertainment and classification of stroke are described elsewhere.¹² Briefly, the stroke diagnosis was assigned according to criteria adapted from the National Survey of Stroke.¹³ Strokes secondary to trauma, neoplasm, hematological abnormality, infection, or vasculitis were excluded, and a focal deficit lasting <24 h was not considered to be a stroke. Out-of-hospital stroke was not ascertained and validated; thus, these potential stroke events were not included. A stroke was classified as ischemic when a brain CT or MRI revealed acute infarction or showed no evidence of hemorrhage. All definite ischemic strokes were further classified as lacunar, nonlacunar thrombotic, or cardioembolic on the basis of the recorded neuroimaging results. A stroke was classified as “lacunar” when two criteria were met: (1) typical location of the infarct (basal ganglia, brain stem, thalamus, internal capsule, or cerebral white matter) and (2) infarct size of ≤ 2 cm or unstated size.¹⁴ Definite or probable “cardioembolic” stroke required either (1) autopsy evidence of an infarcted area in the brain and a source of possible cerebral emboli in a vessel or the presence of an embolus in the brain or (2) medical record evidence of a possible non-carotid source of embolus such as moderate or greater valvular heart disease, atrial fibrillation, cardiac or arterial procedure (e.g., cardiac catheterization, open heart surgery, cerebral angiography, and carotid endarterectomy), or intracardiac thrombus. Definite or probable ischemic strokes that were not deemed lacunar or cardioembolic, including atherothrombotic and unclassified thrombotic strokes, were labeled “nonlacunar.” Along with a computer-based classification, cases were independently reviewed by a physician who was provided with a detailed report of the information abstracted from the medical record as well as the full discharge summary, the CT and MRI scan reports, reports from any neurological consults, and admission history. The final diagnosis was determined by agreement of computer and reviewer classification. In the rare occasion when there was a disagreement between computer and reviewer classifications, the diagnosis was adjudicated by a second physician-reviewer. CT or MRI was available for all the ischemic stroke cases except one cardioembolic stroke classified using carotid artery ultrasound and clinical information. The 92 hemorrhagic strokes identified by ARIC were censored at the time of their occurrence.

Statistical analysis

Of the 15,744 blacks and whites in ARIC, we excluded 1,787 participants (blacks: 365, whites: 1,422) who at baseline had a prevalent stroke, CHD, or cancer since CVD treatment and associated behavioral change or cancer-induced weight loss could confound the association between obesity measures and stroke. Participants lacking baseline measurements of BMI, waist or hip circumference (n=32) were also excluded. Those with missing values of potential confounding variables, including leisure time sport index, smoking status and cigarette-years of smoking, usual ethanol intake, and educational level, were then excluded (n=376) leaving a final sample of 5,930 men and 7,619 women (n=13,549 in total). Sensitivity analyses were also performed after truncating the sample at the 1st and 99th percentile of each obesity measure in an attempt to exclude any impact of extreme values on the associations.

