Ethnic Differences in Risk Factors for Ischemic Stroke
A European Case–Control Study
Background and Purpose— The aim is to estimate the relative risk and population attributable risk (PAR) of risk factors for ischemic stroke by ethnic group.
Methods— In this case–control study, cases of first ischemic stroke were taken from the South London Stroke Register and controls from a cross-sectional prevalence survey covering the same area. PAR was determined for each risk factor by ethnic group. Multivariable analysis was used to examine the association between risk factors and ischemic stroke across all ethnic groups.
Results— 664 cases and 716 controls aged 45 to 74 years were included, with ethnicity of white 78%:42%, black Caribbean 16%:43%, and black African 6%:15%, respectively. For the white group, high PAR was found for ischemic heart disease (IHD) on ECG (56% [95% CI, 49% to 62%]), obesity (49% [95% CI, 40% to 56%]), hypertension (HT) (38% [95% CI, 29% to 46%]), smoking (31% [95% CI, 19% to 41%]), transient ischemic attack (TIA) (23% [95% CI, 19% to 27%]), and atrial fibrillation (AF) (16% [95% CI, 10% to 21%]). In the black Caribbean compared with the white group, PAR was higher for HT (46% [95% CI, 21% to 63%]) and diabetes mellitus (DM) (29% [95% CI, 14% to 42%]), and lower for current smoking (18% [95% CI, 1% to 32%]) and AF (10% [95% CI, 0% to 18%]). In the black African group HT had a higher PAR (59% [95% CI, 91% to 82%]) than the other groups. PAR for AF (11% [95% CI, −11% to 29%]), obesity (30% [95% CI, −20% to 60%]), and DM (4% [95% CI, −25% to 26%]) was low compared with the other groups. In multivariable analysis, risk factors associated with ischemic stroke included TIA, AF, IHD on ECG, smoking, excess alcohol, obesity, HT, and DM.
Conclusion— In the first European case-control study examining risk factors for ischemic stroke in black Caribbean and African populations, some differences were demonstrated in the impact of risk factors between these groups. It may be important to address such differences when developing stroke preventative strategies.
The incidence of stroke in the black population is approximately twice as high in the US black1 and the UK black2 populations compared with the local white populations. Differences in the frequency of risk factors for ischemic stroke may be at least partly responsible.3 Epidemiological studies of ethnic aspects of stroke have been carried out predominantly in America.1,4–6 Data for the UK black population are sparse and characteristics such as risk factor profiles from one ethnic group may not be representative of another.3
The impact of a risk factor on the population, the “population attributable risk,” depends on both the prevalence and risk associated with each risk factor. We have previously demonstrated a higher prevalence of certain physiological risk factors for ischemic stroke in the UK black population3 and in UK black stroke patients,7 compared with their white counterparts. We do not know, however, whether this higher
See Editorial Comment, page 1568
prevalence translates into higher population risk. Also, risk factors may differ between African Caribbean and black African stroke patients as has been demonstrated between these ethnic groups in the general population.3
The aim of this study is to estimate the relative risk and population attributable risk of risk factors for ischemic stroke by ethnic group in the South London population.
Patients and Methods
Patients, registered as part of the South London Stroke Register (SLSR) with ischemic stroke between 1 January 1995 and 31 December 1999, aged 45 to 74 years were included in the study as cases. The SLSR is an ongoing population-based register that enrolls first-ever strokes. Ethical Committee approval was obtained from the Guy’s & St Thomas’ Ethical Committee in accordance with the Helsinki Declaration of 1975, revised in 1983. Detailed methods of patient notification and data collection have been described.2 In brief, all first-ever strokes notified from multiple sources were registered. They were seen by a study doctor who recorded information on demographic details including age, sex, socio-economic status (using the Registrar General’s Codes based on occupation),2 and ethnic origin into white, black Caribbean, black African, black other, and other.2 Only ischemic strokes were included, and these were identified using radiological and necropsy findings or both.2
Data on prestroke risk factors for ischemic stroke for each case were recorded both from the hospital/General Practitioners’ records and self-reported by the patient or carer. Physiological risk factors included ischemic heart disease (IHD), history of transient ischemic attack (TIA), atrial fibrillation (AF), hypertension (HT), and diabetes mellitus (DM). Behavioral risk factors included smoking, heavy alcohol intake, and obesity. Data on risk factors were collected within the first 48 hours after the event, where possible.
A previous history or evidence of AF was noted in addition to a postevent screen for AF using electrocardiography. A previous diagnosis of HT was taken from records or patient recall of high blood pressure requiring treatment (World Health Organization [WHO] classification).8 DM was diagnosed as a record of either diet-controlled, oral hypoglycemic, or insulin-treated disease (WHO classification).9 The patient/carer was questioned about alcohol consumption and smoking. Alcohol consumption was classified as heavy if intake was ≥14 U/week for women and ≥21 U/week for men.10 Smokers were defined as those who had previously smoked either 1 cigarette per day, 1 cigar per week, or 1 ounce of tobacco per month for the duration of 12 months. The quantity of cigarettes smoked was expressed as cigarette pack years (CPY). Waist and hip circumference were measured and a waist to hip ratio (WHR) was determined. Obesity was defined as WHR >0.98 for men and >0.88 for women.11
The control group was taken from a cross-sectional prevalence survey conducted in the same population between 1996 and 1997, full details of which have been published.3 In brief, a postal questionnaire with questions on age, sex, occupation (for socio-economic status, Registrar General’s Classification), and ethnic group (1991 Census question) was sent to 8183 residents aged 45 to 74 years. They were identified from the Family Health Services Authority (FHSA) lists of 16 general practices with the highest proportions of black residents. The responders were stratified by their self-defined ethnic group into white, black Caribbean, and black African. A random sample of 450 subjects in the black Caribbean and white groups were selected and, together with all 193 black African respondents, invited for examination.
A lifestyle questionnaire was administered to controls examining the same risk factors as for cases. Measurement of blood pressure, WHR as a measure of obesity, and recording of an ECG were undertaken at attendance for the interview.
As the control groups were selected on the basis of ethnicity, all univariate analysis was stratified by ethnicity. Univariate analyses of the association between patient demographics and case-control status were performed using χ2 test for each ethnic group. Logistic regression, adjusting for age, sex, and socioeconomic status, was used to determine odds ratios (with confidence intervals) for the association between each risk factor and case-control status within ethnic groups.
Population attributable risk (PAR) was calculated for each risk factor by ethnic group adjusting for age, sex, and social class using the method of Greenland and Drescher as described in the Stata technical bulletin STB42.12 Stata software allows estimation of the PAR from within a logistic regression framework and allows estimation of the summary PARs for a set of exposures.12 PAR measures the proportion of the disease risk in the total population associated with the exposure to that risk factor using Levin’s Formula.13 Levin’s approximation to the PAR assumes that the prevalence of the disease is sufficiently low, so that the proportion with the risk factor in controls is very close to the proportion in the underlying population.14
Unconditional multiple logistic regression was used to determine the odds ratios for each risk factor and case-control status adjusting for age, sex, socio-economic status, and all other risk factors. All ethnic groups were included in this model, but the effect of ethnicity on stroke risk could not be examined because of the sampling frame.
The study included 664 cases of whom 520 (78%) were white, 105 (16%) black Caribbean, and 39 (6%) black African, and 716 controls of whom 300 (42%) were white, 311 (43%) black Caribbean, and 105 (15%) black African.
Table 1 shows characteristics for the 2 groups. Cases were significantly older than controls for all ethnic groups (P<0.001). Cases were more likely than controls to be men (P=0.006 for white; P=0.02 for black Caribbean and African). Socio-economic status differed in the black Caribbean group with significantly fewer manual workers in cases (59% versus 78%; P<0.001).
