The purpose of this appendix is to describe the characteristics and attributes that the committee used to evaluate the exposure-assessment component of epidemiologic studies. The committee first provides introductory information on exposure assessment and its use to understand disease risk in defined populations. The committee then discusses components of an exposure assessment, including defining job titles; measuring exposures using time-weighted averages (TWAs), cumulative exposures, and peak exposures; strategies for exposure sampling; choosing an appropriate summary measure of exposure; and creating a job–exposure matrices (JEM) for use in cohort studies. The committee also discusses differences between exposure assessments for cohort studies of specific industries compared with general-population case–control studies and case–control studies nested within cohorts. The appendix ends with a summary table of the criteria that the committee used to assess the epidemiologic studies cited in Chapters 2 and 3.
INTRODUCTION TO EXPOSURE ASSESSMENT
The fundamental logic of epidemiologic analysis is the 2 × 2 table, in which one axis is the subjects’ disease status (yes–no) and the other is their personal exposures (yes–no). The quality of a study and the strength of its conclusions depend strongly on exposure evaluation, in addition to its epidemiologic aspects. The basic goal of an exposure assessment is to evaluate the qualitative and quantitative discrimination of a study’s exposure assignments. Different methods have different powers of discrimination.
By definition, exposure is personal and external to the individual. The points of entry for chemical exposure are the nose and mouth for inhalation, the mouth for ingestion, and the skin for dermal absorption. All three have the following dimensions: composition, intensity, and time course. Complexity arises along all three dimensions. First, it is rare that a person is exposed to only a single substance, such as formaldehyde; mixed exposures almost always occur. Some components of mixtures, such as formaldehyde vapors and paraformalde-
hyde particles emitted during embalming, may produce similar effects or may modify the effects of other substances, which may serve to confound relationships with disease. The physical form of formaldehyde vapor or particulate paraformaldehyde will strongly affect where it is deposited in the respiratory tract. Second, environmental concentrations are generally not constant in time or location. Sources of airborne formaldehyde are not continuous and steady. As a result, exposure varies in time and location. In addition, concentrations of individual mixture components may vary independently or correlate, depending on their sources. Third, the area near local emission sources, such as embalming fluid in body cavities, produces the highest, variable air concentrations that usually have considerable random temporal variation. These variations are the result of incomplete dilution and mixing processes in the breathing zone air over short periods, which produce approximately lognormal distributions for variations in consecutive concentration measurements. Regional concentrations away from local sources will be lower and are relatively more stable. Outdoor exposures commonly show hourly, daily, seasonal, or annual trends that are associated with weather, climate, source output, exposed subjects’ activities, ventilation, and other factors. Those aspects of exposure are discussed in detail by Lippmann et al. (2003) and Smith and Kriebel (2010).
Epidemiologic researchers seek to exploit natural experiments in which large differences in environmental or occupational exposures occur among large groups of otherwise similar people. The exposure-assessment goal is to identify personal, occupational, or environmental factors that determine differences in exposure to the substance of interest (Checkoway et al. 2004). The gradients in exposure, if sufficiently large, can be used to determine whether there are corresponding gradients in disease risk that might be causally related. Useful occupational-exposure gradients can be produced by the nature of the subjects’ jobs, tasks, or activities in the workplace and by the characteristics of work locations and materials used, such as formaldehyde solutions or paraformaldehyde powder. Similarly, characteristics of subjects’ residences, commuting activities, food sources, and other determinants of environmental exposure can be used to define exposure groups for comparisons of risk. The rationale and quality of data used to assign exposure are important in determining the quality and reliability of the assignments. Blair and Stewart (1992) showed that improved quantitation of formaldehyde exposure tended to increase exposure gradients and sharpen estimates of relative risk. Exposure assignments that are imprecise can result in individuals being categorized into the wrong category of the exposure gradient and the epidemiologic study analysis table. Misclassification reduces the apparent relative risk and may produce misleading conclusions.
An important step in the use of exposure data for an epidemiologic study is the construction of a summary measure of exposure (Smith and Kriebel 2010). When semiquantitative or quantitative data on intensity and duration of exposure for study participants are available, these must be summarized—usually in a single number—to be used in an epidemiologic model to assess the strength of
the exposure–risk association. The choice of which summary measure1 to use should ideally be based on biologic hypotheses about the underlying causal mechanism. In practice, this information is often lacking, so indices are tested and goodness-of-fit data are used to assess which metric is more likely to be (approximately) correct. Unfortunately, the precision of exposure metrics is often low. As a result, it is not possible to determine if one metric is substantially better than another. Such a distinction would be highly useful.
COMPONENTS OF AN EXPOSURE ASSESSMENT
Industrial hygienists are trained to recognize hazards in an occupational setting, how to evaluate those hazards, and how to reduce or control exposures. Part of their expertise includes analyzing workplace organization and defining jobs and their job titles, work activities, or work locations in specific industries. Typically, job titles, department titles, and work locations will be collected from an individual’s job history, which is usually held in company records. Company records also commonly contain extensive data on the site of the industrial operations, including plant maps, locations of major equipment and operations where exposures would have taken place, the raw materials used and products and byproducts produced, and the emission-control equipment used and when it was installed. Industrial hygienists are also trained to take measurements of chemical exposures of individual workers and to assess the quality of available measurement data. Industrial hygienists who are interested in epidemiologic research may also obtain training in the estimation of historic exposures suitable for the extrapolation of long-term past exposures associated with chronic disease. Table C-1 shows how basic knowledge about sources of formaldehyde emissions, the physical setting, the type of job, the job location, and the activities that make up the job can be used to make useful distinctions that discriminate among different levels of exposure. Various approaches have been used to define differences between scenarios of high, medium, low, and no exposure. The various approaches are not equally useful for discriminating exposures with minimum misclassification. High-quality exposure assessments can accomplish that by using the strategies outlined in the section below.
Job titles are labels used by management for personnel functions to organize work activities. In some cases those work activities may have close links with exposure, but the job titles may or may not be associated with exposures depending on how the work activities were distributed across the job titles. The
1Typical summary measures of exposure include average exposure, duration of exposure, cumulative exposure, and various measures of peak exposure.
|High exposure||• Job histories that include job titles, tasks, or activities that take place close to sources of concentrated formaldehyde emissions can provide information on the potential for high exposures.
• Job-site data can provide information on work areas, equipment, and chemicals that are heavily used and handled often.
• Emission and work-area measurement data indicate general high-exposure levels, and poorly mixed, concentrated emissions may produce substantial peaka exposures.
• Absence of emission controls or poor ventilationb in a setting in which vapor can accumulate, such as a warehouse where materials off-gas incompletely or where reactive chemical coatings are present, can lead to high mean concentrations but less extreme peaks.
|Medium exposure||• Job and work-area data identify tasks or activities that take place at a distance from sources of concentrated formaldehyde emissions. These exposures are difficult to define qualitatively or semiquantitatively, and data is often absent. The central tertile of a measured exposure distribution is prone to misclassification into both the high-exposure and low-exposure groups.
• Often insufficient job or work-area data or unevaluated assumptions lead to misclassification of exposures as high or low.
|Low exposure||• Jobs, tasks, or activities with only brief periods when formaldehyde vapors are present, and the work location is distant from the sources.
• Physical separation of work areas from areas with emission sources.
• Good ventilation prevents vapors from accumulating in the area.
|Exposure controls||• If respirators or ventilation engineering systemsc are used, it is important to find documentation in plant records that describe when the controls started to be used and how effective they were at reducing or eliminating exposure.
• Respirators or ventilation systems were usually effective after the middle 1970s. Before then, they were less effective.
|No exposure||• Work in a setting that has no sources of formaldehyde emissions.|
aPeak exposures are short-duration (approximately 15 minutes, but the precise length is often not defined), high-concentration (for example, >2–4ppm) exposures. They may be defined by the limitations of measurement methods.
bVentilation is the amount of air flowing through a work space from windows and doors and by forced ventilation.
cVentilation engineering systems, including fans, ductwork, hoods, and enclosures that provide ventilation, control and minimize airborne emissions.
work histories of study subjects can be a useful link with occupational exposure conditions, but that link and the exposure conditions must be defined with an exposure assessment. A given job title may be associated with substantially different work activities and exposures in different companies or during different historical periods. For example, a chauffeur today may drive a limousine, but a chauffeur before the 1960s was often a truck driver, and that required a chauffer’s license. Therefore, a person with a chauffeur’s license in the 1960s may have had very different exposures compared to a chauffeur today. Industrial hygiene expertise and data from long-term workers is required to translate job and work location information into exposure assignments.
A widely used set of standardized job descriptions—the International Standard Industrial Classification—has been developed by the United Nations (UN 2008). A similar set of more specifically defined occupations—the Dictionary of Occupational Titles—has been developed by the US Department of Labor (DOL 1991). Because the UN and Department of Labor job titles are broad, their link to exposures is often weak, and misclassification is common. A simple and specific job title may be satisfactory for exposure classification if it unequivocally links with an exposure situation. For example, a person whose job title is “embalmer” often uses solutions with high concentrations of formaldehyde while embalming bodies in a small, poorly ventilated room. Those conditions will consistently lead to exposure to high concentrations of formaldehyde vapor. Other, more generic and broad titles, such as “mortician” or “funeral director”, also may involve embalming but less often, and embalming is not one of the main job activities. Thus, the title “mortician” is broader and includes more people but leads to more misclassification and much less discrimination for formaldehyde exposure than the title “embalmer”.
Epidemiologists and exposure assessors have addressed the poor specificity of standard job titles by adding sets of titles that are specific for the industries under study. They have also added questions to questionnaires and interview guides to ask about specific jobs, activities, and substances that are expected to be present, such as “embalmer”, “embalming”, and “formaldehyde and paraformaldehyde”. They may also ask about irritation and odors that distinguish particular substances. The utility of such questions depends on the subjects’ knowing the names and other properties of an agent. It is common for workers not to know the names of substances to which they are exposed; for example, they may know only that they use a clear liquid in a blue can to clean up grease and oil. The identity of the liquid must come from other sources, such as material safety data sheets kept by the company.
Professional requirements, unionization, and certifications can improve the exposure specificity of job definitions. Legal requirements for embalming and preparation of bodies for interment reduce the variation in exposure opportunities. Lower-level nonprofessional jobs—such as laborer, technician, and assistant—often have poorly defined tasks and work locations and are difficult to classify with respect to exposure.
Formal quantitative measurement is the best way to determine to what and where people are exposed. The accuracy and precision of exposure measurements have improved greatly, particularly since the 1970s when extensive exposure surveys and routine monitoring began (Stewart et al. 2000) and when standards were established for allowable exposures in the United States and other countries (Stewart et al. 1996; Symanski et al. 1998). Current methods for exposure measurements were developed by the National Institute for Occupational Safety and Health. The methods have been standardized to measure allowable exposures or emissions for regulatory purposes and they are used by the US Environmental Protection Agency for measuring formaldehyde vapor.2 The numbers of samples collected have also increased because of concern about exposure variability. In many cases, few or no historical exposure data have been available for long-term health studies. However, increasingly sophisticated extrapolation strategies have been developed (discussed below).
Inhaling a time-varying concentration at a fixed rate, such as 10 L/min (light exercise) for a specific time period (such as an 8-hour work day) produces a TWA concentration over the period of exposure. Similarly, drawing air or water into a collector at a fixed flow rate (volume per unit time) for a defined period produces a TWA sample in the collector because an equal volume is passed through each minute (Δt) and each unit of volume contributes material in proportion to the concentration:
TWA = SUM(C[i]Δt)/SUM(Δt) = SUM(C[i])/N,
Where C is the concentration of the substance, i is the period, and N identifies the number of periods; T = NΔt. That is analogous to inhalation at a fixed breathing rate and is a good dose metric for exposed subjects during their work period (shift).
Cumulative exposure is perhaps the most commonly used summary measure of exposure in occupational epidemiology of chronic diseases. Cumulative exposure is defined as the product of the average exposure concentration (C)
multiplied by the duration of exposure (T). The theoretical basis for the widespread use of cumulative exposure is Haber’s rule (also called Haber’s law), which posits that within an appropriate range for inhaled toxicants, all combinations of C and T with the same value will all produce the same effect (Belkebir et al. 2011). The rule breaks down outside narrow ranges of C and T. The rule implies that high exposures for short durations produce the same effects as low exposures for long durations, which may not be true. This rule also implicitly assumes that all toxic processes have no thresholds or lags for responses.
Some of the formaldehyde cancer studies reviewed by the committee used cumulative exposure to summarize occupational exposures across each study subject’s entire work history. If Haber’s rule holds for the carcinogenic effects of formaldehyde, then summarizing exposure histories using cumulative exposure will not introduce any exposure misclassification. However, if a few years of high exposure early in a subject’s work life are more important for cancer risk than many years of low exposure, then using cumulative exposure will introduce misclassification and reduce the likelihood of detecting an association.
In studies of occupational exposure in which the intensity of exposure has not been measured, duration of work is sometimes used as a surrogate for cumulative exposure. Duration of exposure will be proportional to cumulative exposure when the average exposure is approximately the same for all members of the cohort, so that the only person-to-person variability in cumulative exposure derives from differences in duration of exposure. Because this assumption is not likely to be true, there can be substantial misclassification within and between exposure groups in their average durations of work and exposure, which will probably bias the results toward the null (Kriebel et al. 2007). On the other hand, differences in the intensity of exposure among groups in early studies, such as those between embalmers and other funeral workers, were probably quite large so that substantial differences in risk by years of work or exposure would be expected.
High-intensity but short-duration exposures are called peaks. Peaks are of interest because they are much higher concentrations than the mean exposure and as a result may exceed a minimum intensity needed to cause an acute effect that has threshold or nonlinear pharmacodynamics. Peaks are quantified by the product of concentration at the point of entry and duration (C × Δt), where Δt is a short period, such as 15 minutes. The smallest C × Δt for a biologically relevant peak is implicitly the acute dose needed to produce a minimum effect. The range of definitions of a peak used by studies can be broad because the minimum effective dose is not known. The concentration of formaldehyde reported to cause upper airway irritation, 2–4 ppm, has commonly been used to define the minimum peak concentration, but it is not known whether this is relevant for carcinogenesis. As a practical approach, the limitations of the industrial-hygiene
measurement techniques have often been used to define the minimum intensity and Δt of peaks. Allowable peak exposures set by regulatory agencies are based on characteristics and limitations of monitoring methods. For example, formaldehyde concentrations that exceed 2 ppm for 15 minutes exceed the Occupational Safety and Health Administration short-term exposure limit (OSHA 2014). Historically, the 15-minute duration was chosen because 15 minutes of sampling were needed to collect enough material for acceptable measurement precision. Biologically important peaks of shorter duration might produce upper respiratory tissue effects. The acute-dose definition, C × Δt, also breaks down at the extremes of concentration and short duration because of pharmacokinetic and physiologic limits on uptake, transport, activation and deactivation, and removal.
Some jobs or activities have clear opportunities for peak exposures; for example, embalmers work with high-concentration sources nearby, but others do not, such as workers in a garment warehouse. In a garment warehouse, the incomplete polymerization of a fabric’s permanent-press treatment is the source of formaldehyde. Emissions from a single garment are limited, but there are many hundreds or thousands of garments throughout the warehouse. Thus, concentrations do not vary widely, and there are not expected to be high peaks, but average concentrations can be high; high peaks and high-TWA exposures do not necessarily occur together. An exposure assessment specifically designed for the task is needed to determine where peaks may occur.
Assessment of peak exposures for jobs and work activities requires considerable detailed information. Only large, extensive studies have collected the necessary data and measurements to estimate the intensity, frequency, and duration of situations with peak exposures, such as the National Cancer Institute cohort studies of the US chemical industries and the funeral industry (Beane Freeman et al. 2009; Hauptmann et al. 2009). Peaks also contribute to TWA exposures, but they are of short duration and the correlation between peaks and cumulative exposures tends to be weak (Blair and Stewart 1992). Thus, if peaks are causally related to cancer risk, then using average exposure or cumulative exposure metrics will introduce misclassification. However, the peak-exposure metrics are of limited precision and may not be sufficient to distinguish a peak mode of action from a cumulative mode of action. As stated in Chapter 3, it is expected that, on average, choosing the wrong metric will result in an underestimation of an association if one exists (Checkoway et al. 2004).
Exposures are generally highly variable in time and location, but it is impractical to measure them all continuously. Therefore, measurement of personal and location exposures use several types of statistical sampling strategies. Sampling strategies have changed considerably with the development of personal TWA sampling (a small pump and lapel collector) and the implementation of the
Occupational Safety and Health Act of 1970. Therefore, historical sampling data from before the 1970s need to be carefully evaluated and sometimes adjusted (Corn 1992).
The most useful strategy for epidemiology is the collection of random personal samples from different exposure groups. They should be collected at or near the route of entry, such as in the breathing zone, and for the whole duration of exposure. Fixed-location (“area”) samples or stationary samples have been widely collected in places where people may be present at some times, but these may lead to overestimation of exposure if they are taken closer to sources than where people are normally located. They may lead to underestimation of exposures if people are present for only short periods relative to the duration of the sampling or if people normally are closer to sources compared with the location of the monitors. If area samplers are used consistently with the same strategy, they tend to produce samples that are proportional to personal exposures, and the proportionality can be estimated on the basis of the ratio of concurrent personal to area sampling.
Job–Exposure Matrix for a Cohort Study in a Single Company
JEM methods were developed by several investigators, including Stewart et al. (1996). The approach used by industrial hygienists to develop JEM assignments for cohort studies is summarized below.
1. Job titles and plant or worksites associated with jobs are abstracted from company work histories, or cases and controls or their proxies are interviewed.
2. Jobs, worksites, processes, and work rooms are located on plant diagrams, and historical changes are also recorded.
3. Industrial hygienists with knowledge of the industry visit plants for walk-throughs and discuss operations, processes, materials, jobs, and historical changes with long-term workers and supervisors. This information is used to develop a plant history.
4. Industrial hygienists collect all available exposure measurements, personal data, and area data. The amount and quality of data will vary widely by date, plant area, and job. The data also may be limited by plant closures and loss of records.
5. If possible, industrial hygienists conduct field studies to measure exposures and conduct studies of job activities and task exposures, as was done by Stewart et al. (1992) for embalmers.
6. In some cases, sufficient data are available to develop detailed statistical models that can be used to estimate exposures. An example is the work by Hornung et al. (1996). Alternatively, extrapolation models have been developed on the basis of physical principles and extrapolations from current conditions
backward in time (Stewart et al. 1996; Tielemans et al. 2008; Fransman et al. 2011).
7. The exposure information and other data that are collected are used to develop JEM tables for each unique job title and work location by year. Estimates are made of the TWA and of the potential for peak exposures, the frequency of such exposures, and the intensity for each substance of interest. A good example is Blair et al. (1986). Some JEMs may be less complete than others, and this will limit the types of exposure estimates that are possible and may increase the amount of misclassification and thus reduce the ability of a study to detect small risks. The plant-history documentation and exposure estimates are sent to participating plants for technical review by company engineers and industrial hygienists to verify their accuracy.
EXPOSURE ASSESSMENT FOR CASE-CONTROL STUDIES
Exposure assessment for case–control studies that draw their subjects from the general population is difficult because they generally rely on recalled job titles and industries. Even when recall is accurate, there will be a loss of information because the occupation and industry information must be coded using a broad classification system such as the International Standard Classification of Occupations (ISCO) and the International Standard Industrial Classification. An example is a worker reporting he was a salesman for automotive parts. His position might be coded using ISCO code 43 for “male technical salesmen, commercial travelers, and manufacturer’s agents.” That broad grouping will usually have little specificity for a particular chemical exposure of interest, such as formaldehyde. In addition, the distribution of occupations and exposures depends heavily on the distribution of local industries and the prevalence of formaldehyde users in a region. That problem can be reduced by choosing a base population that has a large prevalence of an industry of interest. The study by Luce et al. (2002) drew from areas that had large industries processing wood, which resulted in few subjects who were exposed to formaldehyde without also being exposed to wood dust. Some investigators, such as Luce et al. (2002), improve their specificity by preparing an additional detailed questionnaire on formaldehyde-related jobs. However, as noted earlier, workers or their next of kin often do not know their exposures to specific chemicals with which they worked.
Where there are no exposure data for the study sites, expert or professional industrial-hygiene judgment is often used to estimate who has been exposed and their degree of exposure. Jobs, work activities, and work areas need to be evaluated to achieve specificity. Questionnaire data collected from the subjects, their peers, or next of kin are often evaluated by industrial hygienists familiar with local conditions to assess job or area exposures. There have been a number of evaluations of such expert judgment. For example, Luce et al. (1993) conducted an evaluation of expert judgment used in their population-based case–control study of sinonasal cancer.
Formaldehyde’s irritant properties are readily recognized, which may make identifying the presence of this specific exposure easier. Coggon et al. (1984) used the presence of substantial irritation as a marker of “high” exposure in areas where formaldehyde was known to be used. Unfortunately, this approach is limited by the broad variation in human sensitivity to irritants and by the tendency for people to acclimatize after a period of low to moderate exposure. Also, sensitive individuals may leave the workplace while long-term workers may be self-selected for being relatively insensitive to the irritant effects. As a result, worker appraisals of irritation may underestimate the exposures.
Case–control studies that are drawn from members of an exposed cohort (that is, “nested” case–control studies) have an advantage for exposure assessment because exposures in the source cohort may already have been assessed, and detailed exposure assignments may be available (Checkoway et al. 2004). That can make a study very discriminating for specific agents and long periods.
INFORMATION USED TO EVALUATE EXPOSURE ASSESSMENTS
The committee evaluated five aspects of each epidemiologic study reviewed in Chapters 2 and 3 to determine the quality of discrimination and the utility of an exposure assessment. Those aspects are the expertise of the investigators, the assessment type (such as, personal monitoring or JEM methods), the availability of key data (including job history, site information, and sampling measurements), the potential for misclassification (both qualitative and quantitative), and, where possible, the evaluation of the peak exposures. High quality in the first four aspects of an assessment produces a strong exposure assessment with high discrimination for long-term exposures. Table C-2 shows the information the committee used to review and evaluate the epidemiologic studies cited in Chapters 2 and 3.
|Overall Method||Exposure-Assessment Components||Exposure Assignments||Discrimination of Exposure Differences Between Categories|
|Job-History Data||Site Data and Industrial-Hygiene Evaluation||Sampling Data||Extrapolation of Past Exposures|
|Qualitative—broad occupational groups and industries in a region||None||None||None||None||Yes–qualitative||Low—few exposed in broad job groups; strong tendency to overestimate number exposed; likely large misclassification|
|Semiquantitative—specific jobs in one industry||Yes—job descriptions, interviews, questionnaires, and proxies; industrial hygienist uses professional judgment to assess exposures||None—many worksites or no data on specific sites||None or very limited for the industry||None or maybe some data on time trends; industrial hygienist uses professional judgment to assess past exposures||Yes–Semiquantitative in years of exposure||Moderate—specific job titles and work site data; limited measurements; likely much overlap between categories High—specific jobs with defined exposures and limited overlap of low and high categories|
|Quantitative—specific jobs or areas in a company||Yes—detailed company records||Yes—extensive data on operations, sites, and job activities||Yes—extensive for high-exposure jobs or areas over time||Yes—detailed strategies and modeling; industrial hygienist uses professional judgment to assess past exposures||Quantitative by substance, job or area, and period according to dose metrics||Moderate—if specific job, area, or sampling data are limited; likely overlap between groups High—limited overlap between low- and high-exposure categories|
Beane Freeman, L.E., A. Blair, J.H. Lubin, P.A. Stewart, R.B. Hayes, R.N. Hoover, and M. Hauptmann. 2009. Mortality from lymphohematopoietic malignancies among workers in formaldehyde industries: The National Cancer Institute Cohort. J. Natl. Cancer Inst. 101(10):751-761.
Belkebir, E., C. Rousselle, C. Duboudin, L. Bodin, and N. Bonvallot. 2011. Haber’s rule duration adjustments should not be used systematically for risk assessment in public health decision-making. Toxicol. Lett. 204(2-3):148-155.
Blair, A., and P.A. Stewart. 1992. Do quantitative exposure assessments improve risk estimates in occupational studies of cancer. Am. J. Ind. Med. 21(1):53-63.
Blair, A., P. Stewart, M. O’Berg, W. Gaffey, J. Walrath, J. Ward, R. Bales, S. Kaplan, and D. Cubit. 1986. Mortality among industrial workers exposed to formaldehyde. J. Natl. Cancer Inst. 76(6):1071-1084.
Checkoway, H., N. Pearce, and D. Kribel. 2004. Research Methods in Occupational Epidemiology, 2nd Ed. Oxford: Oxford University Press.
Coggon, D., B. Pannett, and E.D. Acheson. 1984. Use of job-exposure matrix in an occupational analysis of lung and bladder cancers on the basis of death certificates. J. Natl. Cancer Inst. 72(1):61-65.
Corn, M. 1992. Historical perspective on approaches to estimation of inhalation risk by air sampling. Am. J. Ind. Med. 21(1):113-123.
DOL (U.S. Department of Labor). 1991. Dictionary of Occupational Titles, 4th Ed. [online]. Available: http://www.oalj.dol.gov/libdot.htm [accessed Jan. 31, 2014].
Fransman, W., M. Van Tongeren, J.W. Cherrie, M. Tischer, T. Schneider, J. Schinkel, H. Kromhout, N. Warren, H. Goede, and E. Tielemans. 2011. Advanced Reach Tool (ART): Development of the mechanistic model. Ann. Occup. Hyg. 55(9):957-979.
Hauptmann, M., P.A. Stewart, J.H. Lubin, L.E. Beane Freeman, R.W. Hornung, R.F. Herrick, R.N Hoover, J.F. Fraumeni Jr., A. Blair, and R.B. Hayes. 2009. Mortality from lymphohematopoietic malignancies and brain cancer among embalmers exposed to formaldehyde. J. Natl. Cancer Inst. 101(24):1696-1708.
Hornung, R.W., R.F. Herrick, P.A. Stewart, D.F. Utterback, C.E. Feigley, D.K. Wall, D.E. Douthit, and R.B. Hayes. 1996. An experimental design approach to retrospective exposure assessment. Am. Ind. Hyg. Assoc. J. 57(3):251-256.
Kriebel, D., H. Checkoway, and N. Pearce. 2007. Exposure and dose modeling in occupational epidemiology. Occup. Environ. Med. 64(7):492-498.
Lippmann, M., B.S. Cohen, and R.B. Schlesinger. 2003. Environmental Health Science: Recognition, Evaluation, and Control of Chemical and Physical Health Hazards. Oxford: Oxford University Press.
Luce, D., M. Gerin, F. Berrino, P. Pisani, and A. Leclerc. 1993. Souces of discrepancies between a job exposure matrix and a case expert assessment for occupational exposure to formaldehyde and wood-dust. Int. J. Epidemiol. 22(suppl.2):S113-120.
Luce, D., A. Leclerc, D. Begin, P.A. Demers, M. Gerin, E. Orlowski, M. Kogevinas, S. Belli, I. Bugel, U. Bolm-Audorff, L.A., Brinton, P. Comba, L. Hardell, R.B. Hayes, C. Magnani, E. Merler, S. Preston-Martin, T.L. Vaughan, W. Zheng, and P. Boffetta. 2002. Sinonasal cancer and occupational exposures: A pooled analysis of 12 case-control studies. Cancer Causes Control 13(2):147-157.
OSHA (Occupational Safety and Health Administration). 2014. Formaldehyde standards [online]: Available: https://www.osha.gov/pls/oshaweb/owadisp.show_document?p_id=10075&p_table=STANDARDS [accessed June 19, 2014].
Smith, T.J. and D. Kriebel. 2010. A Biologic Approach to Environmental Assessment and Epidemiology. New York: Oxford University Press.
Stewart, P.A., R.F. Herrick, C.E. Feigley, D.F. Utterback, R. Hornung, H. Mahar, R. Hayes, D.E. Douthit, and A. Blair. 1992. Study design for assessing exposures of embalmers for a case-control study. Part I. Monitoring results. Appl. Occup. Environ. Hyg. 7(8):532-540.
Stewart, P.A., P.S. Lees, and M. Francis. 1996. Quantification of historical exposures in occupational cohort studies. Scand. J. Work Environ. Health 22(6):405-414.
Stewart, P.A., R. Carel, C. Schairer, and A. Blair. 2000. Comparison of industrial hygienists’ exposure evaluations for an epidemiologic study. Scand. J. Work Environ. Health 26(1):44-51.
Symanski, E., L.L. Kupper, I. Hertz-Picciotto, and S.M. Rappaport. 1998. Comprehensive evaluation of long-term trends in occupational exposure: Part 2. Predictive models for declining exposures. Occup. Environ. Med. 55(5):310-316.
Tielemans, E., T. Schneider, H. Goede, M. Tischer, N. Warren, H. Kromhout, M. Van Tongeren, J. Van Hemmen, and J.W. Cherrie. 2008. Conceptual model for assessment of inhalation exposure: Defining modifying factors. Ann. Occup. Hyg. 52(7):577-586.
UN (United Nations). 2008. International Standard Industrial Classification (ISIC) of All Economic Activities, Rev. 4, ST/ESA/STAT/SER.M/4/Rev.4. Department of Economic and Social Affairs, Statistics Division, UN, New York [online]. Available: http://unstats.un.org/unsd/publication/seriesM/seriesm_4rev4e.pdf [accessed Jan. 31, 2014].