Science of the Total Environment 443 (2013) 520–529

Contents lists available at SciVerse ScienceDirect

Science of the Total Environment

j ourna l homepage: www.e lsev ie r .com/ locate /sc i totenv

Canadian House Dust Study: Population-based concentrations, loads and loading ratesof arsenic, cadmium, chromium, copper, nickel, lead, and zinc inside urban homes

Pat E. Rasmussen a,b,⁎, Christine Levesque a, Marc Chénier a,b, H. David Gardner a,b,Heather Jones-Otazo c, Sanya Petrovic d

a Exposure and Biomonitoring Division, Healthy Environments and Consumer Safety Branch, Health Canada, 50 Colombine Driveway, Ottawa, ON, Canada K1A 0K9b Department of Earth Sciences, University of Ottawa, 140 Louis Pasteur, Ottawa, ON, Canada K1N 6N5c Regions and Programs Branch, Health Canada, 180 Queen Street West, Toronto, ON, Canada M5V 3L7d Contaminated Sites Division, Healthy Environments and Consumer Safety Branch, Health Canada, 269 Laurier Ave West, Ottawa, ON, Canada K1A 0K9

H I G H L I G H T S

► The Canadian House Dust Study is a nationally representative random sample.► Dust was collected from 1025 urban homes from 13 cities with population>100,000.► Typical indoor dust and metal loading rates and metal concentrations are reported.► Dust mass is the overriding influence on metal loadings and loading rates.► This population-scale study contributes to defining the exposome.

⁎ Corresponding author at: Exposure and Biomonitoringand Consumer Safety Branch, Health Canada, 50 ColombineK1A 0K9. Tel.: +1 613 941 9868.

0048-9697/$ – see front matter. Crown Copyright © 20http://dx.doi.org/10.1016/j.scitotenv.2012.11.003

a b s t r a c t

a r t i c l e i n f o

Article history:Received 11 June 2012Received in revised form 30 October 2012Accepted 1 November 2012Available online 5 December 2012

Keywords:MetalsIndoor environmentExposure assessmentTobacco smokeHouse ageVacuum sampling

The Canadian House Dust Study was designed to obtain nationally representative urban house dust metalconcentrations (μg g−1) and metal loadings (μg m−2) for arsenic (As), cadmium (Cd), chromium (Cr),copper (Cu), nickel (Ni), lead (Pb), and zinc (Zn). Consistent sampling of active dust of known age and prov-enance (area sampled) also permitted the calculation of indoor loading rates (mg m−2 day−1 for dust andμg m−2 day−1 for metals) for the winter season (from 2007 to 2010) when houses are most tightly sealed.Geomean/median indoor dust loading rates in homes located more than 2 km away from industry of anykind (9.6/9.1 mg m−2 day−1; n=580) were significantly lower (pb .001) than geomean (median) dustloading rates in homes located within 2 km of industry (13.5/13.4 mg m−2 day−1; n=421). Proximity toindustry was characterized by higher indoor metal loading rates (pb .003), but no difference in dustmetal concentrations (.29≥p≤ .97). Comparisons of non-smokers' and smokers' homes in non-industrialzones showed higher metal loading rates (.005≥p≤ .038) in smokers' homes, but no difference in dustmetal concentrations (.15≥p≤ .97). Relationships between house age and dust metal concentrationswere significant for Pb, Cd and Zn (pb .001) but not for the other fourmetals (.14≥p≤ .87). All sevenmetals,however, displayed a significant increase in metal loading rates with house age (pb .001) due to the influ-ence of higher dust loading rates in older homes (pb .001). Relationships between three measures of metalsin house dust – concentration, load, and loading rate – in the context of house age, smoking behavior andurban setting consistently show that concentration data is a useful indicator of the presence of metalsources in the home, whereas dust mass is the overriding influence on metal loadings and loading rates.

Crown Copyright © 2012 Published by Elsevier B.V. All rights reserved.

1. Introduction

The question “What constitutes an appropriate dust sample”posed by Paustenbach et al. (1997) captures a variety of concernsabout the difficulty in obtaining a standardized, representative

Division, Healthy EnvironmentsDriveway, Ottawa, ON, Canada

12 Published by Elsevier B.V. All rig

house dust sample (Sutton et al., 1995; Butte and Heinzow, 2002),the heterogeneity of vacuum samples (Aurand et al., 1983), and thelarge inherent variability in dust wipe data (Caplan, 1983). Dustmetal concentrations and dust metal loadings are different measuresof indoor environmental metal levels: dust concentration is a mea-sure of the amount of metal in a given quantity of dust, in units ofμg g−1 or ppm, while loading (μg m−2) is mainly a function of thequantity of dust on the surface sampled and to a lesser extent theconcentration of metal in dust (Sutton et al., 1995). Both loadingand concentration measures are important for understanding and

hts reserved.

521P.E. Rasmussen et al. / Science of the Total Environment 443 (2013) 520–529

assessing the significance of exposure to house dust (Paustenbachet al., 1997), and they provide different but complementary types ofinformation. Loading, which is strongly influenced by housekeepingbefore sample collection (Sutton et al., 1995; Bell et al., 2010), is con-sidered to be an appropriate index of potential exposure (Lioy et al.,1992; Lanphear et al., 1998; Dixon et al., 2009). Concentration infor-mation is useful for identifying residential metal sources and specia-tion (Sutton et al., 1995; Maclean et al., 2011; Beauchemin et al.,2011; Walker et al., 2011), and for direct comparisons of indoordust with outdoor media such as soil and street dust (Rasmussen etal., 2001, 2008; Zota et al., 2011).

Information on everyday exposures to metals in house dust at thepopulation scale contributes to defining the exposome, which is themeasure of all the exposures of an individual in a lifetime and howthose exposures relate to disease (CDC, 2012). Representative base-line information about exposures to chemicals in house dust of thegeneral urban population has been identified as a data gap in residen-tial risk assessments. Arsenic (As), cadmium (Cd), chromium (Cr),copper (Cu), nickel (Ni), lead (Pb), and zinc (Zn) were selected forthe present study as they commonly occur in contaminated sites inCanada (Franz Environ Inc., 2005; Rasmussen and Gardner, 2008),and urban baseline estimates for these elements are needed as apoint of comparison for residential assessments on or near contami-nated sites. Information about these elements is also required forcalculations of population-based Estimated Daily Intakes used inguideline development.

The Canadian House Dust Study (CHDS) was designed to providenationally representative population-based baseline survey informa-tion for typical urban Canadian homes (Rasmussen et al., 2011). In addi-tion to reporting loading and concentration measurements, the presentstudy also reports loading rates, which are rates of deposition expressedin units of mass per unit area per unit time (mg m−2 day−1 for dustand μg m−2 day−1 for metals). Hogervorst et al. (2007) concludedthat metal loading rates should be incorporated into risk assessments,based on observed significant increases in biomarkers of Cd and Pbexposures in adults associated with increased Cd and Pb loading ratesin the vicinity of smelters. There are various methods to determinedust and metal loading rates, many of which involve the collection ofdust which has passively accumulated on an exposed surface area(e.g., plates, beakers, Petri dishes) over a specified period of time(Aurand et al., 1983; Edwards et al., 1998; Meyer et al., 1999; Siefertet al., 2000; Khoder et al., 2010). In the present study, trained techni-cians followed a vacuum sampling protocol designed to collect a consis-tent composite (whole-house) sample of readily accessible “fresh dust”of known age, from dry floor surfaces of known dimensions, to achievea complete set of elemental concentration, load, and loading ratemeasurements.

2. Method

2.1. Sample collection

The goal of the CHDS is to obtain a population-based urban base-line estimate which is representative for Canada, not individual citiesor provinces, as described previously (Rasmussen et al., 2011). Briefly,the statistical design required a stratified random sample of 1025Canadian single family dwellings from a total of 107 disseminationareas (DAs) across 13 Canadian cities having a population greaterthan 100,000 (Rasmussen et al., 2011). Sampling was conducted inthe winter season from January 2007 to March 2010. A compositedust sample was collected in each home using the Pullman Holtmodel 102 ASB-12PD vacuum sampler specified to capture 99.97%of all particles 0.3 μm and larger. The sampling method was basedon the Verein Deutscher Ingenieure protocol (VDI, 2001) for samplingfresh or “active” house dust, defined as dust of known age thatis readily accessible on floor surfaces of living spaces in the home.

A composite sample was collected by vacuuming accessible floor sur-faces in all living areas and connecting rooms, excluding potentiallywet areas (kitchen, garages, workshops and unfinished sections ofbasements) to protect the integrity of the sample. A light touch ofthe vacuum nozzle was used to collect only surface dust from carpetsand avoid dust of unknown age found in joints and cracks in flooring,and areas where the householder does not vacuum on a regular basis(VDI, 2001). To keep costs within the budget, vacuuming was accom-plished in a single household visit rather than the two visits pre-scribed by the VDI protocol. Participants were requested to abstainfrom cleaning floor surfaces for a period of 7 days before the sched-uled fresh house dust sampling, and the date of the last cleaningwas documented.

2.2. Sample processing

Dust was transferred quantitatively from the vacuum bags toweighing paper, with careful avoidance of loss or contamination duringprocessing as described previously (Rasmussen et al., 2011). The vacu-um bag was cut open to allow the dust to air-dry (minimum 24 h),followed by manual removal of pet and human hair and large particles.Each sample was sieved to fine (b80 μm) and coarse (80–300 μm) sizefractions using a sieve shaker (W.S. Tyler RO-Tap RX-94) and stainlesssteel sieves (W.S. Tyler Cat. Nos. 2451, 5209, 5205). The sieved dustsamples were weighed (CP2202S Sartorius Balance; readability 0.01 g)and stored frozen in amber glass bottles with PTFE-lined caps untilsubmission for analysis. The masses of the fine (b80 μm) and coarse(80–300 μm) dust fraction were combined to calculate the total dustload (b300 μm) collected in each home.

2.3. Elemental analysis

The analytical approaches were selected to provide a quantitativedetermination of total metal content in the dust samples. Instrumen-tal Neutron Activation Analysis (INAA) is a direct solid-sample analyt-ical technique which provides adequate detection limits for certainelements (e.g. As and Cr), but other elements require a strong acid di-gestion capable of putting all species of the element into solution withequal efficiency, whether the matrix is soil or dust. Total concentra-tions of Pb, Zn, Cu, Cd, and Ni were determined in 0.5 g subsamplesof the b80 μm dust size fraction by Actlabs Inc. (Ancaster ON, Canada)using a 4-acid digestion (HF, HClO4, HNO3 and HCl) followed by eitherInductively-Coupled Plasma Mass Spectrometry (ICP–MS) or OpticalEmission Spectrometry (ICP OES), as appropriate for the concentra-tion range. Arsenic was determined in 0.5 g subsamples by INAAwith the exception of 62 samples which were determined using theabove acid digestion/ICP method due to insufficient mass for INAA.Chromium was determined by INAA for samples ≥10 μg g−1 Cr,and by acid digestion/ICP–MS for samples containing b10 μg g−1 Cror lacking adequate mass for INAA. Sample duplicates, in-house con-trols, and Certified Reference Materials (CRMs) were inserted (blind)after every 10 samples, for evaluation of reproducibility and extractionefficiency. The following National Institute of Standards & Technology(NIST) CRMswere selected to represent a wide range of total elementalconcentrations in relevant matrices: indoor vacuum dust (NIST 2583and NIST 2584), urban soil (NIST 2586, NIST 2710 and NIST 2711),and sediment (NIST 2702). All results are summarized for the individualNIST CRMs in the Supplementary information (Table S1).

2.4. Questionnaire and data analysis

During the house visit, technicians completed a short question-naire to assist the interpretation of dust loading and composition(e.g. heating practices, dates of construction and renovation activity,proximity to industry, hobbies, number of residents and pets). Outof 1025 homes sampled, there were 3 cases where the dimensions

522 P.E. Rasmussen et al. / Science of the Total Environment 443 (2013) 520–529

of the sampled areas were not measured and 21 cases where theelapsed time since cleaning was unknown. Therefore loadings couldbe calculated for 1022 homes and loading rates could be calculated for1001 homes. Out of 1001 homes, participants in 559 homes adheredto the cleaning protocol (i.e. time elapsed since cleaning=7±1 days)and for the remaining 442 homes the time elapsed since cleaningranged from 1 day to >30 days. House age was documented for 1023homes (unknown for 2 homes) and the average date of constructionwas 1965±27 (mean±sd). Surface characteristics of the area sampled(e.g., carpeted, hardwood, tile, linoleum)were recorded for 1014homes.Data analyses were conducted using SPSS Statistics v.19 and Excel 2007with the statistical add-in Analyze-It. Distributions were evaluated fornormality (Shapiro–Wilk) and homogeneity (Levine); Spearman rankand Mann–Whitney U tests were used where non-parametric analyseswere required. Results have not been weighted geographically norcorrected for water content (typically 1–3%, Rasmussen et al., 2011);laboratory duplicates have been averaged; and results below limit ofdetection (LOD) have been assigned a value of 0.5 LOD. For the metalloading calculations the elemental concentrations in the fine fraction(b80 μm) were applied to the total dust load (b300 μm), based on apreliminary study showing no significant difference (pb .05; 95% CI)between the two size fractions for the range of concentrations(ppm to %) in the present study.

3. Results and discussion

This paper reports on three measures of metals in house dust –

concentration, load, and loading rate – derived from a single vacuummethod. Other types of dust samples collected in the Canadian HouseDust Study (CHDS) included wipes (McDonald et al., 2011) andhousehold vacuum samples (Fan et al., 2010). Wipe samples collectedin accordance with the USEPA/HUD methodology (McDonald et al.,2011) provide a measure of room-to-room variability in metalloading, but sensitivity proved to be inadequate for the presentpopulation-based study, due to the large proportion of wipe resultsbelow limits of quantification (McDonald et al., 2011). Samples fromhousehold vacuum bags can provide a cost-effective measure of con-taminant concentrations (Colt et al., 2008; Fan et al., 2010). For thepurpose of the present study, the chief disadvantage of householdvacuum samples is the lack of information on dust age and prove-nance (area sampled) required for calculating dust and elementalloading rates.

Table 1Elemental concentrations (μg g−1) in fresh vacuum dust (b80 μm) from a nationally reprbackground values included for comparison.

As (μg g−1) Cd (μg g−1) Cr (μg g−1)

Limit of detection 0.1 0.1 0.5Geometric mean 7.7 3.8 101Mean±sd 13.1±14.3 6.0±11.1 117±112Range (min–max) 0.1–153 0.5–223 0.5–2930Percentiles

5 0.3 1.2 49.210 3.0 1.5 58.025 5.2 2.2 74.950 (median) 9.1 3.5 99.075 15.3 5.9 13690 26.7 11.1 17795 40.6 17.2 21497.5 54.5 26.0 26598 59.3 29.7 284NGBa 5.8 0.1 62

a NGB median concentration of Cd in b2 μm fraction glacial till (n=1878) from Kettlesfraction glacial till (n=7398 observations) from Rencz et al. (2006).

3.1. Nationally representative dust metal concentrations

The statistical summary of baseline As, Cd, Cr, Cu, Ni, Pb, and Znconcentrations in fresh house dust, representative of urban Canadiansingle family homes, is presented in Table 1. As none of the 7 elementaldatasets are lognormally distributed, they are most appropriatelysummarized using median and percentile values. Significant positiverelationships between house age and metal concentration are ob-served for Pb, Cd and Zn (Spearman rho, 2-tailed pb .001) but notfor the 4 other elements (discussed in detail later). CHDS results inTable 1 are comparable with international urban house dust concen-trations published since 1996 (Turner, 2011). CHDS median values(Table 1) fall within median/geomean ranges of Turner's compila-tion for Pb (26–662 μg g−1; 9 studies), Zn (419–1999 μg g−1;8 studies), Cu (76–311 μg g−1; 9 studies), and Cd (0.9–4.4 μg g−1;7 studies), whereas CHDS median values are at the high end of thecited range for Ni (16–54 μg g−1; 7 studies), Cr (31–95 μg g−1; 7studies), and As (1.5–4.9 μg g−1; 2 studies). Comparisons are limit-ed by the small size of the cited urban studies, and by differencesin sample preparation, particularly where analytical methods arenot evaluated using CRMs. Although the authors of a large study(n=264) in Syracuse, NY, USA advise that they did not use a “totaldigest” (Johnson et al., 2009), median concentration values for Pb,Cu and Ni (117, 134, 23.6 μg g−1, respectively) in the Syracusestudy are similar to CHDS medians (Table 1). Zinc is the exception:the median concentration value of 1385 μg g−1 in the Syracusestudy is almost double the CHDS median 725 μg g−1 (Table 1).

