Risk Factors and 20-Year Stroke Mortality in Men and Women in the Renfrew/Paisley Study in Scotland
Background and Purpose—The aim of this study was to relate risk factors in middle-aged men and women to stroke mortality over a long follow-up period.
Methods—In the early to mid 1970s, 7052 men and 8354 women from the Renfrew/Paisley prospective cohort study in Scotland were screened when aged 45 to 64 years. Risk factors measured included blood pressure, blood cholesterol and glucose, respiratory function, cardiothoracic ratio, smoking habit, height, body mass index, age, preexisting coronary heart disease, and diabetes. These were related to stroke mortality over 20 years of follow-up.
Results—Women’s stroke mortality rates were similar to men’s, unlike coronary heart disease mortality, in which case women’s rates were lower than men’s. Diastolic and systolic blood pressure, smoking, cardiothoracic ratio, preexisting coronary heart disease, and diabetes were positively related to stroke mortality for men and women, while adjusted forced expiratory volume in 1 second and height were negatively related. Cholesterol and body mass index were not related to stroke mortality. Glucose in nondiabetics was positively related to stroke mortality for women but not men, and there was evidence of a threshold effect at the highest levels of glucose. Former smokers had mortality rates that were similar to those of never-smokers. In sex-specific multivariate models, most variables retained a statistically significant association with stroke mortality, illustrating the multifactorial etiology of stroke.
Conclusions—Overall, findings for women were similar to those for men. Control of risk factors for reduction of stroke mortality should be targeted at men and women in a similar fashion, particularly with reference to smoking cessation and blood pressure control.
Stroke is one of the major causes of death in the United Kingdom, but understanding of stroke epidemiology is less complete than that for coronary heart disease (CHD). Although stroke mortality has declined in the West over the past 50 years, it has increased in other parts of the world, particularly eastern Europe.1 In Scotland, 11% of deaths in men and 15% of deaths in women aged 65 years and older were caused by stroke in 1997.2 Scotland also had higher stroke mortality rates than the rest of the United Kingdom.2 In the United States, stroke mortality is the third leading cause of death.3
Major risk factors for stroke include high blood pressure, smoking, cardiac disease, diabetes, and age.3 4 There is less agreement between studies on the role of blood glucose (outside the diabetic range), blood cholesterol, and body mass index as predictors of incident stroke. Respiratory function and body height have emerged as potentially important risk factors in some of the limited number of studies in which they have been investigated in this regard. A weakness of most of the previous prospective epidemiological studies of stroke is that they have been confined to men; as a consequence of this, there is greater uncertainty regarding risk relationships among women.
Here we present the first comprehensive analyses of risk factors for stroke from a large representative prospective cohort study in west Scotland. The cohort contained both men and women who underwent screening in the early to mid 1970s and have been followed up for stroke deaths over a 20-year period.
Subjects and Methods
The Renfrew/Paisley general population study was performed between 1972 and 1976 and involved 7052 men and 8354 women aged 45 to 64 years who were resident in the towns of Renfrew and Paisley, close to the city of Glasgow in Scotland. A response rate of nearly 80% was achieved. Full details of the study methodology have been reported previously.5 6 7 The number of participants in the study differs slightly from numbers given in previous publications (7058 men and 8353 women) because 13 miscodings of sex and 5 people who attended twice were recently discovered. A questionnaire was completed by each participant, and this was checked when the participant attended a clinic where physical measurements were taken. Smoking habit was defined by the following exposure groupings: never-smoker; pipe or cigar smoker only; 1 to 14, 15 to 24, ≥25 cigarettes per day; and ex-smoker. Participants were asked if they were or had ever been diabetic. Cardiovascular symptoms were assessed with the Rose angina questionnaire,8 with angina defined as definite or possible.9 Severe chest pain was defined as ever having a severe pain across the front of the chest lasting half an hour or more.9 Questions were asked on preexisting stroke symptoms and are being analyzed in a separate publication.
