QUANTITATIVE DETERMINATION OF TRUCKING INDUSTRY WORKERS' EXPOSURES TO DIESEL EXHAUST PARTICLES

ABSTRACT

As part of a case-control mortality study of trucking industry workers, exposures to diesel aerosol were measured among th efour major presumably exposed job groups (road drivers, local drivers, dock workers, and mechanics) in the industry. Eight industrial hygiene surveys were conducted during both warm and cold weather at eight U.S. terminals and truck repair shops. A single-stage personal impactor was used to sample submicrometer-sized diesel particles on quartz fiber filters. Laboratory and field studies demonstrated that the elemental carbon content of the particles is a useful and practical marker of exposure to vehicular diesel exhaust. A thermal-optical analysis technique was used to determine the concentration of elemental carbon in the filter samples. Over­all geometric mean exposures to submicrometer-sized ele­mental carbon ranged from 3.8 ug/m3 in road (long distance) drivers (N = 72) to 13.8 ug/m3  in dock workers (N = 75). Geometric mean background area concentrations, measured in the same cities where workers were sampled, were 2.5 ug/m3  on major highways (N = 21) and 1.1 ug/m3 in residen­tial areas (N = 23). A factorial analysis of variance indicated that exposures in two job groups, dock workers (particularly those exposed primarily via diesel fork lift trucks, introduced relatively recently) and mechanics (working in poorly venti­lated shops during cold weather), were significantly higher than background concentrations and were significantly higher than the exposures in the local and road drivers. The exposures of the truck drivers could not be distinguished from background highway concentrations but were significantly higher than background concentrations in residential areas.

EXPERIMENTAL MATERIALS AND METHODS

Diesel Emissions Studies

In order to investigate the precision and reproducibility of the thermal-optical method, a dilution tube technique was used to generate controlled atmospheres of diesel exhaust at four different concentrations. The experiment was conducted by us­ing a dynamometer and dilution tube apparatus (Figure 1) located at Ford Motor Co.’s Research and Engineering Center in Dear­born, Michigan. The exhaust was injected into the dilution tube from a subcompact automobile equipped with a light-duty diesel engine, cruising on the dynamometer at about 30 mph. Diluent air (filtered with high-efficiency particulate filters and tempered to approximately room temperature) was injected co-current with the exhaust stream, and co-planar with a mixing baffle to ensure rapid and complete mixing. The concentration of diesel exhaust was varied by changing the amount of air mixed with the diesel exhaust. A total of four concentration levels were set by varying the quantity of air injected into the tube: 1200,900,600, and 300 cfm (indicating in each case the volumetric flow rate of air plus injected exhaust moving through the dilution tube).

During each of eight runs (two at each concentration level), simultaneous sampling for total weight of particulate matter, elemental carbon, and organic carbon was conducted. The total particulate samples were obtained by drawing air at a flow rate of approximately 100 Lpm through Pallflex (Putnam, Conn.) TX40 142-mm Teflon®-backed glass fiber filters. The filter hold­ers were connected by vacuum tubing to a flowmeter, a bellows-type pump, and a dry gas meter. The elemental and organic carbon samples (for thermal-optical analyses) were obtained by drawing air at approximately 4 Lpm through 47-mm Pallflex QAOT quartz fiber filters and a stainless steel support screen, both of which were held in aluminum 47-mm in-line filter holders. The holders were again connected by vacuum tubing to a bellows-type pump, a flowmeter, and a dry gas meter. Isokinetic sampling was not considered necessary because diesel particu­lates, mainly submicrometer in size, behave much like a gas under dilution tube conditions/28.

