The spatial distribution of congener-specific human PCB concentrations in a PCB-polluted region

Abstract

Serum PCB congener concentrations were measured in 602 adults living near a PCB pollution source in eastern Slovakia. We created iso-concentration maps for 21 PCB congeners by geocoding each participant’s place of residence and kriging. Concentrations of PCB congeners were inversely associated with the distance of the participants’ residence from the source of pollution. Congener-specific risk factors were derived, particularly for PCBs 52 and 153. We observed that the spatial distribution of serum concentrations was influenced by micro-climatic parameters and physicochemical properties of the congeners. PCB congener profiles strongly correlated with that of the PCB commercial product Delor 106, which was manufactured in the region. The iso-concentration maps indicate that the zones with the highest predicted congener concentration have a mean area of approximately 235.75 ± 188.56 km² and the mean enrichment of concentration of congeners in serum in these zones is about 5.12 ± 1.36. We estimate that depending on congener approximately 23 457±18 762 individuals with PCB concentrations exceeding health-based guidance values live in these zones.

Graphical Abstract

1. Introduction

Polychlorinated biphenyls (PCBs) are synthetic chemicals used for a variety of commercial and industrial applications. Exposure generally occurs as a mixture, rather than as a single PCB. The different types of PCBs, called congeners, are distinguished by the number and location of chlorines. The same properties that made PCBs desirable (e.g., water insolubility and chemical stability) contribute to their persistence in the environment, as they do not easily degrade. The food chain is currently considered the primary source of human exposure to PCBs (ATSDR 2014; Sonneborn et al., 2008). Transfer from mothers to infants is also an important exposure route, as PCBs readily cross the placenta (Lancz et al., 2015a), and transmission also occurs via breastfeeding (Lancz et al., 2015b). PCBs continue to be measurable in a high proportion of samples from the general population, including pregnant women (CDC, 2018). PCBs are persistent, with an estimated in vivo half-life in children of 3–9 years (Grandjean et al., 2008) and 5–10 years in adolescents (Wimmerová et al., 2011). Depending on the congener, PCBs may be carcinogenic to humans (Group 1) (IARC, 2016). The U.S. government has also expressed concern over PCB exposure: According to the U.S. Department of Health and Human Services Agency for Toxic Substances and Disease Registry, in 2017, PCBs ranked 5th out of more than 250 chemicals, given their frequency, toxicity, and potential for human exposure (ATSDR, 2017).

We recently described the spatial distribution of human exposure to PCBs in selected regions around a former production site, the Chemko Strážske plant, in the Michalovce district of Slovakia. Blood serum concentration from participants living in these regions served as an exposure biomarker (Wimmerová et al., 2015). Maps were produced using kriging, an interpolation technique in which the surrounding measured values are weighted to derive a predicted value for an unmeasured location. Weights were based on the distance between the measured points, the prediction locations, and the overall spatial arrangement among the measured points (GIS dictionary, 2018). One of the main conclusions was that in eastern Slovakia, humans can be affected at distances of up to approximately 70 km from the original point source of PCB contamination. For residents in this area, ΣPCB serum concentrations in adults approached limits established by the French Agency for Food, Environmental and Occupational Health & Safety (ANSES, 2010)¹, the limits suggested by the European Food Safety Authority (EFSA, 2005)² and health-based guidance values of the German Human Biomonitoring Commission (HBM-I and -II values) (Apel et al., 2017)³. The extent of exposure was markedly dependent on distance and direction from the source and on microclimatic characteristics of the region (Wimmerová et al., 2015). The present study continues this line of inquiry with the addition of measured congener-specific PCB concentrations, whose addition permits inferences on a congener by congener basis, which may provide additional details about how PCBs are distributed in this region.

The aim of the present work was to study the congener-specific spatial distribution of PCB serum concentrations using geostatistic prediction and to quantify the exposure risk from enrichment of a particular PCB congener in blood serum of subjects living in exposure zones.

One of the objectives of the study was compare the PCB congener signature in blood serum of participants with those of 4 commercial PCB products manufactured in the local, historical PCB manufacturer, Chemko Strážske.

2. Material and methods

2.1. Study population

We recruited 602 volunteers with permanent residence for more than 20 years in equal numbers of 150 in each of the 4 districts, Michalovce, Vranov nad Topľou, Humenné and Trebišov (Figure 1). Participants were equally distributed among sexes and 3 age groups (30–40, 41–50 and 51–60 years). Exclusion criteria were pregnancy, acute injuries, oncological diseases, diseases affecting blood pressure, and diagnosed hypertension of 2^nd or 3^rd stage (systolic blood pressure ≥159 mm Hg and diastolic blood pressure ≥99 mm Hg). We planned to study association between PCB exposure and incidence of hypertension (subclinical, small increases of blood pressure). Overt hypertension has been therefore excluded. Among the participants no one was a former employee of the Chemko plant.

