Body Mass Index, Waist Circumference, and Waist-Hip Ratio on the Risk of Total and Type-Specific Stroke FREE

Adiposity is an established risk factor for cardiovascular disease,^1- 2 but the association between adiposity and cerebrovascular disease is still to some extent unclear. Some prospective studies^3- 9 have found that a high body mass index (BMI) (calculated as weight in kilograms divided by height in meters squared) may increase the risk of total stroke, especially ischemic stroke. However, this association disappeared after adjustment for other risk factors, indicating that the effect of BMI on the risk of stroke is mediated through other factors, especially hypertension.^3- 5,9 Other studies have found no association^10- 11 or a low BMI associated with a higher risk of stroke.¹² In addition, the association between BMI and hemorrhagic stroke remains unclear. Some studies have shown an increased risk of hemorrhagic stroke among lean persons,^5,9,12 but other studies have found a J-shaped association,⁸ no association,⁷ or an increased risk with increasing BMI.⁶

Several studies^13- 15 have also suggested that indicators of abdominal adiposity, such as waist-hip ratio, may be a stronger predictor of stroke than general adiposity, measured by BMI. Limited data exist on the association between different indicators of adiposity and the risk of stroke subtypes in men and women. The aim of this study was to examine the association of 3 indicators of adiposity (BMI, waist circumference, and waist-hip ratio) with the risk of total and type-specific stroke incidence.

Six independent cross-sectional population surveys were carried out in 5 geographic areas of Finland in 1972, 1977, 1982, 1987, 1992, and 1997.¹⁶ In 1972 and 1977, a randomly selected sample making up 6.6% of the population born between 1913 and 1947 was drawn. Since 1982, the sample was stratified by area, sex, and 10-year age group according to the World Health Organization (WHO) MONICA (MONItoring trends and determinants of CArdiovascular disease) Project.¹⁷ The participation rate varied by year from 74% to 88%.¹⁶ The subjects included in the 6 surveys were aged 25 to 64 years, and the 1997 survey also included subjects aged 65 to 74 years. Subjects who participated in more than 1 survey were included only in the first survey cohort. The total sample size of the 6 surveys was 53 166. The final sample comprised 23 967 men and 26 029 women, after excluding the participants with a history of coronary heart disease (n = 1298) or stroke (n = 768) at baseline and the participants with incomplete data on any variables required for this analysis (n = 1104). The participants gave informed consent (verbal in 1972-1992 and signed in 1997). These surveys were conducted according to the ethical rules of the National Public Health Institute, and the investigations were performed in accordance with the Declaration of Helsinki.

A self-administered questionnaire was sent to the participants to be completed at home. The questionnaire included questions on medical history, socioeconomic factors, physical activity, smoking habits, and alcohol consumption. Educational level, measured as the total number of school years, was divided into birth cohort–specific tertiles. Physical activity included occupational, commuting, and leisure-time physical activity, and was merged and regrouped into 3 categories: low, moderate, and high.^18- 19 Based on the responses, the participants were classified as never smokers, ex-smokers, and current smokers. Current smokers were categorized into those participants who smoked less than 20 or 20 or more cigarettes per day. Because questions on alcohol consumption were different between the first 2 surveys (1972 and 1977) and the latter surveys, the participants were categorized into abstainers and alcohol users. Family history of stroke was defined as a history of those whose mothers or fathers were once diagnosed as having stroke. Subjects who reported having diabetes mellitus diagnosed by a physician were classified as having a history of diabetes mellitus at baseline.

