Heritability of Ischemic Stroke in Relation to Age, Vascular Risk Factors, and Subtypes of Incident Stroke in Population-Based Studies
Background— Appropriate design of molecular genetic studies of ischemic stroke requires an understanding of the genetic epidemiology of stroke. However, there are no published population-based data on heritability of aetiological subtypes of ischemic stroke, confounding by heritability of other vascular risk factors, or the relationship between heritability and age of onset.
Methods— We studied family history of stroke (FHxStroke) and of myocardial infarction (FHxMI) in first-degree relatives in 2 population-based studies (Oxford Vascular Study [OXVASC]; Oxfordshire Community Stroke Project [OCSP]). We related FHxStroke and FHxMI to subtype of ischemic stroke, age, and the presence of vascular risk factors and performed a systematic review of all studies of FHxStroke by stroke subtype.
Results— In our population-based studies and in 3 hospital-based studies, FHxStroke was least frequent in cardioembolic stroke (OR=0.74, 95%CI=0.58 to 0.95, P=0.02) but was equally frequent in the other subtypes. In OXVASC and OCSP, FHxStroke (P=0.02), FHxMI (P=0.04), and FHx of either (P=0.006) were associated with stroke at a younger age. Only FHxStroke was associated with previous hypertension (OR=1.59, 95%CI=1.08 to 2.35, P=0.02). FHxMI was more frequent in large-artery stroke (OR=1.63, 95%CI=0.99 to 2.69, P=0.05).
Conclusion— Consistent results in our population-based studies and previous hospital-based studies suggest that inclusion bias is not a major problem for studies of the genetic epidemiology of stroke. Molecular genetic studies might be best targeted at non-cardioembolic stroke and younger patients. However, genetic susceptibility to hypertension may account for a significant proportion of the heritability of ischemic stroke.
Susceptibility to ischemic stroke may be influenced by genetic factors. Some Mendelian disorders have been identified,1 and animal models also suggest that susceptibility to stroke may be genetically determined.2,3 However, human candidate gene studies in apparently sporadic stroke have so far been either inconsistent or negative.3 To target molecular genetic studies appropriately, it is first necessary to understand the basic genetic aetiology of ischemic stroke. Although there have been several epidemiological studies of the heritability of ischemic stroke, results have also been inconsistent.3,4 A recent systematic review of such studies found that although there probably is a genetic contribution to stroke, publication bias and heterogeneity between the studies did not allow reliable interpretation of results.4 Most studies failed to differentiate between ischemic stroke subtypes, and many studies combined ischemic and hemorrhagic stroke.4
It is likely that genetic susceptibility to the pathological mechanisms underlying ischemic stroke differs between the subtypes. However, the few family history (FHx) studies that have differentiated between subtypes were insufficiently powered and were all hospital-based.5–8 There are no published population-based studies of the heritability of aetiological subtypes of ischemic stroke. Population-based studies are worthwhile because hospital-based studies may be subject to inclusion bias.9 Bias might occur in genetic epidemiological studies if, for example, hospital admission or extent of investigation were dependent on age or severity of stroke, both of which might be related to heritability. We therefore studied FHx of stroke (FHxStroke) and FHx of myocardial infarction (FHxMI) in pooled data from 2 population-based studies of incident ischemic stroke (Oxfordshire Community Stroke Project [OCSP];10 Oxford Vascular Study [OXVASC]11).To avoid recall bias associated with case-control comparisons, we confined our study to case–case comparisons. Given that some of the risk factors for ischemic stroke, such as hypertension or hyperlipidemia, are partly genetically determined,12 we also compared the frequency of vascular risk factors in patients with and without FHxStroke. We performed the same analyses for FHxMI to determine whether stroke subtype or risk factor associations were specific to FHxStroke. Finally, to identify any possible bias in previous hospital-based studies and to summarize all currently available data, we conducted a systematic review of all studies of FHxStroke in stroke subtypes.