Figure 1.

The map showing the districts (Vranov n.T., Humenné, Trebišov and Michalovce) from which the study volunteers were recruited and the position of the source of pollution, the Chemko Strážske factory.

2.2. Blood collection, clinical examination, health questionnaires

Each participant underwent a clinical examination (general physical, anthropometric data, blood pressure measurement) and blood sampling. Details on the handling of specimens and isolation of serum have been presented elsewhere (Jusko et al. 2010). The study protocol was approved by the Institutional Review Board at the Slovak Medical University. All participants gave informed consent for participation in the study. After enrollment, participants were administered a questionnaire by trained staff which elicited information about sociodemographic status, including education, occupation, marital status, tobacco and alcohol use, diet, family living environment, and health status. Geographic position of the place of residence of each subject was determined by a commercial GPS device or from Google Maps.

2.3. PCB and lipid measurement

Concentrations of 21 PCB congeners (IUPAC no. 28, 52, 74, 99, 101, 105, 114, 118, 123, 138, 153, 156, 157, 167, 170, 180,187⁺¹⁸², 189, 194, 196⁺²⁰³, and 199) were determined as described elsewhere (Drobná et al. 2011; Čonka et al. 2005; US EPA 2008). The sum of the 21 individual PCB congeners analyzed represents ΣPCBs. Serum samples were treated by modified solid-phase extraction (SPE). Each of the serum samples was thawed, spiked with a known amount of ¹³C-labelled compounds. Serum mixed with an equivalent amount of water: 1-propanol (85 : 15, v/v) mixture was applied to a conditioned SPE column (2 g C18, endcapped; Alltech, Deerfield, Illinois, USA). The analytes were eluted with an n-hexane: dichloromethane (1: 1, v/v) mixture, and the eluate was concentrated. The extract was cleaned-up on a multi-layer florisil–silica/H₂SO₄ column and eluated with n-hexane: dichloromethane (9: 1, v/v). The eluate was concentrated under a gentle nitrogen stream just to dryness. ¹³C -labelled recovery standard solution was added immediately prior to GC injection.

The measurements were performed using isotope-dilution method by a high-resolution mass spectrometer (HRMS, MAT 95 XP; Thermo Finnigan, Bremen, Germany) coupled to an HP 6890 gas chromatograph (Hewlett-Packard, Palo Alto, California, USA) with a DB-5ms column (60 m × 0.25 mm × 0.25 μm) using splitless mode injection. Helium was used as a carrier gas at a constant flow of 0.8 ml·min⁻¹. HRMS was operated at a resolution of 10 000 in the positive ionization mode at 53 eV. The proportion of the two most abundant ions of natural (¹²C) compounds and ¹³C-labelled ones monitored in the selected ion monitoring mode together with retention time matching provided sufficient identification criteria. Calibration was completed through the analysis of five calibration standard solutions, each containing the measured ¹²C and above-mentioned ¹³C-labelled compounds.

2.4. Quality assurance and quality control

The serum samples were treated and analyzed in sets of 10 together with one blank sample. A certified reference material of human serum (1589a, PCBs, Pesticides and Dioxins/Furans in Human Serum, NIST, Gaithersburg, MD 20899) was used for checking the analytical process accuracy. All analytical measurements were carried out at the National Reference Centre for Dioxins and Related Compounds (Department of Toxic Organic Pollutants, Slovak Medical University), which has been certified by the Slovak National Accreditation Service (ISO/IEC 17 025:2005, certification no. S-111) and regularly participates in inter-laboratory studies and proficiency tests on dioxins and PCBs in food and feed (EU-RL for halogenated persistent organic pollutants in Feed and Food, Freiburg, Germany) and interlaboratory comparison program on PCBs and OCPs in blood serum (G-EQUAS) organized by Institute and Out-Patient Clinic for Occupational, Social and Environmental Medicine of the Friedrich-Alexander University in Erlangen, Germany.

For the values below limits of detection (LOD), we took the LOD value divided by the square root of two if the PCB congener had fewer than 20 % of values below the LOD; otherwise, we used LOD values divided by two (Persky et al., 2001; Weisskopf et al., 2005). We estimated total serum lipids using the enzymatic summation method (Akins et al., 1989). PCB concentrations reported in this study are serum lipid adjusted.

2.5. Statistical analysis and kriging

The univariate distributions of serum PCB congener concentrations were positively skewed. We applied Box-Cox transformation to approach normal distributions using Sigma XL version 8.07 (SigmaXL Inc.).

The exposure predictors have been chosen using literature sources and results of bivariate analysis between PCB congener concentrations and age, sex, education, smoking (number of cigarettes /day, passive smoking, rate of smoking at home), consumption of foodstuffs known to contain PCBs (fish, pork, beef, poultry, eggs, pork lard, butter, milk and dairy products), and origin of food (homemade, retail chain, both). Predictors of the PCB serum concentration were examined by Spearman correlation, Mann-Whitney, and Kruskal-Wallis tests. We adjusted the PCB serum concentrations for age, sex, education and smoking by Mixed model from SPSS 19.0 (IBM Support). A p-value <0.05 was considered as statistically significant. We assume that after concentration data adjustment for age, gender, education and smoking the sole significant exposure predictor (determinant) is the distance from the source of pollution.