At the study center, specially trained nurses measured height, weight, and blood pressure using the standardized protocol according to the WHO MONICA Project.¹⁷ Height and weight were measured without shoes and with light clothing. The measurements of height were rounded to the nearest centimeter, and weight to the nearest 100 g. Blood pressure was measured from the right arm after 5 minutes of sitting using a mercury sphygmomanometer for each survey. After blood pressure measurement, a venous blood specimen was taken. Total cholesterol level was determined by using the Lieberman-Burchard method in 1972 and 1977, and by an enzymatic method (CHOD-PAP; Boehringer Mannheim GmbH, Mannheim, Germany) since 1982. Because the enzymatic method gave 2.4% lower values than the Lieberman-Burchard method, the values measured in 1972 and 1977 were corrected by this percentage. All samples were analyzed in the same central laboratory at the National Public Health Institute.

More detailed data on diet and alcohol consumption, and measurement of waist and hip circumferences, were included in the surveys of 1987, 1992, and 1997 (men, n = 9186; and women, n = 10 163). The participants were asked, “How many slices of bread do you eat daily?” The major proportion of fiber (55%) consumption in Finland comes from bread. The frequency of consumption of vegetables and fruits during the past week (< 1 time per week, 1-2, 3-5, and 6-7 times per week) and the frequency of consumption of sausages during the past 12 months (< 1 time per month, 1-2 times per month, 1 time per week, 2 times per week, almost daily, and > 1 time per day) were obtained. Ethanol consumption was categorized into 4 groups (0, 1-34, 35-209, and ≥ 210 g/wk in men; and 0, 1-34, 35-139, and ≥ 140 g/wk in women).

Waist circumference was measured midway between the lower rib margin and iliac crest. Hip circumference was measured at the level of the widest circumference over the greater trochanters. The measurements of waist and hip were rounded to the nearest half centimeter. Waist-hip ratio was calculated as waist circumference divided by hip circumference.

The survey cohorts were followed up until the end of 2004 through computerized register linkage by a unique personal identification number. Mortality data were obtained from Statistics Finland, and data on nonfatal events were obtained from the National Hospital Discharge Register. The 8th, 9th, and 10th revisions of the International Classification of Diseases were used to identify subarachnoid (codes 430 and I60) and intracerebral (codes 431 and I61-I62) hemorrhage, intracerebral infarction (codes 432-438 and I63-I66), and any stroke events (codes 430-438 and I60-I66). International Classification of Diseases, Ninth Revision (ICD-9) code 432 was classified as an intracerebral hemorrhage. In the data analyses, we combined all hemorrhagic strokes because the number of cases with subarachnoid hemorrhage was small. The stroke events that occurred before the baseline survey were identified from the Hospital Discharge Register retrospectively and excluded from the analyses. The validity of the diagnosis of acute stroke in Finland is good for hospital discharge register (agreement in 90% for all strokes, 82% for hemorrhage, and 90% for cerebral infarction) and death register (agreement in 97% for all strokes, 95% for hemorrhage, and 92% for cerebral infarction).²⁰ End points during follow-up were incident stroke events, defined as either the first nonfatal stroke event or stroke-related death without a preceding nonfatal event.

Differences in risk factors at different levels of BMI were tested using analysis of variance or logistic regression after adjustment for age and study year (Table 1). The Cox proportional hazards model was used to estimate the association of BMI, waist circumference, and waist-hip ratio on the risk of stroke incidence. Body mass index was evaluated in the following 2 ways: (1) as the 4 WHO weight categories (underweight, < 18.5; normal weight, 18.5-24.9 [reference]; overweight, 25.0-29.9; and obese, ≥ 30.0) and (2) as a continuous variable. Waist circumference and waist-hip ratio were evaluated in 2 ways: (1) as sex-specific quartiles and (2) as a continuous variable. Different levels of BMI, waist circumference, and waist-hip ratio were included in the models as dummy and categorical variables, and the significance of the trend over different categories of BMI, waist circumference, and waist-hip ratio was tested in the same models by giving an ordinal numeric value for each dummy variable. The proportional hazards assumption in the Cox model was assessed with graphical methods, and with models including time × covariate interactions.²¹ In general, all proportionality assumptions were appropriate. The analyses were first carried out adjusting for age and study year, and then further for smoking, physical activity, educational level, family history of stroke, and alcohol consumption (fruit, vegetable, sausage, and bread consumption in subgroup analyses) (multivariate model 1), and for systolic blood pressure, total cholesterol level, and history of diabetes mellitus (multivariate model 2). To avoid the potential bias due to possible early weight loss in the year before death, additional analyses were carried out excluding the subjects who died during the first 1 year of follow-up (n = 140). All statistical analyses were performed with a commercially available software program (SPSS for Windows, version 14.0; SPSS Inc, Chicago, Illinois).

