Variations in Stroke Incidence and Survival in 3 Areas of Europe
Background and Purpose—Comparison of incidence and case-fatality rates for stroke in different countries may increase our understanding of the etiology of the disease, its natural history, and management. Within the context of an aging population and the trend for governments to set targets to reduce stroke risk and death from stroke, prospective comparison of such data across countries may identify what drives the variation in risk and outcome.
Methods—Population-based stroke registers, using multiple sources of notification, ascertained cases of first in a lifetime stroke between 1995 and 1997 for all age groups. The study populations were in Erlangen, Germany; Dijon, France; and London, UK. Crude incidence rates were age-standardized to the European population for comparative purposes. Case-fatality rates up to 1 year after the stroke were obtained, and logistic regression adjusting for age group, sex, and pathological subtype of stroke was used to compare survival in the 3 communities.
Results—A total of 2074 strokes were registered over the 3 years. The age-standardized rate to the European population was 100.4 (95% CI 91.7 to 109.1) per 100 000 in Dijon, 123.9 (95% CI 115.6 to 132.2) in London, and 136.4 (95% CI 124.9 to 147.9) in Erlangen. Both crude and adjusted rates were lowest in Dijon, France. The incidence rate ratio, with Dijon as the baseline comparison (1), was 1.21 (95% CI 1.09 to 1.34) in London and 1.37 (95% CI 1.22 to 1.54) in Erlangen (P<0.0001). There were significant differences in the proportion of the subtypes of stroke between populations, with London having lower rates of cerebral infarction and higher rates of subarachnoid hemorrhage and unclassified stroke (P<0.001). Case-fatality rates varied significantly between centers at 1 year, after adjustment for age, sex, and subtype of stroke (35% overall, 34% Erlangen, 41% London, and 27% Dijon; P<0.001).
Conclusions—The impact of stroke is considerable, and the risk of stroke varies significantly between populations in Europe as does the risk of death. The striking differences in survival require clarification but lend weight to the evidence that stroke management may differ between northern and central Europe and influence outcome.
Assessing the need for stroke prevention and stroke care is most sensitively achieved with the use of incidence rates.1 Such epidemiological studies are costly and relatively rare compared with studies using mortality data, hospital-based stroke registers, or incidence studies in younger age groups of the population only.
Over the past 20 years, there have been studies of both mortality and incidence indicating significant variations in the outcome of stroke between populations. Comparisons of mortality data in Europe, adjusted for age and sex differences, indicate significant variations in mortality, with the rates being lowest in France and highest in the UK, Portugal, and Italy.2 Incidence studies in the younger groups (aged <65 years) in the Monitoring Trends and Determinants in Cardiovascular Disease (MONICA) project illustrate similar variations in the incidence and case-fatality rates for first in a lifetime strokes in Europe.3 Overviews of population-based stroke registers, including all age groups, show variations in incidence rates, with the lowest again being in France.4 However, comparisons such as that by Sudlow and colleagues4 analyze studies in different time periods that have used differing methodologies. Such biases may cause difficulties with the interpretation of observed differences in the risk of stroke between populations.
In a European BIOMED Program,5 using hospital-based stroke registers, we have observed significant differences in the case-fatality rates from stroke between European centers after adjustment for stroke severity. The highest case-fatality rate was observed in the UK centers, and the lowest was in France. In the International Stroke Trial, Signorini et al6 have reported significant variations in case-fatality rates between countries that could not be accounted for by chance alone, with the United Kingdom again having the poorest outcome.
The European collaboration of registers of stroke (EROS) is composed of centers with population-based stroke registers, with data collection being standardized between them. The aim of the collaboration is to compare the risk of stroke, the natural history of stroke, and the impact of stroke on health services in European populations in a prospective fashion. The present study estimates the incidence and survival of stroke in areas in France, Germany, and England from 1995 to 1997 to establish whether previously observed differences in outcome exist in a prospective, standardized, population-based study.