Tobacco Smoke Studies

An experiment was also conducted in an environmentally controlled chamber to investigate the extent of possible tobacco smoke interference in the measurement of elemental carbon on quartz filters. (This experiment was conducted by NIOSH inves­tigators during an existing study administered concurrently by the John B. Pierce Foundation in New Haven, Connecticut.) During the 8-hr experiment, three men smoked a total of eight cigarettes per hour on a rotating basis. The chamber was approxi­mately 800 ft3 in volume and was ventilated at a rate of 2.5 air changes/hr (recirculation rate 20 ACH). The respirable particu­late concentrations were on the order of 1100 ug/m3. After a 2-hr equilibration period, nine area air samples for analysis of ele­mental and organic carbon were started. Three of the samples were stopped after 4 hr, as were three each at 6 and 8 hr. Three additional control samples were obtained in clean chamber air (in the absence of tobacco smoke). All samples were obtained by drawing air at a flow rate of 2 Lpm through Pallflex QAOT 37-mm quartz fiber filters, supported by stainless steel screens and held in polystyrene cassettes with the caps removed (open- faced). The samples were obtained at a flow rate of 2 Lpm (not 4 Lpm as were the field samples) because no flow-sensitive size-selective device was used and because concentrations of particulates in the chamber were predictably high.

Field Studies

Exposures to diesel aerosol were measured among the four major presumably exposed job groups identifiable from Team­sters Union records (road drivers, local drivers, dock workers, and mechanics) at each of six large, breakbulk (hub) truck terminals. Three of these surveys were conducted during cold weather, arbitrarily defined as daytime highs of less than 10°C (50°F). Three other surveys were conducted during warm weather, during which daytime highs were greater than or equal to 10°C. Limited sampling was also conducted among dock workers at one additional breakbulk terminal and among me­chanics at a small, independent truck repair shop. At the break­ bulk terminals, approximately eight personal samples, each of which was analyzed for both submicrometer-sized Cc and or­ganic carbon (Co), were obtained on each of the two shifts sampled each day.

Generally, four personal samples were ob­tained from each of two of the four jobs (e.g., four samples in dock workers and four in local truck drivers) on one of the two shifts, and an equivalent number of personal samples were obtained from each of the other two jobs (e.g., mechanics and road drivers) during the other shift sampled. At each site the sampling was conducted for three days (six shifts total, three shifts in each of the four jobs). For logistical reasons, personal sampling in road drivers and area sampling in road tractor cabs were limited to “short-turnaround” drivers; i.e., those drivers who delivered their freight to another city and returned 10-12 hr later to the originating terminal.

Except for the small repair shop, all of the sites were break­ bulk (i.e., large, hub) terminals incorporating both line-haul and city freight operations as well as dock and repair shop operations. Typically, these terminals consist of a dock building, adjacent offices, and separate buildings housing repair shops, service/ safety inspection bays, and a truck wash. The function of the dock is to receive large, long-distance loads (inbound freight) and break them down into smaller loads (outbound) for distribution to regional or local destinations. The terminal is thus referred to as a “breakbulk” terminal. The dock is a long (typically 300-400 ft by about 100 ft), open-sided steel structure situated on an elevated concrete slab. Most of the wall space in the dock area (except for the office areas) consists of open bay doors nearly the same size as the rear doors of truck trailers. “Switcher drivers” (not the truck drivers), using special switching vehicles, back truck trailers up to these doors to load and unload freight. Dock workers use forklift trucks to move freight around on the dock and into and out of truck trailers. Topically, there is no mechanically assisted ventilation on the dock, and this was the case at all seven of the facilities visited. Ventilation is mainly natural or by means of passive, dilution­ type ventilators. Air from outdoors enters the building through open doors and openings between the trailer and door.

Repair shops and service areas are usually located in one or more separate buildings on the terminal site. Each area typically consists of a single large room with overhead bay doors at one or both ends. In tractor repair shops, there is at least one center driving lane and a series of repair bays on each side. All of the tractor repair shops visited had some type of mechanical venti­lation equipment in place to remove diesel exhaust emissions from the exhaust pipes of the trucks. These consisted of either flexible ducts connected to a central duct system and exhaust fan or drive-under canopy exhaust hoods located at each repair bay.

