North-South Gradients in Britain for Stroke and CHD
Are They Explained by the Same Factors?
Background and Purpose— The geographic variation in CHD and stroke within Great Britain is well known. We aimed to quantify these variations and to determine the contribution of established risk factors.
Methods— This prospective study consisted of 7735 men 40 to 59 years of age in 24 British towns who were followed up for 20 years from screening in 1978 to 1980. We compared age-adjusted incidences of major stroke and CHD events in southern England and the rest of Britain before and after adjustment for established cardiovascular risk factors.
Results— At least 1 episode of stroke occurred in 467 men (3.54 per 1000 person-years, age standardized) and of CHD in 1299 men (10.05 per 1000 person-years). Event rates varied between towns, from 2.00 to 5.45 per 1000 person-years for stroke and from 6.16 to 12.21 per 1000 person-years for CHD. Incidence for both diseases was highest in Scottish towns and lowest in southern English towns (“north-south gradient”). The hazard ratio for stroke in the rest of Britain compared with southern England was 1.44 (95% confidence interval [CI], 1.16 to 1.78); for CHD, it was 1.32 (95% CI, 1.14 to 1.53). After adjustment for baseline systolic blood pressure, smoking status, physical activity, social class, and height, the hazard ratio was 1.24 (95% CI, 1.00 to 1.54) for stroke and 1.17 (95% CI, 1.02 to 1.35) for CHD.
Conclusions— Similar north-south gradients were observed for major stroke and major CHD events. The magnitude of these gradients was considerably diminished when individual risk variables were taken into account.
- cerebrovascular accident
- cohort studies
- confounding factors (epidemiology)
- coronary heart disease
Geographic variations in coronary heart disease (CHD) and stroke death rates have long been observed in Britain, with lower mortality in the south of England and higher mortality in the north of England and Scotland.1 Attempts to explain these variations have usually consisted of analyzing aggregated data on suspected risk factors over geographical areas. When such data for risk factors are analyzed in relation to disease rates, the relationships may be overestimated because of the ecological fallacy.2 In contrast, the British Regional Heart Study (BRHS) has obtained individual data on risk factors for 7735 individual men in 24 British towns who were followed up for 20 years. It is thus possible to examine differences in disease rates between towns in relation to risk factors measured at an individual level.
The BRHS has now followed up these men for >20 years. We have already demonstrated the geographical variation in incidence of CHD over the first 15 years of follow-up,3 and have shown that 77% of this variation may be explained by 5 established risk factors: smoking, systolic blood pressure, physical activity, social class, and height.
See Editorial Comment, page 2609
Here, we report and compare the magnitude of the difference for the 2 disease end points, namely stroke and CHD, between men living in southern England and men in the rest of Britain. We also report the extent to which such gradients may be explained in terms of established risk factors for the 2 diseases.
The design of the BRHS has been described in detail.4 In the main phase of the BRHS, 24 towns were selected to represent the range of CHD mortality rates and to include all the major regions of Great Britain. Random samples of ≈400 men 40 to 59 years of age drawn from a single general practice in each town were invited for screening. The general practice was chosen to be representative of the socioeconomic composition of the town. A 78% response rate was obtained, and 7735 men were screened between 1978 and 1980.
A single team of 3 trained research nurses visited all towns in succession. Towns in close proximity were visited at different times of year. The London School of Hygiene and Tropical Medicine sphygmomanometer was used to measure blood pressure twice in succession. The mean of the 2 readings was used in analyses, with adjustment for observer variation within each town.5 Height was measured to the nearest millimeter in subjects without shoes and weight to the nearest 0.1 kg in subjects wearing trousers and socks.
The research nurses administered a standard questionnaire that included questions on smoking habits, physical activity, and social class based on the longest-held occupation. Smoking was defined as 1 of 5 categories: never smokers, ex-smokers, and smokers of 1 to 19, 20, or ≥21 cigarettes per day. For physical activity, an established 6-category classification was used6 but was reduced to 4 categories for this analysis: none, occasional, light, and moderate or more (active). Social class was defined by use of the Registrar General’s Classification of Occupations and concerned the longest-held job. Seven categories were defined: I, II, III nonmanual, III manual, IV, V, and armed forces.7 History of diagnosed CHD or stroke was defined as subject recall of ever having had a doctor’s diagnosis of “angina,” “heart attack,” “myocardial infarction,” or “coronary thrombosis” for CHD and “stroke” for stroke.
