Improving alternatives

assessment of plastic additives

Exploring in silico tools to identify less

hazardous flame retardants

Ziye Zheng

Department of Chemistry

Doctoral Thesis

Umeå 2021

This work is protected by the Swedish Copyright Legislation (Act 1960:729)

Dissertation for PhD

ISBN: 978-91-7855-476-8 (print)

ISBN: 978-91-7855-475-1 (pdf)

Cover picture: Ziye Zheng

Electronic version available at: http://umu.diva-portal.org/

Printed by: KBC Service Center, Umeå University

Umeå, Sweden 2021

I believe there is no best, only better

-Akio Toyoda

i

Table of Contents

Table of Contents .............................................................................. i

Abstract .......................................................................................... iii

Enkel sammanfattning på svenska ................................................... v

Abbreviations ................................................................................ vii

List of publications ......................................................................... xi

1 Introduction ............................................................................ 1

1.1 Chemical substitution and alternatives assessment ............................................ 1

1.2 Case chemicals – flame retardants .......................................................................3

1.3 Hazard assessment and in silico tools ..................................................................3

1.4 Exposure and emission of plastic additives ......................................................... 5

1.5 Hazard data and making decisions ...................................................................... 5

1.6 The role of chemical transformation products ................................................... 6

1.7 Study aims ............................................................................................................. 7

2 Materials and Methods ............................................................ 8

2.1 Case chemicals ..................................................................................................... 8

2.2 Data collection ...................................................................................................... 8

2.2.1 Experimental hazard data ......................................................................... 8

2.2.2 In silico hazard data from open source platforms ................................... 9

2.2.3 Chemical transformation data .................................................................. 9

2.3 Establishment of QSAR models .......................................................................... 9

2.4 Establishment of hazard ranking tools .............................................................. 10

2.4.1 Work flow of the studies ............................................................................ 10

2.4.2 Decision approaches with MCDA methods .............................................. 11

2.5 Study of OPFRs emission from polymer films ................................................... 12

3 Results and Discussion .......................................................... 14

3.1 Hazard ranking in alternatives assessment ....................................................... 14

ii

3.1.1 Hazard properties for alternatives assessment ....................................... 14

3.1.2 Availability of experimental data ............................................................. 15

3.2 QSAR models for alternatives assessment ......................................................... 16

3.2.1 Availability of open source QSAR models for alternatives assessment . 16

3.2.2 Development of QSAR models ................................................................... 16

3.2.3 Reliability of QSAR data ........................................................................... 18

3.3 Developed approach for transformation products ............................................ 19

3.3.1 Transformation products predicted by in silico tools ............................. 19

3.3.2 Identify and validate transformation products with high occurrence

potential .................................................................................................................... 19

3.4 Developed approach for MCDAs ....................................................................... 20

3.5 Overall hazard ranking results of alternatives for decaBDE ............................ 22

3.6 Measurement of OPFRs emission and modelling approaches ......................... 24

3.6.1 QCM method development and OPFRs emission measurement ............ 24

3.6.2 Modelling approaches to describe the emission pattern and understand

the emission mechanism ..........................................................................................25

4 Conclusions ........................................................................... 28

5 Future perspectives ............................................................... 30

5.1 Improving in silico tools for hazard assessment............................................... 30

5.2 Measurement of emission process and model description ............................... 31

5.3 Improving prediction for chemical transformation .......................................... 31

5.4 Life cycle aspects in alternatives assessment .................................................... 32

6 Acknowledgements ............................................................... 33

7 References ............................................................................ 35

iii

Abstract

Alternatives assessment is applied for replacing hazardous chemicals with viable,

safer substitutes. High quality experimental hazard data, however, are usually

unavailable for this purpose, and obtaining in silico data is the only approach to

fill in data gaps. In silico tools also have the advantage of providing a large

amount of data with much lower cost and time requirements.

The aim of this PhD project was to explore the use of in silico tools for alternatives

assessment, and develop practical tools for alternatives assessment of organic

plastic additives. For this purpose, flame retardants were used as case chemicals.

