© 2006 British Occupational Hygiene Society Published by Oxford University Press
Original Article |
Identification of markers for PCB exposure in plasma from Swedish construction workers removing old elastic sealants
1 Department of NBC Defence, Swedish Defence Research Agency, FOI, SE-901 82 Umeå, Sweden; 2 Department of Occupational and Environmental Medicine, Örebro University Hospital, SE-701 85 Örebro, Sweden; 3 Department of Chemistry, Environmental Chemistry, SE-901 87 Umeå, Sweden
* Author to whom correspondence should be addressed. Tel: +46 90 10 6740; fax: +46 90 10 6806; e-mail: hakan.wingfors{at}foi.se
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIAL AND METHODSRESULTS AND DISCUSSIONACKNOWLEDGEMENTSREFERENCES |
|---|
The objective of the present study was to identify PCB-indicators of occupational exposure related to the removal of old elastic sealants. Blood samples were collected from workers involved in removing elastic sealants. Samples were also taken from age- and sex-matched controls. A majority of the exposed workers were re-sampled after 10 months. All samples were analysed for 19 PCBs. The levels in the exposed workers were twice as high as those in the controls, 575 and 267 ng g–1 lipid, respectively, but were essentially unchanged at the second sampling. The PCB patterns also differed. Levels of many less chlorinated PCBs were much higher in the exposed workers, compared to the controls, and principal component analysis (PCA) revealed that easily metabolized PCBs decreased in the exposed workers during the study. This finding indicates that elimination exceeded uptake during the study period, and that the safety information given to the workers had been effective. PCA was also used to identify exposure markers. The relatively persistent PCBs 56/60 and 66, the easily metabolized PCBs 44, 70 and 110 (with vicinal hydrogens in meta/para-positions) and the very persistent PCBs 153 and 180 were found to be good markers for occupational, recent occupational and background (dietary) exposure, respectively. A PCA model based on these markers was equally effective in differentiating between exposed individuals and controls, and between recent and less recent exposure, as a model based on all PCBs.
Keywords: multivariate, analysis • PCB • profiles • principal component analysis
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIAL AND METHODSRESULTS AND DISCUSSIONACKNOWLEDGEMENTSREFERENCES |
|---|
During a state-sponsored programme in the 1960s, a million flats were built in Sweden to overcome problems with overcrowding and to raise the general standard of living. Some 70–190 tons of PCBs were used in sealants for these buildings (Sundahl et al., 1999
; Johansson et al., 2003), but in accordance with contemporary legislation concerning PCBs in society the Swedish building sector recently decided to replace these sealants.
Construction workers involved in sealant removal operations may be exposed to PCBs. The PCBs may also leach from the sealants, and people working or living in the buildings may be exposed to them. Indeed, increased levels of PCBs have been found among workers involved in sealant removal in Finland (Kontsas et al., 2004), recent reports from Germany (Gabrio et al., 2000; Schwenk et al., 2002) indicate that airborne PCBs may contribute to the PCB blood levels in people working in buildings with PCB-containing elastic sealants. Furthermore, Johansson et al. (2003) found significantly elevated PCB levels in blood from residents in Swedish PCB-containing buildings. In some of these studies, air levels were also measured and the tri- and tetra-chloro-PCBs (mainly PCBs 28, 52 and 66) were found to be the most abundant, with total PCB concentrations of up to 10 µg m–3. This may seem strange, since primarily highly chlorinated PCB formulations (Arochlor 1248, 1254 or 1260, or equivalent) have been used in the elastic sealants (C Munther, personal communication 2004). However, the PCBs are semi-volatile and their vapour pressure depends on their degree of chlorination (Brorström-Lundén, 1995; Kaupp and McLachlan, 1999). The relatively less chlorinated PCBs will therefore have a greater tendency to evaporate.
It is a great challenge to distinguish occupational exposure from general exposure, and recent occupational exposure from earlier exposure. In our opinion it is essential to carefully select which PCBs to monitor in order to gain maximum information. The congeners selected should include good representatives of general and occupational exposure, respectively. Food consumption, in particular consumption of fatty fish, has been recognized as the major general exposure route for PCBs (Alcock et al., 1998; Schlummer et al., 1998). However, during sealant removal inhalation of vapours and dust generated together with oral (hand to mouth) uptake are the most likely occupational exposure routes.