Trailer repair shops and service/safety check lanes usually consisted of a series of long side-by-side (parallel) bays with overhead bay doors at each end. Many of these shops had no mechanically assisted ventilation systems in place, relying instead on natural ventilation and infiltration of air from the outside.
Four of the six breakbulk terminals used diesel-powered forklift trucks on the dock. One dock used propane-powered forklift trucks exclusively, and one used gasoline-powered fork­ lift trucks exclusively. One oft he sites exclusively used line-haul tractors fitted with horizontal (undercarriage) exhaust systems. The remainder of the sites used line-haul tractors fitted with vertical (stack) exhaust systems.

Worker exposures to submicrometer-sized Ce and Co were determined by obtaining full-shift personal samples with a modi­fied dichotomous sampling cassette developed by researchers at NIOSH’s Division of Respiratory Disease Studies (DRDS), but containing prefired, 37-mm Pallflex QAOT quartz fiber filters. Programmable, battery-operated personal sampling pumps were used to draw air through these cassettes at a flow rate of 4 Lpm. The modifications to the DRDS design entailed resizing the inlet diameter to approximately 0.052 in. in order to accommodate a flow rate of 4 Lpm and using quartz fiber filters (supported by stainless steel pads) instead of PVC filters.

The dichotomous cassette (Figure 2) is essentially a single- stage personal impactor, designed to collect submicrometer­ sized particles and to reject supermicrometer-sized (those larger than 1 um) particles. The dichotomous cassette was used in order to exclude, to the extent possible, nondiesel particulate matter, because almost all diesel particles (about 95%) are smaller than 1 um. An important characteristic of this device is that the particulate matter is uniformly deposited on the quartz filter, which is not necessarily true of commercially available personal cascade impactors. Uniform deposition on the filter is important because only a portion of the filter is removed for analysis by the thermal-optical method.

All of these samples were obtained for a full shift (approxi­mately 8 hr) to maximize sensitivity. The limit of detection is about 2 ug/filter, which translates to a concentration of about 1 ug/m3, assuming a 2 m3 air volume. Prefired quartz fiber filters were loaded in the dichotomous sampling cassettes just prior to each survey. Shortly after each sampling device was removed from the worker, the filter was removed from the cassette, placed in a small petri dish lined with aluminum foil, sealed in an airtight plastic pouch, and stored in a refrigerator until shipment to the laboratory. The filters were also kept cool during shipment to the laboratory by using cold packs in an insulated container, and they were kept refrigerated in the laboratory until analysis. The filters were kept cool primarily to prevent potential losses of some volatile organic carbon compounds.

Subsequent to the survey, the sample filters were submitted to a laboratory for thermal-optical quantitation of both elemental and organic carbon. The analysis used was a modification of one previously described and employs thermal program­ming and optical measurements for analysis of carbonaceous aerosols. This method overcomes limitations associated with many of the existing combustion-based methods by optically correcting for pyrolytically generated elemental carbon (“char”) generated during the first phase of the analysis. This is accom­plished by continuously measuring transmittance of a helium­ neon laser light through the filter punch in the furnace and determination of the amount of elemental carbon oxidation re­ quired to return the instrument to its initial transmittance value.

In the thermal-optical analysis, a 1- x 1.54-cm rectangular portion of the filter (i.e., a “punch”) is removed and placed in a furnace. During the first two of the three major phases of the analysis, the furnace temperature is increased (stepped) one or more times to drive off the various carbon species in stages, resulting in a carbon species profile or thermogram (plot of detector response versus time). The first phase is done in a 100% helium atmosphere, during which time the organic and inorganic (carbonate) carbon are volatilized. The oven temperature begins at room temperature and progresses through four temperature steps (250, 385, 580, and 680°C). The temperature steps are chosen such that the carbonate peak is clearly identifiable and is not confused with organic carbon species. The volatile carbon is oxidized to CO2 (in a bed of granular MnO2 at 950°C) and subsequently reduced to CH4 (in a Ni/firebrick [450°C] methnator), which is then quantitated with a flame ionization detector. The second stage of the analysis begins with a reduction in temperature to 525°C and the introduction of a 2% O2 atmo­ sphere into the furnace and proceeds with two additional tem­perature steps (to 600°C and 750°C). The CO2 formed is again reduced to CH4 and quantitated with a flame ionization detector. The peaks observed during the second stage are reported and quantitated as elemental carbon (Cc). In the third and final stage of analysis, a known amount of methane is injected for calibra­ tion and quality control.