Since no interactions of race or sex with obesity measure quintiles was observed (P>0.1, Wald test with 12 degrees of freedom), analyses were done combining the race- and sex-groups. Age-, sex- and race- adjusted means and proportions among ischemic stroke subtypes were calculated by a general linear model and tested with Tukey-Kramer adjustment. Cox proportional hazards regression was used to calculate age-, sex- and race-, and multivariate-adjusted hazard ratios (HRs) and their 95% confidence intervals (CIs) for subtypes of ischemic stroke incidence in relation to quintiles of the obesity measures with the first quintile as the reference category. Quintile cutoff values of each obesity measure were created by averaging the four race-sex specific quintile cutoff values. They were 23.9, 26.2, 28.6, and 32.0 (kg/m²) for BMI; 86, 92, 99, and 107 (cm) for waist circumference; and 0.87, 0.91, 0.94, and 0.98 for WHR. The first model (model I) adjusted for age, sex, race, smoking status (current, past, or never), cigarette-years of smoking, usual ethanol intake (grams/week), educational level (high school graduate or not) and leisure time sport index (1.0–1.9, 2.0–2.4, 2.5–2.9, 3.0–5.0) score. In a mediation model (Model II, n=288 more excluded), we further adjusted for systolic blood pressure, use of antihypertensive medication, prevalent diabetes, and blood levels of HDL cholesterol, vWF and albumin simultaneously in light of a previous ARIC paper that identified them as predictors of incident stroke.¹⁵ A trend test was performed by assigning the median value of each quintile to corresponding individuals and treating it as a continuous variable in the model. When there was an indication of nonlinear association from quintile categorical analysis, statistical significance of the quadratic term for the obesity measure (continuous) was evaluated. Additionally, for each ischemic stroke subtype, we performed analysis stratified by hypertension, which was defined as either systolic/diastolic blood pressure ≥ 140/90 mmHg or currently using medication for hypertension. Interactions of hypertension with obesity measure quintile were tested in Model I using a Wald test with 4 degrees of freedom and the criterion probability level (p) set at 0.1. The assumption of hazard proportionality was tested by a model including follow-up time by obesity measure quintile interaction. The follow-up time was first examined using a continuous scale, and then dichotomized at year 10. All statistical analyses were performed by SAS.

Results

Waist circumference quintile was positively associated with age, while the proportions of men and blacks were lowest in the first quintile (Table 1). Although the prevalence of current smoking was inversely associated with waist circumference quintile, cigarette-years of smoking among ever smokers was positively associated with waist circumference quintile. The proportion of subjects with more than high school education was higher in lower waist circumference quintiles.

Table 1

Age-, sex, and race-adjusted baseline characteristics according to waist circumference quintile, ARIC 1988–90

During a median of 16.9 years of follow-up (max=19.1 y), there were 598 incident ischemic strokes, 138 of which were lacunar, 338 nonlacunar, and 122 cardioembolic. At baseline, mean BMI, waist circumference and WHR were significantly greater in those who developed nonlacunar or cardioembolic stroke, compared with subjects who did not develop ischemic stroke. Although WHR was significantly higher in those who developed lacunar stroke than those without an ischemic stroke, BMI and waist circumference were not significantly different. Blacks were more likely than whites to develop ischemic stroke, especially lacunar stroke. The crude incidence rate of lacunar stroke was more than four-fold higher in blacks than whites (1.51 versus 0.34 per 1,000 person-years, data not shown in table). Both the proportion of subjects taking antihypertensive medications and mean systolic blood pressure were similarly higher in those who developed any ischemic stroke subtype compared to those free of ischemic stroke. The prevalence of diabetes mellitus was highest in lacunar stroke cases, followed by cardioembolic and nonlacunar ischemic stroke cases.

As shown in Table 2, the HRs of lacunar stroke were generally positively and linearly associated with waist circumference and WHR quintiles, but less so with BMI. The HR of lacunar stroke for the highest quintile of BMI was only 1.43 (95%CI: 0.84–2.45), and continuous BMI showed no significant association (HR1 per standard deviation: 1.15, 95%CI: 0.98–1.34). However, including a quadratic BMI (continuous) variable also did not support departure from linearity (p for quadratic BMI=0.29). Adjustment for potential mediating factors (model II), particularly diabetes, hypertension and HDL cholesterol, significantly attenuated most obesity measure associations with lacunar stroke.

Table 2

Incidence and hazard ratios (95%CI) of lacunar ischemic stroke according to obesity measure quintile or per standard deviation, ARIC, 1987–2005

Nonlacunar ischemic stroke was positively and strongly associated with all obesity measures in quintiles as well as on a continuous scale (Table 3). The association of nonlacunar stroke incidence with obesity measures was attenuated with adjustment for mediating factors but incompletely for WHR (HR1: 1.16, 95%CI: 1.01–1.34 in Model II). The analyses for cardioembolic stroke showed similar results to those for nonlacunar stroke (Table 4). The HR of cardioembolic stroke for the highest quintiles of obesity measures ranged 2.37–2.91.