During an average 19.5 years of follow-up, 1673 men and 1555 women developed an incident stroke event (674 hemorrhagic and 2554 ischemic). General characteristics of the study population at baseline are presented in Table 1.

When we used WHO categories for BMI, men who were underweight (BMI, < 18.5), overweight (BMI, 25.0-29.9), and obese (BMI, ≥ 30.0) had multivariate-adjusted (age, study year, smoking status, physical activity, educational level, family history of stroke, and alcohol drinking [multivariate model 1]) hazard ratios (HRs) of 0.74, 1.23, and 1.59 for total stroke and 0.49, 1.27, and 1.70 for ischemic stroke, respectively, compared with men with a normal weight (BMI, 18.5-24.9) (Table 2). Among women, the corresponding HRs were 1.87, 1.08, and 1.30 for total stroke and 1.81, 1.11, and 1.41 for ischemic stroke. Additional adjustment for systolic blood pressure, total cholesterol level, and history of diabetes mellitus attenuated these associations, particularly in women, but the association remained statistically significant in both sexes. Body mass index did not associate with the risk of hemorrhagic stroke in men and women if we used WHO categories, but a statistically significant U-shaped association was found in women if we divided BMI into 7 categories (< 18.5, 18.5-22.9, 23.0-24.9 [reference group], 25.0-26.9, 27.0-29.9, 30.0-34.9, and ≥ 35.0) (P = .004 for trend). In multivariate analyses (model 1), an increased risk of hemorrhagic stroke was found in women with a low BMI (BMI of < 18.5: HR, 2.90 [95% confidence interval (CI), 1.14-7.40]; and BMI of 20.0-22.9: HR, 1.91 [95% CI, 1.30-2.79]) and a high BMI (BMI of 27.0-29.9: HR, 1.60 [95% CI, 1.08-2.36]; and BMI of ≥ 35: HR, 1.94 [95% CI, 1.16-3.25]) compared with women with a BMI of 23.0 to 24.9 (data not shown). Further adjustment for systolic blood pressure, total cholesterol level, and diabetes mellitus affected the results only slightly.

When BMI was examined as a continuous variable, multivariate-adjusted HRs for each 1-U increase in BMI were 1.04 (95% CI, 1.03-1.06) for total stroke, 1.05 (95% CI, 1.04-1.07) for ischemic stroke, and 1.01 (95% CI, 0.98-1.04) for hemorrhagic stroke in men. Among women, the HRs were 1.03 (95% CI, 1.02-1.04), 1.04 (95% CI, 1.02-1.05), and 1.00 (95% CI, 0.98-1.03), respectively.

When the analysis was restricted to surveys from 1987 to 1997 (men, n = 9186; and women, n = 10 163), the multivariate-adjusted positive associations of total and ischemic stroke across quartiles of waist circumference or waist-hip ratio were found among men but not among women (Table 3 and Table 4). Men with the highest quartile of waist circumference had multivariate-adjusted HRs (model 1) of 1.83 for total stroke and 2.07 for ischemic stroke, compared with men with the lowest quartile of waist circumference (P < .01 for trend for both) (Table 3). Similarly, men with the highest quartile of waist-hip ratio had multivariate-adjusted HRs (model 1) of 1.81 for total stroke and 2.26 for ischemic stroke, compared with men with the lowest quartile of waist-hip ratio (P < .001 for trend for both) (Table 4). After additional adjustment for systolic blood pressure, total cholesterol level, and history of diabetes mellitus, these associations remained statistically significant in men (P < .05 for trend for all). The relative risk of stroke showed almost no change in the multivariate adjustment (model 1), including a dichotomized measure of alcohol consumption compared with another multivariate adjustment, including 4 categories of alcohol consumption plus other dietary variables (fruit, vegetable, sausage, and bread consumption) (data not shown).

