Journal ofNeurology, Neurosurgery, and Psychiatry 1996;61:333-338

NEUROEPIDEMIOLOGY

Epidemiology of stroke

Kay-Tee Khaw

Stroke accounts for about 10% of all deaths inindustrialised countries and is responsible for avast burden of disability in the community.Once a stroke has occurred, treatment islargely focused on care and rehabilitation.However, there is evidence to suggest thatstroke is potentially preventable. Firstly, thereare large international variations in mortalityfrom stroke and migrants between areas wherethe rates of stroke are different generally expe-rience a change in risk; this implies that envi-ronmental factors may be more importantthan genetic susceptibility in determining risk.Secondly, many countries have experienced asteep decline in mortality from stroke in recentdecades; although alterations in diagnosticfashion or certification practice might accountfor a small proportion of this decline, the mostlikely explanation seems to be changes in riskfactor levels over time. Thirdly, epidemiologicalstudies on causes of stroke have indicated thatsome of these are avoidable.

Stroke is often grouped together with coro-nary heart disease as cardiovascular diseasesbut there are substantial differences in the epi-demiology of stroke and coronary heart dis-ease. The most notable of these are thedifferences in time trends and the fact that thepronounced male excess found for coronaryheart disease is not seen in stroke.

University ofCambridge School ofClinical Medicine,Addenbrooke'sHospital, CambridgeCB2 2QQ, UKK-T Khaw

DefinitionsThe World Health Organisation defines strokeas "rapidly developing clinical signs of focal(or global) disturbance of cerebral function,with symptoms lasting 24 hours or longer orleading to death with no apparent cause otherthan of vascular origin". 2 Thus subarachnoidhaemorrhages are included but not transientischaemic attacks, subdural haematoma, andtumour or infection causing haemorrhage orinfarction. Clinical diagnoses have been shownto be reliable.'Most comparisons of stroke using routinely

collected vital statistics rely on mortality rates.These have disadvantages including issues ofdiagnostic reliability, and variable case fatalitywhich have been fully discussed elsewhere.4 5 Itis also not possible in mortality data to makeaccurate distinctions between the variousstroke subtypes, in particular between haem-orrhagic and thrombotic strokes. Morbidity

and disability due to stroke are also of majorconcern, but few routine sources are likely toprovide satisfactory data. A substantial pro-portion of people having a stroke die suddenlybefore reaching hospital or are cared for athome; hospital admission or discharge statis-tics depend on admission policies and accessi-bility and these vary enormously from countryto country and over time. Most studies onincidence of stroke have methodological prob-lems. When community stroke registers havebeen specially set up, as in the World HealthOrganisation (WHO) MONICA Project4 theevidence suggests that, for the purposes ofinternational comparisons, there is goodagreement between mortality rates in officialstatistics and stroke incidence registers. Timetrends also need to be viewed with cautionbecause of changing coding practices butagain, many comparative studies have beenconducted to examine the impact of these.

Case fatalityIn the MONICA study, case fatality rates(defined as the proportion of events thatare fatal within 28 days of onset) averagedabout 30%, although they ranged from15%-50%, with the lowest case fatalities inthe Nordic countries and the highest in mostof the Eastern European populations; therewas no substantial sex difference in casefatality.3

Mortality by age and sexTable 1 shows the mortality rates for stroke byage and sex in the United Kingdom. Rates risesharply with increasing age. In marked con-trast with coronary heart disease, in whichthere is a 5:1 male:female ratio at youngerages, there is no consistent sex differential instroke mortality rates.

Table 1 Mortality ratesllO0 000for stroke, UnitedKingdom 1992 (from jJ906)

Stroke mortality rate/100 000/age group35-44 45-54 55-64 65-74 > 75

Men 7 19 79 291 1351Women 7 17 55 230 1438

333

on March 16, 2022 by guest. P

rotected by copyright.http://jnnp.bm

j.com/

J Neurol N

eurosurg Psychiatry: first published as 10.1136/jnnp.61.4.333 on 1 O

ctober 1996. Dow

nloaded from

Khaw

Figure 1 Age standard-ised stroke mortality rates100 000 by sex 1990-2(from WHO6).

