Exposure to carcinogenic polycyclic aromatic compounds and health risk assessment for diesel-exhaust exposed workers

ABSTRACT

Objectives: Workers’ exposure to diesel exhaust in a bus depot, a truck repair workshop and an underground tunnel was determined by the measuring of elemental carbon (EC) and 15 carcinogenic polycyclic aromatic compounds (PACs) proposed by the US Department of Health and Human Services/National Toxicology Program (NTP). Based on these concentration data, the genotoxic PAC contribution to the diesel- exhaust particle (DEP) lung-cancer risk was calculated.

Method: Respirable particulate matter was collected during the summer and winter of 2001 (except for in the underground situation) and analysed by coulometry for EC and by GC–MS methods for PACs. The use of potency equivalence factors (PEFs) allowed the studied PAC concentrations to be expressed as benzo[a]pyrene equivalents (B[a]Peq). We then calculated the lung-cancer risk due to PACs and DEPs by multiplying the B[a]Peq and EC concentrations by the corresponding unit risk factor. The ratio of these two risks values has been considered as an estimate of the genotoxic contribution to the DEP cancer risk.

Results: For the bus depot and truck repair workshop, exposure to EC and PACs has been shown to increase by three to six times and ten times, respectively, during winter compared to summer. This increase has been attributed mainly to a decrease in ventilation during the cold. With the PEF approach, the B[a]Peq concentration is five-times higher than if only benzo[a]pyrene (B[a]P) is considered. Dibenzopyrenes contribute an important part to this increase. A simple calculation based on unit risk factors indicates that the studied PAC contribution to the total lung-cancer risk attributed to DEPs is in the range of 3-13%.

Conclusions: The 15 NTP PACs represent a small but non-negligible part of lung-cancer risk with regard to diesel exposure. From this point of view, the dibenzopyrene family are important compounds to be considered.

METHODS

Description of workplaces and control strategies

Bus depot: the depot is located inside the urban centre of Lausanne City (Switzerland). It consists of a big hall (about 65,000 m3) in which buses (MAN SL 2000 and Van Hool models) are usually stationed overnight and are driven out at dawn during a period known as the ‘‘run out’’. Some vehicles return to the depot (‘‘run in’’) during the day, depending on the traffic needs; the majority is back in the early–late evening. Activities such as bus maintenance (cleaning, tanking) and some small repairs are done in this depot. The diesel engines may be running within the depot during maintenance and also when the buses are started prior to leaving. There is a tendency for some drivers to start a vehicle and let it run to warm up the engine before run out. There are three distinct groups of employees who may be exposed to DEPs: cleaners, who spend their full working day within the depot, guards and engineering staff, who spend roughly half their time in the depot and, finally, the drivers who come just to warm up and drive the buses. Natural ventilation is maintained by the doors being kept open during all the working day, mainly during the warm season. A forced ventilation system is available, but is used mainly in the mornings during the first run out in winter time. The technical specification of this system corresponds to an air change rate of 2 h-11.

Truck repair workshop: the workplace studied is located in a suburb of Lausanne and consists of a big hall (about 50,000 m3) in which the main activities are truck/semi-trailer maintenance and repair (motors, tyres, electrical circuits), cleaning, and semi-trailers being prepared before they leave. Fork-lift trucks are in use to load and unload goods from lorries and to stock spare parts in the storehouse. Exposure to diesel exhaust is due to the run in–run out of the different vehicles, to displacement inside the hall, and when the repaired motors are tested. Exposed workers inside the workshop are mainly mechanics and, to a lesser extent, truck drivers. Flexible ducts attached to the tailpipe are available at some sites, mainly in the mechanics part. Otherwise, only natural ventilation is achieved, by personnel opening the doors that run along the full length of the workshop, and via some cupolas on the roof. In winter time, such ventilation seldom operates so as to prevent a low temperature in the hall. This lack of ventilation during winter is illustrated by the decrease in air changes rate, which falls from approximately 6 h-1 in summer to 2 h-11 in winter, based on the signal decay recorded by the direct reading PAS 2000 (see below, real-time measurements). These values were obtained by plotting the logarithm of a peak-signal decay vs time. A straight line is obtained, whose slope corresponds to the number of air changes.

