Familial History of Stroke and Stroke Risk
The Family Heart Study
Background and Purpose Although familial history of stroke is generally perceived to be an important marker of stroke risk, very few epidemiological studies have been published to address this hypothesis. We sought to examine whether familial history of stroke is associated with the prevalence of stroke in the Family Heart Study, a National Heart, Lung, and Blood Institute–supported multicenter study of the familial, genetic, and nongenetic determinants of cardiovascular disease in populations.
Methods The personal and familial histories of stroke were assessed in 3168 individuals (probands) who were at least 45 years old and 29 325 of their first-degree relatives with the use of a standardized questionnaire.
Results The age-, ethnicity-, and sex-adjusted stroke prevalences were 4.8%, 4.9%, and 3.9% in probands with a positive familial, paternal, and maternal history of stroke, respectively, in comparison with 2.0% in probands without any positive familial history (P<.01). The age-, ethnicity-, and sex-adjusted odds ratios (95% confidence interval) of stroke were 2.00 (1.13, 3.54) for a positive paternal and 1.41 (0.80, 2.50) for a positive maternal history of stroke. Additional statistical adjustment for the proband’s history of elevated cholesterol level, cigarette smoking status, history of coronary heart disease, hypertension, and diabetes did not alter the associations. A similar pattern was seen for African Americans and European Americans.
Conclusions The increased risk of stroke among persons with a positive familial history of stroke compared with those without a familial history of stroke is consistent with the expression of genetic susceptibility, a shared environment, or both in the etiology of stroke.
Although family history of stroke is often considered to be a predictor of stroke in the offspring, little information is available from family study data documenting such an association in populations, and the findings published by this limited number of studies have been inconsistent.1 2 3 4 5 6 7 8 At present, there are no data to address this issue in African Americans, a group characterized by a higher stroke risk than European Americans.
It is biologically plausible to posit that a family history of stroke can influence a person’s risk of stroke by means of genetic or environmental factors or their combined effects. The purpose of this study is to test whether a history of stroke expressed in first-degree relatives is associated with the reported frequency of occurrence of stroke at the population level and to explore the effects of ethnicity and sex on this hypothesized association.
Subjects and Methods
The population for this study was enrolled by the Family Heart Study (FHS), a population-based, multicenter study of the genetic and nongenetic determinants of cardiovascular disease supported by the National Heart, Lung, and Blood Institute. The design and objectives of the FHS have been reported in detail.9 Briefly, during the initial phase of the study (phase I), approximately 1000 individuals (probands) per field center were identified from the participants of three ongoing parent cohort studies: the Forsyth and Minneapolis cohorts of the ARIC Study,10 the Framingham Offspring Study,11 and the Utah Family Tree Study.12
The Framingham Heart Study is a longitudinal investigation of constitutional, environmental, and genetic factors influencing the development of cardiovascular disease in men and women. The original Framingham cohort was recruited in 1948, and the offspring of the Framingham cohort members were recruited in 1970 to form the Framingham Offspring Study. From this offspring population, 2650 probands from independent families were available for the FHS.
ARIC is a population-based longitudinal study of atherosclerosis and its sequelae sponsored by the National Heart, Lung, and Blood Institute. The ARIC cohort was selected as a probability sample of 15 800 men and women between the ages of 45 and 64 years at entry from four study centers in the United States (Washington County, Maryland; Forsyth County, North Carolina; and selected suburbs of Minneapolis, Minn, and Jackson, Miss). Only the cohort members from Forsyth County, North Carolina, and Minneapolis, Minn, were selected to participate in the FHS as probands. In this process, approximately 4000 probands from each center were available for the FHS.
The Utah Family Tree Study in Salt Lake City began in 1983 and was designed to collect histories of heart disease and other conditions in the family members of 50 000 high school students through questionnaires. The FHS selected its probands from the students’ parents, uncles, or aunts. In this process, approximately 4000 potential probands from independent families were available for the FHS.