International comparisonsFigure 1 shows age-standardised stroke mor-tality rates in men and women for selectedcountries between 1991-2.6 The highest docu-mented rates are now seen in the countries ofEastern Europe and the former USSR. Thereis substantial international variation with menand women in the highest rate countries hav-ing about fivefold the rates in the lowest ratecountries. Whereas men tend to have slightlyhigher rates than women, the sex differentialdoes not compare with the pronounced maleexcess for coronary heart disease noted else-where. Figure 2 shows age specific rates for

3500 o Russian Federation0 United Kingdom

3000 - o United States

2500 -

2000

. 1500

55-64 65-74Age (y)

the United Kingdom, United States, andRussian Federation and indicates that the agestandardised mortality comparisons reflectconsistent differences over all age groups.

Time trendsFigure 3 shows time trends in stroke mortalityfrom 1950 to 19896 and illustrates how thesediffer between selected countries. Japan,which had the highest rates in 1950-4, had anincrease in mortality during the 1960s butthen experienced a substantial decline; thistrend was similar in men and women,although greater in men. Most other countriesexperienced a decline in mortality from strokealthough others, particularly in EasternEurope-of which Hungary is an example-had an increase in rates and only from the1980s was there a decline. The reason forthese trends is unclear. It has often been notedthat the decline in stroke mortality in the west-ern countries preceded, and hence cannot beexplained by, the use of antihypertensive med-ication.7

Figure 4 shows changes in stroke mortalityrates between 1960-4 and 1985-9 in variouscountries. During this period many countriesexperienced substantial declines, more thanhalving stroke mortality; however, in somecountries, notably those in eastern Europe,there was an increase in death rates. Thesetrends suggest that the major determinants ofmortality rates from stroke are likely to bepotentially modifiable factors rather than

Figure 2 Stroke mortalityrates per 100 000 by agefor men in three countries.

334

.... ........................

.......................7..............................

... ... ........... .........

............................................... ....... ...

......

... ............. ........ ..........

..... ...

........

on March 16, 2022 by guest. P

rotected by copyright.http://jnnp.bm

j.com/

J Neurol N

eurosurg Psychiatry: first published as 10.1136/jnnp.61.4.333 on 1 O

ctober 1996. Dow

nloaded from

Epidemiology ofstroke

Figure 3 (A) Agestandardised mortality rates100 000from stroke in men1950-89. (B) Age stan-dardised mortality rates100 000from stroke inwomen 1950-89.

C0000

'._

t0

Figure 4 Changes in agestandardised mortalityrates from stroke in menand women 1960-4 and1985-9 in selectedcountries.

00000

v-:t

0

Age (y)

Japarn

United States

Switzerland

France

Australia

Denmark

Finland

Utnited Kingdom

Sweden

Norwvay

Netherlands

Spain

Italy

Austria

Singapore

BeIgILIm

Portu gal

Hungary

BRu 1garia

Greece

Poland

-80 -60

65-69 70-74 75-79 80-84 85-89Age (y)

:...:: :.. ....: ....i

:::::. WomenMen

77

-40

Decrease

-20%0 Change

0 20

Increase

genetic differences in stroke susceptibility.The results of migrant studies also emphasisethis point; Japanese populations in the UnitedStates experience rates of stroke that are closerto those of American white populations thanto those of Japanese populations in Japan.8Whereas some of the decline in stroke mortalitymay be due to a reduction in case fatality,9possibly due to better medical care or chang-ing severity of disease, the magnitude of the

declines suggest mortality rates also reflectchanging incidence.

Regional and social class differencesWithin any country, rates of mortality fromstroke vary by social class and geographicregion.'0 In Britain, rates vary over threefoldbetween different regions, with the lowestmortality in the south and the highest in the

40

............

...................

335 on M

arch 16, 2022 by guest. Protected by copyright.

http://jnnp.bmj.com

/J N

eurol Neurosurg P

sychiatry: first published as 10.1136/jnnp.61.4.333 on 1 October 1996. D

ownloaded from

Khaw

north. Men in social class V have a risk ofdeath from stroke nearly three times as high asmen in social class 1; a similar pattern is seen inwomen.

Preventive strategiesThere are two dimensions to the problem ofstroke in the community: treatment, rehabili-tation, and care of those who have had astroke; and prevention of the occurrence ofstroke. The large geographic variations, resultsof migrant studies, and changes over time instroke mortality and incidence indicate thepotential for prevention. Major challenges areto identify aetiological factors, and to quantifytheir potential impact on stroke burden.Laboratory, clinical, and epidemiological stud-ies all contribute to our understanding of aeti-ology; this review focuses on theepidemiological evidence. Many biologicaland environmental factors have been impli-cated in the aetiology of stroke. Of these, therole of raised blood pressure has been the bestdocumented. Strategies aimed at preventingstroke may be based on individual subjects orpopulation based." The high risk, or individ-ual based approach aims to identify specificsubjects at high risk of stroke for clinical inter-ventions (such as that of reduction of highblood pressure using pharmacological therapyor behavioural changes). The populationbased approach aims to change levels of riskfactors in the whole population to a more opti-mal distribution (such as by changing dietarypatterns).

Table 2 Risk factors for stroke

Prspective studies Interventionpredictive of trial with

Major risk factors stroke stroke end point

Blood pressure + +Cigarette smoking +Diabetes +Fibrinogen level +Obesity +Blood homocysteine level +Respiratory function +White cell count +

Table 3 Lifestyle factors implicated in stroke

Relation with Predictive ofblood pressure stroke inreported prospective studies

Diet:Adverse

Sodium +Saturated fat +Alcohol + +

ProtectivePotassium + +Vitamin C + +Total calories - +Fruit and vegetables - +co-3-fatty acids +Fibre +

Cigarette smoking - +(adverse)

Physical activity + +(protective)

Obesity (adverse) + +

Biological risk factorsBLOOD PRESSUREThe evidence for the aetiological role of raisedblood pressure in stroke is overwhelming.Numerous prospective studies, recentlyreviewed,'2 have shown that systolic and dias-tolic blood pressure both independently pre-dict stroke; the risk increases steeply andcontinuously with increasing blood pressure,with no threshold: the higher the blood pres-sure, the higher the stroke risk over the wholedistribution of blood pressure, with relativerisks increasing by nearly twofold for every 10mm increase in diastolic blood pressure.Treatment trials of reduction of blood pres-sure have shown that a 5-6 mm Hg reductionin diastolic blood pressure reduces stroke riskby about 33-50%."3

CHOLESTEROLBy contrast, the evidence that raised bloodcholesterol, which is a powerful risk factor forcoronary heart disease, has an effect on stroke ismuch more equivocal.'2 It has been postulatedthat the apparent lack of any consistent rela-tion might be due to a positive associationwith ischaemic stroke and a negative associa-tion with haemorrhagic stroke; trials of choles-terol reduction to date have not had sufficientpower to examine significant effects on strokeone way or another.

Other biological risk factorsMany other biological factors have been impli-cated as predictive of stroke. They provideintriguing evidence for the mechanisms under-lying stroke and raise the possibility of preven-tive interventions. However, for these riskfactors, there is as yet no evidence from inter-vention trials that changes in exposure willreduce risk of stroke. Raised fibrinogen,'4raised homocysteine"5 and lower albumin con-centrations,'6 raised white cell count,'7 infec-tion,'8 poor respiratory function,'9 diabetes orimpaired glucose tolerance,20 and obesity,mainly central adiposity2' 22 have all beenlinked with increased risk of stroke. These fac-tors may have independent effects, may inter-act, or may simply be markers of underlyingprocesses. For example, white cell count orpoor respiratory function may be a marker ofchronic infection or inflammation, which mayraise stroke risk by increasing the likelihood ofthrombosis. Alternatively, white cell countmight be a marker for cigarette smoking.Table 2 summarises these factors.

Lifestyle risk factorsMany lifestyle factors have been linked withstroke (table 3). These include cigarette smok-ing, physical activity, and diet. Some prospec-tive studies have reported direct associationsbetween these factors and stroke mortality orincidence; the results of other studies suggestthat dietary factors may influence risk of strokeby their effect on blood pressure.

SMOKINGCigarette smoking is now well documented as

336 on M

arch 16, 2022 by guest. Protected by copyright.

http://jnnp.bmj.com

/J N

eurol Neurosurg P

sychiatry: first published as 10.1136/jnnp.61.4.333 on 1 October 1996. D

ownloaded from

Epidemiology of stroke

increasing stroke risk; one study reported anapproximate threefold risk in current com-pared with non-smokers.23 It is not clear howsmoking affects risk of stroke but it may bethrough thrombotic mechanisms.

PHYSICAL ACTIVITYSeveral studies have reported that physicalactivity is protective against stroke in bothmen and women24-26; this may be throughinfluencing blood pressure, or through othermechanisms such as platelet aggregation.

DIETMuch interest and debate has focused on therole of dietary factors.26-5 The evidence hasbeen along two lines: dietary factors whichmay influence blood pressure, and therebystroke, and dietary factors which may affectrisk of stroke through other mechanisms, suchas haemostasis or homocysteine concentra-tions. Of these, sodium has figured mostprominently. The hypothesis that sodiummight have an aetiological role in hypertensionand stroke is not at all new but the debateis periodically revived.29-" There is little doubtfrom a wealth of intervention trials andobservational studies that sodium has an effecton blood pressure; the debate is largely overthe clinical and public health importance ofthe effect. Ecological studies have foundthat populations with a high sodium intakehave an increased risk of stroke, but thisassociation has not been convincinglydemonstrated in prospective studies, perhapsdue to the difficulty of measuring sodiumintake. On balance, there is considerablecircumstantial evidence that sodium isimportant in influencing hypertension andstroke risk. The magnitude of effect appliedto the population has been estimated atabout 10 mm Hg blood pressure differencefor a 100 mmol difference in sodiumintake, or about 34% difference in risk ofstroke.34

Other dietary factors which have beenrelated to blood pressure include potassium,magnesium, calcium, dietary fibre, a)-3 fattyacids, saturated fat, vitamin C, protein, andalcohol,26 35-42 but again the relation with strokeis more uncertain. Several prospective studieshave suggested that high intakes of potassiumor vitamin C may be protective for stroke4344and others that high alcohol intake may havean adverse effect, particularly as regards risk ofhaemorrhagic stroke.45 Dietary factors mayhave effects other than on blood pressure.Antioxidant vitamins, for example, may influ-ence haemostasis or endothelial function byprotecting against free radical damage.46Folate may reduce stroke risk by influencinghomocysteine concentrations. Fruit and veg-etables are a rich source of vitamin C, potas-sium, and folate. Several studies have reportedthat a high fruit and vegetable intake seems toprotect against stroke.4347-9 The magnitude ofthe effect was considerable with an estimated25%-40% reduction in risk of stroke for anincrease of two to three servings of fruit or veg-etables daily.

ConclusionsStroke is a devastating condition. However,the evidence from epidemiological and otherstudies suggests that stroke is eminently pre-ventable; the challenge is to identify effectivemethods of prevention. We already haveample evidence that measures to reduce bloodpressure in individual subjects and popula-tions have a substantial impact on stroke rates.Because there is no threshold for the associa-tion with blood pressure, reduction of bloodpressure in the population as a whole by alter-ing lifestyle may be the most effective way toreduce the incidence of stroke. While themechanisms remain to be clarified, interven-tion trials of vitamin supplements and otherdietary modifications would be worthwhile.Such trials are already in progress in the USA.There is circumstantial evidence that lifestylefactors influence stroke risk in individual sub-jects and stroke rates in the community.Current health recommendations to increasefruit and vegetable intake, to reduce sodiumintake, to increase physical activity, and toreduce cigarette smoking are likely to reducerisk of stroke and benefit health in general.

1 Tunstall Pedoe H, for the WHO MONICA ProjectPrincipal Investigators. The World Health OrganisationMONICA Project (monitoring trends and determinantsin cardiovascular disease): a major international collabo-ration. J Clin Epidemiol 1988;41:105-14.

2 World Health Organisation MONICA Project. MONICAManual, revised edition. Geneva: Cardiovascular DiseasesUnit, WHO; 1990.