Subjects and Methods
Community-based stroke registers were established in Dijon, France; Erlangen, Germany; and south London, UK. The registers comply with criteria considered appropriate for such registers4 and have been described in detail elsewhere.7 8 9 The main aspects of the case ascertainment and data collection are summarized in the present study. All data were collected by use of standardized questions developed for the European BIOMED program.5 10
Dijon, France, is an urban area with a population of 143 088.7 Erlangen, Germany, is an urban area with a population of 100 330.8 In south London, the register covers an inner city population of 234 533, with a 21% black population.9 The definition of stroke was that of the World Health Organization with CT confirmation wherever feasible,4 and transient ischemic attacks were not included. All first strokes were registered between January 1, 1995, and December 31, 1997. Multiple overlapping sources of notification were used to increase the chances of full ascertainment. All registers visited the hospitals serving the study area regularly (inpatients and outpatients) and maintained regular (≈3 times monthly) contact with primary care and nursing home sources for notification and scrutiny of death certificates and postmortem reports (International Classification of Diseases, 9th Revision, 430-434 and 436), with validation according to clinical registration criteria. These hot-pursuit techniques advocated by Sudlow and colleagues4 are particularly relevant for community cases, particularly for the elderly and for sudden deaths. Similar techniques were also used in the pediatric cases.
Strokes were characterized as cerebral infarction, primary intracerebral hemorrhage, and subarachnoid hemorrhage and were based on CT scan information within 30 days of the onset of stroke. All patients with stroke in whom brain imaging could not be performed were “unclassified.”
All patients were examined by a study physician within the first week of stroke wherever possible, and information was gained from the patient, their relatives and caregivers, and their healthcare professionals. Information collected at the initial assessment included self-definition of ethnic origin in London only.9 Patients were assessed at 3 months and 1 year after the initial stroke. In the present study, survival to 12 months is reported.
In London, UK, the denominators for the calculation of the incidence rate were the 1996 adjusted estimates of the 1991 census data. In Erlangen, Germany, the denominator was the updated census data from December 31, 1997. In Dijon, France, the denominator was estimated from the 1990 census. Incidence rates were calculated for the whole population and in the following age bands: <45, 45 to 54, 55 to 64, 65 to 74, 75 to 84, and >84 years. Sex-specific and subtype of stroke-specific incidence rates were also calculated and standardized for age to European and World populations.11 CIs for the age-specific rates and age-standardized rates were calculated by the Poisson distribution. Incidence rate ratios comparing the 3 centers were calculated by Poisson regression, adjusting for age group, sex, and pathological subtype of stroke. Incidence rate ratios were also calculated to examine center differences in incidence for each subtype of stroke, adjusting for age group and sex. Age, sex, and subtype of stroke-standardized case-fatality rates were calculated at 28, 90, and 365 days after stroke. Logistic regression adjusting for age group, sex, and subtype was used to compare survival in the 3 centers.
A total of 2074 first strokes were registered in the 3 communities during the 3 years. Table 1⇓ outlines the age, sex, and subtype distribution between communities, with significant variation in the subtypes of stroke, particularly unclassified and subarachnoid hemorrhage.
The age- and community-specific annual incidence rates are detailed in Table 2⇓ and Figure 1⇓. The incidence rates standardized to European and world populations are also displayed. The incidence of stroke rises significantly with age. The European standardized rates differ significantly, with Dijon having the lowest rates, then London, and Erlangen having the highest rates.
Table 3⇓ outlines the crude and standardized sex-specific incidence rates in the 3 communities. Males had higher rates than did females in all 3 communities. Again, significantly lower rates in males and females are reported in Dijon compared with London and Erlangen.
Pathological diagnosis of stroke subtype was confirmed in 1933 cases (93%), which varied between center (95% Erlangen, 96% Dijon, and 88% London; P<0.001) (Table 1⇑). Table 4⇓ and Figure 2⇓ detail the incidence rates by subtype. There were higher rates of cerebral infarction in Erlangen and lower rates of intracerebral hemorrhage and subarachnoid hemorrhage in Dijon.
The age, sex, and subtype of stroke–adjusted 28-, 90-, and 365-day case-fatality rates are detailed in Table 5⇓. There are significant differences in survival at all time points, with poorer survival in London.
Table 6⇓ reports the incidence rate ratios adjusted for age group for particular variables. There is a significant independent effect of sex, center, and subtype of stroke on the incidence rate ratios, with more stroke in men. Cerebral infarction and primary intracerebral hemorrhage are high in Erlangen, and subarachnoid hemorrhage is high in London.