ArcGIS (Esri, Redlands, CA USA) was used to produce iso-concentration maps. The residential address of each participant, along with latitude and longitude coordinates, were entered to create maps providing information on the spatial pattern of each PCB congener quantified in serum of volunteers. The respective zones between the neighboring iso-concentration lines were colored by spectrum colors with the red end denoting the highest serum PCB concentrations and the blue end the lowest serum PCB concentrations. The optional output variance of prediction raster created contains the kriging variance at each output raster cell. Assuming the kriging errors are normally distributed, there is a 95.5 % probability that the actual z value at the cell is the predicted raster value, plus or minus two times the square root of the value in the prediction raster (ArcGIS software). The areas of the respective zones were calculated from the kriging maps after adjustment for Slovakia state borders by combining the Python programming language and the Calculate Geometry tool arcGIS. Finally, we manually cross-checked the selected zones.

Each zone resulting from kriging was defined with an upper and lower serum concentration limit. The areas with the highest concentration were denoted as zone 1, with less concentration as zone 2, etc. We calculated the midpoint PCB congener serum concentration for each zone using the formula for a midpoint of a class interval (Sciencing, 2017) as

Midpoint serum concentration = (Upper limit concentration + Lower limit concentration)/2.

We limited our analysis to the highest 3 zones, as the midpoint serum concentration in zone 3 was close to the median PCB serum concentrations of all participants. For summed zones 1+2 the midpoints were calculated as

Midpoint serum congener concentration for summed zones 1+2 = (Upper limit concentration of zone 1 + Lower limit concentration of zone 2)/2.

Similarly, we proceeded when summing zones 1+2+3.

To assess potential risk resulting from living in the polluted area, we used the outcomes of kriging PCB biomonitoring data. Congener-specific risk was assumed to increase with the size of the zone and with enrichment of the serum concentration. Increasing area of the zone increases the number of exposed individuals. For each PCB congener, we evaluated the risk resulting from the combination of the zone size and the predicted concentration enrichment in serum. We defined the PCB congener enrichment factor (CEF) as

CEF = (Midpoint serum concentration of PCB congener in zone n)/(Median PCB congener serum concentration in subjects of the cohort)

When assessing the risk, we calculated the risk factor (RF) using two methods: 1) expressing the size of the zone either as an absolute value in km² or 2) as a fraction of areas of zones 1+2+3, as follows:

RF (1) = (Area of zone n km²).{ (Midpoint serum concentration of PCB congener in zone n)/(Median PCB congener serum concentration of subjects of the cohort)}

or

RF (2) = {(Area of zone n km²)/(Sum of areas 1+2+3 km²)}.{(Midpoint serum concentration of PCB congener in zone n)/(Median PCB congener serum concentration of subjects of the cohort)}.

3. Results

3.1. Study subjects and exposures

The number of participants in the study was 602, 307 were female and 295 male. Of these, 80% were married, 36% had a basic education (8–9 years), 45% completed high school, and 19% attained a university degree. Forty-six percent of participants reported previously smoking, and 27% were current smokers. The unadjusted serum concentrations of ΣPCBs in volunteers from the four districts of eastern Slovakia are shown in Table 1. The highest serum levels were observed in the Michalovce district, home of the manufacturing facility that once produced PCBs. The congener profile of PCBs in serum of volunteers from the 4 districts is shown in Table 2. PCB 153 was the most abundant congener (by median), followed by (in order) PCBs 180, 170, and 138. The results of bivariate analyses in Table 3 show that statistically significant predictors are older age, male gender, less education, and smoking more than 3 months in life. No association was observed between total PCB serum levels and alcohol intake and some dietary patterns, including food rich in animal fat (eggs, meat, butter, milk and milk products). The results of bivariate analyses between serum concentration of sum PCBs and source of food are in Table 4

Table 1.

Unadjusted concentration of ΣPCBs (ng/g lipids) in blood serum of subjects from the four districts of eastern Slovakia.

District	Population	N	Mean	Standard deviation	Geometric mean	Percentiles
						5	25	50	75	95
Humenné	62 561	150	776.5	717.06	599.1	207.5	371.3	557.2	877.5	1956.8
Michalovce	110 713	150	2871.3	7084.7	1404.5	278.1	720.9	1333.0	2603.4	9154.5
Trebišov	105 605	151	728.3	587.6	597.6	246.6	373.9	590.8	910.8	1683.6
Vranov n.T.	80 607	151	727.8	757.9	527.5	139.7	331.6	509.8	849.6	1758.0

SD - Standard deviation

GM - Geometric mean

Table 2.

Descriptive statistics of PCB congener concentration (ng/g lipids) in blood serum of participants (n=602).

PCB congener	Count	% over LOD	Arithmetic mean serum concentration	SD	GM serum concentration	GM Standard deviation	Percentiles
							5	25	50	75	95
28	591	100	2.27	4.036	1.341	2.8	0.26	0.756	1.378	2.506	6.369
52	525	99.8	0.579	1.989	0.235	3.994	0.018	0.101	0.253	0.646	1.853
74	592	100	19.319	42.612	9.421	2.886	2.052	4.52	8.582	15.798	71.037
99	592	99.5	6.275	9.337	3.497	2.807	0.777	1.786	3.111	6.422	23.765
101	489	94.5	0.562	2.112	0.175	4.249	0.015	0.077	0.185	0.439	1.738
105	592	97.6	2.084	3.477	1.104	3.012	0.248	0.581	1.018	2.008	6.809
114	592	86.6	0.635	0.984	0.281	4.207	0.017	0.156	0.325	0.655	2.571
118	591	100	13.502	23.14	8.175	2.5	2.193	4.222	7.475	14.311	41.4
123	591	41.4	0.708	4.68	0.108	4.051	0.015	0.046	0.096	0.203	2.011
138	592	100	115.53	266.97	70.209	2.461	18.541	39.462	66.428	117.18	331.71
153	592	100	326.24	692.35	204.368	2.349	59.086	114.46	191.96	337.77	962.19
156	592	100	22.492	51.959	13.77	2.334	3.787	7.936	12.966	22.261	53.751
157	592	99.7	2.017	4.239	1.209	2.562	0.32	0.707	1.131	2.023	5.658
167	592	100	6.153	11.179	3.831	2.432	1.029	2.092	3.689	6.671	17.412
170	592	100	171.97	610.26	92.752	2.439	25.514	52.914	86.462	151.15	446.93
180	592	100	335.75	1142.2	180.238	2.437	49.946	101.91	167.09	287.34	882.42
187+182	592	100	75.939	169.87	38.496	2.828	8.719	18.792	33.19	70.776	269.8
189	592	99.8	5.864	36.664	2.679	2.579	0.728	1.482	2.473	4.29	14.387
194	592	100	62.231	569.82	21.273	2.672	5.443	11.297	19.211	33.903	153.92
196+203	592	100	21.265	52.518	10.838	2.708	2.656	5.513	9.654	18.516	74.92
199	592	100	88.291	587.16	32.073	2.925	7.179	15.474	27.876	57.691	267.24
ΣPCB	592		1279.5	3711.9	719.182	2.421	199.59	396.06	658.1	1171.7	3541.2

GM - Geometric mean

SD - Standard deviation

Table 3.

Examination of statistically significant predictors of serum concentration of sum PCBs. Results of bivariate analyses.

	Sum PCBs	N	Mean Rank	p
Age		592	-	<0.001
Gender	Male	290	318.06	0.003
	Female	302	275.79
Education	Primary school and high school without GCSE	212	311.00	0.022
	High school with GCSE	263	299.60
	University	115	257.57
Smoking more than 3 months in life	no	317	277.03	0.003
	yes	275	318.95

Table 4.

Relationship (tested by Kruskal-Wallis test) between origin of consumed food and serum concentration of sum PCBs.

Food	Source	N	%	Mean Rank	p
Fish	homemade	19	3.16	334.21	0.316
	retail	499	82.89	295.16
	both	78	12.96	272.62
	missing	6	1.00
Pork	homemade	73	12.13	281.30	0.805
	retail	420	69.77	295.35
	both	103	17.11	294.76
	missing	6	1.00
Beef	homemade	30	4.98	269.70	0.923
	retail	495	82.23	275.75
	both	34	5.65	285.12
	missing	43	7.14
Poultry	homemade	74	12.29	274.40	0.492
	retail	378	62.79	299.33
	both	145	24.09	290.11
	missing	5	0.83
Eggs	homemade	184	30.56	283.26	0.593
	retail	221	36.71	295.67
	both	191	31.73	300.82
	missing	6	1.00
Pork lard	homemade	133	22.09	281.79	0.225
	retail	320	53.16	292.99
	both	124	20.60	262.89
	missing	25	4.15
Butter	homemade	3	0.50	258.67	0.402
	retail	578	96.01	295.16
	both	15	2.49	237.47
	missing	6	1.00
Milk	homemade	7	1.16	271.71	0.202
	retail	548	91.03	292.72
	both	33	5.48	239.09
	missing	14	2.33
Dairy products	homemade	7	1.16	287.43	0.660
	retail	549	91.20	291.83
	both	40	6.64	317.00
	missing	6	1.00
Fruits and vegetables	homemade	134	22.26	262.53	0.013
	retail	132	21.93	279.24
	both	329	44.19	310.79
missing	7	1.16

3.2. Spatial distribution of PCB congener serum concentrations and exposure risk

Kriging was used to generate iso-concentration maps of blood serum concentrations of the PCB congeners listed in section 2.3. (Figure 2). PCB 123 was included, however owing to only 40 % of samples >LOD was not further evaluated. High resolution maps with Chemko plant and the district town Michalovce marked can be found in supplemental material. Specific information on the spatial distribution of each congener serum concentration and associated exposure risk were ascertained by visual inspection of each map, utilizing the RF and CEF calculations outlined in section 2.4.