At the screening examination, blood pressure was recorded as the mean of 2 measurements taken with the subject in the seated position, with the use of a London School of Hygiene and Tropical Medicine sphygmomanometer.5 Diastolic pressure was recorded at the disappearance of the fifth Korotkoff sound. A nonfasting blood sample was taken for the measurement of plasma cholesterol and glucose.5 Measurements of height and weight were made, enabling body mass index to be calculated as weight (kilograms) divided by height (meters) squared. A 70-mm chest x-ray was taken, and heart and thoracic width were measured. The cardiothoracic ratio was calculated as the ratio of the heart width to the thoracic width.
The forced expiratory volume in 1 second (FEV1) was measured at the screening examination with the use of a Garthur Vitalograph spirometer with the subject standing. The FEV1 score was calculated by obtaining the expected FEV1 from a linear regression equation of age and height for men and women separately, derived from a healthy subset of the population who had never smoked and responded negatively to questions regarding phlegm, breathlessness, wheezy or whistling chest, and whether the weather affected breathing.7 The adjusted FEV1 was defined as the actual FEV1 as a percentage of the expected FEV1.
A 6-lead ECG was made and coded according to the Minnesota system.10 Ischemia on ECG was defined as codes 1.1 to 1.3, 4.1 to 4.4, 5.1 to 5.3, or 7.1. Preexisting CHD was defined as having any of the following: angina, ECG ischemia, or severe chest pain.
Men and women with missing readings for each risk factor were excluded from the analysis for that risk factor only. There were 7 participants with missing systolic blood pressure, 8 with missing diastolic blood pressure, 139 with missing cholesterol, and 24 with missing adjusted FEV1 measurements. Glucose was not measured for some sectors of Paisley and was missing for 4703 people. Cardiothoracic ratio data were missing for 675, height for 13, and body mass index for 15 participants. Risk factors were divided into quintiles for continuous variables. Diabetics were excluded from glucose analyses, which were treated as quintiles and also as the top 5% or 10% versus the rest. Analyses were performed separately for men and women.
Study participants were flagged at the National Health Service Central Register in Edinburgh. A 20-year follow-up period was taken for all participants, and in this time 302 men (4.3%) and 387 women (4.6%) died of stroke (International Classification of Diseases, Ninth Revision codes 430 to 438). This represented 7.8% of male deaths and 12.2% of female deaths occurring in the follow-up period. Insufficient information was available to classify the types of stroke (ischemic or hemorrhagic).
Age-adjusted mortality rates per 10 000 per year were calculated by a life-table approach and were standardized by 5-year age groups, with the age distribution of the Renfrew/Paisley cohort used as the standard. Mortality rates were calculated for the whole 20-year follow-up period and for the first and second 10-year periods of follow-up separately. Cox proportional hazards regression models11 were used to calculate trends with the use of continuous variables and for calculating relative rates associated with 1-SD increase or decrease for each risk factor.
A multivariate model was constructed with the use of Cox proportional hazards regression models and all the risk factors simultaneously. Each variable was entered as a continuous variable, except for glucose, which was treated as a categorical variable (the top 5% versus the rest). Diabetes and preexisting CHD were also treated as categorical variables (no/yes). Smoking was entered by the number of cigarettes smoked per day, with an additional term for ex-smokers (no/yes).
The age-adjusted 20-year stroke mortality rate was similar for men and women (34.3 per 10 000 person years and 30.6 per 10 000 person years, respectively). This contrasts greatly with the all-cause mortality rates (242.7 and 160.1) and the CHD mortality rates (114.1 and 54.7) in the Renfrew/Paisley study, in which men have a far higher mortality rate than women.
Levels of systolic blood pressure and diastolic blood pressure were similar for men and women, as seen by the quintiles (Table 1⇓). There was a clear positive relationship between systolic blood pressure and stroke mortality in both men and women. The mortality rate was approximately 3 times that for men and women in the highest quintile of systolic blood pressure than in the lowest. Similar results were seen for the first and second 10 years of follow-up, although for men the gradient in the rates in the first 10 years was less consistently smooth. A similar positive relationship was seen between diastolic blood pressure and stroke mortality. In women there was a raised rate for those with diastolic blood pressure in the lowest quintile compared with the second and third quintiles, which was not seen in men. The relative rate for 1-SD increase in blood pressure was similar for systolic and diastolic blood pressure. For men the relative rate was higher for systolic blood pressure, whereas for women it was higher for diastolic blood pressure. In a model including systolic and diastolic blood pressure simultaneously, both were significantly associated with stroke mortality for men and women (P=0.0001 for systolic and P=0.024 for diastolic blood pressure for men, P=0.033 for systolic and P=0.0001 for diastolic blood pressure for women).