This large prospective incidence study in Europe identifies significant differences in the risk of and survival from stroke between communities. The present study has used recognized standardized methods of case ascertainment1 4 and confirms previous findings from studies in differing time periods and age groups that the risk of stroke varies significantly.4 It also provides current estimates of the incidence of stroke and, hence, the need for acute services. It also dramatically illustrates the variation in death rates from stroke between communities, already illustrated in younger stroke populations.3 Such findings have implications for the different focus of stroke prevention in different communities and the appropriate ways to manage stroke patients.
Incidence of Stroke
Sudlow and colleagues4 undertook a review of all population-based stroke registers up to the middle of 1994. Unfortunately, their study did not report total population incidence rates, either crude or adjusted to the European population. They did, however, demonstrate variations in the incidence rates in patients aged 45 to 84 years, ranging from 238 of 100 000 in Dijon to 627 of 100 000 in Novosibirsk, Russia. There is consistency with the low rates in Dijon in the study of Sudlow and colleagues and in the present study for all-age strokes. The advantage of the present study is that the biases introduced by comparing across time periods and using different methodologies are reduced. The 3 registers have all used similar registration methods with overlapping multiple sources of registration. There is no gold standard methodology presently in use for estimating the underascertainment of strokes in the centers. The application of capture-recapture techniques has been proposed for stroke to increase the accuracy of estimates, but its application to stroke registry data has not been undertaken in the present study because the methods are not validated for stroke.12 The variation in incidence rates may relate to variations in the prevalence of risk factors in the general populations, and there has been much debate around the “French paradox,” ie, why death rates from cardiovascular disease are lower in France.13 Life expectancy also varies in the 3 countries examined in the present study, with France being the longest at 78 years and south London being the shortest at 74.5 years. This may account for some of the variation seen. The relationship between incidence and case-fatality rates may also indicate ascertainment problems. For example, in the MONICA study of stroke at a younger age in Europe, certain countries had very high case-fatality rates, which, in the absence of case-severity data, indicate an underascertainment of milder strokes.3 In the present study, the low-incidence center did not have the highest case-fatality rate, and the London center had the intermediate incidence rate. Therefore, it would appear that these variations in risk and outcome are real. Another possible source of error when estimating incidence rates is the accuracy of the denominator population. In the 3 centers, the latest census data are used, although it is realized that there may be inaccuracies in them, but no alternative estimates are available to improve the estimation.
Explaining such differences in the risk of stroke requires detailed information on the prevalence of risk factors in the populations, along with quantification of the subtypes of stroke registered. The present study is unable to answer questions regarding the underlying stroke risk, which may vary around Europe. We are able to look at the pathological subtypes of stroke in the 3 populations. The higher levels of unclassified stroke in south London mean that differences in subtype and, hence, underlying risk factors cannot be assessed as satisfactorily as they could be, but the overall CT scan rates in the study are comparable to the more complete recent stroke registry studies. There did appear to be a higher rate of subarachnoid hemorrhage in London and lower rates of all subtypes other than infarction in Dijon. Hypertension is the major risk factor for stroke, and its diagnosis and control may be more effective in France, explaining not only our findings but the generally increased life expectancy in France.13
Variations in outcome after stroke are well documented. In the younger groups, the standardized mortality ratio for stroke varies significantly in Europe, being low in France.2 We have already shown in a hospital-based study that case-fatality rates vary significantly in Europe, with the highest risk of death or disability (after adjustment for case mix) being in 5 UK centers and the lowest being in Dijon, France.5 Unfortunately, it was not possible in these analyses to adjust for confounding factors, such as social class and ethnicity. These weaknesses apply to the current data, and the role of these confounders requires further investigation. In the UK population, the community is among the most deprived in Europe, with a 21% black population,9 and these factors may contribute to the reduced survival rates after adjustment for differences in age, sex, and subtype of stroke. Although we have adjusted for subtype in the case-fatality rates, this is only at a pathological classification level. We have not adjusted for other indicators of stroke severity or clinical classification, which may help to explain these differences.
Another possible explanation for these survival differences is the management of the acute-stroke episode. Data from the International Stroke Trial6 suggests that there are differences in survival between centers after control for confounding factors, with the UK centers again fairing poorly. As outlined in our European hospital-based study, it appears that there are lower intervention rates in the United Kingdom to correct for abnormal physiological parameters in the acute phase.5 In the United Kingdom, the defense of this practice is that there is no randomized trial evidence suggesting that correcting the abnormal parameter improves outcome. The contrary argument might be that if there is an abnormality, it should be corrected as a matter of routine support, eg, hydrating a patient, reducing glucose levels, and treating high temperatures. Further research into the management of these physiological parameters is under way and may provide clues as to what general supportive interventions are effective.
The present study provides current data on the immediate impact of stroke in Europe and is therefore useful when setting targets and developing preventive and management services for stroke. It also raises issues about the tailoring of preventive services to particular communities. The striking differences in survival require further work in terms of identifying confounding factors that may explain such differences and developing new acute-stroke management packages to reduce these differences.
In England, we wish to thank the field workers for their help with data collection and all general practitioners and hospital staff in the area of the South London Stroke Register who have supported the study. The study was funded by the Northern and Yorkshire Research and Development Program, a Glaxo Wellcome R&D grant, and the European Union BIOMED 11 Program, which funded the collaboration travel costs. In Germany, we wish to thank fellow participants of the Erlangen Stroke Project and, in particular, Drs Fahlbush, Hahn, Daniel, Kalden, Sterzl, Hohenberger, E. Lang, W. Lang, Gassman, Ritter von Stockert, Scheider, Szur, Dickschas, and Schupp in the hospitals and Drs Lederer, Lohwasser, and Zeller in the community and all the general practitioners of Erlangen. The study was funded by the Bavarian Ministry of Health and Social Affairs, the German Federal Ministry of Research, and the German Federal Ministry of Health. In Dijon, France, we wish to thank the INSERM, the Directorate Générale de la Santé, the University of Burgundy, the Regional Council of Burgundy, and the University Hospital of Dijon. Special thanks are extended to the collaborators of the Dijon Stroke Register. In all centers, we wish to thank the patients and their families for their participation.
- Received January 10, 2000.
- Revision received May 26, 2000.
- Accepted May 26, 2000.
- Copyright © 2000 by American Heart Association
Holland WW. European Community Atlas of Avoidable Death. Vol 1. 2nd ed. Oxford, UK: Oxford University Press; 1991.
Thorvaldsen P, Asplund K, Kuulasmaa K, Rajakangas AM, Schvoll M. Stroke incidence, case fatality and mortality in the WHO MONICA Project. Stroke. 1995;26:361–367.
Sudlow CLM, Warlow CP, for the International Stroke Incidence Collaboration. Comparable studies of the incidence of stroke and its pathological types: results from an international collaboration. Stroke. 1997;28:491–499.
Wolfe CDA, Tilling K, Beech R, Rudd AG, for the European BIOMED Study of Stroke Care Group. Variations in case fatality and dependency from stroke in western and central Europe. Stroke.. 1999;30:350–356.
Signorini DF, Weir NU, Sandercock PAG. Variations in clinical outcome by country in the International Stroke Trial: implications for international randomised controlled trials. Cerebrovasc Dis. 1999;9(suppl 1):36. Abstract.
Lemesle M, Milan C, Faivre J, Moreau T, Giroud M, Dumas R. Incidence trends of ischemic stroke and transient ischemic attacks in a well-defined French population from 1985 through 1994. Stroke. 1999;30:371–377.
Kolominsky-Rabas PL, Sarti C, Heuschmann PU, Graf C, Siemonsen S, Neundoerfer B, Katalinic A, Lang E, Gassmann KG, von Stockert TR. A prospective community-based study in stroke in Germany: the Erlangen Stroke Project (ESPro): incidence and case fatality at 1, 3, and 12 months. Stroke. 1998;29:2501–2506.
Stewart JA, Dundas R, Howard RS, Rudd AG, Wolfe CDA. Ethnic differences in incidence of stroke: prospective study with stroke register. BMJ. 1999;318:967–971.
Barer DH, Ellul J, on behalf of the ESDB Collaborators. The European Stroke DataBase (ESDB) Project: developing a common clinical language for stroke. Eur J Neurol. 1995;2(suppl 2):11–12. Abstract.
Waterhouse J, ed. Cancer Incidence in Five Continents. Lyon, France: IARC; 1976.
Tilling K, Sterne JA. Capture-recapture models including covariate effects. Am J Epidemiol. 1999;149:392–400.
Law M, Wald N. Why heart disease mortality is low in France: the time lag explanation. BMJ. 1999;318:1471–1476.