Underground tunnel: the underground workplace that was sampled corresponds to the digging of a new tunnel in limestone with a low silica content. Conventional techniques, such as explosive attack followed by removal of the rubble by digger-filled dumpers, are used. The limestone face is then cleaned with a diesel-powered power pick and metallic structures are placed to consolidate the arch. Concrete is then sprayed onto the arch to secure the part that has been dug. Besides rock dust, workers are exposed to DEPs generated by the heavy-duty engines for almost the whole of the shift. Forced ventilation brings fresh air from the outside of the tunnel to the limestone face at a flow rate of 1.5–4 m3/min. Some activities, such as extension of the ventilation tubing, imply a temporary stop of such a system.

Sampling strategies

Season has been shown to play an important role in diesel exposure of garage workers and mechanics . For the bus depot, two campaigns of 2 consecutive days were thus organised, in summer (18–19 of June and 16–17 of July 2001; mean daily temperature 17–23 C) and in winter (3–4 and 17–18 of December 2001; mean daily temperature -6 to 15 C). For the truck repair workshop, two campaigns of 2 consecutive days were carried out during the summer season (25–26 of June and 30–31 of July 2001; mean daily temperature 24–28 C), whereas only one campaign was achieved during the winter season (9–10 of January 2002; mean daily temperature 11–18 C). For the tunnelling site, no seasonal variation was expected; air was thus sampled during two summer campaigns (14 and 22 of August 2001). In this case, samples were taken during various activities: explosion and rubble removal by digger-filled dumpers, limestone face cleaning, and concrete sqraying to secure the arch.

All samples were collected for either 1 or 2 full working days (corresponding to one or two shifts, respectively), depending on the sampling device and the season.

Fixed sampling for the analysis of EC and PACs was performed at a height of between 1 and 1.5 m. Personal sampling was not possible for PAC analysis, due to sensitivity limitations of the analytical method. Two sampling places were chosen during the first campaign in the bus depot and in the truck repair workshop so that the homogeneity of the PAC concentration could be assessed. As no significant differences were observed in the PAC concentrations, air sampling took place only in the vicinity of the main working place, during the subsequent campaigns, so that it was close to the workers’ exposure.

Real-time measurements

A real-time aerosol monitor (miniRAM Model PDM-3, MIE, Billerica) that worked on the light-scattering principle was used to record the profile of particles smaller than 10 lm during the workday. It samples air passively and was connected to a data logger (Eltek SQ-8, Eltek, Cambridge). In parallel, a real-time PAH sensor (PAS 2000, EcoChem, Uberlingen), operating at a 222 nm wavelength and connected to a data logger (Hotbox BV 2, Elpro, Buch) was used. The PAS 2000 gives a signal which is a function of the amount of PAH adsorbed on particles; its response has also been shown to be proportional to the EC content of diesel particulates and also correlates well with the bacterial genotoxicity of air-particle extracts.


Dust sampling and characterisation

As air particles found in the selected working conditions are not only DEPs, the particle size distribution during each sampling campaign was determined by an Andersen cascade impactor (Model 1 ACFM, 9 stages, Andersen, Atlanta, Ga., USA) connected to a sampling pump operating at 28 l/min. The concentrations of the nine fractions were determined by gravimetric measurements of the loaded filters on an analytical balance (Mettler AT-201, Mettler–Toledo, Greifensee; sensitivity ±10 lg). All the filters were weighed after an equilibration period (>24 h) at ambient temperature and stable relative humidity (50% RH, achieved with a saturated solution of Ca(NO3)2 in a glove box).

Total suspended particles and respirable dust concentrations (<4 lm) were determined gravimetrically on the loaded filters by, respectively, high-volume pump and cyclone size selectors (Casella and PM4 type pump) as described below.

EC sampling and analysis

Organic carbon (OC) and EC concentrations were determined in the respirable fraction of the dust. Sampling was done with a size-selector cyclone (Casella, London) connected to a personal pump (Du Pont P2500, Du Pont, Wilmington, Del., USA) set at 1.9 l/min. The airflow was checked at least twice during the day with a calibrated flow meter (DryCal-DC lite, SKC, Eighty Four, Pa., USA). Only pre-cleaned quartz microfibre filters (Whatman QMA, Ø 37 mm, 2.2-lm pore size) were used. Cleaning was achieved by plasma ashing (PlasmaPrep II, SPI, West Chester, Pa., USA) for 3 h at 450 W under 1 mbar O2. The filters were weighed on a microbalance (Mettler M-3, Mettler–Toledo; sensitivity±1 lg), and samples were stored at -20 C until required for analysis within the next 2 months.