See Editorial Comment, page 824
Materials and Methods
We studied FHxStroke and FHxMI in first-degree relatives in 2 population-based stroke incidence studies that conformed to the standard quality criteria for such studies.13 The methods and results of the OCSP have been published previously.10 The OXVASC study started in April 2002 and used identical ascertainment methods to the OCSP.9–11 Briefly, by close collaboration with family practitioners (FPs) (50 FPs in OCSP, 63 FPs in OXVASC), an urban and rural population (105 000 people in OCSP, 91 000 in OXVASC) was studied. Patients with a mild stroke were seen at a daily study clinic, allowing more rapid assessment and treatment than currently the norm in the British health care system. This provided an incentive for FPs to report all patients who might have had a transient ischemic attack or stroke during the study periods. However, we double-checked FP databases and records to identify any strokes that might have been missed. Stroke patients requiring admission to hospital were identified by daily assessment of admission registers and visits to the relevant wards of hospitals within the study catchment. We also reviewed hospital discharge coding, referrals for brain and vascular imaging, and all death certificates and coroner reports relating to our study population. Both studies were approved by the local ethics committee.
In both OXVASC and OCSP, a study neurologist assessed all cases as soon as possible after notification, and CT brain imaging was obtained. Details of the presenting event, clinical characteristics, medical history, and FHx were recorded from the patient, FP records, and/or hospital records. FHx data were obtained separately for stroke and for MI, and FHx was regarded as positive if at least 1 first-degree relative was affected.
In OXVASC, patients routinely undergo Doppler scanning of the cervical arteries and echocardiography. Stroke aetiology is classified prospectively according to the Trial of Org 10172 in Acute Stroke Treatment (TOAST) criteria.14,15 In the OCSP, the subtype of ischemic stroke had been categorized according to the Bamford classification,16 but the investigators had also originally prospectively categorized stroke according to cause, and detailed clinical and imaging data had been collected. It has been shown that the TOAST classification can be applied retrospectively, and that this is accurate and reproducible.15 As reported previously, the details available in the OCSP allowed us to reclassify all ischemic strokes according to the same causative categories as used in the TOAST study.9
Because the OCSP and OXVASC were conducted in the same population in collaboration with the same family practices using very similar methods, we pooled the data to increase statistical power. However, we allowed for possible differences between the studies by adjusting analyses by “study” when appropriate. We studied FHxStroke and FHxMI in relation to stroke subtype and in relation to age, sex, cholesterol level, and history of previous transient ischemic attack. A history of hypertension, hypercholesterolemia, and diabetes mellitus was also recorded and regarded as positive if confirmed by the patient or medical notes, or if the patient was using treatment.
We expressed the prevalence of FHxStroke or of FHxMI as simple proportions and compared these between stroke subtypes by χ2 test. To test for independent associations between FHx and stroke subtypes, we performed a logistic regression analysis, adjusting for age, sex, study, and other vascular risk factors. To study differences in baseline characteristics between patients with and without a FHx, we used the χ2 test and analysis of variance as appropriate. For any factor that was associated with FHx overall or with a particular stroke subtype, we also performed a logistic regression analysis, adjusting for age, sex, and study.
We identified previous published reports of FHxStroke as reported elsewhere.4 Briefly, 2 independent observers searched Medline+, Embase, and the reference lists of all articles that met the inclusion criteria. Studies were included in the current review if they reported FHx by subtype of ischemic stroke according to TOAST criteria14,15 or a similar classification. We calculated the odds for a positive FHx in specific stroke subtypes compared with the remainder within individual studies and, when appropriate, performed fixed-effects meta-analysis according to the Mantel-Haenszel method. SPSS for Windows version 10.0 (SPSS 1999) was used for all statistical analyses.
The OCSP registered 675 patients with a first-ever stroke and 545 patients with a first-ever ischemic stroke; 56 patients were excluded from the analysis because no details were available on FHx, stroke subtype, or risk factors. In the first year of OXVASC (April 2002 to March 31, 2003), 116 patients were registered with a first-ever ischemic stroke, of whom FHx data were available in 107. Of the 596 patients in the combined “Oxfordshire” cohort, 137 (23.0%) had FHxStroke, 137 (23.0%) had FHxMI, 40 (6.7%) had both, and 234 (39.3%) had FHx of either stroke or myocardial infarction. The frequency of FHxStroke or of FHxMI did not differ between hospitalized and non-hospitalized patients (Table 1). Brain imaging or postmortem data was available in 89% of patients (88% in OCSP and 96% in OXVASC).
The distribution of stroke subtypes did not differ between patients with and without FHxStroke or FHxMI (Table 1, Figure 1), and there were no significant associations between stroke subtypes and FHx after adjusting for potential confounders (Table 2). However, FHxStroke tended to be least frequent in patients with cardioembolic stroke (Figure 2).
FHxStroke was associated with a young age of onset (Figure 3), with significant heterogeneity in the frequency of FHx across 10-year age bands (P=0.01) and highest rates in patients aged 60 years or younger (OR=1.73, 95% CI=1.02 to 2.91). The trend toward a higher frequency of FHxStroke in patients younger than age 60 was present for each stroke subtype: large vessel, OR=2.57 (95% CI=0.84 to 7.88), P=0.09; small vessel, 1.43 (0.50 to 4.09), P=0.34; cardioembolic, 2.17 (0.38 to 12.6), P=0.33; and undetermined, 2.51 (1.00 to 6.26), P=0.04. A similar trend toward increasing FHx in younger patients was also present for FHxMI (P=0.04 for trend across 10-year age bands) and for FHxStroke or FHxMI (P=0.001, Figure 3).
The prevalence of vascular risk factors and their association with FHxStroke or FHxMI is shown in Table 1. FHxStroke was associated with a history of hypertension before (OR=1.59, 95% CI=1.08 to 2.35, P=0.02) and after adjusting for age, sex, study, and stroke subtype (OR=1.52, 95% CI=1.02 to 2.26, P=0.04). There was a borderline-significant (P=0.05) trend for FHxMI to be associated with large-vessel stroke (OR=1.63, 95% CI=0.99 to 2.69), particularly in patients with FHxMI in 2 first-degree relatives (OR=1.79, 95% CI=0.85 to 3.77, P=0.09). This trend was present in comparison to all other subtypes: large-vessel stroke versus small-vessel stroke (OR=1.52, 95% CI=0.82 to 2.79, P=0.18), large-vessel stroke versus cardioembolic stroke (OR=1.70, 95% CI=0.91 to 3.16, P=0.09), large-vessel stroke versus stroke of undetermined aetiology (OR=1.61, 95% CI=0.92 to 2.81, P=0.09). These associations were partly accounted for by the association between large-vessel stroke and history of hypercholesterolemia (OR=1.87, 95% CI=1.15 to 3.04, P=0.01).
We identified 4 previous studies that provided data on the prevalence of FHxStroke in subtypes of ischemic stroke according to the TOAST classification, all of which were hospital-based.5–8 Only 1 study also collected details on FHxMI.5 Three studies5–7 defined FHx as at least 1 first-degree relative affected, and 1 study8 did not provide a clear definition of FHx. Two studies were case-control studies,5,6 one of which reported sufficient data to allow re-analysis as a case–case comparison.6 We contacted the authors to obtain the required details for the other study.5 The 2 other studies were cross-sectional studies, in which FHxStroke was one of several factors that were related to stroke subtype.7,8 We excluded 1 study8 because of its very selected population (young Taiwanese stroke patients in a tertiary referral center). In this study, large-vessel stroke was associated with FHxStroke (OR=3.10, 95% CI=1.18 to 8.14), but no other associations were found. Both case–control studies found that in comparison to healthy controls, FHxStroke was more common in large-vessel and small-vessel strokes.5,6 In addition, one study reported that a FHxMI was associated with large vessel strokes.5
In the meta-analysis of the Oxfordshire cohort and the 3 published hospital-based studies, FHxStroke was consistently less frequent in cardioembolic stroke overall (OR=0.74, 95% CI=0.58 to 0.95, P=0.02; Figure 1) and in comparison with each of the other subtypes individually (Figure 2). The prevalence of FHxStroke did not differ between the other stroke subtypes: large-vessel stroke versus lacunar stroke (OR=1.11, 95% CI=0.86 to 1.42, P=0.42), large-vessel stroke versus stroke of undetermined cause (OR=1.17, 95% CI=0.93 to 1.49, P=0.19), small-vessel stroke versus stroke of undetermined cause (OR=1.06, 95% CI=0.83 to 1.35, P=0.64). We did not include strokes of “other determined aetiology” in the meta-analysis because of their heterogeneous aetiology and, in some cases, proven genetic background.
In this study, FHxStroke was consistently less common in cardioembolic stroke than in other subtypes in population-based and in hospital-based cohorts. FHxStroke was equally frequent in the other subtypes of ischemic stroke, but FHxMI was associated with large-artery stroke. In the Oxfordshire cohorts, there were positive associations between FHxStroke and previous hypertension and between young age at stroke onset and both FHxStroke and FHxMI.
One advantage of our study methodology was that by performing case–case comparisons of patients with different stroke subtypes, we are likely to have avoided the recall bias that can undermine case–control studies, with stroke cases being more aware of any FHxStroke than controls. A further advantage of the Oxfordshire studies was that they were population-based, ie, all ages and all degrees of severity of stroke were included. However, our study also had several potential shortcomings. First, we classified stroke subtypes according to the TOAST criteria,14,15 because it is currently the most widely used aetiological classification of stroke. However, it is still relatively crude, with some categories probably comprising several underlying disorders with differing genetic influences. Second, the OCSP was conducted in the early 1980s. The standard of investigations has improved since then, and techniques such as MRI scanning or transesophageal echocardiography are now much more readily available. More detailed investigations might have allowed more accurate subtyping. However, the associations between stroke subtype and FHxStroke were consistent with more recent studies.5–8,11 Finally, we used FHx as a measure of heritability. However, familial clustering of a disease may be due not only to genetic factors, but also to a shared environment.
In the Oxfordshire studies, the frequency of FHxStroke did not differ between subtypes of stroke, although there was a trend toward lower rates in cardioembolic stroke. This was consistent with similar trends in the 3 previous hospital-based studies, was statistically significant when the results of all studies were combined (Figure 1), and was present in comparison with each of the individual subtypes (Figure 2). These results are consistent with the findings of previous case–control studies,5,6 which found that patients with large-vessel disease and patients with small-vessel disease, but not patients with cardioembolic stroke, were more likely to have a FHxStroke than healthy controls. The relatively low heritability of cardioembolic stroke may be explained by the fact that the underlying cardiac disorders, eg, atrial fibrillation or valvular disease, are not highly hereditable and do not invariably cause stroke.
We found a borderline-significant positive association between FHxMI and large-vessel stroke. This was also found by the only other study of FHxMI.5 Large-vessel disease and ischemic heart disease reflect similar pathological processes, and the association with FHxMI may indicate an inherited tendency for atherosclerotic disease to develop. We also found a positive association between a history of hypertension and FHxStroke, which was not present for FHxMI. This is consistent with previous general population studies that have shown that FHxStroke is more common in hypertensive than normotensive subjects17,18 and a previous case–control study in which stroke patients were more likely to have a FHx of hypertension than controls.19 Given that hypertension has a major genetic component,12 heritability of stroke may partly be conferred by an inherited tendency for hypertension.
There are few published data on the relationship between age of stroke onset and FHx of vascular disease, and no study has differentiated between the subtypes of ischemic stroke.4 Studies used different age cut-offs or were restricted to relatively young patients, and results have been conflicting.4 We found significantly higher rates of FHxStroke, FHxMI, and FHx of either stroke or myocardial infarction in younger patients overall and similar trends within each cause subtype. However, given the conflicting trends in the published data, more studies are required to determine the relationship between age and heritability of stroke and to analyze the degree to which these results are influenced by potentially better recall in younger patients versus older patients and the consequent increase in likelihood of stroke of parents and siblings of older patients.
Consistent results in our population-based studies and previous hospital-based studies suggest that inclusion bias is not a major problem for studies of the genetic cause of stroke. Molecular genetic studies might be best targeted at non-cardioembolic stroke and younger patients. However, genetic susceptibility to hypertension may account for a significant proportion of the heritability of ischemic stroke.
Dr Rothwell and Dr Flossmann are funded by the UK Medical Research Council (MRC). Dr Schulz is funded by the Wellcome Trust. We thank Professor Charles Warlow and the OCSP Collaborators for allowing us access to their data, the OXVASC Collaborators for their help with this study, and Hugh S. Markus for providing unpublished data.5
- Received September 23, 2003.
- Revision received December 3, 2003.
- Accepted January 5, 2004.
Okamoto K, Yamori Y, Nagakoa A. Establishment of the stroke-prone spontaneously hypertensive rat (SHR). Circ Res. 1974; 33/34: 1-143–1-153.
Hassan A, Markus HS. Genetics and ischaemic stroke. Brain. 2000; 123: 1784–1812.
Flossmann E, Schulz UGR, Rothwell PM. A systematic review of the methods and results of studies of the genetic epidemiology of ischaemic stroke. Stroke. 2004; 35: 212–227.
Jerrard-Dunne P, Cloud G, Hassan A, Markus HS. Evaluating the Genetic Component of Ischemic Stroke Subtypes. Stroke. 2003; 34: 1364–1369.
Meschia JF, Brown RD, Brott TG, Hardy J, Atkinson EJ, O’Brien PC. Feasibility of an Affected Sibling Pair Study in Ischemic Stroke: Results of a 2-Center Family History Registry. Stroke. 2001; 32: 2939–2941.
Lee T, Hsu W, Chen C, Chen S. Etiologic Study of Young Ischemic Stroke in Taiwan. Stroke. 2002; 33: 1950–1955.
Schulz UG, Rothwell PM. Differences in vascular risk factors between etiological subtypes of ischemic stroke: importance of population-based studies. Stroke. 2003; 34: 2050–2059.
Bamford J, Sandercock P, Dennis M, Warlow C, Jones L, McPherson K, Vessey M, Fowler G, Molyneux A, Hughes T, Burn J, Wade D. A prospective study of acute cerebrovascular disease in the community: the Oxfordshire Community Stroke Project 1981–86. 1. Methodology, demography and incident cases of first-ever stroke. J Neurol Neurosurg Psych. 1988; 51: 1373–1380.
Coull AJ, Silver L, Rothwell PM. Implications of rates of non-fatal acute cerebrovascular events versus acute coronary events for provision of acute clinical services: Oxford Vascular Study (OXVASC). Cerebrovasc Dis. 2003; 16: 1–2.
Hunt SC, Hasstedt SJ, Kuida H, Stults BM, Hopkins PN, Williams RR. Genetic hereditability and common environmental components of resting and stressed blood pressure, lipids and body mass index in Utah pedigrees and twins. Am J Epidemiol. 1989; 129: 625–638.
Sudlow CL, Warlow CP. Comparing stroke incidence worldwide: what makes studies comparable? Stroke. 1996; 27: 550–558.
Adams HP, Bendixen BH, Kappelle LJ, Biller J, Love BB, Gordon DL, Marsh EE and the TOAST Investigators. Classification of Subtype of Acute Ischaemic Stroke: Definitions for Use in a Multicenter Clinical Trial. Stroke. 1993; 24: 35–41.
Goldstein LB, Jones MR, Matchar DB, Edwards LJ, Hoff J, Chilukuri V, Armstring SB, Horner RD. Improving the reliability of stroke subgroup classification using the Trial of ORG 10172 in Acute Stroke Treatment (TOAST) criteria. Stroke. 2001; 32: 1091–1097.
Khaw K, Barrett-Connor E. Family history of stroke as an independent predictor of ischaemic heart disease in men and stroke in women. Am J Epidemiol. 1986; 123: 59–66.
Burke GL, Savage PJ, Sprafka M, Selby JV, Jacobs DR, Perkins LL, Roseman JM, Hughes GH, Fabsitz RR. Relation of Risk Factor Levels in Young Adulthood to Parental History of Disease. The CARDIA Study. Circulation. 1991; 84: 1176–1187.
Nicolaou M, DeStefano AL, Gavras I, Cupples LA, Manolis AJ, Baldwin CT, Gavras H, Farrer LA. Genetic Predisposition to Stroke in Relatives of Hypertensives. Stroke. 2000; 31: 487–493.