METHODS

The rationale and design of EUROSTROKE have been described in detail elsewhere.¹⁵ In short, EUROSTROKE is a collaborative study among European research centres to investigate (1) the variation in incidence of fatal and non-fatal ischaemic and haemorrhagic stroke among populations in different European countries; (2) whether the observed differences in stroke incidence across countries can be explained by differences in prevalence of established cardiovascular risk factors; (3) the relative importance of smoking and some selected dietary factors (potassium intake, alcohol consumption), haemostatic disturbances (fibrinogen) and comorbidity (rheumatic heart disease, atrial fibrillation) compared with established risk factors as determinants of the occurrence of ischaemic and haemorrhagic stroke. The EUROSTROKE database is drawn from ongoing European population-based prospective follow up studies (cohorts) and is designed as a case-control study nested within these ongoing studies. For each stroke case, two controls were sampled. Controls were matched on day of baseline examination only. Apart from its objectives, the EUROSTROKE database allows for aetiological analyses looking into various risk factors for stroke. EUROSTROKE formally started on 1 January 1994. At present, data from four cohorts were available for analysis.

EUROSTROKE case review board

Stroke was defined as a clinical event of rapid onset consisting of neurological deficit lasting more than 24 hours unless death supervenes, or if is lasts less than 24 hours, an appropriate lesion to explain the deficit is seen in a brain image. The event could not be directly caused by trauma to the brain, tumour, or infection. This definition included subjects presenting with signs and symptoms suggestive of a subarachnoid haemorrhage, an intracerebral haemorrhage, ischaemic cerebral infarction, and subjects with a transient ischaemic attack, provided neuroimaging has been performed. A case reviewing panel consisting of four neurologists classified stroke events occurring in EUROSTROKE based on the information that was obtainable such as patients history, results from a lumbar puncture, computed tomography or magnetic resonance images, findings from a necropsy report. Final codes from existing and operating registries, such as, for example, the FINMONICA registry were not reviewed separately, but used as coded. Based on the information present, the neurologist classified the event into first and recurrent stroke, and into subarachnoid haemorrhage, intracranial haemorrhage, intracerebral infarction, or unspecified stroke. In addition, the certainty of the diagnosis was assessed in definite, probable, possible, and no stroke. Events were classified by two neurologists; in case of disagreement a third arbitrated.

This analysis is restricted to definite and probable events. An incident stroke was considered to have occurred when (1) the event had led to a hospitalisation and the hospital discharge record indicated a diagnosis of a new stroke. The clinical diagnosis was based on signs and symptoms, and neuroimaging investigations during hospital stay (definite stroke); or (2) in case of no hospitalisation, signs and symptoms associated with the event obtained from the general practitioner's records were highly suggestive of a stroke according to the neurologists (probable stroke) or (3) in case of out hospital death, when the general practitioner reported that the cause of death was a cerebrovascular accident and a cardiac cause was judged by the general practitioner to be highly unlikely (probable stroke).

Finland

The Finnish contribution to EUROSTROKE comes from the Kuopio Ischemic Heart Disease Risk Factor study, which is a population-based prospective cohort study comprised of an age stratified random sample of 2682 men aged 42, 48, 54 and 60 years. The baseline examination was performed between 1984 and 1989.¹⁶ Fatal and non-fatal stroke cases were collected through the national mortality statistics and the FINMONICA stroke registries. Stroke was defined according to FINMONICA criteria and definitions.¹⁷ Case ascertainment from the baseline examination to 1 January 1993 revealed 74 stroke cases. Controls subjects (n=148) were randomly drawn from the cohort that remained free from stroke during follow up.

The Netherlands

The Dutch contribution to EUROSTROKE comes from the Rotterdam Study, which is a population-based prospective follow up study among 7983 subjects, aged 55 years or over, living in the suburb of Ommoord in Rotterdam, the Netherlands.¹⁸ Baseline data were collected from March 1990 to July 1993. In the Rotterdam Study, information on incident fatal and non-fatal events is obtained from the general practitioners (GPs) working in the study district of Ommoord as described earlier.¹⁹ In short, the GPs involved report all possible cases of stroke to the Rotterdam research centre. Events are presented in coded information following the International Classification of Primary Care (ICPC).²⁰ With respect to the vital status of the participants, information is obtained at regular intervals from the municipal authorities in Rotterdam and also death of a participant is reported. When an event or death has been reported, additional information is obtained by interviewing the GP and scrutinising information from hospital discharge records in case of admittance or referral. All suspected cerebrovascular events reported by the GPs were submitted for review to the EUROSTROKE case review board. From baseline to December 1994, 192 stroke cases were identified and submitted for review and a total of 384 control subjects were drawn from the remainder of the cohort that remained free from stroke during follow up. Altogether 157 events were classified as definite or probable strokes.

Russia

The Russian contribution to EUROSTROKE originates from studies performed in the Octyabrsky, the Kirovsky and Leninsky districts of Novosibirsk, Siberia. The Novosibirsk cohort is based on three population-based surveys, which were conducted between 1984 and 1989 as part of the WHO MONICA project.²¹ The Novosibirsk cohort comprises 9006 men and women aged 25 to 64 years. Stroke cases were collected through a specifically developed stroke registry, aiming to identify fatal and non-fatal hospitalised and non-hospitalised stroke patients.^22,23 Stroke events were defined according to MONICA criteria and definitions.²⁴ From baseline to December 1995, a total of 100 stroke cases had been identified and 200 control subjects were drawn from the database. Finally, 79 subjects proved to be true strokes.

United Kingdom

The British contribution to EUROSTROKE comes from the Caerphilly Heart Disease study in Wales, United Kingdom, in which 2512 men, aged 45 to 59 years are participating.²⁵ Baseline examinations took place from 1979 to 1983. Follow up examinations were performed from 1984 to 1988 (phase II) and from 1989 to 1993 (phase III). Stroke events were registered through national mortality statistics, hospital discharge records, self report and family report. Of the registered events, additional information on signs and symptoms, on neuroimaging, necropsy and a copy of the discharge records were collected. When complete, stroke cases were submitted for review to the EUROSTROKE case review board as descried earlier. Recently, 100 stroke events had been submitted for review and 200 controls subjects had been drawn from the remaining cohort. Eighty four stroke cases were classified as definite/probable stroke by the EUROSTROKE case review board.

Cardiovascular risk factors

Because EUROSTROKE is based on 10 ongoing cohort studies, information on cardiovascular risk factors in each of the participating centres was already collected before the formal start of the EUROSTROKE project. Whenever possible, an exhaustive attempt was made to further harmonise the collected information to make comparison across studies possible. Nevertheless, any baseline measurements could not be further standardised beyond the attempts done in each individual study.

In each of the centres information on smoking, alcohol consumption, and medical history was obtained by questionnaire. The subject's smoking behaviour was categorised into current, former or never. Alcohol consumption was categorised into current drinkers and non-current drinkers (former and never). In addition, an estimate of grams of alcohol per day was obtained. Presence of diabetes mellitus was generally based on the question “Do you suffer from diabetes mellitus?”. In the Rotterdam study diabetes mellitus was considered present when subjects used blood sugar lowering drugs, whereas in the Novosibirsk study presence of diabetes mellitus was not evaluated for the entire cohort. Information on a history of stroke was obtained by direct questioning at baseline “Did you ever suffer from a stroke?”. A similar approach was taken for myocardial infarction. Presence of angina pectoris was based on either the cardiovascular Rose questionnaire or direct questioning.

In general, systolic and diastolic blood pressure were measured twice at one occasion in sitting position. In Cardiff, only one blood pressure measurement was performed. Height and weight were measured and body mass index (kg/m²) was calculated. In all four centres an electrocardiogram was made. In Cardiff, Kuopio and Novosibirsk, LVH was assessed according to the Minnesota classification system (codes 3.1, 3.3, 3.4).²⁶ In Rotterdam, an automated diagnostic classification system of the Modular Electrocardiogram Analysis System (MEANS) was used in which LVH diagnosis was assessed on the parameters voltage, shape and repolarisation, as detailed elsewhere.^27,28 Apart from Rotterdam, a fasting blood sample was taken for determination of serum lipids (total cholesterol, HDL cholesterol). For the present analysis hypertension was defined as a systolic blood pressure of 160 mm Hg or over, or a diastolic blood pressure of 95 mm Hg or over, or current use of blood pressure lowering drugs.

Data analysis

Complete data on both stroke and LVH were available for Cardiff (84 cases/200 controls), Kuopio (60 cases/ 116 controls), Rotterdam (114 cases/334 controls), and Novosibirsk (62 cases/168 controls).

The analyses were first performed for each centre separately. To identify potential confounding variables, cardiovascular risk factors that were related to stroke were examined using logistic regression models and to LVH using logistic regression models. Factors that were related to both stroke and LVH (p<0.10), and that were not assumed to be in the causal pathway of LVH leading to stroke were considered confounding variables. Systolic blood pressure, diastolic blood pressure, body mass index, current smoking, and diabetes mellitus were evaluated as potential confounders. LVH as predictor of stroke was studied as a dichotomeous variable using logistic regression analysis. Initially adjustments were made for age and sex, and subsequently for potential confounding cardiovascular risk factors.

An interaction term (LVH × centre) was used to study whether the association between LVH and stroke differed across the four centres (heterogeneity). Interaction term (LVH × sex) was used to evaluate whether the association differed between men and women. Similar analyses were performed in strata of smoking and hypertension. Results are presented as odds ratios, with their corresponding 95% confidence intervals (95% CI). Separate analyses were performed for total stroke, first ever stroke, fatal stroke (death within 28 days after onset), non-fatal stroke, and for haemorrhagic stroke and cerebral infarction.

In the analysis in which a history of diabetes mellitus was used, analyses were initially performed excluding data from Novosibirsk, because in this centre such data were not available. Additional analyses were performed in which history of diabetes was imputed for the Novosibirsk centre. The imputation was based on the prevalence observed in Cardiff and Kuopio combined, an estimate comparable to what has been reported for Novosibirsk area.¹³ A dummy variable (yes/no) for the imputation was added to the model. As the magnitude of findings did not differ between the two approaches only the latter is presented. The analyses have been performed using STATA version 4.0.

RESULTS

General characteristics of the study populations are given in table 1. Characteristics of the stroke events are presented in table 2. Of all strokes, 73% had been hospitalised. Neuroimaging (CT/MRI) had been performed in 55% of the cases. Among the fatal strokes, 26.2% were a haemorrhagic, 33.8% a cerebral infarctions and 40.0% could not be specified with the available information. The corresponding figures for non-fatal stroke were 9.6%, 80.4% and 10.0%, respectively. In analyses in which information from all centres was combined, age, male sex, systolic blood pressure, diastolic blood pressure, hypertension, current smoking, and history of diabetes mellitus were positively, and significantly (p<0.05) related to stroke. Significant associations with LVH were found for age, systolic blood pressure, diastolic blood pressure and hypertension. Interaction terms indicated no difference in associations between centres (p=0.49) and between men and women (p=0.92).

The association of LVH to stroke for each centre is given in figure 1. In all centres LVH was positively related to the risk of stroke. The centre specific analyses did not always reach the level of statistical significance. The wide confidence interval reflects the limited number of subjects with LVH in each centre. There was no evidence for heterogeneity.

Table 3 describes the association between LVH and risk of stroke for all centres combined. LVH was associated with a twofold increased risk of stroke (odds ratio 2.13 (95% CI 1.29 to 3.50). Results were similar for first ever stroke. The increased risk seemed to be stronger for fatal stroke than for non-fatal stroke (table 3). When differences in systolic blood pressure and body mass index were taken into account, the magnitude of the associations between LVH and all types of stroke was attenuated. Additional adjustment for hypertension, smoking and diabetes mellitus did not materially affect the magnitude, direction and significance of the findings (table 3).

Subjects with LVH had a non-significant increased risk of cerebral infarction compared with those without LVH (age and sex adjusted odds ratio 1.68 (95% CI 0.92 to 3.04), p=0.09). The association was more pronounced for fatal than for non-fatal cerebral infarction: odds ratios 3.66 (95% CI 1.19 to 11.2) and 1.44 (95% CI 0.75 to 2.78), respectively. No significant association was found with haemorrhagic stroke (age and sex adjusted odds ratio of 1.52 (95% CI 0.45 to 5.18), p=0.50) or with fatal haemorrhagic stroke (odds ratio 1.03 (95% CI 0.13 to 7.90)) or non-fatal haemorrhagic stroke (odds ratio 2.03 (95% CI 0.45 to 9.06)).

Stratified analyses among smokers and non-smokers showed different tendencies in magnitude in the associations of LVH with stroke (table 4). In smokers, LVH was associated with a 3.47-fold (95% CI 1.49 to 8.09) increased risk of stroke, whereas in non-smokers the risk was 1.61 (95% CI 0.84 to 3.12). The interaction term, however, did not reach statistical significance (p=0.15).No differences in magnitude in association of LVH to stroke was found for men and women nor for hypertension and normotension.

DISCUSSION

This analysis of the EUROSTROKE project indicates that LVH assessed by electrocardiogram is a predictor of stroke. The association seemed to be stronger for fatal stroke than for non-fatal stroke and may be restricted to smokers. Furthermore, no major differences were seen in the magnitude of the association with haemorrhagic stroke and cerebral infarction, if anything the relation with cerebral infarction was more pronounced.

Several aspects of this study should be considered. Firstly, we cannot be sure about the completeness of case ascertainment in each study. The proportion of fatal cases across the centres ranged between 19.3% in Kuopio and 30.6% in Novosibirsk (table 2). Although it is likely that there are true differences in case-fatality of stroke across these countries this may indicate underrepresentation of non-fatal cases. If the selection of non-fatal cases is associated with a lower prevalence of the exposure of interest (LVH), it may bias a true positive association towards a reduced odds ratio or a null finding. We have no means to evaluate whether this indeed did take place. Secondly, the association with fatal stroke may be biased towards a positive finding when in the assessment a large proportion of the stroke cases was misclassified—that is, had died from coronary heart disease rather than from stroke. Given the diagnostic procedures in the participating centres this is unlikely to have happened. Thirdly, it has been well established that the sensitivity of the ECG to detect LVH is limited. Data from the Framingham Heart Study indicated a sensitivity of 6.9% and a specificity of 98.8% compared with echocardiography as gold standard.²⁹ The ultimate consequence of this notion is that most probably the reported associations under study are an underestimation of the “true” association; as misclassification occurred in both cases and controls, most probably in a random fashion. Alternatively, a recent paper indicated that ECG-LVH and LVH assessed by echocardiography carry to some extent different prognostic information.³⁰ Finally, we acknowledge that the diagnoses of LVH has been made using different approaches. We used whatever approach had been taken in the individual cohort. Some may have a stronger prognostic implication.³⁰ However, in all cohorts, ECG-LVH was related to risk of stroke without showing differences in associations across the cohorts (fig 1). Strengths of this study are the fairly large number of well diagnosed and classified events for both men and women, presence of data on a large number of potential confounders, a wide age range, and risk factors assessment before the occurrence of the stroke event.

Several studies have evaluated the association between LVH and stroke (table 5). Direct comparison between studies is hampered by differences in population distribution, in selection of population, in methods of assessing LVH, in stroke distribution (fatal/non-fatal, type of stroke) and in adjustment for several potential confounding variables. In general, an increased stroke risk was found among subjects with LVH. In reports of the Framingham Heart study, with different samples from the populations and duration of follow up, LVH was associated with a sixfold increased risk of stroke among middle aged men and women ⁷ and with a 2.3-fold increased risk in the elderly.⁸ Results from Dunn and coworkers in the United Kingdom showed that among hypertensives the relative contribution of LVH to death from stroke decreased with age.³ In the Copenhagen City Heart Study in Denmark weaker than in the Framingham Heart Study, associations were found for men and no relation was seen for women (table 5).⁹ Few studies, however, have distinguished between fatal and non-fatal stroke. In a community-based cohort study among 459 subjects aged 75 to 85 years, who were followed up for 10 years, baseline LVH was associated with a threefold increased risk of fatal stroke (p=0.07), whereas a non-significant increased risk of non-fatal stroke was found (odds ratio 1.1).¹¹ The analyses were, however, based on a limited number of fatal (n=15) and non-fatal (n=20) events. The present EUROSTROKE analysis confirms the increased risk of stroke associated with LVH. In addition, our findings point towards a stronger association of LVH with fatal stroke than with non-fatal stroke, a tendency observed for cerebral infarction only. In this analysis, the increased risk was more pronounced in smokers than in non-smokers. As none of the studies reported such a finding, future studies are needed to confirm this potential effect modification of smoking in stroke risk. It has not been clear by which mechanism smoking increases the risk of stroke. This new finding that LVH may be a condition required for stroke to occur in smokers may suggest that the underlying mechanism is related to either arrythmias^31–33 or other cardiac events.

Several hypotheses have been given to explain why LVH predisposes for coronary heart disease and death.⁶ Yet, the mechanisms for stroke may be less clear. Firstly, LVH may be a marker for presence of longstanding increased blood pressure and as such it may reflect cardiovascular end-organ damage, which in itself is associated with increased risk of stroke. Our finding that when differences in blood pressure were taken into account the association between LVH and stroke was reduced, favours this view. Secondly, LVH has been implicated as an important factor in triggering cardiac arrhythmias,^31,32 which may increase the risk of stroke, in particular ischaemic stroke.

Increased blood pressure has consistently been reported as a major determinant of LVH.⁷ The increased stroke risk of LVH among smokers further supports the large potential for prevention of stroke by lowering of blood pressure and discouraging starting and encouraging quitting smoking.

In conclusion, this analysis of the EUROSTROKE project indicates that LVH assessed by electrocardiogram is a powerful predictor of stroke. The association seems to be stronger for fatal stroke than for non-fatal stroke and is more pronounced in smokers.