Women had higher levels of cholesterol than men (Table 2⇓). The level for the second fifth for women was equivalent to that for the third fifth for men. Cholesterol was not significantly associated with stroke mortality in either men or women. There was a suggestion of a U-shaped relationship, with the highest rate seen in the lowest quintile of cholesterol. Similar patterns were seen in the first and second 10 years of follow-up. Levels of glucose (with diabetics excluded) were similar in men and women. Glucose was positively related to stroke mortality in women but not in men. This was not apparent in the first 10 years of follow-up. The top quintile had the highest rate of stroke for both men and women in the whole follow-up period and in the second 10 years. Therefore, glucose was also analyzed by comparing the top 5% of the distribution with the remainder. The rate for men in the top 5% (glucose >7.2 mmol/L) was 46.9, compared with 31.9 for the rest, although the relative rate was not significant (1.31 [95% CI, 0.75 to 2.30]). For women, the top 5% (glucose >6.8 mmol/L) had a rate of 52.4, compared with 28.1 for the remainder. The relative rate was significant (2.18 [95% CI, 1.44 to 3.29]). Analyses of the top 10% versus the rest produced similar results.
Diabetics had a 3-fold increased rate of stroke mortality. Men who reported diabetes on the questionnaire had an age-adjusted mortality rate of 107.7, whereas men without diabetes had a rate of 33.5. The equivalent rates for women were 99.2 and 30.0. The relative rates were significant (2.89 [95% CI, 1.49 to 5.62] for men and 3.98 [95% CI, 2.29 to 6.91] for women).
Adjusted FEV1 was negatively associated with stroke mortality (Table 3⇓). Men and women in the bottom quintile of adjusted FEV1 had approximately double the rate of those in the top quintile, which represents those with the best lung function. Dividing the follow-up period into 2 halves produced results similar to those for the whole follow-up period. A significant positive relationship was seen between cardiothoracic ratio and stroke mortality for men and women. In men but not in women, there was a tendency for the association to be stronger in the first 10 years of follow-up than in the second.
More women than men had never smoked, and there were fewer heavy smokers and ex-smokers among women than men (Table 4⇓). There was a dose-response relationship between cigarette smoking and stroke mortality. Men smoking ≥25 cigarettes per day had almost double the rate of never-smokers. Pipe or cigar smokers had a rate between that of never-smokers and men smoking 1 to 14 cigarettes per day. Ex-smokers had a rate similar to that of never-smokers. The high rate for smokers of ≥25 cigarettes per day was not apparent in the first 10 years of follow-up, although it was high in the second 10 years of follow-up. Results for women were similar, although the highest rate was seen in women smoking 15 to 24 cigarettes per day.
There was no significant relationship between body mass index and stroke mortality in men or women, although there was a suggestion of a U-shaped relationship (Table 5⇓). Height was negatively associated with stroke mortality. Age-adjusted rates for the shortest fifth were approximately 1.5 times greater than for the tallest fifth for men and women. In the second 10 years of follow-up, there were significant downward trends, but this was not as clear in the first 10 years.
Men and women with preexisting CHD had higher rates of stroke mortality than men and women without preexisting CHD. Rates for men with preexisting CHD were 45.6 compared with 30.2 for men without preexisting CHD. Rates for women were 44.3 and 25.4. The relative rates were significant (1.63 [95% CI, 1.28 to 2.06] for men and 1.74 [95% CI, 1.42 to 2.14] for women).