The sampling processes for the FHS included a simple random sample of approximately 500 probands from each of the study sites and another 500 probands with a high family risk score for CHD. This score was calculated based on the self-reported frequency of CHD in first-degree relatives and the expected frequency of CHD estimated from the Framingham cohort experience. In this process, 3168 probands, 6283 parents (3140 fathers and 3143 mothers), 2834 current spouses, 12 140 siblings, and 10 902 children in the probands’ families were recruited to participate in phase I of the FHS. Personal history of stroke, CHD, myocardial infarction, diabetes mellitus, hypertension, and smoking history were obtained for the probands and their family members by self-administered, standardized questionnaires, which were completed by the probands and their first-degree relatives. After the aforementioned phase I data collection, 150 randomly selected families and 150 higher-risk families were invited to a detailed clinical examination (phase II). During the clinical examination, data on medical history, lifestyle factors, electrocardiogram, anthropometry, blood pressure, blood chemistry, pulmonary function, and ultrasound measurement of carotid arterial wall thickness were collected on all available family members. In this report, only the data from phase I of the FHS were used.
The history of stroke for the probands and their family members was based on the response to the question “Have you ever had a stroke?” on a standardized, self-administered questionnaire. The categories of familial history of stroke for probands were defined as follows: (1) familial history of stroke: a positive history of stroke reported by any of the biologically related first-degree relatives (excluding spouse); (2) parental history of stroke: a positive history of stroke reported by one or both biological parents; (3) paternal history of stroke: a positive history of stroke reported by the biological father; (4) maternal history of stroke: a positive history of stroke reported by the biological mother; (5) sibling history of stroke: a positive history of stroke reported by any of the proband’s siblings; (6) history of stroke in the offspring: a positive history of stroke reported by any of the proband’s offspring; and (7) spouse history of stroke: a positive history of stroke reported by the current spouse of the proband. To elucidate differences in paternal and maternal history of stroke on stroke risk of the proband, probands with a positive maternal history were excluded from the analyses when paternal history of stroke was assessed. Similarly, when the maternal history was the focus, probands with a positive paternal history were excluded from the analyses.
Mean age; proportion of African Americans; proportion of female probands; prevalence of CHD, myocardial infarction, hypertension, and diabetes mellitus; and positive parental history of stroke were calculated for all probands and stratified by proband history of stroke. Multivariable adjusted prevalence of proband stroke was estimated according to positive/negative familial history of stroke in the aforementioned categories. The relative odds of a proband stroke (prevalence OR), given a positive familial history of stroke in the aforementioned categories of familial history exposure, were estimated with the use of logistic regression models. Possible interactions of ethnicity, sex, and age with parental history of stroke were tested by −2 log likelihood ratio tests. Effect modifications by FHS sampling strata in the association of familial history of stroke to proband stroke risk were tested in all models and were not found to be statistically significant at the P<.20 level. Consequently, overall results are presented, including an adjustment for the FHS sampling stratum, instead of reporting stratum-specific results.
Twenty-four of the 3168 probands recruited for FHS phase I were excluded from this report because of missing information on personal history of stroke. Among 3144 probands included in this report, 105 (3.3%) reported a positive personal history of stroke. The characteristics of the probands are presented in Table 1⇓. The mean age was 60 years, 50% were female, and 93% were European Americans. As can be seen in this table, probands with a positive history of stroke were older, more likely to be male, African American, and a current smoker, and had a higher frequency of reported history of CHD, diabetes, hypertension, and hypercholesterolemia.
The proband’s age-, sex-, and ethnicity-adjusted frequency of positive family history of stroke according to proband stroke status is presented in the Figure⇓. Probands who reported a personal history of stroke (n=105) had a higher prevalence of familial history of stroke, and specifically of paternal history and maternal history of stroke, than probands without a stroke. Probands who reported a positive history of stroke had a higher prevalence of spouse history of stroke than probands who did not report a history of stroke, but this difference was not statistically significant. No meaningful differences were found between probands with and without a history of stroke in terms of sibling’s history of stroke or in history of stroke in the offspring.