Table 2 shows proportions of risk factors by ethnicity and case-control status. IHD, AF, and smoking are more prevalent in white cases (29%, 20%, and 46%, respectively) compared with other ethnic groups (20%, 10%, and 27% for black Caribbean, respectively; and 22%, 12%, and 26% for black African cases, respectively). HT is more common in black Caribbean and more so in black African cases (52%, 72%, and 76% for white, black Caribbean, and black African cases, respectively). DM is most prevalent in black Caribbean cases (45%; 16% for white cases; and 19% black African cases).
The Figure shows the PAR for each risk factor by ethnic group. For the white group, high PAR was found for evidence of IHD on ECG (56%), obesity (49%), HT (38%), smoking (31%), TIA (23%), and AF (16%). For the black Caribbean group PAR was higher for HT (46%) and DM (29%) and lower for AF (10%) and current smoking (18%) compared with the white group. PAR for IHD on ECG was higher than in the white group (66%) despite a similar PAR for a history of IHD (10% in both groups). In the black African group, HT had a higher PAR (59%) than the other ethnic groups. The PAR for AF (11%) was lower than in the white group, but similar to the black Caribbean group. Also lower was the PAR for obesity (30%) and DM (4%).
Table 3 shows that risk factors independently associated with ischemic stroke include (in decreasing order of risk) TIA, AF, IHD on ECG, current smoking, excess alcohol, obesity, HT, and DM.
Table 4 shows that ex-smoking is associated with a risk reduction of 0.97 in the white groups (P=0.04), but not in black Caribbean or African groups. No cut off time-period, when the risk of stroke equates to that of nonsmokers, could be demonstrated.
White, but not black Caribbean and African, subjects were at higher risk of stroke at greater CPY smoked. Further analysis found the odds ratio to be 3.4 for those with >100 CPY years and 19.4 for those with >200 CPY (P<0.001) compared with nonsmokers.
The risk of stroke from heavy alcohol intake increased markedly at alcohol levels >45 U per week with an odds ratio of stroke of 7.72 (95% CI, 3.70 to 16.11; P<0.001).
This study has shown that overall the presence of obesity and HT are associated with the highest population risk of stroke. The presence of ischemic ECG changes are also highly associated with stroke, but less so with a history of IHD. The greatest ethnic differences were in the PAR from HT. Other risk factors varied less between ethnic groups. AF is an important population risk factor in the white group and DM in the black Caribbean group. The population risk from smoking is less in the black Caribbean and African groups compared with the white group.
This is a case-control study with substantial numbers in all of the study groups, although numbers of black African cases were insufficient for in-depth analysis of some risk factors. Only subjects aged 45 to 74 years were included in the study, in keeping with aims to reduce the years of life lost by any disease. One limitation of this age range is that it will have excluded stroke cases, as the mean age for stroke in this population is 71.3 years (SD 14.4).2 Cases were of higher mean age and had a higher frequency of males compared with controls, as would be expected.
On retrospective comparison of the age–sex distributions with the source population (adjusted 1991 Census data), it was found that the white female age group 65 to 74 and the black male age group 45 to 54 were under-sampled. The prevalence of risk factors in white controls was, therefore, adjusted to the age–sex distribution of the source population (adjusted 1991 Census data). These adjusted prevalence figures were found to be very similar to the unadjusted prevalence figure used here for all risk factors. Thus, it appears that the under-sampling of control subjects identified here should not have a large effect on the results. Separate figures for black Caribbean and African groups, however, were not available for similar analyses.
The black Caribbean control group had a higher proportion of subjects with lower socio-economic status than the stroke group. This is likely to be because of the selection process, whereby controls were selected from practices with the highest ethnic mix. This may have inadvertently selected more deprived areas. However, these differences in demographics were adjusted for in subsequent analyses.
PAR estimates the proportion of disease incidence due to a particular exposure and assumes that the exposure is causally related to the disease. PAR will increase as the population prevalence and the relative risk of the exposure increases or both. Negative PAR and large and negative confidence intervals are seen, particularly in the black African group, as a result of the small proportion of subjects with these risk factors in this small group. This should be taken into account during interpretation of the data.