3 Sandercock P, Molyneaux A, Warlow C. Value of comput-erised tomography in patients with stroke: theOxfordshire Community Stroke Project. BMJ 1985;290:193-7.

4 Thorvaldsen P, Asplund K, Kuulasmaa K, RajakangasAM, Schroll M. Stroke incidence, case fatality and mor-tality in the WHO MONICA Project. Stroke 1995;26:361-7.

5 Asplund K, Bonita R, Kuulasmaa K, et al. Multinationalcomparisons of stroke epidemiology. Stroke 1995;26:361-7.

6 World Health Organisation. World Health statistics annuals1982-1994. Geneva: WHO, 1982-1994.

7 Klag MJ, Whelton PK, Seidler AJ. Decline in US strokemortality. Stroke 1989;20: 14-21.

8 Marmot MG, Syme SL, Kagan A, et al. Epidemiologicstudies of coronary heart disease and stroke in Japanesemen living in Japan, Hawaii and California: prevalence ofcoronary and hypertensive heart disease and associatedrisk factors. AmJIEpidemiol 1975;102:514-25.

9 Bonita R, Beaglehole R. Explaining stroke mortality trends.Lancet 1993;341:1510-1.

10 Central Health Monitoring Unit. Epidemiological overviewseries. Stroke. An epidemiological overview. The health of theNation. London: HMSO, 1994.

11 Rose G. The strategy ofpreventive medicine. Oxford: OxfordUniversity Press, 1992.

12 Prospective Studies Collaboration. Cholesterol, diastolicblood pressure, and stroke: 13,000 strokes in 450,000people in 45 prospective studies. Lancet 1995;346:1647-53.

13 Collins R, Peto R, MacMahon S, et al. Blood pressure,stroke and coronary heart disease. Part 2, short termreductions in blood pressure: overview of randomiseddrug trials in their epidemiologic context. Lancet1990;335:827-38.

14 Wilhelmsen L, Svardsudd K, Korsan-Bengten K, et al.Fibrinogen as a risk factor for stroke and myocardialinfarction. NEnglJ3Med 1984;311:501-5.

15 Perry IJ, Refsum H, Morris RW, Ebrahim SB, Ueland PM,Shaper AG. Prospective study of serum total homocys-teine concentration and risk of stroke in middle-agedBritish men. Lancet 1995;346:1395-98.

16 Gillum RF, Ingram DD, Makuc DM. Relation betweenserum albumin concentration and stroke incidence anddeath: the NHANES 1 epidemiologic follow up study.Am J Epidemiol 1994;140:876-88.

17 Gillum RF, Ingram DD, Makuc DM. White blood cellcount and stroke incidence and death. Am J Epidemiol1994;139:894-902.

18 Syrjanen J, Valtonen VV, Iivanainen M, Kaste M,Huttunen JK. Preceding infection as an important riskfactor for ischaemic brain infarction in young and middleaged patients. BMJ 1988;296: 1156-60.

19 Strachan DP. Ventilatory function as a predictor of fatalstroke. BMJ 1991;302:84-7.

337

on March 16, 2022 by guest. P

rotected by copyright.http://jnnp.bm

j.com/

J Neurol N

eurosurg Psychiatry: first published as 10.1136/jnnp.61.4.333 on 1 O

ctober 1996. Dow

nloaded from

Khaw

20 Barrett-Connor E, Khaw KT. Diabetes-an independentrisk factor for stroke? Am J Epidemiol 1988;128: 116-23.

21 Shinton R, Shipley M, Rose G. Overweight and stroke inthe Whitehall study. J Epidemiol Community Health199 1;45:138-42.

22 Welin L, Svardsudd K, Wilhelmsen L, Larsson B, TibblinG. Analysis of risk factors for stroke in a cohort of menborn in 1913. N EnglJ Med 1987;317:521-6.

23 Khaw KT, Barrett-Connor E. Suarez L, Criqui M.Predictors of stroke-associated mortality in the elderly.Stroke 1984;15:244-8.

24 Gillum RF, Mussolino ME, Ingram DD. Physical activityand stroke incidence in women and men. Am J Epidemiol1996;143:860-9.

25 Wannamethee G, Shaper AG. Physical activity and strokein British middle aged men. BMJ 1992;304:597-601.

26 Manson JE, Stampfer MJ, Willett WC, et al. Physical activ-ity and incidence of coronary heart disease and stroke inwomen. Circulation 1995;91:927.

27 Beilin U. Diet and hypertension: critical concepts and con-troversies. J Hypertens 1987;5:S447-S57.

28 Marmot MG, Poulter P. Primary prevention of stroke.Lancet 1992;339:344-7.

29 Bronner LL, Kanter DS, Manson JE. Primary preventionof stroke. NEnglJrMed 1995;333:1392-400.

30 Frost CD, Law MR, Wald NJ. By how much does dietarysalt reduction lower blood pressure: I: analysis of obser-vational data within populations. BMJ7 1991;302:815-9.

31 Elliott P, Stamler J, Nichols R, Dyer AR, Stamler R,Marmot M, et al. Intersalt revisited: further analyses fromthe Intersalt study on 24 hour urinary sodium and bloodpressure. BMJ 1996;312:1249-53.

32 Thelle DS. Salt and blood pressure revisited. BMJ 1996;312:1240-1.

33 Khaw KT. Sodium, potassium, blood pressure and stroke.In: Poulter NP, Thom S, Sever PS, eds. Cardiovasculardisease: risk factors and intervention. Oxford: RadcliffeMedical Press, 1993:145-52.

34 Law M. Commentary: evidence on salt is consistent. BM31996;312: 1284-5.

35 Swain JF, Rouse IL, Curley CB, Sacks FM. Comparison ofthe effects of oat bran and low-fibre wheat on serum

lipoprotein levels and blood pressure. N Engl Jf Med1990;322:147-52.

36 Criqui MH, Langer RD, Reed DM. Dietary alcohol, cal-cium, and potassium: independent and combined effectson blood pressure. Circulation 1989;80:609-14.

37 Harlan WR, Harlan LC. An epidemiological perspective ondietary electrolytes and hypertension. J Hypertens 1986;4:S334-9.

38 McCarron DA, Morris CD, Young E, Roullet C, DruekeT. Dietary calcium and blood pressure: modifyingfactors in specific populations. Am J Clin Nutr 1991;54:215S-9S.

39 Zhou B, Zhang X, Zhu A, et al. The relationship of dietaryanimal protein and electrolytes to blood pressure: a studyon three Chinese populations. Int J Epidemiol 1994;23:716-22.

40 Salonen JT. Dietary fats, antioxidants and blood pressure.Ann Med 1991;23:295-8.

41 Obarzanek E, Velletri PA, Cutler JA. Dietary protein andblood pressure. JAMA 1996;275:1598-603.

42 Ness A, Khaw KT, Bingham S, Day NE. Vitamin C statusand blood pressure. I Hypertens 1996;14:503-8.

43 Khaw KT, Barrett-Connor E. Dietary potassium andstroke-associated mortality: a twelve year prospectivepopulation study. N EnglJ Med 1987;216:235-40.

44 Gale CR, Martyn CN, Winter PD, Cooper C. Vitamin Cand risk of death from stroke and coronary heart disease ina cohort of elderly people. BM3 1995;310:1563-6.

45 Gill JS, Zezulka AV, Shipley MJ, Gill SK, Beevers DG.Stroke and alcohol consumption. New Engl J Med 1986;315: 1041-6.

46 Khaw KT, Woodhouse PR. Vitamin C, infection, haemo-static factors and cardiovascular disease risk. BM3' 1995;310:1559-63.

47 Gillman MW, Cupples A, Gagnon D, et al. Protectiveeffect of fruits and vegetables on development of stroke inmen. JAMA 1995;273:1113-7.

48 Vollset SE, Bjelke E. Does consumption of fruit and veg-etables protect against stroke? Lancet 1983:ii:742.

49 Acheson RM, Williams DRR. Does consumption of fruitand vegetables protect against stroke? Lancet 1983;i:1191-3.

338 on M

arch 16, 2022 by guest. Protected by copyright.

http://jnnp.bmj.com

/J N

eurol Neurosurg P

sychiatry: first published as 10.1136/jnnp.61.4.333 on 1 October 1996. D

ownloaded from