However, it is also important to realize that the relative PCB levels will change after being taken up by an organism due to metabolic processes. PCB congeners that are resistant to metabolism will therefore be accumulated to a greater extent than easily metabolized congeners. Borlakoglu and Wilkins (1993), Niimi and Oliver (1983), and Andersson et al. (2001) have postulated some general rules concerning the structure of persistent and bioaccumulating PCBs. For instance, high degrees of chlorination in the biphenyl rings and a lack of vicinal hydrogen atoms usually favour enrichment in biota. PCBs with vicinal hydrogen atoms, especially in meta- and para-positions, are more easily metabolized by cytochrome P-450 enzymes. The cytochrome P-450 enzyme capacity and selectivity differs from species to species, resulting in species-specific PCB patterns. Thus, highly chlorinated PCBs with lateral substitution will be abundant in high organisms, including fatty fish, and may be good marker candidates for general exposure. Similarly, less chlorinated PCBs with vicinal hydrogens will be less stable and may be good markers for occupational exposure.
Several attempts have been made to minimize the number of PCB congeners that need to be monitored to assess general exposure (van den Berg et al., 1995; Atuma et al., 1996; Wicklund-Glynn et al., 2000). Usually, for practical and economic reasons, seven PCB congeners (PCBs 28, 52, 101, 118, 138, 153 and 180) are monitored to assess environmental exposure in Europe. These congeners are generally considered to be stable in the environment and may be good markers for human PCB exposure through food. However, additional congeners may be needed to capture additional information on occupational exposure. Luotamo et al. (1993) suggested PCBs 28, 33, 60, 66, and 74 as markers for occupational exposure of Arochlor 1242.
The aims of the work reported here were to characterize PCB exposure through analysis of the PCB patterns in blood plasma from construction workers involved in abatement of PCB-containing sealants and to find markers that can be used to distinguish not only occupational exposure from general exposure but also recent occupational exposure from earlier exposure. A multivariate statistics model was then created using the PCB marker concentrations in workers and controls, which may be used to facilitate the detection and characterisation of occupational exposure within construction workers.
|
MATERIAL AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIAL AND METHODSRESULTS AND DISCUSSIONACKNOWLEDGEMENTSREFERENCES |
|---|
Subjects and plasma samples
In spring 2002 an enquiry was mailed to contractors in central and southern Sweden who were involved in the domestic abatement programme for PCB in buildings. They were asked to report subjects who had had at least 6 months regular experience of PCB removal during the two preceding years. Forty subjects fulfilled the inclusion criterion and 36 of these (90%) gave informed consent to participate in the study, which involved a self-administered questionnaire on possible PCB exposure and other issues as well as blood samples. Twenty-eight of the workers volunteered to submit a second blood sample 10 months after the first and to fill in a questionnaire on intervening exposure. A control group of 33 age- and sex-matched construction workers without previous occupational PCB exposure was collected from construction companies in central Sweden. For comparison, a reference group from a previous Swedish study (Wingfors et al., 2000
) was also used.
Blood was drawn from the PCB-exposed abatement workers during workplace visits, whereas control subjects were sampled at nearby medical facilities. In each case, 
40 ml blood was collected in heparinized Vacutainer® tubes during the morning without previous fasting, but participants were asked to avoid a fatty early meal. Plasma was prepared, transferred to acid-washed glass vials and frozen at –80°C until analysis.
The study was approved by the medical research ethics committee of the Örebro County Council (decision no. 2002/93).
