Variability of personal chemical exposure in eight office ... - Nature

Journal of Exposure Analysis and Environmental Epidemiology (2004) 14, S49–S57 r 2004 Nature Publishing Group All rights reserved 1053-4245/04/$25.00

www.nature.com/jea

Variability of personal chemical exposure in eight office buildings in Sweden BO GLAS,a JAN-OLOF LEVIN,b BERNDT STENBERG,a,c HANS STENLUNDc AND ANNA-LENA SUNESSONb a

Dermatology and Venereology, Department of Public Health and Clinical Medicine, Umea˚ University, Sweden National Institute for Working Life, Department of Work and Physical Environment, Umea˚, Sweden c Epidemiology, Department of Public Health and Clinical Medicine, Umea˚ University, Sweden b

This study focuses on the variability in chemical exposures for individuals working in office buildings. The study involved eight office buildings with 79 participants, and exposures were measured using personal samplers for volatile organic compounds, aldehydes, amines, nitrogen dioxide, ozone, and particles. Ventilation was assessed in each individual office. ‘‘Variability among buildings’’ and ‘‘variability among individuals’’ were evaluated for any component (of the 123) measured in samples from at least 20 persons, using variance component analysis and principal component analysis. Interpersonal differences explained the major part of the variance for 78% of the compounds versus between-buildings differences for 14% of the compounds. For 8% of compounds, the variation was explained in equal amounts by the differences among individuals and among buildings. This study illustrates the necessity for individualised measurements (versus stationary measurements in building) to estimate personal exposures. These results also support the conculsion that in case-referent studies of ‘‘sick building syndrome’’ (SBS), referents to SBS cases can be randomised for building location. Journal of Exposure Analysis and Environmental Epidemiology (2004) 14, S49–S57. doi:10.1038/sj.jea.7500358

Keywords: sick building syndrome, office, epidemiology, variation, chemical exposure, sex

Introduction Sick building syndrome (SBS) has been puzzling scientists for over 25 years. WHO (1983) defined SBS as a combination of general, mucous, and skin symptoms. Several risk factors have been found including low ventilation, presence of photocopiers and laser printers, high psychosocial load, female gender (Redlish et al., 1997), and water damage to the building. These risk factors have been recently reviewed by Bornehag et al. (2001). When people have symptoms believed to be related to the air quality at their work place, the building ventilation is often checked. Studies have shown that low fresh air supply is a risk factor for SBS. It has, however, not been shown as to which offending chemicals are removed or diluted by increased ventilation, thereby improving indoor air quality. The building is often also searched for water damage, mould growth, or other pollution sources. Standard measurements of air quality, such as CO2 concentrations, are of limited use when no damage in the building is found. There is also a lack of knowledge about which chemicals or combinations of chemicals are harmful in concentrations commonly found in office buildings.

1. Address all correspondence to: B. Glas, Dermatology and Venereology, NUS, SE-901 85 Umea˚, Sweden. Tel.: þ 46-90-785-20-64; Fax: þ 46-90-14-36-73; E-mail: [email protected]

In recent years, the use of multivariate methods has been suggested for evaluation of chemical data from indoor environments. Such methods can be used to identify possible systematic differences in the chemical composition of the air in buildings with high and low incidence of SBS. Principles for such strategies have been thoroughly discussed in the literature (Box et al., 1978; Carlson, 1992; Wold et al., 1998). In a study using multivariate methods on air samples in buildings with SBS symptoms, Ten Brinke et al. (1998) found relationships between SBS symptoms and concentrations of groups of volatile organic compounds (VOCs) present at low levels in office air. Sunesson et al. (2002) used principal component analysis (PCA, Eriksson et al., 2001) on VOC concentrations in buildings (homes and offices) with and without known SBS complaints. The data from problem and nonproblem buildings formed two groups in principal component (PC) plots. Pommer et al. (2004) analysed VOC concentrations measured by the Office Illness Project in Northern Sweden. Using PCA, they could separate office buildings with high and low prevalence of SBS into different classes based on VOC concentrations. They were also able to use their models to predict the prevalence of SBS symptoms in buildings, given VOC concentration measurements, thus showing that the predictive models were valid. Published studies using multivariate analysis of building VOC concentrations and patterns have been based on stationary sampling in buildings with high and low prevalence of SBS complaints. However, several studies of

Variability of personal chemical exposure

Glas et al.

SBS show that symptoms are also found among workers in buildings not known as problem buildings (Hodgson et al., 1991; Stenberg, 1994). It is therefore of interest to perform personal exposure measurements of people in indoor environments, in an attempt to find correlations between a person’s exposures and whether or not he/she reports SBS symptoms. It appears that it is possible to separate buildings with high and low prevalence or incidences of SBS symptoms among workers by using multivariate techniques to evaluate chemical data from the buildings. In case-referent studies, it is best to choose referents randomly. To control possible confounding, referents may be matched with cases with respect to sex, age, smoking status, etc. In a case-referent study of individuals classified by the presence or absence of SBS symptoms, it is important to know the variability of VOC concentrations in different buildings and the variability of exposure among individuals within a given building. If the air quality (and thus chemical exposure) is homogenous within buildings, referents must be chosen from different buildings. If the chemical exposure varies more on an individual basis than on a building-by-building basis, referents can be randomly chosen. The primary aim of this study was to confirm that there is variability in air quality among buildings and among individual office workers, and to measure that variability. A secondary aim was to determine whether differences in chemical exposures exist between genders.

Material and methods The Buildings Measurements were performed in eight office buildings in Umea˚, a town in northern Sweden. In five of the buildings, the municipal building office or company health service knew that there were complaints about the indoor air quality. The other three buildings were randomly chosen. Building information is given in Table 1. All buildings had mechanical ventilation. Buildings 2 and 4 had air conditioning. Building 3 had 15% recirculated air. Building 7 was heated by the supply air. Building 2 had minor water damage in one room on the ground floor that occurred 2 years before this study. The damage was repaired as soon as it was discovered.

Participants A questionnaire was mailed to the homes of all the office workers in each building or on a particular floor. The questionnaire included questions about work conditions, health symptoms, and domestic building characteristics. The questionnaire was sent 3 weeks before the planned sampling period so that the answers would coincide as closely as possible with the measurements. By answering the questionnaire, the office workers agreed to participate in the study. A code was assigned to each participant to maintain confidentiality. To be included in the study, a person had to be a nonsmoker who worked at least 20 h a week in their office. Of the 580 questionnaires distributed, 367 (63%) were returned. In total, 79 persons in eight buildings participated in the study. There were 52 females and 27 males. The study was performed between 1 November 2000 and 30 April 2001. At the start of a sampling week, the participants in a building were given an introduction by the study leader. There was a maximum of 10 persons per group (two groups participated in building 2). Participants were given background information on the study. Next, the participants put on the sampling equipment while supervised by the study leader. The study leader gave a thorough demonstration of the use of the different samplers and pumps. Participants were also given detailed written instructions for use of the equipment, and telephone numbers where the study leader could be reached. At the end of the introductory session, the air flow rates through the pump-driven samplers were measured with a rotameter. Besides verifying air flow rates, this test gave the study leader a final opportunity to inspect the samplers. The air flow rates were also measured when the study leader collected the equipment at the end of the sampling period. An average flow rate was calculated for every sampler. To ensure that the samplers were placed in the same position on each person, 50/50 cotton/polyester vests were purchased from Hejco, Sweden. They were reinforced with ribbons that also served to mark the locations where the samplers should be attached (Figure 1). The ribbons were symmetrically placed at the shoulder seams and at the top of the breast pockets. Ribbons were also placed at the back to secure the air tubes connected to the pumps. Each sampler had a prescribed position on the vest. The orientation was reversed for left-handed individuals. The participants were

Table 1. Information on buildings included in the study.

Construction year Renovation/rebuilding Number of floors Number of workers Number of participants

S50

Building 1

Building 2

Building 3

Building 4

Building 5

Building 6

Building 7

Building 8

1955 1987 2 80 9

1976 1992 3 135 17

1959 1975 4 160 10

1910 1999 3 105 9

1974 1997/98 3 160 10

1980 F 3 100 5

1971 1995 3 150 9

1984 1999 2 35 10

Journal of Exposure Analysis and Environmental Epidemiology (2004) 14(S1)


Glas et al.

according to Levin et al. (1988), with a few modifications. Before sampling, the glass fibre filters were coated by dipping them into a solution containing 300 mg 2,4-dinitrophenylhydrazine (DNPH HCl), 0.15 ml 85% phosphoric acid, 1.5 ml 20% glycerine in ethanol, and 9 ml acetonitrile, after which they were dried in formaldehyde-free air for 30 min. Extraction and HPLC analysis was performed as described previously (Sunesson et al., 2001).

Figure 1. Drawing of the vest.

instructed to wear the vest on top of their own clothes. Before the first sampling period and after each sampling period, the vests were laundered in the same way by a laundry and stored together in a plastic bag. The participants were instructed to start sampling when they arrived at work, and to close the samplers when they left the building. The sampling was carried out solely in the building where they worked. Participants noted on a form the time when they started and stopped sampling, and any events that might have influenced the sampling. During the sampling period, the samplers were stored overnight at room temperature at their work site. All chemical analyses were performed within 2 weeks after the sampling period.

The Pumps Two types of pumps were used. Pocket pumps were used for amine sampling. Air Check 2000 pumps were used for particle sampling. (Both pumps were from SKC.) The participants charged the Pocket pumps each night. At each workplace, one person was responsible for charging the Air Check 2000 pumps. This was done on two specified nights during the sampling period. The person responsible for the charging was given a demonstration and written instructions on how to perform the charging. He/she also stored two spare pumps at the workplace. Aldehydes Aldehydes were sampled with diffusive samplers and quantified using HPLC. The procedure was performed Journal of Exposure Analysis and Environmental Epidemiology (2004) 14(S1)

Amines For amine sampling, air was pumped at a flow rate of 200 ml/min and split between two sampling tubes: a 1naphthyl isothiocyanate-coated XAD-2 tube for primary and secondary amines, and a charcoal tube for tertiary amines (both from SKC). The flow rate through each tube was measured at the beginning and end of the sampling period, and an average flow rate was calculated. After sampling, the 1-naphthyl-isothiocyanate-coated XAD-2 tubes were extracted in 3 ml acetonitrile. Primary and secondary amines in this extract were measured as described by Lindahl et al. (1993). Tertiary amines were measured as described by Andersson and Andersson (1989). All data were collected, integrated, and quantified using a Millennium32 lab data system (Waters). Nitrogen Dioxide NO2 was sampled with a Willems badge diffusive sampler impregnated with a solution of triethanolamine and acetone; the analyses were carried out in an FIA analyser with a spectrophotometric detector as described by Hagenbjo¨rkGustafsson et al. (1999). Ozone Ozone was collected using diffusive samplers (Ferm, 2001). This technique measures oxidation of nitrite to nitrate in the presence of ozone. The amount of nitrate produced was measured by ion chromatography. IVL Swedish Environmental Research Institute Ltd. (Gothenburg, Sweden) supplied the samplers and performed the analysis. Particles Particulate matter has been identified as one of the three priority research areas in indoor air quality by the US Environmental Protection Agency (Angell and Grimsrud, 2002). Particles were collected by pumping air at 2 l/min through aerosol monitors (black 3-piece Omega Leak-free Cassettes No. A-002550SAC) equipped with 25-mm preweighed cellulose acetate/cellulose nitrate filters with 0.8 mm pore size (AWP from Millipore). The filters were weighed on a microbalance in a climate-controlled room at the National Institute for Working Life in Solna, Sweden. The filters were conditioned in the room for at least 15 h before being weighed. The micro-balance had a precision of 1 mg. S51


Glas et al.

Volatile Organic Compounds VOCs were collected by diffusive sampling on Perkin Elmer tubes packed with Tenax TA (Varian, 60–80 mesh). The analysis was performed by thermal desorption–gas chromatography (GC)–mass spectrometry (MS) as described by Sunesson et al. (2001), but with a GC oven temperature program initially held at 351C for 5 min, then increased by 101C/min to 2501C and held at 2501C for 5 min. Toluene was used as a standard for quantification and the concentrations of the various compounds were calculated as toluene equivalents. An uptake rate for toluene on Tenax TA of 0.43 ml/min (Wright, 1993; Plant and Wright, 1999) was used to calculate the volume of air sampled. Mass spectra for the different compounds were compared with spectra in the NIST-98 MS library, and the compound whose library spectrum best matched the observed spectrum (according to the NIST-98 algorithm) was accepted as the identity of the measured compound. Samples from the same building were always analysed in the same batch of samples. Ventilation Ventilation was measured in every participant’s office by an authorised technician. All buildings had mechanical ventilation with inlet devices in every room. The inlet airflow in each room was measured in the inlet channel or at the device. The location and type of device determined the method used, either a Swema Air Flow (Swema, Sweden) or Velocity Calc Model, 8355-M-S, (TSI Inc. Mn, USA). A visual inspection of ventilation efficiency was performed with smoke. The room volume was measured at the same time. Statistical Methods The function ‘‘variance component’’ in SPSS was used to calculate the variance of the mean concentrations of chemical species to which at least 20 persons were exposed. The value zero was used in the calculation when a species was not detected for a person. Two sources of variability in chemical exposure were considered: variability among buildings and variability among individuals. The corresponding variance component was estimated for each of these factors. The ratio of the variance among individuals to the variance among buildings was calculated for each compound (RIB). The contributions to the variance from buildings and from individuals were taken as equal when this ratio was between 0.9 and 1.1. To justify a participant pool of randomly chosen referents, the main variability in chemical exposure must be among individuals. We chose a threshold for this condition to be met: the interindividual variability must be greater than the interbuilding variability for at least 70% of the compounds. This threshold was set using the investigators’ judgement; there are no published standards for this value. PCA of the data is another method for determining whether individuals’ chemical exposures were independent of S52

Figure 2. Data (x1, x2, and x3) plotted in three dimensions, and the plane derived from the first (PC1) and second (PC2) principal components.

the building in which they worked. PCA compresses complex data to main structures and describes them by vectors termed PCs. In Figure 2, a set of data for three variables (x1, x2, and x3) is plotted as a simple example. The first PC (PC1) represents the direction of the largest variation in the data. The second PC (PC2) describes the largest variation orthogonal to the first PC, and so on. A PC consists of two parts, the scores and the loadings. The scores describe an observation’s distance to the centre of the PC. The loadings describe the direction of the PC. In this way, a series of vectors (PCs) can describe the variations in a set of data with many variables. This method shows relationships between samples and variables. Data were normalised for PCA such that every substance had the same mean concentration, and each distribution had the same variance. This was done to give all substances the same influence in the analysis. A partial least-squares discriminant analysis (PLS-DA, Eriksson et al., 2001) was performed to determine whether there were any differences in chemical exposure between men and women in the offices studied. In PLS-DA, vectors are constructed orthogonally to compare data in two categories, rather than showing the largest variation regardless of category, as is done by PCA. As in PCA, PLS-DA vectors consist of scores and loadings. Variances were calculated using SPSS v.10 (SPSS Inc., Chicago, USA). Simca v.10 (Umetrics AB, Umea˚, Sweden) was used for multivariate analysis.

Results and discussion In all, 119 VOCs met the criterion of having 20 or more participants exposed to them. Measured spectra were compared to reference spectra in the NIST-98 mass spectral library. The identity of the compound whose NIST-98 Journal of Exposure Analysis and Environmental Epidemiology (2004) 14(S1)


Glas et al.

library spectrum most closely matched the spectrum of a given detected compound was accepted as correct. No further attempt was made to verify the identities, because the aim of the study was to evaluate the variability in exposure that occurs in different buildings and that occurs among individuals and not to identify correctly the detected compounds. Proper identification of compounds will be done in a subsequent case-referent study that investigates possible systematic differences in chemical composition in air samples from people reporting and not reporting SBS symptoms. Compounds with similar mass spectra eluting at the same retention time are likely to be identical and were assumed to be the same in all samples. Likewise, compounds with different retention times and/or different mass spectra were not assigned the same identity. Therefore, the best identity match could not be used in some cases, and another identity was assigned. Owing to the uncertainty in the identities of the detected compounds, the chemical names of the 119 measured VOCs are not given in this paper, except for the examples given in Table 2. The compounds listed in Table 2 span the range of values for the ratio of variance among individuals to the variance among buildings. The identity of these compounds was verified using reference compounds, or the compounds had high spectral similarity to the library spectrum (490%) and GC retention times consistent with the suggested compound. Table 3 shows the results of statistical analysis of concentrations of aldehydes, nitrogen dioxide, ozone, and particles (i.e. those species not collected

by Tenax TA). For each substance included, the tables list the variance in exposure among all individuals, the variance in exposure among individuals in different buildings, and the ratio of the two variances. The sampling time varied between 920 and 2700 min. A total of 518 different VOCs were detected using Tenax TA samplers. Of these, 38 were detected in at least 90% of the samples. Among the 38 were the aldehydes from hexanal to undecanal; the olefins from decane to tetradecane; the aromatics benzene, toluene, m- and p-xylene, ethylbenzene, and styrene; and the terpenes limonene, a-pinene, and 3carene. The four substances found in the highest concentrations were limonene (400 ng/m3), tetrachloroethylene (330 ng/m3), benzene (170 ng/m3), and acetone (130 ng/ m3). All these compounds were also among those detected in at least 90% of the samples. The concentrations are calculated using the uptake rate of toluene on Tenax TA, and toluene standards measured with GC–MS. Thus, concentrations are reported as toluene equivalents. In total, 170 compounds were detected only in one sample. Aromatics are present in vehicle exhaust and can also be emitted from laser-printed and photocopied paper (Wolkoff et al., 1993). Paints, lacquers, glues, etc. can contain aromatic and aliphatic hydrocarbons. Wood and wood products are the main sources of terpenes in indoor air. Limonene is also used in many cleaning products and found in citrus fruit, which people may bring to work. Tetrachloroethylene (TCE) was used as a dry-cleaning agent at the laundry where the vests were cleaned. TCE was used about

Table 2. Variances among individuals and among buildings for a selected number of detected VOCs and the ratio between the two variances. Substance

Variance among individuals

Variance among buildings

RBI

Detected in no. of samples

Propylene glycol Benzene Myrcene a-Pinene Limonene o-Xylene 1,3-bis(1-methenyl) benzene 1-[4-(methylethenyl-phenyl) phenyl] etanone

6.3 76 0.69 8.6 760 2.2 0.13 1.4

0 0.013 0.035 2.6 320 2.2 0.28 5.9

N 5800 19 3.3 2.4 1 0.47 0.24

35 78 49 78 77 64 26 72

Table 3. Concentrations and numbers of samples with concentrations above the detection limit for formaldehyde, acetaldehyde, nitrogen dioxide, ozone and particles.

3

CH2O (mg/m ) CH3CHO (mg/m3) NO2 (mg/m3) O3 (mg/m3) Particles (mg/m3)

Min–max (average)

Detection limit

Found in no. of samples

Variance among individuals

Variance among buildings

RBI

3–18 (9) 2–27 (5) 0.3–84 (28) 11–320 (57) 0.01–0.1 (0.03)

0.44–1.3 1.8–5.3 0.3 o9–o53 o0.01

78 40 75 41 75

5.5 8.3 5.7 2300 0.0002

3.3 8.4 19 140 0.00078

1.7 0.99 0.29 16 2.5

The variance among individuals, the variance among buildings, and the ratio of the two variances are given for all substances.


S53

Glas et al.

30 m away in a closed system, but TCE contamination of the vests at the laundry cannot be excluded. The drinking water in Umea˚ is not chlorinated. Acetone can be emitted from many household products and building materials, but it is also found in exhaled breath. Concentrations of acetone in breath are between 2.9 and 4500 mg/m3 (Fenske and Paulson, 1999). The concentrations of individual compounds measured in this study are low. Wolkoff and Nielsen (2001) refer to several European and North American studies showing mean concentrations for the majority of individual VOCs in nonindustrial buildings of below 10 mg/m3. In our study, the maximum concentration of any single compound was 400 ng/m3 (limonene). One factor contributing to this difference may be the ventilation rate. The American Society of Heating, Refrigerating and Air Conditioning Engineers (62-1999) (ASHARE, 1999) recommends an outdoor air supply of 10 l/(s person), and the Swedish guidelines are 7 l/ (s person) þ 0.35 l/(s m2) floor area (AFS, 2000:42). In our study, the average outdoor air supply was 24 l/(s person), giving an average air exchange rate of 2.8 h1. This exchange rate is higher than the minimum recommended level and may contribute to the relatively low VOC concentrations. Also, the absolute concentrations of all VOCs collected on Tenax TA in this study are only approximate, since they are measured as toluene equivalents. Aldehydes can be formed when ozone reacts with unsaturated hydrocarbons (Atkinson et al., 1995). Accordingly, it has been found that VOCs emitted from fresh carpets react with ozone, yielding aldehydes. Terpenes also react with ozone to form aldehydes (Weschler, 2000). Formaldehyde is released from building materials and furniture. Acetaldehyde is a human metabolite and is found in expired air (Fenske and Paulson, 1999). It is also found in car exhaust. The average and maximum exposures to formaldehyde in this study were 9 and 18 mg/ m3, respectively. The corresponding values for acetaldehyde were 5 and 27 mg/m3. Zhang et al. (1994) sampled six residential houses and found average concentrations of formaldehyde and acetaldehyde of 68 and 5 mg/m3, respectively. Dingle et al. (2000) sampled 18 conventional offices and 20 portable office buildings and found average formaldehyde concentrations of 27 mg/m3 in conventional office buildings and 1138mg/m3 in portable office buildings. The WHO proposes a threshold formaldehyde concentration of 120 mg/m3 in indoor air. Nitrogen dioxide is formed in such combustion processes as the burning of fossil fuels in power plants or in motor vehicles. The WHO (1999) proposes a 1-h maximum exposure concentration of 200 mg/m3 and an annual maximum exposure concentration of 40 mg/m3. The concentrations measured in this study (28mg/m3 average, 84 mg/m3 maximum) are similar to those measured in homes described in studies summarised by Jones (1999). S54


Ozone concentrations varied from 11 to 320 mg/m3 and averaged 57 mg/m3. The WHO’s 1999 guideline for ozone in ambient air is a maximum of 120 mg/m3 for 8 h. In Sweden, the 8-h hygienic threshold concentration is 200 mg/m3 (AFS, 2000). One person in our study had an average exposure of 320 mg/m3 over the 1-week sampling period. That person was contacted, but he did not suffer from any health problems that might be associated with high ozone exposure, no ozone sources were found, and no differences between his room and his colleagues’ rooms were found that could explain his high ozone exposure. According to the laboratory that performed the analysis, there were no problems with analysis of the sample. Eight additional participants (10% of participants) were exposed to concentrations higher than 100 mg/m3. Particle concentrations were between 0.01 and 0.1 mg/m3 (0.03 mg/m3 average). The possible health effects of particles are not clear. Two experimental exposure studies of particles in offices have been published by researchers in the same department (Hauschildt et al., 1999; Pan et al., 2000). Hauschildt concludes that ‘‘dust does not seem to have any objective or subjective effects on humans’’, while Pan found that particles affected tear film break-up time and caused subjective effects among the participants. The mean particle concentration was 439 mg/m3 in the Hauschildt study and 394 mg/m3 in the Pan study. Kemp et al. (1998) performed an intervention study in which the concentrations of respirable suspended particles on two stories of a building were reduced from 310 and 190 to 38 and 33 mg/m3, respectively, by effective cleaning. The effects of this reduction in particle concentration were evaluated with questionnaires distributed before effective cleaning and 4 weeks after regular cleaning began. The most significant effects were reductions in reported eye irritation, throat irritation, dry unproductive cough, and nose irritation. The dust concentrations in the above studies were one order of magnitude higher than those in the offices included in this study. A recently published review by an interdisciplinary group of European researchers (EUROPART) (Schneider et al., 2003) concludes that there is inadequate evidence to use particulate mass or number concentrations in indoor air as generally applicable risk indicators of health effects in nonindustrial buildings.

Statistical Analysis Variances were calculated for concentrations of each of the 123 substances that fulfilled the criterion of having at least 20 persons exposed. The ratio of the variance among individuals and among buildings was calculated for each compound. Differences among individuals were the dominant source of variability (RBIo1.1) for 78% of the 123 compounds, and differences among buildings (RBIo0.9) were the dominant source of variability for 14% of the compounds. For 8% of the compounds, differences among individuals and differences among buildings contributed equally to the variability in concentration (1.14RBI40.9). As variances were Journal of Exposure Analysis and Environmental Epidemiology (2004) 14(S1)


Glas et al.

dominated by individual variation in more than 70% of the compounds considered, we deemed it appropriate to use random selection of referents in our case-referent study. Tables 2 and 3 give ratios of variance among individuals to variance among buildings for concentrations of selected chemicals. Propylene glycol is a component of many emollients, which may explain the total dominance of individual variance in propylene glycol exposure. Myrcene is used in fragrances, and the use of fragranced products varies among individuals. Motor vehicle exhaust and tobacco smoke are important sources of benzene and o-xylene. However, smokers are excluded in this study, and smoking at the workplace is prohibited in Sweden. There may be participants who have benzene adsorbed to their clothes because they live with smokers, but the primary cause of the large variation in benzene exposure among individuals is not yet explained. NO2 also originates from motor vehicle exhaust and other combustion processes, but variability in NO2 concentrations is largely explained by differences among buildings, and the levels correlate with the buildings’ proximity to heavily trafficked streets. This correlation and the lack of such a correlation in aromatics’ concentrations indicate that aromatics must have important sources of exposure besides traffic. Another way to analyse the results is to make a PCA plot of all 124 substances, as shown in Figure 3 (PC1 and PC2). Every point represents the chemical exposure of one person. The points form groups depending on exposure patterns. For example, persons in building 1 are located in the lower central part of the plot. This means that persons working in building 1 received similar chemical exposures. Other buildings, such as buildings 2 and 3, showed large variations in exposures among individuals. Individuals in building 2 had exposure patterns similar to individuals in building 4, and individuals in building 5 had exposure patterns similar to individuals in building 8. Figure 3 shows that there were similarities in the chemical composition of the indoor air within buildings and

differences in the indoor air among buildings. However, there were also similarities in the personal chemical exposures of people working in different buildings and differences in exposures among people working in the same building. Analysis of variance and comparison of variance components, and PCA are two methods for analysis of variability in individuals’ chemical exposures. While variance analysis considers individual substances, PCA detects patterns among all the data. Differences among individuals’ exposures, regardless of the building in which they worked, were the dominant contribution to variance for a majority of the substances. Thus, the chemical exposure for a group of individuals in a building is not homogenous, and performing stationary measurements of a compound can only give a crude estimate of each individual’s exposure to the compound. This is also well illustrated in Figure 3 by the points corresponding to individuals in building 2. People sit in different rooms, perform different tasks, and move around the building encountering different emission sources. We conclude that in a case-referent study, it is acceptable to use referents who work in the same building as cases. A PLS-DA of men’s and women’s exposures to the 123 compounds mentioned above shows a difference in exposures between the men and women (Figure 4). Men’s exposures are clustered in the lower left quadrant of the plot. Three different explanations are obvious: (1) chance; (2) men and women have different work tasks and move around differently in a building; (3) women and men have different habits in the use of perfumed hygiene and cosmetic products. If the difference is real, it is very interesting since the prevalence of SBS symptoms is more common among women than men. This finding will be investigated further, but is beyond the scope of the current paper.

General Considerations We sampled in each location for an entire work week. This provides an estimate of each person’s average exposure. The

Building

15

1 2 3 4 5 6 7 8

5

10 8 6

PC2

5

88 7

8

67

7

8 57 5 5 6

87 5 56 7 5 7

0 7

8

5

6

3 3

78

45 44 4 3

5 4

4

8

-5

8

11

4 4 11

1 1

3

3 3 43 23 3 2 2 2 2

3 2 2

2 2

2 2 2

2

2

2

12 2

1

-10 -1 4

-12

-1 0

-8

-6

-4

-2

0

2

4

6

8

10

12

14

PC1

Figure 3. PCA plot of all participants in the study (PC1 and PC2). Each point represents the chemical exposure of one person, and each number corresponds to a building. Journal of Exposure Analysis and Environmental Epidemiology (2004) 14(S1)

S55


Glas et al.

Women Men

4

8

8

8 3

8

2 7

t[2]

1 0

-2

4 6

8

7

5 67 7 7 5 5 75 78 8 5 7 6 6

5 6

8 4

43

1

1

8 81 4

3

1

4

5

5

2

2

3 3

23 3 2

2 4

23

2

4 4

2

2 2 22

1

1 4

1

7

3

2 2 2

2

2

1 3

-4 5 5

-6 -1 2

- 11

-10

-9

-8

-7

-6

-5

-4

-3

-2

-1

0

1

2

3

4

5

6

7

8

9

10

11

12

t[1]

Figure 4. PLS-DA plot of men and women. Every point represents the chemical exposure of one person. Each number (1–8) correspond to a building.

disadvantage of this approach is that the intervals where the highest exposure occurs cannot be discerned. It is not known whether the average or the peak exposure most affects the incidence of SBS symptoms. Since the concentrations of most compounds, except ozone, are far below all hygienic threshold values, our hypothesis is that long-term exposures are of importance in causing SBS symptoms. One possible source of error is that the participants performed their own measurements. We put great effort into minimising potential bias caused by differences in handling of sampling equipment. The study leader gave the participants a thorough demonstration of the equipment, and participants were supervised the first time they put on their equipment. Airflows through the pumped samplers were measured once the participants were wearing the equipment. This gave the study leader a final opportunity for individual inspection. The participants were also given written instructions on how to handle the samplers, including telephone numbers where the study leader could be reached in case of trouble. All participants at a workplace undertook the sampling during the same week and could therefore observe how others handled their equipment. Participants were instructed to note any deviations in operation that could affect measured concentrations. Noted deviations included pumps that had stopped, and one person who forgot to open one of the samplers at the beginning of one day. During the inspection at the introductory meeting, it was discovered that a few people had not mounted the VOC sampler diffusion caps correctly. They were taught the correct way to handle the caps. Liljelind et al. (2000, 2001) have performed studies on self-assessment of exposure and shown a negligible difference between self-assessment and expert measurements. Our impression is that the participants were able to handle the sampling equipment correctly. The VOCs collected on Tenax TA were reported as toluene equivalents. Toluene was used as a standard for quantification of all compounds, and the toluene uptake rate was used S56

for calculation of all concentrations. This introduces uncertainty in the reported concentrations of the individual VOCs, since both the GC–MS instrument response and the uptake rate on Tenax TA are different for each compound. However, it is not possible to use standards for all compounds in a study like this one. Also, for a majority of the measured compounds, the uptake rate has not been determined. Since the same method was used to quantify all samples, the results from the variance analysis and from the multivariate evaluation were not affected by the uncertainty in absolute concentration caused by the sampling and analysis procedure. Smokers were excluded from this study. This was done to avoid the risk that substances from the smoke would be adsorbed to the smoker’s clothes and interfere with the sampling. Also, it is prohibited to smoke at the workplace in Sweden.

Conclusions There can be large differences in chemical exposures among individuals working in the same building. This result implies that personal exposure measurements must be performed to assess accurately an individual’s chemical exposure. Workers in different buildings can have similar exposure patterns, and exposure patterns can vary significantly within a building. This statement is supported by the finding that for the majority of substances measured, variations in exposure among individuals are larger than variations in exposure from building to building. Our conclusion is that in a casereferent study, referents can be randomly selected, that is, a referent may work in the same building as the corresponding case. Our results show a difference in chemical exposures between men and women. This difference in chemical Journal of Exposure Analysis and Environmental Epidemiology (2004) 14(S1)


exposures may contribute to the difference in the prevalence of SBS symptoms between the sexes.

Acknowledgements Financial support for this project from Va˚rdalstiftelsen and the Centre for Environmental Research (CMF) is gratefully acknowledged. We thank Margit Sundgren, Margaret Rheń, Annika Hagenbjo¨rk-Gustafsson, and Lissi Thomasson for skilful technical assistance, and Ingrid Liljelind for the illustrations used in the informational material for the participants. This study has been approved by the ethical committee of Umea˚ University, dnr 00-101.

References AFS. 2000 Hygieniska gra¨nsva¨rden och a˚tga¨rder mot luftfo¨roreningar, Solna, Arbetarskyddsstyrelsens fo¨rfattningssamling, 2000; p. 3. Andersson B., and Andersson K. Determination of tertiary amines in air. Appl Ind Hygiene 1989: 4: 175–179. Angell W.J., and Grimsrud D.T. Establishing indoor air quality research priorities. Proceedings Indoor Air 2002, Vol. 2, Monterey, CA, 2002, pp. 972–977. ASHARE Standard 62. 1999. Ventilation for Acceptable Indoor Air Quality, American Society of Heating, Refrigerating and Air-Condition Engineers, Inc. Atlanta, GA, 1999. Atkinson R., Tuazon E.C., and Aschmann S.M. Products or the gas-phase reactions of O3 with alkenes. Environ Sci Technol 1995: 29: 1860–1866. Bornehag C.G., Blomquist G., Gyntelberg F., Ja¨rvholm B., Malmberg P., Nordvall L., Nielsen A., Pershagen G., and Sundell J. Dampness in buildings and health. Nordic interdisciplinary review of the scientific evidence on associations between exposure to ‘‘dampness’’ in buildings and health effects (NORDDAMP). Indoor Air 2001: 11: 72–86. Box G.E.P., Hunter W.G., and Hunter J.S. Statistics for Experimenters. John Wiley & Sons, New York, 1978 (ISBN 0-471-09315-7). Carlson R. Design and Optimization in Organic Synthesis. Elsevier, Amsterdam, 1992. Dingle P., Tapsell P., and Hus S. Reducing formaldehyde exposure in office environments using plants. Bull Environ Contamination Toxicol 2000: 64: 302–308. Eriksson L., Johansson E., and Kettaneh-Wold N., et al. Multi- and Megavariate Data Analysis. Principles and Applications. Umetrics Academy, Umea˚, 2001 (ISBN 91-873730-1-X). Fenske J.D., and Paulson S.E. Human breath emissions of VOCs. J Air Waste Mngt Assoc 1999: 49: 594–598. Ferm M. Validation of a Diffusive Sampler for Ozone. Abstract Book from International Conference Measuring Air Pollutants by Diffusive Sampling. Montpellier, France, 2001, p. 71. Hagenbjo¨rk-Gustafsson A., Lindahl R., and Levin J.-O., et al. Validation of a diffusive sampler for NO2. J Environ Monitoring 1999: 1: 349–352. Hauschildt P., Mlhave L., and Kjærgaard S.K. Reactions or healthy persons and persons suffering from allergic rhinitis when exposed to office dust. Scand J Work Environ Health 1999: 25: 442–449. Hodgson M.J., Frohliger J., Permar E., Tidwell C., Traven N.D., Olenchock S.A., and Karpf M. Symptoms and microenvironmental measures in nonproblem buildings. J Occup Med 1991: 33: 527–533. Jones A.P. Indoor air quality and health. Atmospheric Environ 1999: 33: 4535–4564.