Internationally, a variety of semi-quantitative digestion methods arebeing employed for the determination of metals in house dust, some-times resulting in order-of-magnitude variations in median/mean dustmetal concentrations amongst different studies (Rasmussen, 2004a;Butte and Heinzow, 2002). Evaluation of analytical results for NISTCRMs (Fig. 1; Table S1) confirms that recoveries in the present studyare quantitative (i.e. between 90% and 110% of the NIST certifiedvalue) for a wide range of elemental concentrations in a variety of soiland dust matrices. Technical and operational difficulties in obtainingquantitative or “total” metal extractions are recognized (Hassan et al.,2007), but the additional effort reaps significant benefits in minimizinguncertainty, particularly when the results are used as the denominatorin bioaccessibility/bioavailability calculations (Rasmussen et al., 2011);for minimizing bias when evaluating contrasting matrices such as in-door dust and garden soil (Rasmussen et al., 2001, 2008; Beaucheminet al., 2011; Walker et al., 2011; Zota et al., 2011); and for direct

esentative sample of urban Canadian homes (n=1025); NGB = natural geochemical

Cu (μg g−1) Ni (μg g−1) Pb (μg g−1) Zn (μg g−1)

0.2 0.5 0.5 0.2217 73.3 119 749279±361 102±188 210±446 833±47024.0–4880 17.3–2300 14.2–7800 144–6630

97.8 31.8 42.2 383115 36.9 48.5 453147 46.9 66.3 558199 62.3 100 725291 94.5 173 964449 200 357 1290660 322 760 1627851 501 1331 20371025 613 1554 212419 16 8 34

and Shilts (1994); NGB median concentrations of As, Cr, Cu, Ni, Pb and Zn in b63 μm

Arsenic

1

10

100

1000

1000100101

Certified value (mg/kg)

10001001

Certified value (mg/kg)

Obs

erve

d va

lue

(mg/

kg)

1

100

1000

Obs

erve

d va

lue

(mg/

kg)

Avg Recovery = 103% (sd = 9%)Rsq = 0.9999

NIST 2583 NIST 2586

NIST 2584

NIST 2702 NIST 2711

NIST 2710

Cadmium

0

1

10

100

1001010.1

Certified value (mg/Kg)

Obs

erve

d va

lue

(mg/

kg)

NIST 2583NIST 2584

NIST 2586

NIST 2702

NIST 2710

NIST 2711

Avg Recovery = 108% (sd = 11%)Rsq = 0.9979

Chromium Avg Recovery = 95% (sd = 5%)Rsq = 0.9988

NIST 2586

NIST 2583

NIST 2584

NIST 2702

NIST 2710

NIST 2711

Lead

10

100

1000

10000

100000

10 100 1000 10000 100000

Certified value (mg/Kg)

Obs

erve

d va

lue

(mg/

kg)

Avg Recovery = 96% (sd = 3%)Rqs = 0.9999

NIST 2710

NIST 2583

NIST 2584

NIST 2586NIST 2702

NIST 2711

Fig. 1. Observed concentrations for arsenic (As), cadmium (Cd), chromium (Cr) and lead (Pb) determined in 6 NIST certified reference materials plotted against certified values.Details provided for 7 study elements in Supplementary information (Table S2).

523P.E. Rasmussen et al. / Science of the Total Environment 443 (2013) 520–529

comparisons of acid digestion/spectroscopic results with solid-sampleINAA and XRF instrumental measurements (Rasmussen et al., 2007;Niu et al., 2010). Where quantitative data are not available, inter-study comparisons of semi-quantitative measurements are facilitatedby the practice of evaluating and reporting traceable certified referencematerials (such as the NIST CRMs in Fig. 1 and Table S1).

3.2. Potential contributions from natural and anthropogenic sources

Incorporation of “baseline” metal concentrations into risk assess-ments introduces the question of the contribution of natural sourcesto total dust metal concentrations. Particle-bound “geogenic metals”(from natural sources) enter the home via soil and windblown dust,and accumulate in settled dust in combination with anthropogenicmetals from a variety of indoor and outdoor sources. Median estimatesof natural geochemical background (NGB) concentrations were derivedfrom extensive regional till surveys (Rencz et al., 2006; Kettles andShilts, 1994) and are included in Table 1 to provide perspective on thepotential contribution of natural sources at the population level. Com-paring medians for Cd, Zn, Pb and Cu (Table 1), NGB concentrationsare at least an order-of-magnitude lower than “total metal” concentra-tions in house dust, accounting for about 3% of total Cd (0.1 μg g−1 forNGB Cd/3.5 μg g−1 for total Cd), 5% of total Zn (34/725), 8% of total Pb(8/100) and 10% of total Cu (19/199), suggesting that anthropogenicsources of these elements dominate the indoor urban environment.Me-dianNGB values for As, Cr and Ni (Table 1) represent amore substantialfraction of “total metal” concentrations: 64% of total As (5.8 μg g−1

NGB/9.1 μg g−1 total); 63% of total Cr (62/99); and 26% of total Ni(16/62), suggesting that anthropogenic sources of these elements areless dominant in the indoor environment than for Cd, Zn, Pb and Cu.Note that these median ratios do not necessarily apply to an individual

home, as anthropogenic contributions vary widely amongst homes in agiven neighborhood, and even amongst different rooms in the samehouse (Rasmussen, 2004b; Beauchemin et al., 2011; Walker et al.,2011); also natural sources are strongly influenced by site specific geo-logical factors (Rasmussen, 1998). However, at the population scale,thesemedian ratios provide insight into the general influence of anthro-pogenic activities on the contemporary urban indoor environment.

3.3. Duplicate uncertainty

Results of duplicate analyses are presented in Table 2. Duplicateuncertainties (U, defined as one standard deviation) are reported forconcentration ranges typically encountered for each element in thefine dust fraction (b80 μm). Duplicate uncertainty reflects a combina-tion of sample heterogeneity and analytical sources of variability. Byproviding a frame of reference for evaluating uncertainty, Table 2demonstrates that relative uncertainties for b80 μmdust samples com-pare favorably with relative uncertainties for homogenized CRMs.Observed relative uncertainties for CRMs (Table 2) include variabilitybetween analytical batches over the 4 yr sampling campaign, whichcontributes to differences between observed and certified CRM uncer-tainties (for example Pb and Cr in NIST 2584).

Previous research (Rasmussen et al., 2008) showed that smaller dustparticle fractions (e.g., b56 μm) have the advantage of greater homoge-neity and typically higher metal concentrations. In addition to theadvantage of sample homogeneity, finer dust fractions have greater rel-evance for childhood exposure estimates (Duggan and Inskip, 1985;Rodes et al., 2001) and particle transport and resuspension models(Edwards et al., 1998; Young et al., 2002). However, sieving to suchfine size fractions is not always feasible due to the small mass availablein typical “fresh dust” samples. Sampling all living areas (awhole-house

Table 2Relative uncertainty (U) of elemental analysis of fresh vacuum dust samples (b80 μm fraction) compared to relative U observed/certified for two NIST indoor dust CRMs (2583 and2584). NR = not reported on NIST certificate of analysis.

Arsenic Cadmium Chromium Copper Nickel Lead Zinc

Sample duplicatesConcentration range (μg/g) 3 to 12 1 to 8 50 to 100 100 to 300 25 to 100 10 to 150 400 to 950# duplicate pairs in range 56 104 54 75 86 94 72sd from duplicatesa 1.8 0.8 7.3 25 6.2 9.3 3.9Mean concentration of duplicates 7.2 4.1 75 185 52 77 656Relative U (sda/mean) 25% 20% 10% 14% 12% 12% 1%

Indoor dust CRMsNIST 2583

Observed relative Ub 23% 52% 12% 8% 9% 13% 6%Certified relative Uc 21% 51% 28% NR NR 8% NR

NIST 2584Observed relative Ub 21% 16% 13% 10% 12% 10% 8%Certified relative Uc 24% 11% 7% NR NR 1% 6%

a sd calculated from duplicates using equation from Synek (2008).b sd/mean based on analysis of 9–14 replicates of NIST CRMs.c Uncertainty/mean reported on NIST certificates of analysis, listed in Table S1.

524 P.E. Rasmussen et al. / Science of the Total Environment 443 (2013) 520–529

“composite” sample) in the CHDS yielded a sample mass approaching1 g inmost cases,making it possible to sieve to b80 μm,whichprovideda viable, homogeneous house dust sample with respect to elementalcontent (Table 2).

3.4. Dust loading and metal loading

Table 3 summarizes dust loadings (mg m−2) and metal loadings(μg m−2) for all CHDS homes with available area measurements(n=1022). Note that for most metals the loading data are not lognor-mally distributed (with the sole exception of Cd). Fine (b80 μm) andcoarse (80–300 μm) fractions collected in the present study were com-bined to calculate the loading values in Table 3, to facilitate comparisonswith loadings derived fromwipe samples, which trap a full range of dustparticle sizes (Butte andHeinzow, 2002).With few exceptions, the CHDSdust loading values are below 0.8 g m−2 (Table 3), lower than loadingsassociated with home intervention/remediation studies (Roberts et al.,2009), and comparable to floor wipe results reported by Johnson et al.(2009) in Syracuse, New York (range 42 to 2330 mg m−2; geomean311 mg m−2; n=488).

In general the distributions of metal loads obtained using vacuumsampling (Table 3) are consistent with population-based distributionsobtained using wipe sampling methods. Lichtenwalner (1992) reportedmedian values of Cr, Cu, and Pb (approx. 5, 5, and 10 μg m−2, respective-ly) in an evaluation of dust wipes in various indoor, outdoor, residential,and commercial locations. McDonald et al. (2011) reported 95th percen-tile values for As (7.3 μg m−2) and Cd (2.2 μg m−2) for interior wipevalues in a subset of 222 CHDS homes, which fall within the 90th to95th percentile ranges for As and Cd loading in Table 3. It is notablethat the 95th percentile for Pb (124 μg m−2) in Table 3 coincides with

Table 3Dust loading (mg m−2) and elemental loading (μg m−2) based on whole-house floor vacb300 μm particle size range to facilitate comparison with wipe sampling.

Dust (mg m−2) As (μg m−2) Cd (μg m−2) Cr

Geometric mean 79.7 0.6 0.3 8.Mean±sd 147±239 2.2±6.6 0.8±2.1 15Range (min–max) 2.0–4081 0.3×10−3–161 0.2×10−2–26.8 0.Percentiles

5 14.9 0.0 0.0 1.10 21.3 0.1 0.1 2.25 38.3 0.3 0.1 4.50 (median) 75.0 0.7 0.3 7.75 168 1.8 0.8 1690 314 5.0 1.7 3495 489 9.1 3.0 4897.5 711 13.3 5.0 6598 787 14.9 6.2 77

the upper break-point (125 μg m−2) in the distribution of wipe valuesin the CHDS subset (McDonald et al., 2011). Dixon et al. (2009) devel-oped a regression model, based on NHANES data (collected 1999 to2004), which predicts that at a Pb load of 12 μg ft−2 (~129 μg m−2) ap-proximately 5% of children would have blood lead levels >10 μg/dL.

A significant advantage of using whole-house vacuum samples tocalculate population-based loadings was that 100% of results exceededlimits of detection (LOD), even in the most lightly loaded homes(Table 3). In contrast, the wipe method yielded a large proportion ofresults below LOD in the CHDS subset (McDonald et al., 2011). Similarly,Galke et al. (2001) reported that a large proportion of post-interventionPb wipe results fell below typical LODs ranging from 1 to 25 μg ft−2

(11–269 μg m−2).The CHDS metal loadings in Table 3 provide useful baseline data for

comparison with loadings reported in homes impacted by contaminat-ed lands. For example, baseline medians for the CHDS are two- tofive-fold lower than median loadings reported for As (1.7 μg m−2), Cd(1.1 μg m−2), Pb (38 μg m−2) and Zn (288 μg m−2) in 55 homes lo-cated near mine waste (Zota et al., 2011). Cr loadings in homes locatedon or near Cr contaminated waste sites in New Jersey, USA ranged from10 to 120 μg m−2 using vacuum samples and 32.5 to 3200 μg m−2

using wipe samples (Lioy et al., 1992).Non-parametric analyses (Spearman rank) revealed strong corre-

lations between dust mass andmetal load for Cd, Cu, Cr, Ni, Pb and Zn(r=.79, .86, .89, .84, .80 and .92, respectively; pb .001). Correlationsbetween metal concentration and metal load are relatively weaker(r=.58, .17, .30, .25, .57 and .31, respectively) but still significant(pb .001). In the case of As, the correlation between dust mass andAs load (r=.74) is only marginally stronger than the correlation be-tween As concentration and As load (r=.70). The observation that

uum samples of representative urban Canadian homes (n=1022). Calculations used

(μg m−2) Cu (μg m−2) Ni (μg m−2) Pb (μg m−2) Zn (μg m−2)

1 17.4 5.8 9.5 59.7.1±28.8 34.9±56.9 17.7±68.8 31.0±82.6 122±2111–553 0.2–580 0.1–1495 0.2–963 0.8–2437

5 3.2 1.1 1.3 10.01 4.5 1.5 1.8 15.10 7.8 2.5 3.4 26.27 15.3 5.1 8.3 54.9.6 37.8 11.2 23.0 127.3 78.9 28.2 61.0 263.9 129 51.5 124 451.4 215 136 259 721.3 229 171 309 811

525P.E. Rasmussen et al. / Science of the Total Environment 443 (2013) 520–529

dust mass (rather than metal concentration) has the greater influ-ence on metal load is consistent with a previous study of Pb loadingsin high-risk urban communities in California (Sutton et al., 1995),which concluded that 67% of the dust Pb loading measurement isexplained by dust mass, and 33% by the concentration of Pb in thedust. Their observation that dust mass is the overriding influencepointed to the significance of cleaning prior to sample collection(rather than sources of Pb) as the main control on dust Pb loading(Sutton et al., 1995). It follows that keeping dust levels down is theuniversal recommendation to reduce Pb exposure (Sutton et al.,1995). The present study indicates that this advice applies not onlyto Pb, but also to Cd, Cu, Cr, Ni, and Zn (and probably As, althoughthe trend is less clear).

3.5. Dust loading rates

Geomean and median dust loading rates of 11.0 and10.3 mg m−2 day−1, respectively (Table 4) were calculated for thesubset of 559 homes in which floors were cleaned within a period of 6to 8 days prior to sampling (that is, all homes which adhered to the7 day sampling protocol±1 day). The dust loading ratewas lognormal-ly distributed (Shapiro–Wilk p=.114 at 95% CI), therefore analyseswere performed using log-transformed values. Wide variability oc-curred in the calculation of dust loading rates for homes which didnot adhere to the 7 day protocol, with the key source of uncertaintybeing the estimation of time elapsed since cleaning. Nevertheless, thedust loading rate for the whole dataset (geomean and median of 11.1and 10.8 mg m−2 day−1 respectively; n=1001) was very close tothat of the well-constrained subset.