When all the risk factors were simultaneously entered in a multivariate model, most variables retained a statistically significant association with stroke mortality (Table 6⇓). Age, diastolic blood pressure, cigarettes smoked per day, adjusted FEV1, height, body mass index, diabetes, and preexisting CHD were statistically significant risk factors for men and women. Systolic blood pressure, ex-smoker status, and cardiothoracic ratio were significant for men but not women, and high glucose was significant for women but not men. Relative rates for a 1-SD increase or decrease for continuous variables and relative rates for categorical variables are presented in Table 6⇓. Overall, relative rates were similar for men and women for each risk factor.
The relative rate of stroke mortality for men compared with women adjusted for age, diastolic blood pressure, smoking, adjusted FEV1, body mass index, diabetes, and preexisting CHD was 0.997 (95% CI, 0.84 to 1.18). This compares with a relative rate of 1.18 (95% CI, 1.02 to 1.38) adjusted for age only.
Stroke mortality rates were very similar for men and women in this study, in contrast to mortality from all causes and from CHD, for which male mortality rates are much higher than female mortality rates. For most of the risk factors, the risk factor relationships were similar for the whole follow-up period and the follow-up period divided into the first and second 10 years, although for some risk factors, the results in the first 10 years of follow-up were less clear.
Elevated blood pressure is a firmly established risk factor for stroke.12 13 Both systolic and diastolic blood pressure are positively associated with stroke risk. In the present study, systolic and diastolic blood pressure were both highly significantly associated with stroke mortality. Relative rates for 1-SD increase in systolic and diastolic blood pressure were similar, suggesting that either measure is a good indicator of future stroke mortality risk. However, including systolic and diastolic blood pressure in models simultaneously showed them both contributing to stroke mortality risk.
In previous studies the relative influence of diastolic and systolic blood pressure on stroke risk has varied. In a 12-year follow-up of men in the Oslo Study, diastolic blood pressure was a stronger predictor of stroke mortality than systolic blood pressure.14 In a 20-year study of Lithuanian men, both systolic and diastolic blood pressure were strongly associated with risk of stroke mortality.15 The Israeli Ischemic Heart Disease Project found strong associations between systolic and diastolic blood pressure and ischemic stroke in a 21-year follow-up of male civil servants and municipal employees.16 In the Busselton Study in Australia17 and a Finnish study of middle-aged subjects,18 both systolic and diastolic blood pressure were significantly associated with stroke mortality in men and women. The fact that in most studies both systolic and diastolic blood pressure contribute independently to stroke risk suggests that better characterization of the usual blood pressure level results in better prediction of future stroke. Information from both measures should be used in evaluating the level of risk of individuals for stroke.
In the present study there was no relationship between cholesterol and stroke mortality in men and women. The highest rate was seen for men and women in the lowest quintile of cholesterol, and a U-shaped relationship was observed. Collaborative studies have found no relationship between cholesterol and overall stroke,13 but this could mask the positive relation of cholesterol with ischemic stroke and the negative association with hemorrhagic stroke seen when subtypes of stroke are known.19 20
Glucose and Diabetes
Glucose in nondiabetics was found to be positively related to stroke mortality in women but not in men in the present study. Higher rates of stroke were seen in the top 5% of the distributions for men and women, although this was only significant for women. Nonfasting glucose was positively associated with ischemic stroke mortality in the Israeli study,16 but it was not found to be a risk factor for stroke in a 12-year follow-up of men without diabetes in the Oslo Study.14 In the Oslo Study, nonfasting glucose was adjusted for the time since the last meal. After 18 years of follow-up, glucose was found to be a significant risk factor only when diabetics were included in the analyses.21 In the British Regional Heart Study, nonfasting glucose was significantly higher in men who had had a stroke than in other men.22 In the Whitehall Study, middle-aged men were given a 50-g oral glucose load after an overnight fast.23 Blood samples were taken 2 hours later, and glucose was measured. Stroke mortality risk was doubled in men in the top 5% of the (nondiabetic) glucose distribution (5.4 to 11.0 mmol/L), and having a glucose reading in the top 5% was a significant risk factor in a multivariate model. These results suggest a threshold effect of glucose with stroke mortality risk among nondiabetics.