To assess the relative odds of stroke in the probands associated with a positive familial history of stroke and to examine the role of conventional stroke risk factors on the familial aggregation of stroke, three sets of logistic regression models were fit with familial history of stroke in first-degree relatives as main predictors: (1) an unadjusted model; (2) a model that included proband’s age, ethnicity, and sex as covariates; and (3) a model with additional adjustment for proband’s history of hypertension, prevalent CHD, diabetes, elevated cholesterol, current cigarette smoking status, and sampling stratum. The results are summarized in Table 2⇓. It can be seen from the magnitude of the estimates in the three sets of models that a positive familial history, a positive parental history, and a positive paternal history were associated with higher prevalent odds of stroke in the probands. The strengths of the associations were in the order of paternal history, parental history, and any familial history, with age-, ethnicity-, and sex-adjusted ORs (95% CI) of 2.00 (1.13, 3.54), 1.82 (1.20, 2.77), and 1.60 (1.08, 2.39), respectively. Maternal history of stroke was only marginally associated with proband’s risk of stroke, with an age-, ethnicity-, and sex-adjusted OR (95% CI) of 1.41 (0.80, 2.50). A history of stroke in a sibling or in offspring was not statistically significantly associated with proband stroke in any of the three models. History of stroke in a spouse was not statistically significantly associated with proband stroke; the age-, ethnicity-, and sex-adjusted OR (95% CI) was 1.82 (0.61, 5.41). Older age, male sex, and African American ethnicity were significantly associated with higher odds of stroke in the probands. However, adjustment of these three demographic variables did not alter the point estimates seen in the unadjusted model. Although the proband’s history of CHD, hypertension, diabetes, and current smoking was significantly associated with higher odds of stroke, the inclusion of these factors as covariates in the multivariable adjusted models (adjusted model B in Table 2⇓) did not appreciably alter the familial history and proband stroke associations. Table 3⇓ presents the results from a multivariable-adjusted logistic regression model that includes any familial history of stroke as predictor and the conventional risk factors of stroke as covariates. Proband age, ethnicity, and sex were tested as potential effect modifiers (interactions) in the aforementioned multivariable adjusted proband’s familial history and stroke risk models and were not found to be statistically significant at the P<.20 level.
Familial history of stroke has been hypothesized as a risk factor for stroke.1 2 3 4 5 6 7 8 Considering the disease process, it is plausible that a positive parental history of stroke may lead to an increased risk of stroke events through (1) genetic heritability of stroke risk factors, such as elevated blood pressure, elevated serum cholesterol, and diabetes; (2) the inheritance of susceptibility to the effects of such risk factors; (3) familial sharing of cultural/environmental and lifestyle factors, such as higher-sodium and higher-fat diet, lower physical activity, and lower socioeconomic status; and/or (4) the interaction between genetic and environmental factors. Although there is increasing evidence to support the genetic inheritance of elevated blood pressure and documentation of the familial aggregation of cultural, environmental, and lifestyle factors within families, the formal testing of the hypothesis of familial history of stroke and the risk of stroke at the population level has been limited; moreover, the results have been inconsistent. Marshall,1 Khaw and Barrett-Connor,2 Welin et al,3 Carrieri et al,4 and Kiely et al5 have reported some degree of familial and/or parental aggregation of stroke. However, Herman et al,6 Boysen et al,7 and Brass and Shaker8 found no statistical association. The conflicting results from these studies may be a consequence of differences in study design, the methods used to identify stroke events in the family members, small sample size, or the choice of study population. As reported by Kiely and coworkers,5 using validated stroke and transient ischemic attack in the original cohort and in the offspring of the original cohort of the Framingham Study to assess the association of parental history of stroke and the risk of stroke in the offspring generation, a trend suggestive of an association was found (risk ratio of 1.56 and CI of 0.76 to 3.19). Since the study of Kiely et al was based on only 33 stroke events in 2268 offspring, lack of statistical power may explain the width of the 95% CI.
We found a familial history of stroke, specifically parental and paternal history of stroke, to be statistically significantly associated with an increased risk of a prevalent stroke in the probands. Adjustment for proband age, ethnicity, and sex did not change the associations, nor did adjustment of conventional risk factors for stroke. These results suggest that stroke aggregates within families and that the familial aggregation of stroke is independent of age, ethnicity, and sex. The findings also suggest that the familial aggregation of stroke cannot be fully explained by the familial aggregation of conventional risk factors tested in this study. Our data also showed a trend of association between spouse history of stroke and proband stroke, although this did not attain statistical significance. No association was found between stroke in the probands and their offspring in this population. This may be attributable to the young age of the proband’s offspring (mean age=32 years, with only 0.8% of reported stroke history at the time of their participation in this study). Similarly, no association was found between stroke in the probands and their siblings. This may be due to the lack of statistical power, since only 12 probands with a stroke had a positive history of stroke in their siblings in this population.
