Donate Help Contact The AHA Sign In Home
American Heart Association
Stroke
Search: search_blue_button Advanced Search
This Article
Right arrow Abstract Freely available
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Right arrow Citation Map
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow Request Permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Sudlow, C.L.M.
Right arrow Articles by Warlow, C.P.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Sudlow, C.L.M.
Right arrow Articles by Warlow, C.P.

(Stroke. 1996;27:550-558.)
© 1996 American Heart Association, Inc.


Articles

Comparing Stroke Incidence Worldwide

What Makes Studies Comparable?

C.L.M. Sudlow, BM, BCh C.P. Warlow, MD

From the Department of Clinical Neurosciences, University of Edinburgh (Scotland).


*    Abstract
up arrowTop
*Abstract
down arrowIntroduction
down arrowWhy Measure Stroke Incidence?...
down arrowMeasuring Stroke Incidence: A...
down arrowMeasuring and Comparing...
down arrowConclusions
down arrowReferences
 
Background Comparing stroke rates in different parts of the world and at different points in time may increase our understanding of the disease. Comparisons are only meaningful if they are based on studies that use similar definitions, methods, and data presentation.

Summary of Review We discuss the criteria that make such studies comparable, drawing on the experiences of recent studies performed around the world. If only those studies that fulfill the proposed criteria for comparison are considered, comparable data do not exist for vast areas of the world, including Africa, Asia, and South America. The importance of complete, community-based case ascertainment, including strokes managed outside the hospital, is emphasized. An approach for measuring and comparing the incidence of the pathological types of stroke (cerebral infarction, primary intracerebral hemorrhage, and subarachnoid hemorrhage) and subtypes of cerebral infarction is suggested.

Conclusions The "ideal" stroke incidence study does not exist, but studies closely approaching it will reveal the most reliable and comparable results. There is a need for further studies to fill the gaps in our knowledge of the worldwide incidence of stroke, particularly for developing countries.


Key Words: epidemiology • incidence • stroke


*    Introduction
up arrowTop
up arrowAbstract
*Introduction
down arrowWhy Measure Stroke Incidence?...
down arrowMeasuring Stroke Incidence: A...
down arrowMeasuring and Comparing...
down arrowConclusions
down arrowReferences
 
Comparing stroke incidence in different countries and observing changes in incidence with time in a particular place may increase our understanding of the disease. To be comparable, studies of stroke incidence must use the same definitions, methods, and mode of data presentation. However well-conducted a study may be, unless it conforms in each of these three areas, comparison with other studies is not possible. The ideal approach is to develop standard criteria for these components of a stroke incidence study before establishing a series of comparable studies based on the prespecified criteria.

The first attempt to do this on an international scale is the World Health Organization (WHO) MONICA (Monitoring Trends and Determinants in Cardiovascular Disease) study, primarily designed to monitor trends in the incidence of myocardial infarction and stroke, together with the prevalence of various risk factors, in a number of different countries.1 A comparison of incidence and case-fatality rates from those countries monitoring stroke has recently been published.2 Although clearly a landmark study in international cardiovascular epidemiology, the MONICA study has limitations. The centers monitoring stroke are mainly European, extending into Eastern Europe, Russia, and China. This is the most extensive comparison yet, but large areas of the world were not included. In addition, a limited age range was covered, with most centers monitoring stroke up to age 65 years and none routinely collecting data on those older than 75 years, in whom the incidence (and caseload, at least in the areas studied) of stroke is highest. Finally, the MONICA study concentrates on total strokes and is not specifically designed to examine the incidence of pathological types, although some of the centers do have high enough rates of accurate pathological diagnosis for this.

A large number of independent studies have also measured stroke incidence. Most of them did not have comparative incidence as their sole or even main aim, and only a few are suitable for this purpose. In 1987, Malmgren and colleagues3 reviewed studies of stroke incidence from around the world. Of 65 studies, only nine met a set of standard criteria sufficiently to be considered comparable with each other.4 5 6 7 8 9 10 11 12 13 14 Since that time, several further studies approaching the original criteria of Malmgren et al have been published.15 16 17 18 19 20 21 22 23 24 25 26 27 The area of the world covered by these studies is still very limited, and reliable information remains sparse for nonwhite populations. Despite this deficiency, there is a broadening experience of studying the incidence of stroke in different environments and within different healthcare systems. Thus, an increasing number of studies may provide comparable incidence data to add to our knowledge of stroke incidence worldwide. It is important that these valuable data are not ignored simply because they are from studies that were not established as part of an international effort.

This article aims to extend and refine the original criteria of Malmgren et al in the light of further experiences of measuring stroke incidence, the identification of new problems and possible solutions, and the wider availability of imaging technology, which allows accurate pathological diagnosis. It discusses a set of standards for definitions, methods, and data presentation that individual studies of stroke incidence should fulfill to be considered comparable with one another. It does not attempt to cover all the practicalities of establishing such studies but highlights several important areas that may make comparisons difficult. It is based not only on an extensive review of the literature but also on a series of visits around the world in 1994 to meet with investigators in centers where stroke incidence has been studied for detailed discussions of study methods in the context of the particular local healthcare systems. Considering the successes and problems encountered should increase the ease with which further reliable studies can be established. We hope that this article will encourage those establishing stroke registers to consider the factors that will allow their data to contribute to the accumulating worldwide knowledge of stroke incidence. This is particularly true for the many developing countries for which reliable information simply does not exist.

The main concern of this article is to compare incidence studies in different parts of the world. Many of the issues also relate to studies of time trends in incidence. These may be assessed by continuous registration of strokes or by comparing the incidence measured during two or more discrete periods, separated by a number of years. In either case, internal consistency of methodology is essential so that any change detected is real. For comparison between studies in different locations, external consistency is also necessary.

The wider availability of CT scanning in a number of countries since the 1980s means that many more studies than were published at the time of the review of Malmgren et al can now make a reliable diagnosis of pathological type of stroke during life for the majority of cases. This may allow, for the first time, a reliable international comparison of the incidence of the individual pathological types of stroke (cerebral infarction, primary intracerebral hemorrhage, and subarachnoid hemorrhage). We include a discussion of the criteria that should make this possible.


*    Why Measure Stroke Incidence?
up arrowTop
up arrowAbstract
up arrowIntroduction
*Why Measure Stroke Incidence?...
down arrowMeasuring Stroke Incidence: A...
down arrowMeasuring and Comparing...
down arrowConclusions
down arrowReferences
 
Limitations of Mortality Statistics
Mortality is the most readily available variable for comparing stroke rates between countries and monitoring changes with time, so why put enormous efforts into measuring incidence? Official mortality data may be convenient, but they rely on the accuracy of death certificates, which is questionable,28 29 and provide information about fatal cases of stroke only, telling us nothing about the significant proportion of stroke patients who survive, many with major disability. Furthermore, without knowledge of the accuracy of diagnostic methods they cannot tell us about the distribution of the different pathological types of stroke. Several possible explanations exist for the substantial differences in stroke mortality between countries and its trends with time. For example, the apparent recent decline in stroke mortality in many Western countries30 may reflect changes in diagnostic fashions and coding procedures or a real decline, explained by changes in incidence and/or case fatality. The latter may depend on the distribution of pathological types, disease severity, and management. A clearer, more complete picture is only obtained by examining incidence and case-fatality rates separately. If accurate, these measures allow more valid comparisons between countries, may yield data that can be used to follow trends with time, and assist in healthcare planning within communities.