A limitation of the study is the low response rate to the initial questionnaire for controls (45%), which is a feature of sampling-deprived inner-city general practices.3 It may have overestimated the prevalence of risk factors if people with health problems were more likely to respond to an invitation for examination. The use of FHSA lists as a sampling frame for controls may have introduced selection bias caused by problems with incompleteness, wrong addresses and dates of birth, deceased patients, etc. In addition, these problems may all differ between the ethnic groups. However, these problems are unlikely to have had a major effect on the results, as a study in a neighboring district15 with a response rate of 65% found comparable prevalence of risk factors in these ethnic groups. Their higher response rate was probably a result of sampling practices with computerized age–sex registers, which were likely to reflect practices catering to higher social class patients.
Risk Factors for Ischemic Stroke
HT was significantly associated with stroke in keeping with previous studies.6,16 Ethnic differences were demonstrated in the PAR for HT with values of 38% for white, 46% for black Caribbean, and 59% for black African subjects. Previous studies found a higher prevalence of HT in UK black stroke patients7 and in UK and US black populations.16,17,18 This may be, at least partly, because of differences in the definitions used for HT. To our knowledge, there are no previous European data examining differences between black African and black Caribbean stroke patients for comparison. Black African subjects had the highest prevalence of HT in this study (76% versus 72% for black Caribbeans and 53% for whites; figures given are for cases). US black hypertensives have been found to have higher BP than white hypertensives.17,18
In this study, DM was a greater risk in black Caribbean subjects, with more than a 2-fold difference over white and black African groups. These findings are in accordance with the higher prevalence of HbA1c above 6% found in black Caribbean groups in the same population.3 Other studies have found high proportions of DM in black Caribbean populations in the UK15 and in the Caribbean.19
The black Caribbean and African groups had low PAR from AF at 10% to 11% compared with 16% in the white group, in line with previous findings in US black stroke patients.6,20
Behavioral Risk Factors
We have previously shown that UK black stroke patients have a lower prevalence of smoking and heavy alcohol intake.7 This study further investigated this difference between black ethnic groups and both dose–response effects and the effect of smoking cessation.
In multivariable analysis, current smoking was an independent risk factor and its PAR differed between ethnic groups. Previous studies have similarly shown lower levels of smoking in the black general population3,15 and stroke patients.7 This is in contrast to US blacks, shown to have a higher prevalence of smoking,17 despite smoking fewer cigarettes per day.21
A positive dose–effect of cigarette smoking on stroke was found overall, but when analyzed by ethnic group, the dose–effect seemed to exist only in the white group. Ex-smoking was not associated with stroke in multivariable analysis. Further analysis showed that ex-smoking was also significant only in the white group. This is unlikely to be because of small numbers in the black groups, as analysis of the total sample did not give significant results either.
Overall, heavy alcohol intake was not an important risk factor for stroke in any ethnic group in agreement with previous studies.16 However, examination of the dose effect of heavy alcohol intake shows that the risk rises sharply once weekly intake exceeds 45 U.
Alcohol intake has been reported to be higher in US black stroke patients20 but lower in UK black stroke patients compared with their white counterparts.7 The PAR was higher for the black African compared with white and black Caribbean groups. This is likely to be because of small numbers, as the prevalence of heavy alcohol intake for black African cases was lower than that of white cases (17% versus 23%).
Obesity was independently associated with a higher risk of stroke, in agreement with previous studies.8 This study has demonstrated similar PAR for WHR between the different ethnic groups.
In the first European case-control study examining risk factor differences for the UK black Caribbean and African populations, some differences have been demonstrated in the risks and impact of risk factors for ischemic stroke between the different ethnic groups. The treatment of stroke rests largely in the domain of primary and secondary prevention. This study helps us to further understand which risk factors should be targeted to maximize the reduction in stroke risk according to the patients’ ethnic group. It also clarifies issues about behavioral risk factors such as the dose effects of smoking and alcohol intake and ethnic differences in these.
We are only now beginning to understand that different risk factors have different effects on the risk of disease according to other factors, such as the patients’ ethnic background. With this in mind, prevention programs specific for the patients’ ethnic background may be more relevant. For example, in a black Caribbean subject, optimum control of hypertension and diabetes mellitus may be more important than measures to stop smoking. Further work in this area is now needed to complete the picture of risk factors, modifiable and nonmodifiable, for ischemic stroke.
Dr C. Hajat was funded by a UK National Health Service (NHS) Research & Development Grant. This project was funded by the Northern & Yorkshire Research & Development Programme, the Charitable Foundation for Guy’s & St Thomas’ Hospitals, the Stanley Thomas Johnstone Foundation, and the Stroke Association. We would like to thank other members of the South London Stroke Register team who collected data for this project, in particular Dr E. Lawrence, Dr T. Hillen, Dr A. Bhalla, Dr M. Patel, and Ms R. Dundas.
- Received January 2, 2004.
- Revision received April 14, 2004.
- Accepted April 23, 2004.
Friday G, Lai SM, Alter M, Sobel E, LaRue L, Gil-Peralta A, McCoy RL, Levitt LP, Isack T. Stroke in the Lehigh Valley: racial/ethnic differences. Neurology. 1989; 39: 1165–1168.
Stewart JA, Dundas R, Howard RS, Rudd AG, Wolfe CDA. Ethnic differences in stroke incidence; a prospective study using a stroke register. BMJ. 1999; 318: 967–971.
Sacco RL, Boden-Albala R, Gan X, Chen D, Kargman E, Shea S, Paik MC, Hauser WA. Stroke incidence among white, black, and Hispanic residents of an urban community: the Northern Manhattan Stroke Study. Am J Epidemiol. 1998; 147: 259–268.
Sacco RL, Boden-Albala B, Abel G, Lin I-F, Elkind M, Hauser WA, Paik MC, Shea S. Race-ethnic disparities in the impact of stroke risk factors. Stroke. 2001; 32: 1725–1731.
Hajat C, Dundas R, Stewart JA, Lawrence E, Rudd AG, Howard R, Wolfe CDA. Cerebrovascular risk factors and stroke subtypes: differences between ethnic groups. Stroke. 2001; 32: 37–42.
Barker J, Cryer C, Maryon-Davis A. Health Quest South East: Regional Report. South Thames Regional Health Authority, 1993.
Walker SP, Rimm EB, Ascherio A, Kawachi I, Stampfer MJ, Willett WC. Body size and fat distribution as predictors of stroke amongst US men. Am J Epidemiol. 1996; 144: 1143–1150.
Brady, AR. Adjusted population attributable fractions from logistic regression. In: Newton, HJ ed. The Stata Technical Bulletin Reprints, Vol 7. College Station, TX: Stata Corporation; 1998: 137–143
Levin ML, Bertell SR. RE: “simple estimation of population attributable risk from case-control studies.” Am J Epidemiol. 1978; 108: 78–79.
Levin ML. The occurrence of lung cancer in man. Acta Unio Int Contra Cancrum. 1953; 9: 531.
Cappuccio FP, Cook DG, Atkinson RW, Strazzullo P. Prevalence, detection, and management of cardiovascular risk factors in different ethnic groups in south London. Heart. 1997; 78: 555–563.
Spriggs DA, French JM, Murdy JM, Bates D, James OFW. Historical risk factors for stroke: a case control study. Age and Ageing. 1990; 19: 280–287.
Heyman A, Karp HR, Heyden S, Bartel A, Cassel JC, Tyroler HA, Hames CG. Cerebrovascular disease in the biracial population of Evans County, Georgia. Arch Intern Med. 1971; 128: 942–955.
Sacco RL, Kargman DE, Zamanillo MC. Race-ethnic differences in stroke risk factors among hospitalized patients with cerebral infarction: the Northern Manhattan Stroke Study. Neurology. 1995; 45: 659–663.
Mast H, Thompson LP, Lin I-F, Hofmeister C, Hartmann A, Marx P, Mohr JP, Sacco RL. Cigarette smoking as a determinant of high-grade carotid artery stenosis in Hispanic, black, and white patients with stroke or transient ischemic attack. Stroke. 1998; 29: 908–912.