The major results were:

1. Quantitative structure-activity relationship (QSAR) models for endocrine

disruption were developed and explored (Paper I and II). These developed

models were able to identify chemical properties that impact the binding

affinities of brominated organic chemicals with estrogen-related receptor γ

(Paper I), and to predict the androgen receptor activity of several organic

chemicals, including flame retardants (Paper II);

2. A hazard ranking tool was developed for alternatives assessment based on

the hazard properties of persistence (P), bioaccumulation (B), mobility in the

aquatic environment (M) and toxicity (T). The flame retardant

decabromodiphenyl ether (decaBDE) and 16 of its alternatives were taken as

case chemicals to develop the tool. From a comparison of experimental and

in silico data for these case chemicals, hazard data predicted by in silico tools

were identified as the more suitable data source for the hazard ranking tool

as the experimental data were confounded by large data gaps (Paper III);

3. The inclusion of chemical transformation products for the hazard ranking

tool were studied with the case of decaBDE and its alternatives. Several in

silico tools were used to predict transformation products, and a strategy for

prioritizing chemical transformations with high occurrence potential in the

environment was developed (Paper IV);

4. Multicriteria decision analysis (MCDA) tools were used to evaluate diverse

P, B, M and T endpoints of parent compounds (Paper III) and their

transformation products (Paper IV) simultaneously based on in silico data.

Three different MCDA methods were explored, and one of them was

developed to include the consideration of uncertainties of in silico data;

iv

5. In the studied case of decaBDE alternatives, the three different MCDA

methods generally agreed on the most and least hazardous alternatives. With

the consideration of hazard for the studied flame retardants and their in

silico predicted transformation products, two alternatives, melamine and

bis(2-ethylhexyl) tetrabromophthalate, were identified as the least

hazardous of considered alternatives for decaBDE (Paper III and IV);

6. It is critical for the exposure aspect of alternatives assessment to identify the

key properties that influence the emission process. For this, a fast measuring

method for the emission of polymer additives was developed based on a

Quartz Crystal Microbalance (QCM). Empirical linear models were applied

to describe the emission patterns (Weibull model) to better understand the

chemical mechanism behind the emissions of organophosphate flame

retardants from various polymers. The results showed that the octanol-water

partitioning coefficient and molecular size are key parameters for the

emission process, but also showed that the emission process is complex and

is likely driven by a combination of both polymer and additive properties, as

well as their interactions (Paper V).

This research shows how alternatives assessment can make more effective use of

in silico tools. and it also highlights current challenges in the use of these in silico

tools that require further development. The next steps to make a holistic

alternatives assessment would include an exposure assessment procedure based

on the work in Paper V, and combining this with the hazard ranking tools

(developed in Paper III and IV), including information on technical feasibilities,

economic feasibilities, and also life cycle impacts. The MCDA methods for hazard

ranking in Paper III and IV can be further adapted for the decision component of

such a more complete alternatives assessment for specific uses of chemicals.

v

Enkel sammanfattning på svenska

För att ersätta farliga kemikalier mot mer hållbara och säkrare alternativ krävs

en robust substitutionsmetodik. Experimentella farodata av hög kvalitet är dock

vanligtvis inte tillgängliga för detta ändamål och därför erbjuder beräknade (in

silico) data det enda sättet att fylla i dataluckorna. Beräkningsmodeller har också

fördelen att de kan leverera en stor mängd data snabbt och till mycket lägre

kostnader än experiment.

Syftet med detta doktorandprojekt var att utforska användningen av

beräkningsmodeller i substitutionsprocessen och utveckla praktiska verktyg för

detta med fokus på substitution av organiska plasttillsatser. I projektet användes

flamskyddsmedel som fallkemikalier. De viktigaste resultaten var:

1. Kvantitativa modeller för struktur-aktivitetssamband (QSAR) för

hormonstörning utvecklades och undersöktes (papper I och II). Dessa

modeller kunde identifiera kemiska egenskaper som påverkar

bindningsaffiniteten hos bromerade organiska kemikalier till

östrogenrelaterad receptor γ (Papper I) och förutsäga

androgenreceptoraktiviteten hos ett stort antal organiska kemikalier,

inklusive flamskyddsmedel (Papper II);

2. Ett riskrankningsverktyg utvecklades för substitutionsprocessen baserat på

egenskaper för persistens (P), bioackumulering (B), mobilitet i vattenmiljö

(M) och toxicitet (T). Det bromerade flamskyddsmedlet

dekabromodifenyleter (decaBDE) och 16 av dess alternativ valdes som

fallkemikalier för att utveckla verktyget. En jämförelse genomfördes av

experimentella och beräknad data för fallkemikalierna avseende

tillgänglighet på farodata och det var tydligt att modelldata är nödvändig vid

riskranking eftersom stora luckor i data upptäcktes (papper III).

3. Betydelsen av transformationsprodukter i riskrankning studerades med

decaBDE och några valda substitutionskemikalier. Flera modeller användes

för att förutsäga transformationsprodukter, och en strategi för att prioritera

transformationsprodukter med hög förekomstpotential i miljön utvecklades

(Papper IV);

4. Ett antal verktyg för multikriteria-analys (MCDA) användes för att utvärdera

P-, B-, M- och T-egenskaper för modersubstanserna (Papper III) och dess

transformationsprodukter (Papper IV) baserat på modelldata. Tre olika

MCDA-metoder undersöktes, och en av dem utvecklades för att inkludera

osäkerheter i beräknad data;

5. I det studerade fallet med substitutionskemikalier till decaBDE uppnådde de

tre olika MCDA-metoderna liknande resultat angående de mest och minst

farliga alternativen. Avseende risk för studerade flamskyddsmedel och dess

vi

predikterade transformationsprodukter identifierades två

substitutionsprodukter, melamin och bis (2-etylhexyl) tetrabromoftalat, som

de minst farliga alternativen till decaBDE (papper III och IV);

6. Risk för exponering är en viktig del i substitutionsprocessen och att

identifiera de viktigaste kemiska egenskaperna som påverkar

emissionsprocessen. För detta utvecklades en snabb mätmetod för

emissionsbestämning av polymertillsatser baserat på instrumenteringen

Quartz Crystal Microbalance (QCM). Empiriska linjära modeller användes

för att beskriva emissionssmönster (Weibull-modellen) för att bättre förstå

på molekylärnivå emissioner av organofosfater. Resultaten visade att

fördelningskoefficienten oktanol-vatten och molekylstorlek är viktiga

parametrar för processen, men visade också att emissionsprocessen är

komplex och sannolikt drivs av en kombination av egenskaper för både

polymeren och tillsatsämnet liksom deras interaktioner (Papper V).

Sammantaget visar studierna hur substitutionsprocessen kan göras effektivare

med hjälp av beräkningsverktyg men också utmaningar i användningen av dessa

verktyg som kräver ytterligare utveckling. Nästa steg för att göra en

helhetsbedömning vid substitution bör inkludera ett exponeringsbedömnings-

förfarande baserat på arbetet i papper V och att kombinera detta med

riskrankningsverktygen (utvecklat i papper III och IV), inklusive information om

tekniska och ekonomiska aspekter, och även effekter under hela kemikaliernas

livscykel. Utvecklade MCDA-baserade metoder för riskrankning i papper III och

IV kan ytterligare anpassas för att stödja beslut i sådana mer fullständiga

substitutionsprocesser.