Scope of Stroke Incidence Studies
Another advantage of measuring incidence of stroke is the potential of community-based studies to answer additional questions based on an unbiased sample of incident cases (Table 1Down). These may include follow-up for outcome in terms of case fatality, recurrence, and disability; case-control studies of various risk factors; and assessment of stroke services. Expanding the yield beyond the mere identification of cases helps to justify the intensive effort put into the case-finding process. Careful study design is crucial to ensure that the identified aims are achieved without the collection of excess, useless data, which simply create extra work for the investigators and may compromise the quality of the study. As for incidence, the measurement of other variables should be standardized for comparison. For case fatality, it is suggested that the MONICA standard of 28-day case fatality be widely adopted.


View this table:
[in this window]
[in a new window]
 
Table 1. Possible Uses for a Community-Based Stroke Register


*    Measuring Stroke Incidence: A Discussion of Standard Criteria for a Comparable Study (Table 2Down)
up arrowTop
up arrowAbstract
up arrowIntroduction
up arrowWhy Measure Stroke Incidence?...
*Measuring Stroke Incidence: A...
down arrowMeasuring and Comparing...
down arrowConclusions
down arrowReferences
 
Standard Definitions
WHO Definition of Stroke
The WHO definition of stroke, "rapidly developing clinical signs of focal (at times global) disturbance of cerebral function, lasting more than 24 hours or leading to death with no apparent cause other than that of vascular origin,"31 is the broadly accepted standard. Some care is required with its interpretation. First, according to the definition, presentations of subarachnoid hemorrhage with sudden-onset headache, meningism, and no focal or global disturbance of cerebral function should be excluded. However, the definition is generally considered to include all cases of cerebral infarction, primary intracerebral hemorrhage, and subarachnoid hemorrhage. In practice, most investigators include all subarachnoid hemorrhages, and reports of studies should make this clear.


View this table:
[in this window]
[in a new window]
 
Table 2. Core Criteria for a Comparable Study of Stroke Incidence

Second, it is important to rely on the clinical criteria in the definition, with the distinction between stroke and transient ischemic attack (TIA) based on a duration of more (stroke) or less (TIA) than 24 hours. Advances in imaging technology have made us aware that a significant proportion of TIAs are associated with an appropriately located infarct visible on CT or MRI scan. Controversy exists regarding how these "cerebral infarcts with transient signs" should be classified and whether the presence of a visible infarct associated with a TIA alters prognosis.32 33 The availability and sensitivity of imaging technology vary enormously with place and time. Abandoning the clinical distinction between TIA and stroke for one based on imaging would create spurious variability between studies and compound the difficulties in comparing their results.

A similar problem is posed by "silent infarcts"—areas of presumed infarction visible radiologically with no associated symptomatic episode. Again, it would be inappropriate to include these as incident strokes, not only for the reasons outlined in the preceding paragraph but also because, in the absence of symptoms, it is impossible to know when the event leading to the abnormal scan appearance occurred.

First-Ever-in-a-Lifetime Strokes
If we wish comparisons to give us pointers to important determinants of disease etiology, first-ever-in-a-lifetime strokes are the preferred numerator because they most closely reflect these factors. Some studies base their incidence estimates on the number of first stroke events occurring within the study period.31 34 These estimates do not exclude those who had a stroke before the start of the study, thereby including a number of recurrent strokes that are not true incident cases. Others count all strokes within the study period.35 Because there is no consistent relationship between first-ever-in-a-lifetime strokes, first-event strokes, and all strokes,3 it is meaningless to compare incidence measurements based on different definitions (FigureDown). Including recurrent strokes in the numerator makes sense for assessing the total burden of stroke for local healthcare planning. Thus, studies examining stroke services may also register recurrent strokes but will need available data on first-ever-in-a-lifetime strokes for any etiologic comparisons.



View larger version (14K):
[in this window]
[in a new window]
 
Figure 1. Example of incidence measurements based on different definitions: first-ever-in-a-lifetime strokes (S3 and S4); first-event strokes (S2, S3, and S4); all strokes (S2, S3, S4, and S5); recurrent strokes (S2 and S5); and first-ever-in-a-lifetime stroke occurring prior to study (S1).51

In addition, because many intercurrent illnesses can cause neurological deterioration, defining a recurrent stroke is often difficult in clinical practice, making first-ever-in-a-lifetime stroke a more accurate comparator. Its accuracy requires a careful history and review of all available past records for each potential case to ensure that subjects with previous strokes are not included as incident cases.

Standard Methods
Complete, Community-Based Case Ascertainment
Malmgren et al3 found many studies measuring incidence in hospital-based populations. These never represent the whole community because there will always be some cases not admitted to the hospital (Table 3Down). Furthermore, they cannot be used for comparisons of place or time: hospitalization rates for stroke vary considerably between and even within countries and are likely to vary with time as management strategies change and new treatments for acute stroke emerge.13 19 36 37 In the Oxfordshire Community Stroke Project, which was performed in the 1980s, 40% of incident stroke patients were not admitted to the hospital.36 This contrasts with the situation in Swedish studies from the same period and later, which demonstrate admission rates of 95% or more.4 13 38 There is also evidence that hospital-based stroke registers are biased toward younger patients with more severe strokes.36 The authors of some community-based studies of stroke incidence were surprised by the significant proportion of cases they found to be managed outside the hospital (eg, 15% in both Warsaw, Poland, and Umbria, Italy19 24 ), demonstrating that such cases must be rigorously searched for without simply assuming that all will be admitted to the hospital. It is because they are entirely hospital based that many of the studies from developing countries are unsuitable for international comparisons.39 40 41


View this table:
[in this window]
[in a new window]
 
Table 3. Reasons Why Stroke Patients May Not Be Admitted to the Hospital

True community-based studies must therefore use several overlapping sources of information to ensure complete case ascertainment. Possible sources are shown in Table 4Down. The local healthcare system will determine which of these are applicable.


View this table:
[in this window]
[in a new window]
 
Table 4. Possible Sources of Case Ascertainment in Community-Based Studies of Stroke Incidence

Sources must be surveyed with adequate frequency, covering a broad spectrum of possible presentations of stroke so that cases are not missed. For example, labels such as "collapse," "dizziness," "headache," "fits," or "confusion" may reveal stroke cases. The importance of assessing patients labeled as TIA cases (from both hospital and community) as potential stroke cases is demonstrated by results from the Oxfordshire Community Stroke Project: 7% of incident strokes were initially referred to the investigators as TIAs but in fact had symptoms lasting longer than 24 hours.14

Sources outside the hospital, in particular primary healthcare physicians and nursing institutions, should be involved in the study and contacted regularly to ensure complete referral of community cases. Hospital discharge lists and computer-linked records schemes provide a useful check for patients not identified by admission and ward checks. Reviewing death certificates is the only way of finding patients who die without a referral being made. Their importance, like that of other sources, will depend on local practice and culture. Most studies find death certificates an important contributor of cases. However, a recent study in Okinawa, Japan, which examined death certificates in only one subpopulation of the total, found an insignificant number of cases from this source not already ascertained from others.42 Gathering whatever information is available retrospectively from relatives, attending physicians, medical records, and/or autopsy findings is required before possible death certificate cases can be confirmed (or refuted) as first-ever, WHO-defined strokes.

Assessing Completeness of Case Ascertainment
It would be helpful to have some quantitative way of assessing the completeness of case ascertainment. One approach is to perform a cross-sectional survey of the population, or a sample of it, at intervals (usually every one or few years) during the study to find out whether cases have been missed.6 35 43 Because many stroke cases are among the elderly and disabled, who may be housebound, labor-intensive and therefore expensive survey methods are required to ensure that such cases are not excluded. Cross-sectional surveys may also miss those minor strokes that patients tend to forget with time and that may only be included at all if registered at the time of presentation.

Studies that run for a period of several years can assess their own case-finding efficiency. For example, they may note the number of stroke cases found that are recurrences with a first-ever-in-a-lifetime stroke missed at an earlier stage during the study period. In addition, a surprisingly high or low case-fatality rate may alert investigators to the possibility that either fatal or mild, out-of-hospital cases are being missed. Participating centers in the WHO MONICA study are subject to a number of quality-control procedures to assess the validity of their data. Checks for case ascertainment require centers with case-fatality or out-of-hospital case referral rates falling outside certain arbitrarily determined limits to reexamine their data.44

There is currently no perfect, objective indicator of completeness of case ascertainment. Subjective judgment based on knowledge of the appropriateness of case-finding methods to the study population may be as reliable and less likely to provide false reassurance. No one source of cases will be either infallible or all-encompassing. Multiple overlaps between sources may be regarded as an indication of good case-finding methods as long as multiply referred cases are not registered more than once.10 However, it would be wasteful to concentrate resources on a source with negligible yield. We feel that investigators should identify all possible sources in their location and justify the exclusion of any that are not used. For example, quality-control checks of the WHO MONICA study in northern Sweden, where hospital admission rates are very high, revealed that an intensified effort by general practitioners over a 1-month period to refer all cases did not significantly increase the incidence rates.38 The investigators therefore concentrate on hospital cases and death certificates for case finding.

An interesting possibility is the capture-recapture technique, initially developed for estimating the size of animal populations and now being used in some areas of human disease epidemiology.45 46 It recognizes that no survey has complete coverage and allows an estimate to be made of the number of cases not ascertained, with the use of information from two or more overlapping sources of cases of the disease being counted. Complex adaptations of the technique are required in humans to adjust for the nonindependence of sources and variability between subgroups.45 46 47 The information required to attempt such an analysis is not available for any of the stroke incidence studies of which we are aware, except perhaps the Auckland study.10 11 Until this technique has been developed in stroke epidemiology, we must continue to rely on those methods already in use.

However complete we judge case-finding methods to be, we can only expect to find those cases that are presented to the sources surveyed. The Oxfordshire Community Stroke Project found that of those patients whose stroke was preceded by a TIA, only about half had sought medical attention for their TIA.48 Similarly, some persons with mild strokes might never seek medical attention. This is particularly likely among the elderly, who may attribute symptoms of a stroke to age or some age-associated disease such as osteoarthritis. Any individual study may thus underestimate stroke incidence, and this effect may vary between studies because of differences between study populations in the interpretation of symptoms, perception of their severity, and accessibility and cost of health care. These factors are not easy to control or measure, but their potential contribution to differences in incidence rates between studies, particularly in regard to mild strokes, cannot be ignored.

Importance of Prospective Study Design
Stroke is a clinical diagnosis, making accurate clinical assessment by an appropriately qualified specialist an essential part of any incidence study. There is no "gold standard" diagnostic method superior to clinical judgment. For example, CT or MRI scanning cannot necessarily distinguish between a stroke and the transient focal weakness that may follow a seizure. The rate of error, as judged by CT scanning, in the clinical diagnosis of stroke versus nonstroke varies from 1.5% to 16%19 49 50 51 52 53 and appears to be dependent on the degree of skill and training of the assessor. In community-based stroke registers, in which patients are assessed soon after onset of symptoms by a neurologist or stroke specialist with the specific intent of monitoring stroke incidence, the error rate is below 2%.19 49 Only truly prospective studies that pursue cases as they occur—termed "hot pursuit"—provide this degree of accuracy.

Retrospective collection of information ("cold pursuit") is more liable to cause blurring of the margin between TIAs and strokes than is the hot-pursuit method. One of the reasons for increased accuracy with prospective design is that a clinician managing an individual patient has a different agenda from a study epidemiologist or neurologist measuring incidence in the population. The clinician recognizes that making a precise distinction between a TIA lasting 20 hours and a minor stroke with resolution of symptoms by 30 hours is unnecessary because it will affect neither prognosis nor management. Because this information is not considered relevant, it may either be documented inaccurately or not recorded at all. For the investigator wishing to compare different studies of stroke incidence the distinction, albeit artificial, is important. A further example is the assessment of a patient's history. Previous strokes, minor ones in particular, are much more likely to be ascertained prospectively by a study neurologist or epidemiologist aware of the importance of including only first-ever-in-a-lifetime strokes in the incidence figures.

Many centers in the WHO MONICA study use the cold-pursuit approach, with review of records after hospital discharge and death certificates as the main sources of cases.1 This will provide valuable information about incidence trends as long as diagnostic methods remain constant. Cold-pursuit methods have the advantage of consuming considerably less time and resources than hot-pursuit ones. However, it may be invalid to compare studies using this method with each other or with those using the hot-pursuit method without knowing how overall case ascertainment is affected in any particular center. Centers best suited to cold pursuit will require a very reliable records system and should be able to demonstrate that hot-pursuit methods do not materially alter case ascertainment.

Studies based entirely on retrospective reviews of records depend on complete record ascertainment for the whole community and accurate documentation. These are rarely demonstrated, but exceptions may exist. For example, the Mayo Clinic has a records system that appears to be unique in its complete, reliably documented coverage of the whole Rochester community. In this particular case, as long as the information is appropriately accessed, retrospective methods may be considered as reliable as prospective methods.54

The Study Population
The best populations are well defined (usually, but not always, on a geographic basis) and stable, with limited in- and out-migration. Investigators cannot manipulate populations: they simply have to be opportunistic in selecting what is available. Relatively isolated populations, for example, the inhabitants of Japan's Okinawa islands42 or the Valle d'Aosta in northern Italy,20 are particularly suitable. Such situations limit the "cross-boundary effect" in which cases are either missed, because they occur while a resident is outside the study area, or are included erroneously when they occur in visitors who are not permanent residents. Unstable populations, with frequent permanent changes of residence into and out of the study area, are less reliable. However, the more isolated a population is, the less representative of its country it may be because of the effects of such factors as a restricted gene pool and special dietary habits. A sensibly selected population takes these various opposing factors into account. Populations composed of healthy self-selected volunteers or excluding certain disease states (for example, the Framingham Study55 ) are not really suitable for a comparable incidence study.

The size of the population is important because the larger the population, the more strokes are likely to occur within the monitoring period, and the confidence interval around estimates of incidence is dependent on the number of strokes contributing to the estimate.56 There is no perfect population size. Because most strokes occur in the elderly, populations with a high proportion of older people will yield more strokes for the same total population size than will younger populations. Monitoring too large a population is difficult without extensive resources and may lead to undercounting. Some investigators increase the number of stroke cases by monitoring the same population for several years (but ideally not more than 5; see below) for a single estimate based on the number of person-years observed.15 16 17 18 19 24 57 It is suggested that below approximately 250 cases the confidence interval becomes unacceptably wide. For populations with a Western European type of structure, this will require approximately 100 000 person-years of observation. Estimates of incidence from many cohort studies have limited precision because of small population size.55 58 59 It is emphasized that although they provide an idea of the statistical precision of the estimate, confidence intervals do not allow for methodological flaws such as incomplete case ascertainment.

The accuracy of determining the denominator must also be considered. Investing resources in obsessional case finding is inappropriate unless reliable and up-to-date information on the age and sex structure of the population is also available. For an appropriately large population (see above), minor inaccuracies have little effect on incidence rates but major ones must be avoided. Most studies use official census statistics, which are usually (but not always60 ) reliable in developed countries. Their unreliability in developing countries makes them inappropriate for estimating denominator size. Unfortunately, few populations will have census information for every year. Studies performed during the census year can use the census data as they stand.11 25 Those performed between censuses require an estimate of the denominator. Official updates of the census rely on information about births, deaths, and in- and out-migration.4 13 Because these are often inaccurate (Sweden, with accurate updates fortnightly, is an exception), investigators may prefer to rely on interpolation of census data, based on prior trends.15 The error can be estimated when subsequent census statistics become available. In general, it increases with the time from the last census. In populations in which the vast majority of people are registered with a general practitioner, the denominator may be based on the age/sex registers of general practices.14 One potential problem here is the existence of "ghosts": patients who have died or moved but who have not yet been removed from the practice list.

A further consideration is that incidence measurements should really be based on the population at risk. This implies that we should remove from the denominator any prevalent cases, ie, those who have already suffered a stroke. In practice, the information is simply not available to the vast majority of investigators so that adjustments for prevalence are not made. We can assume that the effect in the younger age groups in whom prevalence is low would be minimal. In older age groups the effect in those populations where it has been assessed appears to be significant (R. Brown, unpublished data, 1995).

Standard Data Presentation
Data collected over complete years avoids confounding from the possible effect of seasonal variations in incidence documented in several community-based studies.61 62 63 From the Rochester group's results, it appears that incidence rates can vary secularly over periods of approximately 5 years.5 This implies that rates should not be averaged over time intervals longer than 5 years and that comparisons between different studies should be confined to similar points in time.

Most stroke epidemiologists use mid-decade 10-year age bands (eg, 45 to 54, 55 to 64 years). It seems reasonable to retain this convention in publications. However, it is helpful to collect data for both numerator and denominator in narrower 5-year bands for two reasons. First, in some other areas of epidemiology the convention is to use 10-year age bands spanning a complete decade (eg, 50 to 59, 60 to 69 years).64 The availability of 5-year age bands allows comparisons between studies presenting data by nonmatching 10-year age bands, including those investigating the incidence of other diseases. Second, the distribution of ages within a 10-year age band may vary considerably between populations. Incidence rates in different studies can be compared either by looking at age-specific rates for a particular age band or by comparing age-standardized rates. The use of narrower age bands (5 rather than 10 years) reduces the effect on age-standardized rates of variability of age distribution within each age band, especially that occurring within the open-ended upper age group. Differences in incidence rates between populations may otherwise simply reflect a different distribution of ages within each age band. This effect was regarded by one reviewer as a potential explanation for the variation in incidence rates with time in the Rochester series.65 Thus, while it would be cumbersome to include rates for 5-year age groups in publications, it is helpful to have them available for comparisons with other studies and time trend studies. The population chosen for age standardization should be appropriate to the study populations. A comparison of stroke incidence rates among countries in Europe might use Segi's notional European population.66 Segi's world population would be more appropriate for a true worldwide comparison. Because of potential differences between the sexes, rates should be sex- as well as age-standardized. Age-specific rates should be presented separately for men and women.


*    Measuring and Comparing Incidence of Pathological Types of Stroke
up arrowTop
up arrowAbstract
up arrowIntroduction
up arrowWhy Measure Stroke Incidence?...
up arrowMeasuring Stroke Incidence: A...
*Measuring and Comparing...
down arrowConclusions
down arrowReferences
 
Main Pathological Types
The heterogeneity of stroke is a challenge for clinical epidemiologists. The standard WHO definition31 encompasses cerebral infarction, primary intracerebral hemorrhage, and subarachnoid hemorrhage. Our ability to distinguish the first two pathological types without neuroimaging (usually CT scanning, occasionally MRI) or, in fatal cases, autopsy, is limited.50 It is therefore not possible to draw firm conclusions about the incidence rates of pathological types of stroke in studies using mainly clinical criteria. In the Rochester series, the introduction of CT scanning caused an apparent increase in the rate of intracerebral hemorrhage because of the increased detection rate for small hemorrhages that would previously have been classified as infarcts.67 For many years, particularly high rates of primary intracerebral hemorrhage (22% to 36%) have been reported from countries in the Far East. These reports are from community-based studies with low CT scan rates,6 43 cohort studies with very small study populations,59 or hospital series that have high CT rates but that cannot be assumed to be representative of the whole community.68 69 Newer studies with both high rates of CT scanning and more complete case ascertainment42 are helping to resolve this issue, but there is still a lack of reliable, community-based information on pathological types from this part of the world. It is only when we have confirmation that the apparent excess of cerebral hemorrhages is real that searching for potential causes of this difference will be justified.

Definitions must be standardized for comparisons. We recommend the scheme used in the Oxfordshire Community Stroke Project.57 Appropriately timed CT scan (see below) or autopsy is required to distinguish infarct from intracerebral hemorrhage. The definition of subarachnoid hemorrhage is based on a typical clinical history with CT or lumbar puncture evidence of subarachnoid blood or angiographic demonstration of a source of bleeding. The distribution of pathologies can only be estimated accurately if the proportion with undetermined pathology is small. The problem of falling autopsy rates has been partly overcome in the Dijon Stroke Registry by postmortem CT scans (M. Giroud, unpublished data, 1994). It will never be possible to assign a definite pathological type to every incident case, but an acceptably accurate type can be ascribed to a proportion of cases without CT scan or autopsy. In line with the Oxfordshire Community Stroke Project, we consider it acceptable to use a well-validated clinical scoring system (such as the Guy's Hospital or Siriraj Score) for pathological diagnosis in those cases without CT scan or autopsy, provided the score is at one or the other extreme, giving a diagnostic accuracy for hemorrhage or infarct of 90% or more.70 71 72 For management of individual patients it is often important to be 100% certain about the distinction between infarct and hemorrhage, leading to criticism of these scores in clinical practice.73 However, it seems reasonable to use them for epidemiological studies as long as there is no consistent bias in one direction or the other. Two community-based studies have used clinical scores in combination with high (approximately 70%) CT rates in this way.19 57 Several others have high enough rates of CT scanning (above an arbitrarily selected level of approximately 70%) and autopsy diagnoses to make them suitable for examination of pathological types.16 20 21 23 24

The timing of neuroimaging is important. Several studies consider a CT scan within about 30 days of stroke to be reliable in distinguishing infarct from intracerebral hemorrhage.19 23 57 MR scanning may continue to be helpful after this stage,23 but its availability is often limited. Although it has been demonstrated that a number of small primary intracerebral hemorrhages may be indistinguishable from infarcts on CT scan as early as 2 weeks after onset,74 studies of serially scanned hemorrhages are not large enough to tell us what proportion of hemorrhages we may be misinterpreting as infarcts by performing scans after 2 weeks.75 The 30-day cutoff may still be reasonable for epidemiological purposes, but we do not know. Many centers now scan most cases very early, within 24 hours of onset. However, the issue is important for those countries with many stroke patients managed outside the hospital, with limited scanning facilities and long prescan waiting times.

With early scanning there is concern about misdiagnosis of some cases of hemorrhagic transformation of infarct as primary hemorrhage. There is increasing evidence that such transformation may occur very early, even within hours of stroke onset.76 For the moment this is another unresolved issue but one of which we need to be aware.

The ability of increasing numbers of community-based studies to provide accurate information on pathological types will enhance geographic comparisons. We will be able not only to compare the incidence of these types between countries but also to investigate more closely the difference in incidence between men and women. While it appears that the incidence of myocardial infarction is four to five times greater in men than in women, the incidence of total stroke varies much less between the sexes.77 One reason for this might be that greater differences are hidden until the pathological types are examined separately.

Subtypes of Cerebral Infarction
Cerebral infarction has a number of possible etiologies with differing patterns of clinical presentation and outcome. There is no generally accepted way of classifying these subtypes. We favor the classification developed by Bamford et al,78 dividing cerebral infarction on the basis of simple clinical criteria into four subtypes: total anterior circulation infarction, partial anterior circulation infarction, lacunar infarction, and posterior circulation infarction. This classification is reasonably predictive of etiology,79 prognosis,78 and the size and site of cerebral infarction on CT scanning80 and has reasonable interobserver agreement.81 Furthermore, it is of particular value for epidemiological comparisons because, unlike some classification systems,82 83 it does not require a large number of often invasive investigations. Even in centers with easy access to such investigations, there remain many (40% to 50%) infarcts of undetermined cause. Meticulous searching for a cause may be clinically appropriate in individual cases, but for epidemiological studies the exclusion of a large number of patients from assignment of specific subtypes would make the overall analysis an unrepresentative one.

A few studies fulfilling the core criteria for comparability have been able to apply this classification system to their cases of cerebral infarction.19 78 84 Others have been able to at least define lacunar infarction as a separate group16 24 with the use of the same clinical criteria.


*    Conclusions
up arrowTop
up arrowAbstract
up arrowIntroduction
up arrowWhy Measure Stroke Incidence?...
up arrowMeasuring Stroke Incidence: A...
up arrowMeasuring and Comparing...
*Conclusions
down arrowReferences
 
The above discussion is an attempt to create a framework for establishing future studies and for deciding which of the incidence studies performed to date are suitable for international comparisons. The worldwide visits that contributed so much to its content also led to the formation of a collaborative group of comparable studies of stroke incidence. Both qualitative and quantitative data are being collected and will be presented separately.

Table 2Up summarizes the major standard criteria that make a study of stroke incidence comparable with others. The ideal study does not exist in reality. However, to make sense of geographic and secular comparisons, we should recognize that the studies that come close to the ideal are the most likely to provide reliable and comparable results. In comparing studies that use standard definitions and make a concerted effort to search for out-of-hospital strokes, the major source of potential error is still likely to be variable case finding. Methods must vary between locations because of differences in healthcare systems. Comparing the relative efficiency of these methods is difficult. The possibility of comparing incidence of accurately assigned pathological types and subtypes of cerebral infarction is an interesting new prospect, based on advances in imaging technology that are becoming increasingly widely available.

Our aim should be to use the lessons from previous incidence studies to design even better ones. The challenge is not only to explain the results that we have but also to find out what we can about stroke incidence in Africa, Asia, South America, and other areas about which so little is known but where so many of the world's people live. In developing countries, we might expect that, as life expectancy increases and the high-risk behaviors (eg, smoking, diet) of the Western lifestyle are increasingly adopted, strokes will be an escalating problem. Governments and makers of health policy are in no position to plan for the prevention of diseases that are an unknown quantity.


*    Acknowledgments
 
The worldwide visits for discussions with international investigators were made possible by the kind hospitality of colleagues and friends in Sweden, Denmark, Poland, Italy, France, Japan, Australia, New Zealand, and the United States, with generous financial support from the Edinburgh Stroke Research Fund, the Guarantors of Brain, the Royal College of Physicians of Edinburgh Myre Sim Bequest, Lilly Industries Limited, Bayer plc, and Boehringer Ingelheim Limited. We thank Martin Dennis, Peter Sandercock, and Mark Sudlow for their helpful comments on the manuscript.


*    Footnotes
 
Reprint requests to Dr Cathie Sudlow, Registrar in Medical Neurology, Department of Clinical Neurosciences, University of Edinburgh, Western General Hospital, Crewe Rd, Edinburgh EH4 2XU, Scotland.

Received July 20, 1995; revision received December 19, 1995; accepted December 22, 1995.


*    References
up arrowTop
up arrowAbstract
up arrowIntroduction
up arrowWhy Measure Stroke Incidence?...
up arrowMeasuring Stroke Incidence: A...
up arrowMeasuring and Comparing...
up arrowConclusions
*References
 

  1. WHO MONICA Project Principal Investigators. The World Health Organisation MONICA Project (Monitoring Trends and Determinants in Cardiovascular Disease): a major international collaboration. J Clin Epidemiol. 1988;41:105-114. [Medline] [Order article via Infotrieve]
  2. Thorvaldsen P, Asplund K, Kuulasmaa K, Rajakangas A, Schroll M, for the WHO MONICA Project. Stroke incidence, case fatality, and mortality in the WHO MONICA Project. Stroke. 1995;26:361-367. [Abstract/Free Full Text]
  3. Malmgren R, Warlow C, Bamford J, Sandercock P. Geographical and secular trends in stroke incidence. Lancet. 1987;2:1196-1200. [Medline] [Order article via Infotrieve]
  4. Terent A. A prospective epidemiological survey of cerebrovascular disease in a Swedish community. Ups J Med Sci. 1979;84:235-246. [Medline] [Order article via Infotrieve]
  5. Garraway WM, Whisnant JP, Drury I. The continuing decline in the incidence of stroke. Mayo Clin Proc. 1983;58:520-523. [Medline] [Order article via Infotrieve]
  6. Tanaka H, Ueda Y, Date C, Baba T, Yamashita H, Hayashi M, Shoji H, Owada K, Baba K, Shibuya M, Koa T, Detels R. Incidence of stroke in Shibata, Japan: 1976-1978. Stroke. 1981;12:460-466. [Abstract/Free Full Text]
  7. Herman B, Schulte BP, van Luijk JH, Leyten AC, Frenken CW. Epidemiology of stroke in Tilberg, the Netherlands: the population-based stroke incidence register, I: introduction and preliminary results. Stroke. 1980;11:162-165. [Abstract/Free Full Text]
  8. Herman B, Leyten AC, van Luijk JH, Frenken CW, Op de Coul AA, Schulte BP. Epidemiology of stroke in Tilberg, the Netherlands: the population-based stroke incidence register, II: incidence, initial clinical picture and medical care, and three week case fatality. Stroke. 1982;13:629-634. [Abstract]
  9. Kotila M. Declining incidence and mortality of stroke? Stroke. 1984;15:255-259. [Abstract/Free Full Text]
  10. Bonita R, Beaglehole R, North JD. The long-term monitoring of cardiovascular disease: is it feasible? Community Health Stud. 1983;7:111-116. [Medline] [Order article via Infotrieve]
  11. Bonita R, Beaglehole R, North JD. Event, incidence and case fatality rates of cerebrovascular disease in Auckland, New Zealand. Am J Epidemiol. 1984;120:236-243. [Abstract/Free Full Text]
  12. Ashok PP, Radhakrishnan K, Sridharan R, el-Mangoush MA. Incidence and pattern of cerebrovascular diseases in Benghazi, Libya. J Neurol Neurosurg Psychiatry. 1986;49:519-523. [Abstract]
  13. Terent A. Increasing incidence of stroke among Swedish women. Stroke. 1988;19:598-603. [Abstract/Free Full Text]
  14. Sandercock PAG, Warlow CP, Price SM. Incidence of stroke in Oxfordshire: first year's experience: Oxfordshire Community Stroke Project. Br Med J. 1983;287:713-717.
  15. Broderick JP, Phillips SJ, Whisnant JP, O'Fallon WM, Bergstrahl EJ. Incidence rates of stroke in the eighties: the end of the decline in stroke? Stroke. 1989;20:577-582. [Abstract/Free Full Text]
  16. Giroud M, Milan C, Beuriat P, Gras P, Essayagh E, Arveux P, Dumas R. Incidence and survival rates during a two-year period of intracerebral and subarachnoid haemorrhages, cortical infarcts, lacunes and transient ischaemic attacks: the stroke registry of Dijon: 1985-1989. Int J Epidemiol. 1991;20:892-899. [Abstract/Free Full Text]
  17. Giroud M, Beuriat P, Vion P, A'Athis PH, Dusserre L, Dumas R. Stroke in a French prospective population study. Neuroepidemiology. 1989;8:97-104. [Medline] [Order article via Infotrieve]
  18. Giroud M, Gras P, Chadan N, Beuriat P. Milan C, Arveux P, Dumas R. Cerebral haemorrhage in a French prospective population study. J Neurol Neurosurg Psychiatry. 1991;54:595-598. [Abstract]
  19. Ricci S, Celani MG, La Rosa F, Vitali R, Duca E, Ferraguzzi R, Paolotti M, Seppoloni D, Caputo N, Chirulla C, Scaroni R, Signorini E. SEPIVAC: a community-based study of stroke incidence in Umbria, Italy. J Neurol Neurosurg Psychiatry. 1991;54:695-698. [Abstract]
  20. D'Alessandro G, Di Giovanni M, Roveyaz L, Ianizzi L, Compagnoni MP, Blanc S, Bottacchi E. Incidence and prognosis of stroke in the Valle d'Aosta, Italy: first-year results of a community-based study. Stroke. 1992;23:1712-1715. [Abstract/Free Full Text]
  21. Jorgensen HS, Plesner AM, Hubbe P, Larsen K. Marked increase of stroke incidence in men between 1972 and 1990 in Frederiksberg, Denmark. Stroke. 1992;23:1701-1704. [Abstract/Free Full Text]
  22. Anderson CS, Jamrozik KD, Burvill PW, Chakera TM, Johnson GA, Stewart-Wynne EG. Ascertaining the true incidence of stroke: experience from the Perth Community Stroke Study, 1989-1990. Med J Aust. 1993;158:80-84. [Medline] [Order article via Infotrieve]
  23. Anderson CS, Jamrozik KD, Burvill PW, Chakera TM, Johnson GA, Stewart-Wynne EG. Determining the incidence of different subtypes of stroke: results from the Perth Community Stroke Study, 1989-1990. Med J Aust. 1993;158:85-89. [Medline] [Order article via Infotrieve]
  24. Czlonkowska A, Ryglewicz D, Weissbein T, Baranska-Gieruszczak M, Hier DB. A prospective community-based study of stroke in Warsaw, Poland. Stroke. 1994;25:547-551. [Abstract]
  25. Bonita R, Broad JB, Beaglehole R. Changes in stroke incidence and case-fatality in Auckland, New Zealand, 1981-91. Lancet. 1993;342:1470-1473. [Medline] [Order article via Infotrieve]
  26. Feigin VL, Wiebers DO, Nikitin YP, O'Fallon WM, Whisnant JP. Stroke epidemiology in Novosibirsk, Russia: a population-based study. Mayo Clin Proc. 1995;70:847-852. [Medline] [Order article via Infotrieve]