The multivariate-adjusted HRs of total stroke in men were 1.02 (95% CI, 1.01-1.03) for each 1-cm increase in waist circumference (as a continuous variable), and 1.40 (95% CI, 1.21-1.63) for each 0.1-U increase in waist-hip ratio (as a continuous variable). In men, the corresponding HRs of ischemic stroke were 1.02 (95% CI, 1.01-1.04) and 1.48 (95% CI, 1.25-1.75). Neither waist circumference nor waist-hip ratio as a continuous variable predicted the risk of total or ischemic stroke in women, or hemorrhagic stroke in men and women. No significant (P > .01) interactions between sex and any 1 of BMI, waist circumference, and waist-hip ratio on total, ischemic, and hemorrhagic stroke were identified. The exclusion of the subjects who died during the first 1 year of follow-up did not affect the associations between the adiposity indicators and the risk of stroke (data not shown). We also found a positive association between BMI and ischemic stroke risk in nonsmokers and smokers (Table 5).

General adiposity, measured by BMI, was associated with an increased risk of total and ischemic stroke in both sexes. Body mass index had a U-shaped association with the risk of hemorrhagic stroke in women. Abdominal adiposity, measured by waist circumference or waist-hip ratio, was associated with an increased risk of total and ischemic stroke in men, but not in women.

The results of the previous studies were contradictory because in some studies BMI has been found as a strong risk factor for stroke,^6- 8 but not in others.^10- 11,14 The Honolulu Heart Program did not find an association between BMI and the risk of stroke among relatively lean US men with Japanese origin.¹¹ In a Swedish study⁷ including 7402 apparently healthy men, BMI had a positive association with the risk of total and ischemic stroke, which is in accordance with our results. However, the researchers did not find any association between BMI and the risk of hemorrhagic stroke.⁷ In the Nurses' Health Study⁵ and the Women's Health Study,⁹ BMI was a risk factor for total and ischemic stroke, but this association was highly mediated by hypertension, diabetes mellitus, and elevated serum cholesterol level. In a recent study¹² of 104 928 Japanese men and women, subjects with a BMI of less than 18.5 had a higher risk of total stroke and intraparenchymal hemorrhage compared with those with a BMI of 23.0 to 24.9. Some studies^3,13- 15 have also suggested that abdominal adiposity, in particular, would be associated with the increased risk of stroke. A previous international comparison²² provided evidence that high BMI was more highly correlated with high waist-hip ratio in Finland than in other European populations. In the present study, BMI predicted total and ischemic stroke risk among men and women; however, waist circumference or waist-hip ratio predicted total and ischemic stroke risk only in men, but not in women, although there were no significant interactions between sex and waist circumference or waist-hip ratio on stroke risk.

Several studies have examined the association between BMI and the risk of hemorrhagic stroke, and the results are inconsistent. Most studies^5,9,12 have found an increased risk of hemorrhagic stroke among lean persons. Other studies, however, found a J-shaped association,⁸ no association,⁷ or an increased risk with increasing BMI.⁶ The few hemorrhagic strokes in most of the studies^5- 7,9,12 may explain in part the inconsistent results between BMI and hemorrhagic stroke. Only 1 Korean study⁸ and our study ascertained more than 600 cases of hemorrhagic stroke during follow-up. In the Korean study,⁸ including 234 864 men, a J-shaped association of BMI with hemorrhagic stroke risk was shown. In the present study, a significant U-shaped association between BMI and the risk of hemorrhagic stroke was found among women, but not among men, if we used 7 categories of BMI. The reason behind this difference in our study is not known. High levels of blood pressure and total cholesterol level may explain at least part of the observed association between high BMI and an increased risk of hemorrhagic stroke, and expanded investigation is required.