Chemical analysis
The plasma samples were coded and, thus, their exposure status was unknown to the analyst. They were also extracted and cleaned up in a randomized order. The PCBs were analysed in series of 8-20 samples according to Päpke et al. (1989) utilizing 2-propanol/hexane and diatomaceous earth (Chem-Elut) as a solid support to assist extraction of the lipid fraction containing PCBs. The method, including fat determination, has recently been described in detail (Wingfors et al., 2005). Three stable isotopes (13C12-PCB 28, 13C12-PCB 118 and 13C12-PCB 153; 20 ng each) were used as internal standards and octachlorostyrene was used as syringe (recovery) standard. After clean up on a gravity-fed multi-layer column packed with silica, and acid- and base-modified silica, the extracts were analysed using a high-resolution gas chromatography (Agilent 6890 GC)—mass spectrometry (Agilent 5973 MSD) system. A DB-5 MS column (J&W Scientific, Folsom, CA) was used for the GC, but the column length was reduced to 30 m length as compared to 60 m in Wingfors et al. (2005). Nineteen PCBs were quantified to obtain a comprehensive profile of the exposure. Congeners found in the sample that were not present in the standard were semi-quantified against an isomeric PCB. The most abundant ion was used for quantification, while relative retention times (±2 s) and the isotope ratios (±20%) for the two most abundant ions of each PCB were used for positive identification. A retention time database for Arochlor PCBs was used for the identification of unknown peaks (Harju, 2003). Freshly prepared standard solutions (Larodan-Sweden) of seven PCB-congeners were used for quantification. This standard mixture was also compared to a standard, containing 29 PCBs, which had been assigned values in an inter-laboratory study (Anonymous, 1997). The mean recovery of the three 13C-labelled internal standards was 59–69%, and the detection criteria were met for most samples, although 4% of the values for the monitored PCBs were below the limit of detection. These were mainly found within the control group for PCBs 44, 70, 56/60, 95, 87 and 110. The detection limit was 0.1–1 ng g–1 extracted lipids depending on the PCB congener and noise level. To check the precision of the procedure seven replicate samples were analysed and six replicate instrumental analyses were performed. In addition, seven procedure blanks were included. Analysis of the replicate samples resulted in a mean relative standard deviation (RSD) of 16% for all 19 PCBs. The RSD for the sum of the 19 PCBs was <5% for lipid-adjusted and <3% for wet weight-adjusted values. The deviation from measured and assigned values was <5% for the in-house standard. Blank samples were clean and free from interfering peaks.
Statistical evaluation
Human environmental PCB-data are often log-normally distributed (Wingfors et al., 2000), making arithmetic means and standard deviations inapplicable. In such cases, the geometric mean (GM) is the descriptor of choice (Miller and Miller, 2000). Analysing several PCB congeners in two groups to deconvolute complex exposure patterns is a challenging multivariate task. Ratios of specific congeners are sometimes used to compare groups, which can help to visualize differences between sets. Principal component analysis (PCA) is generally a better tool for simplification, classification and outlier detection and also for prediction and evaluation of variables of importance (Wold et al., 1984). In PCA a complex data matrix may be processed by a least squares method to extract the maximum variance in the data (captured in principal components; PCs), which may be vizualized as score (object-oriented) or loading (variable-oriented) plots. This statistical tool may be used for correlated variables and unbalanced data. In this study cross-validation was used to determine the number of significant components and to measure the predictive ability [Q2 = 1 – (predictive residual sum of squares/total variation in X)] of the resulting model. Prior to PCA analysis all PCB congener concentrations were individually normalized to PCB 153 (the most abundant PCB) to obtain a unit-less profile matrix (Wingfors et al., 2000). All variables were further log-normally transformed, to deal with skewness within the dataset, and finally mean-centred.
|
RESULTS AND DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIAL AND METHODSRESULTS AND DISCUSSIONACKNOWLEDGEMENTSREFERENCES |
|---|
The total level (GM) of the 19 PCBs was about twice as high in the exposed group as in the controls, 575 versus 267 ng g–1 lipid (Table 1). Among the exposed individuals that were sampled twice a concentration of 588 ng g–1 lipid was recorded at the beginning of the study period and 597 ng g–1 lipid after 10 months. Thus, there was no change (Student's t-test) in total PCB concentrations between the two sampling occasions.