These findings are consistent with a familial transmission of susceptibility to stroke, which can be the result of genetic inheritance and/or common cultural and environmental exposures. Our findings that the familial aggregation of stroke is not accounted for by the inheritance of conventional stroke risk factors may be due to the inheritance of host susceptibility or factors other than those included in our multivariable adjustment models. An alternative explanation is that this finding is evidence for heterogeneity between families in the pathogenesis of stroke. Evidently, not all stroke victims have the same risk factor leading to a stroke. Therefore, any linear model that averages the effect of a risk factor, such as hypertension, across a heterogeneous population may show a reduced or absent effect for that risk factor. For instance, if some strokes are atherothrombotic while others are hemorrhagic, the linear estimates of the effects of lipid variables may be diminished after these two types of stroke are combined. By contrast, a variable such as family history of stroke is less affected by heterogeneity, since it is a family-specific variable (a shared risk factor within a family will create a positive family history, as will a different risk factor shared by a different family). Even if different types of families are combined in an analysis, family history will not be affected by heterogeneity, since the family-specific histories are positive, regardless of the reason for their being positive. It can be expected that family history of disease will be a statistically independent risk factor in populations in which multiple heterogeneous risk factors are present, even after multivariable adjustment for causal mechanisms leading to the disease.13 If heterogeneity of the continuous risk factors could be appropriately modeled, only then would the family history effect be decreased in strength. Further study is needed to examine the quantitative roles of genetic and shared environment contribution in the inheritance of stroke, as well as their interaction, and to investigate the mechanisms of the familial aggregation.
Regrettably, even the recruitment of the largest number of African American families into a population-based family study (n=101) of cardiovascular disease reported to date yields a small number of strokes (n=6), which precludes the opportunity to test the hypothesis that a familial history of stroke is a predictor of stroke in this population group. On the basis of tests of the interaction of proband history of stroke with proband age, sex, and ethnicity, the most parsimonious explanation suggests a homogeneous pattern of association across age, ethnicity, and sex strata. However, caution is warranted when the interaction between familial history of stroke and ethnicity is interpreted because the statistical power for a formal test of the interaction of ethnicity was markedly limited.
Personal history of stroke was obtained directly from the probands and from their first-degree relatives, instead of relying on the more usual proband-reported familial history of stroke, since probands who have experienced a stroke may be more likely to learn about stroke events in their family members and to report such events. This would have reduced the possibility of differential misclassification of familial history of stroke to a considerable degree. In clinical settings, the family history of stroke is usually obtained from the proband instead of from family members directly, as in our case. In the FHS population, the overall kappa statistics for the agreement of proband-reported family history and the self-reported personal history of stroke in members of the proband’s family are 0.77 for proband-reported father’s history versus father’s self-reported history of stroke; 0.76 for proband-reported mother’s history versus mother’s self-reported history, and 0.69 for proband-reported sibling’s history versus sibling-reported history of stroke.14 We also analyzed the data using proband-reported family history as predictor of proband’s risk of stroke, and we observed that the patterns of associations did not differ from the current findings (data not shown). This suggests that our findings are applicable to situations in which family history of stroke is obtained from the probands. However, it should be noted that self-reported history of stroke was not validated by medical records, and misclassification cannot be ruled out. In a population-based sample of men and women aged 45 years and older, the question “Have you ever had a stroke?” was reported to have a sensitivity of 95% and specificity of 96% compared with stroke diagnosis made by a panel of neurologists based on medical records.15 Furthermore, it can be speculated that the misclassification of stroke due to disagreement of the questionnaire assessment and the review of medical records is likely to be nondifferential on familial history of stroke in this population. Thus, the estimated association between family history and stroke risk is likely to be underestimated to some degree. Another potential bias in this study is that probands who experienced a positive familial history of stroke were more likely to report a personal history of stroke for themselves, biasing the results toward a positive association. In our data, proband’s age, ethnicity, sex, self-reported history of CHD, and diabetes are significantly associated with the prevalence of proband stroke (Table 3⇑). These results add indirect support to the validity of our self-reported stroke.