Rates of dust loading determined using passive dust collectionsystems (e.g. petri dishes) are not directly comparable with ratesreported in Table 4. This is largely because the passive systems collectonly the fine dust fraction: Edwards et al. (1998) reported that over99% of particles collected on (raised) indoor surfaces are smaller than50 μm. Dust loading determined using polystyrene beakers placed ona shelf 1.7 m above ground for approximately 1-yr accumulation(Meyer et al., 1999) ranged from 1.3 to 48.5 mg m−2 day−1 (geomeanof 8.9 mg m−2 day−1) for 454 children's bedrooms located inHettstedt, a German smelter town. Siefert et al. (2000) reported amuch lower dust loading geomean of 5.41 mg m−2 day−1 for a popu-lation study across thewhole of Germany using the samepassive collec-tion method in 600 children's bedrooms. Had the Table 4 calculationsincluded only the fine fraction, loading rates in the present studywould be comparable to the German population study due to the factthat the mass of the fine fraction (b80 μm) accounted for only 50–60%of the total mass (b300 μm fraction). Fine particles tracked indoors,which are especially abundant in the vicinity of entry areas (Hunt et

Table 4Daily loading rates for dust (mg m−2 day−1) and elements (μg m−2 day−1) calculated forcleaning; n=559).

Dust(mg m−2 day−1)

As(μg m−2 day−1)

Cd(μg m−2 day−1)

Cr(μg m−2

Geometricmean

11.0 0.085 0.042 1.11

Mean±sd 20.3±37.2 0.298±0.718 0.112±0.298 2.12±4.6Range (min–max) 0.558–583 4.20×10−5–9.39 0.002–3.83 0.012–79Percentiles

5 2.10 0.003 0.005 0.19910 3.03 0.010 0.008 0.31825 5.56 0.037 0.015 0.54050 (median) 10.3 0.103 0.041 1.0575 22.4 0.255 0.103 2.1190 43.3 0.720 0.241 4.5295 64.3 1.36 0.373 6.8797.5 95.5 1.88 0.687 9.7798 113 2.16 0.736 11.3

al., 2006) are more likely to be quantitatively captured by vacuumingfloors in the whole house, which is another reason for higher dust load-ing rates in the present study compared to passive accumulation in bea-kers positioned on elevated surfaces in children's bedrooms. Finally,seasonal variations are an important consideration, as dust loadingrates tend to be higher in the summer when increased ventilation al-lows entry of resuspended particles (Edwards et al., 1998). The dustloading rates in the present study are considered a minimum as theyrepresent only the Canadian winter season, when windows and doorsare kept closed and ventilation is at a minimum.

Dust loading rates are not influenced by percentage carpet coverin the CHDS homes, as illustrated by the boxplot in Fig. 2 whichshows median dust loading rates for eleven categories of carpetcover ranging from the minimum of 0% carpet cover (100% barefloors) to the maximum of 90–100% carpet cover. The similarity ofdust loading rates between carpeted and non-carpeted floors in thepresent study is an outcome of the “fresh dust” sampling methoddesigned to collect only the surface dust from carpets, and avoiddeep dust of unknown age. As described by Roberts et al. (2009) thedust that is removable from a carpet is the sum of the surface dustand the extractable deep dust. Regression analysis confirms no corre-lation between loading rate and % carpet cover, whether consideringthe well-constrained subset of homes which followed the cleaningprotocol (R2b .01; p=0.910; n=559), or the entire dataset (R2b .01,p=.471; n=1001).

Although there was no observable geographic trend in indoordust loading rates across the country (R2b .001; p=.603), signifi-cantly higher indoor dust loading rates were associated with prox-imity to industry. An independent t-test showed that dust loadingrates in homes located outside a radius of 2 km from industry(geomean/median=9.6/9.1 mg m−2 day−1; n=580) were signifi-cantly lower (pb .001) than homes located within 2 km of industry(geomean/median=13.5/13.4 mg m−2 day−1; n=421).

3.6. Metal loading rates

Table 4 summarizes metal loading rates (μg m−2 day−1) for thesubset of 559 homes which followed the protocol (7±1 day sincecleaning), out of which 281 are located in or near urban industrialzones and 278 are located in non-industrial zones, as determined bythe field technician during the house visit. These metal datasets arenot normally or lognormally distributed (in contrast with dust load-ing) and therefore non-parametric tests are appropriate.

Metal concentrations and loading rates for the “non-industrial”subset are provided as Supplementary information (Table S2). Toidentify homes in “non-industrial” zones, all homes located within2 km of past or present industry (of any kind) were excluded, and

b300 μm particle size range for homes which followed the protocol (7±1 days since

day−1)Cu(μg m−2 day−1)

Ni(μg m−2 day−1)

Pb(μg m−2 day−1)

Zn(μg m−2 day−1)

2.29 0.754 1.32 8.20

2 4.45±7.37 1.69±4.00 4.52±13.1 16.1±28.1.0 0.146–66.9 0.034–45.0 0.065–138 0.470–348

0.417 0.157 0.173 1.420.611 0.208 0.269 2.101.04 0.343 0.489 3.682.10 0.674 1.18 7.694.56 1.41 3.04 16.79.98 3.34 8.34 34.516.2 5.35 15.5 55.330.6 8.56 44.1 89.632.3 11.5 54.0 96.8

Fig. 2. Box plots showing median dust loading rates for varying percentages of carpet cover (and sample sizes for each category) in 559 homes which followed the protocol (7±1 dayssince cleaning). Variance is homogeneous (Levine's test p=0.114) and means are equivalent (one-way ANOVA F=0.530; p=0.869). There is no significant difference between thelog-transformed means for the end members 100% bare floors and 90–100% carpets (independent t-test, 2-tailed p=0.354) and no significant difference (independent t-test, 2-tailedp=0.615) between end members showing the highest dust loading rate (50–60% carpet) and the lowest dust loading rate (70–80% carpet).

526 P.E. Rasmussen et al. / Science of the Total Environment 443 (2013) 520–529

in addition any homes feasibly impacted by industry outside 2 kmwere also excluded. For example, all homes in the Greater SudburyArea were excluded from Table S2 to avoid potential influence ofthe Sudbury smelter stack (height 380 m; SARA Group, 2008). Notethat in the present study, “industry” is an all-inclusive term whichrefers to urban industrial zones, and is therefore not limited tometals-related industry. Urban industry is represented by a wide va-riety of activities including gas stations, industrial parks, past andpresent manufacturing and recycling plants, factories located in rail-way yards and harbors, and mining and smelting operations.

It was not the study design or purpose of the CHDS to determinesource–receptor relationships, but rather to provide a nationally-representative baseline against which such smaller-scale studies canbe compared. Local- or regional-scale studies of the influence of miningand smelter sources on residential loading rates (e.g., Meyer et al., 1999;Hilts, 2003; Hogervorst et al., 2007) provide information that is comple-mentary to the CHDS, with the caveat that direct comparisons may belimited by sampling and analytical differences, as discussed earlier. Ina Belgian smelter study, Hogervorst et al. (2007) demonstrated an inde-pendent association between the biomarkers of internal exposure andthe loading rates of Cd and Pb in house dust, determined using passivedust collection in Petri dishes set out for a 3-month period in bedrooms.A two-fold increase in the loading rate of cadmium in house dust wasassociatedwith a 2.3% and 3.0% rise in blood and 24-h urinary cadmium,and a two-fold increase in the loading rate of Pb in house dustwas asso-ciated with a 2.0% increase in the blood Pb concentration. Meyer et al.(1999) reported geomean metal loading rates (μg m−2 day−1) of1.14 for Pb, 0.023 for As and 0.024 for Cd for 454 children's bedroomsin a German smelter town. In contrast, lower metal loading rates(μg m−2 day−1) of 0.391 for Pb, 0.008 for As, and 0.016 for Cd werereported for 600 children's bedrooms in the national-scale Germanstudy (Siefert et al., 2000). In a study of the Trail smelter in BritishColumbia, Canada, Hilts (2003) reported a decline in children's

mean blood lead levels from 11.5 μg/dL in 1996 to 5.9 μg/dL in1999 as indoor dust Pb loading rates declined by approximately50% from 0.14 mg m−2 day−1 to 0.07 mg m−2 day−1, attributedto implementation of new smelting technologies in 1997.

