Revisiting the Contributions of Far- and Near-Field Routes to

Subscriber access provided by Kaohsiung Medical University

Environmental Modeling

Revisiting the Contributions of Far- and Near-Field Routes to Aggregate Human Exposure to Polychlorinated Biphenyls (PCBs) Li Li, Jon Arnot, and Frank Wania Environ. Sci. Technol., Just Accepted Manuscript • DOI: 10.1021/acs.est.8b00151 • Publication Date (Web): 17 May 2018 Downloaded from http://pubs.acs.org on May 17, 2018

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 24

1

Environmental Science & Technology

2

Revisiting the Contributions of Far- and Near-Field Routes to Aggregate Human Exposure to Polychlorinated Biphenyls (PCBs)

3

Li Li1*, Jon A. Arnot1,2, and Frank Wania1

4

1. Department of Physical & Environmental Sciences, University of Toronto at Scarborough, Toronto,

5

Ontario, Canada, M1C 1A4

6

2. ARC Arnot Research & Consulting, Toronto, Ontario, Canada, M4M 1W4

7

Corresponding Author: Li L.; E-mail: [email protected]. Phone: +1 (647) 601-4450

8

ORCID:

9

Li Li: 0000-0002-5157-7366

10

Frank Wania: 0000-0003-3836-0901

11

TOC Art

12

ACS Paragon Plus Environment


13

Abstract

14

The general population is exposed to polychlorinated biphenyls (PCBs) by both consuming food

15

from far-field contaminated agricultural and aquatic environments, and inhalation and non-dietary

16

ingestion in near-field indoor or residential environments. Here, we seek to evaluate the relative

17

importance of far- and near-field routes by simulating the time-variant aggregate exposure of

18

Swedish females to PCB congeners from 1930 to 2030. We rely on a mechanistic model, which

19

integrates a food-chain bioaccumulation module and a human toxicokinetic module with dynamic

20

substance flow analysis and nested indoor-urban-rural environmental fate modeling. Confidence in

21

the model is established by successfully reproducing the observed PCB concentrations in Swedish

22

human milk between 1972 and 2016. In general, far-field routes contribute most to total PCB uptake.

23

However, near-field exposure is notable for (i) children and teenagers, who have frequent

24

hand-to-mouth contact, (ii) cohorts born in earlier years, e.g., in 1956, when indoor environments

25

were severely contaminated, and (iii) lighter chlorinated congeners. The relative importance of far-

26

and near-field exposure in a cross-section of individuals of different age sampled at the same time is

27

shown to depend on the time of sampling. The transition from the dominance of near- to far-field

28

exposure that has happened for PCBs may also occur for other chemicals used indoors.


Page 2 of 24

Page 3 of 24


29

Introduction

30

It has generally been accepted that the human population takes up polychlorinated biphenyls (PCBs)

31

predominantly via food originating from contaminated agricultural and aquatic environments (i.e., “far-field” environments) that are mostly distant from where PCB-containing products are used.1-3

32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59

This understanding is rooted in the fact that PCBs are so hydrophobic and persistent that they are able to magnify to appreciable levels along food chains.4 It is also intuitively plausible because the bulk (> 70%) of the PCBs has been used outdoors (e.g., in large electrical transformers)5 and PCBs have long been detected in many outdoor environmental matrices (e.g., soils and sediments).6-8 Reliable information on food contamination therefore is a key requirement when modeling the time-variant human exposure to PCBs. Temporally resolved food contamination information can be either obtained from empirical observations,3, 9 derived semi-empirically from the contamination of the physical environment,10 or modeled mechanistically by considering contaminant accumulation along food chains.11, 12 However, although models often characterize the contamination of food realistically, irritatingly they do not always perform equally well in terms of predicting human exposure or the resulting body burdens. For example, Breivik et al.11 underestimated the lipid-normalized PCB concentrations observed in Swedish mothers’ milk samples taken from 1972 to 1992, despite successfully reproducing the observed time-dependent contamination in fish, beef and dairy products from the region. The underestimation is more notable for lighter chlorinated congeners (e.g., with a difference of up to 10 times for PCB-28 and 2 times for PCB-153), and for observations at earlier times (e.g., before 1980).11 Likewise, a toxicokinetic model underpredicted by a factor of 2 the observed lipid-normalized PCB-101 concentrations among a Welsh population in 1990, in particular for individuals under the age of 35.10 Is it possible that exposure pathways other than the diet are associated with the underestimation? Not until twenty years ago did researchers and regulators recognize the importance of PCB uptake from contaminated indoor environments (i.e., “near-field” environments) where PCBs have been used as additives to paints, caulks and joint sealants.13-15 In a few residential and school buildings, PCB contamination remains startlingly high even four decades after their construction or renovation with PCB-containing materials between the mid-1950s and 1970s.16 Interestingly, some existing biomonitoring evidence suggests that certain near-field exposure routes, e.g. air inhalation and dust ingestion, have features that match those of the overlooked source of exposure. For example, lighter