In assessing the potential differences in paternal and maternal history of stroke on proband’s risk of stroke, we excluded probands with a positive maternal history when assessing paternal history and vice versa. This could introduce bias due to the exclusion of probands for whom both parents reported a personal history of stroke. However, as can be seen in Table 1⇑, such probands represented only 3% of all probands and only 1% of probands with a personal history of stroke. Thus, it is unlikely that our results would be attributable to this exclusion. For the same reason, we cannot assess a “dose-response” relationship using the number of parents reporting a history of stroke.
In summary, familial history of stroke, especially parental history, is significantly associated with an increased risk of stroke in the offspring in this population-based family study. The familial aggregation is not confounded by age, ethnicity, and sex and cannot be fully explained by the conventional risk factors studied, suggesting familial aggregation of host susceptibility or factors other than conventional stroke risk factors. If these findings are replicated in other populations, the use of a simple, low-cost ascertainment of self-reported stroke among first-degree relatives should be of interest to and widely applicable in epidemiological studies.
Selected Abbreviations and Acronyms
|ARIC||=||Atherosclerosis Risk in Communities|
|CHD||=||coronary heart disease|
|FHS||=||Family Heart Study|
This study was supported by National Heart, Lung, and Blood Institute contract N01-HC-25107 and grant U01-HC-56563. The authors wish to acknowledge the valuable contributions made by the FHS staff at the collaborating institutions: The University of North Carolina at Chapel Hill: Phyllis H. Johnson; Bowman-Gray School of Medicine, Winston-Salem, NC: Jeannette Bensen; Washington University, St Louis, Mo: Jeanne C. Cashman; and University of Minnesota, Minneapolis: Gail Murton.
Reprint requests to Dr Duanping Liao, Department of Epidemiology, School of Public Health, University of North Carolina at Chapel Hill, 137 E Franklin St, NationsBank Plaza, Suite 306, Chapel Hill, NC 27514.
- Received April 4, 1997.
- Revision received July 7, 1997.
- Accepted July 21, 1997.
- Copyright © 1997 by American Heart Association
Marshall J. Familial incidence of cerebral hemorrhage. Stroke. 1973;4:38-41.
Khaw KT, Barrett-Connor CE. Familial history of stroke as an independent predictor of ischemic heart disease in men and stroke in women. Am J Epidmiol. 1983;123:59-66.
Kiely DK, Wolf PA, Cupples LA, Beiser AS, Myers RH. Familial aggregation of stroke: the Framingham Study. Stroke. 1993;24:1366-1371.
Herman B, Schmits PIM, Leyten ACM, van Lujik JH, Frenken CWGM, Op de Coul AAW, Schulte BPM. Multivariate logistic analysis of risk factors for stroke in Tiburg, The Netherlands. Am J Epidemol. 1983;118:514-525.
Boysen G, Jorgen N, Appleyard M, Sorensen PS, Boas J, Somnier F, Jensen G, Schnor P. Stroke incidence and risk factors for stroke in Copenhagen, Denmark. Stroke. 1988;19:1345-1353.
Brass LM, Shaker LA. Family history in patients with transient ischemic attacks. Stroke. 1991;22:837-841.
Higgins M, Province M, Heiss G, Eckfeldt J, Ellison RC, Folsom AR, Rao DC, Sprafka M, Williams R. NHLBI Family Heart Study: Objectives and design. Am J Epidemiol. 1996;143:1219-1228.
ARIC Investigators. The Atherosclerosis Risk in Communities (ARIC) Study: design and objectives. Am J Epidemiol. 1989;129:687-702.
Kannel WB, Feinleib M, McNamara PM, Garrison RJ, Castelli WP. An investigation of coronary heart disease in families: the Framingham Offspring Study. Am J Epidemiol. 1979;110:281-290.
Hunt SC, Williams RR. Genetic factors in human hypertension. In: Swales JD, ed. Textbook of Hypertension. Oxford, England: Blackwell Scientific Publications; 1994:519-538.
Kornegay C, Liao D, Bensen J, Province M, Folsom A, Ellison CR. The accuracy of proband-reported family history of stroke: the FHS Study. Am J Epidemiol. 1997;145:S82. Abstract.
O’Mahony PG, Dobson R, Rodgers H, James OF, Thomson RG. Validation of a population screening questionnaire to assess prevalence of stroke. Stroke. 1995;26:1334-1337.