Glas et al.

Kemp P.C., Dingle P., and Neumeister H.G. Particulate matter intervention study: a causal factor of building-related symptoms in an older building. Indoor Air 1998: 8: 153–171. Levin J.-O., Lindahl R., and Andersson K. High performance liquid chromatographic determination of formaldehyde in air in the ppb to ppm range using diffusive sampling and hydrazone formation. Environ Technol Lett 1988: 9: 1423–1430. Liljelind I.E., Rappaport S.M., and Levin J.-O., et al. Comparison of self- and expert assessment of occupational exposure to chemicals. Scand J Work Environ Health 2001: 27: 311–317. Liljelind I.E., Stro¨mba¨ck A., and Ja¨rvholm B., et al. Self assessment of exposure F a pilot study of assessment of exposure to benzene in tank truck drivers. Appl Environ Occupational Health 2000: 15: 195–202. Lindahl R., Levin J.-O., and Andersson K. Determination of volatile amines in air by diffusive sampling, thiourea formation and high-performance liquid chromatography. J Chromatography 1993: 643: 35–41. Pan Z., Mlhave L., and Kjærgaard S.K. Effects on eyes and nose in humans after experimental exposure to airborne office dust. Indoor Air 2000: 10: 237–245. Plant N.T., and Wright M.D. European diffusive sampling initiative: confirmation of BTX uptake rates. UK Health and Safety Executive Report , Ref. OMS/ 99/07, 1999. Pommer L., Fick J., and Andersson B., et al. Class separation of buildings with high and low prevalence of SBS by principal component analysis. Indoor Air 2004: 14: 16–23. Redlish C.A., Sparer J., and Cullen M.R. Sick-building syndrome. Lancet 1997: 349: 1013–1016. Schneider T., Sundell J., Bishof W., Bohgard M., Cherrie J.W., Clausen P.A., Dreborg S., Kildes J., Kjærgaard S.K., Lvik M., Pasanen P., and Skyberg K. ‘‘EUROPART’’. Airborne particles in the indoor environment. A European interdisciplinary review of scientific evidence on associations between exposure to particles in buildings and health effects. Indoor Air 2003: 13: 38–48. Stenberg B. Office illness F the worker, the work and the workplace. Umea˚ University Medical Dissertation, Umea˚, 1994. Sunesson A.-L., Gullberg J., and Blomquist G. Airborne chemical compounds on dairy farms. J Environ Monitoring 2001: 3: 210–216. Sunesson A.-L., Gullberg J., and Olsson-Ko¨hler K., et al. Multivariate evaluation of VOCs in homes and office buildings with and without known SBS complaints. Proceedings Indoor Air 2002, Vol. 4, Monterey, CA, 2002, pp. 84–89. Ten Brinke J., Selvin S., and Hodgson A.T., et al. Development of a new volatile organic compound (VOC) exposure metrics and their relationship to ‘‘sick building syndrome’’ symptoms. Indoor Air 1998: 8: 140–152. Weschler C.J. Ozone in indoor environments: concentrations and chemistry. Indoor Air 2000: 10: 269–288. WHO. Indoor air pollutants: exposure and health effects. EURO Reports and Studies 78, World Health Organisation, Copenhagen, 1983. WHO. Air Quality Guidelines. World Health Organisation, Geneva, 1999 (http:// www.who.int/environmental_information/Air/Guidelines/Chapter3.htm). Wold S., Eriksson L., and Sjostrom M. PLS in Chemistry. Encyclopedia Computational Chemistry. Wiley, New York, 1998. Wolkoff P., and Nielsen G.D. Organic compounds in indoor air F their relevance for perceived indoor air quality. Atmospheric Environment 2001: 35: 4407–4417. Wolkoff P., Wilkins C.K., and Clausen P.A., et al. Comparison of volatile organic compounds from processed paper and toners from office copiers and printers: methods, emission rates and modelled concentrations. Indoor Air 1993: 3: 113–123. Wright M.D. Diffusive uptake rates for the Perkin–Elmer tube F BCR ir sampling intercomparisons at Vito (Mol, Belgium) February 1991 – April 1992. UK Health and Safety Executive Report, Ref. IR/L/IA/93/3, 1993. Zhang J., He Q., and Lioy P.J. Characteristics of aldehydes: concentrations, sources and exposures for indoor and outdoor residential microenvironments. Environ Sci Technol 1994: 28: 146–152.

S57