Development of an Exposure Measurement Database on Five Lung Carcinogens (ExpoSYN) for Quantitative Retrospective Occupational Exposure Assessment

Transcription

1 Ann. Occup. Hyg., Vol. 56, No. 1, pp , 2012 Ó The Author Published by Oxford University Press on behalf of the British Occupational Hygiene Society doi: /annhyg/mer081 Development of an Exposure Measurement Database on Five Lung Carcinogens (ExpoSYN) for Quantitative Retrospective Occupational Exposure Assessment SUSAN PETERS 1 *, ROEL VERMEULEN 1,2, ANN OLSSON 3, RAINER VAN GELDER 4, BENJAMIN KENDZIA 5, RAYMOND VINCENT 6, BARBARA SAVARY 6, NICK WILLIAMS 7, TORILL WOLDBÆK 8, JÉRÔME LAVOUÉ 9, DOMENICO CAVALLO 10, ANDREA CATTANEO 11, DARIO MIRABELLI 12, NILS PLATO 13, DIRK DAHMANN 14, JOELLE FEVOTTE 15, BEATE PESCH 5, THOMAS BRÜNING 5, KURT STRAIF 3 and HANS KROMHOUT 1 1 Environmental Epidemiology Division, Institute for Risk Assessment Sciences, Utrecht University, Utrecht, the Netherlands; 2 Julius Center for Health Sciences and Primary Care, University Medical Center, Utrecht, the Netherlands; 3 International Agency for Research on Cancer, Lyon, France; 4 Institute for Occupational Safety and Health of the German Social Accident Insurance, Sankt Augustin, Germany; 5 Institute for Prevention and Occupational Medicine of the German Social Accident Insurance, Bochum, Germany; 6 Institut National de Recherche et de Sécurité, Vandoeuvre lès Nancy, France; 7 Health and Safety Executive, Bootle, UK; 8 National Institute of Occupational Health, Oslo, Norway; 9 Research Centre of University of Montreal Hospital Centre, University of Montreal, Canada; 10 Department of Chemical and Environmental Sciences, Università degli Studi dell Insubria, Como, Italy; 11 Department of Occupational and Environmental Health, Università degli Studi di Milano and Unit of Epidemiology, Fondazione IRCCS Ca Granda-Ospedale Maggiore Policlinico, Milan, Italy; 12 Cancer Epidemiology Unit, CPO-Piemonte and University of Turin, Italy; 13 The Institute of Environmental Medicine, Karolinska Institutet, Stockholm, Sweden; 14 IGF-BG RCI, Bochum, Germany; 15 Département santé travail, Institut de veille sanitaire, St Maurice, France Received 27 April 2011; in final form 30 June 2011; published online 11 October 2011 Background: SYNERGY is a large pooled analysis of case control studies on the joint effects of occupational carcinogens and smoking in the development of lung cancer. A quantitative job-exposure matrix (JEM) will be developed to assign exposures to five major lung carcinogens [asbestos, chromium, nickel, polycyclic aromatic hydrocarbons (PAH), and respirable crystalline silica (RCS)]. We assembled an exposure database, called ExpoSYN, to enable such a quantitative exposure assessment. Methods: Existing exposure databases were identified and European and Canadian research institutes were approached to identify pertinent exposure measurement data. Results of individual air measurements were entered anonymized according to a standardized protocol. Results: The ExpoSYN database currently includes measurements from 19 countries. In total, personal and stationary data points were available. Measurements were distributed over the five agents as follows: RCS (42%), asbestos (20%), chromium (16%), nickel (15%), and PAH (7%). The measurement data cover the time period from 1951 to present. However, only a small portion of measurements (1.4%) were performed prior to *Author to whom correspondence should be addressed. Tel: ; fax: ; s.peters@uu.nl 70

2 Development of an exposure measurement database on five lung carcinogens The major contributing countries for personal measurements were Germany (32%), UK (22%), France (14%), and Norway and Canada (both 11%). Conclusions: ExpoSYN is a unique occupational exposure database with measurements from 18 European countries and Canada covering a time period of >50 years. This database will be used to develop a country-, job-, and time period-specific quantitative JEM. This JEM will enable data-driven quantitative exposure assessment in a multinational pooled analysis of community-based lung cancer case control studies. Keywords: exposure database; general population; multinational; occupational epidemiology; retrospective exposure assessment INTRODUCTION Recently, a large pooled analysis of case control studies on the joint effects of occupational carcinogens and smoking in the development of lung cancer was initiated (Olssonetal., 2011). Thisprojectiscalled SYNERGY. Up until now, 14 lung cancer case control studies from 13 European countries (Czech Republic, France, Germany, Hungary, Italy, the Netherlands, Poland, Romania, Russia, Slovakia, Spain, Sweden, and the UK) and Canadahavebeenincluded.Together,thesestudiescomprise39 518subjects: casesand21 813controls. Information on job history and lifestyle factors is available for all subjects. Since the occurrence of joint exposure to two or more occupational carcinogens in the generalpopulationisconsideredtobelowduetothewide range of possible workplaces, a large pooled analysis is necessary to provide the essential power for the analysis of possible synergistic effects. Assessment of occupational exposures in community-based studies is complicated by the wide variety of jobs in which the general population will have been employed (Kromhout and Vermeulen, 2001a). Moreover, when studying cancer, the relevant exposures are likely to have occurred at least a decade or more ago due to the long latency between exposure and disease onset. Retrospective assessment of occupational exposures is often qualitative (yes/no) or semi-quantitative (low/medium/high) in communitybased studies. Methods used to assign those estimates are case-by-case expert judgement, job-exposure matrices (JEMs), self-reported exposure, or a combination of these (McGuire et al., 1998). However, to explore exposure response relationships, quantitative exposure assessment is necessary (e.g. in terms of airborne concentrations in milligrams per cubic metre). Detailed exposure monitoring is an option in crosssectional or (prospective) cohort studies, but the retrospective and community-based nature of case control studies renders this impossible. In principle, however, quantitative exposure estimation can be achieved by modelling of available relevant historical exposure measurement data. To facilitate quantitative exposure response modelling within the SYNERGY project, a data-driven JEM will be constructed using historical occupational exposure data from a wide variety of industries, workplaces, countries, and time periods. Such occupational exposure data can be found in several types of databases. Firstly, there are national exposure databases, for instance, the NEDB in the UK (Burns and Beaumont, 1989), COLCHIC in France (Vincent and Jeandel, 2001), MEGA in Germany (Gabriel et al., 2010), and EXPO in Norway (Osvoll and Woldbaek, 1999). However, the purposes of these databases are different: MEGA and COLCHIC serve social security occupational accident and disease prevention programmes, while NEDB and EXPO were set up from a research perspective. Secondly, industry-specific multinational exposure databases exist, for instance, PAPDEM for the paper and pulp industry (Kauppinen et al., 1997), EXASRUB for the rubber manufacturing industry (De Vocht et al., 2005), and AWE for the asphalt industry (Burstyn et al., 2000). Furthermore, there are multinational exposure-specific databases, such as the IMA DMP on respirable (silica) dust (Houba et al.,2009), the WOODEX project for wood dust exposures (Kauppinen et al., 2006), and DERMDAT for dermal exposures (Kromhout and Vermeulen, 2001b). In addition, many (unpublished) measurement data from research institutes and individual research groups exist (Burstyn et al., 2000; De Vocht et al., 2005). The aim of this paper was to describe the structure and contents of the ExpoSYN database. Furthermore, we discuss the feasibility of creating such a multinational exposure database and the methodological issues that arose during the process. METHODS Five established lung carcinogens were selected based on their relatively high expected exposure prevalence in the general population: asbestos, chromium, nickel, polycyclic aromatic hydrocarbons (PAH), and respirable crystalline silica (RCS). Table 1 lists the

3 72 S. Peters et al. Table 1. Selected agents for exposure assessment in the SYNERGY project. Selected lung carcinogen Specification CAREX estimated number of exposed workers European Union a, 139 million workers (%) Canada b, 17 million workers (%) Asbestos Chrysotile (white asbestos), (0.9) (0.9) amphibole, and other type/unspecified Chromium Chromium VI and total chromium c (0.6) c (0.5) Nickel Soluble nickel compounds, insoluble (0.4) (0.3) nickel compounds, and total nickel PAH Benzo(a)pyrene and naphthalene d (0.7) d (1.8) RCS Respirable quartz, respirable cristobalite, respirable tridymite, and RCS (2.3) (2.1) a (Kauppinen et al., 2000). b 2006 (CAREX Canada, 2010). c Chromium VI only. d Disregarding tobacco smoke (passive exposure at work) and diesel motor exhaust. selected agents and their subcategories for which air exposure measurements were collected. Estimates for the number of exposed workers in the European Union ( ) and Canada (2006) are also shown, as provided by the CAREX (CARcinogen EXposure) database (Kauppinen et al., 2000; CAREX Canada, 2010). Collection of individual exposure measurement data points was essential since these, rather than summary statistics, in theory, will give us the opportunity to adjust for differences in the way the historical exposure data were obtained, such as differences in sampling and analytical methods, measurement strategies, and sampling duration. Differences in these elements may result in biased estimates, so besides collecting exposure measurement data, we endeavoured to obtain ancillary data to enable adequate adjustments in statistical modelling (Rajan et al., 1997). A format for structured data entry was developed and input variables were described in a protocol. Table 2 presents the variables included in the Expo- SYN database. These variables represent the a priori defined variables deemed essential for exposure modelling (Rajan et al., 1997). One of the key variables was the job title, which was coded according to the International Classification of Occupations (ISCO), 1968 (International Labour Office, 1968; Mannetje and Kromhout, 2003). Industry was coded by the International Standard Industrial Classification (ISIC Rev.2) (Statistical Office of the United Nations, 1971; Mannetje and Kromhout, 2003). Many data had to be coded into ISCO and ISIC from text fields or recoded from other coding systems. Measurement strategy was divided into a statistical or representative approach (where sampling is performed in order to obtain representative exposure data for all workers with that job title, the preferred strategy) and a worst-case scenario (i.e. sampling under conditions when highest exposures were anticipated). Reasons for measurements included survey, inspection/complaints, or compliance. Categorization schemes for these and other variables were derived in an expert-based manner using available information since the exact same formulations were not utilized in all datasets. All data in the ExpoSYN database were entered without traceable factory or personal identifiers; both were recoded into numeric identification codes. Exposure data collection was performed from September 2007 until August We focused on the countries included in the SYNERGY project, although we have also collected data from neighbouring countries when available in order not to lose valuable information. Existing exposure databases were identified and owners of these databases were approached for collaboration. Both personal (samples collected with a personal portable pump in the breathing zone of the worker) and stationary (samples from a fixed location sampler) measurements were entered into the database. Furthermore, research institutes in Europe and Canada were approached in order to find additional occupational exposure data. Due to cost and time considerations, our search for additional exposure data was more focused on personal rather than stationary measurements since these can be directly linked to jobs. Moreover, personal measurements more accurately capture a workers exposure, which is modified by a worker s position in the work environment and workers behaviour that may modify exposure levels (Cherrie, 2003). No biological monitoring data were included. Data entry according to the protocol was performed mainly (85%) by the institutes where data were originally stored; the rest was provided as raw data and subsequently put in the correct format by one of the authors (S.P.). The data provided by

4 Development of an exposure measurement database on five lung carcinogens 73 Table 2. Variables included in the ExpoSYN database. Type of information Variables Input General Country Text Centre Text Date DD-MM-YYYY Factory ID Text Worker ID Text Sample ID Text Job characteristics Industry code ISIC rev.2 Job code ISCO 1968 Respiratory protective No and yes (type of RPE optional) equipment (RPE) Local exhaust ventilation No and yes Indoors or outdoors Indoors, outdoors, and partially Measurement information Personal or stationary sample Personal and stationary Measured concentration Number Unit of measurement Text (e.g. mg m 3 ) Sampling duration Number (min) Measured fraction Inhalable, thoracic, and respirable (for PAH: particulate and gaseous phase) Sampler type E.g. seven-hole Filter type E.g. glass fibre Digestion method a E.g. acid bomb Analytical method E.g. X-ray diffraction Measurement strategy Statistical approach/representative and worst case Measurement reason Survey, inspection/complaints, and compliance Below or above limit of Below LOD, above, or equal to LOD detection (LOD) Limit of detection for Number measurement Unit of measurement LOD Text (e.g. mg m 3 ) a For chromium and nickel only. contributors were inspected centrally before entry into the ExpoSYN database in order to correct mistakes and to improve consistency in coding. No data were excluded apriorias specific exclusion criteria for statistical modelling of the data differ by agent. RESULTS The ExpoSYN database is a non-hierarchical database comprising five sections: one for each selected agent within the SYNERGY study. To date, the database contains a total of measurements from 19 countries. The majority (90%) of the data originate from national exposure databases [the main sources were MEGA (Germany), COLCHIC (France), NEDB (UK), and EXPO (Norway)], 1% originates from industry-specific databases (AWE and PAPDEM), and the remaining 8% were collected at institutes. The number of measurements differed considerably between the five agents: RCS (n ; 42%), asbestos (71 816; 20%), chromium (57 119; 16%), nickel (52 751; 15%), and PAH (25 954; 7%). The major contributing countries to the complete database were Germany (60% of all measurements), the UK and France (both 10%), Canada (7%), Norway (6%), and Italy (4%). These and other characteristics of the exposure data are presented in Table 3. Information on measurement reason and measurement strategy was missing for 44 and 15% of the data points, respectively, and differed considerably between agents. Missing information for measurement reason ranged from 18% for PAH to 58% for asbestos, while percentages of missing values for measurement strategy ranged from 8% (RCS) to 38% (asbestos). For data for which this information was available, surveys and inspection were the main reasons (28 and 21%, respectively). Measurements were reported to be mainly performed in a representative way (77%).

5 Table 3. Characteristics of the ExpoSYN database by agent. Asbestos (n ) Chromium (n ) Nickel (n ) PAH (n ) RCS (n ) All agents (n ) Agent type (%) Chrysotile (67) Cr VI (43) Soluble Ni (1) B(a)P (65) Quartz (93) Amphiboles (15) Total Cr (57) Insoluble Ni (,1) Naphtalene (35) Cristobalite (2) Unspecified (18) Total Ni (99) Tridymite (,1) Unspecified (4) Time period (years) Number of measurements (%) Personal (39) Stationary (61) Reason for measurements (%) Survey (17) (25) (22) (40) (12) (21) Inspection (16) (32) (34) 7725 (30) (33) (28) Compliance 6750 (9) 6831 (12) 5363 (10) 2204 (8) 7710 (5) (8) Unknown (58) (31) (34) 5530 (21) (50) (44) Measurement strategy (%) Representative (53) (75) (74) (75) (88) (76) Worst case 6665 (9) 9002 (16) 7067 (13) 5443 (10) 5080 (3) (9) Unknown (38) 5302 (9) 5724 (13) 1119 (15) (8) (15) Countries (%) Personal measurements (%) Bulgaria (,1) 28 (,1) Canada (7) (11) Denmark (,1) 2229 (2) Finland (,1) 306 (,1) France (10) (14) Germany (60) (32) Greece (,1) 36 (,1) Iceland (,1) 76 (,1) Italy (4) 4920 (4) The Netherlands (,1) 1797 (1) 74 S. Peters et al.

6 Development of an exposure measurement database on five lung carcinogens 75 Table 3. Continued All agents (n ) RCS (n ) PAH (n ) Nickel (n ) Chromium (n ) Asbestos (n ) Norway (6) (11) Poland (,1) 1671 (1) Romania (,1) 428 (,1) Russia (,1) 203 (,1) Slovakia (1) 1655 (1) Spain 1 1 (,1) 0 (0) Sweden (,1) 1641 (1) Switzerland (,1) 131 (,1) UK (10) (22) The exposure data cover the period from 1951 until 2009, but only a small portion of data points (1.4%) were available before 1975 (see Fig. 1). The oldest data points were of asbestos (both stationary and personal) and RCS (stationary) measurements. Figure 2 gives an impression of the historical coverage of the exposure data (separated by personal and stationary measurements) relative to the working history calendar years in the SYNERGY study population. Measurement data cover the most recent years of the almost 90 working years ( ) of the cases and controls. In total, personal data points were available, mainly from Germany (32%), the UK (22%), France (14%), Norway, and Canada (both 11%), as shown in Table 3. For 90% of these data points, an IS- CO 1968 code was known. Figure 3 illustrates the comparison between the major job groups represented in the SYNERGY study population, and the jobs for which measurements are available in the ExpoSYN database. As expected, the majority of measurements (97%) were performed in occupations in production or transport industries, covering nearly 600 ISCO codes. Fifty per cent of the total working years of the SYNERGY study population involved work in these blue collar occupations. DISCUSSION The ExpoSYN database contains occupational exposure data points from all over Europe and Canada. Information regarding exposure measurements for asbestos, chromium, nickel, PAH, and RCS was coded in a standardized format. The database was set up to provide measurement data for quantitative exposure assessment in a pooled multinational occupational lung cancer case control study. Exposure assessment in such a design is complicated by the fact that, given the long latency time of lung cancer, the relevant exposures will have occurred several decades ago. To perform retrospective exposure assessment in a quantitative way, historical exposure measurements are essential. Accurately converting historical exposure data into currently accepted exposure metrics can be a very demanding task. Ideally, one would use exposure measurements from the population under study (Dahmann et al., 2008). However, as can be seen in Fig. 2, itwilltake several decades before exposure measurements will cover complete work histories of all subjects in retrospective case control studies, provided that exposure measurements are still being conducted. Personal exposure monitoring became widely implemented in the 1970s (Cherrie, 2003). Therefore,

7 76 S. Peters et al. Fig. 1. Distribution of exposure measurements in the ExpoSYN database over time. Fig. 2. Distribution of exposure measurements in the ExpoSYN database over time (separated by stationary and personal sampling) compared to the working years represented in the SYNERGY population. exposure data prior to the 1970s are scarce; an issue that others also faced when collecting historical exposure data for retrospective assessment (Kauppinen et al., 1997; Burstyn et al., 2000; Vermeulen et al., 2010). Most existing exposure databases started in the 1980s; earlier measurement data had generally not yet been digitized. Some exceptions are the MEGA database, for which stationary measurements were already collected in the 1970s, and the personal asbestos exposure measurements from the British Health and Safety Executive. About data points (8%) collected for the ExpoSYN database were not yet included in national or industry-specific exposure databases. Efforts were made to retrieve relevant data from hard copy records and to contact retired experts to help identify and describe old data. Although 8% is only a small portion of the complete ExpoSYN database, these measurements might be highly valuable since these originated from countries that were otherwise not (or sparsely) covered, such as Italy, the Netherlands, Poland, Romania, Russia, Slovakia, and Sweden and moreover comprised

8 Development of an exposure measurement database on five lung carcinogens 77 Fig. 3. Distribution of personal exposure measurement data in the ExpoSYN database over major groups of jobs (1 digit ISCO 1968) compared to number of jobs represented in the SYNERGY population. predominantly less recent data. Searching for unpublished data has been recommended and previously proved to be a highly rewarding effort (Kauppinen et al., 1997; Burstyn et al., 2000; De Vocht et al., 2005; Vermeulen et al., 2010). Many of the researchers involved in obtaining the non-digitized exposure data are, or will soon be, retired. This will likely lead to a loss of valuable information. Therefore, it is becoming increasingly pressing to recover useful exposure data hidden away in archives. Industrial hygienists who collected data and know the details of these datasets might still be available to collaborate and provide the pertinent detailed ancillary information. Consequent digitization of exposure measurements will enable and improve the use of measurement data for exposure assessment in future occupational epidemiological studies. Overall, the ExpoSYN database has good coverage of the past years, which provides opportunities for estimating exposure-specific time trends. Coverage of each job by country is limited. Although exposure circumstances may vary between- and within-jobs, industries, countries, or regions, information can potentially be extrapolated from similar jobs within a country or between countries. Countries with few measurements may especially gain from the exposure information provided by data from the same jobs and industries elsewhere. It can be questioned whether measurements in the ExpoSYN database are really representative of the general working situation and if they reflect true occupational exposure levels considering selection bias. Exposure measurements tend to be objective, but identification and selection of workplaces and workers to be measured involve substantial subjectivity. Although representative sampling should be performed completely at random, measurements are generally only taken when and where exposures are expected. In the most extreme case, a ( white collar ) job which normally does not involve exposure might be monitored in a very unusual situation when exposure occurs. That same job would not be monitored in the normal situation when exposure is not present. Collecting information about measurement reason and strategy enabled us to try to identify such situations. Potentially biases should be considered when using data from any database. Kauppinen et al. (1997) estimated that 94% of the measurement data in the PAPDEM database were not significantly biased, as judged by exposure experts, whereas 5% of the samples overestimated exposure and 1% underestimated it. For the situations where the measurement strategy was known in our database, 89% of data points were classified as being representative. Setting up a prospective exposure database allows for a detailed protocol for data collection and registration. Unfortunately, this is not possible when collecting measurement data retrospectively. In the latter case, one has to work with information often collected for totally different goals or for different purposes. For the ExpoSYN database, it became apparent that it would not be feasible to obtain all ancillary data required by the database protocol for all measurements. Such missing data will complicate interpretation and comparison of the measurement data. For example, it was often unknown if measurements were performed

9 78 S. Peters et al. in the same factory or for the same worker. Betweenand within-subject or factory variability in exposure concentrations will therefore be hard to estimate. Analyses on subsets of the database might give an idea about the variance components, under the assumption that these data are missing at random. Lack of information on measurement strategies or measurement reasons makes it impossible to ascertain the representativeness of all measurements, which could lead to biased results. However, even if this information is available, it is hard to standardize the concept of representativeness and it is questionable whether people taking measurements were always able to assess this consistently. Furthermore, data originated from very different sources. Diversity in purpose of data collection for the original databases or in sampling methods or clustering of data in a few workplaces; all have the potential to bias exposure measurement results (Olsen et al., 1991; Teschke et al., 2002; Agostini et al., 2010). Definitions and standardization of these aspects of measurement strategies are clearly needed. This is the first attempt to set up a large international exposure database covering all types of industries for community-based epidemiological studies. Data collection took a large amount of time and work. Before being entered into the ExpoSYN database, all data had to be transformed into a standardized format and therefore, (re)coding of occupations and industries was often necessary. We aimed to cover a wide range of industries and countries, which complicated coding due to the large amount of possible jobs. Additionally, because of the international character of the database, we had to deal with differences in languages and coding systems. Beside some well-known national exposure databases, not all data were easy to trace and part of these data were not digitized. In the future, the ExpoSYN database can be expanded by adding other agents, although one should realize that the five agents selected for the SYNERGY project were relatively prevalent in occupational settings (affecting % of the working population, as presented in Table 1) and were often measured. It would be more difficult to find sufficient measurement data to study exposure to less prevalent agents. CONCLUSIONS The ExpoSYN database is a unique database, which includes exposure measurement data from 18 European countries and Canada. The large quantity of measurement results will allow for elaboration of a country-, job-, and time period-specific quantitative JEM for exposure assessment in a multinational pooled analysis of community-based lung cancer case control studies from various European countries and Canada. Moreover, these data might be useful for future methodological studies in exposure assessment. FUNDING The SYNERGY project is funded by the German Social Accident Insurance (DGUV) and is coordinated by the International Agency for Research on Cancer, the Institute for Prevention and Occupational Medicine of the DGUV, Institute of the Ruhr-University Bochum (IPA) and the Institute for Risk Assessment Sciences at Utrecht University. Acknowledgements The authors would like to thank the following industrial hygienists, epidemiologists, and other experts from European and Canadian research institutes for their contributions to this database and paper: Cheryl Peters, Martie van Tongeren, Thomas Schneider, Timo Kauppinen, Paolo Boffetta, Rodica Stanescu, Zuzana Vassányi, Ewa Orlowski, Slawek Brzeznicki, Jan Gromiec, Stefano Taronna, Gunilla Klingspetz, Nicole Charrière, Martin Tischer, Paul Swuste, Ton Spee, Evert Meijer, Mieke Lumens, and Virissa Lenters. REFERENCES Agostini M, de Vocht F, van Tongeren M et al. (2010) Exposure to rubber process dust and fume since 1970s in the United Kingdom; influence of origin of measurement data. J Environ Monit; 12: Burns DK, Beaumont PL. (1989) The HSE National Exposure Database-(NEDB). Ann Occup Hyg; 33: Burstyn I, Kromhout H, Cruise PJ et al. (2000) Designing an international industrial hygiene database of exposures among workers in the asphalt industry. Ann Occup Hyg; 44: CAREX Canada. (2010) Report No: 2009/2010 Progress report. Cherrie JW. (2003) The beginning of the science underpinning occupational hygiene. Ann Occup Hyg; 47: Dahmann D, Taeger D, Kappler M et al. (2008) Assessment of exposure in epidemiological studies: the example of silica dust. J Expo Sci Environ Epidemiol; 18: De Vocht F, Straif K, Szeszenia-Dabrowska N et al. (2005) A database of exposures in the rubber manufacturing industry: design and quality control. Ann Occup Hyg; 49: Gabriel S, Koppisch D, Range D. (2010) The MGU a monitoring system for the collection and documentation of valid workplace exposure data. Gefahrst Reinhalt L; 70: Houba R, Vlaanderen J, Jongen R et al. (2009) Building an industry-wide occupational exposure database for respirable mineral dust experiences from the IMA dust monitoring programme. J Phys: Conf Ser; 151. International Labour Office. (1968) International standard classification of occupations. Geneva, Switzerland: International Labour Office Publications. Kauppinen T, Teschke K, Savela A et al. (1997) International data base of exposure measurements in the pulp, paper and

10 Development of an exposure measurement database on five lung carcinogens 79 paper product industries. Int Arch Occup Environ Health; 70: Kauppinen T, Toikkanen J, Pedersen D et al. (2000) Occupational exposure to carcinogens in the European Union. Occup Environ Med; 57: Kauppinen T, Vincent R, Liukkonen T et al. (2006) Occupational exposure to inhalable wood dust in the member states of the European Union. Ann Occup Hyg; 50: Kromhout H, Vermeulen R. (2001a) Application of job-exposure matrices in studies of the general population: some clues to their performance. Eur Respir Rev; 11: Kromhout H, Vermeulen R. (2001b) Temporal, personal and spatial variability in dermal exposure. Ann Occup Hyg; 45: Mannetje A, Kromhout H. (2003) The use of occupation and industry classifications in general population studies. Int J Epidemiol; 32: McGuire V, Nelson LM, Koepsell TD et al. (1998) Assessment of occupational exposures in community-based case-control studies. Annu Rev Public Health; 19: Olsen E, Laursen B, Vinzents PS. (1991) Bias and random errors in historical data of exposure to organic solvents. Am Ind Hyg Assoc J; 52: Olsson A, Gustavsson P, Kromhout H et al. (2011) Exposure to diesel motor exhaust and lung cancer risk in a pooled analysis from case-control studies in Europe and Canada. Am J Respir Crit Care Med; 183: Osvoll PO, Woldbaek T. (1999) Distribution and skewness of occupational exposure sets of measurements in the Norwegian industry. Ann Occup Hyg; 43: Rajan B, Alesbury R, Carton B et al. (1997) European proposal for core information for the storage and exchange of workplace exposure measurements on chemical agents. Appl Occup Environ Hyg; 12: Statistical office of the United Nations. (1971) Indexes to the International standard industrial classification of all economic activities. New York, NY: Publishing Service United Nations. Teschke K, Olshan AF, Daniels JL et al. (2002) Occupational exposure assessment in case-control studies: opportunities for improvement. Occup Environ Med; 59: Vermeulen R, Coble JB, Lubin JH et al. (2010) The Diesel Exhaust in Miners Study: IV. Estimating historical exposures to diesel exhaust in underground non-metal mining facilities. Ann Occup Hyg; 54: Vincent R, Jeandel B. (2001) COLCHIC-occupational exposure to chemical agents database: current content and development perspectives. Appl Occup Environ Hyg; 16: