Associations between PBDEs in Office Air, Dust, and Surface Wipes

Abstract

Increased use of flame-retardants in office furniture may increase exposure to PBDEs in the office environment. However, partitioning of PBDEs within the office environment is not well understood. Our objectives were to examine relationships between concurrent measures of PBDEs in office air, floor dust, and surface wipes.

We collected air, dust, and surface wipe samples from 31 offices in Boston, MA. Correlation and linear regression were used to evaluate associations between variables. Geometric mean (GM) concentrations of individual BDE congeners in air and congener specific octanol-air partition coefficients (K_oa) were used to predict GM concentrations in dust and surface wipes and compared to the measured concentrations. GM concentrations of PentaBDEs in office air, dust, and surface wipes were 472 pg/m³, 2411 ng/g, and 77 pg/cm², respectively. BDE209 was detected in 100% of dust samples (GM=4202 ng/g), 93% of surface wipes (GM=125 pg/cm²), and 39% of air samples. PentaBDEs in dust and air were moderately correlated with each other (r=0.60, p=0.0003), as well as with PentaBDEs in surface wipes (r=0.51, p=0.003 for both dust and air). BDE209 in dust was correlated with BDE209 in surface wipes (r=0.69, p=0.007). Building (three categories) and PentaBDEs in dust were independent predictors of PentaBDEs in both air and surface wipes, together explaining 50% (p=0.0009) and 48% (p=0.001) of the variation respectively. Predicted and measured concentrations of individual BDE congeners were highly correlated in dust (r=0.98, p<0.0001) and surface wipes (r=0.94, p=002). BDE209 provided an interesting test of this equilibrium partitioning model as it is a low volatility compound.

Associations between PentaBDEs in multiple sampling media suggest that collecting dust or surface wipes may be a convenient method of characterizing exposure in the indoor environment. The volatility of individual congeners, as well as physical characteristics of the indoor environment, influence relationships between PBDEs in air, dust, and surface wipes.

1. Introduction

Polybrominated diphenyl ethers (PBDEs) are a class of brominated flame-retardants that were originally manufactured in three commercial mixtures: PentaBDE, which was added primarily to polyurethane foam used in furniture, and OctaBDE and DecaBDE, which were added primarily to plastic polymers used in electronics. In humans, exposure to PBDEs has been linked to disruption of thyroid and reproductive hormone homeostasis (Chevrier et al. 2010; Johnson et al. 2012; Kim et al. 2012; Meeker et al. 2009; Shy et al. 2012; Turyk et al. 2008), adverse birth outcomes (Chao et al. 2007; Harley et al. 2011; Main et al. 2007), and changes in neurodevelopment (Chao et al. 2011; Herbstman et al. 2010; Hoffman et al. 2012; Roze et al. 2009) and reproduction (Chao et al. 2010; Harley et al. 2010).

In light of potential health effects, the accurate assessment of exposure to PBDEs is important in environmental epidemiology, risk assessment, and in determining methods to reduce exposure, but assessing exposure can also be costly and time consuming. Currently, it is unclear which environmental samples (air, floor dust, surface dust, home vacuum bag dust, or others) best characterize PBDEs in a microenvironment, which are most relevant to exposure, and what other factors need to be considered when making decisions regarding sample collection.

In previous work examining pathways of exposure, we reported that exposure to PBDEs in indoor environments occurred via PBDEs on hands (Stapleton et al. 2012; Watkins et al. 2011), which likely resulted from contact with contaminated surfaces (e.g. settled dust). A limited number of studies have measured PBDEs in organic films from windows (Butt et al. 2004; Cetin and Odabasi 2011) or settled dust on televisions, DVD players, and other objects that are likely sources of PBDEs (Schecter et al. 2005; Toms et al. 2009). However, none have used surface wipes to collect settled dust in the office environment.

Weschler and Nazaroff described methods used to estimate partitioning of semi-volatile organic compounds (SVOCs) between air and settled dust in indoor environments using a compound's octanol-air (gas phase) partition coefficient (K_oa) (Weschler and Nazaroff 2008). As PBDEs and other SVOCs may partition between vapor, airborne particles, settled dust on surfaces and floors, as well as on skin and clothing, understanding this process is an important part of determining how people are exposed to these compounds in the indoor environment. Weschler and Nazaroff used this method to predict concentrations of SVOCs in dust based on K_oa values and concentrations in air, and found that it worked well on average but may not be accurate in individual cases. They also suggested that partitioning may be kinetically limited for very low volatility compounds (Weschler and Nazaroff 2010). While important in understanding the partitioning of PBDEs and other SVOCs in indoor environments, this previous study focused mainly on residential environments and did not have adequate data to include BDE209, the only PBDE congener still actively manufactured in the US. As BDE209 is the most highly brominated and least volatile PBDE, it provides an important test of the limits of the equilibrium partitioning model.

The main objectives of this study were to characterize relationships between PBDEs in office air, floor dust, and surface wipes, as well as to evaluate office characteristics and other variables that could potentially influence these relationships. In addition, we aimed to explore associations between measured and predicted concentrations of PBDEs in office dust using methods described by Weschler and Nazaroff (Weschler and Nazaroff 2010). By examining associations between office air, dust, and surface wipes using a variety of methods, we hope to inform exposure assessment methodology by exploring the partitioning of PBDEs among indoor compartments and determine which sample type provides the best measure of PBDE exposure in the office environment.

2. Material and Methods

2.1 Study Design

We recruited a convenience sample of 31 adults who live and work in the Boston, Massachusetts (USA) area, as previously described (Watkins et al. 2011). To be eligible for participation, subjects had to work at least 20 hours a week in an office and be a healthy non-smoker. The field effort was conducted from January through March of 2009. Active participation for each subject lasted one week and included the collection of a dust sample, air sample, and surface wipe from each office, and questionnaire data from each participant.

Participants worked in offices located in seven different buildings, which were grouped into three categories: Building A (n=6), Building B (n=17), and Other (n=8). Building A was newly constructed approximately 1 year prior to sample collection, Building B was an older building that had been partially renovated within 1 year prior to sample collection, and the Other buildings were generally older and had not been recently renovated. The Boston University Medical Center institutional review board approved the study protocol and informed consent was obtained from all participants prior to participation.

2.2 Air Samples

One air sample was collected from each office over four consecutive workdays from 7 am to 7 pm for a total sampling time of 48 hours. Sampling pumps were calibrated to 4 liters per minute, resulting in an average total sampled air volume of 11.52 m³. Flow rates were checked at the beginning and end of each sampling period. Sampling media consisted of a glass fiber filter (GFF) with a nominal pore size of 1 μm followed by a 76 mm polyurethane foam (PUF) plug. The GFF (used to capture PBDEs bound to particulates) and PUF (used to capture PBDEs in the vapor phase) were extracted and analyzed together for one air measurement per office. Sampling media were encased inside a glass tube, placed approximately 1.2 meters above the floor using a tripod, and connected to sampling pumps via ¼ inch tygon tubing. Three field blanks and three pairs of duplicates were collected from offices at random. After collection, samples were wrapped in foil and bubble wrap, sealed in polyurethane bags, and stored at -20°C. We analyzed air samples for 21 PBDE congeners using gas chromatography-mass spectrometry operated in electron capture negative ionization mode (GC/ECNI-MS) (Agilent 6890N/5975), as previously described (Allen et al. 2007).

2.3 Dust Samples

Investigators collected dust samples into cellulose extraction thimbles as previously described (Allen et al. 2008). Each office was vacuumed for approximately 10 minutes at the end of the sampling week, capturing dust from the entire floor surface area of the room including accessible floor space under desks and the tops of immovable furniture. After sample collection, thimbles were wrapped in aluminum foil, sealed in polyurethane bags and stored at room temperature until processed. We collected dust field blanks (n=12) by vacuuming sodium sulfate powder (as a surrogate for dust) from a clean aluminum foil surface. Dust samples were sieved to collect particles <500 μm in size, placed in clean amber glass jars and stored at -20°C. We analyzed dust samples for 37 PBDE congeners using GC/ECNI-MS (Agilent 6890N/5975) as previously described (Stapleton et al. 2008).

2.4 Surface and Hand Wipes

Surface wipes were collected from an undisturbed ‘dusty’ surface in each office, such as a bookshelf or filing cabinet, at the end of the sampling week (prior to floor dust collection). To collect surface wipes, we immersed a 3 inch × 3 inch (58 cm²) sterile gauze pad in 3 ml of isopropyl alcohol and thoroughly wiped a measured area of the designated surface. Templates were used to define a 36 in² (232 cm²) area to be wiped, and were cleaned with a methanol solution between samples. After collection, surface wipes were placed in clean glass vials, wrapped in foil and bubble-wrap, sealed in polyurethane bags, and stored at -20°C. We paired a field blank wipe sample with the collection of each surface wipe by soaking a gauze pad in isopropyl alcohol and placing it directly into the glass vial. However, PBDEs were not detected in most wipe field blanks, suggesting that the collection of fewer wipe blanks would have been sufficient. Surface wipes were analyzed for 37 PBDE congeners using GC/ECNI-MS (Agilent 6890N/5975), using methods described previously (Stapleton et al. 2008). PBDE mass measurements were normalized using the surface area sampled, yielding a measurement of surface loading. Handwipes were collected from participants in their office environment at the beginning of the sampling week using methods as described previously (Watkins et al. 2011).

2.5 Questionnaire

Researchers administered a questionnaire to participants to collect information about work habits and office characteristics, including the age of their office computer and if new carpet had been installed in their office within the past year. Temperature and relative humidity were measured in each office at the beginning and end of the participation week and then averaged.

2.6 Data Analysis

Dust and air samples were blank-corrected by subtracting the mean of the respective field blanks; surface and hand wipes were blank-corrected by subtracting each individual's paired wipe field blank measurements. Limits of detection (LOD) were determined as three times the standard deviation of the appropriate field blanks. In instances where congeners were not detected in field blanks or data were insufficient to calculate a LOD, the laboratory instrument detection limit was used as the LOD. Measurements below the LOD were substituted with a value of LOD/2. PentaBDEs were defined as the sum of BDE 28/33, 47, 49, 85/155, 99, 100, 153, 154, the congeners detected in greater than 50 percent of dust and air samples. A sum measure was appropriate as these congeners were highly correlated within each sample type (data not shown).

Since PBDE data were log-normally distributed they were natural log (ln) -transformed before analysis where appropriate. We used Spearman correlations to explore associations between continuous measures of PBDEs in office air, dust, and surface wipes. Regression models were used to evaluate predictors of ln PBDE concentrations in air, dust, and surface wipes. Predictors included building category, age of computer (<1-2 years vs. 3-6 years), and continuous measures of office temperature and PBDEs in collocated samples (e.g. dust as a predictor of air, or air as a predictor of surface wipes). The beta estimate represents the change in PentaBDEs on the natural log scale per unit change for continuous variables and relative to the reference group for categorical variables. Beta estimates were exponentiated (e^β), producing an estimate of the multiplicative change in PBDE concentration. To minimize the effect of skewed data and outliers, three-level categorical variables (low, medium, high) were created from continuous air and dust concentrations using tertiles. Potential confounders, correlations between independent variables, and sample size were considered when developing multivariable models. All statistical analyses were performed using SAS version 9.1 with statistical significance defined as α=0.05.

Predicted concentrations of BDE congeners in individual floor and surface dust samples (collected by surface wipes) were calculated using BDE congener concentrations in co-located air samples and congener specific octanol-air partition coefficients (K_oa), as described by Weschler and Nazaroff (Weschler and Nazaroff 2010). Calculated log K_oa values at a nominal 25°C previously reported by Weschler and Nazaroff were used with the exception of BDE209, which was calculated using SPARC online calculator (SPARC 2009) again assuming 25°C. Predicted concentrations were calculated using the following equation:

$$\mathit{\text{PBDE}}{s}_{\mathit{\text{dust predicted}}}=\frac{\mathit{\text{volume fraction or ganic matter in dust}}\times K_{oa}\times \mathit{\text{PBDE}}{s}_{\mathit{\text{vapor}}}}{\mathit{\text{density of dust}}}$$

We assumed the volume fraction of organic matter in dust was 0.2 and the density of dust was 2.0 × 10⁶ g/m³ (Weschler and Nazaroff 2010). The concentration of PBDEs in vapor phase was calculated from the measured total concentration of PBDEs in air (vapor and particulate bound) using the following equation:

$$\mathit{\text{PBDE}}{s}_{\mathit{\text{vapor}}}=\frac{\mathit{\text{PBDE}}{s}_{\mathit{\text{airborne particles}}+\mathit{\text{vapor}}}}{1+(\mathit{\text{total suspended particulate}}\times K_P)}$$

where K_p is the particle-air partition coefficient calculated for each PBDE congener and 20 μg/m³ was assumed for total suspended particulate (Weschler and Nazaroff 2010).

To compare measured values of PBDEs in surface wipes (pg/cm²) to predicted concentrations, we converted the measured pg/cm² (mass per unit surface area, or surface loading) to ng/g of PBDEs in surface dust (concentration) using the following equation:

$$\mathit{\text{PBDE}}{s}_{\mathit{\text{concentration}}}=\frac{\mathit{\text{PBDE}}{s}_{\mathit{\text{loading}}}}{\mathit{\text{film thickness}}\times \mathit{\text{density of dust}}}$$

We assumed a film thickness of 100 nm (thickness of the dust layer) (Weschler and Nazaroff 2008) and a dust density of 2.0 × 10⁶ g/m³ (Weschler and Nazaroff 2010). Weschler and Nazaroff assumed a film thickness range of 10 – 100 nm when sampling window surfaces. As thicker films may be possible on horizontal surfaces compared to windows, we used the higher film thickness value.

Measured and predicted PBDE concentrations in floor and surface dust were compared by plotting the measured geometric mean of each congener verses the predicted geometric mean of each congener, and comparing these points to a hypothetical 1:1 line. We performed Pearson correlations between predicted and measured geometric means and reported R² values.

3. Results

Summary statistics for PBDEs in office dust, air, and surface wipes are presented in Table 1. Due to low recovery of the BDE209 standard in some samples (< 45% for ¹³C-BDE209, compared to an average of 84% recovery for F-BDE69), BDE209 measurements were only available for 14 surface wipes. PBDE concentrations in office dust samples collected in this study were previously described in Watkins et al. 2011 and Watkins et al. 2012.

Table 1. Measurements of PBDEs in Office Dust, Air^a, and Surface Wipes (n=31).

	Office Dust (ng/g)			Office Air (pg/m3)			Office Surface Wipes (pg/cm2)
	% Detection	GM (GSD)	Range	% Detection	GM (GSD)	Range	% Detection	GM (GSD)	Range
BDE 28/33	87	7.5 (4.9)	<0.4 - 207	100	57 (2.0)	18 - 433	48	NC	<1.0 - 35
BDE 47	100	697 (3.7)	37 - 19494	100	263 (3.1)	25 - 2895	97	35 (3.1)	<4.1 - 639
BDE 49	87	19 (8.4)	<0.4 - 612	58	10 (2.6)	<8.6 - 101	58	0.47 (7.0)	<0.09 - 23
BDE 66	84	9.0 (10)	<0.2 - 504	39	NC	<4.3 - 71	35	NC	<0.14 - 14
BDE 75	87	40 (6.1)	<0.4 - 227	3	NC	<8.6 - 42	83	3.2 (2.9)	<1.3 - 21
BDE 85/155	97	50 (5.6)	<0.2 - 3085	68	3.4 (2.2)	<2.6 - 67	65	0.74 (6.8)	<0.10 - 87
BDE 99	97	915 (6.1)	<0.4 - 32831	100	82 (2.8)	14 - 1804	100	21 (3.3)	3.4 - 992
BDE 100	100	195 (4.3)	13 - 8672	94	19 (3.0)	<4.3 - 458	100	5.6 (3.2)	1.1 - 267
BDE 138	100	18 (4.9)	1.6 - 958	3	NC	<2.6 - 10	29	NC	<0.06 - 20
BDE 153	100	138 (4.9)	11 - 5973	90	5.8 (2.2)	<2.6 - 79	84	2.1 (5.5)	<0.07 - 158
BDE 154	100	115 (4.5)	7.6 - 5202	81	4.6 (2.4)	<2.6 - 85	84	2.4 (7.1)	<0.10 - 140
BDE 183	100	81 (4.0)	15 - 12970	6	NC	<4.3 - 4.4	61	0.83 (4.3)	<0.37 - 16
BDE 196	100	29 (3.0)	6.7 - 2858	NA	-	-	29	NC	<0.09 - 10
BDE 197	100	32 (4.1)	4.2 - 6109	NA	-	-	19	NC	<0.07 - 31
BDE 201	97	4.9 (3.0)	<1.0 - 359	NA	-	-	10	NC	<0.10 - 3.0
BDE 206	100	153 (2.7)	29 - 3395	0	NC	<17	29	NC	<0.52 - 38
BDE 207	100	125 (3.0)	22 - 4312	NA	-	-	42	NC	<0.39 - 19
BDE 208	100	62 (2.9)	10 - 1710	NA	-	-	39	NC	<0.42 - 11
BDE 209	100	4204 (2.9)	912 - 106204	39	NC	<113 - 423	93b	125 (3.8)	<8.5 - 619
PentaBDEsc	-	2411 (4.2)	154 - 70163	-	472 (2.7)	72 - 5522	-	77 (3.1)	9.2 - 2314

^aSum of vapor and particulate. GM = Geometric Mean, GSD = Geometric Standard Deviation, NA = not analyzed, NC = not calculated due to detection in <50% of samples.

^bOffice Wipes n=14 for BDE209;

^cPentaBDEs = 28/33, 47, 49, 85_155, 99, 100, 153, 154. Congeners with <50% detection in office dust not reported (BDE 17, 25, 30, 71, 116, 119, 156, 171, 176, 179, 181, 184, 188, 190, 202, 205). BDE 191 and 200/203 not reported due to interfering substances.

3.1 Predictors of PentaBDEs in Air

Figure 1 shows that in all buildings combined PentaBDEs in office air were positively correlated with PentaBDEs in office dust (Spearman r=0.60, p=0.0003), a correlation that was strongest in Building B (r=0.60, p=0.01). Correlations in Building A and Other buildings were also positive (r=0.49, p=0.33; r=0.52, p=0.18 respectively) but not statistically significant possibly due to the smaller number of participants in these building categories.

Figure 1.

Scatterplot of PentaBDEs in office dust vs. office air. Spearman correlation for all buildings was 0.60 (p=0.0003). Spearman correlations by building were as follows: Building B r=0.60, p=0.01; Building A r=0.49, p=0.33; Other r=0.52, p=0.18.

Parameter estimates, standard errors, and p-values from regression models examining predictors of PentaBDEs in office air are shown in Table 2. Building category was a univariate predictor of PentaBDEs in air (p=0.01) with concentrations in Other buildings that were 4.5 (i.e., e^1.51) times those in Building A (p=0.004), and 2.3 (e^0.83) times those in Building B (p=0.04). PentaBDEs in air from Building B were 1.97 (e^0.68) times those in Building A, but this difference was not statistically significant (p=0.12). PentaBDE concentrations in dust were a univariate predictor of PentaBDE concentrations in air, both as a continuous variable (p=0.008) and as tertiles (p=0.004). Those in the high dust group had 4.1 times the PentaBDEs in air compared to the low dust group (p=0.0009), and those in the medium dust group had 2.0 times the PentaBDEs in air compared to the low dust group (p=0.07). Temperature, relative humidity, having new office carpet installed in the past year, and age of computer (<1-2 years vs. 3-6 years) were not significant predictors of PentaBDEs in office air. In a multivariable model, categorical variables for dust and building together explained 48% of the variation in PentaBDE air concentrations (p=0.001).

Table 2. Predictors of PentaBDEs in Office Air (pg/m³).

Univariate Models	n	βa (SE)	p-value	R2
Dust (ng/g)
continuous	31	0.00003 (0.00001)	0.01	0.22
Temperature (°C)
continuous	31	0.14 (0.16)	0.40	0.02
Humidity (%)
continuous	31	0.011 (0.036)	0.76	<0.01
New Carpetb
No	12	0.57 (0.32)	0.10	0.17
Yes	5	ref
Age of Computer
3-6 years	20	0.13 (0.40)	0.75	<0.01
<1 - 2 year	10	ref
Building			0.01	0.26
Other	8	1.51 (0.48)	0.004
Building B	17	0.68 (0.43)	0.12
Building A	6	ref
Dust (tertiles)			0.004	0.33
High	10	1.42 (0.38)	0.001
Med	11	0.71 (0.37)	0.07
Low	10	ref
Multivariable Model	n	βa (SE)	p-value	R2
Model			0.001c	0.48
Dust			0.01
High	10	1.18 (0.36)	0.003
Med	11	0.52 (0.38)	0.19
Low	10	ref
Building			0.04
Other	8	1.19 (0.44)	0.01
Building B	17	0.53 (0.41)	0.21
Building A	6	ref

^aBeta coefficients represent the change in natural-log transformed PentaBDE concentrations in air relative to the reference group for categorical variables, and per unit change for continuous variables.

^bNew carpet installed in office within past year. Restricted to Building B as all offices with new carpeting were located in this building.

^cGlobal p-value.

3.2 Predictors of PentaBDEs in Surface Wipes

PentaBDEs in surface wipes were positively correlated with PentaBDEs in office air and dust (Spearman r=0.51, p=0.003 for both associations), although correlations varied by building category (Figure 2). Surface wipes and air were strongly correlated in Building A and Other buildings (r=0.83, p=0.04; r=0.64, p=0.09 respectively), although the latter relationship was not statistically significant. This is probably due to the small number of participants in these buildings. In contrast, PentaBDEs in surface wipes and air were not correlated in Building B (r=0.05, p=0.84). A similar pattern was seen with office dust and surface wipes with correlations again in Building A and Other buildings (r=0.77, p=0.07; r=0.81, p=0.01 respectively) and no correlation in Building B (r=0.24, p=0.35).

Figure 2.

Scatter plots of PentaBDEs in office air (pg/m³) vs. office surface wipes (pg/cm²). a. Spearman correlation for all buildings was 0.51 (p=0.003). b. Spearman correlation for offices in Building B was 0.05 (p=0.84). c. Spearman correlation for offices in Building A was 0.83 (p=0.04). d. Spearman correlation for offices in Other buildings was 0.64 (p=0.09).

Table 3 presents parameter estimates, standard errors, and p-values from regression models examining predictors of PentaBDEs in surface wipes. Continuous variables for PentaBDEs in dust and air were both significant univariate predictors of surface wipes (Dust: β=0.00004, p=0.003; Air: β=0.0006, p=0.0003). PentaBDE dust tertiles were univariate predictors of surface wipes (p=0.001), with wipes in the high dust group that were 3.9 times higher than wipes in the low dust group (p=0.002). PentaBDE air tertiles were also predictors of PentaBDEs in surface wipes (p=0.007), with wipes in the high air group that were 3.6 times higher than wipes in the low air group (p=0.006). Building category was a marginally significant, univariate predictor of PentaBDEs in surface wipes (p=0.06), with wipes from Other buildings that were 3.5 times higher than those from Building A (p=0.04), and 2.7 times higher than those from Building B (p=0.03). Surface wipes from Building A and Building B were not significantly different (p=0.61). Temperature, humidity, new carpet installed within the last year, and age of computer (<1-2 years vs. 3-6 years) were not predictors of PentaBDEs in surface wipes.

Table 3. Predictors of PentaBDEs in Office Wipes (pg/cm²).

Univariate Models	n	βa (SE)	p-value	R2
Dust (ng/g)
continuous	31	0.00004 (0.00001)	0.003	0.27
Air (pg/m3)
continuous	31	0.0006 (0.0001)	0.0003	0.37
Temperature(°C)
continuous	31	-0.035 (0.18)	0.85	<0.01
Building			0.06	0.18
Other	8	1.25 (0.56)	0.04
Building B	17	0.25 (0.50)	0.61
Building A	6	ref
Dust (tertiles)			0.001	0.38
High	10	1.36 (0.41)	0.002
Med	11	-0.18 (0.40)	0.66
Low	10	ref
Air (tertiles)			0.007	0.30
High	10	1.28 (0.44)	0.006
Med	11	0.001 (0.42)	1.00
Low	10	ref
Multivariable Models	n	βa (SE)	p-value	R2
Model 1			0.003b	0.45
Dust			0.04
High	10	0.92 (0.50)	0.08
Med	11	-0.25 (0.45)	0.58
Low	10	ref
Air			0.25
High	10	0.73 (0.50)	0.16
Med	11	0.036 (0.45)	0.94
Low	10	ref
Model 2			0.001b	0.50
Dust			0.002
High	10	1.13 (0.39)	0.008
Med	11	-0.40 (0.42)	0.35
Low	10	ref
Building			0.07
Other	8	1.16 (0.48)	0.02
Building B	17	0.57 (0.45)	0.22
Building A	6	ref

^aBeta coefficients represent the change in natural log transformed PentaBDE measurements in office wipes relative to the reference group for categorical variables, and per unit change for continuous variables.

^bGlobal p-value.

The results of two multivariable regression models predicting PentaBDEs in surface wipes are presented in Table 3. The first model (Model 1) includes categorical variables for PentaBDEs in dust and air, the two strongest univariate predictors of PentaBDEs in surface wipes. This model explained 45% of the variation in surface wipes (p=0.003), however beta estimates for both variables decreased substantially, such that air was no longer statistically significant.

The second multivariable model (Model 2) shown in Table 3 includes the categorical variables for dust and building as predictors of PentaBDEs in surface wipes. Since building category was a significant predictor of PentaBDEs in air independent of dust, building is likely a surrogate for factors that affect PentaBDEs in air but not dust (e.g. ventilation). Overall, Model 2 was statistically significant and predicted 50% of the variation in surface wipes (p=0.0009). PentaBDE dust tertiles were significant predictors of PentaBDEs in surface wipes (p=0.008), and building category was a marginally significant predictor of PentaBDEs in surface wipes (p=0.07). Including building category in this model explained an additional 12% of the variation in surface wipes compared to the univariate dust model.

3.3 Carpeting as a Predictor of PentaBDEs in Dust

Having carpet installed within one year prior to sample collection (“new” carpet) was a significant predictor of PentaBDEs in office dust. Within Building B, where all offices with carpeting less than one year old were located (5 of 17 offices), concentrations among those with older carpet were 3.4 times higher than those that had new carpet (p=0.04).

3.4 Predictors of BDE209 in Dust and Surface Wipes

Since BDE209 was detected in 39% of air samples and BDE209 measurements were available for only 14 of the 31 surface wipe samples, data analyses evaluating associations between different sampling media were limited. Among the 14 participants with detected concentrations, BDE209 in surface wipes and in office dust were significantly correlated (Spearman r=0.69, p=0.007).

Regression models were used to evaluate univariate predictors of BDE209 in both office dust (Table 4A) and surface wipes (Table 4B). Having carpet installed within one year prior to sample collection (“new” carpet) was a significant predictor of BDE209 in dust, specifically within Building B where all offices with carpeting less than one year old were located. BDE209 concentrations in dust among those with older carpet were 2.6 times higher than those that had new carpet (p=0.006). BDE209 in dust and new carpet were not significant predictors of BDE209 in office wipes, although these analyses were limited by small sample size (n=14). Temperature, humidity, age of computer (<1-2 years vs. 3-6 years), and building category were not predictors of BDE209 in office dust or surface wipes.

Table 4A. Predictors of BDE209 in Office Dust (ng/g).

Univariate Models	n	βa (SE)	p-value	R2
Temperature (°C)
continuous	31	0.23 (0.17)	0.19	0.06
New Carpetb
No	12	0.98 (0.31)	0.01	0.40
Yes	5	ref
Age of Computer
3-6 years	20	0.016 (0.42)	0.97	<0.01
<1 - 2 year	10	ref
Building			0.83	0.01
Other	8	0.36 (0.59)	0.54
Building B	17	0.19 (0.52)	0.71
Building A	6	ref

Table 4B. Predictors of BDE209 in Office Surface Wipes (pg/cm2)
Univariate Models	nc	βa (SE)	p-value	R2
Dust (ng/g)
continuous	14	0.00001 (0.00001)	0.32	0.08
Temperature (°C)
continuous	14	-0.43 (0.39)	0.30	0.09
New Carpetb
No	5	-0.42 (0.96)	0.68	0.03
Yes	4	ref
Building			0.41	0.15
Other	3	-0.11 (1.21)	0.93
Building B	9	-1.09 (1.04)	0.32
Building A	2	ref

^aBeta coefficients represent the change in natural log transformed BDE209 measurements in office dust and wipes relative to the reference group for categorical variables, and per unit change for continuous variables.

^bNew carpet installed in office within past year. Restricted to Building B as all offices with new carpeting were located in this building.

^cBDE209 not reported in 17 office wipe samples due to low recovery of standard.

3.5 Correlation of Handwipes with Office Air, Dust, and Surface Wipes

PBDE measurements in serum and handwipe samples collected from participants in this study have been presented elsewhere (Watkins et al. 2011). PentaBDEs in handwipes collected in the office environment were similarly correlated with PentaBDEs in office dust (r=0.34, p=0.06), surface wipes (r=0.32, p=0.08), and air (r=0.31, p=0.08). BDE209 measurements in handwipes were also similarly correlated with BDE209 in office dust (r=0.33, p=0.07) and in office wipes (r=0.42, p=0.14).

3.6 Measured vs. Predicted Dust and Surface Wipe Concentrations

In Figure 3, measured geometric mean concentrations of each PBDE congener in floor dust (panel A) and surface wipes (panel B) are plotted against the predicted value for each congener. In floor dust, most lower brominated congeners were on or slightly above the dashed 1:1 line, meaning measured dust values were lower than predicted based on K_oa and the concentrations in air, and higher brominated congeners were below the dashed line, meaning measured dust values were higher than predicted based on K_oa and the concentrations in air.

Figure 3.

Measured versus predicted geometric mean concentrations of BDE congeners in a. floor dust and b. surface wipes. Solid line is data regression line; dashed line represents the predicted 1:1 relationship based on equilibrium partitioning and certain parameters. For example, the 1:1 line for surface wipes assumes a high surface film thickness of 100 nm; a lower value would shift the line downward. Points above the dashed line signify measured values that are less than predicted. Points below the dashed line signify measured values that are greater than predicted.

Measured concentrations of almost all congeners in surface wipes were higher than predicted based on concentrations in air, K_oa, and other assumptions (e.g., the assumed film thickness). A smaller film thickness would shift the 1:1 line downward (Figure 3). The natural log of measured and predicted geometric mean concentrations were strongly correlated (Dust: Pearson r=0.97, R²= 0.95, p<0.0001; Surface Wipes: Pearson r=0.94, R²= 0.89, p=0.0005). Measured and predicted dust and surface wipe concentrations for each individual office are plotted in Figure 4 by congener. Although this plot demonstrates that in general measured and predicted values agree fairly well, there are many individual points that vary from the dashed 1:1 line of agreement by at least an order of magnitude.

Figure 4.

Measured vs. predicted concentrations of BDE congeners in individual a. floor dust and b. surface wipe samples. The dashed line represents the predicted 1:1 relationship based on equilibrium partitioning and certain parameters. For example, the 1:1 line for surface wipes assumes a high surface film thickness of 100 nm; a lower value would shift the line downward.

4. Discussion

We previously reported that PBDEs measured in handwipes were significantly correlated with PBDEs in serum, suggesting that exposure occurs via hands (Stapleton et al. 2012; Watkins et al. 2011). In the work presented here, we explore factors which may influence the transport of PBDEs from consumer products, into the office environment, and onto hand surfaces, ultimately affecting human exposure. Our results demonstrate that most PBDEs in air, floor dust, and settled surface dust in the same microenvironment are on average related in ways that are predictable based on the volatility of individual congeners and characteristics of the environment.

4.1 Associations between Sample Types

The consistency of associations between PBDE concentrations in air, dust, and surface wipes, suggests that the distribution of these compounds in indoor microenvironments may be driven by physical processes such as particle transport or partitioning, and that PBDEs in these compartments may have similar sources. PBDEs are likely released from products either through volatilization into surrounding air or physical abrasion of particles into surrounding dust, and then partition from air to floor or surface dust, or vice versa. However, this process may differ by PBDE congener, as dust concentrations of the relatively non-volatile BDE209 were much higher than would be expected based on partitioning. Possibly BDE209 is distributed in the indoor environment via particles, explaining the association between surface wipes and floor dust, and lower levels of BDE209 in air. In addition, there may be kinetic constraints that prevent BDE209 concentrations in dust and air from reaching equilibrium.

Correlations between sample types varied by building category, which also predicted PentaBDEs in both air and surface wipes. Differences in ventilation could be one possible explanation for this variation between the three building categories. Building A was a newly constructed building that had a centralized HVAC system and windows that did not open. In contrast, Building B was an older building in which each office had an individually controlled heating and cooling unit and windows that could be opened. The types of heating and ventilation systems included in the Other building category were not recorded by researchers, but most buildings were older and not known to have undergone recent renovations. Increased ventilation could potentially lead to decreased PBDEs in office air as room air may be mixed with outside air. However, removing PBDEs from office air could also result in the partitioning of PBDEs in room surface films into air to reestablish equilibrium (Xu et al. 2010). In this study, the lowest concentrations of PentaBDEs in air were seen in Building A, a new building that may not have yet developed a significant reservoir of PBDEs in surface films and with a modern HVAC system.

In Building B, the installation of new carpet within the year prior to sampling was associated with decreased concentrations of PentaBDEs and BDE209 in dust. Older carpet may be a source of PBDEs or a reservoir for dust and PBDEs that have accumulated over time. Replacing the carpet likely removed this reservoir, resulting in lower concentrations of PBDEs in office dust. The association between building category and PentaBDEs in air and surface wipes may also be partially explained by potential PBDE reservoirs in offices. Building A was newly constructed, including new office furniture, carpet, and other furnishings, approximately 18 months prior to sample collection and after PentaBDEs were phased out of use in commercial products in the US in 2004. Building A likely had few reservoirs with accumulated dust or surface films, and had the lowest levels of PentaBDEs in air, dust, and surface wipes. Other buildings, which had not been recently renovated and may have had substantial reservoirs for accumulated dust, surface films, and PentaBDEs, also had the highest levels of PentaBDEs in air, dust, and surface wipes.

4.2 Measured and Predicted PBDE concentrations in Dust and Surface Wipes

Measured concentrations of PBDEs in dust and surface wipes were highly correlated with concentrations predicted based on PBDEs in air and congener-specific K_oa values. On average, agreement between measured and predicted dust concentrations was better for lower brominated congeners (BDE28/33 through BDE100), which have a lower K_oa, compared to higher brominated congeners (BDE153, 154, and 209), which have a higher K_oa. BDE209, which is completely brominated and has the highest K_oa (i.e. least volatile), had the worst agreement between measured and predicted concentrations of all congeners. In addition, measured dust concentrations of congeners with a higher K_oa, such as BDE209, tended to be higher than predicted (Figure 3), possibly suggesting that equilibrium is limited by kinetics for these congeners. In addition, as they do not readily volatilize into surrounding air, these congeners may instead be released from sources into dust via physical weathering or abrasion (friability) as suggested by a recent study using electron microscopy images of dust (Webster et al. 2009).

Weschler and Nazaroff used SVOC concentrations in air and dust from multiple studies to calculate predicted concentrations in dust, but BDE209 was not included (Weschler and Nazaroff 2010). Our findings are consistent with their work as they reported an overall R² of 0.76 for measured versus predicted concentrations for all SVOCs, while we report a slightly higher R² of 0.95 and 0.89 for PBDEs in floor dust and surface dust respectively. Our findings are also consistent with those reported by Zhang et al. which suggest partitioning between air and dust among lower brominated congeners and release of higher brominated congers from sources through abrasion instead of volatilization (Zhang et al. 2011).

4.3 Implications for Exposure Assessment Methodology

A goal of this work was to inform methods used to assess exposure to PBDEs in the indoor environment. As collecting air samples for PBDE analysis is both costly and time consuming, alternative methods of characterizing PBDE exposure through this medium would be useful. The research presented here demonstrates that most PBDEs in air, floor dust, and settled surface dust in the same microenvironment are on average related in ways that are predictable based on the volatility of individual congeners and characteristics of the environment. However, predicting PBDEs in dust using the methods described here and by Weschler and Nazaroff are not always accurate when applied to individual data points (Figure 4), and cannot substitute for more detailed exposure assessment. Furthermore, they do not work equally well for all congeners, particularly the close to nonvolatile BDE209 (Weschler and Nazaroff 2010). For example, if PBDEs are measured only in air, exposure to BDE209 may be underestimated as the relationship of this congener in air to dust is not reliable, with concentrations in dust often higher than expected. In addition, it is difficult to make assumptions regarding the influences of ventilation and other environmental characteristics on the relationship between PBDEs in air and dust in a specific environment.

The pathway by which people are exposed to PBDEs in the indoor environment is also a factor in examining exposure assessment methods. If exposure occurs through inhalation or deposition from air to skin, dust concentrations may not represent potential exposure if air and dust are not at equilibrium. However, other studies have suggested that inhalation is only a minor exposure route compared to incidental dust ingestion and diet (Allen et al. 2007; Lorber 2008). In our previous work, we found correlations between PBDEs measured in handwipes and in serum, suggesting that exposure occurs via hands either through incidental ingestion or dermal absorption (Stapleton et al. 2012; Watkins et al. 2011). PBDEs may attach to hands and other exposed skin either through contact with dust and sources of PBDEs or potentially via deposition from air to skin (Weschler and Nazaroff 2008).

PentaBDEs in handwipes collected from study participants in their office environment were equally correlated with PentaBDEs in office air, dust, and surface wipes. Handwipes are presumably a measure of personal exposure to PBDEs and were significantly correlated with PentaBDE body burden in this study population (Watkins et al. 2011). However, serum PentaBDE measurements were not significantly correlated with measurements of PentaBDEs in the office environment, possibly due to our small sample size and limited power to detect weak associations. This is consistent with our recent finding that PBDEs in dust from the home environment are the strongest predictor of PentaBDE body burden in this population (Watkins et al. 2012). Consequently, we are unable to establish whether air, dust, or surface wipes are the best measure of exposure to PBDEs in the indoor environment from the data presented here. Similar research should be performed in the home environment to determine which exposure assessment method is most relevant to PBDE body burden.

4.4 Comparisons with Other Studies

The correlation between PentaBDEs in collocated air and dust samples reported here (r=0.60 for all buildings) is similar to those seen for predominant individual PentaBDE congeners from Michigan offices (r= 0.59 - 0.92) (Batterman et al. 2010), as well as individual PBDE and PCB congeners in offices in Toronto Canada (r=0.55 - 0.86) (Zhang et al. 2011). A similar correlation was also found between PentaBDEs in bedroom dust and air (r=0.62) in Boston, but a significant association was not found between dust and air from main living areas of homes (Allen et al. 2008). A recent study in Denmark also did not see a correlation between PBDEs in residential air and dust (Vorkamp et al. 2011).

4.5 Limitations

Our small sample size limited the number of variables we were able to evaluate simultaneously, as well as our ability to detect weak associations and address potential confounding. Associations between BDE209 concentrations in the different sampling media were especially difficult to explore. BDE209 measurements were available in 45% of surface wipes, and due to high limits of detection, detected in only 39% of air samples. Only two participants had detectable BDE209 levels in both air and surface wipe samples. Our population was also a relatively homogeneous sample, with over half of participants working in offices located in Building B. This may limit the generalizability of this study to populations in other types of office buildings but shouldn't affect the internal validity of our study.

Our cross-sectional study design could also be a limitation, as we collected air, dust, and surface wipes samples over one short period of time. Although we would expect PBDEs in floor and surface dust to be relatively stable (Allen et al. 2008), we have no comparable data on how much PBDEs in air fluctuate over time. In the present study, air samples were collected over one week in the middle of winter. As office temperatures and ventilation likely change over seasons, the relationship between PBDEs in dust, air, and surface wipes in the office microenvironment may change. Measuring ventilation rates in offices would have been ideal given the observed differences by building category, but we were unable to collect this information. PBDEs in dust, air, and on surfaces may also differ according to usual cleaning schedules and practices. Arrangements were made through each building's facilities management office to ensure that study offices were not vacuumed during the sampling week. However, the amount of time since the last previous cleaning for each office is unknown.

There are many uncertainties associated with calculating predicted concentrations of PBDEs in dust from concentrations in air, most notably the calculation of K_oa values for individual PBDE congeners. Weschler and Nazaroff compared calculated K_oa values to measured K_oa values and found that there were large discrepancies for BDE153 and BDE154 (Weschler and Nazaroff 2010). We used calculated K_oa values in our calculations as experimental values were not available for all BDE congeners and can vary a great deal across experiments (Weschler and Nazaroff 2008). This may be particularly important for BDE209. As the K_oa is used to calculate vapor and particle fractions of PBDEs in air, inaccuracies in the K_oa also lead to inaccuracies in estimated concentrations of PBDEs in the vapor phase. There is additional uncertainty in predicting PBDE concentrations in surface wipes as we measured the mass of PBDEs per unit surface area, which we then converted to a concentration based on various assumptions, most notably the thickness of the dust layer. Assuming a constant thickness across all collected samples introduces error into the measured value, and uncertainty into the comparison with predicted values.

5. Conclusions

The relationships between PBDEs in air, floor dust, and settled surface dust in the office environment are related to the K_oa of individual congeners, and are likely influenced by characteristics of the environment. In addition, surface wipe and dust samples provide comparable, quick and easy measures of PBDEs in an indoor environment that may be useful in exposure assessment. This research demonstrates that handwipes, which are associated with PentaBDE body burden (Watkins et al. 2011), are similarly correlated with office air, dust, and surface wipes, suggesting that exposure in the office environment may contribute to PentaBDE body burden among office workers.

Highlights.

• PentaBDEs in office air, floor dust and surface wipes were significantly correlated.

• PentaBDEs in both air and surface wipes varied significantly by building.

• Dust and surface wipe PBDE concentrations can be predicted by K_oa and air concentrations.

• Predicted and measured concentrations of BDE congeners were highly correlated.

• BDE209 was employed to test the limits of the equilibrium partitioning model.

Abbreviations

GC/ECNI-MS

gas chromatography/electron capture negative ionization mode - mass spectrometry

GFF

glass fiber filter

HVAC

heating, ventilation, and air conditioning

K_oa

octanol-air partition coefficient

K_pa

particulate-air partition coefficient

LOD

limit of detection

PBDE

polybrominated diphenyl ethers

PUF

polyurethane foam

SVOCs

semi-volatile organic compounds