All men have been followed up for major nonfatal and fatal CHD events (myocardial infarction and sudden cardiac death) and major fatal and nonfatal stroke.8 Deaths have been flagged through the National Health Service Central Registers in Southport for England and Wales and in Edinburgh for Scotland. Fatal events resulting from CHD or stroke were recorded if the International Classification of Diseases, ninth revision, codes were 410 to 414 and 430 to 438, respectively. Regular reviews of general practice records have been carried out biennially throughout the follow-up period. Nonfatal myocardial infarctions were defined according to standard criteria.8 Nonfatal strokes comprised all those cerebrovascular events that produced a neurological deficit present for >24 hours.9 Follow-up rates have been >99%. A cross-check between medical records and subjects’ recall of diagnoses confirmed that 97% of CHD diagnoses and 77% of stroke diagnoses were correctly identified by medical records.10 Also, only 5% of strokes and <1% of CHD events initially identified by medical records were false-positives. The database was updated accordingly. A record review carried out in 2000 completed at least 20 years of follow-up for every subject. We have therefore calculated event rates for all 24 towns over a 20-year follow-up period for both CHD and stroke events.
Event rates for first CHD or stroke were calculated per 1000 person-years; follow-up time was counted as the elapsed years between screening and the first event for men who experienced the event and as 20 years for other men unless they died of other causes before 20 years, in which case they were censored at that time. For each town, these rates were calculated within the age groups of 40 to 44, 45 to 49, 50 to 54, and 55 to 59 years and then averaged to obtain an age-standardized event rate.
Five risk factors (smoking, systolic blood pressure, physical activity, social class, and height) were selected. These factors had explained 77% of the variation in incidence of CHD over a 15-year follow-up in the BRHS.3 The first 2 have well-established relationships with both CHD and stroke.11 The latter 3 have been demonstrated to be related to either CHD or stroke in the BRHS and other studies.12 Serum total cholesterol, although a classic risk factor for CHD, is not related to stroke incidence,12 and our previous work found that it failed to explain geographical variation in CHD.3 Height was taken as a proxy marker for deprivation in various life stages before adulthood, and social class was a marker for social disadvantage in adult life.
Pearson correlations of age-standardized CHD and stroke incidence with mean levels of systolic blood pressure and height and prevalence of current cigarette smoking, moderate or vigorous physical activity, and manual social class were calculated for the 24 towns.
Because a sample of 24 towns was chosen from a larger number of possible towns and because subjects were chosen from each of the towns, the data formed a multilevel structure.3,13 The statistical package MLwiN was used, including a macro for fitting Cox’s proportional-hazards model for survival data grouped into years of follow-up. All models were adjusted for subject age as a continuous variable and included region. The 2 chief subject-level variables (smoking status, systolic blood pressure) were entered. Then, physical activity, social class, and height were added, and the residual between-town variance was noted.
Each town was classified according to whether it was in southern England or the rest of Britain. This dichotomous classification was included in our models as a town-level variable rather than an individual-level variable. Hazard ratios were calculated for the rest of Britain compared with the south of England.
Over 20 years of follow-up, major CHD events occurred in 1299 of 7735 men (16.8%), equivalent to 10.05 first events per 1000 person-years of follow-up. Stroke events occurred in 467 men (6.0%, 3.54 first events per 1000 person-years). The average age at first stroke was 65.1 years and at first CHD event was 62.3 years. For 97 men, both a CHD event and a stroke event were recorded.