Table 3

Incidence and hazard ratios (95%CI) of nonlacunar ischemic stroke according to obesity measure quintile or per standard deviation, ARIC, 1987–2005

Table 4

Incidence and hazard ratios (95%CI) of cardioembolic stroke according to obesity measure quintile or per standard deviation, ARIC, 1987–2005

Sensitivity analyses excluding subjects with extreme values of obesity measures produced similar findings except for the association between BMI and lacunar stroke incidence, where the HR per 5.4 kg/m² increment (corresponding to HR1 of Model I in Table 2) became statistically significant (HR: 1.29, p=0.006).

Hypertension modified the associations of obesity measure quintiles with lacunar stroke (Table 5, p for interaction<0.05), but no such effect modification was observed for nonlacunar or cardioembolic stroke (p for interaction>0.3). Significant positive associations between obesity measures and incidence of lacunar stroke were present only in subjects without hypertension. In contrast, obesity was not associated with the incidence of lacunar stroke among hypertensives. Subjects with hypertension did have significantly higher BMI than those without (29.5 versus 26.7 kg/m²).

Table 5

Incidence and hazard ratios (95%CI) of lacunar stroke according to obesity measure quintile in participants with or without baseline hypertension, ARIC, 1987–2005

Discussion

In this analysis of data from the ARIC cohort, all obesity measures were positively and similarly associated with all ischemic stroke subtypes. These associations were slightly stronger than in a previous ARIC report with fewer events and no examination of waist circumference.⁶

The present findings are consistent in part with the Hisayama Study, which found a positive association between BMI and lacunar stroke incidence in Japanese women.⁷ However, in men BMI associations were not found with any ischemic stroke subtype or for atherothrombotic or cardioembolic stroke in women. The large difference in the degree of obesity between ARIC and Hisayama studies precludes direct comparisons (mean BMIs in the Hisayama Study were 21.5 kg/m² in men and 21.7 kg/m² in women, compared with 27.7 kg/m² for both genders in the ARIC Study). ARIC finding may also be consistent with a cross-sectional study that found a positive association between abdominal obesity and the prevalence of silent lacunar infarct identified by MRI in apparently healthy Japanese men and women aged 40 to59.⁸ Another study¹⁶ found that BMI ≥ 27 kg/m² was not associated with any ischemic stroke subtype after adjusting for age, sex, social class, hypertension, and hypercholesterolemia. Since mediating variables were included, these results are consistent with our findings from the mediation analyses. Nevertheless, further investigations in populations with various degrees of obesity that included different ethnicities and sexes, are essential to confirm the present findings.

Mediating factors together explained all the association of obesity measures with each ischemic stroke subtype, but no single mediating factor accounted totally for the associations except in the case of lacunar stroke analyses where any of hypertension, diabetes mellitus, or HDL cholesterol attenuated the significant association. Coefficients for waist circumference (HR1) changed 44% after adding systolic blood pressure and antihypertensive medication to Model I. The amount of change was 72% in the case of diabetes mellitus, and 35% for HDL cholesterol. Statistically significant associations of waist circumference with lacunar stroke incidence disappeared in all cases. Each of these variables explained some associations between waist circumference and nonlacunar thrombotic or cardioembolic stroke (changes in the coefficients from Model I ranged from 12% to 38%), but the statistical significance of the waist circumference variable remained. It was only when all these variables were entered simultaneously that the significant association was attenuated (p=0.13 for both nonlacunar thrombotic and cardioembolic stroke). Given the strong and probably causal association between obesity and hypertension, diabetes mellitus, and HDL cholesterol, obesity would be an important target for the prevention of any subtype of ischemic stroke.

