methods of - a bmj journal · the methods of assessment of the health impact can be presented with...

Occupational and Environmental Medicine 1997;54: 145-151

METHODOLOGY Series editors: T C Aw, A Cockcroft, R McNamee

Methods for assessing the extent of exposure andeffects of air pollution

Michal Krzyzanowski

AbstractBackground and objectives-In manyplaces in Europe, the ambient air pollu-tion exceeds the levels considered to besafe for health. The objective of the paperis to review and summarise the methodsof assessment of its impact on health, andto indicate the contributions of variousresearch disciplines, particularly envi-ronmental epidemiology.Methods-The framework for assessmentof impact is based on a four stage model:assessment of release of pollutant; assess-ment of exposure; assessment of the con-sequence; and risk estimation.Results-Epidemiology is crucial in pro-viding the data for the assessment of con-sequence. The criteria that determine theuse ofepidemiological studies for this taskinclude lack of bias, correct control ofconfounding, and measured estimates ofexposure. At present, those criteria areeasier to satisfy for studies of short termeffects on health than for the delayed con-sequences of exposure, or exposure accu-mulated over a prolonged period.Combinations of results from variouspopulations through meta-analysis ofexisting studies or conducting multicen-tre studies is often necessary to increasethe reliability of the consequence assess-ment stage.Conclusion-To assess the impact onhealth systematically helps to focus onactions to limit air pollutants with thegreatest impacts on human health and onthe most affected populations. Thismethod allows identification of the mostpertinent questions which have to beanswered by studies on relations betweenpollution and health and on exposure ofpopulations to air pollutants.Epidemiology has considerable potentialto contribute to this research.

WHO EuropeanCentre forEnvironment andHealth, Bilthoven, POBox 10, 3730 AA DeBilt, The NetherlandsM KrzyzanowskiCorrespondence to:Dr Michal Krzyzanowski,WHO European Centre forEnvironment and Health,Bilthoven, PO Box 10, 3730AA De Bilt, TheNetherlands.

Accepted 29 July 1996

(Occup Environ Med 1997;54:145-151)

Keywords: risk assessment; air pollution; environmen-tal epidemiology

Health risks posed by air pollution are wellrecognised. Based on experience fromepisodes of high pollution, occupationalhealth, and controlled human or animal expo-sure studies, individual countries have recom-

mended baseline controls on various pollu-tants in the atmosphere. The World HealthOrganisation (WHO) has recommended airquality guidelines.' It was thought that"... inhalation of an air pollutant in concentra-tions and for exposure times below a guidelinevalue will not have adverse effects onhealth ...". However, it has been admitted that"Compliance with recommendations regard-ing guideline values does not guaranteeabsolute exclusion of effects at levels belowsuch values" because a section of the popula-tion may be exceptionally sensitive, andbecause of a "... combined exposure to variouschemicals, or the exposure to the same chemi-cal by various routes". Recent research onhealth effects of air pollution has indicatedthat adverse health effects in normal urbanpopulations of several of the most common airpollutants can be found at concentrationsclose to, or below, the WHO guidelines.2Based on this new data, the WHO has initi-ated an update of the guidelines with the aimof producing the revised assessment by theend of 1996.3Ambient concentrations of several air pollu-

tants, such as SO2, suspended particulate mat-ter, NO2, and 03 regularly exceed therecommended air quality guideline concentra-tions in many places in Europe.4 This causesconcern about the impacts of the pollution onthe health of the population. The relevantquestion is not "Does the impact exist?" but"What is its magnitude, what proportion ofthe population is affected, what is the severityof the health effects in individual subjects, andwhat is the public health significance of theexposure consequences?". These questions areasked by the general public and by the publichealth services. Often, they are also asked byrepresentatives of various economic sectorsthat cause the air pollution. Estimated impacton health is compared with the costs to thesociety related to the measures to be taken toreduce air pollution. This comparison shouldsupport management of the risk. When riskelimination is not realistic, risk reduction to anacceptable level can be proposed.The objective of this paper is to summarise

the methods of assessment of health impactwhich can be used to support decisions on riskmanagement. As explained later, the impactassessment uses the existing data fromresearch and from routine monitoring; it is,therefore, an interpretation of the scientificdata, and not research by itself. At the same

145

on October 3, 2020 by guest. P

rotected by copyright.http://oem

.bmj.com

/O

ccup Environ M

ed: first published as 10.1136/oem.54.3.145 on 1 M

arch 1997. Dow

nloaded from

http://oem.bmj.com/

Krzyzanowski

time, the framework of the assessment identi-fies topics which require more, or morefocused, study. Some elements of the method-ology have been applied in the preparation ofthe WHO-ECEH (European Centre forEnvironment and Health) report "Concern forEurope's tomorrow".4

Framework for assessment ofthe healthrisk of air pollutionThe methods of assessment of the healthimpact can be presented with the frameworkof risk assessment adapted from earlierapproaches by Covello and Merkhofer.o Fourmain stages are: (a) assessment of release ofpollutants, (b) assessment of exposure, (c)assessment of the consequences, and (d) riskestimation. All the stages are preceded byidentification of the hazard.

In this paper, the discussion focuses on thenon-cancer end points and the term impactassessment is used instead of risk assessmentto distinguish two situations. The impactassessment considered here evaluates healtheffects of existing exposure. Provided theassumptions used in the assessment are valid,this should reflect impacts which haveoccurred in the population of concern. Therisk assessment refers to the potential healthimpacts, assuming a hypothetical exposurescenario. In the analysis of components of airpollution with carcinogenic properties, riskassessment has been used; this reflects theassessment of the probability of the health out-come with a (long) latency period.7With the multitude of substances with haz-

ardous properties which pollute ambient andindoor air, the first step of impact assessmentis to select the pollutants to be considered.These can be individual chemicals-such asSO2, Pb, or 0,,-or mixtures-such as sus-pended particulate matter or environmentaltobacco smoke. Sometimes an individual sub-stance is considered to be an indicator repre-senting a more complex mixture.8 Theselection is often decided by the availability ofdata-both on the health effects of the pollu-tant and on its concentrations in the environ-ment. Usually, the list of pollutants reduces toa few routinely monitored hazards or to sub-stances for which good validated models areavailable. In investigations of specific pollutionsituations-for example, in the neighbourhoodof a source of hazardous pollutants-the selec-tion can be more specific because of thesource. Knowledge of the human activity ornatural process that is the source of risk maybe necessary to select the hazardous sub-stances for assessment. For risk management,the knowledge of those processes is essential.

ASSESSMENT OF RELEASE OF POLLUTANTSOne of the features of the well known, socalled, classic pollutants is their presence inmost locations where fossil fuels are used forheating, cooking, energy production, or trans-port. Their release is proportional to the inten-sity of these activities, to the quality of fuelused, the technology of combustion, and thepresence of the measures to control emis-

sion-such as filters in the end of pipes.Besides the emission from local sources, trans-portation of the pollutant in the atmospherealso contributes to pollution.The release of secondary air pollutants

depends on the presence of the precursor pol-lutants and the conditions (processes) promot-ing the transformations. Ozone is a secondarypollutant formed in the atmosphere from NO2through a photochemical reaction involvingshort wavelength light. Ozone is a highly reac-tive gas; its reactions with NO reduces its con-centrations near the sources of NO. Thepresence of volatile organic compounds influ-ences the production/scavenging ratio andincreases the steady state concentration of 03.The presence of high concentration peaks,with normal concentrations of primary pollu-tant, is usually dependent on the weather con-ditions-such as high amounts of sunshine,and low wind velocity. This identifies geo-graphical areas where high peak concentra-tions of 03 can be expected.

ASSESSMENT OF THE CONSEQUENCESAt this stage, scientific information on thequantitative relation between the exposurelevel and extent of health impairment is sum-marised and its reliability is assessed. The rela-tion can be expressed as a relative risk of aqualitative health outcome or a magnitude ofchange in a continuous health variable at acertain exposure level. Establishment of thisrelation is in the domain of toxicology and epi-demiology.

Animal experiments have been the mostcommon method to study the biologicalresponses to chemicals. However, extrapola-tion of their results involves numerous uncer-tainties and assumptions.9 For the mostcommon air pollutants, the evidence fromhuman studies, both from the experimentswith controlled exposure and from populationstudies, is crucial to assess their health conse-quences. Occupational studies or assessmentof the health impact of episodes of high airpollution relate to unusual situations.Controlled human exposure studies are lim-ited to observation of immediate health conse-quences of the exposures, and for ethicalreasons cannot include people who are poten-tially the most susceptible to air pollution.Often, the few subjects included in the studylimits the study power and precludes measure-ment of effects which vary between, andwithin, the subjects. Nevertheless, controlledstudies of exposure to 03 in concentrationssimilar to those in ambient air have signifi-cantly contributed to the assessment of therelation of pulmonary function to the expo-sure. Io

The study of the relation between healthand low level air pollution is in the domain ofepidemiological studies, investigating largepopulations to establish (sometimes weak)associations." However, the integrity of theresults of the various methods used is impor-tant in indicating of the validity of the estima-tion of the relation between exposures andtheir consequences.

146



.bmj.com

/O

ccup Environ M


arch 1997. Dow

nloaded from

http://oem.bmj.com/

Methods for assessing the extent of exposure and effects of air pollution

For studies designed to establish causalityand measure the health effect of the factor it isadvisable to find and measure a pollutant (orits metabolite) close to the target organ.'2However, this requires invasive methods and israrely possible in a human population study.Therefore, most studies designed to estimateof the association between the pollutant andhealth have used environmental indicators ofcontact of the study subjects with the pollutantassuming that there is a fairly good correlationbetween those indicators and the causativefactor."3 Ambient concentrations of the pollu-tants have often been used as such indicatorsdue to the availability of routinely collected airmonitoring data. A potential drawback of thistype of data results from the variability of con-centrations of pollutants in time and space.Validity of exposure estimates based on ambi-ent air monitoring should, therefore, be thor-oughly evaluated in each epidemiologicalstudy.'4 15 Several pollutants, however, havebeen assessed for effects, and even measuredwithout full understanding of the biologicalmechanisms leading to the health damage.This is the case for the important air pollutant,particulate matter.'6

Exposure indicators must be validated sothat the effects of the pollutant can be properlymeasured. Errors in the measurement of expo-sure, or misclassification of exposure status,may distort the real association and lead tofalse conclusions. When the errors in the mea-surement of exposure are similar for studysubjects with various health outcomes (non-differential misclassification), the error usuallyleads to underestimation of the true associa-tion between the pollution and outcome.'2Greater variability in measured than in trueexposure leads to a smaller range in the mea-sured exposure, and consequently to a smallerrelative measure of effect-such as relative riskor odds ratio. The statistical power of suchstudies is also reduced, and it may be neces-sary to increase the sample size to achieve suf-ficient power.'7 However, in some situationsnon-differential misclassification of exposuremay result in an increase in the apparent riskestimate. One example of this occurs when theerror in exposure estimate depends on the trueexposure-for example, is greater with lowpollution than with high pollution.'8 When thebias in exposure measurement is different ingroups with different health-for example,when the exposure is overestimated moreoften in sick than in healthy subjects-theresults of the study may both underestimate oroverestimate the true relation. Such errors,therefore, make the study unusable for assess-ment of health risk.

Errors in measurement of confounding vari-ables may also bias results of the study, andthe adjustment with a poor measure of theconfounder may be equivalent to no controlfor confounding. The bias may cause the asso-ciation to seem stronger or weaker than inreality. 12

In summary, several criteria determine theuse of an epidemiological study for quantita-tive risk assessment, and for assessment of its

consequences to health in particular.'9Assuming that there is a positive associationbetween the exposure and health variables, thestudy must be unbiased, confounding must beeliminated, and quantitative, and valid, expo-sure estimates must be available. Although,until recently, epidemiological studies havebeen criticised for not being able to cope withthese criteria sufficiently well, there are agrowing number of studies which can be, andare, used for impact assessment.Most of those studies deal with health con-

sequences of short term variability of air pollu-tion levels.2 Temporal studies with aggregateddata on mortality or hospital admissions assesshealth and exposure variables at the popula-tion level." The "all cause" mortality is con-sidered in many studies as it is often available asthe only indicator. The specificity of this indi-cator of health outcome is far from satisfac-tory. When the cause specific rates areconsidered, the disease which may be causallyrelated to the increased air pollution may bementioned as a contributing cause of deathand so would not be available for analysis.2Less severe health outcomes may be moreappropriate for the studies of the effects of airpollution; reports from hospitals on admis-sions for respiratory diseases have been usedfor this purpose. The morbidity indicatorsbased on health services may be related to theactivity pattern of the services to a greaterextent than to health of the population andthis must be carefully considered in the analy-sis.22 The indicator of air pollution usually isan average concentration of the pollutant cal-culated from the data routinely collected in thearea of residence of the studied population.2'This method certainly lacks precision and, inmost cases, the analyses that use it will under-estimate the effect of specific concentrations ofa pollutant on health.

Panel studies consider information collectedover a particular time (weeks or months) froma selected group of people. In comparison withthe temporal studies of aggregated data, panelstudies are more labour intensive and, there-fore, less common. Specially collected infor-mation gives a more specific and moresensitive description of the health variablespotentially related to air pollution, and pro-vides potentially a better exposure indica-tor. 24 -26

In the temporal studies with aggregateddata and in the panel studies, a health variableis correlated, after adjustment for confoundingby a variety of time related factors, with the airpollution indicator relevant to the time unit.Recent developments of statistical techniqueswhich are able to account for peculiarities ofthe data (autocorrelation, periodic changes)have contributed to measurement of these cor-relations. Considering types of errors in assess-ment of health and exposure variables, as wellas the possible overadjustment for the timechanging covariates, it can be argued that suchstudies provide a conservative estimate of thehealth consequences of the pollution.

Less common are the studies on health con-sequences of measured exposures that are pro-

147



.bmj.com

/O

ccup Environ M


arch 1997. Dow

nloaded from

http://oem.bmj.com/

Krzyzanowski

longed, and have a latency period. These stud-ies have numerous technical difficulties suchas the study duration (in cohort designs),problems with measurement of exposure overa prolonged period (or validity of retrospectiveexposure assessment), or appropriate controlof confounding in studies comparing popula-tions living in different environmental condi-tions. Nevertheless, such studies have beencompleted and have provided important indi-cations for the long term health impacts of airpollution.27 28 Information on health conse-quences is usually better defined in the longi-tudinal studies than in temporal studies withaggregated data. Often, individual data onhealth and on potential confounders is avail-able in such studies allowing for a better defin-ition of the health end point, affectedsubgroups of the population, and possibleindividual characteristics correlated with theresponse to air pollution-for example, atopy.Most of the common air pollutants are asso-

ciated with health outcomes of multifactorialaetiology. Epidemiological studies aim to com-pare the frequency of those outcomes (or dis-tribution of a continuous health variable)between groups exposed to different amountsof the studied pollutant. The desired outcomeof the studies is a curve indicating the risk ateach level of exposure (exposure-effect curve),and not only the estimated increase in riskbetween two exposure groups.An important question pertinent to the

studies of health effects of air pollutants at lowconcentrations is the shape of the exposure-effect curve, and in particular, the existence ofa concentration below which no adverseeffects are found. It is a common task of toxi-cological analysis to find the "no observedadverse effect level (NOAEL)", used in assess-ment of health criteria for environmental fac-tors.9 The assumption of the existence of such asafe level is a basis of threshold limit values(TLVs) established for occupational settings.For epidemiological studies of the generalpopulation, it is difficult to find the thresholdlevel. It is mostly due to the uncertainty of thelow (or no) exposure status in a general popu-lation study: even if the threshold exists, the(usually non-specific) health outcome can befound in the low exposure group. It may beimpossible to find if the effect is related to thelow level of pollution, to factors other thanpollution, or to misclassification of the expo-sure. In studies of the most common air pollu-tants, it may not be possible to establish anon-exposed group as the whole population is,or was in the past, exposed to some concentra-tion of the pollutants above zero. A furthercomplication in finding the threshold is thedifferent individual sensitivities. The thresholdmay exist but may be different in each person.The overall picture for the population maydepend on the proportion of sensitive subjectsin this population. Nevertheless, if an appro-priate exposure range is measured with suffi-cient precision, epidemiological studies canattempt to assess the existence of the threshold.29

Results from individual studies may be suffi-cient if risk assessment relates to the popula-

tion considered in the given study. Calculationof attributable proportion, together with theestimate of its confidence interval, may beincluded into the interpretation of studyresults. However, in most cases, the impactassessment is conducted in populations with-out the relevant data from epidemiologicalstudies. The estimates of the strength of associ-ations must be extrapolated from studies con-ducted elsewhere with the underlyingassumption that the shape, and the magni-tude, of the association remains unchanged invarious populations. The validity of such anassumption should be supported by observa-tions actually conducted in various popula-tions and reviewed by individual experts orgroups of experts.30 33

Formal review, and estimation of summarystatistics reflecting the relation of selected andwell specified health consequences of exposurebased on several studies, are done in a meta-analysis.3435 A paper published recently sum-marises the main criteria to apply to thistechnique.36 Between the other relevant fea-tures, the ability of the meta-analysis to iden-tify heterogeneity of effects among manystudies should be emphasised. However, thedivergence of methods, health end points, andexposure indicators used in various studiesseverely restrict present possibilities for meta-analysis in the epidemiology of air pollution.An important development of this is theorganisation and implementation of multicen-tre studies, which by design allow comparisonof associations between well defined and uni-formly measured air pollution and health vari-ables.2337 Combined estimates of effect of airpollution based on such studies provide thebest input to the assessment of the conse-quences to health.

EXPOSURE ASSESSMENTThe main elements considered when exposureinformation is used in impact assessment are:* Representativeness of the available envi-

ronmental data for the (exposure of) popu-lation at risk

* Averaging time appropriate for the linkwith health

* Correspondence of the exposure indicatorwith the indicator of the exposure-response(consequence) function used in the assess-ment.An additional criterion, not necessary for

the impact assessment but potentially usefulfor the recommendations related to the riskmanagement after the assessment, is the abilityto link the exposure with the source of pollu-tion.

These points differ considerably from theaspects of exposure assessment recommendedfor epidemiological research studies mainlydue to the feasibility of collection of the opti-mal data.To assess whether the exposure information

is representative of the population at riskrequires that the monitoring stations of the airpollutant are located in residential areas and inother sites where substantial proportions of thepopulation exposure to outdoor air pollution

148



.bmj.com

/O

ccup Environ M


arch 1997. Dow

nloaded from

http://oem.bmj.com/


may occur. Stations located next to importantsources of pollution or stations located faraway from built up areas (impact stations orbackground stations, according to theEuropean Commission terminology38) shouldnot be (directly) used for health impact assess-ment. Based on the monitoring results, infor-mation on concentrations specific to time andspace should be derived.The simplest method to estimate the popu-

lation exposure is based on the average valuecalculated from data measured at the stationsoperating in the region where the populationlives-for example, the town. This one value(in one time unit) is then assigned to the entirepopulation of concern. Such a method may beacceptable and has been used in several epi-demiological studies of relations between thesource of exposure and the consequences.2' 23Also, the town specific mean concentrationmay well reflect the mean concentration expe-rienced by the residents moving within thetown during the time unit.39 This approach hasalso been used to estimate exposure of urbanpopulations in Europe for the WHO reportConcern for Europe's tomorrow.4More complex methods that use mapping

of the pollutant concentrations often supportthe monitoring data with statistical or disper-sion modelling. Modelling and mapping tech-niques are well developed to describe meanconcentrations of pollutants in large ruralareas.40 Their application in densely populatedurban areas is difficult due to the requiredscale and precision, but several studies haveattempted this approach.4' The area specificestimates of concentrations are then super-imposed-for example, with geographicalinformation systems techniques-on the spa-tial distribution of the population to producepopulation distribution of exposure in a giventime unit. Although that type of modellingmay improve spatial definition of the pollutionand can account for spatial differences in den-sity of the resident population, it does not usu-ally account for movement of the populationwithin the area. However, if improved data,better than presently available, on patterns ofactivity were to be included, such modellingmay significantly improve exposure assess-ment in the population where impact assess-ment is conducted. Such data will benecessary when sufficient numbers of studiesthat use personal exposure data are availablefor assessment of health response, and areincluded in the consequence assessment partof the risk assessment.

IMPACT ESTIMATIONThis stage combines information from the pre-ceding stages with the aim of measuring theimpact, to indicate its uncertainty or certainty,and to indicate the elements influencing theprecision of the estimates. Further, estimationof impact identifies factors which may influ-ence the risk-such as special exposure condi-tions or susceptible groups.

Quantitative estimate of the impact can bebased on the calculation of the attributableproportion, indicating the fraction of the

health outcome which can be attributed to theexposure in a given population (provided thatthere is a causal association between the expo-sure and the health outcome). With the distrib-ution of exposure of the population found inthe exposure assessment stage, and the rela-tion between exposure and consequence iden-tified, the attributable proportion can becalculated with the formula:

AP = £{ [RR(c)- 1] * p(c)}[ RR(c) *p(c)]

where: RR(c) = relative risk for the healthoutcome in category c of exposure, and

p(c) = proportion of target population incategory c of exposure. Knowing (or, often,assuming) a certain underlying frequency ofthe outcome in the population, I, the rate (ornumber of cases per unit population) attrib-uted to the exposure in the population can becalculated as:

IE = I * AP

For a population of a given size, this can beconverted to the estimated number of casesattributed to the exposure.When the response to the harmful exposure

is expressed as a change in a continuous vari-able of health status-for example, lung func-tion, body weight, intelligence quotient-theestimate of impact corresponds to the esti-mated mean change of the variable in theexposed population. Using the information onthe variability of the response (from the assess-ment of the consequences), the range of theimpact should be estimated as well for exam-ple, as the 90th or 95th percentile of theresponse-to show the impact in the particu-larly sensitive people.

Each of the values used to derive this num-ber is associated with an estimation error orother sources of uncertainty (related to the useof assumptions and expert judgement). Thoseuncertainties have to be considered at eachstage of the estimation to produce lower andupper boundaries of the estimate as well as thebest expected value. More elaborateapproaches include modelling of the distribu-tion of impact probability-for example, withMonte Carlo techniques. Input data for suchprocedures include information on the rangeof probabilities of the estimates of exposureand their consequences.6 The same proce-dures can also be used in sensitivity analysis toassess the most important elements in estima-tion of the precision of the impact-such asthe relation between exposures and their con-sequences or the distribution of the exposurein the studied population.An approach similar to that used in sensitiv-

ity analysis can be used to estimate changes inthe impacts due to various scenarios of modifi-cation of exposure. This can increase the use-fulness of assessment for risk management.

DiscussionExtrapolation and practical application ofresearch involves simplification, generalisa-

149



.bmj.com

/O

ccup Environ M


arch 1997. Dow

nloaded from

http://oem.bmj.com/

Krzyzanowski

tion, and application of assumptions whichmay be difficult to test. This is often criticisedby researchers as the tight quality standardsare mixed with less stringent methods of infer-ence. However, these applications are themain justification of the studies. As a matter ofprinciple, the research on environmentalhealth is oriented towards finding the factorsaffecting the health of populations and pre-venting adverse impacts on health. Obviously,not all populations can be studied. Therefore, itis necessary to generalise the results obtainedin one population to other groups. The preci-sion of this generalization may be relativelylow, but even this may be sufficient to identifylocal priorities and to find preventive actions.The approximate result based on impactassessment from the best available knowledgeis always more systematic and clearer than asubjective judgement based on emotions orarbitrary assumptions.42The procedure of risk assessment may indi-

cate important gaps in knowledge and imposesquality criteria for further, focused research.Reduction of the scientific uncertainties per-taining to the risk assessment should be acommon goal of the scientists and the decisionmakers. Improved data from epidemiologicalstudies and from the studies describing expo-sure patterns in the populations are needed torefine risk assessment. Without the progress inresearch, the assessment of health benefitsfrom the reduction or prevention of exposuresto air pollution will remain uncertain.43Environmental epidemiology, as a disciplinestudying the associations of interest in the gen-eral population, can contribute most relevantdata provided the studies are conducted to thedesired standards.44

Particular caution in the interpretation ofthe impact assessment should be exercisedwhen the analysis is based on "indicator pollu-tants", or on other exposure indicators whichonly indirectly correspond to the pollutantcausing the health effect. The correlation ofthis indicator with the true level of the expo-sure in one population may not be repeated inanother population. This problem shows aneed for more relevant exposure monitoring inthe assessment of health risks.An important limitation of the present

methodology is its reliance on the "one pollu-tant-one outcome" approach. To someextent, this approach is dictated by the scarcityof studies on health effects of a combination ofpollutants characteristic of ambient exposures,and particularly on their interactions.Measurement of the components of mixturesresponsible for the health effect is important topropose actions which may efficiently reducethe impact on health." However, it is possiblethat the same pollutant may have differentimpacts in the presence, or absence, ofincreased concentrations of other pollutants.46Further, it is possible that a susceptible sectorof the population responds to various types ofair pollution in a similar way. In such casesreduction of one of the pollutants may resultin a smaller than expected decrease in healthproblems as the effect of the other pollutants

will remain. Therefore, in situations in whicheffects of individual pollutants are difficult toseparate, a reduction of the pollution mixture,and of all its indicator components, should beadvised. For a more precise description of therelation between air pollution and exposure,and for more specific preventive actions, betterepidemiological studies and understanding ofmechanisms of action of the relevant pollu-tants are necessary.

In conclusion, application of a systematicimpact assessment helps to focus on preven-tive actions for air pollutants with the greatestimpacts on human health and on the mostaffected populations. Use of this methodologyenables identification of the most pertinentquestions which have to be answered by studieson relations between pollution and health andon population exposure to air pollutants.Epidemiology has a great potential to con-tribute to this research.

Several of the issues presented in this paper have been dis-cussed at the meetings of the Working Group "RiskAssessment" of the European Concerted Action "Air pollutionEpidemiology". Helpful comments on an earlier version of thispaper were provided by R van Leeuwen and A Kuchuk. Theviews presented are those of the author and not necessarily theviews of the WHO.

1 World Health Organisation. Air quality guidelines forEurope.Copenhagen: WHO Regional Office for Europe,1987. (WHO Regional Publications, European Series No23.)

2 Brunekreef B, Dockery DW, Krzyzanowski M.Epidemiologic studies on short-term effects of low levels ofmajor ambient air pollution components. Environ HealthPerspect 1995;103(suppl 2):3-13.

3 World Health Organisation. Update and revision of the airquality guidelines for Europe. Copenhagen: WHO RegionalOffice for Europe, 1994. (EUR/ICP/CEH 230, 00104.)

4 World Health Organisation. Concern for Europe's tomorrow:health and the environment in the WHO European Region.WHO European Centre for Environment and Health.WHO-Euro. Stuttgart: Wiss Verl-Ges, 1995.

5 Edwards R. Industry denies dangers of particle pollution.New Scientist 1995;2002:5.

6 Covello VT, Merkhofer MW. Risk assessment methods.Approaches for assessing health and environmental risks.New York: Plenum Press, 1993.

7 Hemminki K, Pershagen G. Cancer risk of air pollution:epidemiological evidence. Environ Health Perspect1994;102(suppl 4):187-92.

8 Bostrom CE, Almen J, Steen B, Westerholm R. Humanexposure to urban air pollution. Environ Health Perspect1994;102(suppl 4):39-47.

9 World Health Organisation/IPCS. Assessing human healthrisks of chemicals: derivation of guidance values for health-based exposure limits. Geneva: World Health Organization,1994. (Environmental Health Criteria 170.)

10 Horstman DH, Folinsbee LJ, Ives PJ, Abdul-Salaam S,McDonnell WF. Ozone concentrations and pulmonaryresponse relationships for 6-6-hour exposures with fivehours of moderate exercise to 0-08, 0-10 and 0-12 ppm.Am Rev RespirDis 1990;142:1158-63.

11 Katsouyanni K, ed. ECA Air Pollution Epidemiology ReportSeries. Study designs. Luxembourg: CEC DG XII, 1993.(Report No 4. EUR 15095 EN.)

12 Armstrong BK, White E, Saracci R Principles of exposureassessment in epidemiology. Oxford: Oxford UniversityPress, 1994.

13 Lebret E. Models of human exposure based on environ-mental monitoring. Sci Total Environ 1995;168:179-86.

14 Seifert B. Validity criteria for exposure assessment meth-ods. Sci Total Environ 1995;168:101-7.

15 World Health Organisation European Centre for Environ-ment and Health. Methodology for assessment of expo-sure to environmental factors in application toepidemiological studies. Sci Total Environ 1995;168:93-100.

16 Seaton A, MacNee W, Donaldson K, Godden D.Particulate air pollution and acute health effects. Lancet1995;345:176-8.

17 McKeown-Eyssen GE, Tibshirani R. Implications of mea-surement error in exposure for the sample sizes of case-control studies. Am Y Epidemiol 1994;139:415-21.

18 Wacholder S. When measurement errors correlate withtruth: supprising effects of non-diferrential misclassifica-tion. Epidemiology 1995;6:157-61

19 Hertz-Picciotto I. Epidemiology and quantitative riskassessment: a bridge from science to policy. Am Y PublicHealth 1995;85:484-91.

150



.bmj.com

/O

ccup Environ M


arch 1997. Dow

nloaded from

http://oem.bmj.com/


20 Schwartz J. What are people dying of on high air pollutiondays? Environ Res 1994;64:26-35.

21 Schwartz J. Air pollution and hospital admissions for respi-ratory disease. Epidemiology 1996;7:20-8.

22 Quenel P, Medina S, Pirard P, Momas I, Le Moullec Y,Festy B, Dab W. Health service based morbidity indica-tors as a measure of health effects of air pollution. EuroJPublic Health 1994;4:201-6.

23 Katsouyanni K, Zmirou D, Spix C, Sunyer J, Schouten JP,Ponka A, et al. Short-term effects of air pollution onhealth: a European approach using epidemiological time-series data. The APHEA project: background, objectives,design. Eur RespirJ 1995;8:1030-8.

24 Hoek G, Brunekreef B. Effect of photochemical air pollu-tion on acute respiratory symptoms in children. Am JRespir Crit Care Med 1995;151:27-32.

25 Krzyzanowski M, Quackenboss J, Lebowitz MD. Relationof peak respiratory flow rates and symptoms to ambientozone. Arch Environ Health 1992;47:107-15.

26 Quackenboss JJ, Krzyzanowski M, Lebowitz MD.Exposure assessment approaches to evaluate respiratoryhealth effects of particulate matter and nitrogen dioxide. JExpo Anal Environ Epidemiol 1991;1:83-107.

27 Abbey DE, Lebowitz MD, Mills PK, Petersen FF, BeesonWL, Burchette RJ. Long-term ambient concentrations ofparticulates and oxidants and development of chronicdisease in a cohort of non-smoking California residents.Inhal Toxicol 1995;7:19-34.

28 Dockery DW, Pope CA III, Xu X, Spengler JD, Ware JH,Fay ME, et al. An association between air pollution andmortality in six US cities. N Engl J Med 1993;329:1753-9.

29 Ulm K. A statistical method for assessing a threshold inepidemiological studies. Stat Med 1991,10:341-9.

30 Dockery DW, Pope CA III. Acute respiratory effects ofparticulate air pollution. Annu Rev Public Health 1994;15:107-32.

31 Department of Health. Advisory Group on the MedicalAspects of Air Pollution Episodes. Third report: oxides ofnitrogen. London: HMSO, 1993.

32 Department of Health. Committee on the Medical Effectsof Air Pollutants. Non-biological particles and health.London: HMSO, 1995.

33 World Health Organisation. Update and revision of the airquality guidelines for Europe. Meeting of the Working Group

on Classical Air Pollutants. Copenhagen: WHO RegionalOffice for Europe, 1995. (EUR/ICP/EHAZ 94 05/PBO 1.)

34 Hasselblad V. Meta-analysis of environmental health data.Sci Total Environ 1994;1601161:545-58.

35 Schwartz J. Air pollution and daily mortality: a review andmeta-analysis. Environ Res 1994;64:36-52.

36 Blair A, Burg J, Foran J, Gibb H, Greenland S, Morris R, etal. Guidelines for application of meta-analysis in environ-mental epidemiology. Regul Toxicol Pharmacol1995;22:189-97.

37 Brunekreef B, Hoek G, Roemer W, Kalandidi A, BaldiniG, Vondra V, et al. PEACE Study. (1) Background anddesign. Epidemiology 1995;6:S65.

38 European Commission for Europe (ECE). Conference ofEuropean Statistician. Readings in international environmen-tal statistics.Geneva: United Nations, 1993. (UnitedNations Publication GE.93-32468.)

39 Pirard P. Quenel P. Lameloise P, Le Moullec Y. Etude del'utilisation d'une moyenne arithmetique des mesuresd'un reseau de surveillance comme indicateur de niveaude pollution atmospherique en milieu urbain. PollutionAtmosphirique 1995;Avril-Juin:59-66.

40 Simpson D. Photochemical model calculations overEurope for two extended summer periods: 1985 and1989. Model results and comparison with observations.Atmospheric Environment 1993;6:921-43.

41 Briggs DJ, van der Veen AA, Pryl K, Collins S. Air pollu-tion mapping in SAVIAH study. Epidemiology 1995;6:S32.

42 Goldstein BD. The need to restore the public health basefor environmental control. Am J Public Health 1995;85:481-3.

43 Sexton K. Science and policy in regulatory decision mak-ing: getting facts right about hazardous air pollutants.Environ Health Perspect 1995;102(suppl 6):213-22.

44 Shore RE. Epidemiologic data in risk assessment-imper-fect but valuable [editorial]. Am 7 Public Health 1995;85:474-5.

45 Dockery DW, Schwartz J. Particulate air pollution andmortality: more than Philadelphia story. Epidemiology1995;6:629-32.

46 Lebowitz MD, Quackenboss ll, Krzyzanowski M,O'Rourke MK. Multipollutant exposures and healthresponses to particulate matter. Arch Environ Health1992;47:71-5.

Occupational and Environmental Medicine and theelectronic age

OEM has an Email address which is100632.3615(5compuserve.com. We wel-come contact by Email, including letters tothe editor. Some of our reviewers alreadysend us their reports by Email, helping tospeed up the peer review process.We are moving towards electronic pub-

lishing and for some months now we havebeen asking authors to send us their revisedpapers on disk as well as a hard copy. I amdelighted to report that nearly all ourauthors are managing to comply with this

request. Oddly enough, the few authorswho have not sent us a disk version of theirrevised papers have been almost exclusivelyfrom the United Kingdom. I would beinterested in suggestions for why this mightbe. Perhaps United Kingdom basedauthors read our correspondence andinstructions less assiduously? Watch forrevised Instructions to Authors.

The Editor

151



.bmj.com

/O

ccup Environ M


arch 1997. Dow

nloaded from

http://oem.bmj.com/

methods of - a bmj journal · the methods of assessment of the health impact can be presented with...

Documents