Association Between Incidences for 2 End Points in 24 Towns
Table 1 shows the age-standardized 20-year event rates of major CHD and stroke in the 24 towns ordered by region. CHD event rates varied from 6.16 per 1000 person-years in Guildford to 12.21 per 1000 person-years in Dewsbury, whereas stroke event rates varied from 2.00 per 1000 person-years in Guildford to 5.45 per 1000 person-years in Falkirk. There was a correlation of 0.49 between the rates for the 2 diseases (the Figure). Table 1 also shows risk factor distributions for each town. For CHD incidence, strong correlations were observed with the baseline prevalence of current smoking (positively) and mean height (negatively), whereas moderate positive correlations existed with mean systolic blood pressure and prevalence of manual social class. For stroke incidence, moderate correlations were observed with baseline prevalence of current smoking and manual social class (positively) and mean height (negatively). A weak positive correlation existed with mean systolic blood pressure, and no correlation existed with physical activity.
Role of Individual Risk Factors
Complete data on smoking, blood pressure, physical activity, social class, and height were available for 7609 men. Table 2 shows hazard ratios for each individual risk factor for the 2 diseases after adjustment for each other and age. Strong associations were found with smoking and blood pressure for both CHD and stroke, with physical activity and height for CHD, and with social class for stroke. Only weak, nonsignificant associations with social class for CHD and with physical activity and height for stroke were noted.
Tables 3 and 4⇓ show the hazard ratios for men living in the rest of Britain compared with southern England for CHD and stroke, respectively. After adjustment only for age, the hazard ratio for the rest of Britain compared with the south of England was 1.32 (95% confidence interval [CI], 1.14 to 1.53) for CHD and 1.44 (95% CI, 1.16 to 1.78) for stroke. When smoking and blood pressure were included in the model, the magnitude of the log (hazard ratio) was reduced by 29% for CHD and by 28% for stroke. After additional adjustment for physical activity, social class, and height, the magnitude was reduced by 42% and 40% for CHD and stroke, respectively. For CHD, smoking and systolic blood pressure were the 2 variables most effective in reducing the magnitude of the regional difference (although the other 3 variables were almost as effective), whereas for stroke, systolic blood pressure and social class were most effective for reducing this difference.
Exclusion of Subjects With Prior Diagnosis of CHD or Stroke
At the initial screening, 322 of the 7609 men recalled a doctor diagnosis of CHD or stroke, and for another 6, information on recall of these diagnoses was not available. The analysis was repeated for each end point including only those 7281 subjects who did not recall a doctor diagnosis of CHD or stroke. For stroke, the results were very similar to those obtained before exclusion of the 328 subjects. Hazard ratios for living outside the south of England were 1.46 (95% CI, 1.23 to 1.68; P=0.001) and 1.27 (95% CI, 1.04 to 1.50; P=0.037) before and after adjustment for the 5 variables, and the magnitude of the log (hazard ratio) was reduced by 36% when the 5 variables were included in the model. For CHD, the hazard ratios were somewhat reduced when the 328 subjects were omitted. The hazard ratio was 1.25 (95% CI, 1.10 to 1.40; P=0.003) when adjusted only for age. This was reduced to 1.10 (95% CI, 0.96 to 1.24; P=0.18) when the 5 variables were included. The magnitude of the log (odds ratio) was reduced by 59%.
The present analysis has shown that the incidence of stroke was increased in regions of Britain outside the south of England, just as for CHD. The increased incidence outside the south of England was slightly greater for stroke both before and after adjustments for individual risk factors and after exclusion of subjects who recalled a doctor diagnosis of CHD or stroke at baseline. Generally, towns with a high incidence of 1 disease also had high incidence for the other. However, only 97 men experienced both types of disease over the 20-year follow-up; this amounted to about one fifth of the strokes and <1 in 13 of the CHD events. Lower mean blood pressure was the variable that most accounted for the decreased incidence of stroke in the south (18%), whereas lower mean blood pressure and lower smoking prevalence accounted to a similar extent for the decreased incidence of CHD (15% and 14%, respectively). Physical activity and height made very little contribution to explaining the geographic variation in stroke, but their contribution, and that of social class, was almost as great as for smoking and blood pressure in explaining variations in CHD.