Additional area sampling was conducted during the survey to measure concentrations of (1) respirable particulates by gravimetry with a 10-mm nylon cyclone and a 37-mm, 5-p.rn pore size PVC filter (NIOSH method No. 0600'34’) and (2) submicrometer-sized elemental and organic carbon. This was done in order to allow comparisons of the elemental carbon data collected in this study with another recently completed study of exposures in railroad workers.'18’ Area sampling was conducted on the dock platform, in road and local truck cabs driven by nonsmokers, and in various areas of the mainte­nance shops. Finally, for purposes of comparison, area sam­ples were obtained to measure background concentrations of submicrometer-sized elemental and organic carbon on a major  state or interstate highway located in or near each city in which exposure evaluations were done (hereafter referred to as highway samples). Additional background area samples were obtained in a residential area of each city (residential defined as located at least 1 mile from any major highway— hereafter referred to as residential samples).


RESULTS

The main purpose of the study was to estimate current exposures of the four a priori job categories to diesel exhaust aerosol by using an appropriate surrogate marker of exposure. Descriptive statistics (means, standard deviations, and confidence limits) were first compiled by the categories being examined for both elemental and organic carbon concentrations. Both arithmetic and geometric means were calculated and reported. The statisti­cal analyses consisted of a preliminary analysis that investigated assumptions for the analysis of variance (ANOVA), a factorial analysis of variance that investigated the presence or absence of differences between group means (job type and weather), and a posteriori multiple range tests (by using 95% Scheffe intervals) to test the presence or absence of differences between group means two at a time. Weather was treated as a dichotomous variable (cold, defined as less than 10°C [50°F] and warm, greater than or equal to 10°C).

Several specific aspects of the data (e.g., methods comparisons) were explored by using linear regression techniques. A small number of the sample results (approximately 1%) were below the limit of detection. These re­sults were included in the overall statistical analysis by substitut­ing a value of one-half of the limit of detection for each “less- than” value.

Dilution Tube Studies

The results of the dilution tube experiments demonstrated strong linear relationships be­ tween measurements of elemen­tal carbon (dependent variable) and the independent variables— total weight of particulates (r2 of 0.98, Figure 3a) and the inverse of the volumetric flow rate in the dilution tube (1/cfm * Ce, r2 = 0.99, Figure 3b). The mean coef­ficient of variation (expressed as a percent of the arithmetic mean) of the elemental carbon method was 7.3%, indicating a relatively high level of precision, because this value includes both sam­pling and analytical errors. Ele­mental carbon constituted an average of 62% of the total particulate loading (by weight) on the filters (Table I) and 64% of the total carbon loading. Similarly, total carbon (elemental plus organic carbon) constituted an average of 91% of the total particulate loading on the filters (Table I).

Cigarette Smoke Studies

The results of the controlled smoking chamber study (Table II) indicated that elemental carbon comprised an average of 1.8% of the total carbon (from tobacco smoke) collected on the filter (i.e., 98.2% of the total carbon was organic carbon). Even in the presence of extremely high concentrations of tobacco smoke, the measured concentrations of elemental carbon were extremely low, thus confirming that tobacco smoke should not seriously interfere in the measurement of elemental carbon from diesel exhaust. Assuming a respirable dust concentration of 1100 ug/m3, as measured in the chamber during the experiment, elemental carbon would comprise ap­proximately 1% or less of the airborne respirable dust from pure tobacco smoke.

Field Studies

Statistical analyses of the field data distributions in general indicated that the exposures to elemental carbon in submicrometer­ sized particulates were lognormally distributed. For example, Figure 4 is a frequency histogram for the log-transformed ele­mental carbon concentration obtained in all personal samples (N = 327), upon which is superimposed the theoretical normal dis­tribution with that data set’s mean and standard deviation. A Kolmogorov-Smirnov (K-S) test for the data set illustrated indi­cated that the distribution was not significantly different (p = 0.27) than the theoretical normal distribution. In addition, the residu­als, based on the analysis of variance of the log-transformed data,followed a normal distribution. Thus, all subsequent statistical analyses were done with the log-transformed data.


