1. Institute of Occupational Medicine, 8 Roxburgh Place, Edinburgh EH8 9SU, United Kingdom.

2. University of Aberdeen, Department of Environmental and Occupational Medicine, Foresterhill Road, Aberdeen AB25 2ZP, United Kingdom.

A reliable reconstruction of past exposure to diesel exhaust particulate is an essential part of estimating the possible cancer risks. This paper describes a theoretical model of the exposure process and a limited investigation of the validity of the estimates derived from the model. We conclude that this approach could be used in a retrospective epidemiological study to assess risk.

In 1989 the International Agency for Research on Cancer (IARC) classified exposure to diesel engine emissions as probably carcinogenic to humans. This conclusion was based there being sufficient evidence of carcinogenicity in experimental animals and limited evidence of carcinogenicity in humans (IARC, 1989). In the period since this decision there have been a number of additional epidemiological studies published that provide further information on cancer risk for people exposed to diesel emissions. A recent meta-analysis has shown an overall relative risk of lung cancer of 1.33 (95 percent confidence interval 1.21-1.46) amongst exposed workers (Lipsett & Campleman, 1999).

However, a quantitative assessment of the risk, in particular of lung cancer, cannot be confidently undertaken on the basis of the available studies. The main reason being the absence of reliable information on lifetime exposure to diesel exhaust emissions. There is therefore the need for further large-scale retrospective epidemiological studies, coupled with a quantitative assessment of past exposures. The United States Health Effects Institute (HEI) funded a number of feasibility studies to investigate whether it was possible to identify cohorts where such studies could be undertaken. This investigation was undertaken as part of one of the feasibility studies funded by the HEI. The work was coordinated by IARC and the aim was to assess the feasibility of enrolling a multicentre historical cohort of workers exposed to diesel engine emissions in the Czech Republic, Estonia, Hungary, Latvia, Lithuania, Poland, Russia, Slovakia, and Slovenia and, in addition, to assess whether past exposure could be reconstructed for the cohort. In this paper we report the work to develop a method to retrospectively assess exposure to diesel exhaust particulate.

At the end of sampling, the filter cassette was removed from the sampling head and the size selective foam removed and discarded. The cassettes were then analysed gravimetrically to determine the thoracic dust concentrations. Each filter sample was analysed for organic carbon and elemental carbon using Evolved Gas Analysis with Thermal Optical Sensor in accordance with NIOSH method 5040 (NIOSH, 1996).

It was not practical to use the thoracic dust samplers in the oil-shale mine and cyclone type respirable samplers were used instead. These samplers were prepared for gravimetric determination but losses from the edges of the filter media due to the brittle nature of the quartz fiber filters prevented this from being reported. However, the samples were analysed for elemental and organic carbon as described above and the filter losses did not affect the results of the elemental carbon analysis.

We have developed a theoretical description of the exposure process from which we plan to reconstruct past exposures in a full-scale epidemiological study. This builds upon work for an earlier epidemiological study investigating the risks of lung cancer from man-made vitreous fibers and other research currently underway (Cherrie et al, 1996; Cherrie, 1999; Cherrie & Schneider, 1999).

In this scheme we use elemental carbon as a marker for diesel exhaust particulate. The exposure level is defined as the average concentration of elemental carbon inhaled by a worker during a task or some other defined period of time. It is assumed that the exposure level (C) is the sum of contributions from: a multiplicity of background sources (C_B); other more local sources in the far-field (C_FF); and sources close to the workers (C_NF), that is in his or her near field.

An arbitrary distinction is made between the workers near-field, which is defined as a cubic space with 2 m sides centered on the subject's head, and the far-field, which is the remainder of the immediate environment. For indoor spaces the far-field is defined by the boundary walls of the space, while outdoors the far-field is defined as being within 10 m of the subject. In modeling exposure we have normally assumed that emissions are dispersed uniformly from the source in all directions. However, in the case of diesel emissions this is not the case because of the high initial velocity of the exhaust gasses.