Adiposity has been found as a strong risk factor for hypertension,²³ diabetes mellitus,²⁴ and high serum cholesterol level,²⁵ and has been the key or an important component of the metabolic syndrome.²⁶ All these factors were associated with an increased risk of stroke or cardiovascular disease,^26- 30 and were considered as mediating factors for the physiologic effects of adiposity on stroke risk. In the present study, the adjustment for systolic blood pressure, total cholesterol level, and history of diabetes mellitus attenuated the association between BMI and stroke, but BMI remained a statistically significant predictor of total and ischemic stroke in the multivariate model. The failure to control for smoking may also distort the association between adiposity and stroke risk. In the present study, smoking status was considered a confounding factor in the multivariate model and the positive association between BMI and the risk of ischemic stroke was found in nonsmokers and smokers.

There are several strengths and limitations in our study. First, many men and women from a homogeneous population participated in the study. Second, the mean follow-up was sufficiently long to ascertain many type-specific stroke end point events. Third, we had data on standardized measurement of 3 different indicators of adiposity, and many other adiposity-related risk factors, which may modify the association of adiposity with the stroke risk. Because our data allowed for only a dichotomized measure of alcohol consumption in the whole sample, we may not be able to fully control for the effect of this variable on the risk of stroke. Also, we did not have detailed dietary variables in the whole sample. To evaluate the impact of this shortcoming, we performed separate subgroup analyses (surveys of 1987, 1992, and 1997) in the multivariate-adjusted model of a dichotomized measure of alcohol consumption compared with another multivariate-adjusted model of 4 categories of alcohol consumption plus fruit, vegetable, sausage, and bread consumption. In general, the relative risk between BMI and total and ischemic stroke was not influenced substantially or systematically. We cannot completely exclude either the effects of residual confounding due to measurement error in the assessment of confounding factors or some unmeasured factors, such as triglycerides, apolipoprotein B, and other dietary factors.

In conclusion, the present study demonstrates that BMI is associated with an increased risk of total and ischemic stroke in men and women. Furthermore, a substantial additional risk is mediated through adiposity-related risk factors, such as blood pressure, total cholesterol level, and history of diabetes mellitus. Body mass index has a U-shaped association with the risk of hemorrhagic stroke in women. Abdominal adiposity is a risk factor for total and ischemic stroke in men only.

Correspondence: Gang Hu, MD, PhD, Department of Health Promotion and Chronic Diseases Prevention, National Public Health Institute, Mannerheimintie 166, FIN-00300 Helsinki, Finland (hu.gang@ktl.fi).

Accepted for Publication: March 18, 2007.

Author Contributions:Study concept and design: Hu, Tuomilehto, Silventoinen, and Jousilahti. Acquisition of data: Hu, Tuomilehto, and Jousilahti. Analysis and interpretation of data: Hu, Sarti, Männistö, and Jousilahti. Drafting of the manuscript: Hu, Sarti, and Jousilahti. Critical revision of the manuscript for important intellectual content: Hu, Tuomilehto, Silventoinen, Sarti, and Männistö. Statistical analysis: Hu and Sarti. Obtained funding: Hu, Tuomilehto, and Jousilahti. Administrative, technical, and material support: Hu, Tuomilehto, and Jousilahti. Study supervision: Tuomilehto and Sarti.

Financial Disclosure: None reported.

Funding/Support: This study was supported by grant 108297 from the Finnish Academy and by the Finnish Foundation for Cardiovascular Research.

Role of the Sponsors: The funding bodies had no role in data extraction and analyses, in the writing of the manuscript, or in the decision to submit the manuscript for publication.