Document ZJqNbr22LOX5KxO2D3jeyEM8d

Occupational and Papers 721 Future impact of genetic screening in occupational and environmental medicine Roger G Rawbone 725 Applications of new technology in molecular epidemiology and their relevance to occupational medicine David Koh, Adeline Seow, Choon Nam Ong 730 Molecular methods of measurement of hepatitis B virus, hepatitis C virus, and human immunodeficiencyvirus infection: implications for occupational health practice ]ia-Horng Kao, lulia Heptonstall, Ding-Shinn Chen 735 Exposure assessment of high molecular weight sensitisers: contribution to occupational epidemiologyand disease prevention Dick Heederik, Gert Doekes, Mark ]Nieuwenhuijsen 742 Flow cytometry in the exploration of the physiopathology of occupational lung disease A D Curran 747 Measurement of human exposure to biologically relevant fractions of inhaled aerosolsJohn W U~mie, Robert J Aitken 753 New technologies in the use of exhaled breath analysis for biological monitoring H K Wilson, A C Monster 758 Development of an expert system for the interpretation of , serial peak expiratory flow measurements in the diagnosis of \ occupational asthma P S Burge, C F A Pantin, D T Newton, P F G Cannon,P Btight,J Belcber, J McCoacb, D R Baldwin, C B S G Burge, and the M i d W Thoracic Soctety Research Group 765 Conceptual model for assessment of dermal exposure Thomas Schneider, Roe1 Venneulen,Derk H Brouwer, ]ohn W Chmie, Hans Krotnhout, Christian L Fogh 774 Analysis of incidence of childhood cancer in the West Midlands of the United Kingdom in relation to proximity to main roads and petrol stations Roy M Harrison, Pei-Ling Leung, Lillian Sornerwaille, Ralph Smith, Estelle Gilman 781 Correspondence 788 Risk of stomach cancer associated with 12 workplace hazards: analysis of death certificates from 24 states of the United States with the aid of job exposure matrices PierfuigiCocco, Mary H Ward, Mustafa Dosemeci Measurement of vitamin D, metabolites in smelter workers exposed to lead and cadmium Howard Mason, David Chettle; Sylvia Chalkley 788 Exposure-response relations of a-amylase sensitisation in British bakeries and flour mills T A Smith, N Bunnett, M T Bithell; Anne Linehan; Mark ]Nieuwenhuijsen, Dick Heederik, Gert Doekes, Katherine M Venables, Anthony] Newman Taylor 790 Health effects among workers in sewage treatment plants Sheila Andren, Rualri Brugha 790 Notice 790 Book reviews .- .-~ c Measurement of human exposure to biologically relevant fractions of inhaled aerosols John W Cherrie, Robert J Aitken Abstract holders that were thought to collect "total Aerosol sampling has evolved with dust". More complex devices, such as cyclones changes in technology and our under- designed to collect only the finer fractions of standing of the importance of assessing airborne dust, became available only later. the biologically relevant fractions of the However, these improvements in technology ! total aerosol. During the past decadethere enabled the links to be identified between has been international agreement on the occupational exposure to a range of aerosols definitions for the inhalable, thoracic, and and lung disease. respirable h c t i o n s and instruments have Developments in monitoring of exposure to been developed to collect samples accord- aerosols have generally been focused towards ing to these conventions. These measure- sampling of occupational exposure, although ment techniques are now well established the principles are equally applicable to nonin the workplace and are increasingly occupationalexposures. In the past, limitations being applied to assessments of non- on analyticalsensitivityhave dictated that envi- occupational exposure, as the practical ronmental sampling of aerosols should be ditficulties in obtaining samples over 24 undertaken with high volume fixed location hours have been solved. It is argued that monitoring equipment. This is no longer the multiple aerosol size h c t i o n s should be case and personal exposure measurements in measured for hazardous substances- non-occupational situations should help im- such as inorganic lead-in which inhaled prove risk assessments. material may either be absorbed in the In this paper we briefly review some of the i alveoli (respirable) or cleared from the recent developments in sampling methods used ciliated airways in the lung to the gut and to assess human exposure to aerosols, and in then absorbed (thoracic and extratho- particular, we look at issues of transferring racic). Such measurements should im- experience h m workplace sampling to non- prove the evaluation of the risk for inhaled occupationalexposure monitoring. We use non- lead, particularly for non-occupational occupational exposure to inorganic lead aerosol exposure ofchildren. Also,passive aerosol as an example of the issues involved in develop- sampling techniques may enable meas- ing an appropriate measurement strategy. urements ofnon-occupationalexposure to be made over severalweeks and this would also help improve the reliability of the risk evaluation for aerosol toxicants with a long biological half life. However, further development of these techniques is required if they are to be practicable and biologically relevant. (OccupEmiron Med 1999;56:747-752) Early developments in aerosol sampling Up to the mid- 1920s the principal method of sampling aerosols was the sugar tube, later described by Walton.' This device comprised a 32 mm diameter tube filled to a depth of about 100 mm with 10-20 mesh sugar granules. Air was drawn through the granules by a hand operated pump and the collected dust was analysed by dissolving the sugar, and filtering Keywords: exposure; thoracic; lead the residue, which was then weighed to provide an estimate of the airborne mass. As well as being difficult to use, this method was criticised It has long been recognised that the presence of because reductions in the measured dust particles in the air may adversely affect the concentration in mines and other dusty envi- health of exposed people. However, it is only in ronments were not matched by corresponding University of Aberdeen the past 100 years that serious attempts have falls in dust related diseases. J W Cherrie been made to measure the size, shape, and More complex instruments soon became Institute of Occupational Medicine, 8 Roxburgh Place, Edinburgh concentration of airborne particles and to then relate these measures to human health. Inevitably, almost all of the early work was involved with measurement of aerosols in occupational available. One such device was the konimiter, which was developed in 1916 by Kotze.' In this device, a small sample of air was drawn through a jet nozzle and directed towards a plate EH8 9SU, UK settings such as mining, as it was in these situ- covered with petroleum jelly where the parti- J W Cherrie R J Aitken ations that effects on health were most obvious. cles deposited by impaction. The plate was Most of the early aerosol monitoring devices removed and the particles counted with a Correspondence to: Dr J W Cherrie, Institute of Occupational Medicine, 8 Roxburgh Place, Edinburgh EH8 9SU, UK. were designed to be placed in one place while the sample was collected. However, the advent of small and light battery operated pumps in the 1960s meant that personal exposure measurements became possible. Early personal microscope. The operating principles of this device recognised that both the particle numbers and the particle size could be important determinants of risk. However, it was only able to collect a sample over a very short duration Accepted 15 June 1999 samplers comprised simple open faced filter (about 1 second). Cherrie,Aizken By the mid-1930s there were several alterna- mouth (the inhalable fraction). This concept tives to the konimiter, the most important was based on measurements of the aspiration being the thermal precipitator.' This overcame efficiency of life sized human models reported the limitations of the konimiter to collect very by Ogden and his co-workers" l2 and subse- small particles (<lpm) and samples could be quently by Vincent and his colleague^.^^"^ For collected over much longer periods. The large particles (>50p,aerodynamic diam- thermal precipitator had an inlet leading to a eter) only about 50% of the total aerosol enters narrow channel where there was a heated wire the respiratory tract according to the inhalable oriented perpendicular to the direction of flow. criteria, and this increases to almost 100% for As the contaminated air passed the wire, ther- small (-1 pdiameter) particles. The thoracic mal gradients caused the particles to be and respirable conventions are subfractions of deflected towards the walls of the sampler the inhalable aerosd and have their basis in a where two glass disks were located. After sam- wide range of human volunteer deposition and pling, the glass disks were removed and the clearance studies.16However, the final forms of particles counted as with the konimiter. The the definitions were a compromise in the case thermal precipitator was a heavy bulky instru- of the respirable convention between the ment that could only be used for fixed location BMRC curve and the previous ACGIH measurements. definiti~n.'~ Partly because of ease and reliability of The development of size conventions for analysis, gravimetric sampling with filters to measurement of ambient aerosol has been collect the dust from the air largely superseded through a separate process to that of the work- particle counting methods. An early example of place definitions and has been driven primarily such an instrument is the Mine Research by the deliberations of the United States Envi- Establishment (MRE) type 113A gravimetric ronmental Protection Agency. Greater empha- sampler` which has been used in the mining sis in these considerations has been placed on industry in the United Kingdom to collect the the sue distribution of aerosol in the ambient British Medical Research Council (BMRC) air but potential penetration into the human respirable dust fra~tion.T~he respirable frac- respiratory tract was also considered. The two tion was defined qualitatively as the fraction of commonly used environmental sampling con- the airborne dust that approximates to that ventions are PM,, and both of which are which would reach and deposit in the alveoli, curves falling from unity to zero with, in the \ reflecting the belief that these particles were of case of PM,, a 50% (0.5) value at 10 pm and, greater importance in causing occupational for PM,,, a 0.5 value at 2.5 p.Although not lung disease, especially pneumoconiosis. With specificallydesigned to be so, the thoracic con- this device a close relation was established vention is almost identical to the PM,, size between the prevalence of pneumoconiosis and selection curve. There is, however, a sigmficant the mass concentration of respirable coal mine difference between the respirable convention dust: whereas earlier attempts to correlate dis- and the Pw,,m e . ease with the number of particles of dust had It is now considered that personal samplers been less successf~lT.~he development of the are much more effective tools for assessing BMRC respirable fraction was the first of what exposure to aerosols in the workplace than have become known as the "biologically static samplers. The types and performance of relevant" aerosol size fractions that now form these devices have been described by other the basis of modern sampling methods. authors.'* In the main, these samplers have been targeted to meet either the inhalable Sampling of biologically relevant and convention-for example, the Institute of other aerosol size fractions Occupational Medicine personal inhalable Over the past 10 years substantial progress has sampler"--or the respirable convention-for been made towards standardising the sampling example, the Higgins cyclone.zoAs yet there are of aerosols to ensure that the fractions sampled no commercial samplers for the thoracic are biologically relevant. In 1993 a new convention although several research devices European standard for the definition of size have or are being developed." 22 The clear fractions for measuring airborne particles was rationale is that sampling personal exposure approved by the European Committee for provides a better evaluation of potential risks, Standardisation.' This standard defines three as it has also been shown that personal sampling conventions: inhalable fraction (the mass fraction of airborne particles which is inhaled into the nose or mouth); thoracic frac- h tion (the mass fraction of inhaled particles penetrating beyond the larynx) and respirable ffaction (the mass fraction of inhaled particles penetrating to the unciliated airways). These three criteria are shown in figure 1. The same definitions, for workplace atmospheres, have 0.3 k also been adopted by the International Stand- ards Organisation' and The American Confer- ence of Governmental Industrial Hygienists (ACGIH).'" Aerodynamic diameter (pm) These conventions recognise that only part Figure 1 Inhalable, thoracic, and respirable sampling of the total aerosol is inhaled into the nose or Criteria. exposures are almost invariably higher than the size selective sampling criteria described those found with simultaneously operating earlier. Even with such systems there is still the fixed location monitor~.'~ possibility that wearing a noisy pump may alter Measurement of exposure in ''the person's behaviour and bias the measured exposure leveLZ6 non-occupational groups There are a few studies, mostly carried out Procedures for monitoring personal exposure over the past 10 years, which provide reliable to airborne particles in non-occupational situ- measurements of personal exposure to air- ations have been adapted from those used in borne particles. These studies, which have the workplace. The equipment generally com- been carried out in North America and more prises a small battery operated pump con- recently in Europe, do not generally measure nected to a size selective sampling head the metal content of the aerosol. Most of the containing a filter paper or some other particle data refer to urban or suburban populations, collection system. The main problems associ- generally excluding workers in occupations ated with these techniques arise because of the where there is known exposure to aerosols, and low airborne mass concentrations in non- most measurements have been made over 24 occupational situations and the necessity to hours. In total, there are about 2000 published sample over relatively long periods, typically 24 measurements from various subgroups within hours or more. As the person should wear the society and only about 1200 measurements pump throughout the sampling period, except made in relation to currently accepted biologi- during sleep, it must be lightweight and quiet cally relevant sampling conventions-such as and this limits the maximum airflow. As a con- thoracic aerosol (table). sequence, the minimum detectable mass con- In these studies the mean thoracic aerosol centration is generally about 10 &m' over 24 concentration ranged from 42 &m3 for the hours?` The detection limit for components of study of Seaton et a P of elderly volunteers in the aerosol, such as lead, will be determined by Belfast to 105 pglm' for children in The the sensitivityof the analyticaltechniques used, nether land^.^' Most of these measurements but could be much less than for gravimetric were made over 24 hour periods and where this assessments, perhaps about 10ng/m3for lead.25 was not the case (Birmingham, UK) the data ! During personal sampling the entry into the have been adjusted to estimate 24 hour sampling head is normally located close to the exposure by using the average fixed point person's nose and mouth, typically on the out- monitoring data to represent the unmeasured side of their clothing on a lapel to give an esti- time. Other studies have shown that during the mate of the concentration of particles inhaled. night indoor, outdoor, and personal concentra- During the hours of sleep the sampler is gener- tions agree quite well and so this adjustment ally placed in the individual's bedroom. To seems justifiable.MThe maximum individual ensure that the noise of the sampler does not thoracic concentration was 971 pglm' (24 hour disturb the person during sleep some investiga- average), measured by Loiy et al in Phil- tors have devised sound insulated boxes to lipsburg, NJ." muffle the pump noise.u These systems may also be used at other times when the noise Sampling human exposure to lead aerosol would be distracting-for example, while Sampling of lead aerosol in occupational situa- watching television or reading. It is necessary tions is one of the few cases where current to replace the pump battery about every 8 sampling practice does not require biologically hours or to use mains electricity overnight. relevant size selective sampling, although the There are several sampling heads that have method has been standardised. There is no been used to measure personal exposure to air- agreement about how lead should be sampled borne particles. Measurements made more outside the workplace. than 15 years ago may have been made as total There is also very little information about suspended particulate (TSP), although this is a the size range of lead aerosols in ambient air in misnomer and these instruments generally had the United Kingdom. From the Quality of an undefined sampling characteristic which Urban Air Review Group (QUARG) report on would have included some, but not all, of the airborne particulate matter in the United larger particles suspended in the air. More Kingdom3'about 90% of the lead aerosol is in recently measurements have been made with the fine fraction (aerodynamic diameter c2.5 Measuremmts of thorack aerosol exposure in various groups of non-occupationally exposed subjects oerer 24 hours SdY Location Iioy ez UP' Phillipsburg, USA P e U i et CP Riverside,U S A Janssen et aP9 Amsterdam and Wageningen, Netherlands Mark et d2 Central Birmingham, UK Mark er dz East Birmingham, UK Janssen et d3 Amsterdam, Netherlands Seaton et ha Edinburgh, UK Seaton et aP' Belfast, UK Non-smoking volunteers Representative selection of non-smokers Children aged IO- 12 y Volunteers, smokers, and non-smokers Volunteers,smokers, and non-smokers Non-smoking volunteers aged 50-70 y Volunteers aged >60 y Volunteers aged > 60 y `Population weighted median exposure rather than geomemc mean. tMaximum is the estimated 98th percentile. SGeomemc mean estimated from arithmetic mean (SD). Estimatedgeometricmean and maximum, see text for explanation. 14 178 45 22 15 37 57 52 189 171 301 127 178 262 57 54 66 98' 1OS+ 709 479 59$ 57 42 97 1 285t 195 260 120 113 418 288 C h d ,Aitken pm), probably associated with particles <lpm in diameter.33It is possible that there may be greater concentrations of lead in the coarse fraction when lead contaminated dust, which is attached to other settled dust in the environment, is then resuspended. Because of the bonding of particles to surfaces,.most dust which is resuspended will have a diameter >1-5 p ~ n . 'M~easurements of lead in household dust suggest that there may be substantial amounts of lead associated with larger particles." When considering which size fraction or fractions are most appropriate for sampling human exposure to lead aerosol, we should consider the fate of the different size particles. Ideally, the measurement system should be capable of collectirig the inhalable and respir- able fractions. For lead particles which deposit in the alveolar region there is almost 100% absorption. However, only about 10% of the lead that is swallowed by adults is absorbed into the blood, although the proportion absorbed by children is higher, perhaps 30% to 40%.'6 The inhalable fraction less the respirable fraction represents that part of the aerosol which could potentially be absorbed through the gut and the respirable fraction that part which would be absorbed through the lung. These two measurements could then be combined as a weighted sum, based on the approximate efficiency of intake from each route of exposure. Alternatively, such data would provide reliable input data for developing the physiologically based pharmacokinetic models available for lead.37 The United Kingdom Expert Panel on Air Quality Standards has recommended that there should be a standard of 0.25 pglm' for lead, averaged over a year." This standard is intended to protect the intellectual development of young people from inhaled lead aerosol. Unfortunately, the standard does not precisely specify the sampling procedure to be used and so it is unclear what size fraction it includes. For the fixed point measurements made by the United Kingdom Department of Environment, Transport, and the Regions this is probably unimportant because their samplers are generally located far from sources of coarse particles and most available samplers should give comparable results. However, differences in instrumentation would be much more important if the air quality standard were applied to personal exposure monitoring. There is very little information on exposure to lead aerosol in the general population and the data that do exist have been obtained with samplers that do not conform to the agreed size selective sampling riter ria.'^ Nevertheless, these data suggest that week long average lead exposures in Swedish adults range between 0.04 and 0.09 @m3, as total dust. In the United Kingdom measurements of lead in the general atmosphere suggest that it makes up about 0.2% by mass of the thoracic (I'M,,) aerosol.Applying this factor to the data shown in the table, assuming most of the variability in measured values is associated with the people rather than the day of sampling, suggeststhat the average exposure level would be about 0.1 &rn3 and only about 5% of adults might exceed the United Kingdom air quality standard for lead. However, with the data from children there might be about half who could exceed the standard and it is this group who are especially at risk fiom lead exposure. Also, children living in homes where there was excessive lead contamination from flaking paint or from localised industrial sources might have personal exposures well above the standard. Passive aerosol samplers The appropriate averaging time for measurement should be dictated by the biological half life of the pollutant in the body with short term variations in exposure level being less relevant.39In the case of lead the biological half life is about 35 days'3 and this would be an appropriate averaging period. Clearly, measurement of personal lead exposure with conventional sampling strategies over such a long period is impracticable and the available data do not easily 'allow extrapolation to estimate such exposure. It might be possible to sample 24 hour average exposure on several occasions throughout a 5 week period and this would approximate to the long term exposure level, but this would still be costly and inconvenient for the subjects. Sampling of gases and vapours has been revoluuonised by the introduction of small lightweight samplers that do not have a pump but rely on the principle of diffusion to collect the material. Over the past 10 years there have been attempts to develop a similar approach for aerosols. The most promising of these, shown in figure 2, comprises a 25 nun diameter electret material held within an electrically conducting holder.40 Electrets are polymers that have been treated to induce a permanent elecmcal charge within their structure, with the charge at the surface of the polymer as high as 1000 V. The sampler has a 1 cm gap between the surface of the electret and the front plate of the sampler and any pamcles that pass into this volume will drift towards the electret with a velocity proportional to their electrical mobility. The rate of collection of particles is relatively independent of their size, charge Figure 2 (A) Passive aerosol sampler deweloped by the United Kingdom Health and S a f e 5 Laboratory. s i distribution, or velocity, but this is highly Finally, selecting an appropriate averaging dependent on the electrical mobility and this time provides a particular challenge for all haz- depends on the chemical composition of the ardous aerosols that have a chronic effect on aerosol.*' health. Here there is an advantage in having The sampler is small, lightweight (about 15 sample averaging times as long as possible, g), and not intrusive. It may be worn for consistent with the biological half life of the extended periods and requires almost no pollutant, so as to reduce the variability associ- operator maintenance over that period. There- ated with short term samples. Continuous per- fore it would be an excellent device to assess sonal sampling for aerosols over several days or aerosol exposure of non-occupational groups weeks is currently impracticable, but further and could be used to measure exposure to development of passive samplers offers the lead, although there are several limitations that possibility of adopting such strategies in the make its use problematic. The rate of collec- future. The key difficulty will be to develop tion by the device is very low. In a study which systems which combine the need to obtain size is currently underway to assess the use of the selective samples with that of measuring over i device in assessing urban domestic aerosols, prolonged periods. collected masses over 2 week periods were of the order of 50 pg. Also, in its present form the sampler is not size selective. In principle at least, it collects all sizes with equal efficiency, although the range of sizes over which this holds has not been measured. Finally, as We are grateful to Dr Nigel Crawford and Mr Fintan Hurley at the Institute of Occupational Medicine for their helpful comments, and to Dr Lee Kenny at the Health and Safety Laboratory (Sheffield)for the photograph of the passive aerosol sampler. already described, the collection rate is dependent on the electrical mobility, and hence composition of the particles. It is therefore necessary to calibrate the collection rate for different types of aerosol by comparing the passive sampler with a conventional pumped sampler in each case. This may lead to the requirement for material specific designs, coupled with appropriate analytical methods to be produced. Further work is necessary if these limitations are to be overcome. 1 Walton WH. The name, hazards and assessment of occupational exposure to airborne asbestos dust: a review. Ann Ocnrp Hyg 1982;25: 117-247. 2 Koae RN. Final npon of the M i m ' Phthisis Comminee. Pretoria: GPSO, 1919. 3 Green HL,Watson HH. P h y W mezhodrfor the estimation of dust hazard in industry. London: Medical Research Council, 1935. (Special Repons Series No 199.) 4 DunmoreJH, Hamilton RJ, Smith DSC. An instrument for the sampling of respirable dust for subsequent gravimetric assessment. Journal of Scientific Immmenu 1964;41:669- 72. 5 Orenstein AJ, ed. Recommendations adopted by the Pneumoconiosis Conference. In: Aoceedings of the BwwnomnZiOsir Confemce, 3ohannesberg, 610-21. London: Churchill, Conclusions There have been many developments in the sampling of aerosols over the past decade. In this paper we have highlighted the important attempts to standardise the fraction of the aerosol that is sampled so that it roughly corresponds to the fraction penetrating to sections of the respiratory tract. There is now, for the first time, agreement between all those involved in sampling workplace aerosols to use the same definitions so that measurements of respirable aerosol made in the United States, Europe, and other parts of the world are comparable. It would be advantageousfor those involved with sampling non-occupational aerosols to ensure that their measurements are obtained to the same standard. A consequence of moving to biologically relevant size fractions is the desirability of obtaining measurements of multiple size fractions for some aerosols. We have used the example of lead, where measurement of the inhalable and respirable aerosol might be appropriate. Use of this type of approach would certainly allow a more precise estimate of the biologically relevant exposure to be made and would enable more accurate estimates of risk from inhaled aerosol. It is more important to select an appropriate sampling system for personal exposure measurements than for tixed point monitoring because of the 1959. 6 Jacobsen M, Rae S, Walton WH,ez al. The relationship between pneumoconiosis and dust exposure in British coal mines. In: Walton WH, ed. Inhaled panidcs III. Old Woking: Unwin, 1971:903-17. 7 Rogan JM, Attlield MD, Jacobsen M, n al. Role of dust in the working environment in the development of chronic bronchitis in British coalminers. Br 7 Ind Med 196730: 217-26. 8 British Standards Institute. RMpluce amtospherez: size fiaction dejiniriar for measuremenrr of airborne panicles. London: BSI. 1993. (BS EN 481.) 9 International 'Standaids Organisation (ISO). Air qual%; parride size f m c t i a definitions for heath-dated samplhg. Geneva: ISO, 1993. ( I S 0 CD 7708.) 10 American Conference of Governmental Industrial Hygien- ists (ACGIH). Threshold limit valuesfor chemical subsrnnces andphysical agenrr. Cincinnati, Ohio:ACGM, 1993. 11 Ogden TL, Birken JL.The human head as a dust sampler. Walton WH, ed. Inhaled panicles IV. Oxford: Pergamon, 1977. 12 Ogden TL, Birken JL,Gibson H . Impmmenrr fo dust meoruring rechniques. Edinburgh: Institute of Occupational Medicine, 1977. Final report on CEC Contract 6253-221 8/016. 13 Vincent JH,Armbruster L. On the quantitative inhalability of airborne dust. Ann O m p Hyg 1981;24:245-8. 14 Vincent JH,Mark D. Application of blunt sampler theory to the definition and measurement of inhalable dust. In. W H Walton, ed. Inhaled parricks V. Pergamon Press, Oxford 1982:319. 15 Armbruster L Bruer H (1982). Investigations into defining inhalable dust. In: WH Walton, ed. Inhaled particles V. Oxford: Pergamon Press, 1982:21-32. 16 International Commission on Radiological Protection. Human respiratory tract model for radiological protection. Ann ICRP 1994;24:1482. 17 Soderholm SC. Proposed international conventions for particle sue selective sampling.Ann Oca@ Hyg 1989;33:30120. 18 Vincent JH.Aerosol science for indusmnl hygienuu. Oxford: Elseivier, 1995. 19 Mark D, Vincent JH. A new personal sampler for airborne total dust in workplaces. Ann Occup Hyg 1986;30:89-102. 20 Higgins RI, Dewell P. A gravimemc size-selective personal dust sampler. Davies EX,ed.Inhaled particles N.Oxford: Pergamon, 1967575-86. 21 Aitken RJ, Vincent JH, Mark D. Application of porous likelihood of resuspended dust playing a greater part and this should be carefully considered when sampling aerosols outside the workplace. foams as sue selectors for biologically relevant samplers. Applied Occupational and Envimnmentnl Hygiene 1993;8: 363-9. 22 Kenny LC, Stancliffe JD,Crook B et al. Development and evaluaaon of a personal bioaerosol sampler. Yournal Of Aerosol Science 1998;29:S497-8. 752 Cherrie,A h 23 CherrieN.The effect of room size and general ventilation on the relationship between near and far-field concenuadons. Applied Occupariaal and Environmental H y & e (in press). 24 Janssen N, Hoek G, Harssema H, et al. Personal samplingof airborne panicles: method performance and data quality. Journal ofExposureAndysir and Envimnmenrnl Eprdemlorogy 1998;8:37-49. 25 Vahter M,Berglung M,Lind B, et al. Personal monitoring of lead and cadmium exposure: a Swedish study with reference to methodological aspects. ScandJ ubrk E m ' m Health 1991;t7:65-74. 26 Cherrie JWL,ynch G, Bod BS, et al. Does the wearing of samplingpumps affect exposure?.Ann Occup Hyg 1993;38: 827-38. 27 Fugas M. Monitoring of personal exposures to air pollutants: subjects' experience.A d i v za Hy&u Rad0 I Tmsikdogiiu 1995;46:23741. 28 Seaton A, Godden D, Cherrie J, er al. Air Popuria and cnrdiovascular disease: an inuesrigonbn of the d a n h h i p be- parriculore air poUurion and Mood coa&ation factors. Final repon to the Deparnnenr 4 H e a W fmprojeu 121/6281. Aberdeen: University of Aberdeen, 1999. 29 JanssenNAH, Hoek G, Harssema H, e6 al. Childhood exposure to PM,,: relation between personal, classtoom, and outdoor concentrations. Occup E m ' m Med 1997;5488t?94. 30 Pellizzari ED, Thomas KW, Clayton CA, et al. Particle total exposure ass~~smenr nt e t h c d o b ~(PTEAM); Rimmi&, Cdiibrnia pilot S M .Vol 1. Research Triangle Park Environmental Protection Agency, 1993. (EPA/600/SR-93/ 050.) 31 Lioy PJ, WaldmanJM, Buckley T, et d.The personal,indoor and outdoor concentrationsof PM,,measured in an indusmal community during winter. Atmospheric Envimmnent 1990;24:5746. 32 Harrison RM,Brirnblecombe P, Dement RG, et al. Airbonv pami& muor in the United Kingdom. 7Xrd Rcpm ofhe Quality of UhanAir RNinu Gmup.Birmingham: University of Bumingham, 1996. 33 International Programme on Chemical Safety. I n 0 6 1ead.Geneva: World Health Organisation, 1995.(Environ- mental health criteria 165.) 34 Thatcher TL, Layton DW. Deposiaon, resuspension and oenetration of oarticla wirhin a residence. Atmdwnc Em'mmenr 1965;291487-97. 35 Wang EY, Willis RD, Buckley TJ,et al. The relationship between the dust lead concenmtion and the partide sizes of household dusts collected io Jersey City residences. Appl &cup E m n m Hyg 1996;11:99-206. 36 Ziegler EE,Edwards BB, Jensen RL,et al. Absorption and retention of lead in infants.Adriur Res 1978;12:29-34. 37 White PD, Van Leeuwen P, D w i a BD, et al.The conceptupl s t r u ~ ~ h omf the integrated exposure uptake biokinefic model for lead in children. Em'm Healrh Anpra 1998;106(suppl6):1513-30. 38 Expert Panel on Air Quality Standards. Lad. London: W S O , 1998. 39 Rappaport S.M.Assessment of long-termesposum to toxic substances in air. Ann 0ccup Hyg 1991;35:61-121. 40 Brown RC, Wake D, Thorpe M, et al. A passive sampler for airborne dust using an elecuct. Journal of A d Science 1992;23S623-6 41 Brown RC, Wake D, Tborpe M, et al. Prelimiapry assessment of a device for passive sampling of airborne pnrdculate. Ann 0ccup Hyg 1994;38:303-18. 42 Mark D, Upton SL, Lyons CP, et d.Personal exposure measurements of the general public to aunospbaic panicles. Ann O a p Hyg 1997;41(suppl 1):7o(M. 43 Janssen NAH, Hoek G, Brunekreef B, n al. pmoanl sampling of particles in adults: relanon among personal, indoor and outdoor concentrations. Am 3 Epidmriol 1998;147:537-47. Vancouver style All manuscripts submitted to Occup E m ' m Med should conform to the uniform requirements for manuscripts submitted to biomedi- cal journals (known as the Vancouver style.) Occup Environ Med, together with many other international biomedical journals, has agreed to accept articles prepared in accordance with the Vancouver style. The style (described in fullin the3AMA[11) is intended to standardise requirements for authors, and is the same as in this issue. References should be numbered consecutively in the order in which they are first mentioned in the text by Arabic numerals on the line in square brackets on each occasion the reference is cited (Manson[l] confirmed other reports[2] [3][4][5]). In future references to papers submitted to Occup E n v i m Med should include: the names of all authors if there are three or less or, if there are more, the first three followed by et al; the title of journal articles or book chapters;the titles of journals abbreviated according to the style of Index Medicus; and the first and final page numbers of the article or chapter. Titles not in Index Medicus should be given in full. Examples of common forms of references are: International Committee of Medical Journal Editors. Uniform requiremars for manuscripts submitted to biomed journals. 1993;2692282-6. Soter NA, Wasserman SI, Austen KF. Cold urticaria: release into the circulation of hismaine and eosinophil chemotactic factor of anaphylaxis during cold challenge. N EnglJ Med 1976;294:607-90. Weinstein L,S w a m MN.Pathogenicproperties of invad- ingmicrosgsnisms. In:Sodeman WA Jr. Sodeman WA, eds. Pathdogrc physwbgy, mechaniFnr of duecue. Philadei- phia: W B Saunders, 1974:457-72. Conceptual model for assessment of dermal exposure Thomas Schneider, Roe1 Vermeulen, Derk H Brouwer, John W Cheme, Hans Kromhout, Christian L Fogh Abstract exposure limits in an analogous way to those Dermal exposure, primarily to pesticides, for inhalational exposure.*'However, there has has been measured for almost half a been criticism of the existing methods of century. Compared with exposure by measurement of dermal exposure because they inhalation, limited progress has been determine the mass of contaminant either made towards standardisation of methods depositing on the skin or retained on the skin at of measurement and development ofbio- the end of the exposure period.' logically relevant exposure measures. It is Hazardous substances on the dermal surface suggested that the absence of a consistent will be taken up continuously into the body terminology and a theoretical model has through the stratum corneum and the epider- been an important cause of this lack of mis towards the dermis where they or their progress. Therefore, a consistent termi- dermal metabolites will be removed by the nology based on a multicompartment blood flow. The transport process is driven by model for assessment of dermal exposure the concentration gradient between the dermal is proposed that describes the transport of surface and the perfused tissue. The risk arising contaminant mass from the source of the from dermal exposure is thus firstly related to hazardous substance to the surface of the the time dependent concentration of a sub- skin. Six compartments and two barriers stance on the dermal surface rather than the together with eight mass transport proc- mass of material on that surface at any given esses are described. With the model struc- time. Mass is nevertheless important when ture, examples are given of what some there is little material available for uptake. existing methods actually measure and Contamination of the skin may arise in many where there are limited, or no, methods different ways. It is possible for hazardous sub- $ National Institute of Occupational Health, Copenhagen, Denmark T Schneider for measuring the relevant mass in a compartment or transport of mass. The importance of measuring the concentra- stances to land on or be absorbed into the skin directly from the air. They may be transferred to the skin from contact with contaminated tion of contaminant and not mass per area surfaces or by submersion of part of the body Environmental and in the skin contaminant layer is stressed, into the substance. Also, the contaminant may Occupational Health Group, Wageningen University, Wageningen, The Netherlands as it is the concentration difference between the skin contamination layer and the perfused tissue that drives uptake. Methods for measuring uptake are cur- be lost from the skin, either by evaporation or some other mechanisms such as washing or abrasion, without being taken up into the body. Finally, the presence of clothing or protective R Vermeulen rently not available. Measurement of garments may m o w the rate at which hazard- H Kromhout mass, concentration, and the transport ous substances come into contact with the skin. TNO Nutrition and Food Research Institute, The Netherlands D H Brouwer processes must be based on a theoretical model. Standardisation of methods of measurement of dermal exposure is strongly recommended. (OCCUEPtwiron Med 1999556~765773) All of these processes are important to consider when making an assessment of dermal exposure and a complete understanding of these complex processes will help in developing an appropriate control strategy. In this paper we have attempted to produce a University of Aberdeen, and Institute of Occupational Medicine, Edinburgh, UK J W Cherrie Rise National Laboratory, Roskilde, Denmark C L Fogh Keywords: dermal exposure; model; measurement methods Exposure to hazardous substances most commonly occurs either by inhalation, ingestion, dermal contact, or some combination of these routes. Occupational hygiene has traditionally focused on exposure by inhalation because it was almost invariably considered to be the most important pathway. Many methods have consistent terminology for assessment of dermal exposure. The terminology is based on a conceptual model of the processes leading to exposure (from the source of a hazardous substance to the surface of the skin). We have also defined several terms related to exposure, which provide a valid basis for investigatingthe risks posed by dermal exposure. It has not been our intention to consider the process of uptake into the body or the derivation of dose estimates from dermal exposure, although oth- Correspondence to: Dr T Schneider, National Institute of Occupational Health, Lerse Parkalle 105, DK 2100 Copenhagen, Denmark. Telephone 0045 been developed to measure exposure levels from inhalation and there is a clear understanding of how such levels should be interpreted to help reduce risk. The situation is less clear for the dermal route of exposure. ers have developed models that could be used in such c o n t e ~ t s'. ~ Conceptual model of dermal exposure 39165295; fax 0045 Practical methods of measurement have been A consistent terminology has to be based on a 39165201; email: ts@ami.dk developed to assess dermal exposure' and pro- coherent and systematic description of dermal Accepted 14 June 1999 posals have been made to develop dermal exposure scenarios. A multicompartment \ \ 766 Schneider, Vermeulen, Bwruer, et al model is an appropriate basis for a terminology as it comprises distinct physical objects or compartments connected by mass transport processes. Models of this type are concerned with what happens-for example, where fingers get contaminated by touching a surfaceand not why it happens-for example, particle adhesion. COMPARTMENTS All compartments are assumed to be well and instantaneously mixed. As a result the concentration in a compartment is described by the amount of mass and distribution volume of the compartment. Six principle compartments are being distinguished in the model (figure): Source (S) Processes or activities, from which a mass is being introduced into any of the compartments, will be considered as sources. Air The air compartment contains vapours and dispersed particles, which are assumed to be homogeneously distributed in the compartment. The total mass of a given substance in the air compartment is well defined and can, in ,.R..d...s..u... ii DAir ,R..d..A..i.r...., LS" It ' Personal behaviour R..d..C..l.o..O..u..t I! I /i ii ,.....a .. ii ii ii ,........... .- ii !j LSk 1; i tRSk.Cloln TCloln,Sk Ij i; RdSk t ipSk Stratum corneum barrier i O v e r - ofthe conceptual model,'comparnenu and rare constants.E=emission (-); Dp=deposirimt (-1; -Ln==re~suosmpennstionnmor;envnar-pto;.rn.act.io.n).(IS--.-...---T..=:t-r--a....nA.Isfe;t.(..-... -1; R=removal +:-. I (---); Rd=redism'burion ( ); - principle, be measured. The compartment volume is given by the size of room or other boundaries, either physical or virtual. Surface contaminant layer (Su) Contaminants on a surface form a layer, which delineates the compartment called the surface contaminant layer. The compartment is assumed to be homogeneous. In principle, all substances belonging to the surface contaminant layer can be identified and thus mass in this compartment can be assessed. The compartment volume is given by the three dimensional volume of this layer. For many practical purposes, a two dimensional representation of this layer will suffice. Outer and inner clothing contaminant layer (CloOut, Cloln) Solid or liquid contaminants at the boundary between outside and the surface of the outer clothing are modelled as the compartment called outer contaminant layer. The fabric separates this compartment from the inner clothing contaminant layer. For simplicity of the model the fabric is described as a barrier (mass transfer rate limiting) having the property to retain mass (M,,). If the mass of the hazardous substance in the outer clothing contaminant layer compartment is MCmt then some of this material will be transported through the clothing to the compartment called inner clothing contaminant layer. The mass in the inner clothing compartment is M,-loln=M,-ld)ur-MRaF4ovromu'any practical purposes a two dimensional representation of these compartments will suffice. Skin contaminant layer (Sk) On the skin, contaminants, sweat, skin oil, and barrier cream (if applied) form a layer. This layer constitutes the skin contaminant layer. The compartment is assumed to be homogeneous. In principle, all substances belonging to the skin contaminant layer can be identified and thus the mass in this compartment can be identified and measured. The compartment volume is given by the three dimensional volume of this layer. The conventional two dimensional representation of this layer is an oversimplification, which has contributed to the confusion about the principles involved in the choice of measurement and interpretation of dermal contamination in terms of dermal uptake. MASS TRANSPORT PROCESSES The mass transport from the source to the compartments and sinks in the system is shown in the figure. Below the horizontal dotted line in the figure a person's movement begins to influence the transport processes. Two units are used to measure transport of mass; g d and g.event-'. For processes called events, it is important to describe the number of events within a reference period, typically 8 hours. The mass transport can be divided into eight distinct processes as described later and in table 1. Emission (E)-Emission (E) is the transport of substances into the air, onto surfaces, outer clothing, and the skin contaminant layer from all primary sources. Evaporation of liquids or emission of droplets or particles into the air gives rise to emission of contaminant mass to the air. For aerosols we restrict this emission pathway to those with aerodynamic diameter <lo0 J.UTI so that sedimentation is relatively unimportant. Emission to the different surfaces in the model can arise from splashing, spilling, immersion, and impaction of large particles. Splashing is the emission of large droplets the trajectory of which towards the surfaces is unaffected by air movement, whereas spilling is the event by which a liquid or powder is spilled on a surface. Immersion is an event whereby a part of the body is submerged into a liquid or a powder. Impaction is the process by which large particles are generated at the source and ejected from that source to impact onto surfaces. The emission rate is given as either g.s-l or as g.event-'. Deposition (Dp)-Deposition @p) is the transport of substances from the air to surfaces, Table 1 Compartment desmptnrs Compamnent Dejinilion of memi Source Air Surface contaminant layer Outer clothing Contaminant layer Inner clothing contaminant layer Skin conraminantlayer Mass of hazardous sibstance available for emission Concenaation of a hazardous substance in the source M a s s of substance in the air compartment Volume of the air compartment Concenuationof hazardous substance in the air Mass of a hazardous substance in the surface contaminant layer Concentrationof a hazardous substance on the surface Area of surface which is contaminated with hazardous substance Mass of hazardous substance in the outer clothmg contaminantlayer compartment Concentrationof a hazardous substance in the outer clothing compartment Area of the outer clothing which is contaminatedwith hazardous substance Mass of hazardous substance in the inner clorhing contaminantlayer compartment Concentration of a hazardous substance in the inner clothing compamnent Area of the inner clothing which is contaminated with hazardous substance Mass of hazardous substance on the skin surface Concentration of hazardous substance in the skin contaminantlayer Area of the skin which is contaminatedwith hazardoussubstance &= mass of all other substancesin a particular compartment. Symbol h.1 C. MA" VA" CA" WV CS" 4, &!&I CC, ,&I& Ma, c,,, &bob Wb CS, & Relacion L)M sJ(Ms,+ Units g g.g-', g.m-' B m3 g/m' g g.kg-` cm2 B MC,ai(Lz+k*.3g.kg-' Md(%,+.%3 Md(Ms,+*%-) cm2 g g.k&-' an2 g e-ki-' an' , Schneider, Vermeulen,Brouwer, et al outer clothing, and the skin contaminant layer. Deposition can be of mass as either solid, liquid, or vapour. The time dependent mass deposition rate Dp(t), in units of g.a' ,can be represented by the deposition velocity v(t) in units of cm.s-' as follows: Dp(Q = v(t) .C(t) .A , (1) where C(t) is concentration in air outside the boundary layer and A, is the area of the two dimensional representation of the compart- ment, where X is either Su for surface, Clo for clothing, or Sk for skin). Resuspension or evaporation (L)-Resuspen- sion or evaporation (L)is the transport of sub- stances from surfaces, outer clothing, and the skin contaminant layer to the air, as partic (resuspension), vapours, or both. Evaporation is a continuous process driven by diffusion. Mechanical forces cause resuspension. If resus- pension is caused by a single mechanical impact then the transport of mass is conven- iently modelled as a transfer of mass per single event. If resuspension is caused by a sudden flow of air along a surface, there is a short ini- tial peak of resuspended mass followed by a decay of resuspended mass per time.' Thus resuspension can, in general, be meaured as a transport of mass per single event. Transfer (TI-Transfer 0is the transport of substances by direct contact between surface, skin, and outer and inner clothing Contaminant layers in a direction towards the worker. Trans- fer from the surface to the skin contaminant layer is event based and takes place &om a small area of actual contact. That area could be considerably smaller than the surface area of the body part involved in the contact. The actual surface area of the skin that will be con- taminated, as well as the efficiency of mass transfer, will depend on the actual contact-for example, single pressure, or movements of the skin along the surface. Removal (R)-Removal (R) is the transport of substances by direct contact between skin, inner and outer clothing, and surface contami- nant layer in a direction away from a worker. Removal thus is defined as an event based transport in the opposite direction of transfer. Redistribution (Rd)- If contaminants in the air, on the surface, clothing, or skin are not homogeneously distributed, or different parts of the body such as palm, neck, and trunk need to be distinguished, the compartments can be subdivided initially into subcompartments. Redistribution then is the transport of sub- stances from a subcompartment to another subcompartment of the same type. Redistribu- tion of contaminants from one part of the skin contaminant layer to another can occur as a result of touching the face with contaminated fingers. Also, fabric wetting can redistribute liquid contaminants. Decontamination (D)-Decontamination (D) is the deliberate transport of contamination from the system-for example, ventilation of room air, cleaning of room surfaces and outer clothing, or washing material off the skin. The air compartment is decontaminated by the combined effect of natural And mechanical ven- tilation. Cleaning of taminated surfaces, chang- ing of clothing, and cleaning of skin all result in permanent loss of mass from the system and thus are decontamination processes. By con- trast, brushing dust off clothing transports particulate mass to the room air and thus is resuspension, not decontamination. Penetration and permeation (P)-Penetration and permeation (P) both involve transport of substances through a rate limiting barrier- such as clothing or the stratum corneum. Pen- etration is transport caused by external pres- sure, capillary penetration, and evaporation- condensation. Permeation always involves diffusion. Transport of contaminants through perrne- able clothing occurs by aerosol penetration and liquid transport. External air pressure can be considered to be the driving force for penetra- tion of aerosols through fabric,8 whereas the mechanisms of liquid transport are capillary penetration, pressure penetration, impact pen- etration, and evaporation-condensation? Mass transport through non-permeable clothing is a diffusion process driven by concentration. The rate of mass transport through the stra- tum corneum PSk(t)can be represented by a permeability coefficient K(t), the concentra- tion difference over the stratum corneum C,(t), and skin contaminant layer area A(t) as: &(t) = K(t) Csk(t) * A(t) (2) The area of the skin contaminant layer can be time dependent, which for example is the case for a drop that dries while on the skin. Roed et d ohave shownthat very small partic)es may penetrate into the stratum corneum but their fate is not known. Concentration The model describes transport of mass of a given substance, which is a conserved quantity provided we neglect chemical reactions. How- ever, several transport processes are driven by the compartment concentration. For the air compartment concentration is readily obtained from mass because an air compartment can be defined so that it has a constant volume. For the surface and skin contaminant layers, volume is defined by the total amount of mate- rial present (table 2). If the skin contaminant layer compartment only contains liquids the concentration C, of a hazardous substance of mass M,, is given by the ratio Csk=MSb!MSk+MS,ochcr (3) where M,, orher is the mass of all other liquid substances. This concentration can be readily transformed into the molar concentration, which is more relevant for the skin contami- nant layer. As a lirst approximation and for substances of low volatility C,, can be consid- ered to be the concentration in the bulk liquid. The concept of concentration is more complicated for particles as they are discrete entities and solid substances must dissolve to diffuse through the stratum corneum. For soluble or leachable substances it is the concentration in the wet layer around the indi- vidual particle that has to be used for C, in equation 2.This may mean that uptake will be limited by the rate of dissolution rather than diffusion -through the stratum corneum. Fur- Table 2 Mass tranzport process descriptions Emission Deposition Mass of hazardous substance emitted into air from primary sources per unit time Mass of hazardous substance emitted to surface contaminant layer by splashing, spilling and ejection of particles from ,E primary sources per unit time per event % M a s s of hazardoussubstanceemitted to the outer clothing contaminantlayer for a particular worker by splashing, spilling and ejecdon of particles from primary sources per unit time per event ,E Mass of hazardous substance emitted to skin contaminant layer for a p a r d d a r worker by splashing,spilling and ejecdon of particles from primary sources per unit time per event L Mass of hazardous substance deposited h m air compartment to the surfaces per unit time DPS" Mass of hazardous substance deposited from the air compamnenr to the outer clothing contaminant layer for a particular Resuspension or evaporation Transfer Removal Redismbution Decontamination worker per unit time DPDdhn Mass of hazardous substance deposited from the air compamnent to the skin contaminant layer for a particular worker per unit time DP, Mass of hazardous substance lost from the surface contaminant layer to the air compartment by evaporation per unit time or by resuspension or evaporation per event L" Mass of hazardous substance lost from the outer clothmg contaminantlayer of a particular worker to the air compartment by evaporation per unit time or by resuspension or evaporation per event Mass of hazardous substance lost from the skin contaminant layer of a particular worker to the air compamnent by Lcloom evaporation per unit time or by resuspension or evaporation per event r, Mass of hazardous substance uansferred from the surface contaminant layer to the skin contaminant layer for a particular worker by direct contact per event Ts, SL Mass of hazardous substance uansferred from the surface contaminant layer to the outer clothing contaminant layer for a particular worker by direct contact per event Mass of hazardous substance uansferred from the inner clothing contaminant layer to the skin contaminant layer for a TS".oochn particular worker by direct contact per event M a s s of hazardous substance transferred from the outer clothing contaminant layer to the skin contaminant layer for a Ta, particular worker by direct contact per event Mass of hazardous substance removed from the skin contaminant layer of a parricular worker and uansferred to the surface Tclme contaminant layer by direct contact per event %b Mass of hazardoussubstance removed from the skin contaminant layer and transferred to the inner clothing contaminant layer of a particular worker by directcontact per event %, Mass of hazardous substance removed fromthe skin contaminant layer and transferred to the outer clothing contaminant layer of a particular worker by direct contact per event M a s s of hazardous substance removed from the outer clothing contaminant layer of a particular worker and transferred to %,oochn the surface contaminant layer by direct contact per event L Mass of hazardous substance transferred from one pan of air compartment to other part per event Rd, M a s s of hazardous substance transferred from one part of contaminantlayer to other pan per event Us Mass of hazardous substance transferred from one part of inner clothing contaminant layer to other part per event Rd, Mass of hazardous substance transferred from one pan of outer clothing contaminant layer to other part per event Mass of hazardous substance transferred from one pan of skin contaminant layer to other pan per event %dJw ma Mass of hazardous substance removed from the air compartment by vendtation per unit of rime Dk Mass of hazardous substance removed from the surface contaminant layer by deliberatedecontaminationper event D, Mass of hazardous substance removed from the outer clothing concaminantlayer by deliberatedecontaminationper event D, Mass of hazardous substance removed fromthe inner clothing contaminantlayer by deliberatedecontaminationper event D, Mass of hazardous substance removed from the skin contaminant layer by deliberatedecontaminationper event Da Mass of hazardoussubstance transported through stratum corneum Mass of hazardous substance transported from the outer clothing contaminant layer to the inner clothing contaminantlayer p?p Penetration and permeation per unit time P~~ Mass of hazardous substance transported from the inner clothing contaminant layer to the outer clothing contaminant layer per unit time PckoULaaln g.s -I g.s-' g.evenf ' g.s-1 g.ewnt-' gd gment-' g.s-l g.s-' g.s-1 g.s-' g.eventP g.s-' g.eVmtP g.9 g even-' g.event-' g.event" g.evenf' g.evenr-' g.evmt-' g.event-' g.eveni' g.event-' g.eVent+ g.eveot-1 g.eveni' g.went-l g.event-' gs-' g.event-L g.cvent-1 g.event-' g.event" g.s" g.s-' g.s-1 &air comparment; Su=surfacecontaminantlayer; Sk=skincontaminantlayer; CloOut=outer clorhingconfaminant layer; cfoh=hcelorthins layer, thermore, if uniform deposition takes place a monolayer of particles may be built up until the entire surface is covered. As a multilayer forms, the additional mass will have less and less influence on uptake and may be likely to dislodge and fall off. For this reason the mass of particles in the skin contaminant layer may only have limited relevance to uptake. Measurement methods of skin and surface contamination Depending on the perspective from which the exposure scenario is investigated a range of measurement approaches have been developed from the 1950s onwards." Table 3 lists methods in common use for skin and surface contamination. Several of these methods are used to assess both compartment mass-for example, total mass in surface layer-and mass transport processes-for example, dislodgeable mass. A clear distinction is not always made and this can create confusion about determination and interpretation of sampling efficiencies. As an example, measurement methods should aim to have 100% sampling efficiency where they intend to assess compartment mass. On the other hand, a 100% sampling efficiency for a wipe test to measure the transfer to skin upon contact with a surface is not necessarily desirable.' The amount that can be transported from surfaces to the skin or outer clothing contaminant compartments depends on the type of surface, the contaminant, and on the forces acting on the contaminant layer, rather than how much mass is present in the compartment. Measurement of transport must therefore be based on a model. Several routes may be followed. The transport process fiom surfaces could be simulated with,for example, a standardised instrumental sampling method or standardised events. Transport can also be measured during actual field conditions for an exposed population, with the result summarised in the form of a distribution of mass transport. Such reported distributions have been used for risk assessment-for example, transfer from soil giving the resdts as soil to skin adherence." Deposition from air is more predictable and Schneider and S t ~ k h o l mh~av~e proposed a theoretical model dependent on particle size for the transport of airborne dust onto the ocular surface, in relative deposition velocities. This could be called the ocular deposition fkaction, by analogy with the inhalable fiaction. Roed et a l ' O have given experimental data for 770 Schneder, Vermeulen,Brouwer, et al Table 3 Measurement methods @or a review see Carson et al") Method Principle of sampling Samplzng area definim Measured iomparcmnu mass* Mearured zranspon pmcerrt Refmerue W Fluorescence In situ Portable x ray fluorescence monitor In sim 0.1-2 m' Instrument defined MS, Wet wipe ,Manual wiping Wet wipe Mechanised wiping Gelatine foil Surface dust lifting Fixed pressure dislodgeable None or template Instrument defined 1 :1 transfer of dust layer residue sampler Mechanical transfer in situ 10-20 an x length sampled Dislodgeable foliar residue sampling Surface removal Punches 0.1-2.5an', total 100 cm' Adhesive tape skin stripping 1:1 transfer of dust layer Hand wash Wash with water or alcohol Total hand surface Patch Passive Sampler defined Whole body Passive Body parts SUR Resuspensionby air jet Instrument defined STEPP Resuspensionby impact Instrument defined Microvacuuming FLEC Resuspensionby suction Instrument defined Evaporationby aidow Instrument defined TEWL Evaporation Instrument defined MS".MS, MSU .%" MS" All processes related to the surface contaminant layer TSGWTs,c-, &u, k t C b l n ) kkC, TS".SW TS,'l&, Tsu.sw Tsu.c-t 13-17 18 19 20 21 T S G " T,c-r 22 23 24 25 26 26 27 28 29 30 31 'Defmiaon of symbols as for table 1 . +Mass transport process descnptions as for table 2. mean deposition velocities on arms and other body parts for people seated in a test room, which likewise could form a basis for defining other dermal deposition fractions. There is a close analogy to requirements for sampling inhalable airborne particles. Historically, fractions with biologically relevant parti- cle size have been based on experimental data that reflect mean deposition efficiencies in the airways. An alternative approach has been to define the fraction as that obtained by a given sampling instrument. Initially there were different proposals for the size fkactions, and samplers were not available which could measure according to these criteria. However, better experimental data on deposition efficiencies has resulted in the adoption of an international standard for the definition of the respirable, thoracic, and inhalable fraction^.'^ Recent development has resulted in samplers, which have aspiration efficienciesthat match the definition of the fraction^.'^ In the following section some strategies for measuring with the model compartment mass and mass transport processes will be discussed, highlighting several problems in conventional approaches to measurement and possible routes forward. Surface contaminant layer For surfaces, several in situ methods are available, but they can only measure a limited range of substances, and never total mass. Portable x ray fluorescence analysis is one example of this type of analysis."Methods based on removalsuch as adhesive tape sampling-can have a high sampling efficiency. The efficiency in field use is usually estimated with consecutive sampling of the same area. However, this is based on a circular argument, as only the contaminant that the tape can actually remove will ever be sampled.'' Clothing contaminant layers In the model clothing was divided into an outer and inner contaminant layer, with a buffer to represent the mass residing inside the clothing which does not come into contact with surfaces or skin, respectively. Measurement of the entire mass in the clothing would in principle be no problem and if this were done for nonpermeable fabrics theoretically the outer contaminant layer would be measured. To our knowledge a method which measures the inner contaminant layer for permeable layers is not available. i ! COMPARTMENTS The mass in a compartment at a given time can in principle be measured by sampling all of the contaminant present at that time in the compartment. A strategy often used in monitoring is to estimate the entire compartment mass by sampling a small proportion of the whole.26" Air Methods for measurement of concentration of hazardous substances in the air compartment are well established relative to estimation of inhalation risk. For aerosols specifications are available for sampling fractions of biologically relevant inhalable size,'4 but methods for measuring the concentration of non-inhalable pr a- r-t.ic-Jle_-s., _which :*:-.- are L relevant for estimating sur- --.--I-.-- a - . . . l - - - - 1 Skin contaminant layer The skin rinse methodz5has been used to assess the mass in the skin contaminant layer compartment. The method typically recovers 40%-90% of contaminants spiked onto the skin. Recoveries can be measured experimen- tally. However, spiking experiments have inher- ent problems as it is unclear whether this result should be interpreted as a 40%-90% sampling efficiency for the skin contaminant layer or if it reflects partitioning between this compartment and the stratum corneum or perfused tissue. Measurement methods based on UV fluores- cence of tracer compounds mixed with the con- taminant at source'^" indirectly measure the mass of the contaminant in the skin contami- nant layer. However, fluorescenttracers have the a+._b_ility--t-o--b.i.n-d ..with the cell proteins ---Irl. _. --*L- -1 :* _ * in the stra- " L L f ^ I:+- ferentiate between the skin contaminant layer and the stratum corneum. MASS TRANSPORT PROCESSES Transport can in principle be measured with a mass balance calculation for the change in mass in all relevant compartments per event or per time. Transport can more directly be determined by methods which standardise a set of external factors that are considered to be important determinants of the transport process (table 3). Deposition Adhesive tape can measure the particle deposition unaffected by loss. This has been done for particle deposition onto the face," but the method does not work for parts of the body that get into contact with other surfaces. Charcoal cloth has been used to measure skin deposition of volatile compounds. A problem is that the method does not differentiate between vapours from air and liquid ~ p l a s h e s3.9~ ~ Resuspension or evaporation Resuspension is usually best described as a series of single events. If events are repeated, say at frequency y, and the ratio y / T tends to infinity, where T is a suitable reference time interval, resuspension can be treated as a con- tinuous process. In this case it is customary to define a resuspension rate R by the equation where the mass of substance lost from the sur- face per unit time &) (table 2) is: L(gs-') = R(s-').M,,(g) (4) This definition assumes that R is independ- ent of hQk(see also discussion for transfer). An approachto assess resuspension based on events would be to collect all dust resuspended as the result of a standardised mechanical impact. Kildese er ap" have developed the sur- face total emission potential of particles (STEPP) tester, which simulates resuspension caused by a walking person from a carpet. Resuspension caused by air movements is simulated in the SMAIR tester." In this instru- ment a well defined air jet is directed at a sur- face contaminant layer and the resuspended particles collected on a filter. Vacuuming tech- niques could be used to simulate resuspension caused by strong air velocities, provided the nozzle does not touch the surface. Evaporation is driven by diffusion and thus it is simpler to speafy external factors. Wolkoflo has developed a field and laboratory emission cell (FLEC),which in principle is a small air compartment that is placed on a surface. A well defined flow of conditioned air is passed through the compartment and the exhaust air is analysed for evaporated substances. The diffusive trans- port of water vapour from the skin (transepider- mal water loss) can be measured simply with instrument^.^' A similar principle could be used for other vapours lost from the skin. Transfer The increase in mass over time in the skin contaminant layer is the measured net balance between transport to and h m the compartment and thus is affected by deposition, resuspension or evaporation, removal, and uptake. Surrogate skin methods such as pads, gloves, and coveralls are based on the assumption that they are estimates of the transport of mass to the skin contaminant layer from surfaces. However, the materials used do not have the roughness, stickiness, and other properties of human skin and so may not meet this assumption. Dislodgeable foliar residue, a procedure first described by Iwata et a1F determines the pesticide remaining on foliage after spraying. It involves destructive sampling of foliage by punching or removing leaves to provide a sample leaf area of about 100 cmz and subsequent, partly mechanical partly chemical, removal of pesticide residue by shaking the leaves in water with some drops of surfactant. This process is intended to mimic transfer to skin. Then pesticide is extracted from the water and analysed chemically. Sometimes an organic solvent is used in the first step, but this then measures the mass in the compartment rather than the dislodgeable foliar residue. Both methods, however, have been used to indicate the strength to transfer by direct contact from the source, which may be relevant to predict exposure by contact. A more general form of dislodgeable foliar residue is transferable residue (TR). If T R is measured simultaneouslywith an increase per event of mass per area h&/kkin the skin contamination layer, a dermal transfer coefficient (DTC) can be defined as` 40: DTC (cm'. event-') = M,,(g.event-I)/ &,TR(g.cmW2) (5) For this equation to be useful it has to be assumed that TR is independent of h4JAsw However, Brouwer et ul " have shown that T R depends non-linearly on the mass per area in the surface contaminant layer, &JbuA. better approximation would be to assume a functional relation: TR=TR(MJA,J (6) determined by non-linear regression. In reality a stochastic relation must be expected, and thus for given &J&, ,TR has a given dis- tribution with variables (mean and variance) being functions of %,,ItIis&to b.e.expected that liquid contaminants can be assessed with the simple relation, grease or other paste-like contaminants an intermediate relation, and particulate contaminants with the most com- plex relation. Transfer of particles will depend on contaminant properties-such as particle size, distribution, shape, and humidity or cohe- sion. In all cases it must be remembered that TR will depend on the method used for its measurement. There are several versions of samplers designed to exert a constant force on a defined area in a wiping action. These methods have potential for simulating the transfer of mass and are non-destructive.One such monitor is a tixed pressure dislodgable residue sampler- such as a polyurethane foam ROSS e t al 42 used standardised movements (aerobic dance routines following an instructor and music) to measure transfer of pesticides from fogged carpets onto the skin. Transfer from the inner clothing contaminant layer to the skin contaminant layer is 772 usually estimated with patches or underwear. No other method seems to have been developed specifically for measuring this transfer. Removal Dissipation of contaminants from the clothing compartment has been the subject of many studies, however, the studies were focused on the transport of mass from the clothing compartment to the skin compartment-that is, the opposite direction to removal. Yang and Li43reported a frictional transport of pesticides of 1%-5% from contaminated clothing to underwear. As mechanical or frictional transport may be an important mechanism of removal, these data indicate the range of mass transport from clothing in the opposite direction. Removal from skin to surface has been studied by Brouwer et aL4' They reported a mean removal efficiencyof 38%during a single pressure contact of a contaminated hand with an uncontaminated flat surface. Redistributwn Information on the area exposed and the redistribution of contaminant could be obtained by observation. However, the method of choice is the fluorescent tracer technique with a quantitative image analysis system.'*'' These methods have high spatial resolution. Decontamination Decontamination of the air compartment is the removal of mass from the air compartment by either local or general ventilation. Methods to assess the efficiency of ventilation in reducing the concentration of gaseous and particulate contaminants are well developed.44 Decontamination of surfaces can be assessed by measuring the mass removed-for example, the quantity of dust in a vacuum cleaner filter bag-or by the decrease in mass of the surface contaminant layer. A measurement strategy for decontamination of surface dust in offices has been described by Schneider et aP' and a standard procedure to assess the efficiency of handwashing has been developed by Fenske and Lu." Penetrakn and permeation Penetration and permeation stand for transport of mass through either the clothing barrier or the stratum corneum. These mass transport pathways have been studied in great detail relative to the effectiveness of protective and percutaneous ~enetration.I~n' both cases a test cell design is used to measure permeation. However, several different methods are used in the assessment of percutaneuos penetration ranging from in vitro models with static diffusion cells and flow through diffusion cells with animal or human skin (viable and non-viable). General discussion and conclusions In this paper we propose a conceptual model of both the important pathways leading to dermal exposure and the intermediate compartments where the hazardous substances may reside. This model has allowed us to define consistent terminology, which should form the basis for improved comparability between studies of dermal exposure or surface contamination. Consistent use of the model will ensure that most appropriate variables are measured in any situation. The model has been constructed at a conceptual level and omits much of the detail, which is evident in real situations. It is possible to extend the model as a series of interlinked compartments for the skin, surfaces, etc. In this way compartments for hands, arms, torso, etc could represent the skin surface; or if greater resolution is required, fingers, palms, etc could represent the skin surface. In this way the existing model encompasses sufficient detail to enable terminology to be developed, and it could be generalised for any particular need. The model can cope with special exposure scenarios-such as immersion of body parts in liquid or powder. In this case the liquid or powder constitutes the skin contaminant layer compartment and has infinite volume. For interpretation of contaminated substances derived Erom soil and paste-like substances in the skin contaminant layer a refinement of the model is necessary. The skin contaminant layer could be subdivided into an outer layer and an inner layer in intimate contact with the stratum corneum. The possibility exists that it is the supply of hazardous substance from the outer layer to the inner layer, which limits the rate of uptake. This identifies a need to refine existing methods of measuring mass in the skin contamination layer compartment. We have not considered the stratum corneum in much detail, as it is generally not available for surface sampling methods. Skin smpping with adhesive tape is a possible exception. Consecutive smpping allows semiquantitative depth profiling, but if the skin contaminant layer is not empty, the mass in this compartment and in the stratum corneum will be partially mixed up. The fmite thickness of the stratum corneum constitutes a buffer capacity and introduces a lag time for the transport process to reach equilibrium after a step change in c~ncentration.~T' he lag time may vary from a few minutes up to days and its significance for risk assessment would be determined by duration of exposure compared with lag time. The inventory of existing methods for measurement of skin and surface contamination (table 3) is only illustrative. Compilation of an exhaustive inventory and discussion relative to the model is beyond the scope of the present paper. However, with the model structure, limited or non-existing methods for measuring relevant compartment mass or transport processes have been identified: 0 The model stresses the importance of measuring concentration in the skin contaminant layer but methods relevant for uptake are lacking 0 The mass of particles in the skin contaminant layer may only have limited relevance for uptake. Particles are less likely to result in uptake than liquids and solubility of particles is an important qualifier which should be i r: i i i specified along with results of measurements of the mass and composition of the skin contami- nant layer 0 A clear distinction must be made between mass in a compartment and transport of mass. Direct measurement of mass transport must be based on an appropriate theoretical model. We envision that our paper will stimulate discussions and help in the development of more appropriate methods for the assessment and control of dermal exposure. This work was facilitated by the Dermal Exposure Network, supported by European Commission Contract SMT4-CT967502 (DGl2-RSMT). 1 Fenske RA. Dermal exposure assessment techmques. Ann O m p Hyg 1993;37:687-706. 2 Fenske RA, van Hemmen J. Occupational skin exposure to chemical substances: setting limits. Ann O m p Hyg 1994;38:3336. 3 Bos PMJ, Brouwer DH, Stevenson H, et al. Proposal for the assessment of quantitativedermal exposure limits in occu- pational environments: part 1. Development of a concept to derive a quantitative dermal exposure limit. O m p Ern.- nm Med 1998;55:795-804. 4 Cherrie JWR,obemon A. Biologically relevant assessment of dermal exposure.Ann Occup Hyg 1995;39387-92. 5 Sartorelli P, Aprea C, Cenni A, LT al. Prediction of percuta- neous absorption from physicochemical data: a model based on data of in vim exwriments. Ann ocnri, HVP I 1998;42:267-77. 6 McKone TE, Howd RA. Estimatingdermal uptake of non- ionic 0-c chemicals from water and soil: I. Unified fugacity-%ased models for risk assessments. Risk -Anal 1992;12:543-57. 7 Hall D, Reed J. The time dependenceof the resuspension of particles. Ann Occup Hyz 1989;20839-42. 8 Fedele PD. Model of aerosol protection offered by permeable protective garments. In: MacBriaty JP,Henry h'W, eds. Perjbnnance of pmrecriw dothing. Vol 4. Philadelphia: American Society for Testing Materials, 1992. (ASTM STP 1133.) 9 Crouse JL,DeJonge JO, Calogero F. Pesticide barrier performance of selected non-woven fabrics in laboratory capillary and press penetration testing. Tex& Research Journal 1990;60:137-42. 10 Roed J, Anderssnn KG, Bell KF, LT al. Quandratiw m e a r e - ment of m s o l deposition 011 skin, hair nd clothingfor dosirnet- ric usses- Roskild, Denmark Rise National Labora- tory, 1998. (RiSeR-l028(EN).) 11 Carson B, Wells F, Stola M, et aL Dnmal expasure assessment. A litemnrre review. Las Vegas, Nv: Environmen- tal MonitoringSystems Laboratory, Office of Research and Development, United States Environmental Protection Agency, 1992. (Repon:EPA 6M)K-92/002.) 12 Durham WF,Wolfe HR.Measurement of the exposure of workers to pesticides. Bull W Health Orpan 1962;26:75- 91. 13 Roff MW. A novel lighming system for the measurement of dermal exposure using a fluorescent dye and a image proc- essor.Ann Oenrp Hyg 1994;38:903-19. 14 Roff MW. Accuracy and reproducibility of calibrations on the skin using the FIVES Auorescencemonitor. Ann O m p Hyg 1997;41:313-24. 15 Fenske RA, Bimbaum SG. Second generation video imaging techniquefor assessing dermal exposure (VITAE- system).Am Ind Hyg Asroc3 1997;5863&46. 16 Archibald BA, Solomon KR, Stephenson GR A new proce- dure for calibrating the video imagingtechnique for assess- ing dermal exposure to pesticides. Arch Em'nm Caram T b l 1994;26:398-402. 17 Bierman EPB, Brouwer DH, van Hemmen JJ. Implementa- tion and evaluation of the fluorescent tracer technique in greenhouse exposure studies. Ann Occup Hyg 1998;42: 467-75. 18 Dost AA. Monitoring inorganic airborne and surface contamination by a portable XRF spectrometer. Epidenrwl- 001995;6:547. 19 Occupational Safety and Health Administration. Samp4mg for surface urnramination. Indurrial hygiine technical manual chapter 2. Washington, DC: OSHA, 1990. 20 Wheeler JP, Stancliffe JD.Comparison of methods for monitoring solid particulate surface contamination in the workplace. Ann Occup Hyg 1998;42:477-88. 2 1 Schneider T,Kildess J, Petersen OH, er al. Design and cali- bration of a simple instrument for measuring dust on sur- faces in the indoor environment. IndoorAir 1996;6:204-10. 22 Ness SA. Surface and dermal moniwring for r0xic exposum. Chapter 7. New York Van Nosuand Reinhold, 1994: 191-4. 23 lwata Y,Knaak JB, Spear RC, et al. Worker re-entry into pesticide- treated crops: I Procedure for the determination of dislodgeable residues on foliage. Bull Enwinm Conram Tmi2011977;10649-55. 24 Rougier A, Lone C, Dupuis D. An original predictive method for in vivo percutaneous absorption studies.3Soc Cosmet Chem 1987;38:397-417. 25 Environmental Protection Agency. Pesticide assessment guide- lines. Subdierirwn U.Applicarm exposure moniwring. Washington, DC: US EPA, 1986. 26 Organisation for Economic Co-operation and Develop ment. Guidance documentfor rhe conduct of studies of occupa&mal exposure w pes& during agricultural application. Paris: OECD, 1997. (Series on Testing and Assessment No 9.) 27 Royster GW, Fish BR. Techniques for assessing removable surface contamination.In: Fish BR, ed. Surface conramination,proceedings of a rymporium held at Gdingsburg, Tennessee. New York Pergamon Press, 1967,201-7. 28 Kildess J, Vinzents P, Kloch NP, er al. A simple method for measuring the potential resuspension of dust from carpets in the indoor environment. 72xtile Research 3ournal 1999;69:169-75. 29 Farfel M R , Lees PSJ, Rohde CA, er al. Comparison of a wipe and a vacuum collection method for the determination of lead in residential dusts. Environ Res 1994;65:291301. 30 Wolkoff P.An emission cell for measurement of volatile organic compounds emitted from building materials for indoor use: the field and laboratory emission cell FLEC. Gefahrswfle-Reinhalcungder LufZ 1996,56:151-7. 31 Morrison BM, Scala DD. Comparison of instrumental measurements of skin hydration.Jburnal of T d g y Cummnu and oculm T & b 1996;15:305-14. 32 F d e y BL, Scott PK, MayhaU DA. Development of a standard soil-to-skin adherence probabiliry density function fur use in Monte Carlo analyses of dermal exposure. Risk Anal 1994;14555-69. 33 SchneiderT, StokholmJ. Accumulation of fibres in the eyes of workers handling man-made mineral fibres. S c a d 7 llrbTk Envimn Health 1981;7:271-6. 34 European Committee for Standardisation. llrbTk@ atmospheres: rize fmcda defininbnr f m measurements of airborne @nides. Brussels: European Committee for Standardization,l993. (CEN. EN 481.) 35 Vincent J. Amsol sampling science and practice. Chichester: John Wiley, 1989. 36 Anzai I, Kikuchi T. A new monitoring technique of surface contamination-the test surface method. H& Pkv& 1978;34:27 1-3. 37 Gudmundsson A, Schneider T, Bohgard M , et d.Deposition of airborne particles onto the human eye: wind tunnel studies of rhe deposition velocity onto the eyes of a manne- quin. 3 Amsol Sn' 1997381085-100. 38 Cohen BM, Popendod W. A method for monitoring dermal --ucosure IO volatile chemicals. Am Ind HVP A s m 7 1989;50214-23. 39 Perkins JL,Vescial K. An evaluation of charcoal cloth as a potential field monitor for the efficacv of chemical 0rote~- hve clothing.Appl Occup Envimn Hyg1997;12:3626. 40 Zweig G, Leflinwell JT, Popendorf WJ. The relationship between dermal pesticide exposure by fruit harvesters and dislogeable foliar residues. Em'ron Sa.Health 1985;202759. 41 Brouwer DH, Kroese R, van Hemmen JJ. Transfer of contaminants from surface to hands: experimental assessment of linearity of the exposure process, adherence to skin, and area exposed during frxed pressure and repeated contact with surfaces contaminated with a powder. Appl Occup Em'ron Hyg 1999;14231-9. 42 Ross J, Thongsinthusak T, Fong HR, er al. Measpotential dermal uansfm of surface pesticide residue generated from indoor fokger use:an interim report. Chonosphere 19982034940. 43 Yang Y,Li S.Frictional uansition of pesticides h m protective clothing.Arch Envinm Contam Tmicd 1993;25:279-84. 44 Breum NO, Brhede E. Dilution versus displacement ventilation: environmental conditions in a garment sewing plant. Am b d HygAssocJ 1994;55:140-8. 45 Fenske RA, Lu C. Determination of handwash removal efficiency: incomplete removal of the pesticide chlorpyrifos from skin by standard handwash techniques. A m Ind Hyg Asroc3 1994;55:425-32. 46 Anna DH, Zellers ET, Sulewski R. ASTM F739 method for testing the permeation resistance of protective clothing materials: critical analysis with proposed changes in procedure and test-cell design. A m Ind HygAsrocJ 1998;59;54765. 47 Roper CS, Howes D, Blarn PG, er al. Percutaneous penewnon of 2-phenoxyethanolthrough rat and human skin.Food Chem Toxuol 1997;35;1009-16. 774 Occup Envimn Med 1999;56:774-780 Analysis of incidence of childhood cancer in the West Midlands of the United Kingdom in relation to proximity to main roads and petrol stations ROY M Harrison, Pei-Ling Leung, Lillian Somervaille, Ralph Smith, Estelle Gilman Department of Environmental Health R M Hamson P-L Leung West Midlands Cancer Intelligence Unit L Somervaille R Smith Department of Public Health and Epidemiology, Institute of Public and Environmental Health, University of Birmingham, Edgbaston, Birmingham B15 2 T I ; UK E Gilman Correspondence to: Professor R M Harrison, Department of Environmental Health, Instirute of Public and Environmental Health, University of Birmingham, Edgbaston, Birmingham B15 2 l T , UK Accepted 23 June 1999 Abstract Objectiwes-To investigate whether there is an excess of leukaemias in 0-15 year old children among those living in close proximity (within 100 m) of a main road or petrol station. Methods-Data for 0-15 year old children diagnosed between 1990 and 1994 in the United Kingdom West Midlands were used. Postcode addresses were used to locate the point of residence which was compared with proximity to main roads and petrol stations separately, and to both together. Odds ratios (ORs) were calculated with solid tumours as a control, and incidence ratios (I&wi)th population density as a control. Results-The method based on solid tum o m as a control showed ORs of 1.61 (95% confidence interval (95% CI) 0.90 to 2.87) and 1.99 (95% CI 0.73 to 5.43), for those living within 100m ofa mainroad or petrol station respectively. When population was used as a control, the estimated for leukaemia were 1.16 (95% CI 0.74 to 1.72) and 1.48 (95% CI 0.65 to 2.93) for residence within 100 m of a main road or petrol station respectively, but neither reached significance at the 95% level. Results for residence in close proximity to both a main road and petrol station were inconsistent, but there were few. The influence of socioeconomic factors as represented by the Townsend deprivation index on leukaemia incidence was not significant and the results were not explicable on the basis of impact of social class. Conclusions-The results are suggestive of a small increase in risk of childhood leukaemia, but not solid tumours, for those living in close proximity to a main road or petrol station. This increase in risk is not, however, significant and a larger study is warranted to establish the true risk and causes of any increase in risk. (Occup Envirm Med 1999;56:774-780) Keywords: leukaemia; children; road traffic; petrol station It is now accepted that occupational exposure to high concentrations of benzene can lead to an excess of non-lymphocytic leukaemias.' For example, workers exposed to in excess of 400000 ppb-years benzene in the Goodyear Pliofilm plant showed a standardised mortality ratio (SMR) of 6.64 (95% confidence interval (95% CI) 1.33 to 19.39).*Despite the fact that most leukaemias in children are of the acute lymphocytic variety, a type not associated with occupationalbenzene exposure, there have been suggestions that environmental exposure to benzene vapour may be a cause of childhood leukaemia. Wolf? showed significant correla- tions between acute myeloid leukaemia, acute lymphoblastic leukaemia, chronic lymphoid leukaemia, low grade non-Hodgkin's lymphoma, and all lymphoproliferative disease and the number of cars per household in each United Kingdom county.Wolfp postulated that benzene exposure within the car might be responsible for these correlations. In a more complex study, Alexander et aF estimatedrelative risks (RRs)for childhood acute lymphoblastic leukaemia by Poisson regression for groups of electoral wards classified by isolation and car ownership. Multivariate analyses include adjustment for s0Ci0economic status. No evidence of increased risk of childhood acute lymphoblastic leukaemiawas found in areas where more households own cars. It was concluded that the previous explanation for the small area variation of childhood acute lymphoblastic leukaemia in terms of geographical isolation and interpretation in terms of exposure to common mfections continued to be justified. Other geographical analyses of risk of childhood cancer have looked for evidence of involvement of an infectious process5 or for an association of high risk with proximity to either nuclear or indusmal installation^.'-^ Apart from the well known excess of childhood leukaemias around the Sellafield nuclear reprocessing plant, none of the nuclear sites examined in England and Wales showed evidence of an increased risk of childhood leukaemias and non-Hodgkin's lymphomas in the area covered by a 25 krn circle around each plant.8Knox and Gilman' found an association of increased risk of childhood leukaemias and solid cancers with proximity to a range of indusmal sites. Relative excesses were consistently associated with proximity to sites producing two main types of indusmal atmospheric effluent: ( a ) volatiles derived from petroleum, and (b) kiln and furnace smoke and gases, and effluents from internal combustion engines. However, no evidence was found of an association with proximity to benzene works. Much is known about personal exposure to benzene. In the United States the total exposure assessment methodology (TEAM) study" '' measured the sources of personal I exposure, showing a wide range of sources those of uncertain behaviour, and for solid including active and passive smoking, personal cancers and benign neoplasms. In the United activities-such as use of paints, solvents, other Kmgdom all cancer registries, together with consumer products, and refuelling the car, as the Office for National Statistics use the ICD well as breathing outdoor air and air in the car. for both childhood and adult tumours to be Personal exposures generally exceeded, but consistent across databases rather than the were correlated with concentrations in outdoor international CCC 1996 coding classification. air. In a study in the United Kingdom, h u n g and Harrison`* carried out both direct GEOREFERENCING DATA USING THE UNIT measurement and modelling of personal expo- POSTCODE AND CENSUS GEOGRAPHY sure to benzene and other monoaromatic The areas of enumeration districts (EDs), hydrocarbons. The results showed that al- wards and health authorities are used in the though some microenvironments-such as study for aggregation of data. The EDs are the adjacent to petrol pumps when refuelling the smallest areas that can be used to use car-showed highly increased concentrations, population census data. Wards are constructed integrated personal exposure was influenced from EDs and in the West Midlands have an far more by microenvironments in which the average population of 15 600. Health authori- subjects spent large periods, the home being ties are constructed from wards and form the the most important, with the workplace also administrative areas by which health services being significant for those who worked outside are organised in the United Kingdom. To inte- the home. Measurements of the spatial distri- grate data into a geographic information bution of benzene in the atmosphere have system, both the petrol stations and the cancer shown that the highest outdoor concentrations cases and controls were allocated a grid within urban areas tend to occur adjacent to reference using their unit postcode. This was main roadsJX3or in the vicinity of petrol accomplished with a table called the Central station~.T'~he influence of such ground level Postcode Directory (CPD). This consists of all sources usually extends for about 100 m before the postcodes in the West Midlands and their concentrations fall to the local ba~kgr0und.I~grid references. The postcodes from the cancer This result derives from severalmonitoring and registration database are matched to identical modelling studies of traffic generated pollut- entries in the CPD. A unit postcode (the full ants in open country. The dispersion of pollut- seven or eight digit version-for example, B15 ants in urban terrain is more complex, but none 2lT) represents about 13 household children the less the same general guidelines can be in the West Midlands. Grid references were expected to apply. Concentrations within allocated to unit postcodes in the following houses are strongly influenced by those out of way. A 100 mzgrid is used to identify in which doors, and hence the occupants of homes grid square the first household of a unit i located within 100 m of a main highway or postcode is located. This unit postcode is then petrol station would be expected to experience assigned the grid reference of the south west higher exposures to benzene and related corner of the grid square. This means that all hydrocarbons than occupants of comparable postcodes whose first house is situated in a homes at greater distances from such sources. particular grid square are allocated the same As the epidemiological studies already re- grid reference. If a postcode happens to be ferred to have not given a wholly consistent towards the north west of the grid square then answer to the proposition that benzene expo- the reference will be inaccurate. This inaccu- sure may influence incidence of childhood leukaemia, and some have been criticised on methodological grounds, it was thought appropriate to conduct a further analysis of incidence of childhood cancer relative to sources of racy was offset by adding 50 m to the easting and northing of the grid reference as recommended by Gatrell et at6and used in a study of cancer around a TV transmitter mast by Dolk et d." hydrocarbons. In our study, residence within Full postal addresses of petrol stations 100 m of a main highway or petrol station was operational in 1997 within the West Midlands taken as a surrogate for increased exposure to were purchased from Thames Communica- benzene and other hydrocarbons present in tions (Gravesend, Kent). The local authority evaporative losses of petrol, or to exhaust emis- planning departments provided information on sions from road traffic. the closure and construction of petrol stations during the period 1990-7, which allowed Methods development of a database containing ad- CASES OF LEUKAEMIA AND SOLID CANCER dresses of stations fully operational during the Data from the West Midlands Cancer Intelli- years 1990-4 inclusive. Each petrol station was gence Unit for 0-15 year old children diag- then allocated a grid reference with Matchcode nosed between 1990 and 1994 were used. (a software package developed b y Capscan, Cases were taken from the Birmingham, Walsall, Solihull, Dudley, Wolverhampton, and Sandwell District Health Authorities (total population 2 251 000) as full data on petrol stations were available only for these areas. Classifications from the ninth revision of the international classification of diseases (ICD-9) were included for lymphoid, myeloid, monocytic, and unspecified leukaemia, including London). Matchcode can improve address data by adding a postcode if one is misskg or improving the address if it is slightly inaccurate. At the same time it can add a grid d m c e to an address. This is done by the sofiwpre referring to a CD ROAM h a t contains d the addresses in the United b g d o m . ?hir, is called the Post Office address file (PAF) and is what the CPD is derived from. After the p t - I 776 Harrison, Dung,Somervaille, et al code was allocated to the stations with Matchcode, the 50 m adjustment was made. The CPD is provided free to NHS organisations and Universities and is often used in geographical studies as a method of grid referencing postcodes. Resources are available that allow users to grid reference a postcode to an claimed accuracy of 0.1 m (ordnance survey address point product), but the expense of this product prohibits its use in the NHS and academia. ROADS OF HIGH TRAFFIC DENSITY Traffic flow data for the West Midlands was obtained from the Joint Data Team, Birmingham. This exists as counts of vehicles along links of road throughout the West Midlands. The data pertaining to the roads used in this study were extracted and average traffic flows calculated. The Ordnance Survey dataset STRATEGI was used as a source for the digital road network. The following road categories were chosen as being of high traffic density: motorways, primary routes (dual and single carriageways), and single and dual carriageway A roads. In all, this selected 755 sections of road with mean traffic flow 23 400 vehicles a day (10% percentile 11 500; 90% percentile 38 800) as derived .from the Joint Data Team statistics for the individual road links. DATA ANALYSIS Spatial analysis was performed with a geographic information system (ArcView V2.1). Other manipulation of data was carried out with the software package Microsoft Excel. To investigate prevalence of childhood leukaemia relative to locality of a petrol station and proximity of main roads, the geographic information system was used to perform two different kinds of spatial analysis with different control groups. In the first kind of analysis, the spatial distribution of leukaemias was analysed with solid tumour cases as a control. Leukaemia and control cases were plotted (with grid references Table 1 Population (0-15years old) within each disnict health authoricy and esrimated to live within 100 m of source Total population in age -g= Population living <lo0 m from petrol station Population living ~ 1 0 m0 from major road Population living <IO0 m fiom petrol station and road Birmingham Dudley Sandwell Solihull Walsd Wolwhampton 217583 8760 33226 59297 59665 2118 2739 9425 10137 40263 54610 51170 760 3102 2221 3792 8958 8561 5955 1332 1994 300 1900 1575 Table 2 Numbers of leukaemia and solid cancers relative to bcatwn of residence, and Calculated oddc ratios (OR) Location of residence Leukaemias Major road c 100 m Major road z 100 m Petrol station c 100 m Petrol station > 100 m Major road and petrol station e 100 m Major road and petrol station > 100 m Total casedcontrols 24 106 8 122 3 127 130 Solid cancers (controls) 31 220 8 243 1 250 251 OR 1.61 1.99 5.91 95%CI 0.90to 2.87 0.73 to 5.43 0.61 to 57.3 obtained from addresses at the time of diagnosis) and a spatial query was then performed to determine the number of cases and controls within a 100 m radius of a petrol station or a zone 100 m from a main road. In the second kind of analysis, leukaemia cases, petrol stations, and main roads were again entered into the GIs. The number of new leukaemia cases within 100 m of a source was again evaluated, but hrther separated according to the District Health Authority area. To include the population data as a control, the method of Dolk et ul" was adopted. To calculate the expected number of leukaemia cases within 100 m of a possible source, a population for such an area has to be calculated. Actual population counts do not exist for small user defined areas so they have to be estimated. The Postcode to Enumeration District Directory from the Office of National Statistics was used which contains the number of household children in each unit postcode. With census information for the ED that the postcode falls in, an average number of people per household can be calculated. From this we calculate the number of people per postcode. Finally with ED census information the number of 0-15 year old children can be estimated. Townsend deprivation indices were available for electoral wards from the NHS Executive West Midlands Regional Office. To evaluate any influence of socioeconomic factors, the leukaemia incidence was plotted against the Townsend score for each ward. The Townsend index'* is a measure of multiple deprivation calculated from census variables. High scores reflect severe deprivation. Some other researchers in the United Kingdom have used the Carstairs index derived from Scottish data, whereas the Townsend index is derived from English data. Both indices have four census variables, three of which are common to both, and we would not anticipate any main differences from the use of the Carstairs index. STATISTICAL. METHODS Two different approaches were used to investi- gate whether an excess incidence of leukaemia was present around main roads or petrol stations. In the first method odds ratios (ORs) were calculated with a case-control method. Each subject was classified as positive or negative according to their proximity to a main road or petrol station and defined as a case if they had been diagnosed with leukaemia or a control if the diagnosis was of a solid tumour. The OR was then calculated according to equation (1) where A were patients with leukaemia living within 100 m of a road or petrol station; B were patients with solid tumours living within 100 m of a road or petrol station; C were patients with leukaemia living away from a road or petrol station; and D were patients with solid tumours living away from a road or petrol station. OR=AD/BC (1) The 95% confidence interval (95% CI) for the OR was calculated from Table 3 Incidence ratio(IR (95% GI)) of childhood leukaemia if living < I 00 m from source Petrol station Obs Exp IR 9S%CI Roads of high naBc &nri@. Obs Exp IR 9S%CI Birmingham Dudley Sandwell Solihull Walsall Wolverhampton All disuicts 4 1 1 0 1 1 8 2.6 1.54 0.42 to3.92 14 0.6 1.67 0 to 9.33 4 0.6 1.67 0 to9.33 0.2 - - 3 0 0.9 1.11 Oto6.22 1 0.5 2.00 0 to 11.2 2 5.4 1.48 0.65 to 2.93 24 9.8 1.43 0.79 to 2.40 2.6 1.54 0.42 to 3.92 2.2 1.36 0.27 to4.00 0.8 - - 2.6 0.38 0 to2.23 2.7 0.74 0.07 to 2.67 20.7 1.16 0.74 to 1.72 Peml scation and mads Obs EXP IR 1 1.8 0.56 1 0.4 2.50 1 0.4 2.50 0 0.1 - 0 0.6 - 0 0.4 - 3 3.7 0.81 95% CI 0 to 3.22 0 to 14.5 0 to 14.5 - - 0.16 to 2.38 95% CI=exp[ln RRk1.96(1/A+l/B+l/C+l/ D)1 (2) An OR of 1.O would indicate that there is no effect on incidence of childhood leukaemia of living in close proximity to main roads or petrol stations. The second method used indirect age and sex standardisation to examine the observed number of cases of leukaemia around main roads and petrol stations compared with what would be expected for the population of the health authority as a whole. An incidence ratio (IFQ of 1.0 would indicate that there is no dif- ference between the observed number of cases in the exposed group and that which would be expected from the comparison population. We have chosen not to use the term standardised IR as full age distributions were not known and therefore full standardisation has not been carried out. To estimate the IR, the number of leukaemias in the 0-15 age group expected within a region within 100 m for a petrol station or main road was first calculated from equation ( 3 ) where E,,,,=expected number of cases, C,=total number of cases in District Health Authority A (0-15 year age group), PA= total population of 0-1 5 year old children in District Health Authority A, P,,,,=population of 0-1 5 year old children residing within 100 m of petrol station or road. The IR was then calculated from where O,-,=observed number of leukaemia cases within 100 m of a petrol station or road. Due to the small number of cases, the 95% confidence interval on the observed number of cases was read directly from standard statistical tables applying to the Poisson distribution. These were used due to the non-integer nature of the expected values. Additionally, the Poissonprobability of from 0 to (r-1)cases of diseasewas calculated, where r is the number of observed cases, and h is the expected number. Probability (n cases)=e-'..h"/n! These probabilities were summed from n=O to n=r-1 and subtracted from 1.0 to give the probability of S r events occurring. Resdts The study was limited to 0-15 year old children for several reasons. Most important is the relatively short time interval between disease induction and diagnosis which makes spatial analysis more realistic than for adult cases. Secondly, important factors which can confound the analysis of adult cancer data are not present for children. Most important among these are smoking and occupational exposure to carcinogens. Childhood cases of cancer are also more accurately diagnosed and more completely registered than adult cancers." Table 1 shows basic data relating to the populations within the six district health authorities which were included in this study. The area is densely populated and as many as 17% of the population were classified as residing within 100 m of a main road within some health authority areas. Table 2 shows the number of cases and controls according to location of residence when solid tumours were used as controls. The table also shows the ORs and 95% CIS for leukaemia relative to solid cancer. While we were conducting this study, the work of Knox and Gilman' showed that both leukaemias and solid tumours had similar excesses relative to the sources examined in that work. It was therefore thought prudent to use some other form of analysis, and with the procedures outlined above, estimated populations of 0-15 year old children were used to calculate expected numbers of cases of leukaemia. The results of this study expressed as IRs appear in table 3 . This procedure was carried out also for solid tumours, showing for all districts combined IRvalues of 0.77,0.8, and 0.24 for residences close to petrol stations, main roads, and the two jointly, respectively. None of these was significantly different from 1.O at the 95% level. Discussion Both methods show an increased risk of childhood leukaemia for those living within either 100 m of a petrol station, or within 100 m of a main road. For the case-control method, standard 95% CISappear in table 2. For the IR method, 95% confidence limits on the observed value were taken from statistical tables for the Poisson distribution, and used to calculate a 95% CI on the IR. Probabilities were calculated for the combined health authorities for proximity to a main road @=0.28) and proximity to a petrol station (p=O.18). In neither instance, therefore, is the result significant at the 95% level although the data for all district health authorities together approach significance for the case-control method. This fmding is broadly consistent with a study of the incidence of leukaemia and lymphoma in young people (0-24 years) within 3 km of the petrochemical plant at 778 Harrison, b u n g , Somervaille, et al Baglan Bay, South Wales." By contrast, a study in Britain by Knox and Gilman' found excesses near to sites producing two main types of industrial atmospheric effluent (petro- leum derived volatiles; and kiln and furnace smoke and gases, and effluents from internal combustion engines), but a non-significant excess of childhood leukaemias and solid can- cers near to petrochemical works, and no association of increased risk with proximity to benzene installations. Reasons for the differ- ences in findings between studies may be related to problems obtaining accurate popu- lation denominators on which to base calcula- tions of expected numbers or in the definition of exposed populations. The study in South Wales took as its exposed population those resident within EDs, wholly or partially within concentric circles of 1.5 or 3 km radius of the centre of works. This included areas outside these zones, and took no account of the wind direction frequencies and pattern of dispersion of gaseous effluents from the petrochemical plant. Because of the long period covered by cases used in the Knox and Gilman study (1 953-80), census-based population estimates could not be used, and would not have been available at the fine spatial resolution required; instead a method based on postcodes was used.' This could have created a false apparent excess of cases near to industrial sources if postcodes close to those sources also system- atically had higher densities of children living in them than did postcodes which were distant from industrial sources. Aware of this prob- lem, the authors investigated the distribution of postcodes. Areas nearest to indusmal sources had a lower density of postcodes per kilometre than other locations, indicating that fewer people lived in these postcodes, and hence that fewer cases would be expected in postcodes near to industrial locations. This, together with the consistency in the nature of the sites around which excesses were found, makes it unlikely that their findings were due to an artefact. None the less, some of the con- clusions of Knox and Gilman are surprising. In particular, the finding of a relative excess of solid cancers and leukaemias within 4 km of a motorway, with a significant deficit beyond 4 km, is hard to reconcile with the fact that the impact of main roads on air quality is not measurable above the local background air pollution at distances beyond about 100-200 m.I5 Our study was not without its problems, the population data for the 0-15 year age group was inferred from information for EDs which are considerably larger than postcode districts (by a factor of about 16). It was because of the uncertainties in defining exactly the population numbers in the exposed and control groups that we thought the case- control method would provide valuable sup- porting information. Another limitation of our study is the lack of age and sex standardisa- tion. This is a problem for childhood cancers as previous studies have shown a slight increase of incidence for males relative to females (1.3:l male to female ratio) and for the under Although main geographic variations in the age and sex distribution of the 0-15 year old children within the West Midlands are unlikely, the lack of discrimination by age and sex may obscure aetiological insights which could be gained from a more refined analysis. There is little published evidence to suggest that pollutant emissions from road traffic or petrol stations are connected with solid tumours in the general population, excluding occupationally exposed groups. Occupational exposure to polycyclic aromatic hydrocarbons is linked with an excess risk of lung cancer. AS road vehicle emissions contain polycyclic aromatic hydrocarbons, adults residing alongside main roads may run a slightly increased risk of lung cancer from this source. This condition, however, represents only a tiny proportion of cancers in the 0-15 year age band for whom the brain and spinal cord are the main loci of solid tumours.21Exposure to gasoline vapour in adult workers has been associated with an increase in kidney cancer? but this is also a rare condition in children. We therefore think it likely that solid cancers represent a good control for this study and that the ORs presented in table 2 represent a genuine representation of increased risk. The 1% for solid tumours show no excess risk and support this view. The combined use of the two complementarymethods of data analysis in our view adds to the confidence with which the results can be viewed. A plot of incidence of childhood leukaemia in 1000 children by electoral ward versus Townsend deprivation index shows a poor correlation (t)=0.04; n=130), but there does seem to be a slight decrease in incidence of leukaemia with increased deprivation (higher Townsend index) as reported in previous ~tudies.'I~n the West Midlands Health Region the more deprived wards tend to be sited in the central conurbation, whereas the more aWuent wards tend to be in the rural shire counties. Thus, this finding is consistent with earlier findings of higher incidence of childhood leukaemia associated with lower population densities. It does not indicate any role for exposure to benzene from passive smoking in the aetiology of childhood leukaemia as smoking is more prevalent in homes of lower socioeconomic status (higher Townsend scores) and would produce the opposite slope to that found. One question rarely, if ever, considered in the spatial analysis of disease in relation to pollution sources is the magnitude of exposure. Thus, many studies have related incidence of disease to distance from source without regard to the spatial distribution of pollutant concentrations. In this work the use of a criterion of 100 m separation between source and point of residence (as indicated by postcode) is consistent with knowledge of the range over which air pollution from a ground level source sustains concentrations above the local background. It is, however, extremely difficult to estimate the increased magnitude of exposure for those liv- ing within the 100m band. The work of h u n g and Harrison" shows that exposure within the I , home is likely to be the main contributor to exposure to aromatic hydrocarbons, including benzene. The extent of exposure arising from the proximity of the road is very hard to estimate, but some indications can be gained from the work of Leung and Harrison." This work showed that roadside concentrations of aromatic hydrocarbons, including benzene, were highly variable depending on the sampling site. In particular, the openness of the site, and therefore the ease of pollutant dispersion, has a major influence on atmospheric concentrations.Taking the mean of 53 samples collected alongside six main roads in Birming- ham indicates a typical ratio of roadside to urban background concentration of benzene of roughly 5, and for toluene roughly 2.5. Allowing for the fact that indoor concentrations often slightly exceed those out of doors due to indoor sources, and the slightly greater distance of most homes from the traffic than the roadside samplers used in the work of Leung and Harri~on,i't~is likely that the exposure to benzene for those living within 100 m of a busy road is around twice that of those living at greater distance from the road. Long term average concentrations of benzene in the vicinity of petrol stations are broadly similar to those at the roadside, and therefore a similar magni- tude of increase might be expected. Petrol stations are unlikely to generate suffi- cient traffic on their own to cause a substantial increase in pollutants from vehicle exhaust. It might be expected that petrol stations would be located mostly on main roads, and the data in table 1 indicate that this is the case for >50% of the exposed population. The data for populations living within 100 m of a petrol station does, however, include a substantial additional population and justifies the separate treatment of this category of exposure. Petrol stations are most notable as a source of evaporated hydrocarbons from petrol. On the other hand, main roads are the source of a wider range of pollutants including nitrogen oxides and polycyclic aromatic hydrocarbons as well as volatile hydrocarbons such as benzene. If our finding of a higher RR associated with proximity to a petrol station than to a main road were borne out by a study of higher statistical power, it would be strongly suggestive of an influence of hydrocarbons upon childhood cancer. The results of this study are broadly consistent with those of other published work. Savitz and FeingoldZ4reported an OR of 1.7 (95% CI 1.O to 2.8) for cancer among children exposed to more than 500 vehicles a day. They reported ORs also for leukaemia of 2.1 (95% CI 1.1 to 4.0) and brain tumours of 1.7 (95% C1 0.8 to 3.9). Their results showed a dose-response gradient with traffic in excess of 10 000 vehiclesadaygivinganORof3.1 (95% CI 1.2 to 8.0) for total cancer and of 4.7 (95% CI 1.6 to 13.5) for leukaemia. Feychting et UP' used calculated nitrogen dioxide concentrations as an estimate of exposure to traffic pollution and found a relation between RR of cancer in 0-1 5 year old children relative to exposure to nitro- gen dioxide. With those exposed to a 99 percentile of hourly average concentrations over 1 year of nitrogen dioxide of >49 pg m" defined as an Rl7 of 1, those exposed to concentrations in the 50-79 pg m" range were subject to an RR of 1.9 (95% CI 0.8 to 4 . 3 , and for a concentration of 3 80 pg m" the RR was 3.8 (95% CI 1.2 to 12.1). Increased but imprecise risk estimates were found for leukaemia and central nervous system tumours. Background concentrations of nitrogen dioxide in the United Kingdom West Midlands conurbation lie within the highest band (99 percentile 380 pg.m") used in this study. These workersz5also used data published by Wertheimer and LeepeF to calculate the association between traffic density and childhood cancer from a study designed to investigate the association between exposure to electromagnetic fields and childhood cancer where traffic density was included for control of confounding. In this study exposure to heavy traffic was defined as homes within 40 m of road with a daily traffic count of C 5000. The effect of thus defined heavy traffic on total cancer mortality in children was calculated as an OR of 1.6 (95% CI 1.1 to 2.3). Mormation for specific cancer sites was not provided. The results of these studies both for total childhood cancer and for leukaemias are broadly consistent with the result of our own research strongly suggesting that although the results are just below the level of significance normally required, they are reflective of a true effect of road traffic on cancer in children. Conclusions The total population of the area studied is 2.25 million and the population of 0-15 year old children included in the data analysis is 482 588. The study included all diagnosed cases of leukaemia in the 0-15 year age group over a period of 5 years. The results suggest a slight increase of risk of leukaemia for those 0-1 5 year old children living in close proximity to a main road or petrol station, although in no case do the results reach significance. The two methods used based on use of solid tumours and population numbers as controls give broadly similar results except in the case of proximity to both main road and petrol station where the number of cases is very small and the results have very wide 95% CIS. Exposure of the population within 100 m of source is likely to be of the order of double that of other dwellers within the conurbation not living in such high proximity. In the case of petrol stations, increased exposures are primarily to hydrocarbons, including benzene, whereas in the case of main roads, pollutants include combustion products such as NO, and polycyclic aromatic hydrocarbons as well as volatile hydrocarbons such as benzene. The fact that increased risks are found at both kinds of site (although not significant) is suggestive of volatile hydrocarbons as a causal agent. Although the results of the study are suggestive of a possible slight association of exposure to pollutant (not necessarily benzepe) with increased incidence of leukaemia, the data are overall rather reassuring in showing that any such effect, if real, is likely to be small in magnitude. A larger 3 Harrison, L u n g , Somervadle, et al study is warranted to establish whether there is truly an increase in risk for those living in close proximity to roads and oetrol stations and to determine the causes. We are grateful to Tim Marshall for advice on statistical methods. 1 Expert Panel on Air Quality Standards (EPAQS). Benzene. London: The Stationery Office, 1994. 2 Rinsky RA, Smith AB, Hornung R, et al. Benzene and leukaemia: an epidemiologicrisk assessment.N End3 Med 1987;316:1044-50. 3 Wolff SP. Correlation between car ownership and leukaemia: is non-occupational exposure to benzene from petrol and motor vehicle exhaust a causative factor in leukaemia and lymphoma. Experionria 1992;48:301-4. 4 Alexander FE, Leon DA, Cartwright RA. Isolation, car ownership and small area variation in incidence of acute lymphoblastic leukaemia in children. Paediatric Perinarol Eptdemiol1996;10411-17. 5 Kinlen LJ. Epidemiologicalevidencefor an infectivebasis in childhood leukaemia. BrJ Cancer 1995;71:1-5. 6 Stiller CA, Boyle PJ. Effect of population mixing and socioeconomic status in England and Wales, 1979-85, on lymphoblasticleukaemia in children. BMJ 1996;313:1297300. 7 B e d V, Roman E, Bobrow M, eds. Childhood cancer and nuclear insralhrimt. London: BMJ PublishingGroup, 1993. 8 Bithell JF,DuttOn SJ,Draper GJ, et al. Dismbution of childhood leukaemias and non-Hodgkin's lymphomas near nuclear installations in England and Wales. BM3 1994;309: 501-5. 9 &ox EG, Gilman EA. Hazard proximities of childhood cancers in Great Britain from 1953-80. 3 Epidemwl CommunityHealth 1997;Sl:151-9. 10 Wallace LA. Major sources of benzene exposure. Envimn Health Perspat 1989;82:165-9. 11 Wallace LA. The exposure of the general population to ben- zene. CrUBiol T&I 1989;5:297-314. 12 h u n g P-L, Harrison RM.Evaluation of personal exposure to monoaromatic hydrocarbons. Occup Em.m Med 1998; SS:249-57. 13 Leung P-L, Harrison RM.Roadside and in-vehicle concennations of monoaromatic hydrocarbons. Atmos E m m n 1999;33:191-204. 14 Stanger Scienceand Environment.A pilor srudy 10 mess ben- zene~concenrrarions in the vuinity of pen01 srarias. Croydon, London: SSE, 1997. (SSVAQilO85.) 15 Quality of Urban Air Review Group (QUARG). First report: urban air qualiry in the United Kingdom. QUARG, January 1993. 16 Garrell A, Dunn C, Boyle P. The relative utility of the cen- tral postcode director and pinpoint address code in applicationsof geographical information systems. E m ' m menr and finning A 1991;23:1447-58. 17 Dolk H, Shaddick G, Walls P, et al. Cancer incidence near radio and TV transmitters in great Britain. Am 3 Eprdenriol 1997;145:10-1 7. 18 Townsend P, PhillimoreP, Beanie A. Health and inequaliw in the norrh. London: Routledge, 1989. 19 Parkin DM, Stiller CA, Draper GJ, er d. Inrernon'onal incidenceof childhoodcancer. Lyon: International Agency for Research on Cancer, 1998. (IARC Sci Pub1No 87.) 20 Lyons R4,Monaghan SP, Heaven M,er al.Incidence of leu- kaemia and lymphoma in young people in the vicinity of the penochemical plant at Baglan Bay, South Wales, 19741991. Ocmp Encimn Med 1995;52:225-8. 21 W t h s R,Cummins C, Kirk A, er al, eds. NHS Execunw, IEst M & d . Jotnt Rejxm of the War M&nds Dinctor of Rcblic Healrh and rhe West Midlands Regtnol C a m Regisny. Birmingham: West Midlands Regional Health Authority, 1995. 22 Dement JM, Hensley L, Gitelman A. Carcinogenicity of gasoline: a review of epidemiological evidence. Ann NY Acad Sci 1997;837:53-76. 23 Office of Population Censuses and Surveys. Gerald Draper, ed. The geogaphhical epidemiology of childhood leukaemia and m-hodgliin lymphomas in Great Brirain, 1966-83, studies on medical popularion subjects No 53. London: The Stationery Office, 1991;5-6:37-56. 24 Saviu D,Feingold L. Association of childhood cancer with residential traffic density. S c a d 3 E%d E m ' m Health 1989;15:360-3. 25 FeychtingM,Ahlbom A. Magnetic fields and cancer in cfiildren residing near Swedish high-voltage power lines.Am 3 Eebl1993;138:467-81. 26 Wertheimer N, Lecper E. Electrical wiring con6gurations and childhood cancer.AmJ Epidenriol 1979;109:273-84. I Rejected manuscripts From February 1994, authors whose submit- be returned to them. The Journal will destroy ted articles are rejected will be advised of the remaining copies of the article but corresdecision and one copy of the article, together pondence and reviewers' comments will be with any reviewer's comments, will kept. j , Risk of stomach cancer associated with 12 workplace hazards: analysis of death certificates from 24 states of the United States with the aid of job exposure matrices Pierluigi Cocco, Mary H Ward, Mustafa Dosemeci Abstract dust might act through non-specific Objective-To investigate the risk of gas- mechanisms, similar to those proposed tric cancer associated with 12 workplace for salt, aspirin, and heat by other au- exposures suspected or discussed as aetio- thors. logical agents in previous reports. ( O C C UEf~iviron Med 1999;56:781-787) Metho&-A case-control study was con- ducted based on the death certificates of Keywords: stomach; neoplasms; occupational exposure several million deaths in 24 states of the United States in 1984-96. Overall, the data base included 41 957 deaths e o m stomach In a review of occupational risk factors for cancer among subjects aged 325 years. stomach cancer we suggested that various These were 20 878 white men, 14 125 white occupational exposures may cause or contrib- women, 4215 Afirican American men, and ute to gastric carcinogenesis.' Ionising radia- 2739 Wean American women. Two con- tion and N-nitroso compounds, either directly trols for each case were selected &om or through the formation of free radicals or among subjects who died from non- nucleophilic intermediates, may damage the malignant diseases, Erequency matched to DNA of the cells of the gastric mucosa, acting cases by geographicregion, race, sex and 5 as initiators of the carcinogenic process. Other i year age group. Each three digit occupation and industry code listed in the 1980 United States census was classified for probability and intensity of exposure to asbestos, inorganic dust, metals, lead, polycyclic aromatic hydrocarbons (PAHs), nitrogen oxides, nitrosamines, sulphuric physical agents, such as asbestos and other inorganicdusts, could be irritants to the gastric mucosa and act as cocarcinogens in a way similar to the mechanism proposed for salt, aspirin, and heat." These agents cause a superficial gastritis and may increase cell acid, f d s e r s , herbicides, other pesti- proliferation, thus promoting initiated clones. cides (including insecticides and fungi- Dust could also act as a carrier delivering car- cides), and wood dust. These job exposure cinogens to the gastric mucosa. The effective- matrices were subsequently applied to the ness of this mechanism has been experimen- occupation-industry combinations in the tally proved in lung carcinogenesis.` death certificates of study subjects, sepa- A major cause of concern is that most stud- rately by sex and race. ies of gastric cancer and occupation rely on Resuh-Risk of stomach cancer showed a poor environmental data. Often, only occupa- Istituto di Medicina del Lavoro, Universith di Cagliari, Cagliari, IdY P Cocco Occupational Epidemiology Branch, Division ofCancer Epidemiology and Genetics, National Cancer Institute, NIH, Bethesda, MD, USA M H Ward M Dosemeci Correspondence to: Dr Pierluigi Cocco, Isdtuto di Medicina del Lavoro, Universita di Cagliari, via San Giorgio 12,09124 Cagliari, Italy. Telephone 0039070602 8278;fax modest association with occupational exposure to inorganic dust (odds ratio (OR)=1.06; 95% confidence interval (95% CI) 1.03 to 1.11) with significant increas- ing trends by probability and intensity of exposure overall and by cross classifi- cation of the two exposure metrices. Workplace exposure to nitrosamines also showed a modest association (OR=1.06; 95% CI 1.01 to l . l l ) , but the excess risk was even smaller after adjusting for inorganic dust exposure. Risk of gastric cancer was not associated with any of the other workplace exposures considered in t h i s study. Conclusions-Non-differentialmisclassification of exposure may have caused negative findings in this study, and inorganic dust may be a partial surrogate for exposure to other unknown risk factors. tional titles are available to imply exposure to suspected gastric carcinogens. Even surrogates for exposure-such as duration of employment-are seldom used to calculate dose-response trends. Besides, small study size has been a limiting factor in interpreting findings. In the absence of industrial hygiene measurements, use of job exposure mamces, which has provided a useful tool in other studiesY5might be profitable. To test the hypothesis of a role of occupational risk factors in the aetiology of gastric cancer, we established job exposure matrices for 12 suspected gastric carcinogens in the workplace and applied them to the occupation-industry combinations in a large data base including death certificates from 24 states ftom the United States in 1984-96. Poor detail of the occupational information is still a concern in this 0039 070 654350;email coccop@pacs.unica.it Alternatively, our results suggest that study, but both the large size of the study occupational factors contribute little to population and the use of already designed job Accepted 13 July 1999 the aetiology of gastric cancer. Inorganic exposure matrices are new contributions. ~ 782 Cocco, Ward, Dosemeci Methods Since 1984, the National Cancer Institute, the National Institute for Occupational Safety and Health, and the National Center for Health Statisticshave supported the coding of occupation and industry titles on death certificates from 24 states from the United States accord- ing to the 1980 United States census occupa- tion and industry codes6 Details on the data base of death certificates from 24 states of the United States have been reported elsewhere.' We extracted data of 41 957 subjects who died from stomach cancer at age 325 from several million death certificates in 1984-96. These were 20 878 white men, 14 125 white women, 4215 African American men, and 2739 African American women. We used a case-control design to evaluate the association of risk of stomach cancer with 12 workplace exposures, separately by sex and race. Two controls for each case were selected from among subjects who died from non-malignant diseases, frequency matched to cases by geographic region, race, sex, and 5 year age group. One occupation and industry is reported on the death certificate. Information on duration or other characteristics of employment is not available. By applying job exposure matrices to the occupation-industry combinations in the death certificates of study subjects, we evaluated occupational exposure to 12 workplace hazards, that we previously discussed in a review on occupational risk factors for stomach cancer'-namely asbestos, inorganic dust, met- als, lead, polycyclic aromatic hydrocarbons (PAHs),nitrogen oxides, nitrosamines, sulphuric acid, fertilisers, herbicides, other pesticides (including insecticides and fungicides), and wood dust. An estimate of intensity (none=O, low=l, medium=2, high=3) and probability (none=O, low=l ,mediumz2, high=3) of expo- sure to each of the 12 occupational hazards was developed by two authors (MD and PC) for each three digit 1980 United States census occupation and industry code. Intensity of exposure was estimated based upon industrial hygiene and occupational health teXtbooksJ8 computerised exposure data bases (OSHA files, NIOSH inspections data base), unpublished industrial hygiene reports, and personal experience. The probability index was esumated based on the proportion of exposed Table 1 Risks of swmoch cancer by probabilityof exposure to suspectedgastric carcinogens, by sex and race Expxure S W gmtrp tinupored (OR=I) fn) RobabJity of exposure [n OR (95% GI)) Low Medium High Asbestos+ Inorganic dust Metals Lead Polycyclic aromatic hydrocarbons Nitrosamines Nitrogen oxides Sulphuric acid Herbicides Other pesticides Fertilisers Wood dust WM AAM ww AAW WM** AAM* ww** AAW WM AAM ww AAW WM AAM ww AAW WM AAM ww** AAW WM** AAM ww" AAW WM AAM ww*' AAW WM AAM ww AAW WM AAM ww .4AW WM AAM ww AAW WM AAM ww AAW WM AAM ww AAW 16859 2856 13356 2538 11974 1990 13482 2558 16572 3473 13696 2678 17276 3365 13753 2677 14232 2815 13736 2614 16478 3322 13810 2608 14794 2835 13815 2612 15228 2678 13128 2103 19190 3817 14062 2688 18099 3318 13930 2634 19058 3760 14012 2648 19107 3523 14022 2715 1361 274 175 28 662 120 67 9 1572 312 224 43 1356 256 201 33 792 163 176 40 788 216 147 37 4531 1212 675 219 146 62 5 1 1104 508 110 43 151 37 48 16 903 482 75 18 0.98 (0.92to 1.05) 1.16 (0.99to 1.36) 1.17 (0.96to 1.41) 1.12 (0.70to 1.79) 1.05 (0.96to 1.16) 1.08(0.86to 1.35) 0.86(0.65to 1.15) 1.41 (0.60to 3.32) 1.0(0.94to 1.07) 1.04 (0.90to 1.20) 1.01 (0.86to 1.19) 1.50(1.01to 2.24) 1.04 (0.97to 1.12) 1.02 (0.87to 1.20) 1.16(0.97to 1.38) 0.96 (0.63to 1.46) 1.01 (0.93to 1.10) 0.84 (0.70to 1.02) 1.09 (0.91to 1.32) 1.23(0.82to 1.83) 1.04 (0.95to 1.14) 0.99 (0.84to 1.18) 1.20 (0.97to 1.48) 1.49 (0.97to 2.29) 1.0(0.96to 1.04) 1.02 (0.94to 1.12) 0.91 (0.83to 1.0) 1.06(0.88to 1.28) 1.12 (0.91to 1.37) 1.13 (0.82to 1.55) 2.05 (0.59to 7.08) 0.64 (0.07to 6.20) 0.99 (0.91to 1.07) 1.17(1.03to 1.32) 1.41 (1.10to 1.81) 0.78 (0.54to 1.12) 0.90 (0.74to 1.09) 0.76 (0.52to 1.12) 1 .O (0.70to 1.41) 1.28(0.68to 2.42) 0.96 (0.89to 1.05) 1.17(1.03to 1.33) 0.91 (0.69to 1.20) 0.86 (0.49to 1.50) 4007 1357 769 201 2666 825 337 71 2142 448 292 46 527 85 83 9 1896 325 90 13 2760 567 123 88 669 219 38 8 675 223 310 416 19 10 4 0 121 58 23 12 148 100 7 25 195 80 22 6 0.97(0.93to 1.02) 1.08 (0.99to 1.17) 1.06(0.96to 1.17) 1.19 (0.98to 1.45) 1.11 (1.05 to 1.17) 1.19(1.07to 1.32) 1.22 (1.06to 1.41) 0.90(0.67to 1.20) 1.04 (0.98to 1.10) 1.01 (0.90to 1.14) 1.22 (1.05to 1.41) 1.14(0.79to 1.65) 1.01 (0.91to 1.13) 0.93(0.72to 1.21) 0.90 (0.69to 1.16) 1.37 (0.58 to 3.21) 1.06 (1.Oto1.13) 1.02 (0.88to 1.17) 1.13(0.87to 1.46) 1.48(0.72to 3.02) 1.08(1.02to 1.14) 0.97 (0.86 to 1.09) 1.40(1.11 to 1.78) 1.27 (0.96 to 1.69) 0.98(0.89to 1.08) 0.98(0.83to 1.16) 1.35(0.89 to 2.04) 1.48 (0.59to 3.70) 1.12(1.01to 1.23) 1.12(0.94to 1.33) 1.02 (0.87to 1.20) 0.96(0.79to 1.17) 0.91 (0.53to 1.55) 0.68(0.33 to 1.40) _2.1_0(0.52to 8.41) 1.05 (0.84to 1.31) 0.95 (0.69 to 1.30) 0.87 (0.53to 1.42) 0.71(0.37to 1.37) 1.30(1.06 to 1.60) 0.95(0.74to 1.22) 0.86 (0.36to 2.08) 0.97(0.59to 1.57) 0.91 (0.77to 1.08) 0.96(0.73to 1.26) 0.71 (0.44101.16) 1.33 (0.47to 3.76) 22 2 0 0 4877 1126 131 82 1502 174 70 6 1503 453 65 10 3394 819 98 79 848 163 16 3 4627 945 125 82 444 102 12 1 1523 326 54 50 1554 331 62 50 1521 318 58 50 673 130 6 0 1.09 (0.54to 2.20) -0._78(0.15to 4.03) __ 1.08 (1.03to 1.13) 1.08 (0.98to 1.20) 1.37(1.lOto 1.72) 1.16 (0.87to 1.55) 1.06 (0.99to 1.13) 1.08(0.89to 1.31) 0.99 (0.74to 1.32) 0.95 (0.35to 2.54) 0.92(0.86to 0.99) 1.15 (1.01to 1.32) 1.53 (1.10to 2.12) 1.76 (0.74to 4.16) 0.99 (0.95to 1.05) 1.01 (0.91 to 1.11) 1.57 (1.20to 2.06) 1.16 (0.87to 1.56) 1.05(0.97IO 1.15) 1.09(0.89to 1.32) 1.42(0.74to 2.70) 1.47 (0.33to 6.59) 1.0(0.96to 1.05) 1.03 (0.93to 1.13) 1.41 (1.12to 1.78) 1.17 (0.88 to 1.56) 0.96 (0.85 to 1.08) 1.16 (0.90to 1.50) 0.87 (0.44to 1.72) 2.07 (0.13to 33.3) 1.06(0.98to 1.15) 0.91 (0.78to 1.06) 1.71 (1.18to 2.46) 1.25 (0.86to 1.80) 1.04 (0.96 to 1.12) 0.94(0.80 to 1.09) 1.78(1.26to 2.51) 1.22(0.85 to 1.77) 1.06 (0.98to 1.15) 0.89 (0.76to 1.03) 1.77 (1.23to 2.53) 1.27(0.88 to 1.84) 1.0(0.90to 1.10) 0.93 (0.75to 1.16) _0.8_4(0.32to 2.18) *p<0.05; "p<O.Ol test for trend. tThe reference group for subjects exposed to asbestos includes those with a low probability of exposure, due to the difficulty in estimating a baseline exposwe. WM=white men; AAM=African .American men;WFZwhite women; AAW=fican American women). f i I 8 OB 91 LZ 6 OP 1 58I 65 09 I e 0651 os E5 LLE PLSI P El LL 92 92 96 L9Z PI11 01 89 6LI E6P 61 ZL szz 919 E LE zs 062 62 LEI 982 IPS1 oz L9 82 L9ZI Z 0 96 I EP 61 ZL 901 LE8 9 26 562 6ZS1 8 E 09 1 LES I 8 LI 26 OZP 6IP 8ZZ PLZI oz 86 60E PEE1 111 OEZ 999 PLPE KZ 85I OZE 861 SE 591 ZPZ PPI I sz ZPZ 862 60G I 9E 96Z ZP8 ZOO 1s EIZ 8P8 b9LZ 0 EZ 90P ILL I ZI oz 901 8E 20 I 95E 259 0 Z P ZZ ZIZ 59s 2621 E lIP 18 911 PO8 9E9E 0 LI 8P E+ 58 65I 558 LPOP PZ OLI 955 8912 L OS 85 I 95ZI SZI 082 0011 SE9P SILZ ZZOPI EZSE LO161 8P9Z ZIOPI 09LE 85061 PE9Z OE6E1 81EE 6608 I 8892 Z90P1 L18E 06161 EOIZ 8ZIEI 8L9Z 8ZZSI 2191 S18EI saz P6LPI 8092 OI8E1 ZZEE 8LP91 PI91 9ELEI swz ZEZPI LL9Z EGLEI 59% 9LZLI 8192 969EI ELPE ZL59I 8552 Z8bCI 0661 PL611 9892 Z16EI I LZE E8LLI uopednmo LUOJJ Bu!~uaplaylaqM 'apruyu%eur 30 iapio awes a q ~q sai03s lenppupq ~ A B Y o i SEM Chiens s y 3~0 adoss q e u ~atp clcpms masaid atp UI , ; a i a q ~ a ~ ppaqpssap uaaq Seq 1UaUISSaSSe amsodxa ahp3adsonai aA!leip -uenb!uras u! saioss Bupi[dnpw i o Syienbs io3 aleuo!iei arU, .apes d a s n p y atp 01 painq!nie ietp saurp apos uopednsso atp o i painqpm aioss aapsadsai ay, %u$Idppw woy p a l p a i sa~ossh p a i u ! pue &pqeqoid s'isalqns a y l uatp-dnsnpq uopnnsuos atp u! iamoqei io 'dnsnpq iaqqni a y l u! ia~!ipysnn 'aIdwe -xa io~-iuaurssasse ainsodxa 01 ioinqyuos qew atp SBM dnsnpqj! (9)faioss Ieuopadnsso atp 30 aienbs atp se sasoss hyqeqoid pue h ! s u a i q s,isa[qns a y l paieinsles am-~appm io 'iaqumld 'aldwexa 103-dnsnp~ 30 s s a ~ -p~e%auiopednsso atp urog paynsai ainsodxa UO UO!lEWiOJu! 3! ( D ) : S h M O W U! paiap!SUOJ a i m d n s n p q atp pue uopednsso atp tppi paie -posse saioss Qyqeqoid pue &!suaiq 'isa!qns dpms qsea ainsodxa U ~ A $ e io3 aioss h y q e -qoJd auo pue &!suaiy auo uyiqo o~ -~S!M~Y?I papos sawsnpq io suopednsso 30 iaqurnu a y l pue 'Ansnpy io apy qo! u~A!% e "VIM S J ~ ~ J O M i 784 Cocco, Ward, Dosemeci Table 3 Risks of stomach cancer by probability and intensity of exposure to asbestos, inorganic dust, metal dust, lead, nitrosamines, nitrogen oxides, PAHs, sulphuric acid, and fernhim (white men) Robabdizy of exposure (n OR (9S% CI.) Intensity of exposure All exposed Lau Medium High Errfor trend Inorganic dust: All levels None LOW Medium High p Value test for trend Metals: All levels None LOW Medium High Test for trend Nitrosamines: All exposed None Low Medium High p Value test for trend Sulphuric acid: AU levels None LOW Medium High p Value test for trend 8904 11974 4635 3002 1267 co.01 4306 16572 1.0 1256 1509 1541 0.01 4400 16478 433 3474 493 <0.01 5650 15228 4113 1274 263 0.01 1.07 (1.03 to 1.11) 1.o 1.05 (1.0 to 1.10) 1.07 (1.01 to 1.12) 1.13 (1.05 to 1.21) 1361 11974 971 302 88 NS 1.05 (1.0 to 1.09) 1.02 (0.95 to 1.09) 1.04 (0.97 to 1.11) 1.09 (1.02to 1.16) 662 16572 1.o 359 268 35 NS 1.06 (1.01 to 1.11) 1.o 0.99 (0.88 to 1.12) 1.06 (1.01 to 1.11) 1.16 (1.03 to 1.30) 792 16478 201 487 104 NS 1.01 (0.97 to 1.05) 1.O 0.99 (0.95 to 1.03) 1.03 (0.96 to 1.11) 1.23 (1.05 to 1.44) 4531 15228 3624 845 62 NS 0.98 (0.92 to 1.05) 1.0 0.97 (0.90 to 1.06) 1.04 (0.91 to 1.20) 0.94 (0.73 to 1.22) 2666 11974 1124 1125 417 <O.Ol 1.05 (0.96 to 1.16) 2142 1.06 (0.93 to 1.21) 1.06 (0.91 to 1.23) 1.01 (0.67 to 1.51) 16572 683 857 602 0.05 1.01 (0.93 to 1.10) I .o 0.94 (0.79 to 1.11) 1.01 (0.90to 1.13) 1.21 (0.95 to 1.54) 2760 16478 169 2313 278 co.01 1.O (0.96 to 1.04) 1.o 0.99 (0.95 to 1.04) 1.04 (0.95 to 1.13) 0.93 (0.68 to 1.26) 675 15228 144 408 123 <0.05 1.11 (1.05 to 1.17) 1.o 1.09 (1.0 to 1.18) 1.13 (1.04to 1.22) 1.10 (0.97 to 1.25) 4877 11974 2540 1575 762 <0.01 1.04 (0.98 to 1.10) 1502 1.o 1.0 (0.91 to 1.10) 1.02 (0.94 to 1.11) 1.12 (1.01 to 1.24) 16572 214 384 904 NS 1.08 (1.02 to 1.14) 1.o 1.05 (0.87 to 1.27) 1.07 (1.01 to 1.14) 1.16(1.0to 1.35) 848 16478 63 674 111 NS 1.12 (1.01 to 1.23) 1.o 1.13 (0.92 to 1.39) 1.05 (0.93 to 1.19) 1.42 (1.12 to 1.79) 444 15228 345 21 78 NS 1.08(1.03 to 1.13) 1.o 1.08 (1.01 to 1.16) 1.03 (1.0 to 1.10) 1.18 (1.08 to 1.30) pCO.01 pCO.01 NS pC0.01 1.06 (0.99 to 1.13) p=NS 1.o 1.01 (0.86 to .20) 1.06 (0.94 to 21) 1.07 (0.99 to 17) p=NS p=NS p<0.05 1.05 (0.97 to .15) 1.o 1.06 (0.78 to 1.45) 1.04 (0.94 to 1.14) 1.12 (0.89 to 1.41) p<0.05 p=NS pe0.05 p=NS 0.96 (0.85 to 1.08) 1.o 0.91 (0.80 to 1.04) 0.79 (0.48 to 1.30) 1.29 (0.97 to 1.72) p=NS p=NS p=NS p<O.Ol n=Number of exposed cases. distribution with a single degree of freedom, and it probes specifically linear trends in log relative risk with increasing exposure.'2 p Values were two tailed. Results Mean age at death from stomach cancer was lower among men than among women and among African Americans than white people (African American men: 67.3 (SD 13.0); Afiican American women: 71.4 (SD 14.6); white men: 69.4 (SD 12.3); white women: 74.2 (SD 12.8)). Compared with controls, cases were more often married in all four study groups, and they defmed their ethnic origin (mainly European, or Hispanic) twice as often as controls among white men and white women. Other variables-such as metropolitan residence and socioeconomic s t a t u s d i d not show a consistent frequency distribution by case-control across the four study groups (not shown in the tables). Risks by probability of exposure to 12 occupational risk factors combining all levels of exposure are shown in table 1. None of the investigated exposures showed an unequivocal trend in all study groups. Only inorganic dust Table 4 ORs (95% GI) for stomach cancer associared with any exposure w inorganic dust, metals, nitrosamines, and sulphuric acid ad'ted for all the other exposures and by marital and socioeconomic status, and metropolitan residence Expmre Inorganic dust Metals Nitrosamines Sulphuric acid OR 1.05 1.01 1.04 , 0.99 95% CI ~ ~~ 1.0 to 1.10 0.96 to 1.07 0.99 to 1.09 0.95 to 1.03 and nitrosamines were associated with ORs >1.0 for the high probability of exposure category in all four study goups. Two of the four risks associated with high probability of exposure to inorganic dust were significant and another was of borderline significance. Although absolute increases in risk were small, trends for probability of exposure to inorganic dust were significant among white men, white women, and African American men, and for exposure to nitrosamine among white men and white women. Table 2 shows risks by intensity of exposure to the same 12 occupational risk factors. High intensity exposure to the risk factors was rare among women. However, a significant positive trend in at least one study group was found for inorganic dust, metals, nitrosamines, nitrogen oxides, and sulphuric acid. Again, although absolute risk increases were small, the significant trends by intensity of exposure to inorganic dust and nitrosamines among white men replicated similar findings that used probability of exposure. Patterns of increasingrisk of stomach cancer by probability within intensity category and vice versa were explored for inorganic dust, metals, nitrosamines, and sulphuric acid (table 3). Although excess risks were quite small, trends by probability of exposure to inorganic dust were significant in low and high categories of intensity, and trends by intensity of exposure were significant for subjects classified in the medium and high probability of exposure. Results were less consistent for metals, nitrosamines, and sulphuric acid. No such pattern was found for exposure to the other workplace i Table 5 Risks of stomach cancer by probability and intensity of exposure to inorganic dust among subjects unexposed to metals and nimsamines (white men) h b a b i l i y of exposure (n OR (95% Go) Intenrizyofexposure Lau Medium High p value testfor trend Inorganic duit: Unexposed LOW Medium High p Value test for trend 10955 701 59 13 NS 1.0 0.93 (0.8410 1.02) 0.94 (0.69 to 1.28) 1.14 (0.58 to 2.24) 10955 705 293 59 NS 1.0 1.23 (1.10 to 1.38) 1.13 (0.97 to 1.30) 1.10 (0.80 to 1.51) 10955 180 680 359 C0.05 1.0 0.96 (0.80 to 1.15) 1.01 (0.91 to 1.12) 1.20 (1.04 to 1.38) NS NS CO.01 hazards considered in this study. To see probability and intensity of exposure overall whether reciprocal confounding was responsi- and by cross classification of the two exposure I ble for some of the observed associations, we meuics. Other suspected risk factors for gastric i combined all categories of exposure and fitted cancer did not show the same association or a model with the basic variables plus exposure showed a even weaker association after adjust- to inorganic dust, metals, nitrosamines, and ment for exposure to inorganic dust. Our data sulphuric acid to reciprocally adjust the do not provide information about whether respectiverisk estimates. As reported in table 4, ingestion of generic inorganic dust or specific this analysis resulted in a very modest increase dust components play a part in gastric carcino- in risk of stomach cancer for exposure to inor- genesis. Therefore, one possible explanation ganic dust and to nitrosamines, but no excess for our findings is that exposure to inorganic risk was associated with ever exposure to met- dust may have behaved as a partial surrogate als and sulphuric acid. for exposure to other unknown risk factors. Cross tabulation of probability and intensity Alternatively, dusty workplace environments of exposure to inorganic dust among workers could play a non-specific role in gastric unexposed to metals and to nitrosamines carcinogenesis. If the association were con- 3 showed significantly positive trends by prob- firmed with exposure to generic inorganic dust, ability in the high intensity category and by a plausible explanation would be that physical intensity in the high probability category (table properties of dust might be important, by 5). The analysis of such a risk pattern among causing local irritation or absorbing gastric subjects with isolated exposure to metals or to carcinogens on its surface and delivering them i I nitrosamines did not provide interpretable to the target cells in the gastric mucosa. Such a results due to small numbers and empty cells. mechanism was experimentally proved long Among white men, the most numerous ago in lung carcinogenesis.' occupation contributing to the excess risk associated with high probability and high intensity of exposure to inorganic dust was mining machine operators represented by 1.34% cases and 1.25% controls. Other less numerous conmbuting occupationswere: plasterers (0.05% cases and 0.04% controls), concrete and terrazzo finishers (0.12% cases and 0.07% controls), mining occupations not elsewhere classified (0.10% cases and 0.09% controls), excavating and loading machine opera- Previous case-control studies have reported an association between occupational exposure to dust and stomach cancer: l*" not explained by ethni~ity'~or diet.5 Results were less consistent for a role of specific dusts. A case-control study based on the Los Angeles County cancer registry found the greatest increase in risk of cancer of the antrum or pylorus for exposure to mineral dust." The association was also found in an Italian multicentre case-control study of 640 hystologically tors (0.04% cases and 0.03% controls), and crushing and grinding machine operators (two cases and no controls). In the same study group, the most represented industry contributing to the excess risk of stomach cancer in the htgh probability and high intensity of exposure to inorganic dust was mining and quarrying (1.49% cases and 1.35% controls). Coal mining, metal miniig, and nori-metallic mining and quarrying contributed equally to the modest excess risk. Other contributing industries were: construction (0.21% cases and 0.13% controls), and structural clay products (three cases and three controls). confinned male cases of stomach ~ a n c e r , ~ although no attempt was made to investigate subsites in more detail. In a large multicancer site, multifactor case-control study in Mon- treal, Canada, silica was the only inorganic dust," and wood the only organic dust" to show a positive association with stomach cancer. Disentangling exposure to silica from other inorganic dust would require additional information on workplaces and possibly industrial hygiene measurements, which was not the case in the present study. On the other hand, we did not find an association with wood dust, the only organic dust we tested, and trends associated with exposure to metals (which Discussion included dust and fumes) were less consistent In this large case-control study based on death than with the more generic category of certificates from 24 states from the United inorganic dust (including also metal dust). States we found a modest association of risk of Exposure to lead was not associated with a risk stomach cancer with occupational exposure to of stomach cancer, and adjustment for expoinorganic dust. Although the increase in risk sure to inorganic dust weakened the positive was M y , increasing trends were calculated by association found with nitrosamines. Cross 706 Cmo, Ward, hsemeci tabulation of probability and intensity of isolated exposure to nitrosamines did not provide further clues, because of small numbers and empty cells. Endogenous synthesis of nitrosamines from dietary precursors was suggested as the crucial event in gastric carcinogenesis.' Exposure to nitrosamine precursors or to preformed nitrosamines in the work environment was also suggested as a possible explanation for the repeatedly found increase in risk of gastric cancer in industries where such exposures may occur.' Further study including monitoring of exposure to nitrosamines in the work environment and in biological fluids of exposed workers are warranted. Occupational exposures in agriculture, such as to herbicides, other pesticides, and fertilisers, were not associated with a risk of gastric cancer in this study. A review of cancer among farmers suggested a possible increase in risk of stomach cancer.I6 The epidemiological evidence for this seems to be conflicting,' and confounding by rural residence was suggested as a possible e~planation.~ Limitations in this study are mainly related to the poor occupational information that may be extracted from death certificates and the insufficient specificity of the coding system. Use of the most prevalent occupation and industry in the working history of each person as reported in the death certificate may result in important loss of information on exposures experienced in other jobs, mainly among short term workers who usually experience the highest workplace exposure^.^' Besides, the three digit census code of occupations and indusmes may incorporate very heterogeneous workplace conditions within the same code, which prevents a reliable classification of exposure. Although this characteristic was considered in classifying occupations and indusmes by probability of exposure, a substantial amount of non-differential misclassification could have affected our results, which could have lowered the association between stomach cancer and occupational exposure to inorganic dust." Among possible confounders, ethnic origin, marital status, and socioeconomic status were included in the logistic regression model. As rural residence is another important confounder in occupational studies of stomach cancerYb5ut an urban or rural characterisation of residence was not available from the death certificates, we used metropolitan versus nonmetropolitan residence as a surrogate. Region of residence was controlled for by matching cases and controls. The analysiswas conducted separately by race and sex groups. As controlling for diet was not possible, residual confounding may have biased our findings in cases of important differences in dietary habits asso- ciated with the occupational exposures considered in this study. Gastric cancer risk is reportedly correlated with low educational level and low socioeconomic status.' l9 'O Therefore, risks associated with exposures typical of occupations classified in the lower socioeconomic groups, such as inorganic dust, might have been affected as well. Low socioeconomic status was not asso- ciated with stomach cancer risk in our study, which was conducted using a death certificate data-base. This source of information is not suitable to evaluate the association between low socioeconomic status and mortality from specific diseases, as bias can result from a more accurate definition of the cause of death and a lower mortality from all causes among the wealthier socioeconomic groups. As a conse- quence, in our study, when socioeconomic status was not included as a covariate in the logistic regression model to adjust risk esti- mates, risks associated with intensity and probability of exposure to inorganic dust were 1.O or very close to unity, and no positive trend was found (not shown in the tables). On the other hand, risk associated with high probabil- ity of exposure to inorganic dust was increased in three out of four socioeconomic groups comprising exposed cases (lowest socioeconomic status: 1.12; second lowest socioeconomic status: 1.11; medium socioeconomic status: 0.89; medium to high socioeconomic status: no cases; highest socioeconomic status: 1.08), although linear increases with probability of exposure were found only in the two lowest socioeconomic groups. Odds ratios for high intensity exposure to inorganic dust were increased in three socioeconomic groups (no cases and no controls were classified in this cell among the medium to high and the highest socioeconomic status category), and linear increases in risk with intensity of exposure were found in the second lowest and medium socioeconomic status (not shown in the tables). Therefore, low socioeconomic status was not an explanation for the modest increase in risk of stomach cancer associated with exposure to inorganic dust found in our study. In conclusion, our results confirm previous reports of an association between occupational exposure to inorganic dust and risk of stomach cancer. Further research is warranted to explore mechanisms and dose-response relations possibly with measurements of environmental dust. 1 Cocco P, Ward M, Buiam E. Occupational risk factors for stomach cancer: an overview. Epldnnwl Rev 1996;18:21834. 2 Correa P. Human gasmc carcinogenesis: a multistep and multifactorial process (First American Cancer Society Award Lecture on Cancer Epidemiology and Prevention). Cancer Res 1992;52:6735-40. 3 La Vecchia C, Negri E, D'Avamo B, et al. Food temperature and gasmc cancer. InrJ Cancer 1990;46:432-4. 4 SafEottiU,Cefis F, Kolb LH.A method for the experimental induction of bronchogenic carcinoma. Cancer Res 1968; 28:1 0 6 2 4 . 5 Cocco PL, Palli D, Buiam E, et al. Occupational exposures as risk factors for gamic cancer. Cancer Causes Caml 1994;5:241-8. 6 Bureau of the Census. Alphaberid index of i n d u s h and occupariar. Washington, DC, US: Department of Commerce, 1982. (PHC80-R3.) 7 Bumen CA, MaurerJ, Dosemeci M. M o d & by ocnrpatia, indusmy, and cause of &&. Cincinnati, OH: US Depanment of Health and Human Services,National Institutefor Occupational Safety and Health, 1997. (DHHS N O S H Pub1 No 97-1 14). 8 Parmeggiani L, ed. Encyclopedia of occupational health and safecy, 3rd ed. Geneva, Switzerland: International Labour Office, 1985. 9 Zenz C. Occupational medicine. Rinciples and practical a&cations, 2nd ed. Chicago, I L Yearbook Medical Publishers, 1988. 10 Gomez MR, Cocco PL, Dosemeci M, er al. Occupational exposure to chlorinated aliphatic hydrocarbons and brain cancer risk job-exposure matrix. Am 3Ind Med 1994;26: 171-83. 11 Green LW. Manual for scoring socioeconomic status for research on health behavior. Pub& Health Rep 1970;85: 815-27. 12 Breslow NE,Day NE.SrOrirticol methods in c a m research. Vol I. The analysis of care-umml snrdier. Lyon, France: International Agency for Research on Cancer, 1980. (IARC Sci Pub1 No 32.) 13 Siemiatycki J, Richardson L, Grin M, et al. Associations between several sites of cancer and nine organic dusts: results from a hypothesis-generating case-control study in Montreal, 1979-83. A m 3Epidemiol1986;123:235-49. 14 Wright WE, Bernstein L, Peters JM, et al. Adenocarcinoma of the stomach and exposure to occupational dust. Am 3 Epldemwll988;128:64-73. 15 SiemiatyckiJ, Dewar R, Lakhani R,et al. Cancer risks asso- ciated with 10 inorganic dusts: results from a casesonml study in Montreal. Am 3Ind Med 1989;16:547-67. 16 Blair A, Zham SH,Pearce NE,et al.Clues to cancer etiology from studies among farmers. ScandJ l&k Em'm Hedzh 1992;18:209-15. 17 Gardner MJ.Considerations in the choice of expected numbers for appropriate comparisons in occupational cohort studies. Med Law 1986;77:23-47. 18 Rothman KJ. Modern eprdemwlogy. Boston, MA. Little Brown, 1986. 19 Pearce NE,Howard JK. Occupation, social class and male cancer mortality in New Zealand, 1974-8. Int 3Epidmrior 1986;15:456-62. 20 Buiani E, Palli D, DeCarli A, et al. A case-conuol study of gastric cancer and diet in Italy. InrJ Cancer 1989;44:6116. i