60

chlorinated congeners (notably tri- and tetrachlorobiphenyls) are significantly more abundant in individuals living in contaminated indoor environments than those living elsewhere.17-21 Near-field

61

exposure may contribute more to the total PCB exposure among some vulnerable groups, such as

62

toddlers and K12 children.22 Furthermore, Meyer et al.23 observed PCB concentrations among

63 64

residents living in contaminated indoor environments four times higher than in a reference population, which cannot be explained by differences in food contamination. We therefore

65

hypothesize that neglecting near-field exposure is responsible for the underestimation in existing



66

PCB exposure modeling. However, to date, the historical indoor emissions and contamination of

67

PCBs have yet to be characterized mechanistically. While the human uptake via inhalation of

68

contaminated urban outdoor air has been taken into account in an earlier modeling effort,3 the human uptake from more contaminated indoor environments has not been incorporated into current models.

69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86

As such, the relative importance of far- and near-field uptake routes remains largely unexplored. In addition, human exposure is dynamic in nature, not only because the magnitude of individual exposure routes is age-dependent (e.g., due to anthropometric characteristics and human behavior that changes with age), but also because the contamination of exposure relevant environments varies in response to time-variant emissions. We therefore need to explore how the route-specific contributions vary over time. Here, we revisit the case of aggregate PCB exposure among Swedish women, previously investigated by Breivik et al.11 using only a far-field exposure model, by additionally taking into account emissions and contamination of PCBs in near-field environments and near-field uptake routes. To this end, we develop a mechanistic, dynamic modeling framework, which allows for an integrated simulation of PCB contamination of far- and near-field environments, accumulation along food chains, and ultimately time-variant concentration in humans. While we recognize some high-throughput screening models also couple and reconcile co-occurring far- and near-field exposure scenarios (e.g., the USEtox24 and SHEDS-HT25 models, and refs.26, 27), they are not ideal for this investigation because those models are static or steady state. While this work focuses on PCBs, which are “legacy” contaminants, its conclusions may also be beneficial for predicting, even

87

reducing and preventing, human exposure to a wider range of “emerging” contaminants presently used indoors.

88

Methods

89

Overview of modeling framework

90

Figure 1 overviews the modeling framework, which incorporates a dynamic substance flow analysis (Block I), a nested multimedia module describing chemical transport and fate in indoor, urban and

91 92 93 94 95 96

rural environments (Block II), a food-chain bioaccumulation module calculating contamination in food (Block III), and a one-compartment human toxicokinetic module calculating aggregate exposure and the resulting concentration in bodies (Block IV). Construction, parameterization and testing of Blocks I and II have been described in detail in Li et al.28

97

The model depicts the technosphere and the physical environment of the western Baltic drainage basin, which includes the territory of Sweden. The starting point of the modeling is the information

98

on annual import of PCBs to the modeled region (since 1931), and the respective consumption

99

indoors (e.g., mainly as additives to paints, caulks and joint sealants) and outdoors (e.g., mainly as

100

coolants and lubricants in electrical transformers and capacitors). As shown in Figure 1, the dynamic