OC–EC analysis was achieved on the whole filter by use of a thermal desorption system coupled to a coulometric titration device, described elsewhere (Perret et al. 1999). Briefly, after the filter has been introduced into the heating part of a coulomat (Strohlein 720 DR/C, Dusseldorf, Germany), the system is purged and allowed to stabilise under a nitrogen atmosphere for 3 min. The OC determination is performed by a fast temperature increase to 800 C. This temperature is stepped until no change in electrolysis current is observed for 2 min. The corresponding total electrolysis current corresponds to the OC content of the filter. In a third step that allows EC determination, the atmosphere is switched to an oxygen, the temperature is held at 800 C and the current zeroed. The EC content corresponds to the amount of electricity needed when no more counts are observed on the current meter for 2 min. The samples from the tunnelling site were treated with HCl (200 ll HCl 1.5%, evenly distributed on the filter and left for a minimum of 2 h at ambient temperature) because of the possible presence of interfering carbonate particles. Blank filters were analysed for each series, and quality control was achieved by re-analysis of a diesel-loaded filter used during an inter-laboratory round-robin test within the framework of CEN/TC 137 (filter QF-708). The limit of detection (LOD) for EC with this method is in the range 2–5 lg/m3.

PAC sampling and analysis

A Gravikon PM-4 (GSM, Neuss-Norf) with a size-selector cyclone head was used for respirable-particle sampling. The flow was maintained at a constant 66.7 l/min, and glass fibre filters (Macherey–Nagel MN 85/90, Ø 70 mm, retained particles >0.5 lm), which were cleaned as described for EC analysis, were used. As mentioned before, during one summer campaign, we used two high-volume pumps (Model 353, Sierra, Carmel Valley, Calif, USA; flow set at 230 l/min) to assess the PAC homogeneity in the bus depot and truck repair workshop. The quartz filters that we used (Whatman QMA, 20·25 cm) were cleaned by ultrasonication in methanol (30 min), dichloromethane (30 min) and, finally, toluene (30 min). The cleaned filters were dried at 150 C and kept separately in aluminium foil until required for use. Samples were stored at )20 C until required for analysis.

PAC analysis was achieved on the total filter by a method described by Sauvain et al. (2001). Briefly, after each sample has been spiked with deuterated internal standards, the filter is Soxhlet extracted with toluene for 24 h. The extract is then concentrated, purified through a 10% deactivated SPE silica cartridge with dichloromethane:acetone 39:1. A semi-preparative HPLC fractionation on a silica column is further performed, and two fractions containing PAH and polycyclic aromatic nitrogen heterocyclic compounds (PANH), respectively, are obtained. A liquid–liquid partition is achieved on the PAH fraction and the PANH fraction is chromatographed on a polyvinylbenzene co-polymer column. The purified extracts are concentrated and finally injected on a Varian 3800 gas chromatograph coupled to a Varian 4D MS ion trap detector. On-column injection (1 ll) was performed on a 30 m·0.25 mm·0.25 lm BPX-50 column (50% phenyl equivalent polysilphenylensiloxane, SGE, Weiterstadt). Certified diesel particulate matter SRM 1650 and field blanks were analysed with each sample series. No significant contamination was observed with the blank analysis.

Assessment of PAC contribution to the total DEP lung-cancer risk 

One can assess the theoretical occupational lung-cancer risk due to the inhalation of DEPs or B[a]P by multiplying the DEP or B[a]P concentration by its corresponding unit risk factor. Such a calculation has been achieved for miners exposed to DEPs by Stayner et al. (1998). The unit risk is defined in our case as the risk corresponding to an occupational continuous exposure(assumed to be for 45 years, 8 h per day), to 1 lg/m3o f DEPs or B[a]P

For carcinogenic compounds, a non-threshold dose–response curve is generally considered. In order for this dose–response curve to be extrapolated in these low concentrations, two kinds of models are used: statistical or mechanistic. The unit risks used in this study are mainly obtained from this last model, based on the so-called linearised multistage model. The unit risk corresponds to the slope of such an extrapolated line, when it is linear. Unit risks given in the literature are often representative of a lifetime exposure. Since we are interested in the occupational situation, and the occupational exposure is shorter than an entire lifetime, we must, of necessity, make an adjustment to this unit risk factor. We have thus applied a multiplicative factor of 0.21 to the lifetime unit risk value, as described by . Whereas EC represents only a small part of the ill-defined DEPs, it has been retained as surrogate for DEPs in this study. This implies that the DEP unit risk factor has to be further corrected in order for this fact to be taken into account. One will thus calculate the EC-corrected DEP unit risk factor by dividing the DEP unit risk factor by a DEP/EC ratio of 2.5. This ratio was used by and is similar to the median value of 2.6 reported by for an underground mining environment. Table 2 summarises the unit risk factors obtained in this study, corrected when necessary to be representative for occupational exposure, and expressed for EC instead of DEPs. They are based mainly on values presented by Scheepers and Bos (1992) and Bostrom et al. (2002).

