Workplace exposure assessment is concerned with identifying and evaluating agents with which a worker may come in contact, and exposure indices can be constructed to reflect the amount of an agent present in the general environment or in inhaled air, as well as to reflect the amount of agent that is actually inhaled, swallowed or otherwise absorbed (the intake). Other indices include the amount of agent that is resorbed (the uptake) and the exposure at the target organ. Dose is a pharmacological or toxicological term used to indicate the amount of a substance administered to a subject. Dose rate is the amount administered per unit of time. The dose of a workplace exposure is difficult to determine in a practical situation, since physical and biological processes, like inhalation, uptake and distribution of an agent in the human body cause exposure and dose to have complex, non-linear relationships. The uncertainty about the actual level of exposure to agents also makes it difficult to quantify relationships between exposure and health effects.
For many occupational exposures there exists a time window during which the exposure or dose is most relevant to the development of a particular health-related problem or symptom. Hence, the biologically relevant exposure, or dose, would be that exposure which occurs during the relevant time window. Some exposures to occupational carcinogens are believed to have such a relevant time window of exposure. Cancer is a disease with a long latency period, and hence it could be that the exposure which is related to the ultimate development of the disease took place many years before the cancer actually manifested itself. This phenomenon is counter-intuitive, since one would have expected that cumulative exposure over a working lifetime would have been the relevant parameter. The exposure at the time of manifestation of disease may not be of particular importance.
The pattern of exposure—continuous exposure, intermittent exposure and exposure with or without sharp peaks—may be relevant as well. Taking exposure patterns into account is important for both epidemiological studies and for environmental measurements which may be used to monitor compliance with health standards or for environmental control as part of control and prevention programmes. For example, if a health effect is caused by peak exposures, such peak levels must be monitorable in order to be controlled. Monitoring which provides data only about long-term average exposures is not useful since the peak excursion values may well be masked by averaging, and certainly cannot be controlled as they occur.
The biologically relevant exposure or dose for a certain endpoint is often not known because the patterns of intake, uptake, distribution and elimination, or the mechanisms of biotransformation, are not understood in sufficient detail. Both the rate at which an agent enters and leaves the body (the kinetics) and the biochemical processes for handling the substance (biotransformation) will help determine the relationships between exposure, dose and effect.
Environmental monitoring is the measurement and assessment of agents at the workplace to evaluate ambient exposure and related health risks. Biological monitoring is the measurement and assessment of workplace agents or their metabolites in tissue, secreta or excreta to evaluate exposure and assess health risks. Sometimes biomarkers, such as DNA-adducts, are used as measures of exposure. Biomarkers may also be indicative of the mechanisms of the disease process itself, but this is a complex subject, which is covered more fully in the chapter Biological Monitoring and later in the discussion here.
A simplification of the basic model in exposure-response modelling is as follows:
exposure uptake distribution,
elimination, transformationtarget dosephysiopathologyeffect
Depending on the agent, exposure-uptake and exposure-intake relationships can be complex. For many gases, simple approximations can be made, based on the concentration of the agent in the air during the course of a working day and on the amount of air that is inhaled. For dust sampling, deposition patterns are also related to particle size. Size considerations may also lead to a more complex relationship. The chapter Respiratory System provides more detail on the aspect of respiratory toxicity.
Exposure and dose assessment are elements of quantitative risk assessment. Health risk assessment methods often form the basis upon which exposure limits are established for emission levels of toxic agents in the air for environmental as well as for occupational standards. Health risk analysis provides an estimate of the probability (risk) of occurrence of specific health effects or an estimate of the number of cases with these health effects. By means of health risk analysis an acceptable concentration of a toxicant in air, water or food can be provided, given an a priori chosen acceptable magnitude of risk. Quantitative risk analysis has found an application in cancer epidemiology, which explains the strong emphasis on retrospective exposure assessment. But applications of more elaborate exposure assessment strategies can be found in both retrospective as well as prospective exposure assessment, and exposure assessment principles have found applications in studies focused on other endpoints as well, such as benign respiratory disease (Wegman et al. 1992; Post et al. 1994). Two directions in research predominate at this moment. One uses dose estimates obtained from exposure monitoring information, and the other relies on biomarkers as measures of exposure.
Exposure Monitoring and Prediction of Dose
Unfortunately, for many exposures few quantitative data are available for predicting the risk for developing a certain endpoint. As early as 1924, Haber postulated that the severity of the health effect (H) is proportional to the product of exposure concentration (X) and time of exposure (T):
H=X x T
Haber’s law, as it is called, formed the basis for development of the concept that time-weighted average (TWA) exposure measurements—that is, measurements taken and averaged over a certain period of time—would be a useful measure for the exposure. This assumption about the adequacy of the time-weighted average has been questioned for many years. In 1952, Adams and co-workers stated that “there is no scientific basis for the use of the time-weighted average to integrate varying exposures …” (in Atherly 1985). The problem is that many relations are more complex than the relationship that Haber’s law represents. There are many examples of agents where the effect is more strongly determined by concentration than by length of time. For example, interesting evidence from laboratory studies has shown that in rats exposed to carbon tetrachloride, the pattern of exposure (continuous versus intermittent and with or without peaks) as well as the dose can modify the observed risk of the rats developing liver enzyme level changes (Bogers et al. 1987). Another example is bio-aerosols, such as α-amylase enzyme, a dough improver, which can cause allergic diseases in people who work in the bakery industry (Houba et al. 1996). It is unknown whether the risk of developing such a disease is mainly determined by peak exposures, average exposure, or cumulative level of exposure. (Wong 1987; Checkoway and Rice 1992). Information on temporal patterns is not available for most agents, especially not for agents that have chronic effects.
The first attempts to model exposure patterns and estimate dose were published in the 1960s and 1970s by Roach (1966; 1977). He showed that the concentration of an agent reaches an equilibrium value at the receptor after an exposure of infinite duration because elimination counterbalances the uptake of the agent. In an eight-hour exposure, a value of 90% of this equilibrium level can be reached if the half-life of the agent at the target organ is smaller than approximately two-and-a-half hours. This illustrates that for agents with a short half-life, the dose at the target organ is determined by an exposure shorter than an eight-hour period. Dose at the target organ is a function of the product of exposure time and concentration for agents with a long half-life. A similar but more elaborate approach has been applied by Rappaport (1985). He showed that intra-day variability in exposure has a limited influence when dealing with agents with long half-lives. He introduced the term dampening at the receptor.
The information presented above has mainly been used to draw conclusions on appropriate averaging times for exposure measurements for compliance purposes. Since Roach’s papers it is common knowledge that for irritants, grab samples with short averaging times have to be taken, while for agents with long half-lives, such as asbestos, long-term average of cumulative exposure has to be approximated. One should however realize that the dichotomization into grab sample strategies and eight-hour time average exposure strategies as adopted in many countries for compliance purposes is an extremely crude translation of the biological principles discussed above.
An example of improving an exposure assessment strategy based on pharmocokinetic principles in epidemiology can be found in a paper of Wegman et al. (1992). They applied an interesting exposure assessment strategy by using continuous monitoring devices to measure personal dust exposure peak levels and relating these to acute reversible respiratory symptoms occurring every 15 minutes.A conceptual problem in this kind of study, extensively discussed in their paper, is the definition of a health-relevant peak exposure. The definition of a peak will, again, depend on biological considerations. Rappaport (1991) gives two requirements for peak exposures to be of aetiological relevance in the disease process: (1) the agent is eliminated rapidly from the body and (2) there is a non-linear rate of biological damage during a peak exposure. Non-linear rates of biological damage may be related to changes in uptake, which in turn are related to exposure levels, host susceptibility, synergy with other exposures, involvement of other disease mechanisms at higher exposures or threshold levels for disease processes.
These examples also show that pharmacokinetic approaches can lead elsewhere than to dose estimates. The results of pharmacokinetic modelling can also be used to explore the biological relevance of existing indices of exposure and to design new health-relevant exposure assessment strategies.
Pharmacokinetic modelling of the exposure may also generate estimates of the actual dose at the target organ. For instance in the case of ozone, an acute irritant gas, models have been developed which predict the tissue concentration in the airways as a function of the average ozone concentration in the airspace of the lung at a certain distance from the trachea, the radius of the airways, the average air velocity, the effective dispersion, and the ozone flux from air to lung surface (Menzel 1987; Miller and Overton 1989). Such models can be used to predict ozone dose in a particular region of the airways, dependent on environmental ozone concentrations and breathing patterns.
In most cases estimates of target dose are based on information on the exposure pattern over time, job history and pharmacokinetic information on uptake, distribution, elimination and transformation of the agent. The whole process can be described by a set of equations which can be mathematically solved. Often information on pharmacokinetic parameters is not available for humans, and parameter estimates based on animal experiments have to be used. There are several examples by now of the use of pharmacokinetic modelling of exposure in order to generate dose estimates. The first references to modelling of exposure data into dose estimates in the literature go back to the paper of Jahr (1974).
Although dose estimates have generally not been validated and have found limited application in epidemiological studies, the new generation of exposure or dose indices is expected to result in optimal exposure-response analyses in epidemiological studies (Smith 1985, 1987). A problem not yet tackled in pharmacokinetic modelling is that large interspecies differences exist in kinetics of toxic agents, and therefore effects of intra-individual variation in pharmacokinetic parameters are of interest (Droz 1992).
Biomonitoring and Biomarkers of Exposure
Biological monitoring offers an estimate of dose and therefore is often considered superior to environmental monitoring. However, the intra-individual variability of biomonitoring indices can be considerable. In order to derive an acceptable estimate of a worker’s dose, repeated measurements have to be taken, and sometimes the measurement effort can become larger than for environmental monitoring.
This is illustrated by an interesting study on workers producing boats made of plastic reinforced with glass fibre (Rappaport et al. 1995). The variability of styrene exposure was assessed by measuring styrene in air repeatedly. Styrene in exhaled air of exposed workers was monitored, as well as sister chromatid exchanges (SCEs). They showed that an epidemiological study using styrene in the air as a measure of exposure would be more efficient, in terms of numbers of measurements required, than a study using the other indices of exposure. For styrene in air three repeats were required to estimate the long-term average exposure with a given precision. For styrene in exhaled air, four repeats per worker were necessary, while for the SCEs 20 repeats were necessary. The explanation for this observation is the signal-to-noise ratio, determined by the day-to-day and between-worker variability in exposure, which was more favourable for styrene in air than for the two biomarkers of exposure. Thus, although the biological relevance of a certain exposure surrogate might be optimal, the performance in an exposure-response analysis can still be poor because of a limited signal-to-noise ratio, leading to misclassification error.
Droz (1991) applied pharmacokinetic modelling to study advantages of exposure assessment strategies based on air sampling compared to biomonitoring strategies dependent on the half-life of the agent. He showed that biological monitoring is also greatly affected by biological variability, which is not related to variability of the toxicological test. He suggested that no statistical advantage exists in using biological indicators when the half-life of the agent considered is smaller than about ten hours.
Although one might tend to decide to measure the environmental exposure instead of a biological indicator of an effect because of variability in the variable measured, additional arguments can be found for choosing a biomarker, even when this would lead to a greater measurement effort, such as when a considerable dermal exposure is present. For agents like pesticides and some organic solvents, dermal exposure can be of greater relevance that the exposure through the air. A biomarker of exposure would include this route of exposure, while measuring of dermal exposure is complex and results are not easily interpretable (Boleij et al. 1995). Early studies among agricultural workers using “pads” to assess dermal exposure showed remarkable distributions of pesticides over the body surface, depending on the tasks of the worker. However, because little information is available on skin uptake, exposure profiles cannot yet be used to estimate an internal dose.
Biomarkers can also have considerable advantages in cancer epidemiology. When a biomarker is an early marker of the effect, its use could result in reduction of the follow-up period. Although validation studies are required, biomarkers of exposure or individual susceptibility could result in more powerful epidemiological studies and more precise risk estimates.
Time Window Analysis
Parallel to the development of pharmacokinetic modelling, epidemiologists have explored new approaches in the data analysis phase such as “time frame analysis” to relate relevant exposure periods to endpoints, and to implement effects of temporal patterns in the exposure or peak exposures in occupational cancer epidemiology (Checkoway and Rice 1992). Conceptually this technique is related to pharmacokinetic modelling since the relationship between exposure and outcome is optimized by putting weights on different exposure periods, exposure patterns and exposure levels. In pharmacokinetic modelling these weights are believed to have a physiological meaning and are estimated beforehand. In time frame analysis the weights are estimated from the data on the basis of statistical criteria. Examples of this approach are given by Hodgson and Jones (1990), who analysed the relationship between radon gas exposure and lung cancer in a cohort of UK tin miners, and by Seixas, Robins and Becker (1993), who analysed the relationship between dust exposure and respiratory health in a cohort of US coal miners. A very interesting study underlining the relevance of time window analysis is the one by Peto et al. (1982).
They showed that mesothelioma death rates appeared to be proportional to some function of time since first exposure and cumulative exposure in a cohort of insulation workers. Time since first exposure was of particular relevance because this variable was an approximation of the time required for a fibre to migrate from its place of deposition in the lungs to the pleura. This example shows how kinetics of deposition and migration determine the risk function to a large extent. A potential problem with time frame analysis is that it requires detailed information on exposure periods and exposure levels, which hampers its application in many studies of chronic disease outcomes.
In conclusion, the underlying principles of pharmacokinetic modelling and time frame or time window analysis are widely recognized. Knowledge in this area has mainly been used to develop exposure assessment strategies. More elaborate use of these approaches, however, requires a considerable research effort and has to be developed. The number of applications is therefore still limited. Relatively simple applications, such as the development of more optimal exposure assessment strategies dependent on the endpoint, have found wider use. An important issue in the development of biomarkers of exposure or effect is validation of these indices. It is often assumed that a measurable biomarker can predict health risk better than traditional methods. However, unfortunately, very few validation studies substantiate this assumption.