vii

Abbreviations

4'-PeBPOBDE208 tetradecabromodiphenoxybenzene

ASNN associative neural networks

B bioaccumulation

BAF bioaccumulation factor

BCF bioconcentration factor

BEH-TEBP bis(2-ethylhexyl) tetrabromophthalate

BFRs brominated flame retardants

BPA-BDPP bisphenol A diphenyl phosphate

BPBPE decabromodiphenoxyethane

BTBPE bis(tribromophenoxy) ethane

CERAPP Collaborative Estrogen Receptor Activity Prediction

Project

CoMPARA Collaborative Modelling Project for Androgen Receptor

Activity

CTS Chemical Transformation Simulator

CTSA Cleaner Technologies Substitutes Assessment

DBDPE decabromodiphenyl ethane

decaBDE decabromodiphenyl ether

DGE European Commission’s Directorate General for

Employment, Social Affairs and Inclusion

DHSP distance between the Hansen solubility parameters

viii

DP bis(hexachlorocyclopentadieno) Cyclooctane

EAWAG-BBD/PPS EAWAG-biocatalysis/biodegradation database pathway

prediction system

EBTEBPI ethylene Bis-Tetrabromophthalimide

EH-TBB 2-ethylhexyl tetrabromobenzoate

ELECTRE III Elimination Et Choix Traduisant la REalité

ERRγ estrogen-related receptor γ

EU European Union

GHS Globally Harmonized System of Classification and

Labeling of Chemicals

HSP Hansen solubility parameters

Kd dissociation constants

KOW octanol-water partition coefficient

M mobility

MA melamine

MAUT multi-attribute utility theory

MCDA multicriteria decision analysis

MOE Molecular Operating Environment

NAS National Academy of Science

OCHEM Online Chemical Database

OH-PBDEs hydroxylated polybromodiphenyl ethers

ix

OPFRs organophosphorus flame retardants

P persistency

p preference thresholds for ELECTRE III

P2OSH Pollution Prevention-Occupational Safety and Health

PBDEs polybromodiphenyl ethers

PBDPP resorcinol bis(diphenyl phosphate)

PLS partial least-squares

POPs persistent organic pollutants

q indifference thresholds for ELECTRE III

QCM quartz crystal microbalance

QSAR quantitative structure-activity relationship

REACH Registration, Evaluation, Authorisation and Restriction of

Chemicals

SD standard deviation

T toxicity

TBBPA-BDBPE tetrabromobisphenol A Bis (2,3-dibromopropyl) Ether

TBEP tris(2-bromoethyl)phosphate

TPHP triphenyl Phosphate

TSCA Toxic Substance Control Act

TTBNPP tris(tribromoneopentyl) Phosphate

TTBP-TAZ tris(tribromophenoxy) Triazine

x

USEPA Environmental Protection Agency of the United States

v veto thresholds for ELECTRE III

WEIA Work Environment Impact Assessment

α scale parameter of one compartment Weibull distribution

model

β shape parameter of one compartment Weibull

distribution model

xi

List of publications

The thesis is based on the following papers, which are referred to in the text by the

corresponding Roman numerals.

I. Lin-Ying Cao, Ziye Zheng, Xiao-Min Ren, Patrik L. Andersson, and Liang-Hong

Guo. "Structure-Dependent Activity of Polybrominated Diphenyl Ethers and Their

Hydroxylated Metabolites on Estrogen Related Receptor γ: in Vitro and in Silico

Study." Environmental Science & Technology 52.15 (2018): 8894-8902.

II. Kamel Mansouri, Nicole Kleinstreuer, Ahmed M. Abdelaziz, Domenico Alberga,

Vinicius M. Alves, Patrik L. Andersson, Carolina H. Andrade, Fang Bai, Ilya Balabin,

Davide Ballabio, Emilio Benfenati, Barun Bhhatarai, Scott Boyer, Jingwen Chen,

Viviana Consonni, Sherif Farag, Denis Fourches, Alfonso T. García-Sosa, Paola

Gramatica, Francesca Grisoni, Chris M. Grulke, Huixiao Hong, Dragos Horvath, Xin

Hu, Ruili Huang, Nina Jeliazkova, Jiazhong Li, Xuehua Li, Huanxiang Liu, Serena

Manganelli, Giuseppe F. Mangiatordi, Uko Maran, Gilles Marcou, Todd Martin,

Eugene Muratov, Dac-Trung Nguyen, Orazio Nicolotti, Nikolai G. Nikolov, Ulf

Norinder, Ester Papa, Michel Petitjean, Geven Piir, Pavel Pogodin, Vladimir Poroikov,

Xianliang Qiao, