1Journal of Epidemiology and Community Health 1997,51:138-143

Lifelong exposures and the potential for strokeprevention: the contribution of cigarettesmoking, exercise, and body fat

Roger Shinton

AbstractStudy objective - To examine the potentialfor stroke prevention by avoidance of haz-ards related to lifestyle. Assessment of thelifelong contribution of both individualand combined exposures was a particularfeature of the study. An estimate wassought of the proportion (population at-tributable risk fraction) of strokes likelyto be caused by one or a combination ofcigarette smoking, lack of exercise, andobesity.Design - Case-control study.Setting - Eleven general practices in westBirmingham.Participants - Altogether 125 men andwomen who had just had their first strokeand were aged 35-74 and 198 controls fre-quency matched for age and sex recruitedover 24 months during 1988-90.Main results - The hazards of cigarettesmoking for stroke were confirmed andnoted to persist for one to two decadesafter stopping smoking. Cigarette smokingwas estimated to have caused 49% (95%confidence interval 22, 67%) of the strokesin this population with roughly equal con-tributions from current and former smok-ing. Combinations of cigarette smoking(current and former) with lack of exerciseor previous obesity apparently caused 62%and 72% of the strokes respectively. If ci-garette smoking, lack ofexercise and obes-ity were all avoided, 79% (32, 94%) of thestrokes could potentially have been pre-vented. No stroke patient reported a com-bination of never smoking, taking regularexercise aged 15 to 25, and avoidance ofoverweight (body mass index >25 kg/m2).Conclusions - Taken together, the com-bination of cigarette smoking, excessivebody fat, and lack of exercise accountedfor a major proportion of stroke cases inthe population studied. It appears thatthese easily identifiable factors related tolifestyle are a major and possibly pre-dominant cause of stroke, at least until theage of 75.

University ofBirmingham,Department ofMedicine (Elderly),BirminghamHeartlands Hospital,Yardley Green Unit,Birmingham B9 5PXRoger Shinton

Correspondence to:Dr Shinton.

(7 Epidemiol Community Health 1997;5l:138-143)

The extent to which lifestyle can prevent strokehas received surprisingly little attention. Ci-garette smoking is now widely recognised as a

risk factor for stroke.' 4There have, however,been few estimates of the proportion of strokes

likely to have been caused by cigarette smokingin populations at risk (population attributablerisk fraction). Considerable doubt also remainsconcerning the speed with which those stoppingsmoking return to the risk of those who neverstarted. An Australian case-control study ofhospital patients with "non-embolic" cerebralinfarction has suggested the increased risk per-sists at least 10 years.2 No increased risk ofoverall stroke, however, was seen in formersmokers in the British regional heart studycohort and the risk quickly returned to baselinein the American nurses cohort.34There is increasing evidence that other po-

tentially avoidable hazards are also importantcauses of stroke, for example lack of exerciseand excessive body fat.5`8 Case-controlmethods, allowing a lifelong view, have not yetreported the extent to which stroke is pre-ventable by avoidance of particular com-binations of hazards which are within thepotential control of the individual. The westBirmingham stroke project was designed toexamine the contribution of single and com-bined lifestyle related risk factors in causingstroke. Data from the study have already in-dicated that lifelong exercise and and the avoid-ance of overweight and obesity protect againstthe risk of stroke.78 The community setting ofthis study has the added advantage of allowingestimates of the proportion of strokes causedby particular exposures (the population at-tributable risk fraction) or combinations of ex-posures. These estimates, although strictly onlyappropriate for the study population, will reflectthe picture in populations with similar char-acteristics.

MethodsThe detailed methods of this case-control studyhave been previously described.7 Those in-dividuals currently on the register of 11 generalpractice partnerships in west Birmingham com-prised the study base population (n = 101 000).