Table 5 summarizes dust and metal loading rates in smoking andnon-smoking homes, based on the subset of 580 homes whichreported no past or present industrial activity within 2 km. Cadmi-um is the only element displaying lognormal distribution of loadingrates in non-industrial homes (Shapiro–Wilk p=.227). The median/geomean Cd loading rate in 494 homes occupied by non-smokers(33.6/34.3 ng m−2 day−1) is significantly lower (2-tailed t-test, p=.034) than 86 homes occupied by smokers (47.3/49.5 ng m−2 day−1).Non-parametric tests confirmed that loading rates for the other six ele-ments are significantly higher (p≤ .035) in smoking homes as summa-rized in Tables 5 and 6. In contrast, there are no significant differences inelemental concentrations (μg g−1) in dust of smoking vs non-smokinghomes for any of the 7 elements studied (Table 6).

3.7. The influence of dust mass on metal loading rates

As discussed earlier, the relative strength of the relationship betweendust mass (mg m−2) and metal load (μg m−2) for Cd, Cu, Cr, Ni, Pb, andZn (r values from .8 to .9), compared to the relationship between metalconcentration (μg g−1) and metal load (r values from .3 to .6) indicatesthe dominating influence of dust mass on metal loading in the CHDShomes (n=1022). Dust mass is also the dominant influence on rates ofmetal loading (μg m−2 day−1) in smoking vs non-smoking homes(Table 5), evidenced by the significant difference in dust loading rate(Table 5) and lack of significant difference in metal concentrations be-tween the two groups (Table 6). This point is illustrated by a multilinearregression analysis for Cd (the only lognormally distributed element inTable 5) which indicates that the Cd loading rate is only 32% influenced

Table 5Smoking vs non-smoking homes: daily loading rates for dust (mg m−2 day−1) andelements (μg m−2 day−1) calculated for b300 μm particle size range in all homeslocated more than 2 km from industry (n=580). Differences are significant for allelements at 95% CI (see Table 6).

Non-smoking Smoking

Dust Median 8.70 13.0(mg m−2 day−1) Geomean 9.20 12.6

Mean±sd 21.7±51.4 28.6±60.2Range (min–max) 0.199–598 0.833–491

As Median 0.076 0.136(μg m−2 day−1) Geomean 0.068 0.121

Mean±sd 0.318±0.998 0.651±2.70Range (min–max) 2.24×10−5–13.5 1.37×10−4–22.4

Cd Median 0.034 0.047(μg m−2 day−1) Geomean 0.034 0.049

Mean±sd 0.107±0.381 0.176±0.391Range (min–max) 5.37×10−4–7.48 0.001–2.64

Cr Median 0.856 1.46(μg m−2 day−1) Geomean 0.931 1.40

Mean±sd 2.38±5.93 3.35±7.76Range (min–max) 0.018–79.0 0.068–65.3

Cu Median 1.78 2.51(μg m−2 day−1) Geomean 2.02 2.76

Mean±sd 5.62±15.1 8.16±19.7Range (min–max) 0.064–188 0.189–128

Ni Median 0.553 0.848(μg m−2 day−1) Geomean 0.674 0.883

Mean±sd 3.33±23.4 2.22±5.24Range (min–max) 0.014–498.189 0.082–38.6

Pb Median 0.948 1.41(μg m−2 day−1) Geomean 1.11 1.62

Mean±sd 5.06±17.8 7.68±23.3Range (min–max) 0.014–206 0.078–163

Zn Median 6.55 10.1(μg m−2 day−1) Geomean 6.81 10.2

Mean±sd 16.5±40.4 30.8±66.3Range (min–max) 0.137–604 0.420–451

527P.E. Rasmussen et al. / Science of the Total Environment 443 (2013) 520–529

by the Cd concentration of dust, and 68% by the dust mass (n=580;pb .001).

The observed dominating influence of dust mass in the presentstudy is consistent with the California homes studied by Sutton etal. (1995), but contrasts with “hotspot” studies where dust metal con-

able 6ignificance (2-tailed p values) of house age, proximity to industry and smoking activity associations with elemental concentrations (μg g−1) and loading rate (μg m−2 day−1).haded values are significant at 95% CI.

As Cd Cr Cu Ni Pb Zn

House age

Concentration p = 0.260 p < 0.001 p = 0.866 p = 0.143 p = 0.467 p < 0.001 p < 0.001

r = 0.048 r = 0.280 r = 0.007 r = 0.062 r = 0.031 r = 0.498 r = 0.219

Loading rate p < 0.001 p < 0.001 p < 0.001 p < 0.001 p < 0.001 p < 0.001 p < 0.001

n = 559

7 ± 1 day since cleaning

Spearman Rank test

r = 0.251 r = 0.409 r = 0.301 r = 0.318 r = 0.336 r = 0.521 r = 0.381

Proximity to industry

Concentration p = 0.917 p = 0.812 p = 0.291 p = 0.330 p = 0.445 p = 0.629 p = 0.602

Loading rate p = 0.002 p < 0.001 p < 0.001 p < 0.001 p < 0.001 p < 0.001 p < 0.001

n = 281 close to industry

n = 278 non industry

7 ± 1 day since cleaning

Mann Whitney U test

Smoking in the home

Concentration p = 0.295 p = 0.701 p = 0.089 p = 0.970 p = 0.700 p = 0.440 p = 0.145

Loading rate p = 0.005 p = 0.034a p = 0.003 p = 0.035 p = 0.021 p = 0.038 p = 0.010

n = 86 homes with smokers

n = 494 homes no smokers

No industry within 2 km

Mann Whitney U test

2-tailed p value of t-test used because Cd loading rate dataset is lognormally distributed (Shapiro Wilk p=0.227).

TSS

a

centrations and loading rates are elevated in homes located in con-taminated areas (Meyer et al., 1999; Seifert et al., 1984; Aurand etal., 1983). In the Meyer et al. (1999) smelter study, for example,62% of the variance in dust Pb loading rates was explained by dustPb concentration and 38% by dust loading rate. This is the oppositetrend to the present study and shows that observations that applyto the general population may not apply in “geochemical hotspots”where associations and trends are dominated by elevated metalconcentrations.

Spearman rank analysis of house age provides another example ofthe influence of dust mass on metal loading rates (Table 6). Note thatmetal loading rates increase significantly with house age for all 7 ele-ments studied (pb .001 for 559 homes which followed the protocol),even though the concentration–age relationship is significant onlyfor Pb, Cd and Zn (Table 6). The apparent discrepancy in the case ofthe other 4 elements (As, Cu, Ni and Cr), whose concentrationsshow no relationship to house age, is explained by the significantlyhigher dust loading rates associated with older homes (pb .001; n=559). This demonstrates that concentration data are more likely to in-dicate the presence of metal sources inside and around the home,while loading measurements are more likely to be influenced bydust mass than by concentration.

Mann–Whitney U tests were used to further investigate the influ-ence of proximity to industry on elemental concentrations and load-ing rates using the 559 homes which followed the protocol (Table 4).Unlike local- or regional-scale smelter studies, elemental concentra-tions (μg g−1) in the population-based CHDS are not significantlyinfluenced by proximity to urban industrial zones (p≥ .646 for all 7metals studied, Table 6). This is due to the all-inclusive definitionof the “urban industry” term, which is not specific to metals, andalso to the randomized population-based sampling design. Metalloading rates, however, do significantly increase with proximity toindustry: for all 7 metals studied, loads (μg m−2) and loading rates(μg m−2 day−1) are significantly higher in homes near industry(p≤ .009 and p≤ .008 respectively; n=559). It follows that, forthe CHDS homes, the increase in dustiness in industrial areas isthe main control on metal loading rates (μg m−2 day−1). Mann–Whitney U tests applied to the entire dataset of CHDS homeswith available measurements (n=1001) yield the same results,confirming that differences in loads and loading rates are significant

528 P.E. Rasmussen et al. / Science of the Total Environment 443 (2013) 520–529

for all 7 elements (pb .001 and pb .003 respectively) but not concen-trations (.29≥p≤ .97).

Meyer et al. (1999) did not find an association between metalloading rates and smoking activity, which they attributed to twofactors: (1) the positioning of the samplers in children's bedrooms,and (2) elevated environmental metal levels in the smelter town.Sampling the whole house in the CHDS (rather than an isolated bed-room as in the passive sampling studies) assisted in detecting the in-fluence of smoking on loading rates, as did excluding homes locatedin urban industrial zones. When the same calculations in Table 5were conducted without considering proximity to industry, the dif-ferences in metal loading rates between smoking and non-smokinghomes were no longer significant at the 95% CI (n=1001; p=.074).Thus, in the CHDS the increased dustiness of homes located inurban industrial zones is a key factor obscuring the smoking vs non-smoking relationship, which does not emerge until homes within2 km of industry are excluded.

In a detailed study of the influences of home location, constructiondate, and tobacco smoking on the amount of indoor dust and its metalcontent in 8 homes in Giza, Egypt, Khoder et al. (2010) underscoredthe important influence of exterior sources on indoor settled surfacedust. In the Egyptian study, dust from unpaved streets was determinedto be a key contributor to indoor dust loading (Khoder et al., 2010). De-spite the differences in Canadian climate, house construction, and urbanlandscape, the CHDS also showed that exterior sources have a signifi-cant influence on indoor dust. In the CHDS, increased dust loadingrates in homes located near industry may be due to the combinationof increased vehicular traffic and scant vegetation cover associatedwith industrial zones compared to residential zones. House agedoes not provide an explanation, as there is no significant differencein house age between CHDS homes located in industrial vs. non-industrial zones (2-tailed p=.323). An important and related findingthat emerged from the German smelter studies (Seifert et al., 1984;Aurand et al., 1983) was the wide variability of metal loadings amongsthomes which were located at the same distance from the smeltersource, and even between adjacent homes (Aurand et al., 1983). In allsuch cases it was noted that the lower values were found in the housesthat were surrounded by gardens or other green spaces or wereequipped with tight windows (Seifert et al., 1984). Their observationthat the rate of dust loading and metal loading is strongly influencedby the percentage of dust-emitting land around the house led to theconclusion that covering open, dusty spaces with vegetation is animportant step in reducing exposures.

3.8. Summary and conclusions

In recognition of the need for a reliable house dust samplingmethodology, the CHDS focused on quality assurance during all as-pects of sampling and laboratory processing, including: measure-ments of area sampled and time elapsed since cleaning; avoidanceof particle contamination and losses; duplicate reproducibility; andanalytical method evaluation using NIST certified reference materials.

A rigorous random sampling approachwas used to obtain nationallyrepresentative urban dustmetal concentrations (μg g−1),metal loading(μg m−2) andmetal loading rates (μg m−2 day−1), to address the needfor population-based baseline estimates for metals commonly targetedin residential risk assessments. Nationally representative dust loadingrates (mg m−2 day−1) presented in this paper will also assist in esti-mating chemical loading rates where only dust concentration data areknown (e.g. sampled from the household vacuum bag) in future resi-dential assessments.

Collection of a complete set of measurements using a single sam-pling protocol (concentration, load, and loading rate) brings clarity tothe relative importance of metal sources, frequency of cleaning, orgeneral dustiness in a given exposure scenario. The results revealthat associations between metal loading and smoking activity are

mainly driven by increased dust loading in homes occupied bysmokers. Concentration data indicate that Pb, Cd and Zn concentra-tions increase with house age, which indicates a greater presence ofresidential sources of these metals in older homes. However, it isthe increased dustiness of older homes, not necessarily the presenceof metal sources, which drives the increase in metal loading rateswith house age for the other four elements (Cr, Cu, Ni, and As). It isconcluded that, while concentration data are useful for indicatingthe presence of metal sources inside and around the home, dustload has an overriding influence on metal load and metal loadingrates.

These results also point to the significance of dustiness in urbanindustrial zones: approximately half the CHDS homes were locatedwithin 2 km of industrial zones and were characterized by elevateddust and metal loading rates compared to homes in non-industrialzones. The fact that there is no significant difference in dust metalconcentrations in non-industrial vs industrial zones (broadly definedin this study) indicates that it is the higher dust loading rate that isthe driver for the higher metal loading rates observed in homes locat-ed in urban industrial zones.

Trends emerge in the CHDS results (such as correlations betweensmoking activity and metal loading) which may otherwise be ob-scured in geochemical hotspot areas where the signal is swampedby locally elevated environmental metal concentrations. Moreover,the CHDS provides a baseline, or point of comparison, for local scalesource-receptor type studies of homes located in metal processing,mining and smelting regions.

Supplementary data to this article can be found online at http://dx.doi.org/10.1016/j.scitotenv.2012.11.003.

Acknowledgments

The Canadian House Dust Study was approved by Health Canada'sResearch Ethics Board. Special thanks go to the study participants,and to Water & Earth Science Associates (WESA) Inc., Carp, ON. Theauthors are very grateful to T. Mischki and L. Seed for their internalreviews and to M. Lanouette, L. McDonald, and T. Roselli for valuabletechnical assistance. Funding and support from Health Canada'sChemicals Management Plan (Monitoring and Surveillance), Contam-inated Sites Division and the Federal Contaminated Sites Action Plan(FCSAP), HECS Branch and RAP Branch are gratefully acknowledged.

References

Aurand K, Drews M, Seifert B. A passive sampler for the determination of the heavymetal burden of indoor environments. Environ Technol Lett 1983;10:433–40.

Beauchemin S, MacLean LCW, Rasmussen PE. Lead in indoor dust: an in situ speciationstudy to assess old paint contribution. Environ Geochem Health 2011;33:343–52.

Bell T, Campbell S, Liverman DGE, Allison D, Sylvester P. Environmental and potentialhuman health legacies of non-industrial sources of lead in a Canadian urbanlandscape: the case study of St. John's, Newfoundland. Int Geol Rev 2010;52(7–8):771–800.

Butte W, Heinzow B. Pollutants in house dust as indicators of indoor contamination.Rev Environ Contam Toxicol 2002;175:1-46.

Caplan KJ. The significance of wipe samples. Am Ind Hyg Assoc J 1983;53:70–5.CDC (Centers for Disease Control and Prevention-NIOSH). Exposome and exposomics:

overview. NIOSH>Workplace Health and Safety Topics>Hazards and Exposures;2012) [http://www.cdc.gov/niosh/topics/exposome (accessed April 4, 2012)].

Colt JS, Gunier RB, Metayer C, Nishioka MG, Bell EM, Reynolds P, et al. Household vacuumcleaners vs. the high-volume surface sampler for collection of carpet dust samples inepidemiologic studies of children. Environ Health 2008. http://dx.doi.org/10.1186/1476-069X-7-6.

Dixon SL, Gaitens JM, Jacobs DE, StraussW, Nagaraja J, Pivetz T, et al. Exposure of U.S. chil-dren to residential dust lead, 1999–2004: ii. The contribution of lead-contaminateddust to children's blood lead levels. Environ Health Perspect 2009;117(3):468–74.

Duggan MJ, Inskip MJ. Childhood exposure to lead in surface dust and soil: a commu-nity health problem. Public Health Rev 1985;13:1-54.

Edwards RD, Yurkow EJ, Lioy PJ. Seasonal deposition of housedusts onto household sur-faces. Sci Total Environ 1998;224(1–3):69–80.

Fan X, Kubwabo C, Rasmussen PE, Jones-Otazo H. Simultaneous quantitation ofparabens, triclosan, and methyl triclosan in indoor house dust using solid-phaseextraction and gas chromotography-mass spectrometry. J Environ Monit2010;12:1891–7.

529P.E. Rasmussen et al. / Science of the Total Environment 443 (2013) 520–529

Franz Environ Inc. Canadian Federal Contaminated Sites Action Plan: contaminantmixture study. Report 1045-0401 prepared for Health Canada, March 31FranzEnvironmental Inc., 329 Churchill Avenue, Suite 200, Ottawa, Ontario; 2005[12 pp.].

Galke W, Clark S, Wilson J, Jacobs D, Succop P, Dixon S, et al. Evaluation of the HUD leadhazard control grant program: early overall findings. Environ Res A 2001;86:149–56.

Hassan NM, Rasmussen PE, Dabek-Zlotorzynska E, Celo V, Chen H. Analysis of environ-mental samples using microwave-assisted acid digestion and inductively coupledplasma mass spectrometry: maximizing total element recoveries. Water Air SoilPollut 2007;178:323–34.

Hilts SR. Effect of smelter emission reductions on children's blood lead levels. Sci TotalEnviron 2003;303(1–2):51–8.

Hogervorst J, Plusquin M, Vangronsveld J, Nawrot T, Cuypers A, Van Hecke E, et al.House dust as possible route of environmental exposure to cadmium and lead inthe adult general population. Environ Res 2007;103:30–7.

Hunt A, Johnson DL, Griffith DA. Mass transfer of soil indoors by track-in on footwear.Sci Total Environ 2006;370:360–71.

Johnson DL, Hunt A, Griffith DA, Hager JM, Brooks J, StellaLevinsohn H, et al. Geographicpatterns of non-carpeted floor dust loading in Syracuse, New York (USA) homes.Environ Geochem Health 2009;31:353–63.

Kettles IM, Shilts WW. Composition of glacial sediments in Canadian Shield terrane,southwestern Ontario and southwestern Quebec: applications to acid rain researchand mineral exploration. Geological Survey of CanadaBulletin, 463; 1994 [58 pp.].

Khoder MI, Hassan SK, El-Abssawy AA. An evaluation of loading rate of dust, Pb, Cd, andNi and metals mass concentration in the settled surface dust in domestic housesand factors affecting them. Indoor Built Environ 2010;19:391–9.

Lanphear BP, Matte TD, Rogers J, Clickner RP, Dietz B, Bornschein RL, et al. The contri-bution of lead-contaminated house dust and residential soil to children's bloodlead levels: a pooled analysis of 12 epidemiologic studies. Environ Res Sect A1998;79:51–68.

Lichtenwalner CP. Evaluation of wipe sampling procedures and elemental surface con-tamination. Am Ind Hyg Assoc J 1992;53:657–9.

Lioy PJ, Freeman NCG, Wainman T, Stern AH, Boesch R, Howell T, et al. Microenviron-mental analysis of residential exposure to chromium laden wastes in and aroundNew Jersey homes. Risk Anal 1992;12:287–99.

MacLean LCW, Beauchemin S, Rasmussen PE. Lead speciation in house dust fromCanadian urban homes using EXAFS, Micro-XRF, and Micro-XRD. Environ SciTechnol 2011;45(13):5491–7.

McDonald LT, Rasmussen PE, Chénier M, Levesque C. Extending wipe samplingmethods to elements other than lead. J Environ Monit 2011;13:377–83.

Meyer I, Heinrich J, Lippold U. Factors affecting lead, cadmium, and arsenic levels inhouse dust in a smelter town in eastern Germany. Environ Res 1999;81:32–44.

Niu J, Rasmussen PE, Wheeler A, Williams R, Chénier M. Evaluation of airborne partic-ulate matter and metals data in personal, indoor and outdoor environments usingED–XRF and ICP–MS and co-located duplicate samples. Atmos Environ 2010;44:235–45.

Paustenbach DJ, Finley BL, Long TF. The critical role of house dust in understanding thehazards posed by contaminated soils. Int J Toxicol 1997;16(4–5):339–62.

Rasmussen PE. Long-range atmospheric transport of trace metals: the need for geosci-ence perspectives. Environ Geol 1998;33(2–3):170–82.

Rasmussen PE. Elements and their compounds in indoor environments. In: Merian E,Anke M, Ihnat M, Stoeppler M, editors. Elements and their compounds in the envi-ronment. 2nd ed. Weinheim: Wiley-VCH; 2004a. p. 215–29.

Rasmussen PE. Can metal concentrations in indoor dust be predicted from soil geo-chemistry? Can J Anal Sci Spect 2004b;49:166–74.

Rasmussen PE, Gardner HD. Earth and health— building a safer Canadian environment.Geosci Can 2008;35(2/3):61–72.

Rasmussen PE, Subramanian KS, Jessiman BJ. A multi-element profile of housedust inrelation to exterior dust and soils in the city of Ottawa, Canada. Sci Total Environ2001;267:125–40.

Rasmussen PE, Wheeler AJ, Hassan NM, Filiatreault A, Lanouette M. Monitoring person-al, indoor, and outdoor exposures to metals in airborne particulate matter: risk ofcontamination during sampling, handling and analysis. Atmos Environ 2007;41:5897–907.

Rasmussen PE, Beauchemin S, Nugent M, Dugandzic R, Lanouette M, Chénier M. Influ-ence of matrix composition on bioaccessible copper, zinc and nickel in urban resi-dential dust and soil. J Hum Ecol Risk Assess 2008;14(2):351–71.

Rasmussen PE, Beauchemin S, Chénier M, Levesque C, MacLean LCW, Marro L, et al.Canadian House Dust Study: lead bioaccessibility and speciation. Environ SciTechnol 2011;45(11):4959–65.

Rencz AN, Garrett RG, Adcock SW, Bonham-Carter GF. Geochemical background in soiland till. Geological Survey of Canada, Open File 5084; 2006. p. 1-64.

Roberts JW, Wallace LA, Camann DE, Dickey P, Gilbert SG, Lewis RG, et al. Monitoringand reducing exposure of infants to pollutants in house dust. Rev Environ ContamToxicol 2009;201:1-39.

Rodes CE, Newsome JR, Vanderpool RW, Antley JT, Lewis RG. Experimental methodol-ogies and preliminary transfer factor data for estimation of dermal exposures toparticles. J Exp Anal Environ Epidemiol 2001;11:123–39.

SARA Group. Summary of volume II: human health risk assessment, SARA (SudburyArea Risk Assessment); 2008. p. 1-44. [http://www.sudburysoilsstudy.com/EN/media/Volume_II/Volume_II_SummaryReport/FINAL_SARA_HHRA_Summary_May01_08.pdf (accessed October 30, 2012)].

Seifert B, Drews M, Aurand K. Indoor heavy metal exposure of the population around asecondary lead smelter. Indoor Air, 2. Stockholm Sweden: Swedish Council forBuilding Research; 1984. p. 177–81.

Siefert B, Becker K, Helm D, Krause C, Schulz C, Seiwert M. The German EnvironmentalSurvey 1990/1992 (GerES II): reference concentrations of selected environmentalpollutants in blood, urine, hair, house dust, drinking water and indoor air. J ExpoAnal Environ Epidemiol 2000;10:552–65.

Sutton PM, Athanasoulis M, Flessel P, Guirguis G, HaanM, Schlag R, et al. Lead levels in thehousehold environment of children in three high-risk communities in California.Environ Res 1995;68(1):45–57.

Synek V. Evaluation of the standard deviation from duplicate results. Accred Qual Assur2008;13:335–7.

Turner A. Oral bioaccessibility of trace metals in household dust: a review. EnvironGeochem Health 2011;33:331–41.

Verein Deutscher Ingenieure (VDI). German Protocol VDI 4300. Part 8 — measurementof indoor air pollution: sampling of house dust. Berlin: Beuth; 2001.

Walker SR, Jamieson HE, Rasmussen PE. Application of synchrotron microprobemethods to solid-phase speciation of metals and metalloids in house dust. EnvironSci Technol 2011;45(19):8233–40.

Young TM, Heeraman DA, Sirin G, Ashbaugh LL. Resuspension of soil as a source of air-borne lead near industrial facilities and highways. Environ Sci Technol 2002;36:2484–90.

Zota AR, Schaider LA, Ettinger AS, Wright RO, Shine JP, Spengler JD. Metal sources and ex-posures in the homes of young children living near a mining-impacted Superfundsite. J Expo Sci Environ Epidemiol 2011;21:495–505.