Men and women in the present study who were diabetic had a 3-fold greater stroke mortality rate compared with nondiabetics. Increased risk of stroke mortality for diabetics has been seen in other studies.15 16 17 21 24 25 Diabetes was the strongest risk factor for stroke mortality in men and women in eastern Finland.18
Forced Expiratory Volume
Adjusted FEV1 was found to be a significant risk factor for stroke mortality in men and women in the present study. Earlier results from this study with 15 years of follow-up showed that adjusted FEV1 was also a risk factor for stroke (and other causes of death) in lifelong nonsmokers.7 FEV1 was found to be associated with stroke mortality in an 18-year follow-up of men from the Whitehall Study.26 Other studies of risk factors for stroke did not include FEV1,14 15 17 22 25 although it is a relatively simply measured risk factor that is potentially modifiable.27
Cardiothoracic ratio, an indicator of left ventricular mass and left ventricular systolic function, is a predictor of CHD28 29 and cardiovascular30 mortality. In a secondary prevention stroke trial, cardiothoracic ratio has been found to predict recurrence,31 but it was not found to be an independent risk factor in a small prospective study among elderly men and women.32 The present study therefore provides the first data from a large-scale prospective study demonstrating that cardiothoracic ratio can aid in the identification of individuals at increased risk of stroke. However, the reduced use of routine chest radiography in screening examinations will mean that the necessary information will often not be available for this purpose. Future studies using echocardiography, as well as intervention studies in which cardiothoracic ratio may be reduced by drug therapy for hypertension or exercise, will add to better characterization of the usefulness of information about heart size and function both in predicting stroke risk and in monitoring the potential effectiveness of interventions.
Smoking has been found to be associated with stroke in many epidemiological studies.33 In a meta-analysis of 32 studies, cigarette smoking independently contributed to the risk of stroke, and there was a dose response with number of cigarettes smoked.34 Smoking cessation resulted in reduction of risk of stroke.
Rastenyte et al15 found that 19% of stroke deaths in men in a Lithuanian study were attributed to smoking, and smokers were almost twice as likely to die of stroke than nonsmokers. In a 14-year follow-up of the Finnmark Study in Norway of men and women aged 20 to 49 years, daily smokers had a 3-fold greater risk of dying of stroke than never-smokers and ex-smokers.25 In the present study, a clear dose response with cigarette smoking was seen, and pipe or cigar smokers had stroke rates similar to those of smokers of 1 to 14 cigarettes per day. Former smokers had stroke rates similar to those of never-smokers. Encouragement of smoking cessation is an important way to reduce the burden of stroke.
Other Risk Factors
Although potentially modifiable, body mass index has not been found to have a relationship with stroke mortality,14 15 17 but in a Finnish study, body mass index was associated with a significantly increased risk of stroke death among women (but not men).18 In the present study, body mass index was not related to stroke mortality, although there was a suggestion of a U-shaped relationship for both men and women.
Height was inversely related to stroke mortality in this cohort, and an inverse relationship was also seen in the Oslo Study14 and the Finnmark Study.25 In the British Regional Heart Study,35 a weaker association was found, and increased risk was restricted to the shortest quintile. Short height may serve as an indicator of socioeconomic deprivation and consequent interrupted growth in childhood, since adverse social environment in childhood has been related to an increased risk of stroke in later life.36
Men and women in this study with preexisting CHD had a higher rate of stroke mortality than men and women without preexisting CHD. Higher rates were also seen in the Busselton Study17 and the British Regional Heart Study.22
Overall, results were similar for men and women. Interaction tests between sex and each risk factor were not statistically significant, with only glucose attaining a nearly significant interaction term (P=0.08). Most quintiles were similar for men and women, with the exception of cholesterol, in which case women had higher levels than men. Although the risk of stroke mortality appeared similar for men and women, especially when compared with the far higher CHD and all-cause mortality rates for men compared with women, the age-adjusted relative rate was significantly higher for men than women (1.18 [95% CI, 1.02 to 1.38]). However, when adjusted for risk factors that were significant in both the male and female multivariate models (with the exception of height, which serves as a proxy for sex), there was no difference in the relative rates (0.997 [95% CI, 0.84 to 1.18]). The risk factor that behaved differently for men and women was glucose, which was significant for women but not for men, in both the quintile-based analysis and the top 5% analysis. Large studies including women are needed to verify this result. Diabetes, however, was a significant risk factor for both men and women.