STROKE PATIENTSStroke patients were assessed for inclusionusing the standard World Health Organisationcriteria - rapidly developing clinical signs offocal and at times global (applied to patientsin deep coma and to those with subarachnoidhaemorrhage) loss of cerebral function, withsymptoms lasting more than 24 hours or lead-ing to death, with no apparent cause other thanthat of vascular origin."' Both hospital and

138 on M

arch 26, 2020 by guest. Protected by copyright.

http://jech.bmj.com

/J E

pidemiol C

omm

unity Health: first published as 10.1136/jech.51.2.138 on 1 A

pril 1997. Dow

nloaded from

Lifelong exposures and stroke

community stroke patients, if fit, were offereda computerised tomography head scan (withoutcontrast). Cerebral infarction was diagnosed ifa head scan or autopsy was performed within21 days of stroke onset and there was eitherno evidence of haemorrhage or a pattern ofpredominant infarction. Those included weremale and female patients with their first stroke,between the ages of 35 and 74 years, occurringbetween 1st October 1988 and 30th September1990. The following characteristics excludedstroke patients from the study; a prior historyof stroke (but not transient ischaemic attack),a history of mitral valvular heart disease com-bined with atrial fibrillation, a history of prim-ary or metastatic cerebral neoplasm, a historyof coagulation disorder or myeloproliferativedisease, and patients dying before assessmentcould be made.

CONTROL SUBJECTSSubjects were randomly selected from the gen-eral practice population to match broadly byage and sex the incident stroke case distribution(frequency matching). The same exclusionsapplied as for stroke patients.

ASSESSMENTThe cases and controls were all anthro-pometrically assessed and interviewed by asingle observer (RS) using a standard struc-tured questionnaire. Stroke patients were seenas soon as possible following the event. Whendisability prevented an adequate response theclosest relative or friend was interviewed.

Cigarette smoking status was classified intonever, former, and current smokers by re-sponses to to the questions; "Have you eversmoked cigarettes regularly, that is at least onecigarette per day for one year?" and "Haveyou smoked cigarettes during the past year?"Former smokers were also asked how long agothey stopped smoking. General practice andhospital notes were also examined for a priorrecord of cigarette smoking status.To assess physical exercise, subjects were

divided by their response to the question; "Didyou often take part in vigorous sports or ac-tivities between the ages of 15 and 25 years?"Examples of vigorous exercise provided by thequestionnaire were digging, running, swim-ming, cycling, playing tennis or squash, andkeeping fit.

Standing height was measured, when pos-sible, by standard methods using a portablestadiometer." Lying height was measured in allstroke patients admitted to hospital. For lyingheight the subject lay on his/her back andan assistant ensured the heels were correctlypositioned at the base of the stadiometer. Asimple regression equation which related stand-ing and lying height amongst stroke patientswho could stand enabled standing height to beestimated for those unable to stand. Cases andcontrols were asked; "What is the most youhave ever weighed?" The response was con-verted into kilograms and combined withmeasured/derived standing height to provide

maximum reported body mass index: weightin kilograms/(height in metres)2. Harpendenskinfold calipers were used to measure sub-scapular skinfold thickness using standardmethods." Two records for each subject weremade and the mean calculated. Sex specificinterquartile ranges amongst controls were es-tablished to subdivide subjects by the variousmeasures of body fat. Other variables whichcould confound or interact with observed re-lations were assessed by questionnaire and havebeen detailed previously.79

STATISTICAL ANALYSISA computer software programme (EGRET)'2using methods described by Breslow & Day'3was employed to calculate odds ratios (relativerisk) and their 95% confidence intervals. Man-tel-Haentzel methods were used to provide anestimate of the odds ratio after adjustment forone or two other variables. Adjustment forseveral variables was made by use of the mul-tiple logistic regression model. Regressionmodels were fitted using exposure levels of agiven risk factor as discrete (factored) variables.The population attributable risk fraction(PARF) and 95% confidence intervals for agiven exposure was estimated from the oddsratio and the prevalence of the exposure in thecontrol population according to the equation"':

PARF =prevalence (odds ratio-i)

prevalence x (odds ratio-i) + 1

Inspection of this formula demonstrates thatraising either the odds ratio or the prevalenceof the exposure will increase the size of thepopulation attributable risk fraction. The popu-lation attributable risk fraction ranges from 0(no effect) to 1 (all cases explained by theexposure).

ResultsAltogether 125 eligible stroke patients wereidentified over 24 months. Forty three (34%)were managed at home. No stroke patient (orrelative) declined to participate. Cerebralinfarction was diagnosed in 81 patients,intracerebral haemorrhage in five, and sub-arachnoid haemorrhage in four. In the re-maining 35 patients computed tomographywithin 21 days was either impossible or de-clined by the patient, relative or attending phys-ician. Twelve first event stroke patients wereexcluded, eight because they died before as-sessment was possible, three because ofhaematological disorders, and one due to mitralvalve disease with atrial fibrillation. 198 (96%)of the 207 eligible control subjects agreed toparticipate.

