J7ournal of Epidemiology and Community Health 1991; 45: 4-10

Impact of work environment on cardiovasculardiseases in Denmark

Ole Olsen, Tage S Kristensen

AbstractStudy objective-The aim was to estimate

the quantitative impact of workingconditions on cardiovascular diseases inDenmark.Design-The study was based on a review

of recent epidemiological researchpublications in which relative risks wereestimated and risk factor prevalences weredetermined. The impact of workingconditions was quantified by means ofaetiological fractions.Setting-The aetiological fractions were

estimated on the Danish population.Main results-16% of the premature

cardiovascular mortality in men and 22% inoccupationally active women is avoidablethrough interventions in the workenvironment. If "sedentary" work isincluded in the occupational risk factors,the aetiological fractions reach 51% for menand 55% for women. Taken separately,the major aetiological fractions forcardiovascular risk factors at work arerespectively (women in parentheses) 6%(14%) for monotonous high paced work, 7%(7%) for shift work, and 2% (2%) for passivesmoking.Conclusions-The aetiological fractions

show that working conditions play aconsiderable role in cardiovasculardiseases. Furthermore they might widen thefocus of preventive cardiology frominterventions directed only at individualand lifestyle risk factors to interventionsdirected also at working conditions.

Institute of SocialMedicine, Universityof Copenhagen,Blegdamsvej 3,DK-2000Copenhagen N,DenmarkO OlsenT S Kristensen

Correspondence to:Dr Olsen

Accepted for publication July1990

The epidemiology of cardiovascular disease hasfor many years focused on the significance of the

conventional individual risk factors for the

incidence of cardiovascular diseases, while factors

in the social and occupational environment have

attracted less interest. Looking at occupationalmedicine we find a similar picture: much researchhas been performed on work and cancer,

reproductive failures, musculoskeletal diseases

etc, but little attention has been paid to

cardiovascular diseases.In 1989 two systematic and extensive reviews of

the epidemiologic literature on work environment

exposures and cardiovascular diseases were

published.1 2 In these reviews it was concluded

that the research on work environment and

cardiovascular diseases-although at many pointsstill insufficient and of unsatisfactory quality-has today reached a level that makes it possible to

identify a number of risk factors in the work

environment with reasonable certainty. Anoccupational risk factor for cardiovasculardiseases is in this article defined as any factor inthe work environment that is causally related tocardiovascular diseases, that is any factor thatdirectly or through one of the conventional riskfactors (such as blood pressure) increases the riskof getting a cardiovascular disease.The aim of the present paper is to some extent

similar to the aim of the paper by Fox andAdelstein: "Occupational mortality: work or wayof life?".3 But the calculations in the present paperare-though still crude-based on more explicitcriteria. The purpose of the present paper is toestimate the relative impact of the workenvironment on the incidence of prematurecardiovascular disease. This is done by estimatingthe aetiological fractions of the relevant workenvironment factors for men and women inDenmark. By substituting the estimates used forthe different exposures with estimates for othercountries the results can be applied to thesecountries as well. The aetiological fractions can beconsidered as estimates of the potential benefit tobe gained by eliminating a given occupational riskfactor. This measure is therefore essential informulating a rational strategy for the preventionof cardiovascular disease.

MethodsMATERIALTwo kinds of information are needed to quantifythe influence of the work environment onpremature cardiovascular diseases in a particularcountry: (1) the risk associated with each riskfactor; and (2) the prevalence of each risk factor.In the following section the basis on which the riskis quantified in terms of relative risk is described.The subsequent section contains an assessment ofthe prevalence of the risk factors.

Origin of the applied relative risksIn the above mentioned reviews by Kristensen' 2

the methodological quality of the epidemiologicalstudies on work environment factors andcardiovascular disease was evaluated using fivecriteria: (1) control with the time dimension; (2)control with confounding variables; (3) controlwith selection processes; (4) quality ofmeasurements of exposures and diseases; (5)quality of design and statistical analysis. On thebasis of this evaluation each study was given a

score between "x" (very poor) and "xxxxx"(extremely good). The next step was to evaluatethe overall evidence (for causality) concerningeach potential risk factor. Here three criteria wereused: (1) the number of studies; (2) the quality ofthe studies (with more weight given to the best

on February 6, 2021 by guest. P

rotected by copyright.http://jech.bm

j.com/

J Epidem

iol Com

munity H

ealth: first published as 10.1136/jech.45.1.4 on 1 March 1991. D

ownloaded from

Work environment and cardiovascular disease

studies); (3) the consistency of the study results.This two step evaluation resulted in theclassification ofoccupational risk factors shown intable I. Here quantifications will be made for therisk factors in the two groups: "Very definite" and"Quite definite". The following approach doesnot assume consensus regarding the causalrelationship between each work environmentfactor and cardiovascular diseases. By presentingthe assumptions and calculations explicitly, it isthen possible for readers who disagree with thecategorisation of the relationships used here toconduct their own calculations.

Table I Classification of possible cardiovascular risk factors in the work environment.

Type of risk factor

Causal relation to Non-chemical Chemicalcardiovascular diseases

Very definite Physical inactivity at work Carbon disulphideNitroglycerin/lnitroglycol

Quite definite Work strain (high demands Leadaand low influence) Passive smoking

Shift workNoisea

Quite possible CobaltArsenicCombustion products

Possible Heatb OrganophosphatesIrradiation DinitrotoluenePower frequency magnetic Antimony

fields BerylliumLow frequency noise Cabon monoxideb

Probably no relationship Microwaves CadmiumColdb Organic solventsc

a Increases the risk for cardiovascular disease through increased blood pressure.b High level exposure may be fatal, especially when combined with other risk factors.c High level exposure may cause cardiac arrhythmia and sudden death.

When (or if) the causal relationships areaccepted the next step is to quantify the strengthsof these relationships. This has to some extentbeen done by Kristensen' 2 by means of relativerisks. Of course this is a rough and simplisticmeasure: the relative risks might depend on age,on other risk factors, on the specific dose, etc. Butit has been judged to be an adequate simplificationin order to reach a first quantification, which canbe used in the discussion of different preventivestrategies. The relative risks proposed int 2 arebased on a thorough evaluation of theepidemiological quality of the studies and aneducated "weighted" average of the relative risksin the individual studies taking into accountmethodological strengths and weaknesses,biological plausibility, consistency amongstudies, etc. The influence of occupational riskfactors can operate either directly on the incidenceof cardiovascular diseases or throughintermediate individual risk factors. Theindividual risk factors, blood pressure and serum

cholesterol, were regarded as being causal riskfactors for cardiovascular diseases. So was

smoking, but as both starting smoking andstopping smoking are determined by many othersocial and cultural factors than workingconditions, smoking was not considered a

mediator between work environment andcardiovascular diseases. Consequently an

occupational factor is considered a risk factor forcardiovascular diseases if it either raises the bloodpressure or the level of serum cholesterol, or

directly raises the incidence of cardiovascular

diseases. The relative risks are presented in tableII. The relative risks for work strain and for noiseare not explicitly stated in reference.' Theestimates of the former vary between 13 andabout 4 in different studies, with a slight tendencyfor higher values to be found in the better studies.Here the relative risk is fixed at 2. The quality ofthe epidemiological studies concerned with noiseis relatively lower and the bias is mostly towardsthe null hypothesis. The causal relationshipbetween noise and cardiovascular diseases seemsto be through an increase in blood pressure of 5 to15 mm Hg among those exposed. A reasonableestimate might be a relative risk of 1 2, which isused here.As mentioned above, the main interest is in

premature cardiovascular disease. Butcardiovascular diseases have many manifestationsand are categorised in many different ways andsubgroups: angina pectoris, ischaemic heartdisease (lethal or non-lethal), stroke, etc. Therelative risks and therefore the calculations basedon them have to be interpreted as relevant to abroad perception of cardiovascular diseases. Wehave chosen-or to some extent been forced-todo this because the studies reviewed' 2 are notconsistent with regard to end points. The choice isto some extent supported by Kannel,4 p 282:"The major atherosclerotic diseases share mostprecursers and a common pathogenesis ... Thesefacts strongly suggest that prophylactic measuresdesigned to prevent one may prevent the others aswell". With regard to early mortality (below theage of 70), which is of particular concern, mostcardiovascular disease is ischaemic heart disease.In Denmark this proportion is 7500.The prevalence of the occupational risk factorsBesides the relative risks the prevalence figures ofthe occupational risk factors are needed toquantify the impact on cardiovascular diseases.Contrary to expectations the task of obtainingthese figures turned out to be a difficult one. Thechemical exposures of interest here are rare inDenmark (with lead being perhaps the onlyexception) and an inquiry to the Danish NationalInstitute of Occupational Health and the DanishLabour Inspection Service revealed that the dataconcerning the prevalence of these chemicalexposures are quite poor. These risk factors havetherefore been omitted in our calculations. Theestimated prevalences of the rest of theoccupational risk factors have been obtained invarious ways, which will be described below.The population of interest is the occupationally

active part of the population which consists of 1 5million men aged 15-74 and 1-25 million womenaged 15-74. Overall within this age range, thereare 1 9 million of each sex. The figures presentedbelow are prevalence figures, ie, they indicate howmany people are being exposed at a given point intime. But mobility in the labour market meansthat substantially larger numbers have beenexposed at some time in their lives.Monotonous high paced work-The prevalence

of monotonous high paced work has beenestimated according to the numbers employed inthe relevant job categories, such as drivers:55 000; waiters: 5000; etc. In total these figuressum to 110000 men. However not everyone in

5

on February 6, 2021 by guest. P

rotected by copyright.http://jech.bm

j.com/

J Epidem

iol Com

munity H

ealth: first published as 10.1136/jech.45.1.4 on 1 March 1991. D

ownloaded from

Ole Olsen, Tage S. Kristensen

these categories necessarily has monotonous highpaced work, so the figure is reduced to 90 000,which is equivalent to 6 00 of occupationally activemales. Among women, high paced work is morewidespread. An estimation similar to that for menleads to an estimated prevalence amongoccupationaly active women of 1600. Comparedto the definition of monotonous high paced workin the research referred to in', the aboveenumeration is based on a more restrictivedefinition.

Working hours-In a representative survey ofthe Danish population aged 20-69 yearsconducted by the Danish Institute of SocialResearch in 1976,5 5000 people were asked abouttheir working hours. This showed that theprevalence of shift work and other unsocialworking hours was approximately 20), for bothsexes.Noise-Regarding noise, no good surveys have

been performed in Denmark. Based on Dutch andNorwegian investigations, the Danish LabourInspection Service estimates that 150 000 personsare exposed to sound levels above 90 dB. Thisis a very imprecise estimate. Despite theinappropriateness of measuring sound instead ofnoise as a risk factor for cardiovascular diseases,'much research has used a limit of 90 dB. As moremen than women are exposed to noise, it has beendecided to use the figures of 100 000 men and50 000 women equivalent to 70o and 40()respectively exposed to noise among the workingpopulation.

Passive smoking-No surveys regarding passivesmoking at the workplace existed prior to thepresent investigation. Through the DanishNational Institute of Social Research, a surveywas performed in 1987,6 120) of working menwere exposed to passive smoking and 1300 of thewomen. Passive smokers include only non-smokers; the percentage is of all persons(including smokers). To be classified as passivesmokers, the non-smokers had to be exposed thewhole day or part of the day and not just atmeetings of during breaks.

Physical activity-Physical activity has to beboth vigorous7 8 and dynamic to reduce the risk ofcardiovascular diseases. Static muscle work doesnot have any beneficial influence on the risk ofcardiovascular diseases. As society has changed, itis hard to find any job categories with sufficientlydynamic muscle work. Most work is sedentary orstatic muscle work: office work, transportation,surveillance, etc. Postmen and ballet dancers areamong the few with sufficient dynamic musclework. We have estimated that at most 10 ( of theworking force have enough dynamic, aerobicmuscle work during working hours. In otherwords, 900O of the working population is judgedto have "sedentary" work-the quotation marksindicating that sedentary work is defined toinclude static muscle work. As people with highphysical activity during leisure time will have onlya minor beneficial effect from physical activityduring working hours, figures showing theprevalence and intensity of physical activity inleisure time are needed. In a recent survey9 theDanish Institute of Clinical Epidemiology hasestimated that approximately 200) of theoccupationally active population with "sedentary

work" indulge in sports, digging the garden or thelike, for at least four hours a week.

PROCEDURE

The impact of working conditions on publichealth is expressed by means of the aetiologicalfraction. The concept was introduced by Levin in1953,10 the term "etiological fraction" and theabbreviation, "EF", by Miettinen in 1974."1 Theconcept has sometimes 12 been defined by thefollowing simplistic formula: (RR-1)*b/[(RR-1)*b + 1], where RR is the relative risk and b is thefraction exposed. This is, however, in manycircumstances inappropriate. 13 It is far morecorrect to state a verbal definition, as done byseveral authors.'0 11 14 The phrasings aresomewhat different but the meaning is the same asthat of the following definition:

The aetiological fraction is a concept defined for aspecific disease or group of diseases (here:cardiovascular diseases) in a given population(here: the Danish) and has the followingtheoretical meaning: "The aetiological fraction isthat proportion of the disease that would not haveoccurred had the risk factor not occurred in thepopulation."If a certain factor is not a causal agent, but only arisk indicator, the aetiological fraction is zero, asremoval of the factor would not change the diseaseincidence. Because it is impossible to reachabsolute certainty as to whether a factor has acausal relationship to a disease or not, we cannever be certain that the numerical calculation ofan aetiological fraction is correct. However, byagreeing upon causality and the strength of thecausal associations as well as the distribution ofrisk factors in the population, it is possible toobtain an estimate of the aetiological fraction. Asthe definition of the aetiological fraction is madein the context of a hypothetical situation ("if therisk factor had not occurred"), caution is neededin the interpretation of the numerical calculationand in the translation into preventive potential. Ifeverybody stopped smoking today neither theprevalence nor the incidence of cardiovasculardiseases would drop immediately to the levelfound in persons who have never smoked.

Ideally we ought to know both the simultaneousdistribution of all risk factors in the populationand the risk associated with each combination ofrisk factors appearing in the population in order tocalculate the aetiological fraction. This is not thecase, so we shall have to make some assumptions.

All the occupational risk factors are assumed tobe simultaneously independent. Furthermore theoccupational risk factors and the conventional riskfactors which are not intermediate are alsoassumed to be independent. The only exception isactive and passive smoking, since passive smokingis defined in such a way that it occurs only amongnon-smokers.The strength of associations has already been

reported by means ofone relative risk for each riskfactor. This implies that a multiplicative model isassumed, or in other words that, eg, the relativerisk of working in noise is the same irrespective ofwhether the work is "sedentary" or not, shiftwork or not, etc. or whether the exposed person isa smoker, a vegetarian, etc. The reported relative

6

on February 6, 2021 by guest. P

rotected by copyright.http://jech.bm

j.com/

J Epidem

iol Com

munity H

ealth: first published as 10.1136/jech.45.1.4 on 1 March 1991. D

ownloaded from

Work environment and cardiovascular disease

risks have to be interpreted as adjusted relativerisks which means that, in the "educatedaveraging" of the estimated relative risks, mostemphasis has been put on studies with the bestcontrol for confounding.As the occupational risk factors in general have

not been recognized, the studies of oneoccupational risk factor seldom consider otheroccupational factors explicitly. But in a way theyare implicitly kept under control because thestudies are either randomized or restricted in sucha way that the occupational risk factors not underconsideration are often the same in the exposed(index) and the unexposed (reference) group.With the assumptions made it is possible to use

the simplistic formula to estimate the aetiologicalfractions. Two exceptions have to be maderegarding the straightforward application of theformula. When the aetiological fraction for"sedentary" work is estimated, the value of b hasto be defined somewhat differently, because thosepersons who are physically active in their leisuretime are assumed to gain no positive effect offurther exercise. So b is set at 72°,, namely the900o with "sedentary" work minus 180o (200o of90%0) having "sedentary" work but beingphysically active in their leisure time. Theaetiological fraction for passive smoking iscalculated according to the following expression:

(RRp-l)*bp / [ba*RRa + bp*RRp + (1-bp-ba)]where RRp and RRa are the relative risks forpassive and active smoking respectively, and bpand ba are the fractions of passive and activesmokers. The values RRa=2 5 and ba = 50° areused in the calculations for both sexes.

Additionally aetiological fractions for "allworking conditions" are reported. These requirea minor extension of the definition of aetiologicalfractions to handle the fraction of a set of riskfactors and not just of a single risk factor.'4 Thedefinition then reads: "The aetiological fraction isthat proportion of the disease that would not haveoccurred, had the set ofrisk factors not occurred inthe population". With the assumptions madeabout the mutually independent distribution ofthe occupational risk factors and theirmultiplicative effect, it is possible to calculate theaetiological fraction for a set of risk factors (herecalled a, b,. . ., n) according to the followingformula: 10EF(all) = 1-[I-EF(Note that the aetiol4risk factors is not th

Table II Aetiological fractions of work environmcardiovascular diseases in Denmark.

Prevalence ofexposures (0

Risk factor Men Women

Monotonous, high paced workShift work, etcNoiseChemical exposuresPassive smoking"Sedentary" work

6 1620 207 4

low low12 1390a 90a

All occupational risk factorsAll occupational risk factors except "sedentary" worka But only those 72", who are also physically active duringfrom physical activity at work.

fractions involved. Nor is there any reason forexpecting this'5 16 as many events can beprevented in more than one way, but never morethan once.

ResultsThe relative risks, the prevalence of exposures,and the calculated aetiological fractions for eachrisk factor are presented in table II. Additionallytwo proposals for the aetiological fraction for "allworking conditions" are reported. The first ofthese is calculated for the set of all occupationalrisk factors. The second one is calculated for thesame set, excluding "sedentary" work. Thereason for calculating the second aetiologicalfraction is the possibility of compensating for"sedentary" work by being physically active inleisure time. It is not possible in the same way toavoid the increased risk associated with the otherworking conditions in leisure time.As already mentioned the disease entity

"cardiovascular diseases" is only loosely definedin the above estimations. Therefore the reportedaetiological fractions can be translated intoabsolute figures in different ways, depending onthe purpose, eg, number of first myocardialinfarctions, number of disability pensions, etc.Morbidity is relevant when quantifying loss inquality of life or burden on the health care system,but morbidity is not very well registered.Mortality is described in death registers andmight serve as an example of the calculation ofpreventive potential.

In Denmark approximately 4200 males and1700 females die from cardiovascular diseaseseach year before age 70. Assuming that all themale deaths and 700o of the female deaths occur

among persons who have been occupationallyactive, this means that 2100 (5100 of 4200)premature deaths among males and 700 (5500 of700o of 1700) deaths among females could havebeen avoided if none of the occupational riskfactors had occurred. Ifwe disregard "sedentary"work, the numbers would be 700 males and 300females per year. With an age limit of 60 the latternumbers would be 200 and 100.

DiscussionTo our knowledge this is the firct time- thpimnbntA VJ %JLA L%11W1%A -6W 111D lb L11C. l.LbL LillllC LIIC llllPdbL

a)][1-EF(b)] ...[I-EF(n)] of the work environment on the incidence ofcardiovascular diseases has been estimated by

ogical fraction for the set of means of an aggregate aetiological fraction,e sum of all the aetiological although the question has been discussed

previously.3 Aetiological fractions for some of the7ent risk factors for premature specific occupational risk factors have been

calculated for populations similar to the Danish.Aetiological In Swedish calculations based on different followfractions 0, up studies"' monotonous high paced work had

Relative aetiological fractions in the range 3-7"0 for menrisks Men Women and 9-130. for women. These are comparable to2 0 6 14 our figures. For lower age limits the estimate for12

7 7

men was larger. In a study from New Zealand'8> 1.0 0-1 0 passive smoking in the work place had aetiological20 42 42 fractions of 10"o in men and 70, in women. This

51 55 is larger than our findings, and mostly due to16 22 application of a larger relative risk based on the

leisure time will gain any benefit assumption that exposure to passive smoking inthe work place is of greater intensity than

7

on February 6, 2021 by guest. P

rotected by copyright.http://jech.bm

j.com/

J Epidem

iol Com

munity H

ealth: first published as 10.1136/jech.45.1.4 on 1 March 1991. D

ownloaded from

Ole Olsen, Tage S. Kristetnsen

exposure to spousal smoking. Without thisassumption the estimates would be in closeragreement with ours. Aetiological fractions for"sedentary work" specifically have not beencalculated elsewhere, but figures for physicalinactivity in general in the US population'9 lead toaetiological fractions in the same order ofmagnitude as here.

For comparison, some aetiological fractionscalculated for non-occupational risk factors arelisted below without further discussion. Theestimates are based on samples which are notrepresentative for the countries, but relaterespectively to middle aged Swedish men fromGothenborg20 and US college alumi21:hypertension: 17(, and 900; smoking: 39()0 and25 " o; parental history: 1 2 "o and 11",; serumcholesterol (in the Swedish study only): 380,.Both sets of authors note that the sum ofaetiological fractions may exceed 100°0. This is infact the case in the Swedish study, but not in theAmerican.

Theoretical papers'5 16 also stress that the sumof the aetiological fractions may exceed 100",,.There are three important interpretations of this.First, the conceptual interpretation is that eventhough the aetiological fractions for some riskfactors are large, this does not exclude thepossibility that other risk factors may also have alarge influence on disease incidence. In otherwords the fact that occupational risk factors have alarge impact on cardiovascular diseases does notcontradict the fact that the traditional risk factors(smoking, cholesterol, etc) have a large impact.Second, the preventive interpretation is that it ispossible to choose between completely differentpreventive strategies to obtain similar decreases indisease incidence, when the sum of aetiologicalfractions is large, ie, interventions directed only atsmoking, only at the work environment, etc.Third, the scientific interpretation is that eventhough the sum has exceeded 100(0, the researchought not to stop. It is still possible that more riskfactors might be identified, and that these couldlead to cheaper or more convenient preventivealternatives.

Interpretations concerning only a singleisolated aetiological fraction also have to beconsidered. The definition of the aetiologicalfraction is made in the context of a hypotheticalsituation ("if the risk factor had not occurred"), socaution is needed in the interpretation of theestimate and in the translation into preventivepotential. Even if the risk factor were completelyremoved at a given moment, some time may elapsebefore the incidence level among the previouslyexposed reaches the level among the neverexposed. Nonetheless the fraction is informative.It provides information about the aetiology of thedisease in the actual population as well as what canmaximally be obtained by a change ofhabits in thenext generation; but it does not indicate whetherthis change can easily be obtained.

If some of the assumed causal relations are notagreed upon due to lack of "final proofs" ofcausality, the calculated aetiological fractions arestill of some interest. They can then be regardedas indicators of the possible size of impact onpublic health of the risk factor. A large aetiologicalfraction calls for further research to clarify

whether the observed associations really arecausal.

BIAS

It has to be stressed once again that the intrinsicinterpretation of the aetiological fractions iscrucially dependent on the assumed causalrelationships. If these assumptions are agreed, thenext question is whether the estimatedaetiological fractions are biased in any direction.We have tried to use conservative estimates of

both the relative risks and the fractions of exposedindividuals in the calculations in this paper. Theremaining question is whether the assumptionsmade introduce any particular bias. Theinteractions between the studied risk factors haveseldom been studied, but when this is done theinteractions have been shown to be small.22 23This is in accordance with the general experiencethat the multiplicative model is a good descriptionof most data. 15 Deviation from the assumption ofno interaction probably introduces only minorbias.

Possible deviations from the assumption ofindependent distribution of risk factors can be intwo directions. The risk factor is either more orless frequent in the presence of some other riskfactors. An example of the first would be that shiftwork was more prevalent among workers withmonotonous high paced work. An example of thelatter would occur if "sedentary" workers wereless exposed to noise than non-sedentary workers.When a multiplicative model is assumed and therisk factors are positively correlated, insertion ofan adjusted relative risk into the simplisticformula leads to an underestimation. Similarly anegative correlation would lead to anoverestimation. The first type of situation isprobably the more common, but it has not beenpossible to calculate the correlation between therisk factors in Denmark. Consequently the bias inthe estimate of aetiological fractions is notcalculated either.The most serious bias is probably due to the

way in which the estimation of the number ofpeople exposed has been performed and to theway in which the studied population has beendelimited. The proportion of people exposed isthe proportion of currently exposed people. Butfor some of the risk factors it might be morerelevant to use the proportion of the populationthat has been exposed. Unfortunately these figuresdo not exist. Monotonous high paced work andshift work are probably more common amongyounger people. A selection and healthy workereffect like this biases cross sectional studiestoward no association while cohort studies willstill show an association. This is in fact the case inthe studies of shift work.2 To what extent theincreased risk is fading out and how fast ithappens when people change from shift work toregular hours is not well described. We are notable to quantify the bias, but it might lead toconsiderable underestimation. We consider thisselection bias to be negligible in relation to passivesmoking and small in relation to noise.We have been dealing with the working

population and not the entire population.Applying the estimated aetiological fraction to theentire population would overestimate the impact

8

on February 6, 2021 by guest. P

rotected by copyright.http://jech.bm

j.com/

J Epidem

iol Com

munity H

ealth: first published as 10.1136/jech.45.1.4 on 1 March 1991. D

ownloaded from

Work environment and cardiovascular disease

of working conditions on cardiovascular health.This bias would be larger among women, as aconsiderable proportion has not worked regularlyoutside the home, and minor among men, asalmost all men are (or have been) working.

CONCLUSIONThe calculations presented in this paper give aprovisional impression of the quantitative impactof working conditions on cardiovascular diseases.The impact appears to be rather large. We feelthat work environment factors ought to play abigger role in cardiovascular epidemiologicalresearch in the future and also that thecardiovascular diseases should be acknowledgedas work related diseases in the same way as cancer,reproductive failures, lung diseases, and manyothers. It is evident that more-and better-research is needed, but this fact ought not be usedas an excuse to postpone preventive programmesaiming at improving the work environment.

The authors thank Lis Olsen for language revision andProfessor Gavin Mooney for helpful comments on thepaper. Supported by the Danish Work EnvironmentFund, the Danish Heart Foundation, and the Universityof Copenhagen.

1 Kristensen TS. Cardiovascular diseases and workenvironment-a critical review of the epidemiologicliterature on non-chemical factors. Scand J Work EnvironHealth 1989; 15: 165-79.

2 Kristensen TS. Cardiovascular diseases and workenvironment-a critical review of the epidemiologicliterature on chemical factors. Scand J Work EnvironHealth 1989; 15: 245-64.

3 Fox AJ, Adelstein AM. Occupational mortality: work or wayof life? J Epidemiol Community Health 1978; 32: 73-8.

4 Kannel WB. Some lessons in cardiovascular epidemiologyfrom Framingham. Am J Cardiol 1976; 37: 269-82.

5 Morkeberg H. Arbejdstidens placering. Meddelelse nr. 29.K0benhavn: Socialforskningsinstituttet, 1980.

6 Olsen 0, Kristensen TS. Hjerte/karsygdomme ogarbejdsmiljo. Bind 3. Kobenhavn: Arbejdsmiljofondet,1988.

7 Paffenbarger RS, Hyde RT, Wing AL, et al. A naturalhistory of athleticism and cardiovascular health. J7AMA1984; 252: 491-5.

8 Morris JN, Everitt MG, Pollard R, et al. Vigorous exercisein leisure-time: Protection against coronary heart disease.Lancet 1980; ii: 1207-10.

9 Groth MV. Fysisk aktivitet og motionsvaner i den danskebefolkning 1987. Kobenhavn: Dansk Institut for KliniskEpidemiologi, 1988.

10 Levin ML. The ocurrence of lung cancer in man. Acta UnioInt Contra Cancrum 1953; 9: 531-41.11 Miettinen OS. Proportion of disease caused or prevented by

a given exposure, trait or intervention. Am J Epidemiol1974; 99: 325-32.

12 Walter SD. Effects of interaction, confounding andobservational error on attributable risk estimation. Am JEpidemiol 1983; 117: 598-604.

13 Greenland S, Robins JM. Conceptual problems in thedefinition and interpretation of attributable fractions. AmJ7Epidemiol 1988; 128: 1185-97.

14 Bruzzi P, Green SB, Byar DP, et al. Estimating thepopulation attributable risk for multiple risk factors usingcase control data. Am J7 Epidemiol 1985; 122: 904-14.

15 Breslow NE, Day NE. Statistical methods in cancer research.Vol 1. The analysis of case-control studies. Lyon: IARC,1980.

16 Rothman KJ. Modern epidemiology. Boston: Little, Brown,1986.

17 Theorell T, Karasek R. Kan vi minska antalet hjartinfarktergenom at forbattra den psykosociala arbetsmiljon?Lakartidningen 1989; 86: 1455-6.

18 Kawachi MB, Pearce NE, Jackson RT. Deaths from lungcancer and ischaemic heart disease due to passive smokingin New Zealand. NZ Med J 1989; 102: 337-40.

19 Anonymous. Protective effect of physical activity oncoronary heart disease. MMWR 1987; 36: 426-30.

20 Wilhelmsen L. Kranskarlsjukdom og cerebrovaskularsiukdom-Kunskapsunderlag till 1987 drs folkhalso rapport.Stockholm: Socialstyrelsen 1987.

21 Paffenbarger RS, Hyde RT, Wing AL, Hsieh CC. Physicalactivity, all-cause mortality and longevity of college alumni(letter). N Engl J Med 1986; 315: 400-1.

22 Hattis D, Richardson B. Noise, general stress response, andcardiovascular disease processes: review and reassessment ofhypothesized relationships. Massachusetts: MIT, 1980.

23 Gotto AM. Interactions of the major risk factors forcoronary heart disease. Am 7 Med 1986; 80(suppl 2A):48-55.

CommentaryThe effects of occupation on the incidence ofcardiovascular disease have been seriouslyunderreported in published work. Public healthepidemiologists tend to concentrate almostexclusively on life style factors while occupationalhealth professionals prefer to report the influenceof work on the incidence of cancer, reproductiveoutcome, or musculoskeletal disorders. Theimbalance is partly restored by this thoughtprovoking paper of Olsen and Kristensen.The standard texts in occupational medicine

give little space to cardiovascular disease despitethe fact that it is a major contributor to Westernmortality and morbidity. Two recent andthorough reviews by, incidentally, one of theauthors of the current paper, cite a number ofchemical and non-chemical workplace factorswhich can (or might) influence cardiovasculardisease.' 2 Of the chemical factors, carbondisulphide and nitroglycerine/nitroglycerol arewell recognised causes, although recent studiessuggest that carbon disulphide may act morethrough a reversible, direct cardiotoxic orthrombotic influence than through previouslyassumed atherogenic effect.3 Nitroglycerine cancause both acute and chronic effects and is wellknown for the disturbing "Monday morning

death" syndrome. Some chemicals, such as

methylene chloride, cause carboxyhaemo-globinaemia while other chlorinatedhydrocarbons appear to cause cardiacarrhythmias. Inconclusive effects, possibly insome cases secondary to hypertension, have beenascribed to lead, cadmium, arsenic, and antimony.Recent reports suggest that epichlorohydrin5 andhydrogen sulphide6 might be added to the list.Peripheral vascular effects certainly arise fromworkplace exposure to cold, vibration and vinylchloride monomer.While chemical exposures are relatively rare,

the influence of putative circumstances such aswork "stressors", shift work, and sedentaryoccupations has received less attention but mightbe of greater preventive health importance. Thedifficulty has always been how to tease out theseinfluences from life style when reviewing the massof health data related to working populations.Olsen and Kristensen have attempted to do justthat.Using a scoring system for the quality of

epidemiological studies, they calculated cruderelative risks using quality in combination withquantity and consistency of findings in publishedreports. Combining this with a rough guide to the

9

on February 6, 2021 by guest. P

rotected by copyright.http://jech.bm

j.com/

J Epidem

iol Com

munity H

ealth: first published as 10.1136/jech.45.1.4 on 1 March 1991. D

ownloaded from