Racial differences in the incidence of intracerebral hemorrhage

Methods.

NHANES I was conducted by the National Center for Health Statistics, Centers for Disease Control and Prevention from 1971 through 1975 to collect health-related information on a probability sample of the US civilian, noninstitutionalized population. To increase the sample size in select subgroups, the elderly, persons living in poverty areas, and women of childbearing age were oversampled.8-10

The objective of the NHEFS was to provide follow-up to the 14,407 examinees in NHANES I who were between the ages of 25 and 74 years at the baseline survey.11 Data collection for this analysis included tracing all NHANES I participants for morbidity and mortality through 1992.12 Persons who reported a history of stroke during the baseline NHANES I survey were excluded from this analysis.

ICH events were determined by hospital record or death certificate diagnosis that included one of the International Classification of Diseases (ICD-9), Clinical Modification codes 431-432. There were 78 ICH events during the 20-year follow-up (60 among whites and 18 among African Americans).

Potential baseline risk factors for ICH included age, sex, race (African American or white), education (<12 or ≥12 years), systolic blood pressure, diabetes mellitus, serum cholesterol, smoking status (current nonsmoker or smoker), body mass index (weight [kg]/height [m²]), and consumption of any alcohol (yes, no). All measures were obtained prospectively during the NHANES I baseline interview. Socioeconomic data and smoking status were self-reported. Diabetes mellitus was determined by either patient self-report or by physician coding. Measurements of height and weight were taken by trained examiners during the NHANES I physical examination. Three sitting blood pressure determinations were obtained, and the mean of the second and third blood pressure determinations were used for this analysis.

Because the interval of follow-up varied among individuals, Cox proportional hazards analysis was used to estimate the relative risk (RR) for ICH among African Americans compared with that among whites. Ninety-five percent confidence intervals (CIs) were calculated using a Taylor series approximation for the standard error of the RR.13

Results.

Baseline characteristics differed markedly by race (table 1). The proportion of men and persons with 12 or more years of education was significantly higher among the white subjects; African Americans had higher mean systolic blood pressure and body mass index than their white counterparts. The proportion of subjects with diabetes mellitus and nonsmokers was also higher among African Americans. The prevalence of alcohol consumption and mean cholesterol level was higher among white than African American adults.

Persons who developed ICH were more likely to be older, have diabetes mellitus, have higher levels of systolic blood pressure and cholesterol, and were less likely to have 12 or more years of education (table 2). Body mass index, cigarette smoking, and alcohol consumption did not differ between persons who did or did not develop ICH during follow-up. The estimated annual incidence of ICH was 50 per 100,000 among African Americans and 28 per 100,000 among whites. The age-adjusted risk for ICH among African Americans was almost twofold higher than that among whites (table 3). After adjustment was made for differences in age, gender, education, systolic blood pressure, cholesterol, body mass index, diabetes mellitus, smoking status, and alcohol consumption, African Americans had a 50% increased risk for developing ICH; however, this increased risk failed to reach statistical significance.

Finally, we sequentially added cardiovascular disease risk factors to the Cox proportional hazards model to determine the effect each risk factor had on the excess ICH risk observed in African Americans (figure). After we adjusted for age and gender, African Americans were significantly more likely to develop ICH (RR = 1.9, 95% CI 1.1 to 3.2). With the addition of educational attainment to the latter model, the RR was reduced to 1.7 (95% CI 1.0 to 2.9); the addition of systolic blood pressure further reduced the RR to 1.6 (95% CI 0.9 to 2.7).

• Download figure
• Open in new tab
• Download powerpoint

Figure. African American–white relative risk for intracerebral hemorrhage after adjusting for selected risk factors (First National Health and Nutrition Examination Survey Epidemiologic Follow-up Study). The first model is adjusted for age and sex only; the second model for age, sex, and education (edu); the third model for age, sex, education, and systolic blood pressure (bp); and the fourth model is the multivariate model adjusted for age, sex, education, systolic blood pressure, serum cholesterol, body mass index, diabetes mellitus, smoking status, and alcohol consumption.

Discussion.

During the 20-year follow-up, we found that risk of ICH for African Americans was twice that of their white counterparts. Most of this excess could be attributed to differences in educational attainment and systolic blood pressure. Interestingly, after adjustment for other known risk factors, the RR for ICH among African Americans was still 50% higher than that for whites. Even though the excess risk for ICH has been identified in previous population-based case-control studies,1-5 the effect of various risk factors, including systolic blood pressure, on the development of ICH has not been documented in a longitudinal study. This longitudinal study attempted to identify the racial differences in the prevalence of certain risk factors that may predispose African Americans to a higher risk of ICH.

The estimated annual incidence of ICH among African Americans was 50 per 100,000, which was almost twofold higher than that among whites (28 per 100,000). The overall estimated incidence was 30 per 100,000, which is similar to the 23 per 100,000 reported from south Alabama3 and 22 per 100,000 reported from Oxfordshire, England.14 The incidence in this study was higher than those reported from Cincinnati (15 per 100,000)1 and Dijon, France (14 per 100,000).15 The incidence among African Americans was similar to that among the Japanese population (55 per 100,000).16 Differences between study populations in the incidence of ICH may reflect regional variations in case ascertainment, age and ethnic distributions, access to health care, and prevalences of predisposing risk factors.