RECALL OF PREVIOUSLY RECORDED CIGARETTESMOKING HISTORYCigarette smoking status had been recorded inmedical records prior to the study assessmentin 97 cases and 116 controls. Of the 75 cases

139

=

on March 26, 2020 by guest. P

rotected by copyright.http://jech.bm

j.com/

J Epidem

iol Com

munity H

ealth: first published as 10.1136/jech.51.2.138 on 1 April 1997. D

ownloaded from

Table 1 Relation between current and former cigarette smoking and stroke

Cigarette smoking status (OR (95% CI))

Never Ex>19y Ex 1-19y Ex 1-9y Current Test for trend

Age and sex adjusted 1.0 1.85 (0.61, 5.80) 2.44 (0.78, 8.78) 3.43 (1.37, 8.76) 2.57 (1.28, 5.51) p = 0.002Multiple risk factor adjusted* 1.0 3.49 (1.52, 8.03) 2.29 (0.86, 6.10) 2.92 (1.14, 7.43) 3.02 (1.48, 6.13)Cases/controls 22/69 15/20 12/21 27/28 49/60

* Covariables included in the logistic models were; age group, sex, race (black/non-black), vigorous exercise aged 15-25 and dietary saturated fat (four levels).

with a record of cigarette smoking, 72 (96%)could recall a history at interview. Amongst the81 controls with a recorded cigarette smokinghistory 79 (98%) could recall the history atinterview.

CIGARETTE SMOKING AND STROKECurrent smokers were grouped together foranalysis but former smokers were subdividedinto three groups by number of years sincecigarette smoking cessation (1-9, 10-19, and>19). This was to enable assessment of howsoon after stopping smoking the risk of strokedeclines to that of the never smoker. The ageand sex adjusted odds ratios indicated the riskof stroke in current smokers was more thandouble that in never smokers (fig 1, table 1).The decrease in the risk with time since stop-ping smoking appears to be slow.

Multiple risk factor adjustment including vig-orous exercise aged 15-25 and dietary saturatedfat (4 levels) along with age group, sex, and

* Age/sex adjustedo Multiple risk factoradjusted

Never

_±

>19 10-19

Years since stopping-

Cigarette smoking status

Figure 1 Relation between cigarette smoking and stroke (covarlogistic models were; age group, sex, race [blacklnon-black], vigiand dietary saturated fat ifour levels]). Bars indicate 95% conj

Table 2 Relation between cigarette smoking and stroke by sex

Cigarette smoking status (OR (9

Never Former

Males:Age adjusted 1.0 1.99 (0.61, 6.80)Multiple risk factor adjusted* 1.0 2.87 (0.91, 9.10)Cases/controls 6/17 36/48

Females:Age adjusted 1.0 3.50 (1.30, 10.4)Multiple risk factor adjusted* 1.0 3.95 (1.51, 10.3)Cases/controls 16/52 18/21

* Multiple risk factor adjustment included age group (4 age grouvigorous exercise aged 15 to 25 (yes/no) and dietary saturated fat (

race (black/non-black) tended to increase thesize of the risk of both current and formersmokers (fig 1, table 1). Smokers tend to bethinner and less hypertensive than neversmokers but these observations may reflect aneffect of smoking itself. Including markers ofbody fat and hypertension history in theadjusted analysis, therefore, would tend tofalsely elevate the apparent risks of smoking.Separate univariate adjustment by the followinghad no appreciable effect on the relation be-tween cigarette smoking and stroke: familyhistory of stroke, social class, alcohol con-sumption, diabetes history, and cardiac isch-aemia history. The hazards of smoking wereseen in both sexes (table 2) and in all agegroups (results not presented). When the datawere analysed excluding surrogate responders,leaving 95 cases and 196 controls, there wasno major change in the age/sex adjusted oddsratios; current - 2.84, ex 1-9 years - 4.00, ex10-19 years - 3.88, ex >19 years, 1.52. Whenproven cerebral infarction was examined sep-arately using the same methods of analysis asfor overall stroke the results were similar tothose for all stroke (results not presented).A possible explanation for the high risk in

those recently stopping smoking was that theyhad given up because of diagnosed vasculardisease other than stroke. 18 (67%) of the27 stroke patients who had stopped smokingwithin the previous 10 years had a historyof either; ischaemic heart disease, peripheralvascular disease, transient ischaemic attack,hypertension, or diabetes mellitus. This com-pared with 66 (33%) of the 198 controls.

LIFESTYLE COMBINATIONS AND POPULATION

ATTRIBUTABLE RISK FRACTION1-9 Current Because lifestyle related risk factors such as

smoking, lack of exercise, and excessive bodyfat are closely interrelated it was felt multiplying"independent" effects would be more likely to

riables included in the have mathematical appeal than to reflect trueorousexercise aged 15-25 biological effects. Body fat, for example, isidence intervals, influenced by both physical exercise and ci-

garette smoking. For this reason lifestyle relatedrisk factors in combination were assessed inorder to examine their possible overall pre-

?5% CI)) ventive potential for stroke. As cigarette smok-Curent ing remains the best established lifestyle related

risk factor it was used to form the basis of other2.78 (0.86, 9.74) combinations. If both former (prevalence in3.23 (0.99, 10.5) controls=35%) and current (prevalence in

controls = 30%) smoking were used to estimate2.44 (0.98, 6.67) the population attributable risk fraction, ci-

18/27 garettes appeared to be responsible for ap-

pa), race (black/non-black), proximately half of all strokes in the population4 levels) as covariables. studied (table 3). The risk fraction con-

9

8

7

6

5

4

3

0

C/u

0

2

1

0

140 Shinton

WI_-_ -

on March 26, 2020 by guest. P

rotected by copyright.http://jech.bm

j.com/

J Epidem

iol Com

munity H

ealth: first published as 10.1136/jech.51.2.138 on 1 April 1997. D

ownloaded from

Lifelong exposures and stroke

Table 3 Relative risk and population attributable risk fraction for stroke of combinations of cigarette smoking history,exercise history, and history of obesity

Variable OR* (95% CI) Caseslcontrols Population attributablerisk fraction (95% CI)

Cigarette smoking Never 1.0 22/69Current/former 2.6 (1.4, 5.2) 103/129 0.49 (0.22, 0.67)

Maximum reported BMIt Quartiles 1-3 1.0 78/146Upper quartile 1.7 (1.0, 2.8) 46/50 0.16 (0.02, 0.28)

Vigorous exercise aged 15- 25 Yes 1.0 55/130No 3.0 (1.8, 5.5) 70/68 0.33 (0.16, 0.46)

Cigarette smoking and/or lack of vigorous No 1.0 8/33exercise aged 15-25

Yes 2.7 (1.1, 7.0) 117/165 0.62 (0.20, 0.82)Cigarette smoking and/or upper quartile No 1.0 8/45maxirnum reported BMIt

Yes 4.3 (1.9, 12) 116/151 0.72 (0.43, 0.86)Cigarette smoking and/or lack of vigorous No 1.0 3/23exercise aged 15 to 25 and/or upper quartilemaximum reported BMIt

Yes 5.3 (1.4, 23) 121/173 0.79 (0.32, 0.94)

BMI = body mass index* Odds ratio adjusted for age and sex.t Data restricted to 124 cases and 196 controls with recall of weight. Upper quartile; males - >30.5 kg/mi, females - >31.3kg/m .

tributions offormer (0.26) and current smoking(0.24) were roughly equal.Data from this study have been published

previously indicating physical activity at allstages of adult life offers protection againststroke.7 Cohort studies examining activity levelsat the time of assessment are consistent withthese observations.6 History of exercise aged15-25 years was added to the analysed com-bination to represent physical activity both be-cause it reflected lifelong exercise habits andbecause it was less likely to be influenced byillnesses in later life.7 This combination of ci-garette smoking and/or lack of exercise in-creased the attributable risk fraction to 62%(table 3). The age/sex and multiple risk factoradjusted (for age group, sex, race, maximumreported body mass index and dietary saturatedfat) odds ratios for this combination were sim-ilar - 2.73 and 2.83 respectively.

Overweight and obesity are now more widelyaccepted as stroke risk factors which reflectdietary and exercise habits.58 In this study themost appropriate marker of obesity and over-weight appeared to be recall of lifelong max-imum body mass index. A combination,therefore, of cigarette smoking and/or max-imum reported body mass index in the upperquartile was examined. This combination ac-counted for 72% of all strokes (table 3). Theage/sex and multiple risk factor adjusted (forage group, sex, race, exercise aged 15-25, anddietary saturated fat) odds ratios for this com-bination were, again, similar - 4.31 and 4.49respectively. If upper quartile of subscapularskinfold thickness was added to cigarette smok-ing in place of maximum reported body massindex similar figures were reached: odds ratiofor cigarette smoking and/or upper quartile ofsubscapular skinfold thickness compared withthinner never smokers - 3.45 (1.55, 7.43). Itis of note that only one stroke patient reportednever smoking and remaining in the lowestmaximum reported quartile ofbody mass index(<25 kg/M2).Combining cigarette smoking, a history of

obesity and lack of exercise increased the popu-lation attributable risk fraction towards 80 %(table 3). As before the age/sex and multiple

risk factor adjusted (for age group, sex, race,and dietary saturated fat) odds ratios for thiscombination were similar - 5.29 and 5.94 re-spectively. It is, finally, worth noting that nostroke patient reported a combination of neversmoking, never being overweight (>25 kg/M2)and taking regular vigorous exercise aged 15-25.

DiscussionThe results indicate the substantial potentialfor stroke prevention which could be achievedthrough easily identifiable measures related tolifestyle.

CIGARETTE SMOKINGThe results of this study confirm that currentcigarette smokers are two to three times morelikely to have a stroke than those who havenever smoked. The study, also, suggests thatthe risks of cigarette smoking for stroke lastone, two, or possibly more decades after sub-jects have given up smoking. The calculatedpopulation attributable risk for a history ofcigarette smoking indicates that around half ofall strokes in the age group and populationstudied could have been prevented if peoplehad never started smoking.The persistent nature of the risk of cigarette

smoking for stroke in this study is consistentwith other studies ofboth stroke in those under75 years of age and cerebral infarction.'2 An-other British case-control study found a similarslow decline in risk but a smaller overall riskamongst female former smokers.'5 The findingsare also consistent with the results ofthe Britishregional heart study for myocardial infarction- another disease related to atherosclerosis.'6For stroke the latter study revealed only aminimal excess risk in former smokers, al-though confidence intervals would overlap withthose reported here.' The weaker impact offormer smoking seen in the nurses cohort maybe partly explained by the younger age/fewerpack years that would have been prevalentamongst their former smokers.4

141 on M

arch 26, 2020 by guest. Protected by copyright.

http://jech.bmj.com

/J E

pidemiol C

omm

unity Health: first published as 10.1136/jech.51.2.138 on 1 A

pril 1997. Dow

nloaded from

Shinton

The risks of current and past smoking aremore clearly seen in this study than in someothers. There are several possible reasons forthis that relate to difficulties that can arise withcohort and hospital based case-control studies.Accurate case definition is a potential problemin large cohort studies as clinical informationfrom possibly less reliable sources has to beused, for example; death certificates, patientrecall or a review of clinical records. This is aparticular problem as age increases - mostcohort studies have included a substantial pro-portion of elderly stroke patients. Mis-classification in these studies would havetended to attenuate the real risks. Anotherproblem raised by cohort studies is the in-complete nature of the exposure history; cur-rent smokers at screening may be former-smokers by the time the stroke occurs. Hospitalbased case-control studies may also under-estimate the true risks of cigarette smoking dueto selection biases operating on the controlsubjects. This is because it is quite plausiblethat those with a history of cigarette smokingare more likely to be under some form ofhospital care and thus available to act as acontrol. This study offered a detailed as-sessment of all cases in a defined populationby a physician, an up to date history of exposureto cigarette smoking and minimal selectionbiases because of the almost complete responserate and the community based nature of thestudy. Examination of smoking history re-corded in general practice/hospital notes priorto assessment in the study provided little evi-dence of recall bias. Pipe and cigar smokingwill not have exerted any appreciable influenceon the results as these habits were virtuallyunknown in this population (reported by onecase and one control).

COMBINATIONS OF LIFESTYLE RELATED RISKFACTORSTaken together, the combination of cigarettesmoking, excessive body fat, and lack of regularvigorous exercise may have accounted for be-tween 70 and 80 % of stroke cases in thepopulation studied. The size of the preventivepotential indicated by these results has not beenpreviously described. This estimate can onlybe a guide as the 95% CI are wide and theexcluded stroke patients have not been in-corporated into the figure.

Potential sources of error in a study such asthis may have come from misclassification ofthe exposures and confounding, as well as thelimited size ofthe study. There is good evidenceon the accuracy of cigarette smoking clas-sification in this study and data previously pub-lished are reassuring as regards misclassificationin relation to lifelong body fat.8 It is moredifficult to be confident about reports of pre-vious exercise but the contribution of the latterto the overall combined population attributablerisk fraction is small. Any unbiased mis-classification in the assessment of exercise (po-tentially considerable) will have led to anunderestimate of its impact. If a biased tend-ency existed for stroke patients to reply in the

negative about past exercise the attributablerisk fraction could have been falsely inflated.

It seems unlikely that the results were dueto confounding by any well established riskfactors. If unadjusted, age and sex adjusted ormultiple risk factor adjusted odds ratios werecalculated for the three combined lifestyle re-lated factors the odds ratio remained the same.Although hypertension, diabetes, and cardiacdisease are known to be associated with bothobesity and stroke, it is difficult to make thecase that these risk factors should be regardedas potentially important confounders. Hyper-tension, diabetes, and cardiac disease are likelyto lie in the causal pathway of stroke as clinicalsyndromes which result from both genetic andenvironmental influences. This study was dir-ected towards the latter potentially avoidablerisks which could be more easily identifiable bythe population at large. Hypertension, diabetes,and heart disease all require identification bya clinician - cigarette smoking, overweight, andinactivity do not. Data from the prospectivestudies collaboration do indicate that diastolicblood pressures greater than 70 mmHg, if con-sidered causal (data were unadjusted), accountfor around 60% of all stroke cases.17 A recentreport from the Framingham cohort, however,notes that around 70% of hypertension is dir-ectly attributed to obesity.'8 These inter-re-lations mean that high population attributablerisk fractions for both hypertension and thelifestyle related factors addressed in this studyare likely to be compatible. For these reasonsresults adjusted for hypertension, diabetes, andcardiac ischaemia were not computed. Roth-man argues it is inappropriate to adjust forvariables in the causal pathway except to ex-plore the mechanisms involved.'9 The size ofany contribution of specific dietary intakes (egsalt), interuterine nutrition, or genetic pre-disposition to the population attributable riskfraction for stroke remains unclear. A degreeof confounding by these variables is possiblebut unlikely to be substantial. Problems withassessment of these variables and limitationson the interpretation of blood pressuremeasurement in patients presenting with anacute stroke mean some of these inter-re-lationships are not easily explored in studies ofa case-control design.The results of this study, particularly if taken

together with the results of the Whitehall study,suggest that there is an interaction betweenthe effects of cigarette smoking and body fat.5Dangers, therefore, would exist in merely add-ing or multiplying these two risks together toobtain a combined effect. A more reliable (al-though statistically low powered) approachwas, therefore, adopted by examination of therisk factors in combination. An approach tocombinations using regression models would,perhaps, tend to underestimate the preventivescope of these two potentially avoidable haz-ards.The data of American college alumni, using

combinations of cigarette smoking and over-weight are consistent with the findings of thisstudy - smoking/overweight students had astroke morbidity ratio of 2.6 compared with

142 on M

arch 26, 2020 by guest. Protected by copyright.

http://jech.bmj.com

/J E

pidemiol C

omm

unity Health: first published as 10.1136/jech.51.2.138 on 1 A

pril 1997. Dow

nloaded from

Lifelong exposures and stroke

other colleagues.20 The stroke patients in thealumni study, however, were self reported andexposures had only been measured once,around the age of 20.The observation that a history of inactivity

does not appear to contribute much over andabove a history of cigarette smoking is worthcomment. There was a trend for the benefitof exercise to be greater amongst cigarettesmokers7 but the small number of never smok-ing stroke patients in this study means that it isdifficult to draw conclusions until the potentialbenefits of exercise in a larger never smokingpopulation have been assessed.The finding that a substantial proportion of

strokes in the age group studied could probablybe prevented by avoidance of cigarettes, ex-cessive body fat, and inactivity has implicationsfor health promotion. It conveys a positivemessage to younger people about the potentialto avoid premature cerebrovascular disease. Itis also consistent with recent findings aboutthe limitations of advising and encouragingpeople to stop smoking and lose weight in lateradult life.2' The results support the currentpreventive strategy addressing these risks butadd weight to the case for concentrating effortsat those in younger age groups. Working to-wards creating an environment in which it iseasy to adopt these health promoting patternsof living would seem to offer large gains forstroke prevention.The results highlight particular concern for

long term cardiovascular health from the in-creasing prevalence of obesity in western popu-lations and the associated decrease in activitylevels. The fall in stroke mortality attributed todeclining smoking in men, better managementof hypertension, anticoagulation for atrial fib-rillation, carotid endarterectomy and aspirinuse could be reversed. The situation is evenmore alarming in developing countries wheresmoking is also increasingly prevalent.

I thank the patients and staff of the 11 participating generalpractices (see below) for their cooperation, Professor D GBeevers and Dr I J Perry for their advice, and Dr G Sagar for

computed tomography. The project was funded by the WellcomeTrust. The participating practices: Bearwood Road Surgery;Smethwick, Birchfield Medical Centre, Cape Hill MedicalCentre, Church Lane Surgery; Handsworth, Hamstead RoadSurgery; Handsworth /Great Barr, Lee Bank Health Centre,Newtown Health Centre, Plough and Harrow Medical Centre;Ladywood, Sherwood House Surgery; Bearweed, Tower HillHealth Centre, Victoria Health Centre; Smethwick.

1 Shinton R, Beevers G. Meta-analysis of the relation betweencigarette smoking and stroke. BMJ 1989;298:789-94.

2 Donnan GA, Adena MA, O'Malley HM, McNeil JJ, DoyleAE, Neill GC. Smoking as a risk factor for cerebralischaemia. Lancet 1989; ii:643-47.

3 Shaper AG, Phillips AN, Pocock SJ, Walker M, MacfarlanePW. Risk factors for stroke in middle-aged British men.BMJ7 1991;302:1111-5.

4 Kawachi I, Colditz GA, Stampfer MJ, Smoking cessationand decreased risk of stroke in women. _AMA 1993;269:232-6.

5 Shinton R, Shipley M, Rose G. Overweight and stroke inthe Whitehall study. J Epidemiol Community Health 1991;45:138-42.

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

7 Shinton R, Sagar G. Lifelong exercise and stroke. BMJ1993;307:231-4.

8 Shinton R, Sagar G, Beevers G. Body fat and stroke:Unmasking the hazards of overweight and obesity. J Epi-demiol Community Health 1995;49:259-64.

9 Shinton R, Sagar G, Beevers G. The relation of alcoholconsumption to cardiovascular risk factors and stroke.The west Birmingham stroke project. J Neurol NeurosurgPsychiatry 1993;56:458-62.

10 World Health Organisation. Cerebrovascular diseases - pre-vention, treatment and rehabilitation. Technical report seriesno. 469. WHO: Geneva, 1971.

11 Weiner JS, Lourie JA. Human Biology: a guide tofield methods.IBP Handbook No.9. Oxford: Blackwell Scientific Pub-lications, 1969:27-52.

12 EGRET. Reference manual. Seattle: Statistics and epi-demiology research corporation, 1990.

13 Breslow NE, Day NE. Statistical methods in cancer research.Volume I - The analysis of Case-control studies. Lyon: Inter-national agency for research on cancer, 1980:122-159.

14 Schlesselman JJ. Case-Control Studies. Design, Conduct andAnalysis, New York 1982:220-222.

15 Thompson SG, Greenberg G, Meade TW. Risk factors forstroke and myocardial infarction in women in the UnitedKingdom as assessed in general practice: a case-controlstudy. Br Heart j 1989;61:403-9.

16 Cook DG, Shaper AG, Pocock SJ, Kussick SJ. Giving upsmoking and the risk of heart attacks. A report from theBritish Regional Heart Study. Lancet 1986; ii:1376-80.

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

18 Kannel WB. Blood pressure as a cardiovascular risk factor.Prevention and treatment. JAAMA 1996;275: 1571-1576.

19 Rothman KJ. Modern Epidemiology. Little Brown and Co-pmany, Boston 1986:89-94.

20 Paffenbarger RS, Wing AL. Chronic disease in former col-lege students: early precursors of non-fatal stroke. Am JEpidemiol 1971;94:524-30.

21 Multiple Risk Factor Intervention Trial Group. Multiple riskfactor intervention trial: risk factor changes and mortalityresults. _AMA 1982;248:1465-77.

143 on M

arch 26, 2020 by guest. Protected by copyright.

http://jech.bmj.com

/J E

pidemiol C

omm

unity Health: first published as 10.1136/jech.51.2.138 on 1 A

pril 1997. Dow

nloaded from