Tables III and IV summarize the overall descriptive statistics for submicrometer-sized elemental and organic carbon, respectively. The data were grouped by job (or sampling location in the case of area and background samples) and ambient weather conditions (warm or cold). These statistics include data from all eight study sites. For the organic carbon results (Table IV), the de­scriptive statistics were com­puted (and the statistical analyses were carried out) with only nonsmoking samples be­ cause tobacco smokew ould have provided an unknown and potentially large positive bias.

Figure 5 compares the means of the log-transformed submicrometer-sized elemental carbon concentrations for each of four major job categories (dock work­ers, city drivers, mechanics, and line-haul drivers), together with the means of background con­centrations on the highway and in residential areas within each city. The error bars surrounding each mean are the 95% Scheffe intervals. On the basis of this comparison and the results of the ANOVA and follow-up multiple range tests, all job means were greater than the mean residential concentration (1.1 ug/m3), but only dock workers’ (13.8 ug/m3) and mechanics’ (12.1 ug/m3) mean exposures were greater than highway background concen­trations (2.5 pg/m3). Road and local drivers’ exposures (3.8 and 4.0 ug/m3, respectively) were not discernible from background highway concentrations measured, and dock workers’ exposures were not different than mechanics’ exposures. Both job type and climate factors were significant in the ANOVA (p < 0.0001 and p = 0.023, respectively). In addition, the interaction term between job and climate was highly significant (p < 0.0001).

Figure 6 shows the same data broken down further by sam­ples obtained in cold weather (leftmost bar in each pair) versus those obtained in warm weather (rightmost bar). In mechanics, exposures were clearly higher in cold weather (28 ug/m3) than they were in warm weather (4.8 ug/m3). In road drivers, expo­sures were significantly higher in warm or moderate weather (7.0 ug/m3 versus 2.0 ug/m3 in cold weather). In dock workers and local drivers, a trend appeared to be toward higher exposures in warm weather, but no significant differences were observable. The changes in mean exposures in the mechanics and road drivers by weather were undoubtedly responsible for the signifi­cant interaction term in the ANOVA.

Two of the breakbulk terminals studied used nondiesel fork-lift trucks on the dock. One terminal used propane-fueled forklift trucks, and the other used gasoline-fueled forklift trucks. Because the type of forklift trucks used should have been the primary determinant of the level of dock workers’ elemental carbon exposures, a second analysis of variance was performed comparing dock workers’ exposures by forklift truck engine type: diesel, gasoline, and propane. Dock workers’ mean expores to submicrometer-sized elemental carbon by engine type areshown in Table V. The ANOVA and the a posteriori multiple range test indicated that all three group means were significantly different (p < 0.0001) from each other. Geometric mean exposures to submicrometer-sized elemental carbon were by far the highest where diesel engines were used (27.2 ug/m*), followed in order by gasoline (5.46 ug/m3) and propane  (1.30 ug/m3). The latter two exposure means were of the same order of magnitude as the residential and highway background sample means (2.0 and 3.4 g/m3, respectively).

Table IV and Figures 7 and 8 summarize and compare exposures to submicrometer-sized organic carbon by job and weather, excluding personal samples obtained on workers who smoked. The ANOVA for this dependent variable (C,) again indicated that job (p <0.0001), weather (p = 0.003), and interaction (job * weather) terms (p <0,0001) were all highly significant.

Figure 7 indicates that all Co job means were significantly higher than either highway (3.4 ug/m3) or residential (2.0 ug/m3) background area samples. Figure 8 indicates that both highway and residential background levels were significantly higher in warmer weather. This change could have been caused by either or both the location (city) effects, or the temperatures, because surveys were not done during both warm and cold weather at
each terminal (resources were not sufficient to do this). These trends may have been caused by a number of factors, such as increased levels of ambient pollutant levels during the summer months or decreased suspension of soil particles (which contain a high proportion of organic carbon) in winter because of the frozen ground. However, it appears that all job means were significantly higher than both highway and residential back-
ground samples during cold weather and were also higher than residential background samples during warm weather. However, none of the Co job means were significantly higher than highway samples during warm weather.

Because both organic and elemental carbon fractions were quantitated from the same filters, the data were examined for the possibility of a significant relationship between submicrometer-sized elemental and organic carbon sampling results. This was done because the focus of most toxicity studies has been the organic carbon species adsorbed onto diesel particulate matter. It is known that the proportion of organic to total carbon typically varies from about 10% to 30% in diesel particulates, although in at least one study of diesel engine exhaust from passenger cars extract-
able (organic) fractions were as high as 90%.

Figures 9, 10, and 11 are plots of elemental versus organic carbon for (respectively) all nonsmoking personal samples, non-smoking dock workers only, and nonsmoking mechanics only. The comparisons excluded samples in which the C, concentrations were less than 20 ug/m3. This was done in order to make it more likely that a substantial proportion of diesel exhaust particles was present on the sample filter and not a large proportion
of ambient particles. As indicated, calculated correlations were 0.71 for all nonsmoking samples (p < 0.00001), 0.62 for non- smoking dock workers (p = 0.00086), and 0.71 for nonsmoking mechanics (p = 0.0030).

Although these relationships are highly significant, the degree of association appears to be only moderate. The reasons for lack of better relationships are many, including other sources of organic carbon such as sidestream tobacco smoke, diesel fuel, degreasing solvents, and grease and oil, all of which are strongly adsorbed by the quartz fiber filters. Even where a personal sample was recorded as having been obtained from a nonsmoking employee, those employees invariably spent part of each shift in a smoke-filled break room.

Table VI contains the mean percentages of elemental to total carbon (Ce/Ct x 100) in nonsmoking personal samples obtained in each of the four jobs. In each case, the statistics exclude samples in which the concentration of elemental carbon was less than 10 ug/m3 (using a cutoff value of 20 ug/m3 as above excluded too many samples from the analysis). In this sample subset, the overall mean was about 41%. Within the four jobs, the mean percentage ranged from a low of 33% in road drivers to a high of 47% in dock workers. A one-factor ANOVA indicated a highly significant difference (p = 0.0024) between these percentages. A multiple-range analysis of the data indicated differences only between dock workers (higher) and mechanics
(lower) but no differences between any other two of the jobs. These percentages are somewhat lower than are typically reported for diesel particles and indicate the probable presence of sources of nondiesel particles. In dock workers, there were no obvious external sources of nondiesel organic carbon exposure other than sidestream tobacco smoke. However, in the mechanics, sources included frequent exposure to diesel fuel, oils, greases, and degreasing solvents (the quartz filters also adsorb organic vapors to a variable extent, particularly if a layer of diesel exhaust—derived carbon soot is present on the filter), as well as sidestream tobacco smoke.

Fifty paired area samples consisting of one of each pair for respirable dust and the other for submicrometer-sized elemental carbon were obtained in various areas (on the docks, in road and city driver cabs, and in repair shops) to determine whether a usable relationship could be demonstrated between gravimetric measurements for respirable dust and submicrometer-sized Ce, by thermal-optical analysis. This relationship is important because a major exposure evaluation of diesel exhaust was recently conducted by using ARP as the principal surrogate index. Because the sample
pairs were area samples or were obtained in nonsmoking cabs, the influence of tobacco smoke on the results should have been minimized.

Figure 12 is a scatter diagram of a subset of 18 pairs (pairs were excluded if the elemental carbon concentration was less than 10 ug/m3) and the results of a linear regression analysis. As shown, the strength of the relationship was moderate with a correlation of 0.84 and a coefficient of determination of about 71%. On the basis of the regression coefficients, the Ce results averaged slightly less than 20% of the respirable dust concentrations.

