A field survey was conducted at a copper electrorefinery facility. Tasks involved in this facility were loading/unloading large racks of copper cathodes and anodes from the electrolyte tanks, washing copper cathodes down with hoses, processing finished cathodes, and preparing cathode starter sheets for electrorefining. Sampling took place over the course of four days during the summer; temperatures ranged from 30°C to 47°C during the day in the tankhouse, with an average relative humidity of 64%. The facility was dependent upon natural ventilation (i.e., opening entrances and windows) along with fans, sporadically placed and turned on and off at irregular intervals. Photos were prohibited to protect production trade secrets. Typically, as with most copper refineries, workers in this facility were exposed to airborne metals including copper, arsenic, silver, lead, and selenium.

We collected 48 paired exposure measurements (23 personal and 25 area) using the IOM and the DIAS. Personal exposure measurements were collected from the following job categories: stripper crane man, cleaning stripper, cellar man, scrap operator, scrap washer, crane operator, and supervisor. Area sampling was utilized in aisles between electrorefining tanks and sections in which the acidic electrolyte was heated to a prerequisite temperature and pumped throughout the tankhouse. A mixed-cellulose ester (MCE) filter (37-mm for the prototype and 25-mm for the IOM) with 0.8 μm pore size was mounted in the sampler. In addition, the DIAS has a capsule (made of thin-film polycarbonate) attached to the MCE filter. In previous studies tested this prototype (L’Orange et al., 2016; Stewart et al., 2017; Anthony et al., 2017), the perimeter of a capsule’s base was attached to a filter using toluene (i.e., as glue) so that the capsule and filter could be weighed together. In the present study, we did not glue the capsule with the filter because the capsule material could not be dissolved. Instead, we wiped the internal walls of the sampler to conduct wet chemical analyses. Exposures to airborne particles were collected at 2 L min⁻¹ for both sampler types with sampling times ranging from 256 to 529 minutes. We placed samplers randomly on the opposite sides of the worker’s torso (i.e., one on the left lapel and one on the right lapel). The sampling pumps were calibrated prior to the sampling and checked after the sampling with a DryCal® DC-Lite device (BIOS International Corporation, Butler, NJ, USA) to ensure the nominal flow rate of 2 L min⁻¹ ± 5%. Nine field blank samples for each sampler type were collected.

For each IOM sampler, a filter was removed with tweezers and placed in a tube. After removing the filter, we wiped the interior surfaces of the filter holder twice with a quarter-sized clean MCE filter, soaked in isopropyl alcohol, and placed in another tube. Then, the inside of the cap was wiped twice with another quarter-sized MCE filter and placed in a third tube. For each DIAS, the same steps as for the preparation of IOM samples were repeated for separate analyses. These tasks were done by one lab personnel to minimize the variations among different lab personnel for the wiping procedures.

All collected samples were analyzed according to the NIOSH NMAM 7303 method, with some modification. In order to dissolve the MCE filters, a solution of 4 ml concentrated nitric acid and 1 ml 20% hydrogen peroxide was used. After dissolving an MCE filter, water was then added to a total volume of 40 ml and 1 ml of this solution was pipetted into a 15 ml tube to which was added a water solution of 1 % nitric acid and 1% ethanol containing 10 ppb of Yttrium (Spex Claritas standard solution). Calibration solutions were made by diluting a copper standard (Spex Claritas) solution of 1% nitric acid, 1% ethanol, and 10 ppb Yttrium. Calibration solutions included method blanks and copper concentrations ranging from 1 to 100 ppb. All solutions were analyzed by Inductively Coupled Plasma Mass Spectrometry (ICP-MS; Perkin Elmer 300D, Waltham, MA, USA). In this study, we initially obtained airborne metal concentrations including copper, arsenic, silver and selenium. However, only copper concentrations were used for further data analyses since the amount of mass determined for the other metals were less than the limit of quantitation (LOQ) for most sample pairs when analyzing the wiped samples from the interior surfaces and/or inside the cap. The LOQs are 158.21 ng for copper. For copper, 288 samples (48 pairs x 2 sampling types x 3 tubes [filter, internal wall wipe and cap wipe] for each sample) were analyzed. Only 3 samples (1 filter and 2 cap wipes) from the DIAS showed analyzed masses less than the LOQ. All field blank samples (including filters and wiped samples from the interior walls and inside the cap) showed masses less than the LOQ and thus, no subtraction of the field blank mass was conducted.

The analyzed masses from filter, internal wall wipe and cap wipe were combined to calculate total mass concentrations for each sample. Prior to conducting data analyses, all exposure measurements were log-transformed to meet the assumptions of the statistical tests. We performed data analyses with two data sets, exposure data with LOQ treated by replacing mass below LOQ with LOQ/sqrt(2) and exposure data without LOQ treatment. The results showed no differences of conclusions. Thus, we only reported the results of exposure measurements without LOQ treatment in this paper. In addition, we calculated the proportion of particle mass detected on the cap (determined from wet chemical analysis) and internal walls of each sampler.

For assessing agreement between the DIAS and the IOM, we conducted a concordance correlation coefficient (CCC) test expressing the results as the product of precision and bias coefficients. The precision coefficient represents variation by measuring the distance of each measurement from the best-fit line, while the bias coefficient measures the distance between the best-fit line and the unity line (Lin 1989; Lin et al., 2002; Barnhart et al., 2002; Carrasco and Jover, 2003). Unlike the Bland-Altman test, the results of CCC tests can determine a source of disagreement whether it is from precision or bias. For this test, we used a validated SAS macro provided by Lin et al. (2002). The CCC results were assessed using ± 35% acceptance criterion (0.878 = [1–0.35²]), selected from previous studies to compare different sampler types by Lee et al. (2011) and Lee et al. (2017). Two additional statistical tests were conducted to test 1) if the slope between the DIAS and the IOM sampler is 1 (H₀: slope [β]=1) using Proc Reg procedure and 2) if the average concentration of the DIAS is the same as that of the IOM using Proc Mixed procedure. All statistical analyses were performed with SAS/STAT software version 9.3 (SAS Institute, Cary, NC, USA). A p-value of 0.05 was used for testing the hypotheses.

Table 1 shows a summary of exposure measurement data, presenting the combined results of filter samples and wiped samples (interior walls and inside the cap). For the combined data (personal and area), the geometric mean (GM) concentrations were 36.1 μg m⁻³ for the DIAS and 26.5 μg m⁻³ for the IOM sampler. For both sampler types, the average concentrations were considerably higher than the corresponding GM concentrations revealing that the measurement data were positively skewed. The range of exposure measurements was wider for the IOM sampler compared to that for the DIAS. A similar pattern was observed when the exposure measurements were separated by personal and area exposures.

The GM ratio of exposure measurements (DIAS/IOM) was 1.1 for the personal exposures, 1.6 for the area exposures, and 1.4 for the combined personal and area exposures. The average ratio of exposure measurements was not noticeably different compared to the corresponding GM ratio (1.5, 2.1 and 1.8 for the personal, area and combined exposures, respectively). Overall, the variation of area exposure data is similar to that of personal exposure data (coefficient of variation [CV] = 0.7 for the personal and 0.6 for the area exposure data). Figure 1 shows the individual mass concentrations between the pairs of samples for copper. Regardless of the sampling method (personal or area), overall, the DIAS showed higher concentrations compared to the IOM sampler. About 81% of the combined data (~ 78% of the personal exposures and ~ 84% of the area exposures) showed the concentration ratios (DIAS/IOM) greater than one.

As shown in Table 2, the CCC-total results for the personal, area and combined exposure measurements were lower than 0.878, indicating disagreements between the two types of samplers. Regardless of the sampling method, the CCC-Precision was considerably lower than the corresponding CCC-Bias, suggesting precision (i.e., variation) as the source of disagreement. The CCC-Bias ≥ 0.889 suggests little deviation from the unity line. The results of linear regression analyses to develop slope and test if slope (β) = 1 revealed statistically significant differences of exposure measurements between the two types of samplers: all p-values for the personal, area and combined data were < 0.05 with adjusted R² ≤ 0.51. The comparison of exposure measurements between the DIAS prototype and the IOM resulted in no overall statistical difference for the personal exposure data (p-value = 0.487) and statistically significant differences for the combined data (p-value = 0.035) and area exposure data (p-value = 0.031); the estimated average concentrations were always higher for the DIAS prototype compared to the IOM (Table 2).

We obtained the amount of copper mass collected on the cap by analyzing the samples separately to determine the proportion of particle loss due to the transportation of samples to a laboratory. As described in Methods, although we are interested in the proportion of particle mass deposited on the inside of cap, the proportions of particles collected on filter and internal walls were also presented to provide additional information. As shown in Figure 2, the geometric mean percent proportion of mass (cap/total) was similar for both sampler types (6.4% for the DIAS prototype and 7.4% for the IOM sampler). The comparison results of the average proportion are also similar for both sampler types (13.7% for the DIAS prototype and 11.6% for the IOM sampler).

The copper exposure measurements for the combined data (personal and area) were 0.8 – 337.7 μg m⁻³ (GM = 36.1 μg m⁻³) for the DIAS and 5.1 – 760.8 μg m⁻³ (GM = 26.5 μg m⁻³) for the IOM (Table 1). Regardless of the sampling time ranging from 256 minutes to 529 minutes, none of individual exposure measurements exceeded the occupational exposure limit of 1000 μg m⁻³ by Occupational Safety and Health Administration (OSHA) - Permissible Exposure Limit (PEL), NIOSH-Recommended Exposure Level (REL) and ACGIH-Threshold Limit Value (TLV). For both DIAS and IOM samplers, the GM exposure concentrations were higher for the personal sampling method compared to the area sampling method, leading the difference due to the distance of samplers’ location from the emission source.

Overall, the DIAS generated higher exposure data than the IOM sampler. About 81% of sample pairs (39 out of 48 sample pairs) showed the ratios of exposure data (DIAS/IOM) greater than the unity (Figure 1). For the area exposure measurements, the DIAS resulted in about 1.6 times higher concentrations than the IOM along with the results of disagreement (CCC-Total < 0.878) and statistically significant differences (p-values < 0.05 for the slope test and the comparison of overall average concentrations between two sampler types) (Table 2). Anthony et al. (2017) conducted the performance of the DIAS (running at 10 L min⁻¹) against the IOM (reference, running at 2 L min⁻¹) in a livestock production facility by collecting area exposure measurements; 36 sample pairs of inhalable dust and 44 pairs of inhalable endotoxin were collected. They reported no significant difference of the DIAS compared to the IOM for the comparison of inhalable dust, whereas the DIAS produced higher inhalable endotoxin concentrations compared the IOM. The findings of the present study are inconsistent with the conclusion of inhalable dust but consistent with that of inhalable endotoxin (i.e., overestimation of DIAS) reported by Anthony et al. The sampling flow rate of DIAS utilized in this study was 2 L min⁻¹, not the same as 10 L min⁻¹ that Anthony et al. employed. However, because Stewart et al. (2017) reported no difference of mass concentrations of the DIAS between 10 L min⁻¹ and 2 L min⁻¹, the difference of sampling flow rates between this study and Anthony et al. might not be a reason for causing the difference of the findings.

For the personal exposure measurements, although the comparison of overall average concentrations between the DIAS and IOM revealed no statistically significant difference (p-value =0.487 > 0.05), the disagreement (CCC-Total = 0.692 < 0.878) and lack of a linear relationship of exposure data between the two sampler types indicates that that the performance of the DIAS is not comparable to that of IOM. The range of exposure measurement ratios (DIAS/IOM) between the personal and area exposure data was similar. Interestingly, the GM value of ratios was higher for the area exposure data compared to that for the personal exposure data (Table 1). It is expected to have a lower GM value of ratios for the area samples compared to the personal samples because the paired samplers located for the area sampling were stationary and thus no influence of exposure concentrations from a worker’s behavior. Probably, this high ratio might happen because of abrupt changes in the wind direction from a fan (running on and off, irregular) directing an air stream into the IOM sampler.

As far as the authors are aware, this study is the second field evaluation (after Anthony et al., 2017) using area exposure measurements and the first evaluation using personal exposure measurements for testing the performance of the DIAS. The present study is limited to only one metal component at a workplace with the inhalable exposure range of 5.1 – 760.8 μg m⁻³using IOM samplers. In addition, there might be some errors arising from the wiping process (i.e., applying inconsistent pressure to wipe internal walls and inside of cap), although the wiping process was conducted by one lab personnel to minimize variations among different personnel. In order to make a firm conclusion about the performance of the DIAS, it will be necessary to characterize the performance of the DIAS in various environments sampling numerous chemical components.

The proportion of copper mass collected on the cap over the total copper mass (i.e., sum of copper masses from filter, internal wall wipes and cap wipes) was similar for both sampler types (GM [mean] proportions = 6.4% [13.7%] for the DIAS prototype and 7.4% [11.6%] for the IOM), shown in Figure 2. Demange et al. (2002) investigated metal deposits on sampling cassette walls due to transportation disturbances and reported that 1.8% of the sampled mass of barium and 7.9% of the sampled mass of iron were deposited on cassette walls during transportation. On the other hand, Stacey et al. (2013) also checked the weight of respirable dust on 12 filter samples after transporting to Italy and South Africa from the UK but reported no significant differences (the ratio close to 1). Although the methods to determine particle losses due to the transportation were different between the present study and a study by Demange et al., it is obvious that the loss of particles could happen from the transportation of samples. It should be noted that the proportion of particle loss reported in this study is not representative because particle losses from other parts (e.g., from filter to internal walls and/or vice versa), which was not considered here, could happen during the transportation of samples. The findings of this study indicate that occupational professionals should be aware of potential errors caused from transferring samples from a field to a laboratory and be careful not to exclude particles collected on a cap.