Unadjusted hazard ratios for the 2 diseases were 1.32 and 1.44 for men living in the rest of Britain compared with the south of England or, inverting, 0.76 and 0.69 for those living in the south of England compared with the rest of Britain. This suggests the potential to avoid about one quarter of CHD and one third of stroke events occurring in the rest of Britain if conditions enjoyed by men in the south of England could be replicated elsewhere. Decreases in rates of smoking and levels of blood pressure to levels prevalent in the south would reduce the inequality by almost one third.
Strengths and Weaknesses of the Present Study
The BRHS represents towns in all major regions of Britain. The cardiovascular mortality for these towns differed markedly when the men were screened.4 Other major British prospective studies of CHD or stroke are based on a restricted geographical location. The known geographic variation in disease rates potentially allowed the study to explore the contribution of known risk factors that may differ in their geographic distribution. Thus, the potential benefit when risk factors are modified at the population level may be estimated. Other studies have examined geographic variation by use of aggregated data. Overall rates of disease have been calculated from routinely collected statistics and related to prevalence of risk factors in different areas.14–17 In contrast, the BRHS has been able to avoid possible ecological biases inherent in aggregated data2 by relating relationships at the individual level to variations in incidence at the town level through multilevel modeling.
In using stroke data ascertained from death certificates (for fatal strokes) and general practice medical records (for nonfatal strokes), we have been unable to distinguish between ischemic and hemorrhagic strokes. It has been suggested that height may be more strongly related to hemorrhagic strokes than ischemic strokes.18 However, the vast majority of strokes of unknown origin are probably ischemic in older British populations.19
The ecological correlations observed between towns for risk factors with stroke events were lower than with CHD, probably because the lower stroke event rates led to incidence estimates that were more liable to sampling variation. The real strength of our analysis was in quantifying the geographic gradient within a multilevel model.
It has been shown that associations between individual risk variables and outcome may be underestimated by taking baseline measures to represent usual levels over the follow-up period.20 An analysis that could account for such imprecision may explain the regional differences even more fully than demonstrated by these findings. However, the use of baseline measurements themselves would be more relevant for prediction of regional differences.
Comparison With Findings of Other Studies
Other studies have pointed out trends in risk factor distribution among areas where stroke mortality was known to differ. Hypertension prevalence and glucose intolerance mirrored a gradient observed for stroke mortality in 3 areas in the United States.21 Among men in 22 Scottish health districts, hospital admission rates for stroke were associated with prevalence of hypertension, alcohol consumption, and absence of fruit consumption.22 Common to both is the importance of raised blood pressure. Our study has shown that this single factor was the strongest explanation for the variation in stroke rates between towns.
Occupational social class was the second-most-effective variable in explaining geographic variation. The relationship has previously been demonstrated at an individual level both in the BRHS cohort and in a Scottish cohort (Renfrew/Paisley study), but in both cases, its effect has been at least partially explained by other risk factors.23 Because height has not explained geographic variation in stroke rates in the present study, the contribution of adult social class may reflect exposures to risk factors acting in adult life that occur more commonly outside the south of England.
This study was limited to British middle-aged men, but a recently mounted study of British women 60 to 79 years of age drawn from almost the same general practices demonstrated similar findings using cross-sectional data on geographical variation in cardiovascular disease prevalence.24
Public Health Implications
Although differences in the prevalence of the established risk factors appear to make a substantial contribution to the modest increase in risk of stroke and CHD associated with living beyond southern England, it should be recognized that mean levels of serum total cholesterol and body mass index are uniformly high throughout Britain and that the prevalence rates of other risk factors even in southern England are far from desirable. The relatively small changes in blood pressure, cigarette smoking, and physical activity required to bring the rest of Britain down to the levels of CHD and stroke encountered in southern England cannot be seen as a major public health objective when the levels of cardiovascular risk in southern England are still high by international standards. Clearly, any public health actions on diet, body weight, smoking, physical activity, and control of blood pressure must be directed to the whole population throughout Britain and not targeted at specific geographical groups.
The BRHS is funded by the British Heart Foundation with additional support from the Department of Health. Opinions expressed in the article are those of the authors and not necessarily those of the funding bodies.
- Received March 6, 2003.
- Revision received June 26, 2003.
- Accepted July 8, 2003.