The multivariate models were performed with all participants, with those with missing values given mean values for continuous variables and those with missing categorical variables placed in the “no” category. However, performing the models with the exclusion of participants with any missing variables, including the large numbers of men and women whose glucose levels were not measured, did not change the results. Combining the risk factors in the multivariate model suggested that several risk factors may act together to increase the risk of stroke mortality. Other studies have found similar risk factors in multivariate models retaining statistical significance. Age, blood pressure (either systolic or diastolic), and smoking were consistently found to be independent risk factors for stroke mortality or incidence.14 15 16 17 25 37 Some studies found other independent risk factors that were not measured in the present study (physical activity at leisure,14 alcohol consumption,17 and triglycerides25 ). Likewise, FEV1, cardiothoracic ratio, and height were found to be independent risk factors for stroke mortality in the present study but have not generally been analyzed in relation to stroke risk in previous studies.
This study has found significant associations between blood pressure, FEV1, smoking, glucose, diabetes, cardiothoracic ratio, height, and preexisting CHD and stroke mortality for men and women. No associations were seen with cholesterol or body mass index. Since many of these risk factors are potentially modifiable, there is scope to further reduce mortality from stroke in men and women. Overall, results for women were similar to those for men. This would suggest that any protective influence of female hormones on CHD mortality is not having a similar effect on stroke mortality. Control of risk factors for reduction of stroke mortality should thus be targeted at men and women in a similar fashion, in particular through encouraged smoking cessation and blood pressure control.
This study was supported by grants from Chest, Heart, and Stroke Scotland and the Stroke Association. Victor M. Hawthorne and Charles R. Gillis were responsible for the original study of the Renfrew/Paisley cohort, and Pauline MacKinnon is responsible for updating mortality.
- Received May 18, 1999.
- Revision received July 8, 1999.
- Accepted July 8, 1999.
- Copyright © 1999 by American Heart Association
Warlow C. Epidemiology of stroke. Lancet. 1998;352(suppl III):1–4.
Registrar General for Scotland. Annual Report of the Registrar General for Scotland 1997. Edinburgh, Scotland: General Register Office for Scotland; 1998.
Sacco R, Benjamin E, Broderick J, Dyken M, Easton J, Feinburg W, Goldstein L, Gorelick P, Howard G, Kittner S, Manolio T, Whisnant J, Wolf P. American Heart Association Conference IV: stroke—risk factors. Stroke. 1997;28:1507–1517.
Wolf PA. Prevention of stroke. Lancet. 1998;352:15–18.
Hole DJ, Watt GM, Davey-Smith G, Hart CL, Gillis CR, Hawthorne VM. Impaired lung function and mortality risk in men and women: findings from the Renfrew and Paisley prospective population study. BMJ. 1996;313:711–716.
Hart CL, Watt GCM, Smith GD, Gillis CR, Hawthorne VM. Pre-existing ischaemic heart disease and ischaemic heart disease mortality in women compared with men. Int J Epidemiol. 1997;26:508–515.
Prineas RJ, Crow RS, Blackburn H. The Minnesota Code Manual of Electrocardiographic Findings: Standards and Procedures for Measurement and Classification. Boston, Massachusetts: John Wright; 1982.
Cox DR. Regression models and life tables. J R Stat Soc B. 1972;34:187–220.
MacMahon S, Peto R, Cutler J, Collins R, Sorlie P, Neaton J, Abbott R, Godwin J, Dyer A, Stamler J. Blood pressure, stroke, and coronary heart disease, part 1: prolonged differences in blood pressure: prospective observational studies corrected for the regression dilution bias. Lancet. 1990;335:765–774.
Haheim LL, Holme I, Hjermann I, Leren P. Risk factors of stroke incidence and mortality: a 12-year follow-up of the Oslo Study. Stroke. 1993;24:1484–1489.