|
At first glance it was also difficult to detect any dramatic difference in the PCB patterns. The highly chlorinated PCBs (e.g. PCBs 153, 138/163, 187 and 180) dominated in both exposed individuals and controls. This indicates that a significant proportion of the total PCB body burden stemmed from general background exposure (mainly dietary intake), which is rich in such persistent PCB congeners. However, it was still possible that the relative proportions of minor PCB congeners differed between the groups. Therefore, the GM concentration of each congener in the exposed group was divided by the corresponding value for the controls (Fig. 1) and it became apparent that two congeners, PCB 66 and PCB 56/60, were clearly elevated in the exposed group. High ratios were also observed for PCBs 28, 44, 52, 74, 101, and 105. Clear differences in PCB patterns between exposed and unexposed individuals were also visible in the reconstructed ion chromatograms. Figure 2 shows the tetra- and penta-CB traces of a control subject and an exposed individual. Many of these less-chlorinated PCBs are hardly detectable in the plasma of the control individual, but abundant in plasma from the exposed individual. The abundant hexachloro- and heptachloro-PCBs varied less between the two groups, indicating that background exposure may have been the major source of these substances.
|
|
In order to further analyse the information captured in the PCB patterns, PCA was applied. A PCA model of the PCB 153-normalized data (Table 1) resulted in two significant components (model A in Figs 3 and 4). The first component (PC1) explained 83% of the variance and was primarily influenced by the difference of highly chlorinated PCBs 138/163, 180, and 187 (Group 1 in Fig. 4), and the tetrachloro-PCBs, in particular PCBs 66 and 56/60 (lower right corner of Fig. 4) in the data. The second component (PC2) explained 8% of the variation and reflected the PCB persistency, with less stable (less chlorinated, with vicinal H in meta- and para-positions) PCBs in the upper portion. The predictive ability (Q2) of model A was 0.84.
|
|
In general, PCA model A was able to distinguish between the exposed and control groups. The exposed workers had higher PC1 scores and are therefore found to the right in Fig. 3. However, seven individuals of the exposed group appeared among the controls and the reference group. These subjects had PCB blood profiles resembling those of the controls, indicating that they had been less exposed than the rest of the workers in the exposed group. This may be because these workers were exposed for less time, made better use of protective equipment (or better equipment was available), or worked in buildings that were relatively lightly polluted with PCBs. However, the participants in this subgroup of exposed workers and in the control group are hereafter referred to as ‘less exposed’ individuals and the remaining exposed workers as ‘exposed’ individuals.
Along PC2 the exposed individuals were distributed depending on their relative proportions of easily metabolizable (Group 2 in Fig. 4) and ‘persistent’ tri-, tetra- and penta-CB congeners (Group 3 in Fig. 4). Individuals in the top right corner of Fig. 3 (model A) have relatively high percentages of easily metabolizable congeners, indicating recent exposure, and individuals in the lower right corner have higher percentages of persistent tri- through penta-CBs, indicating a stronger influence of past occupational exposure compared to recent exposure.
Interestingly, all but five of the individuals sampled twice appear to have been less exposed during the 10-month study period than prior to that period, as the percentage of easily metabolizable tri- through penta-CB congeners (with vicinal meta- and para-hydrogens) decreased between the two sampling occasions. The solid arrows in Fig. 3 illustrate this change (the arrow heads and tails correspond to the values from the second and first samplings, respectively). The change is also evident from the chromatograms in Fig. 5. The other individuals appear to have been more exposed during the study period than prior to it (dashed arrows in Fig. 3). However, three of these individuals seem to have had lower exposure at the beginning of the study as they were oriented close to the ‘less exposed’ group of individuals. Altogether, these findings indicate that the safety information given to the workers in connection with the study was effective.
|
The structure-persistency relationships discussed above agree well with the relative human accumulation (RHA) factors reported by Brown (1994)
for an extensive set of 146 PCB congeners, all of which are present in any of the Arochlor formulations used. The PCB structures and characteristics, as well as the RHAs, are given in Table 2. Easily metabolizable PCBs with RHAs lower than 0.05 are indicated in bold. All but one of these has adjacent H-atoms in the meta- and para-positions of the biphenyl rings. The data gathered in this study do not allow a fully quantitative measure of the metabolic rate of individual congeners. However, the mean change in individual PCB concentrations over the study period was calculated for the exposed individuals whose exposure apparently declined during that period, and it was found that the congeners that decreased the most also had the lowest RHAs. Thus, the levels of easily metabolizable PCBs declined more than the levels of more persistent congeners in these individuals. On the other hand, the individuals who seem to have had higher exposure during the study period than prior to it had a higher percentage of persistent tetra-CBs, in particular PCBs 56/60 and 66. This indicates that these are good marker candidates for general occupational exposure, whilst the PCBs with vicinal H in meta- and para-positions, such as PCBs 44, 70 and 110, are good candidates for recent occupational exposure. The levels of more highly chlorinated PCBs did not change significantly and abundant hexachloro-CBs like PCBs 153 and 180 may therefore be used as markers for general (dietary) exposure.
|
If these conclusions are generally true, it may be possible to categorize the dominant routes of exposure within a group of workers by only analysing seven components, i.e. PCBs 44, 56/60, 66, 70, 110, 153 and 180. To test this hypothesis, a new PCA model was created (model B in Figs 3 and 4).The resulting score plot was almost identical to that of model A, with a similar degree of explanation (PC1, 82% and PC2, 8%; Q2 = 0.84). Thus, it may also be used to differentiate between workers who are primarily subject to general background (dietary) exposure from occupationally exposed workers, as well as between occupationally exposed workers with primarily recent and primarily earlier exposure.
We therefore suggest that it will be more fruitful to study the proposed set of marker PCBs (PCBs 44, 56/60, 66, 70, 110, 153 and 180) to trace occupational PCB exposure, e.g. during the removal of PCB-containing sealants, than to monitor the traditional set of indicator PCBs (PCBs 28, 52, 101, 118, 138, 153 and 180). Alternatively, five PCB components (PCBs 44, 56/60, 66, 70 and 110) might be analysed in addition to the indicator PCBs. Whichever of these approaches is adopted, much more information on the type and degree of PCB exposure would be obtained at the same or slightly higher cost than if the traditional markers are used.
|
ACKNOWLEDGEMENTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIAL AND METHODSRESULTS AND DISCUSSIONACKNOWLEDGEMENTSREFERENCES |
|---|
The authors are indebted to all workers, PCB-exposed and unexposed, and to their employers for participating in the study. Lisbeth Viklund, Sibylla Robertson and Rigmor Fredriksson (at the Department of Occupational and Environmental Medicine, Örebro University Hospital) and Birgitta Johansson (at Haluxa Occupational Health Service, Örebro) organized and conducted the epidemiological fieldwork, and Ing-Liss Bryngelsson organized the dataset. The Department of Occupational and Environmental Medicine, Örebro University Hospital, supported the study.
Received March 17, 2005; in final form June 22, 2005
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIAL AND METHODSRESULTS AND DISCUSSIONACKNOWLEDGEMENTSREFERENCES |
|---|
Alcock RE, Behnisch PA, Jones KC et al. (1998) Dioxin-like PCBs in the environment–human exposure and the significance of sources. Chemosphere; 37: 1457–72.[Medline]
Andersson PL, Berg AH, Bjerselius R et al. (2001) Bioaccumulation of selected PCBs in Zebrafish, three-spined Stickleback, and arctic Char after three different routes of exposure. Arch Environ Contam Toxicol; 40: 519–30.[CrossRef][Web of Science][Medline]
Anonymous. (1997) IUPAC—Working Group: Halogenated Hydrocarbon Environmental Contaminants. ‘Study on the quality of methods for simultaneous determination of toxicologically relevant PCB congeners occurring in two fish oils and an analyte solution’.
Atuma SS, Linder C-E, Andersson Ö et al. (1996) CB153 as indicator for congener specific determination of PCBs in diverse fish species from Swedish waters. Chemosphere; 33: 1459–64.[CrossRef]
Borlakoglu JT, Wilkins JPG. (1993) Correlations between the molecular structures of polyhalogenated biphenyls and their metabolism by hepatic microsomal monooxygenases. Comp Biochem Physiol; 105C: 113–17.[CrossRef]
Brorström-Lundén E. (1995) Measurements of semivolatile organic compounds in air and deposition. Gothenborg, Sweden: Chalmers University of Technology.
Brown JF Jr. (1994) Determination of PCB metabolic, excretion, and accumulation rates for use as indicators of biological response and relative risk. Environ Sci Technol; 28: 2295–305.
Gabrio T, Piechotowski I, Wallenhorst T et al. (2000) PCB-blood levels in teachers, working in PCB-contaminated schools. Chemosphere; 40: 1055–62.[Medline]
Harju, M. (2003) Analysis of PCBs with special emphasis on comprehensive two-dimensional gas chromatography of atropisomers. Umeå, Sweden: Umeå University.
Johansson N, Hanberg A, Wingfors H et al. (2003) PCB in building sealant is influencing PCB levels in blood of residents. Organohalogen Compd; 63: 381–4.
Kaupp H, McLachlan MS. (1999) Gas/particle partitioning of PCDD/Fs, PCBs, PCNs and PAHs. Chemosphere; 38: 3411–21.[CrossRef]
Kontsas H, Pekari K, Riala R et al. (2004) Worker exposure to polychlorinated biphenyls in elastic polysulphide sealant renovation. Ann Occup Hyg; 48: 51–5.
Luatamo M, Patterson DG Jr, Needham LL et al. (1993) Concentrations of PCB congeners in Sera from workers with past and present exposure. Chemosphere; 27: 171–7.[CrossRef]
Miller JM, Miller JC. (2000) Statistics and chemometrics for analytical chemistry. 4th edn. Harlow, UK: Pearson Education Limited. ISBN 0 130 22888 5.
Niimi AJ, Oliver BG. (1983) Biological half-lives of polychlorinated biphenyl (PCB) congeners in whole fish and muscle of Rainbow Trout (Salmo gairdneri). Can J Fish Aquat Sci; 40: 1388–94.
Päpke O, Ball M, Lis ZA. (1989) PCDD/PCDF in whole blood samples of unexposed persons. Chemosphere; 19: 941–8.[CrossRef]
Schlummer M, Moser GA, Mclachlan MS. (1998) Digestive tract absorption of PCDD/Fs, PCBs, and HCB in humans: mass balances and mechanistic considerations. Toxicol Appl Pharmacol; 152: 128–37.[CrossRef][Web of Science][Medline]
Schwenk M, Gabrio T, Päpke O et al. (2002) Human biomonitoring of polychlorinated biphenyls and polychlorinated dibenzodioxins and dibenzofurans in teachers working in a PCB-contaminated school. Chemosphere; 47: 229–33.[Medline]
Sundahl M, Sikander E, Ek-Olausson B et al. (1999) Determinations of PCB within a project to develop cleanup methods for PCB-containing elastic sealant used in outdoor joints between concrete blocks in buldings. J Environ Monit; 1: 383–7.[Medline]
van den Berg M, Sinnige TL, Tysklind M et al. (1995) Individual PCBs as predictors for concentrations of non and mono-ortho PCBs in human milk. Environ Sci Pollut Res; 2: 73–82.
Wicklund-Glynn A, Wolk A, Aune M et al. (2000) Serum concentrations of organochlorines in men: a search for markers of exposure. Sci Total Environ; 263: 197–208.[CrossRef][Medline]
Wingfors H, Lindström G, van Bavel B et al. (2000) Multivariate data evaluation of PCB and dioxin profiles in the general population in Sweden and Spain. Chemosphere; 40: 1083–8.[Medline]
Wingfors H, Hansson M, Päpke O et al. (2005) Sorbent-assisted liquid-liquid extraction (Chem-Elut) of polychlorinated biphenyls, dibenzo-p-dioxins and dibenzofurans in the lipid fraction of human blood plasma. Chemosphere; 58: 311–20.[Medline]
Wold S, Albano C, Dunn WJ et al. (1984) Multivariate data analysis in chemistry. In Proceedings of the NATO Advanced Study on Chemometrics: Mathematics and Statistics in Chemistry, Cosenza. Bruce R, Kowalski R, editor, September 1983.
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||