Charlton J, Murphy M, Khaw K, Ebrahim S, Davey Smith G. Cardiovascular diseases. In: Charlton J, Murphy M, eds. The Health of Adult Britain 1841–1994. London, UK: Stationery Office; 1997; 2: 60–81.
Greenland S. Ecologic versus individual-level sources of bias in ecologic estimates of contextual health effects. Int J Epidemiol. 2001; 30: 1343–1350.
Morris RW, Whincup PH, Lampe FC, Walker M, Wannamethee SG, Shaper AG. Geographic variation in incidence of coronary heart disease in Britain: the contribution of established risk factors. Heart. 2001; 86: 277–283.
Shaper AG, Pocock SJ, Walker M, Cohen NM, Wale CJ, Thomson AG. British Regional Heart Study: cardiovascular risk factors in middle-aged men in 24 towns. BMJ. 1981; 283: 179–186.
Shaper AG, Wannamethee G, Weatherall R. Physical activity and ischaemic heart disease in middle-aged British men. Br Heart J. 1991; 66: 384–394.
Office of Population Censuses and Surveys. Classification of Occupations and Coding Index. London, UK: HM Stationery Office; 1980.
Walker M, Shaper AG, Lennon L, Whincup PH. Twenty year follow-up of a cohort based in general practices in 24 British towns. J Public Health Med. 2000; 22: 479–485.
Shaper AG, Phillips AN, Pocock SJ, Walker M, Macfarlane PW. Risk factors for stroke in middle aged British men. BMJ. 1991; 302: 1111–1115.
Walker MK, Whincup PH, Shaper AG, Lennon LT, Thomson AG. Validation of patient recall of doctor-diagnosed heart attack and stroke: a postal questionnaire and record review comparison. Am J Epidemiol. 1998; 148: 355–361.
Shaper AG. Causes of coronary heart disease: new concepts. In: Sharpe I, ed. Looking to the Future: Making Coronary Heart Disease an Epidemic of the Past. London, UK: Stationery Office; 1999.
Goldstein H. Multilevel Statistical Models. 3rd ed. London, UK: Hodder Arnold; 2003.
Howard G, Howard VJ, Katholi C, Oli MK, Huston S. Decline in US stroke mortality: an analysis of temporal patterns by sex, race, and geographic region. Stroke. 2001; 32: 2213–2220.
Pickle LW, Mungiole M, Gillum RF. Geographic variation in stroke mortality in blacks and whites in the United States. Stroke. 1997; 28: 1639–1647.
Rodriguez AF, Guallar-Castillon P, Gutierrez-Fisac JL, Ramon BJ, del Rey CJ. Socioeconomic level, sedentary lifestyle, and wine consumption as possible explanations for geographic distribution of cerebrovascular disease mortality in Spain. Stroke. 1997; 28: 922–928.
He J, Klag MJ, Wu Z, Whelton PK. Stroke in the People’s Republic of China, I: geographic variations in incidence and risk factors. Stroke. 1995; 26: 2222–2227.
McCarron P, Hart CL, Hole D, Davey Smith G. The relation between adult height and haemorrhagic and ischaemic stroke in the Renfrew/Paisley study. J Epidemiol Community Health. 2001; 55: 404–405.
Clarke R, Shipley M, Lewington S, Youngman L, Collins R, Marmot M, et al. Underestimation of risk associations due to regression dilution in long-term follow-up of prospective studies. Am J Epidemiol. 1999; 150: 341–353.
Stolley PD, Kuller LH, Nefzger MD, Tonascia S, Lilienfeld AM, Miller GD, et al. Three-area epidemiological study of geographic differences in stroke mortality, II: results. Stroke. 1977; 8: 551–557.
Starr JM, Thomas B, Whalley LJ. Population risk factors for hospitalization for stroke in Scotland. Int J Epidemiol. 1996; 25: 276–281.
Lawlor DA, Bedford C, Taylor M, Ebrahim S. Geographical variation in cardiovascular disease, risk factors, and their control in older women: British Women’s Heart and Health Study. J Epidemiol Community Health. 2003; 57: 134–140.