The background concentration of elemental carbon arises from diesel engines and other sources in the general urban environment. It will be dependent on many factors, including the number of vehicles actively emitting, the types of engine, meteorological conditions, etc. It is likely that the background concentrations are well represented by fixed location monitors situated some distance away from localized sources, such as busy roads, and that at any one time they are relatively constant across a town or city.

For diesel exposure the far-field contribution to exposure may come from the exhaust plume of the vehicle being driven by the individual or may come from other vehicles in the general vicinity.

For a source in the far-field of the worker similar considerations apply and so the far-field contribution to the exposure level is given by the following equation:

The dispersion of the pollutant is dependent on the proximity of the source to the person exposed, that is whether the source is in their near- or far-field, and on the directional nature of the initial dispersion. In addition, if the source is inside a building, then the volume of the enclosed space and the quantity of general ventilation will also determine exposure level.

For outdoor diesel sources in the far-field we have assumed a Gaussian type dispersion pattern. This allows estimation of the reduction in concentration to be made along with an adjustment of the "time source active" term to allow for the variation in the plume trajectory because of changes in wind direction. For example, it is assumed that moving away from the source will reduce the particulate concentration approximately according to the inverse square of the distance from the source. In addition, we assume that the direction of the exhaust plume may vary according to the wind direction with the probability that it might blow from the source towards the worker being reflected in the model in the time source active term. If there was no preferred wind direction in relation to the line between the source and worker, then it has been assumed that t_a,FF would be best estimated as the angle of expansion.

We have used the model to estimate the average exposure level for the five groups of workers where measurements of exposure were undertaken. Estimates were made as elemental carbon, with no adjustment made to estimate eight-hour time-weighted average level. The exposure assessments were carried out by one of us (JWC), who was unaware of the measurement data at the time of the assessments. Information about the circumstances of the exposure was obtained partly by observation of the work activities (train drivers in Russia) or from descriptions provided by one of the other investigators who had visited the sites, and partly form the questionnaire information or other descriptive information acquired during the study.

Little objective information was available to determine the magnitude of the parameters that should be input to the model and so most were assigned based on the judgement of the assessor. Since many of the samples were obtained on drivers who, because of the nature of their job, were for most of the sampling duration away from the investigators, it was not possible to have detailed contextual information about the circumstances of the exposure. For this reason the exposure assessments were completed for all of the broad job categories identified, rather than for each individual measurement.

The estimated exposure level for the bus mechanics was 43 µg/m³ and the mean measured exposure level was 39 µg/m³. The main source of exposure for this group was judged to be the exhaust from busses moving around the garage and we assumed that there were buses moving for 8 percent of the work shift. Drivers were assumed to be exposed for a longer period of time but their exposure was lower because we assume that the cab reduces the exhaust particulate entering the drivers near-field. The estimated exposure level for driver was 5 µg/m³ and the average measured exposure level was 9.5 µg/m³.

The final group of workers was the shale miners. The estimated exposure level for this group was 500 µg/m³ with the average measured exposure being 220 µg/m³. The main reason for this group having the highest exposures was the relatively continuous work in a confined space. Part of the reason for the overestimation in this case may have been the difficulty in judging the importance of the high level of general ventilation in controlling exposure. It is possible that this was underestimated. We used the data for large, well-ventilated spaces, shown in Table 16, although the actual ventilation rates were much higher than this.

The model provides a scheme for describing the exposure in a way that can help in estimating past exposure. The estimates are, however, clearly dependent on the quality of information about the work and the work environment that is available. Information about the work processes can be obtained from records and by interview with long-service employees or retired workers. This information may then provide the basis for estimating exposure levels in the past.

Combination of measurements of present day exposure levels with the exposure reconstruction method provides an opportunity to minimize any bias in the estimates and to refine the magnitude of the component factors in the theoretical model. Further refinement of the model parameters may also be obtained from further experimental investigations.

Estimates of historical background concentrations of diesel exhaust particulate could be obtained from information about diesel traffic density, obtained either from historical records or interviews with knowledgeable local people. This information along with the theoretical model outlined here could be used to reconstruct past exposure for groups of workers exposed to diesel exhaust particulate.