However, we did not find positive associations between obesity measures and lacunar stroke incidence in subjects with hypertension. This finding may be consistent with a prospective study of hypertensive older adults aged 60 or over that did not find associations between BMI and any ischemic stroke subtype.¹⁷ We did several additional analyses in an attempt to find a possible mechanism related to the null association. However, neither additional adjustment for potential mediating variables, excluding subjects with less than five years of follow-up, nor limiting analyses to either race or sex, changed the null association in hypertensives. Other established risk factors, such as elevated systolic blood pressure, diabetes mellitus, or low HDL cholesterol, were all positively associated with lacunar stroke incidence in subjects with hypertension. Although the underlying mechanism remains unknown, it may have been due to a high incidence rate of lacunar stroke even in subjects in the lowest BMI and waist circumference quintiles among hypertensives. Other dietary or lifestyle factors may have existed in these individuals that predisposed them to lacunar stroke over the effect of obesity.¹⁸ On the other hand, confidence intervals were wide and it is possible that null finding was by chance.

There are several limitations of this study that warrant discussion. Although neuroimaging reports and clinical features were used to classify ischemic stroke cases into subtypes, some cases may have been misclassified. Subtypes were classified via a review of medical records and neuroimaging reports, rather than direct examination of patients or images. The medical record classification process did not use the obesity status or other risk factors of stroke cases, but reviewers were not blinded to such information. This might have lead to some potential biases in the associations. Further, since we used a maximum diameter of 20 mm in our study, there is the possibility of misclassification of nonlacunar thrombotic stroke into lacunar stroke; however we minimized this possibility by prioritizing neuroimaging reports that explicitly stated that the infarct was not lacunar even the image itself met the criterion. Any misclassification might have distorted the obesity and lacunar stroke association. Nevertheless, the size criterion of lacunar infarct has also been debated;¹⁹ future studies utilizing advanced imaging would be ideal. Although we did not find any effect modification by race or sex, the number of cases was limited, so we could not conduct sex- or race-specific analyses. The present sample consisted of US blacks and whites, most of which were from a single center, limiting the generalizability of our findings to other cultural or socio-economic contexts. Finally, we used regression rather than structural equation modeling to examine mediation.

A strength of the present study is that we analyzed the association of obesity measures and ischemic stroke subtypes using prospective population-based data, including both blacks and whites, with a relatively large numbers of events. There are no prior prospective studies that specifically addressed the association between obesity and ischemic stroke subtypes in detail.

In conclusion, lacunar, nonlacunar and cardioembolic stroke were all significantly positively associated with obesity measures. Although different pathophysiological mechanisms may exist relating obesity and each subtype, the prevention and control of obesity has the potential to reduce the burden of stroke in the US. Further studies are needed for other populations.

Acknowledgement

The Atherosclerosis Risk in Communities Study is carried out as a collaborative study supported by National Heart, Lung, and Blood Institute contracts N01-HC-55015, N01-HC-55016, N01-HC-55018, N01-HC-55019, N01-HC-55020, N01-HC-55021, and N01-HC-55022. The authors thank the staff and participants of the ARIC study for their important contributions. The author (HY) is also grateful to the Uehara Memorial Foundation which supported his research activities at Division of Epidemiology and Community Health, School of Public Health, University of Minnesota.

References

1. Konishi M, Iso H, Komachi Y, Iida M, Shimamoto T, Jacobs DR, Jr., et al. Associations of serum total cholesterol, different types of stroke, and stenosis distribution of cerebral arteries. The Akita Pathology Study. Stroke. 1993;24:954–964. [PubMed]

2. Kurth T, Gaziano JM, Berger K, Kase CS, Rexrode KM, Cook NR, et al. Body mass index and the risk of stroke in men. Arch Intern Med. 2002;162:2557–2562. [PubMed]

3. Jood K, Jern C, Wilhelmsen L, Rosengren A. Body mass index in mid-life is associated with a first stroke in men: a prospective population study over 28 years. Stroke. 2004;35:2764–2769. [PubMed]

4. Song YM, Sung J, Davey Smith G, Ebrahim S. Body mass index and ischemic and hemorrhagic stroke: a prospective study in Korean men. Stroke. 2004;35:831–836. [PubMed]

5. Park JW, Lee SY, Kim SY, Choe H, Jee SH. BMI and stroke risk in Korean women. Obesity (Silver Spring) 2008;16:396–401. [PubMed]

6. Ohira T, Shahar E, Chambless LE, Rosamond WD, Mosley TH, Jr., Folsom AR. Risk factors for ischemic stroke subtypes: the Atherosclerosis Risk in Communities study. Stroke. 2006;37:2493–2498. [PubMed]

7. Tanizaki Y, Kiyohara Y, Kato I, Iwamoto H, Nakayama K, Shinohara N, et al. Incidence and risk factors for subtypes of cerebral infarction in a general population: the Hisayama study. Stroke. 2000;31:2616–2622. [PubMed]

8. Park K, Yasuda N, Toyonaga S, Tsubosaki E, Nakabayashi H, Shimizu K. Significant associations of metabolic syndrome and its components with silent lacunar infarction in middle aged subjects. J Neurol Neurosurg Psychiatry. 2008;79:719–721. [PubMed]

9. The Atherosclerosis Risk in Communities (ARIC) Study: design and objectives. The ARIC investigators. Am J Epidemiol. 1989;129:687–702. [PubMed]

10. Ferrario M, Carpenter MA, Chambless LE. Reliability of body fat distribution measurements. The ARIC Study baseline cohort results. Atherosclerosis Risk in Communities Study. Int J Obes Relat Metab Disord. 1995;19:449–457. [PubMed]

11. Baecke JA, Burema J, Frijters JE. A short questionnaire for the measurement of habitual physical activity in epidemiological studies. Am J Clin Nutr. 1982;36:936–942. [PubMed]

12. Rosamond WD, Folsom AR, Chambless LE, Wang CH, McGovern PG, Howard G, et al. Stroke incidence and survival among middle-aged adults: 9-year follow-up of the Atherosclerosis Risk in Communities (ARIC) cohort. Stroke. 1999;30:736–743. [PubMed]

13. The National Survey of Stroke. National Institute of Neurological and Communicative Disorders and Stroke. Stroke. 1981;12:I1–I91. [PubMed]

14. Ay H, Furie KL, Singhal A, Smith WS, Sorensen AG, Koroshetz WJ. An evidence-based causative classification system for acute ischemic stroke. Ann Neurol. 2005;58:688–697. [PubMed]

15. Chambless LE, Heiss G, Shahar E, Earp MJ, Toole J. Prediction of ischemic stroke risk in the Atherosclerosis Risk in Communities Study. Am J Epidemiol. 2004;160:259–269. [PubMed]

16. Markus HS, Khan U, Birns J, Evans A, Kalra L, Rudd AG, et al. Differences in stroke subtypes between black and white patients with stroke: the South London Ethnicity and Stroke Study. Circulation. 2007;116:2157–2164. [PubMed]

17. Davis BR, Vogt T, Frost PH, Burlando A, Cohen J, Wilson A, et al. Systolic Hypertension in the Elderly Program Cooperative Research Group. Risk factors for stroke and type of stroke in persons with isolated systolic hypertension. Stroke. 1998;29:1333–1340. [PubMed]

18. Reed D, Jacobs DR, Jr., Hayashi T, Konishi M, Nelson J, Iso H, et al. A comparison of lesions in small intracerebral arteries among Japanese men in Hawaii and Japan. Stroke. 1994;25:60–65. [PubMed]

19. Cho AH, Kim JS. Reply to the letter by Lodder: Size criterion for lacunar infarction. Cerebrovascular Diseases. 2007;24:156–157. [PubMed]