Figure 2.

Iso-concentration maps for individual PCB congeners. The zones with highest enrichment are marked with the deepest orange color. The midpoint concentrations of a particular PCB congener in the zones and the size of the zones are shown in Table 3.

3.2.1. Visual inspection of iso-concentration maps

The spatial distribution of PCBs is congener, residence, and local climate dependent. The exposure decreases with increasing distance from the exposure source (the former PCB manufacturing facility), and this distance is the dominant predictor of exposure. The prevailing winds blow predominantly from the north-west to south-east and this pattern can be observed in the distribution pattern of serum concentrations. While the exposure gradient (width of zones) with most congeners steeply decreases in north-west direction, in the opposite direction, the zones are elongated and frequently fall beyond the examined area. Such configuration prevents determination of areas of several zones 2 (except PCBs 114, 153, 156, 167, 189 and 196) and most zones 3. In spite of the complexity of the entire system, the external environment, and human subjects living in it (Beyer and Biziuk, 2009), marked similarities in the pattern of spatial distribution of the serum concentration of PCB congeners 153, 156, 170, 180, 189, 194 and 196 are visible. Relationships between physicochemical properties of the congeners and serum concentration distribution pattern can also be inferred. The areas of the zones 1 of the low chlorinated congeners (PCBs 28, 52, 74, and 99) are larger compared with the high chlorinated ones, most probably owing to higher volatility of low chlorinated PCBs.

3.2.2. Assessment of risk parameters

The number of inhabitants to be exposed increases with the area of the central zones. The areas of the zones, the midpoint congener serum concentrations in the given zone/s and CEFs are shown in Table 5. We were not able to determine the zone areas of PCB 105. PCB 52 had the largest area of zone 1. For summed areas of zones 1+2 it was replaced by another low chlorinated congener PCB 28. The RFs were calculated two ways. RFs (1) expressing absolute value of the area are shown in Figure 3. The greatest value of RF (1) for zone 1 was attained by PCB 52, whereas for zones 1+2 by PCB 105 and 1+2+3 by PCB 157. RFs (2), expressing zone areas as a fraction of areas of zones 1+2+3, are shown in Figure 4. PCB153 presents the greatest exposure risk for residents living in combined zones 1+2+3.

Table 5.

The congener enrichment factors (CEFs), area in km² of zone 1 and combined areas of zones 1+2 and 1+2+3 for individual PCB congeners and the midpoint congener serum concentration in ng/g lipid. The congeners were ranked based on the size of zone/s. The mean±SD area for all congeners of zone 1 is 235.75 ± 188.56 km², zones 1+2 1523.7 ± 902.44 km² and zones 1+2+3 3196.8 ± 1535.92 km².

PCB Congener	Zone 1			PCB Congener	Zones 1+2			PCB Congener	Zones 1+2+3
	Area km2	Midpoint concentration	CEF		Area km2	Midpoint concentration	CEF		Area km2	Midpoint concentration	CEF
52	839	1.265	4.99	28	3269	3.504	2.54	157	6980	4.962	4.39
28	507	4.042	2.93	105	3198	4.208	4.13	28	5777	3.275	2.38
105	421	4.775	4.69	52	2775	1.12	4.43	138	5650	279.74	4.21
187+182	337	123.25	3.71	157	2548	5.183	4.58	187+182	4836	105.535	3.18
194	326	78.678	4.09	138	2344	291.59	4.39	105	4737	4.008	3.94
74	324	50.637	5.9	187+182	2041	110.11	3.32	52	3182	1.065	4.21
199+201	282	125.31	4.5	74	1936	44.52	5.19	99	3285	12.59	4.05
180	267	610.96	3.66	118	1762	30.59	4.09	74	2964	42.268	4.92
170	238	325.04	3.76	180	1403	545.96	3.27	118	2903	29.218	3.91
118	195	34.302	4.59	99	1320	13.484	4.33	180	2795	552.68	3.13
99	189	15.545	4.99	194	1301	70.166	3.65	194	2787	67.125	3.49
196+203	127	61.24	6.34	101	1256	0.8	4.32	170	2716	278.35	3.22
138	126	328.4	4.94	170	1125	290.58	3.36	167	2574	20.335	5.51
101	125	0.906	4.89	199+201	1069	111.569	4.0	199+201	2268	106.68	3.83
157	96	5.882	5.2	153	662	1578.9	8.22	196+203	2053	51.703	5.36
156	82	62.094	4.79	196+203	646	54.175	5.61	156	1971	52.645	4.06
114	81	1.908	5.88	167	543	21.496	5.83	101	1965	0.756	4.08
153	65	1765.5	9.2	156	523	55.156	4.25	153	1735	1515.7	7.9
189	57	16.616	6.72	189	428	14.774	5.97	189	1603	14.174	5.73
167	31	24.599	6.67	114	325	1.675	5.16	114	1155	1.584	4.88

Figure 3.

Risk factors, RFs (1), for population of the polluted area computed from the absolute size of the zones predicted by kriging and the enrichment of the PCB congeners (CEFs) in the blood serum. The congeners were ranked with respect to the RFs in Zone 1.

Figure 4.

Risk factors, RFs (2), for population of the polluted area computed from the relative size of the zones predicted by kriging and the enrichment of the PCB congeners (CEFs) in the blood serum. The congeners were ranked with respect to the RFs in Zones 1+2+3.

3.2.3. Behavior of congeners from the pollution source to human exposure

To understand the relationship between historical PCB production in Michalovce and human exposure, we compared the four commercial PCB products (Delors 103,104, 105, and 106) manufactured by Chemko Strážske to the congener-specific concentrations among the participants in our study. Based on the homologue composition, the approximate chlorine contents by weight of Delors 103,104, 105, and 106 were estimated to be 44, 50, 57 and 60%, respectively (Taniyasu et al., 2003). We then correlated the % (w/w) of PCB congeners of each of the Delor mixtures against the median PCB congener concentration (ng/g lipid) in blood serum of participants in our cohort (Figure 5, A-D). The serum PCB congener pattern in study participants correlated strongly with that of Delor 106. In addition, we observed that the Delor 106 signature correlated with the midpoint concentrations of PCB congeners in zone 1. The resistance to metabolism and the long survival of the high chlorinated PCBs in the human body is also reflected in correlation pattern of PCB congeners in serum (Figure 6). The matrix shows that the correlation between concentrations of congeners in serum increases with increasing chlorination. The correlation level between the concentration of the hexachlorobiphenyls, heptachlorobiphenyls and octachlorobiphenyls was r ≥0.7 with the exception of PCB 167 and PCB congeners 189, 194 and 196 with r slightly less than 0.7. The serum concentration of congeners with lower chlorination was less interrelated with highly chlorinated congeners.

Figure 5.

Correlation between percentage of PCB congeners (w/w) in various types of Delors and median PCB congener concentrations in blood serum of participants of the cohort (Figures A-D). Figure E shows correlation between percentage of PCB congeners (w/w) in Delors 106 and midpoint serum concentration of PCB congeners in zone 1.

Figure 6.

Pearson’s correlation coefficients between the concentration of PCB congeners in blood serum of particpants (n=602) of the cohort. For all colored fields p≤0.001.

3.2.4. Estimate of exposed population residing in zone 1

To estimate the mean number of persons residing in the highest exposure zones 1 has been attempted based on general demographic data on Slovakia. The population density for the entire examined area (the 4 districts shown in Figure 1) of 3616.45 km² is 99.5 (population per km²). Applying this value to the mean areas of 235.75±188.56 km² of zone 1, assuming an even population distribution across the whole area, we obtain a population estimate of 23 457±18 762 persons.

4. Discussion

We have determined a series of PCB congeners in blood serum of 602 adults residing in an area of 3616.45 km² (7.4% of the territory of Slovakia) polluted by organochlorines (Wimmerová et al., 2015). For each PCB congener, an iso-concentration map defining zones with predicted congener serum concentration was developed. We purposely restricted quantification of exposure risk to most central zones due to methodical difficulties with determining the areas of zones >2. The mean± SD area of zones 1 was 235.75 ± 188.56 km² which represents 6.52 ± 5.21 % of the total examined area of 3616.45 km². The mean ± SD CEFs in zones 1 was 5.12 ± 1.36 (Table 5). A person permanently living in zone 1, with a probability defined in section 2.4., attains an enrichment of a particular serum congener concentration by its CEF value. The greatest enrichment was observed with PCBs 153, 189, 167, and 196⁺²⁰⁸, 9.2, 6.72, 6.67, and 6.34, respectively.

We understand that the current PCB body burden in our cohort is a result of historical and recent intakes. A previous study demonstrates that PCB congeners 153, 138, and 180 contain a phenyl group with 2,4,5-substitution, and are resistant to biotransformation and elimination (Megson et al., 2013). Correspondingly in the present study, serum concentrations of these congeners were the highest among all congeners: PCBs 180, 153, and 138 (335.75, 326.24, and 115.53 ng/g lipid, respectively). Such behavior corresponds to our data on interrelations between PCB congeners in serum shown in the correlation matrix (Figure 6). These correlations indicate exposure to highly chlorinated biphenyls and their long survival in environment and consequently in human body.