Page 4 of 24

Page 5 of 24


101

substance flow module (Block I) calculates the accumulation of PCBs in indoor and outdoor

102

in-service products (i.e., in-use stocks), landfills and dumps (i.e., waste stock), as well as emissions

103

from industrial processes, use phase and waste disposal. Emissions from the indoor in-use stock serve as inputs to the indoor air compartment, and the remaining emissions are inputs to the urban air

104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130

and fresh water (Figure S1). Despite recognizing the fact that PCBs were used in some, but not all, Swedish buildings, we assume that the entire indoor environment in the modeled region is a single uniform space receiving homogeneous indoor emissions. Thus, our modeled concentrations represent an “average” indoor contamination level across Sweden. Next, based on the indoor and outdoor emissions, the nested multimedia module (Block II; Figure 1) calculates time-variant concentrations in various compartments (e.g., air, fresh water, estuarine water, soil, carpet, vinyl floor and organic film) of near-field (i.e., indoor) and far-field (i.e., urban and rural) environments.28 The module considers the mass exchange between the indoor and urban environments through ventilation (based on an observed median air exchange rate of 1.23 h-1), and that between the urban and rural environments through atmospheric (based on an average atmospheric residence time of 27.2 h) and fresh water advection (based on precipitation rate). It also describes permanent loss of the emitted PCBs from the environment via anthropogenic (e.g., cleanup of indoor surfaces) and natural (e.g., irreversible burial in sediments) removal processes. Furthermore, the food-chain bioaccumulation module (Block III; Figure 1) takes the rural environmental concentrations and simulates the bioaccumulation through terrestrial and aquatic food chains. PCB concentrations in organisms at the time of slaughter represent the contamination levels in food, which, along with the contamination in indoor and urban environmental compartments, serve as inputs to the human toxicokinetic module (Block IV; Figure 1). The human toxicokinetic module calculates the time-variant uptake of PCBs via far- and near-field routes, and the evolution of PCB concentrations by taking into account major elimination processes (egestion, urination, respiration, biotransformation, etc.). The modeled results are for an “average” Swede living in an “average” house, i.e., represent the overall contamination level of the general Swedish population. Both Blocks III and IV are based on the fugacity approach,29 whereby the concentration of a chemical (C; in mol m-3) in an organism or human compartment is the product of a fugacity (f; in Pa) and a compartmental fugacity capacity (Z-value; in mol m-3 Pa-1). The mass transfer rate (N; in mol h-1) is expressed as the product of the fugacity (f; in Pa) in the compartment from which the chemical

132

is coming, and a transfer parameter (D-value; in mol h-1 Pa-1). The model framework is applied to two PCB congeners (Table S1): PCB-28 is relatively volatile and degradable, and PCB-153 is much

133

less volatile and more persistent.

134

In the following, all organisms including the human, excluding the lumen of their gastrointestinal

135 136

tracts, are treated as single well-mixed compartments of aqueous and lipid phases. Accordingly, all exposure results are represented in terms of the chemical amount that penetrates the absorption

137

barriers (e.g., gastrointestinal tract wall, respiratory tract lining and skin) and is truly absorbed by

131



139

wildlife and humans (i.e., uptake or internal exposure30), instead of the amount that crosses an outer exposure surface but does not necessarily pass the absorption barriers (i.e., intake or external

140

exposure30).

138

141 142 143 144 145

Figure 1. Schematic overview of the modeling framework applied in this study. A substance flow analysis module (Block I) estimates the stocks and emissions of PCBs. The transport and transform of PCBs in indoor, urban and rural environments are simulated with a nested

146

multimedia module (Block II). Note that compartments irrelevant to chemical uptake by organisms and humans are hidden. A food-chain bioaccumulation module (Block III) and a

147

human toxicokinetic module (Block IV) calculate the far- and near-field exposures.

148

Food-chain bioaccumulation module

149

The bioaccumulation module builds upon the earlier ACC-Human model12 and considers two food


Page 6 of 24

Page 7 of 24


150

chains (Figure 1). The aquatic food chain consists of plankton, a planktivorous fish (represented by

151

Baltic herring) and a piscivorous fish (represented by Baltic cod) living in the estuarine environment.

152

The terrestrial food chain comprises vegetation (i.e., pasture grass for herbivores and vegetables for humans), and herbivorous animals raised either for meat (represented by beef cattle) or milk

153 154 155

(represented by dairy cattle). Note that the vegetation here is different from the vegetation compartment (representing deciduous and coniferous trees) defined in Block II.28 All the organisms

156

live in the rural environment.

157

In contrast to ACC-Human12 which uses equations of different type to describe the contaminant

158 159

dynamics in different organisms, we parameterize a generic equation (Eq. 1) incorporating processes applicable to different organisms (a process is set to zero if it is not applicable to an organism):

160

d (VB ⋅ Z B ⋅ f B ) = ( IN diet + IN water + IN air + IN soil ) dt − ( OUTegestion + OUTwater + OUTair + OUTbiotransformation + OUTlactation ) = ∑ ( DfB,i ⋅ f B,i ) + DwaterB ⋅ f W,rural + DairB ⋅ f A,rural + DsoilB ⋅ fS,rural

(Eq. 1)

i

− ( DegestB + DwaterB + DairB + DbiotransB + DlactB ) ⋅ f B 161 162 163

where, VB is the body volume of an organism, which can be calculated from its age-dependent weight (Figure S2) and an organism density (assumed to be identical to that of water); ZB is the

164

fugacity capacity of an organism, which is averaged from the fugacity capacities of water and lipid phases weighted by their volumetric fractions in the organism (Text S1); fB is the fugacity in the

165

target organism; t is time (in h). Other denotations are elaborated in Text S1 unless explained below.

166

As indicated in Eq.1, four uptake routes are considered (Text S1). (i) Uptake through eating prey

167

(INdiet) is applicable to species other than plankton and vegetation. (ii) Uptake from water (INwater) describes either gill ventilation of estuarine water by aquatic species or the drinking of freshwater by

168 169 170

terrestrial species. (iii) Uptake from air (INair) includes foliar mass exchange between the atmosphere and vegetation, and respiratory uptake by terrestrial animals. (iv) Uptake from soil (INsoil) results

171

from the unintentional ingestion of soil particles on the pasture or harvested feed by terrestrial

172

species, as well as root uptake by vegetation. We assume that intake rates of food, water, air and soil

173

are proportional to body weight (Text S1). Three absorption efficiencies are applied to convert the calculated intake to uptake: For gastrointestinal uptake (INdiet and INsoil), a gastrointestinal absorption

174 175 176 177 178 179 180

efficiency (ED) is organism-specific and calculated from the octanol-water partitioning coefficient (KOW; Text S1).31-33 For uptake from water (INwater) and air (INair), absorption efficiencies (EW and EA) are assumed to be 100% unless indicated in Text S1. In addition, we also consider five removal processes from organism bodies (Text S1). (i) Egestion via feces (OUTegestion) is applicable to organisms other than plankton and vegetation. (ii) Loss to water (OUTwater) includes gill ventilation for aquatic species and urination for terrestrial species. (iii) Loss to air (OUTair) includes foliar mass



Page 8 of 24

181

exchange between the atmosphere and vegetation, and exhalation by terrestrial animals. (iv)

182

Biotransformation (OUTbiotransformation) is applicable to all organisms. (v) Lactation (OUTlactation) is the

183

most notable process for dairy cattle.

184

For each organism, Eq.1 links its fugacity with the fugacities in prey (fB,i), water (fW, estuarine water

185

for aquatic species and fresh water for terrestrial species), rural air (fA,rural) and soil (fS,rural). Therefore, solution of this set of sequential equations yields the time-variant accumulation of a chemical in

186 187 188 189

individual organisms. We define the ages at which organisms are either eaten by predators, or slaughtered or harvested by humans (Figure S2). For example, the cattle are slaughtered in the 28th

190

month after birth. Thus, the chemical concentration in each kind of food is calculated from its fugacity at the moment the prey dies. For simplification, contamination of milk is assumed to equal

191

the fugacity in dairy cattle in the 28th month.

