Inclusion criteria for this study were non-smoking males 45 years of age or younger who were instructors with the Chicago Fire Department. The Coordinator of Research and Development at the Chicago Fire Department distributed our study information sheet to eligible Chicago firefighters. After receiving volunteers, he coordinated with the fire chiefs from each station to schedule five firefighter participants for each day of the study. Participants were instructed to not eat char-grilled foods and avoid second-hand tobacco smoke for 2 days prior to the start of the study. Scheduling was also done to ensure that the participants had at least one day off from firefighting activities before reporting to the study site.

Round 1 of the study was in August 2010; round 2 was 1 year later. Firefighters were consented before each round. Fifteen firefighters participated in each round (five firefighters each day). Twelve firefighters from round 1 repeated the study during round 2. Each round consisted of three controlled structure burns (one burn each day).

This study was conducted at the University of Illinois Fire Service Institute training facility. We had five sample collection periods: pre-exposure (~1h before the controlled burn), exposure (during the controlled burn), post-exposure (10–40min after the controlled burn), 3h after the controlled burn, and 6h after the controlled burn. The timing of the samples was based in part on previous studies involving firefighters that found maximum excretion of PAHs 4–8h after exposure (Caux et al., 2002; Laitinen et al., 2010) and rapid excretion of benzene (Laitinen et al., 2010).

Table 1 summarizes the controlled burns for each round, with some important differences between these rounds noted. Prior to each day of the study, a burn room and target room were constructed of drywall inside these structures. These rooms were connected to each other by an open doorway. Fuel packages typical of family room furniture were placed inside the burn room. Thermocouples were placed at various heights inside the target and burn rooms.

Fires were started by the interior safety officer by igniting newspapers inside a wastebasket. Firefighters were inside the target room during active fire (when the fires grew from the ignition stage to the fully developed stage) and knockdown (when the nozzleman suppressed the fires with the hosestream) and then moved into the burn room for overhaul (when the firefighters searched for and extinguished any residual flames). Water was applied to the burn room to control the fire size if ceiling temperatures inside the burn room exceeded 425°C or 1.2-m height temperatures inside the target room exceeded 120°C or if the company officer/safety officer felt it necessary for crew safety. Firefighters were instructed to suppress the fires within 15min after ignition.

The period of active fire was intended to simulate the time period encompassing a typical interior attack when firefighters would be exposed to products of combustion. For round 1, firefighters stood or crouched in the target room during this period. For round 2, the firefighters (except for the nozzleman and company officer who stood or crouched in attack position with a charged hoseline) rotated between three stations in the target room that were intended to simulate searching for victims, pulling down a ceiling to expose overhead fires (Fig. 1), and resting at the back of the structure.

The firefighters wore protective ensembles that complied with the 2007 editions of the NFPA 1971 and 1981 standards that were applicable at the time of this study (NFPA, 2007a,b). The gloves and turnout gear were laundered before use. The hoods used in round 2 were brand new and ~7.5cm longer than the laundered hoods used in round 1. We tested the turnout gear before each burn to ensure that the gear had minimal PAH contamination. Participants did not remove their SCBA until overhaul was completed, and they were at least 30 m northwest of the burn structure. Prevailing winds were out of the southwest. Following each burn, we instructed the participants to stay inside a climate-controlled building for the rest of the day (6h) to minimize any potential exposures from other training exercises taking place at the facility. During round 1 only, training exercises involving wood smoke were performed at the facility several hours after our study’s controlled burns.

We conducted personal air sampling for PAHs during the controlled burns (from ignition to completion of overhaul). We sampled respirable particles using an aluminum cyclone and polytetrafluoroethylene filter (SKC Inc., Eighty Four, PA, USA) and gases and vapors using an attached SKC XAD-2 sorbent tube. The flow rate we used (2.5 l min⁻¹) provides a 4-µm aerodynamic diameter cut point for the cyclone. To prevent fire damage to the sampling train, we wrapped it in Nomex® flame-resistant material (DuPont, Wilmington, DE, USA) and replaced the cyclone’s plastic grit pot with a fabricated aluminum grit pot. The sampling pump and majority of the tubing were positioned on the interior of the turnout gear.

Of the 30 personal sampling pumps used in this study, only 2 functioned properly through the completion of the exercises. Three pumps had disconnected tubing and 25 faulted before completion of overhaul. Of the 25 that faulted, 12 stopped working before knockdown. The median amount of time the pumps ran after fire ignition was 15min, ranging from 2 to 24min. For the 12 pumps that faulted before knockdown, air concentrations were calculated based on the amount of time the pumps ran during active fire; we assumed that the PAH air concentrations during active fire were fairly constant over time. For the 15 pumps that continued to run after knockdown, air concentrations were calculated based on the total time period of active fire (see Table 1); we assumed that the PAH air concentrations after knockdown were negligible. These calculations were done to provide estimates of airborne PAH exposures during the most contaminated time period of active fire.

NIOSH Method 5506 (NIOSH, 2013a) was used to analyze these air samples, which employs high-performance liquid chromatography with photodiode array detection (HPLC-PDA). This method is able to differentiate between 17 PAHs. However, for this article, we report air concentrations of total PAHs. The calculation for total PAHs accounts for all peaks in the PAH response region of the chromatogram; we combined these results for both the particulate and gas phase.

Wipe sampling was used to measure dermal exposure to PAHs on the forearms, hands, neck, face, and scrotum. Samples of both forearms and the neck and scrotum were collected pre- and post-exposure. Samples of both hands and the face were only collected post-exposure. We sprayed corn oil on skin using spray bottles to facilitate the collection of lipophilic PAHs. Four to six sprays (~0.5ml per spray) were applied evenly depending on the surface area of the skin site and then wiped off using AlphaWipes® (TX1004, Texwipe®, Kernersville, NC, USA). For the scrotal wipe samples, wipes impregnated with 1ml of corn oil and nitrile gloves were given to the firefighters with instructions on how to collect the samples. This is similar to the technique used by Väänänen et al. (2005). The wipe samples were analyzed by NIOSH Method 5506 (NIOSH, 2013a), which employs HPLC-PDA.

The corn oil that we used for the dermal wipe samples had a complex chemical matrix. Several peaks were present in the media blanks (wipes impregnated with 2ml of corn oil) in the region of the chromatogram where the PAHs would typically elute, so we could not use total PAHs as a metric for the dermal wipe samples. Instead, we selected six PAHs (anthracene, benzo[a]pyrene, chrysene, fluoranthene, phenanthrene, and pyrene) to sum as a surrogate of the total PAHs. The PAHs we selected had the highest rates of detection and mostly eluted in the region outside of the interfering peaks. They also corresponded with the PAHs toward which the urine enzyme-linked immunosorbent assay (ELISA) method had the greatest sensitivity. The analytical results were blank corrected.

Dermal exposure levels of PAHs were standardized by the surface area of the skin collection site. The surface areas of the forearms (0.15 m²) and hands (0.11 m²) were based on mean dermal exposure factor data for adult males (U.S. Environmental Protection Agency [EPA], 2011). We estimated surface areas of the face (0.068 m²) and scrotum (0.054 m²) by dividing the surface areas of the head and hand, respectively, by two. The surface area of the neck (0.042 m²) was determined based on data from Lund and Browder (1944) showing the neck accounts for 2% of the total body surface area, which is 2.1 m² for adult males 30–39 years of age (EPA, 2011).

Exhaled breath can be used to assess volatile fraction of the systemic uptake, regardless of route of entry (presumably dermal in this case) (Pleil and Lindstrom, 1998; Pleil, 2008). We measured the concentrations of combustion products in exhaled breath samples collected pre-, post-, and 6-h post-exposure. The firefighters were instructed to take a deep breath in and then forcefully exhale into the Bio-VOC™ sampler (Markes International, Wilmington, DE, USA) until they had fully expired their breath, permitting the sampler to collect alveolar air. We then pushed the collected alveolar air through Markes Carbograph 2TD/Carbograph 1TD thermal desorption tubes using a plunger. The samples were analyzed for aromatic hydrocarbons (benzene, toluene, ethyl benzene, xylene, and styrene) and semi-volatile PAHs (naphthalene, anthracene, phenanthrene, fluoranthene, and pyrene) using a gas chromatography/mass spectrometry (GC/MS) method described in Sobus et al. (2008).

We assessed systemic PAH and benzene exposure of participants by measuring their PAH and benzene metabolite levels in urine. We also measured urinary creatinine and cotinine. Creatinine was used to normalize the PAH and benzene metabolite results. Urine samples were collected pre-, post-, 3-h, and 6-h post-exposure. The participants were given sterile 100-ml collection cups for their specimens. All urine samples were aliquoted into labeled tubes and stored on dry ice while in the field. On arrival in the lab, samples were stored at −20°C for those pending PAH analysis and at −80°C for those pending cotinine and creatinine analyses.

The urinary PAH-metabolite assay was performed using a modified version of a commercial ELISA (PAH RaPID Assay®, Strategic Diagnostics Inc., Newark, DE, USA) to detect polycyclic aromatic compounds in aqueous samples (Smith et al., 2011). Urine samples diluted 25% with methanol at collection were treated with the enzyme β-glucuronidase to cleave glucuronide conjugates of PAH metabolites followed by 1/20 dilution with kit diluent to diminish urine matrix effects and then assayed according to the instructions. The concentrations were reported as phenanthrene kit equivalents corrected for a dilution factor of 28.8. This method has been shown to correlate well (r = 0.89) with the sum of eight PAH metabolites measured by GC/MS in exposed workers’ urine (Smith et al., 2011). Although it may be considered a broad screening tool, it is most sensitive toward the six PAHs measured on skin. The assay detection limits for phenanthrene, fluoranthene, benzo[a]pyrene, pyrene, chrysene, and anthracene were 0.93, 0.43, 0.67, 0.27, 0.53, and 0.72ng ml⁻¹, respectively.

The urine samples were not analyzed for the benzene metabolite, s-phenylmercapturic acid (s-PMA), until 1–2 years after collection. We added this test to the study after detecting elevated post-exposure concentrations of benzene in breath compared with the pre-exposure concentrations. These samples were analyzed by NMS Labs (Willow Grove, PA, USA) using an internal LC/MS-MS method. Prior to analysis, 1 drop of 12 N hydrochloric acid was added to each 5-ml urine sample aliquot.

Creatinine was measured using a Vitros Autoanalyzer (Johnson & Johnson, New Brunswick, NJ, USA) with a Vitros CREA slide. Cotinine, a metabolite of nicotine, was measured using Diagnostic Products Corporation (Siemens Corporation, Washington, DC, USA) Immulite® 2000 analytical platform. Analysis of cotinine concentrations was used to confirm current non-smoking status of the participants and to quantify possible exposure to environmental tobacco smoke (Suwan-ampai et al., 2009).

SAS 9.3 statistical software (SAS Institute, Cary, NC, USA) was used for statistical analysis. Because of varying parameters of the two rounds, round 1 and round 2 data were analyzed separately. Non-detectable (ND) measurements were assigned values of the minimum detectable concentration (MDC) divided by the square root of two. MDCs are the lowest concentrations of a particular analyte that can be detected with a sampling method. We used the average volume of air sampled (or average other denominator) to calculate the MDCs. All personal air and breath samples measured detectable levels of PAHs and benzene. However, several wipe samples measured ND levels of PAHs. Of particular interest in this article, 10 of 30 post-exposure wipe samples of the neck measured ND levels for all six PAHs and 5 of 30 urine samples (3-h post-exposure) measured ND levels of PAH metabolites. The MDCs for PAH levels on the neck varied slightly for each of the six PAHs, ranging from 7 to 14 µg m⁻² in round 1 (57 µg m⁻² total) and from 7 to 21 µg m⁻² in round 2 (76 µg m⁻² total). The MDC for PAH metabolites in urine was 22 µg g⁻¹ creatinine. A full list of MDCs is provided in the NIOSH report (NIOSH, 2013b).

We excluded a few personal exposure data from the statistical analyses. The pre-exposure level of PAHs on the neck of one subject (380 µg m⁻²) was excluded from analyses because it was well above all other pre and even post-exposure measurements. One pre-exposure breath measurement of benzene (378 µg m⁻³) was excluded from statistical analysis because we had reason to believe the thermal desorption tube used to make this measurement had not been thoroughly cleaned from prior use. Three air samples were excluded because the tubing became disconnected (as mentioned previously).

New variables representing the change in dermal exposure, urine, and exhaled breath levels over different collection time periods were generated. The distributions of these and other variables were explored. Because the assumption of normality was not reasonable for many of our paired comparisons, non-parametric sign tests were used to explore possible significant changes. Spearman correlations were used to explore relationships between exposure and biomarker variables.

The personal air concentrations of PAHs are summarized in Fig. 2. Because the firefighters wore SCBA, these do not represent inhalation exposures. The air concentrations of PAHs appear higher and more variable during round 1 than round 2. Of the PAHs measured in air, >95% were in the particulate phase or adsorbed to particles. Table 2 provides the average proportion of known, probably, and possibly carcinogenic PAHs (according to IARC), and other PAHs collected on the personal air samples for each burn. These proportions did not vary substantially within a round (coefficient of variations < 38%). Overall, the round-1 burns produced a greater proportion of ‘potentially’ carcinogenic PAHs than the round-2 burns.

Table 3 provides the PAH dermal exposure results by skin site. The number of detectable levels on the arm, neck, and scrotum was generally greater in round 2 than in round 1. All median levels were below the sum of MDCs for the 6 PAHs, except for the exposures on the arm in round 2.

The neck was the only skin site where we measured a statistically significant pre- to post-exposure increase in PAH levels (for round 1 only). During round 2, the greatest pre- to post-exposure increase in median PAH levels was measured on the neck, but the difference in PAH levels was not statistically significant (P = 0.07). The neck also had the greatest range of post-exposure values for both rounds. For these reasons, subsequent data analyses involving dermal exposure data were focused on PAH levels measured on the neck. Phenanthrene was detected on the neck post-exposure in more than twice the number of samples (n = 15) of any other analyte.

Because the corn oil provided a complex matrix, we explored the contribution of the corn oil matrix to the variability in the PAH levels measured on the neck. Analysis of the field blanks (wipe plus corn oil) showed standard deviations of 16 µg m⁻² (n = 6) and 6.5 µg m⁻² (n = 9) for rounds 1 and 2, respectively. This analysis suggests that the corn oil matrix could account for 22–39% of the variability in the PAH measurements on the neck, with the exception of the round 1 pre-exposure measurements that had much less variability (standard deviation = 0.72) than the field blanks. Low pre-exposure variability was not unexpected because the firefighters were instructed to avoid sources of PAHs prior to starting the study.

Of the compounds measured in breath, we focus on benzene for this article. This is because benzene represented the predominant aromatic hydrocarbon measured post-exposure in breath with the largest change in median concentrations over time. For both rounds, the median breath concentrations of benzene were elevated post-exposure and subsequently decreased at the 6-h collection. Although not shown here, this trend was repeated for most other aromatic hydrocarbons measured in breath. Other than for naphthalene, this trend was not observed for the semi-volatile PAHs (NIOSH, 2013b).

Figure 3 shows the individual benzene breath concentrations across time. Most round 1 participants (12 of 14) exhibited a post-exposure increase (P = 0.01) in exhaled breath concentrations, spanning up to two orders of magnitude. Most round 2 participants (12 of 15) also exhibited a post-exposure increase (P = 0.04), spanning up to one order of magnitude. For round 1, nine participants had higher 6-h breath concentrations than their post-exposure levels (versus three in round 2), including seven with benzene ≥ 290 µg m⁻³ (connected by dotted lines in Fig. 3). Tobacco smoke did not appear to be a cause of these elevated 6-h measurements because 6 of 7 of these participants had ND levels of cotinine in their urine.

We ascribe these elevated measurements to a true effect possibly from some uncontrolled post-firefighting benzene exposure. For example, gaseous contaminants produced by training fires that took place later in the day in round 1 (several hours after the exposure period) could have been drawn through the air intake of the building where firefighter participants were stationed. Additionally, the participants might have gone outdoors unbeknownst to us. Such exposure would have mainly affected the 6-h exhaled breath levels of aromatic hydrocarbons and possibly the 6-h urinary PAH levels. This bimodal trait in benzene in breath was not observed in round 2 where no concurrent training exercises took place and where we ensured participants remained indoors after the controlled burns. Figure 3 also includes the high pre-exposure breath concentration (380 µg m⁻³) measured in round 1 that was excluded from statistical analyses. Including this data point resulted in a sharp decrease in post-exposure breath concentrations of benzene for one individual (Fig. 3).

Although we found statistically significant post-exposure increases of benzene in breath, all the urine concentrations of s-PMA were below the LOD of 5 µg l⁻¹. After correcting by the average creatinine concentration, the MDC was 8.5 µg g⁻¹. Therefore, on average, the systemic exposure to benzene was below the American Conference of Governmental Industrial Hygienists (ACGIH®) Biological Exposure Index (BEI®) of 25 µg g⁻¹ creatinine (ACGIH, 2013). This suggests that although some biological uptake occurred (as shown in the breath samples), the total dose of benzene over the short exposure period (≤30min) was not enough to increase urinary excretion of benzene metabolites above exposure criteria.

Figure 4 summarizes the urinary PAH metabolite levels over time. Median urinary PAH metabolite levels appear higher during round 1 than round 2. The highest median urinary PAH metabolite levels were measured during the 3-h collection for both rounds, but the temporal pattern varied between rounds. The PAH metabolite levels in the 3-h samples did not differ significantly from the pre-exposure levels for either round.

Urinary cotinine concentrations above 30 µg l⁻¹ may indicate passive or active light cigarette smoking (Wall et al. 1988). The majority of the firefighters (23 of 30) had urinary cotinine concentrations below 30 µg l⁻¹. Of the seven firefighters with elevated levels, two had urinary cotinine concentrations ranging from ND (<MDC of 10 µg l⁻¹) to 66 µg l⁻¹ (with levels above 30 µg l⁻¹ measured only in the first two collection periods) and five had urinary cotinine concentrations ranging from 45 to 990 µg l⁻¹.

To explore relationships between external exposures and biomarkers, we created three variables that generally represented the greatest increase in median exposures from one collection point (e.g. pre-exposure) to another collection point (e.g. post-exposure). Table 4 presents the median values of these three variables and the personal air concentrations of PAHs by burn. The firefighters participating in the two burns with the highest median personal air concentrations of PAHs (burns 1 and 3 of round 1) also had the highest median levels of the other exposure and biomarker variables. It is also interesting to note that a ranking of the change in PAH levels on the neck from lowest to highest corresponds with the same ranking of the change in urinary PAH metabolite levels.

Table 5 provides the descriptive statistics for these four variables by round. We saw statistically significant increases in the change in PAH levels on the neck for round 1 and the change in breath concentrations of benzene for both rounds. Although we had 15 or less measurements per variable per round, we explored the correlations among these four variables (Table 6). We chose a sign test P value < 0.15 as the inclusion criterion for the correlation analysis. Although benzene has one ring and PAHs have multiple rings, we explored correlations between the benzene and PAH variables because other firefighter studies have shown that emissions of benzene increase with increasing emissions of PAHs (QFRS, 2011a,b). We found two significant correlations; personal air concentrations of PAHs were significantly correlated with the change in urinary PAH metabolite levels in round 1 and with the change in breath concentrations of benzene in round 2.