Before analyzing the concentration data by kriging, we adjusted the serum concentrations for potential confounders. We confirmed the following exposure predictors: age, sex, education, and smoking more than 3 months in life. No association was observed between total PCB serum levels and alcohol intake and some dietary patterns, including food rich in animal fat (eggs, meat, butter, milk and milk products).

It is interesting that in surveys made on a site polluted by PCBs in Brescia, Italy in 2003 and 2013, in agreement with our data, no association was found between total PCB serum levels and tobacco smoking, alcohol intake and some dietary patterns (eggs, meat, butter, milk and cheese) (Magoni et al., 2016; Apostoli et al., 2005; Donato et al., 2006). Questionnaire data from an independent study has shown that most participants even from rural region had urban way of life and purchased in supermarkets most of the food items known as potential vectors of PCBs, however with low PCB level (Salgovicová and Pavlovicová, 2007).

The currently predicted exposure to PCBs in zones 1 is several orders of magnitude higher than health-based guidance values for sum of marker PCBs in blood (ANSES, 2010; EFSA, 2005; Apel et al., 2017). In particular, from the data for the Michalovce district (population 110 713) the 50^th percentile is 1333 ng/g lipids. Conversion of wet weight HBM-II: 7 μg/L to lipid-based limit gives 1000 ng/g serum lipid (sum of the three most abundant congeners 138, 153 and 180 times two) (Apel et al., 2017). The value of 1333 ng/g lipid indicates that every second inhabitant exceeds the HBM-II limit in the Michalovce district. Note that a similar comparison with the French (ANSES, 2010) and EFSA limits (EFSA, 2005) for two other exposed population groups has been done previously (Wimmerová et al., 2015).

Moreover, the “background” PCB level in eastern Slovakia, the starting level for CEF calculations, is high compared with other world regions. So, while NHANES data (CDC, 2011) for PCB 153 are approximately 20 ng/g lipids depending on age, gender, and race, the currently measured median PCB 153 concentration (Table 2) in our cohort was approximately 190 ng/g lipids.

The current study was targeting spatial distribution of serum concentration of PCB congeners in environmentally exposed adult population living around a source of pollution. The PCB body burden is an end stage of the largely unknown process following the entrance of the chemical stressor into the environment. Our prior contribution to this issue has shown from data gathered during years 2002–2004 on delivering women that the PCB level in serum was associated with the consumption of fat from locally sourced food products, however distance from pollution source was not taken into consideration (Sonneborn et al., 2008). At that time on average, more than a third of the fat consumed by these women from pork, lard and eggs came from locally produced foods. However, since that time the consumption pattern of the Slovak population has dramatically changed, especially after joining the EU in 2004, which has been reflected in questionnaire data of the current project summarized in (Table 4). A trend towards consumption of food purchased in retail chains has been observed. Bivariate analysis between food source and concentration of PCBs in serum (Table 4) did not show an association between consumption of any food products and PCB exposure. According to our data from the area of the 4 studied districts, the suspected PCB vectors, fruits and vegetables (Donato et al., 2006; Magoni et al., 2016), do not play role in PCB concentration in blood (Table 4).

In spite of this change in nutritional habits the PCB serum level in population of the most polluted district Michalovce during a period of more than a decade did not decrease markedly. Comparison of the current mean concentration of ΣPCBs of 2871 ng/g lipids for adults (n= 149, age (mean ±SD) 45.3±8.4) of Michalovce district (Table 1) with the value of 3105 ng/g lipids for a corresponding population sample (n= 1008, age (mean ±SD) 44.6±12.5) of Michalovce district in year 2001 (Petrik et al., 2006) shows that the exposure of adults in Michalovce district decreased negligibly over 14 years. This is in contrast to kinetics of the PCB exposure of residents around a former PCB production facility in Brescia, northern Italy, where a ban on locally produced food of plant origin was associated with a decrease of PCB serum levels within 10 years to about one third, the level observed in industrialized countries in last decades (Raffetti et al., 2017; Magoni et al., 2016). The differences between the exposure scenarios in Brescia and Strážske have to be stressed. While PCB polluted water was the source of exposure in Brescia, which was discharged in irrigation channels and hence accumulated in the soil of a nearby agricultural area, (Donato et al., 2006) in Strážske, such practice was never used.

Considering the significantly reduced oral intake of PCB vectors, the question arises which sources and mechanisms contribute to preserve the present (2015) almost constant PCB serum level of the adult population of the Michalovce district, comparable to the level observed in an age- and region-matched adult population 13 years ago (Petrik et al., 2006)? It may be inferred that in this population, redistribution of historic pollution is expected to be the major source of PCB exposure. This redistribution involves volatilization from soil and water into the atmosphere with subsequent transport in air and removal from the atmosphere via wet/dry deposition of PCBs bound to particulates (CIRCABC EU, 2011). In support of identification exposure routes other as oral intake may serve several observations. Tree bark samples were collected to identify the relative amounts and congener profiles of atmospheric PCBs dissolved into bark lipids from the gas phase in Anniston, Alabama, USA, where PCBs were manufactured from the 1920s until 1971. Results from Anniston show that organisms living near the PCB plant and landfills were exposed to very high concentrations of atmospheric PCB and a mixture of PCB congeners that may have included high molecular mass compounds near the plant and landfills (Hermanson, et al., 1989; Hermanson and Hites, 1990; Hermanson et al., 2003). Important from our perspective is that exposures since the end of production have remained high near that area, PCB congener profiles show persistent congeners 31 + 28, 52, 66, 153, 138, and 180 and bark PCB concentrations were dropping exponentially at a distance of about 7 km (Hermanson and Johnson, 2007). These results are applicable to other sites where PCBs were produced in a high temperature process. The Strážske exposure scenario bears many similarities to that of Anniston site. The same technology has been used, there are big stores of PCB distillation residues at the neighborhood of the producing facility and there is an adjacent creek which has been heavily contaminated with PCBs from surface drainage at the PCB plant. PCB off-gassing from these sources may contribute to uptakes alternate to the oral route.

The importance of the Brescia Caffaro contaminated site and its surrounding areas have been confirmed as primary source in driving PCB concentrations in air (DiGuardo et al., 2017). Importance of atmospheric pollution by PCBs demonstrates a study on outdoor air at 34 homes surrounding New Bedford Harbor during dredging of highly contaminated harbor sediments. Air concentrations were higher in neighborhoods closest to the harbor and contained slightly greater proportions of volatile PCB congeners (Vorhees et al., 1997). Increased PCB soil concentrations were described in Michalovce area (Kocan et al., 2001; Dömötörová et al., 2012) however, data suggests that soil contaminated with PCBs contribute little to human body burden as measured by serum concentrations (Kimbrough et al., 2010). PCBs may be taken up by some plants, but that is not typical (ATSDR, 2000). If dermal absorption can be excluded as an important contributor to PCB body burden in our cohort we consider inhalation an alternate exposure route. Air as a source of PCB exposure was nearly completely ignored until a decade ago (Robertson and Ludewig, 2011). Indeed, recently much attention is paid to inhalation exposures to low chlorinated PCBs (Basra et al., 2018; Carpenter 2015; Lehmann et al., 2015; Marek et al., 2017). These exposures are clearly dependent on congener volatility related to boiling point. However, there are many observations that the hexa, hepta, octa, and nona chlorinated biphenyls contribute to environmental exposures by inhalation. The atmospheric PCBs are predominantly in the gaseous phases showing prevailing occurrence of slightly chlorinated congeners, whereas the highly chlorinated ones are mainly under particulate form (Blanchard et al., 2006). Several authors reported significant concentrations of the highly chlorinated congeners in air samples (Vilavert et al., 2014; Norström et al., 2010; Cetin et al., 2017; Hao et al., 2017; Ampleman et al., 2015). Airborne emissions from sources of legacy pollutants may lead to inhalation exposure at levels comparable to, and sometimes higher than, dietary exposure (Currado et al., 1998; Harrad et al., 2006; Ampleman et al., 2015).

In our previous paper on spatial distribution of PCB serum concentrations (Wimmerová et al., 2015) we stated that no reports were published on distance from the point source as the main exposure determining factor. Interestingly, in a recent study an inverse correlation between total WHO-TEQ and distance to source of pollution was confirmed by multiple linear regression models (Chen et al., 2015). An argument for air transport of PCB may be the correspondence of orientation of the elongated exposure zones for all congeners with prevailing winds. It is not plausible that distance from the pollution source may drive the outcomes of PCB oral intake.

Conclusions

We assume that after adjustment to age, sex, smoking and education the serum concentration of 20 PCB congeners in serum of environmentally exposed adults is dependent on the distance of the subjects’ residencies from the source of pollution. Iso-concentration maps demonstrate the influence of micro-climatic parameters and physicochemical properties of the congeners on spatial distribution of their serum concentration. Highest congener-specific risk factors considering size of the zone with highest exposure and the congener serum enrichment were observed for PCBs 52 and PCB 153. The PCB serum signature correlated best with manufactured PCB product containing 60% of chlorine. We suggest examine the hypothesis that airborne emissions from local sources may lead to inhalation exposure at levels comparable to dietary exposure.

Highlights.

We determined PCB congeners in serum from 602 adults living in PCB polluted area.

We created PCB congener-specific iso-concentration maps using kriging.

The mean area of zones with highest predicted concentration was 235.75 ± 188.56 km².

The mean serum congener concentration enrichments in these zones was 5.12 ± 1.36

Depending on the congener 23 457±18 762 inhabitants live in these particular zones.