192

Human toxicokinetic module

193

The uptake and elimination of PCBs by humans are modeled using the following toxicokinetic module (Eq. 2):

194

195

196 197 198 199 200

d (VH ⋅ Z H ⋅ f H ) = N far-field + N near-field dt − ( DeH + DurH + DreH + DmH + DperH + DlH + DchH ) ⋅ f H

(Eq. 2)

in which, the product of an “effective” body volume (VH in m3), the fugacity capacity of the human body (ZH in mol Pa m-3) and the human fugacity (fH in Pa), represents the total amount of a chemical within the human body (in mol); t is time (in h); DeH, DurH, DreH, DmH, DperH, DlH, and DchH are D-values (in mol Pa−1·h−1) representing chemical loss from humans via egestion, urination,

201

respiration, biotransformation, percutaneous excretion, lactation, and childbirth, respectively. The parameters above are calculated as in ACC-Human.12 The temperature of the human body (including

202

skin and hands) is constant at 37 °C.

203

In Eq.2, Nfar-field and Nnear-field sum up the far- and near-field uptake rates (in mol h-1), respectively.

204

The far-field uptake includes:

205

(I) Diet: Ndiet =

∑(D

fH,i

⋅ f fH,i ) . For infants younger than 6 months, human milk is the only dietary

i

206

item; the consumption rate is taken from ref.12 and assumed to be constant throughout lactation. Note

207 208

that while uptake through breast-feeding is accounted for in our calculation it is not discussed in this paper. Children and adults eat fish, beef, dairy, and vegetables, whereby Sweden-specific

209

consumption rates of different kinds of food are taken from refs.12, 12

34-36

and are extrapolated

to account for age dependence. To remain consistent with earlier studies,11, 37, 38

210

according to ref.

211

we do not consider (i) losses or additions of PCBs during food processing and preparation, (ii)


Page 9 of 24


212

consumption of food imported from outside the modeled region, and (iii) temporal changes in diet

213

composition;

214

(II) Ingestion of soil particles (including soil and settled dust39) from the urban environment: Nsoil =

215

DsoilH ⋅ fS,urban . We adopt the age-dependent soil ingestion rate recommended by the USEPA;39

216

(III) Drinking water that originates from rural fresh water: Nwater = DwH ⋅ f FW,rural . An age-dependent

217

water consumption rate is taken from ref.12;

218

(IV) Inhalation of outdoor air (gaseous and particulate phases): Nair,outdoor = DreH,outdoor ⋅ f A,urban . As

219 220

recommended, we assume that an individual spends on average 10% of the time outdoors.39 An age-dependent respiration rate is taken from ref.12.

221

The near-field routes include:

222

(I) Dermal permeation: Ndermal = DdeH ⋅ f A,indoor , which is described as two processes in series (Text

223

S2):40 (i) diffusion across a stagnant air layer above the skin, quantified with an air-side mass transfer

224

coefficient and a bathing removal efficiency, and (ii) permeation through the skin (specifically the stratum corneum and/or viable epidermis), quantified with a mass transfer coefficient within the skin

225 226 227

(also named a dermal permeability coefficient). The relationship between age and the total body surface is parameterized as recommended in ref.16, 18. For simplification, we do not consider the

228

impact of clothing on dermal permeation;

229

(II) Non-dietary ingestion of PCBs from indoor compartments via (i) “hand-to-mouth transfer”,

230 231

which involves mouthing or sucking hands that had contact with contaminated indoor surfaces and dust thereon, and (ii) “object-to-mouth transfer”, which involves licking or chewing contaminated

232

indoor objects (e.g., smoking) and dust thereon.41, 42 Non-dietary ingestion is calculated by Nnondiet =

233

DXH ⋅ f X (X = carpet, vinyl floor, organic film on hard surfaces). Hand-to-mouth transfer is treated