  27. Asberg KH, Parrow A. Event, incidence and fatality rates of cerebrovascular diseases in Enkoping-Habo, Sweden, 1986-1988. Scand J Soc Med. 1991;19:134-139. [Medline] [Order article via Infotrieve]
  28. Corwin LE, Wolf PA, Kannel WB, McNamara PM. Accuracy of death certification of stroke: the Framingham Study. Stroke. 1982;13:818-821. [Abstract]
  29. Hasuo Y, Ueda K, Kiyoham Y. Accuracy of diagnoses on death certificates for underlying causes of death in a long-term autopsy-based population study in Hisayama, Japan: with special reference to cardiovascular diseases. J Clin Epidemiol. 1989;42:577-584. [Medline] [Order article via Infotrieve]
  30. Bonita R, Stewart A, Beaglehole R. International trends in stroke mortality: 1970-1985. Stroke. 1990;21:989-992. [Abstract/Free Full Text]
  31. Aho K, Harmsen P, Hatano S, Marquardsen J, Smirnov VE, Strasser T. Cerebrovascular disease in the community: results of a WHO collaborative study. Bull World Health Organ. 1980;58:113-130. [Medline] [Order article via Infotrieve]
  32. Dennis M, Bamford J, Sandercock P, Molyneux A, Warlow C. Computed tomography in patients with transient ischaemic attacks: when is a TIA not a TIA but a stroke? J Neurol. 1990;237:257-261. [Medline] [Order article via Infotrieve]
  33. Giroud M, Gras P, Milan C, Vion P, Essayagh E, Dumas R. Prognostic des accidents ischemiques transitoires relevant un infarctus. Rev Neurol (Paris).. 1992;148:576-579. [Medline] [Order article via Infotrieve]
  34. Hansen BS, Marquardsen J. Incidence of stroke in Frederiksberg, Denmark. Stroke. 1977;8:663-665. [Abstract/Free Full Text]
  35. Yuling H, Bots ML, Pan X, Hofman A, Grobbee DE, Chen H. Stroke incidence and mortality in rural and urban Shanghai from 1984 through 1991. Stroke. 1994;25:1165-1169. [Abstract]
  36. Bamford J, Sandercock P, Warlow C, Gray M. Why are patients with acute stroke admitted to hospital? Br Med J. 1986;292:1369-1372.
  37. Tuomilehto J, Sarti C, Narva EV, Salmi K, Sivenius J, Kaarsalo E, Salomaa V, Torppa J. The FINMONICA Stroke Register: community-based stroke registration and analysis of stroke incidence in Finland, 1983-1985. Am J Epidemiol. 1992;135:1258-1270.
  38. Stegmayr B, Asplund K. Measuring stroke in the population: quality of routine statistics in comparison with a population-based stroke registry. Neuroepidemiology. 1992;11:204-213. [Medline] [Order article via Infotrieve]
  39. Rosman KD. The epidemiology of stroke in an urban black population. Stroke. 1986;17:667-669. [Abstract/Free Full Text]
  40. Matenga K, Kitai I, Levy L. Strokes among black people in Harare, Zimbabwe: results of computed tomography and associated risk factors. Br Med J. 1986;292:1649-1651.
  41. Joubert J. The MEDUNSA Stroke Data Bank: an analysis of 304 patients seen between 1986 and 1987. S Afr Med J. 1991;80:567-570. [Medline] [Order article via Infotrieve]
  42. Kinjo K, Kimura Y, Shinzato Y, Tomori M, Komine Y, Kawazoe N, Takashita S, Fukiyama K. An epidemiological analysis of cardiovascular diseases in Okinawa, Japan. Hypertens Res. 1992;15:111-119.
  43. Chen D, Roman GC, Wu GX, Wu ZS, Yao CH, Zhang M, Hirsch RP. Stroke in China (Sino-MONICA-Beijing study) 1984-1986. Neuroepidemiology. 1992;11:15-23. [Medline] [Order article via Infotrieve]
  44. Asplund K, Bonita R, Kuulasmaa K, Rajakangas A, Feigin V, Schaedlich H, Suzuki K, Thorvaldsen P, Tuomilehto J. Multinational comparisons of stroke epidemiology: evaluation of case ascertainment in the WHO MONICA Stroke Study. Stroke. 1995;26:355-360. [Abstract/Free Full Text]
  45. McCarty DJ, Tull ES, Moy CS, Kwoh CK, LaPorte RE. Ascertainment corrected rates: applications of capture-recapture methods. Int J Epidemiol. 1993;22:559-565. [Abstract/Free Full Text]
  46. Bruno G, LaPorte RE, Merletti F, Biggeri A, McCarty D, Pagano G. National Diabetes Programs: application of capture-recapture to count diabetes? Diabetes Care. 1994;17:548-556. [Abstract]
  47. Hook EB, Regal RR. Effect of variation in probability of ascertainment by sources (`variable catchability') upon `capture-recapture' estimates of prevalence. Am J Epidemiol. 1993;137:1148-1166. [Abstract/Free Full Text]
  48. Dennis MS, Bamford JM, Sandercock PA, Warlow CP. Incidence of transient ischemic attacks in Oxfordshire, England. Stroke. 1989;20:333-339. [Abstract/Free Full Text]
  49. Sandercock P, Molyneux A, Warlow C. Value of computed tomography in patients with stroke: Oxfordshire Community Stroke Project. Br Med J. 1985;290:193-197.
  50. Allen C. Clinical diagnosis of acute stroke syndrome. Q J Med. 1983;52:515-523. [Abstract/Free Full Text]
  51. Britton M, Hindmarsh T, Murray V, Tuden S. Diagnostic errors discovered by CT in patients with suspected stroke. Neurology. 1984;34:1504-1507. [Abstract/Free Full Text]
  52. Giroud M, Binnert D, Dumas R. Les erreurs diagnostic etiologique corrigees par la scanographie au cours d'une hemiplegie aigue. La Presse Medicale. 1986;15:2061-2063.
  53. Norris J, Hachinski V. Misdiagnosis of stroke. Lancet. 1982;1:328-331. [Medline] [Order article via Infotrieve]
  54. Whisnant JP, Melton LJ, Davis PH, O'Fallon WM, Nishimaru K, Schoenberg BS. Comparison of case ascertainment by medical record linkage and cohort follow-up to determine incidence rates for transient ischemic attacks and stroke. J Clin Epidemiol. 1990;43:791-794. [Medline] [Order article via Infotrieve]
  55. Wolf PA, D'Agostino RB, O'Neal A, Sytkowski P, Kase CS, Belanger AJ, Kannel WB. Secular trends in stroke incidence and mortality: the Framingham Study. Stroke. 1992;23:1551-1555. [Abstract/Free Full Text]
  56. Gardner MJ, Altman DG. Statistics With Confidence. London, England: British Medical Journal Publications; 1989:61-62.
  57. Bamford J, Sandercock P, Dennis M, Burn J, Warlow C. A prospective study of acute cerebrovascular disease in the community: the Oxfordshire Community Stroke Project 1981-1986, II: incidence, case fatality rates and overall outcome at one year of cerebral infarction, primary intracerebral haemorrhage and subarachnoid haemorrhage. J Neurol Neurosurg Psychiatry. 1990;53:16-22. [Abstract]
  58. Lindenstrom E, Boysen G, Nyboe J, Appleyard M. Stroke incidence in Copenhagen, 1976-1988. Stroke. 1992;23:28-32. [Abstract/Free Full Text]
  59. Hu H, Sheng W, Chu F, Lan C, Chiang B. Incidence of stroke in Taiwan. Stroke. 1992;23:1237-1241. [Abstract/Free Full Text]
  60. Raleigh VS, Balarajan R. Public health and the 1991 census. Br Med J. 1994;309:287-288.[Free Full Text]
  61. Christie D. Stroke in Melbourne, Australia: an epidemiological study. Stroke. 1981;12:467-469. [Abstract/Free Full Text]
  62. Ricci S, Celani MG, Vitali R, La Rosa F, Righetti E, Duca E. Diurnal and seasonal variations in the occurrence of stroke: a community-based study. Neuroepidemiology. 1992;11:59-64. [Medline] [Order article via Infotrieve]
  63. Shinkawa A, Ueda K, Hasuo Y, Kiyohara Y, Fujishima M. Seasonal variation in stroke incidence in Hisayama, Japan. Stroke. 1990;21:1262-1267. [Abstract/Free Full Text]
  64. Meneghini F, Rocca WA, Grigoletto F, Morgante L, Reggio A, Sarvettien G, Di Perri R, Anderson DW, for the Sicilian Neuro-Epidemiologic Study (SNES) Group. Door-to-door survey of neurological diseases in a Sicilian population. Neuroepidemiology. 1991;10:70-85. [Medline] [Order article via Infotrieve]
  65. Kuller LH. Incidence rates of stroke in the eighties: the end of the decline in stroke? Stroke. 1989;20:841-843. [Free Full Text]
  66. Waterhouse J, ed. Cancer Incidence in Five Continents. Lyon, France: IARC; 1976.
  67. Drury I, Whisnant JP, Garraway WM. Primary intracerebral haemorrhage: impact of CT on incidence. Neurology. 1984;34:653-657. [Abstract/Free Full Text]
  68. Suzuki K, Kutsuzawa T, Takita K, Ito M, Sakamoto T, Hirayama A, Ito T, Ishida T, Ooishi H, Kawakami K, Hirota K, Ogasawara T, Toshida J, Tamura T, Hattori S, Iwabuchi S, Karouji Y, Waga T, Oosato Y, Yazaki K, Saito T, Oouchi T, Kojima S. Clinico-epidemiological study of stroke in Akita, Japan. Stroke. 1987;18:402-406. [Abstract/Free Full Text]
  69. Kay R, Woo J, Kreel L, Wong HY, Tech R, Nicholls MG. Stroke subtypes among Chinese living in Hong Kong: the Shatia Stroke Registry. Neurology. 1992;42:985-987. [Abstract/Free Full Text]
  70. Sandercock P, Allen C, Corston R, Harrison M, Warlow C. Clinical diagnosis of intracranial haemorrhage using Guy's Hospital Score. Br Med J. 1985;291:1675-1677.
  71. Poungvarin N, Viriyavejakul A, Komontri C. Siriraj Stroke Score and validation study to distinguish supratentorial intracerebral haemorrhage from infarction. Br Med J. 1991;302:1565-1567.
  72. Celani MG, Righetti E, Migliacci R, Zampolini M, Antoniutti L, Grandi FC, Ricci S. Comparability and validity of two clinical scores in the early differential diagnosis of acute stroke. Br Med J. 1994;308:1674-1676. [Abstract/Free Full Text]
  73. Weir CJ, Murray GD, Adams FG, Muir KW, Grosset DG, Lees KR. Poor accuracy of stroke scoring systems for differential clinical diagnosis of intracranial haemorrhage from infarction. Lancet. 1994;344:999-1002. [Medline] [Order article via Infotrieve]
  74. Dennis M, Bamford J, Molyneux A, Warlow C. Rapid resolution of signs of PICH in CT of brain. Br Med J. 1987;95:379-381.
  75. Dolinskas CA, Bilaniuk LT, Zimmerman RA, Kuhl DE. Computed tomography of intracerebral hematomas, I: transmission CT observations on hematoma resolution. AJR Am J Roentgenol. 1977;129:681-688. [Abstract]
  76. Bogousslavsky J, Regli F, Uske A, Maeder P. Early spontaneous hematoma in cerebral infarct: is primary cerebral hemorrhage overdiagnosed? Neurology. 1991;41:837-840. [Abstract/Free Full Text]
  77. Wolf PA, Lewis A. Conner lecture: contributions of epidemiology to the prevention of stroke. Circulation. 1993;88:2471-2478. [Free Full Text]
  78. Bamford J, Sandercock P, Dennis M, Burn J, Warlow C. Classification and natural history of clinically identifiable subtypes of cerebral infarction. Lancet. 1991;337:1521-1526. [Medline] [Order article via Infotrieve]
  79. Naylor AR, Sandercock PAG, Sellar RJ, Warlow CP. Patterns of vascular pathology in acute, first-ever cerebral infarction. Scottish Med J. 1993;38:41-44.
  80. Wardlaw JM, Dennis MS, Lindley RI, Sellar RJ, Warlow CP. The validity of a simple clinical classification of acute ischaemic stroke. J Neurol. In press.
  81. Lindley RI, Warlow CP, Wardlaw JM, Dennis MS, Slattery J, Sandercock PAG. Interobserver reliability of a clinical classification of acute cerebral infarction. Stroke. 1993;24:1801-1804. [Abstract/Free Full Text]
  82. Adams HP, Bendixen BH, Kappelle J, Biller J, Love BB, Gordon DL, Marsh EE, and the TOAST Investigators. Classification of subtype of ischemic stroke: definitions for use in a multicenter trial. Stroke. 1993;24:35-41. [Abstract/Free Full Text]
  83. Sacco RL, Ellenberg JH, Mohr JP, Tatemichi TK, Hier DB, Price TR, Wolf PA. Infarcts of undetermined cause: the NINCDS Stroke Data Bank. Ann Neurol. 1989;25:382-390. [Medline] [Order article via Infotrieve]
  84. Anderson CS, Taylor BV, Hankey GJ, Stewart-Wynne EG, Jamrozik KD. Validation of a clinical classification for subtypes of acute cerebral infarction. Stroke. 1994;57:1173-1179.



This article has been cited by other articles:


Home page
StrokeHome page
P. U. Heuschmann, A. P. Grieve, A. M. Toschke, A. G. Rudd, and C. D.A. Wolfe
Ethnic Group Disparities in 10-Year Trends in Stroke Incidence and Vascular Risk Factors: The South London Stroke Register (SLSR)
Stroke, August 1, 2008; 39(8): 2204 - 2210.
[Abstract] [Full Text] [PDF]


Home page
StrokeHome page
B. Hallstrom, A.-C. Jonsson, C. Nerbrand, B. Norrving,