The high risk for ICH among African Americans was recognized in the last two decades. Epidemiologic studies of stroke from south Alabama,3 northern Manhattan,2 and Cincinnati1 have consistently reported a greater incidence of ICH among African Americans compared with whites, particularly those aged 30 to 50 years. Sacco et al.2 reported an incidence among African Americans residing in northern Manhattan of 64 per 100,000 for men and 60 per 100,000 for women. In a population-based cross-sectional study, Broderick et al.1 reported that the risk for ICH was twofold higher for African Americans than it was for whites. The increased risk was predominately observed in young and middle-aged persons. They concluded that the racial differences in incidence of ICH became less prominent with advancing age. Because 75% of the participants in the present study were under age 65, our results are likely to be a reflection of ICH occurring among young and middle-aged adults.

We found that most of the excess risk for ICH among African Americans was accounted for by differences in educational attainment and blood pressure. Previous studies have postulated that the excess risk of ICH may reflect the higher prevalence of hypertension among African Americans.1,3 Elevations in blood pressure were a major risk factor for ICH in this and other studies.17-19 Previous studies using population-based screening have also already documented a higher prevalence of hypertension among African Americans.20-22

The role of educational attainment as a risk factor for ICH is less well described. However, low educational attainment has been identified as a risk factor for stroke in other studies,23-25 and is an important risk factor for cigarette smoking and the development of hypertension, obesity, and high blood cholesterol. Furthermore, persons of lower socioeconomic status are substantially less likely to have access to health care,23 which may also affect the risk for ICH.

Even after adjustment for all risk factors, the incidence of ICH was 50% higher among African Americans, suggesting that the addition of other, unidentified risk factors, such as limited access to health care, physical inactivity, and gene-environmental interactions, to the Cox proportional hazards model may have provided further explanation for racial differences in the incidence of ICH. This hypothesis is supported by findings from a previous report that assessed the determinants of racial differences in the risk of cerebral infarction among NHANES I participants during the 13-year follow-up.26 Young African Americans in that study had a 70% increased risk for cerebral infarction even after adjustment for traditional risk factors. In addition, Kittner et al.27 studied reasons for racial differences in stroke incidence using 10-year follow-up from the NHANES I survey. Interestingly, they reported that the RR for stroke associated with hypertension and diabetes did not differ by race.

Limited information is available regarding the associations between either cholesterol or diabetes mellitus and the likelihood for ICH. Iribarren et al.28 reported that there was an inverse relationship between total cholesterol levels and the likelihood of ICH; however, this finding was limited to men over the age of 65. Sharma et al.29 examined the association between hemorrhagic stroke and glycemic control and reported that diabetics had a higherprevalence of severe hemorrhagic stroke than euglycemic controls. Additional studies need to examine the associations between cholesterol, diabetes mellitus, and the likelihood for ICH.

This study is subject to a number of potential limitations. The number of stroke events among African Americans was small (n = 18), limiting our ability to reach statistical significance in some of our analyses. We were unable to examine racial differences by gender or age group for the same reason. African Americans were more likely than white participants to be unavailable for follow-up or to have had only vital status ascertained.26 However, follow-up was available for more than 90% of the cohort, and the race-related differences in follow-up were most prominent among persons under age 35 and were not seen among persons over age 55. Because the majority of ICHs occur among older adults, participants lost to follow-up could not have substantially biased the results presented in this report. In addition, these results are consistent with those from a number of other investigators.1-5 Another potential limitation is the use of ICD-9 codes from discharge abstracts and death certificates to identify incident cases of ICH. Broderick et al.30 reported a positive predictive value of 83% for ICD-9 diagnoses of ICH, and Leibson et al.31 reported a positive predictive value of 87% in the Rochester Stroke Registry. Iso et al.32 reported that the sensitivity for a death certificate diagnosis of hemorrhagic stroke was higher than that for ischemic stroke, and that death certificate diagnoses are sufficiently accurate to justify their use in epidemiologic studies. Probably most important, the validity of ICD-9 diagnoses does not differ by race. A further limitation is that data on stroke risk factors were limited to information collected during the NHANES I baseline interview. Change in risk factor status during the 20-year follow-up was not considered. Finally, the analysis is likely to underestimate the contributions of risk factors with measurement error such as blood pressure33 and cholesterol.34 Because of this underestimation, we may overestimate the magnitude of the adjusted African American–white difference in ICH risk.

Despite these limitations, this is one of the first studies to examine racial differences in the incidence of ICH. Given the high mortality associated with ICH, an increased emphasis on primary prevention is key if racial disparities of ICH are to be eliminated. These primary prevention efforts should include an increased emphasis on the prevention and treatment of hypertension. Key components of hypertension prevention strategies should include efforts to increase physical activity, limit sodium and alcohol intake, and promote the maintenance of adequate calcium and potassium intake.35 Furthermore, unless racial differences in socioeconomic status are eliminated, the African American–white disparities in ICH will likely not be eliminated.