As the carcinogenicity of DEPs is assumed to be due to a non-genotoxic pathway (particulate core) and a genotoxic pathway (adsorbed organics, mainly PACs), the total occupational lung-cancer risk, determined by the EC-corrected DEP unit risk factor times the EC concentration, can be estimated in a first approximation as the contribution of both pathways. Even if the interactions between these two pathways are difficult to assess, the main hypotheses are that each pathway contributes in an additive way to the total cancer risk, and that the PAC concentrations that are determined in the air result from diesel emissions only. For the genotoxic pathway, as a PAC mixture is involved, the potency equivalence factor (PEF) scheme is used to calculate the B[a]Peq concentration, expressed in microgrammes per cubic metre. This B[a]Peq concentration time the B[a]P unit risk factor gives an estimation of the lung-cancer risk attributable to the PACs considered. One can thus evaluate the contribution of this genotoxic risk to the lung-cancer risk by dividing the PAC lung risk by the total lung risk based on EC measurement. In this calculation, we neglected the PAH contribution from the gas phase; the contribution of the volatile PAHs with two and three benzene rings has been shown to be biologically inactive when implanted into the lungs of rats (Grimmer et al. 1987).

Statistics

Some non-parametric tests, such as the Wilcoxon signed rank test and the Wilcoxon rank sum test, have been used to assess the difference between the selected working situations. These tests have been chosen due to the fact that the normality distribution of our small number of measurements cannot be confirmed.

RESULTS

Direct-reading measurements and particle-size characteristics 

Except for the tunnelling conditions, total dust given by the direct-reading miniRam did not provide useful information. This instrument is probably not sensitive enough to detect the low concentrations and rather small particle sizes in the bus depot and truck workshop. Figure 1 presents an example of the recorded signal from the real-time PAS instrument in the bus depot. A fairly good association is observed between this signal and the total numbers of vehicles leaving or entering the depot. Such a correlation is less marked for the truck workshop (data not shown), probably due to a response more influenced by local activities (displacement of a truck, air flow and so on). The mean value of the PAS signal is smaller in summer than in winter in the case of the bus depot (as shown in Fig. 1) and for the truck workshop. In summer, the PAS signal may return to a near-zero value, corresponding to a situation where PAC concentrations are low. In winter, such a situation seldom occurs; as soon as the first vehicles left or entered the bus depot or the mechanics’ workshop, the PAS signal did not reach a baseline value anymore.

The use of an Andersen impactor allowed us to determine the size distribution during the different sampling campaigns. As no difference was observed between summer and winter distribution (Wilcoxon signed rank test, P<0.05), the contribution of each mean-size fraction to the total dust was calculated from these two sets of seasonal data and is presented in Fig. 2. The particle size distribution follows a log-normal law for the bus depot and the truck repair workshop. We verified this distribution by obtaining a straight line when the cumulative percent particulate of each impactor stage vs the corresponding cutoff were plotted on log-probability paper. The determination of the geometric mean size was based on the log-probability plot of the particle size. For the tunnelling conditions, where a bimodal distribution was observed, we determined the geometric mean size by following the procedure described by Knutson and Lioy (1989; Table 3). 

Diesel aerosol is characterised by very small particles, with a mass median aerodynamic diameter of 0.2 lm and 90% of the particles being less than 1 lm (Watts 1995). Thus, a respirable-aerosol sampling device such as a cyclone collects all diesel and non-diesel aerosol particles (oil mist, pollen, cigarette smoke and so on) that fall into the respirable size range. As illustrated in Fig. 2 and Table 3, each working condition differs, clearly based on the particle size distribution. Based on the Andersen impactor results, the bus depot is characterised by a high amount of respirable particles that reach approximately
75% of the total sampled mass, with a diesel contribution of around 40% (corresponding to particles smaller than
1.1 lm). The truck repair workshop presents a rather constant distribution of particle size, with diesel particles
representing approximately 25% of the total dust mass. As expected, the coarse particulate matter (>7 lm, probably mineral dust) is in the majority, in term of mass, in underground mine air compared with respirable (20%) or diesel particulates (4%).

The mean seasonal air concentrations for total suspended particles (TSPs), determined with the high-volume pump during the summer sampling campaign, and for the respirable fraction, determined with cyclone head, are presented in Table 3. The percentage of particles smaller than 4.7 lm that were obtained with the Andersen impactor corresponds rather well with the concentration ratio of particles sampled with cyclone head to the TSPs. 

An increase in the respirable-dust concentration between summer and winter for the two transport-industry environments (bus depot and truck repair workshop) is suggested by Table 3; it is significant (Wilcoxon rank sum test, P<0.05) only for the truck repair workshop. 

Measurement of OC and EC concentrations

Table 4 presents the mean concentrations for OC and EC content of the different analysed samples. The total carbon (TC) corresponds to the sum of OC and EC concentrations. A statistically significant increase of EC and OC air concentration (Wilcoxon rank sum test, P<0.05) is observed between summer and winter in the transport-industry environment. Such an increase between these two seasons is also statistically significant for the EC/TC ratio (Wilcoxon rank sum test, P<0.05).

Measurement of PAC concentrations

Table 5 presents the median concentrations determined for carcinogenic PACs adsorbed on the dust particles sampled in the three working environments. Only detected compounds are presented in this table and their sum outlined in the column "P15 PACs". The air volume that was sampled with the high-volume pump for 2 consecutive days ranged between 1,730 and 2,000 m 3 and, for the PM4 pump, ranged between 87 and 100 m 3 . For the tunnelling conditions, the sampled volume was smaller, between 2.9 and 10.3 m 3 , depending on the mining activities. No high-volume sampling was achieved in this working environment, due to the high particle concentration in the air and the risk of filter clogging. Detection limits presented in Table 5 have been calculated based on the analyte mass that gives a mean signal 3r above the mean sample baseline signal (r corresponds to the standard deviation of the baseline signal).

Based on these concentrations, the relative contribution of each compound to the P15 PACs, expressed in percent, has also been calculated (Fig. 3). We carried out statistical tests on these data mainly to evaluate if differences exist in the PAC distribution profile for the three working environments that were surveyed. The Wilcoxon signed rank test indicated no difference (P<0.05) between the carcinogenic PAC profile in the bus depot and truck repair workshop, either according to season (summer or winter) or sampling-pump type (total suspended or respirable particles). All the distribution values corresponding to this working branch have thus been grouped and are presented in Fig. 3A. To test whether the individual PAC median value for the
transport-industry distribution and the underground mine distribution (Fig. 3B) are different, we used the Wilcoxon rank sum test. A statistically significant difference (P<0.05) was observed between these medians for all the detected PACs, except for dibenz[a,h]-anthracene (DB[a,h]A), dibenzo[a,l]pyrene (DB[a,l]P), dibenzo[a,e]pyrene (DB[a,e]P) and dibenzo[a,i]pyrene (DB[a,i]P). The percentages of benz[a]anthracene (B[a]A) and 5-methyl chrysene (5-MeC) are higher for the underground mine, whereas the percentages for the three benzofluoranthenes, B[a]P and indeno[1,2,3- c,d]pyrene (In[c,d]P) are smaller than for the transport branch profile.

Assessment of PAC contribution to the total DEP lung-cancer risk

The PEF scheme, briefly described in the introduction, has been applied to the obtained PAC concentrations. One can calculate an individual B[a]P eq by multiplying each compound concentration with the corresponding PEF value given in Table 1. Summation of all these B[a]P eq concentrations (Table 5, "P15 PAC eq ") gives an evaluation of the carcinogenic potency of all the
quantified PACs and allows us to compare this with the one due to only B[a]P. Table 5 indicates that if this is taken into account, these biologically active PAHs increase the B[a]Peq concentration by a mean factor of 5±2.

A predicted total lung-cancer risk and PAC lung-cancer risk among bus-depot workers and mechanics exposed to DEPs can be calculated by analogy to the procedure described by Stayner et al. (1998) and Tsai et al. (2001) and is presented in Table 6. The estimated contribution of the 15 NTP PACs to the total DEP lung-cancer risk corresponds to the ratio PAC risk/DEP risk
and is between 3% and 13%. This value indicates that the PAC concentrations may contribute, in a relatively small but non-negligible part, to the total lung-cancer risk attributable to the diesel emissions.