234

as two sub-processes in series (Text S2): (i) surface-to-hand skin transfer, which depends on

235

frequencies and contaminant transfer efficiencies for contact between hands and different surfaces, and is corrected for removal by hand washing; and (ii) hand-to-mouth contact, which depends on

236 237 238

frequency and contaminant transfer efficiency of mouthing. We assume that the fugacity in indoor objects is the same as that in the indoor organic film. Age-dependent contact frequencies are taken

240

from ref.43 (for children under the age of 11) and ref.41 (for other ages). The area of the hands is 5% of the total body surface,39 with a palm accounting for 40% of the area of a hand.42 The lip area

241

available for mouthing is assumed to be 0.5% of the total body surface;42 and

239



242

(III) Inhalation of indoor air (gaseous and particulate phases): Nair,indoor = DreH,indoor ⋅ f A,indoor .

243

For gastrointestinal absorption (dietary and non-dietary ingestions, and ingestion of soil particles), we use a KOW-dependent absorption efficiency (ED)33 to convert calculated intake to uptake. For

244 245 246 247 248 249 250 251 252 253

inhaled air, the absorption of gaseous contaminants is associated with an efficiency (EA) of 70%, and that of contaminants in the particulate phase is based on particle size-specific deposition fractions in the human respiratory system.42 Contaminants in ingested water are assumed to be fully bioavailable, i.e. EW = 100%. The model simulates the lifetime (up to the age of 80) aggregate exposure and the resulting body burden of females in different birth cohorts. To include her perinatal exposure, we additionally simulate the body burden of her mother, who gives birth at a user-defined age, up to the end of breast-feeding. The initial fugacity of every newborn and the fugacity of milk are equal to the

254

fugacity of the mother. The initial fugacity of the mother and the fugacity of the milk she was fed are assumed to be negligible, as the perinatal exposure of a woman has little effect on her body burden

255

during child-bearing age.44

256

Results and discussion

257

Evaluation of model performance

258

Modeled concentrations in the indoor, urban and rural environments (outputs of Blocks I and II) have

259

been evaluated in Li et al.28 Briefly, the modeled concentrations agree with the means or medians of observed concentrations in indoor air, as well as in air, fresh water, estuarine water and soil in the

260 261 262

rural environment of the modeled region, with differences of less than a factor of two. In particular, our modeled peak concentrations in indoor air (~45 ng m-3 for PCB-28 and ~13 ng m-3 for PCB-153)

263

are within the ranges (14 – 296 ng m-3 for PCB-28 and 1.5 – 45 ng m-3 for PCB-153) measured in

264

rooms with known PCB sources.45-47

265

We evaluate here the performance of the food-chain bioaccumulation module (Block III) and human toxicokinetic module (Block IV) by comparing model predictions with literature-reported

266 267 268 269 270 271 272

observations (available since the late 1980s). Figures S3 and S4 compare the predicted and observed PCB-28 and PCB-153 concentrations in herring and cod; Figure S5 presents the comparison for PCB-153 in beef (data for PCB-28 not available in the literature). Overall, our predictions are in reasonable agreement with observations. Figures S3 – S5 also compare our predictions with those by Breivik et al.,11 demonstrating that our predictions peaked earlier, and are higher, than theirs before the mid-1970s, but the two predictions converge for the period after the mid-1970s. Whereas the

274

herring data indicate that our predictions fit better with earlier observations than Breivik et al.11, the scarcity of congener-specific food contamination observations before the 1980s prevents us from

275

determining which model’s prediction is more reasonable.

273


Page 10 of 24

Page 11 of 24


276

Figure 2 compares the modeled time course of lipid-normalized PCB concentration in mothers’ milk

277

with observations.48, 49 The mothers were 27 to 31 years old and 55% – 75% of them were nursing

278

their first infant.48 We therefore use concentrations predicted in women aged 27, 29 and 31 immediately after bearing their first child for comparison. Modeled concentrations agree with

279 280 281 282 283 284 285 286 287 288 289 290

observations within a factor of two. It is worth mentioning that our calculation is for a hypothetical “average” Swede living in a hypothetical “average” house, whereas in reality, only some Swedes lived in houses with indoor PCB sources. Figure S6 presents results of our additional calculation, which predicts that individuals living in PCB-contaminated buildings would be 13 (PCB-28) and 3 (PCB-153) times more contaminated than their counterparts living in PCB-free buildings, if we assume that PCBs had been used in 10% of Swedish buildings in which 10% of Swedes lived. Earlier monitoring work observed that the median concentrations of PCB-28 and PCB-153 in blood of Swedes living in flats with PCB-containing sealants are 30 (p < 0.001) and 1.3 (p = 0.35) times of those of Swedes living in uncontaminated flats.50 Furthermore, despite the remarkable difference between the concentrations predicted in the two groups of Swedes, the average concentration

291

weighted by the relative size of the two groups is quite close to our modeled result for the hypothetical “average” Swede. This suggests that the modeled result is insensitive to the assumption

292

as to what fractions of the Swedish population lived in buildings with and without PCB emissions.

293

Our model performs much better than the model by Breivik et al.11 (Figure 2), which not only

294

underestimated the observations but also failed to reproduce the observed temporal trend. Such an improved modeling performance is mainly attributed to two of our modifications to work by Breivik

295 296 297 298 299 300 301 302 303 304 305 306

and colleagues.11 First, instead of assuming constant emission factors for a total in-use stock combining both indoor and outdoor uses,11 we distinguish between indoor and outdoor emissions and consider a rapid decline in emissions from indoor in-use stocks (see description of Block I in Li et al.28). Therefore, compared with Breivik et al.11, our approach predicts heavier contamination of the regional environment prior to mid-1970s. Second, in addition to the dietary exposure considered by Breivik et al.11, our approach incorporates the contribution of near-field routes to the human body burden. As demonstrated in the following sections, near-field exposure is more notable for individuals born in earlier times due to the remarkable indoor contamination at the time. The above two reasons point to severe regional emissions and contamination in the past, as well as higher relevance of near-field exposure routes to historical PCB concentrations, thus explaining the reason

307

that Brevik et al. “tends to significantly underestimate the earlier values while being more in line with recent measurements”.11 In fact, Breivik et al.11 and Wood et al.51 also realized that their

308

emissions prior to the 1970s had been too low, as modeling performance was improved by

309

additionally assuming a 5% PCB loss during industrial processing. However, this 5% lacks a

310 311

mechanistic explanation, and it is likely an overestimate according to a follow-up substance flow analysis.52 Instead of incriminating industrial processes, our work attributes the underestimation to

312

higher emissions from indoor in-use stock in early years.



313 314 315 316

Figure 2. Comparison between modeled and observed48, 49 PCB-28 (a) and PCB-153 (b) concentrations in Swedish first-time mothers’ milk. Modeled results in Breivik et al.11 are also

317

presented for comparison. The single observational value for each year before 1996 represents a pooled sample comprising over 100 individual milk samples,48 whereas

318

observations after 1996 are presented in means and standard deviations.49

319

Additional calculations were performed to evaluate the relative contribution of the two modifications

320

to our improved model performance relative to Breivik et al.11 (Figure S7). For PCB-28, the update

321 322

of emission estimate is insufficient to explain the better fit with observations, because the modeled concentrations are still only half of the observed ones even if all atmospheric emissions occur in the

323

rural environment, as assumed in Breivik et al.11 This suggests the importance of taking into account

324

near-field exposure routes. For PCB-153, using the updated emission estimate but moving all atmospheric emissions to the rural environment can also lead to a satisfactory agreement between

325 326 327

modeled and observed concentrations. Whereas ignoring the indoor and urban emissions omits the contribution from near-field exposure, it also adds the part of the PCB emissions that is predicted to


Page 12 of 24

Page 13 of 24


328

be retained in the indoor and urban environments (~31% of the cumulative emissions by 201528) to

329

the rural contamination. However, other parameters, e.g., the human biotransformation half-life

330

(HLB), can be highly uncertain and variable. It is possible that equally good agreement between model and observations could also be achieved from parametrizations that do not consider near-field

331 332 333

exposure, e.g., if the rural environment is assumed to have received all atmospheric PCB-28 emissions and a HLB at the lower end of the known range is applied (Figure S8). Therefore, further

334

investigations may still want to thoroughly examine possible alternative explanations.

335

Longitudinal aggregate exposure of cohorts born at different times

336

We begin our discussion by illustrating how the age-dependent change in human exposure factors (i.e., anthropometric characteristics and human behavior) influences the relative contributions of

337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358

near- and far-field routes to the aggregate human exposure to the two congeners. To eliminate the confounding effect of time-variant environmental contamination, we assume hypothetical constant emissions (e.g., 1 unit per time) of PCBs to either indoor (Figures 3a and 3c) or urban air (Figures 3b and 3d), which results in steady-state contamination of indoor, urban and rural environments. Figure 3 indicates that far-field routes contribute similarly to human total exposure with either indoor or urban emissions. This is because the contamination of the rural environment, where food originates, responds to the total regional emissions, irrespective of whether the emissions take place in indoor or urban environments.28 However, near-field exposure to PCB-28 almost equals far-field exposure in the case of indoor emissions (Figure 3a). Indoor emissions also increase the aggregate exposure to PCB-153 but the increment is less prominent than in the PCB-28 case (Figure 3c). In addition, when emissions occur outdoors (Figures 3b and 3d), near-field exposure (mainly inhalation of indoor air) does exist, albeit very small, because of ventilated contamination from outdoor air. During the lifetime of an individual, dietary uptake and inhalation peak in young adulthood; non-dietary ingestion is notable in childhood because mouthing behavior is frequent before adolescence (Figures 3a and 3c). For all ages, diet dominates the human far-field exposure to the two congeners. Inhalation of indoor air is always the single dominant route for near-field exposure to the more volatile PCB-28 (Figure 3a), whereas non-dietary ingestion contributes most to children’s near-field exposure to the less volatile PCB-153 (Figure 3c). Our identified dominant near-field exposure routes agree with an earlier study,42 in which PCB-28 was assigned to a chemical cluster (defined by molecular weight, partitioning properties and indoor degradation half-life) for which

359

inhalation of indoor air is of great importance in determining the total “indoor intake fraction”, whereas PCB-153 belonged to a cluster where non-dietary ingestion dominates. In addition, because

360

the two congeners have very low solubility and permeability, other routes make rather minor

361

contributions to the total exposure.



362 363 364 365

Figure 3. The lifetime total uptake rate of PCB-28 (Panels a and b) and PCB-153 (Panels c and d) under constant emissions solely to indoor air (Panels a and c) and urban air (Panels b

366

and d). Note that the absolute values of the lifetime total uptake rates are arbitrary in this illustrative calculation; for each congener, the total uptake rates under two scenarios are

367

normalized to the same magnitude thus no scale is provided.

368

When realistic time-variant emissions are taken into consideration, the age-dependent route-specific

369

contributions to aggregate human exposure differ between birth cohorts. Figure 4 shows the longitudinal, annually averaged exposure to PCB-28 and PCB-153 of female cohorts born in 1956

370 371 372 373

(when indoor emissions started to increase), 1986 (when declining indoor emissions were exceeded by outdoor emissions), and in 2016 (when indoor and outdoor contamination is minimal) (Figure 4).

374

The total uptake rates in recent years are lower than historical ones (moving from left to right in Figure 4), in accordance with the declining PCB contamination in different scales of the environment

375

since the mid-1970s.28


Page 14 of 24

Page 15 of 24


376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391

Figure 4. Age dependence of the total uptake rates of PCB-28 (Panels a, b, and c) and PCB-153 (Panels d, e, and f) by different exposure routes for females born in 1956, 1986 and 2016. In Panels b, c, e and f, the uptake rates at the age

Revisiting the Contributions of Far- and Near-Field Routes to

Recommend Documents