Rastenyte D, Tuomilehto J, Domarkiene S, Cepaitis Z, Reklaitiene R. Risk factors for death from stroke in middle-aged Lithuanian men: results from a 20-year prospective study. Stroke. 1996;27:672–676.
Tanne D, Yaari S, Goldbourt U. Risk profile and prediction of long-term ischemic stroke mortality: a 21-year follow-up in the Israeli Ischemic Heart Disease (IIHD) Project. Circulation. 1998;98:1365–1371.
Tuomilehto J, Rastenyte D, Jousilahti P, Sarti C, Vartiainen E. Diabetes mellitus as a risk factor for death from stroke: prospective study of the middle-aged Finnish population. Stroke. 1996;27:210–215.
Ben-Shlomo Y, Neaton J, Wentworth D, Davey Smith G. Serum cholesterol and mortality from haemorrhagic and non-haemorrhagic stroke: 16 year follow up of a third of a million men screened for the Multiple Risk Factor Intervention Trial. J Epidemiol Community Health. 1998;52:683–684. Abstract.
Haheim LL, Holme I, Hjermann I, Leren P. Nonfasting serum glucose and the risk of fatal stroke in diabetic and nondiabetic subjects. Stroke. 1995;26:774–777.
Shaper AG, Phillips A, Pocock S, Walker M, Macfarlane P. Risk factors for stroke in middle aged British men. BMJ. 1991;302:1111–1115.
Fuller J, Shipley M, Rose G, Jarrett R, Keen H. Mortality from coronary heart disease and stroke in relation to degree of glycaemia: the Whitehall Study. BMJ. 1983;287:867–870.
Manolio TA, Kronmal RA, Burke GL, O’Leary DH, Price TR. Short-term predictors of incident stroke in older adults. Stroke. 1996;27:1479–1486.
Njolstad I, Arnesen E, Lund-Larsen PG. Body height, cardiovascular risk factors, and risk of stroke in middle-aged men and women: a 14-year follow-up of the Finnmark Study. Circulation. 1996;94:2877–2882.
Strachan D. Ventilatory function as a predictor of fatal stroke. BMJ. 1991;302:84–87.
Strachan D. Predicting and preventing premature mortality. BMJ. 1996;313:715–716.
Hemingway H, Shipley M, Christie D, Marmot M. Is cardiothoracic ratio in healthy middle aged men an independent predictor of coronary heart disease mortality? Whitehall Study 25 year follow-up. BMJ. 1998;316:1353–1354.
Hart C, Davey Smith G. Knowledge of cardiothoracic ratio adds to cardiovascular risk stratification. BMJ. 1998;317:1321–1322.
Rautaharju P, LaCroix A, Savage D, Haynes S, Madans J, Wolf H, Hadden W, Keller J, Cornoni-Huntley J. Electrocardiographic estimate of left ventricular mass versus radiographic cardiac size and the risk of cardiovascular disease mortality in the epidemiologic follow-up study of the First National Health and Nutrition Survey. Am J Cardiol. 1998;62:59–66.
Van Latum J, Koudstaal P, Venables G, Van Gijn J, Kappelle L, Algra A. Predictors of major vascular events in patients with a transient ischaemic attack or minor ischemic stroke and with nonrheumatic atrial fibrillation. Stroke. 1995;26:801–806.
Aldoori MI, Rahman SH. Smoking and stroke: a causative role. BMJ. 1998;317:762–763.
Shinton R, Beevers G. Meta-analysis of relation between cigarette smoking and stroke. BMJ. 1989;298:789–794.
Wannamethee G, Shaper A, Whincup P, Walker M. Adult height, stroke and coronary heart disease. Am J Epidemiol. 1998;148:1069–1076.
Davey Smith G, Hart CL, Blane D, Hole D. Adverse socioeconomic conditions in childhood and cause specific adult mortality: prospective observational study. BMJ. 1998;316:1631–1635.
Wolf PA, D’Agostino RB, Belanger AJ, Kannel WB. Probability of stroke: a risk profile from the Framingham Study. Stroke. 1991;22:312–318.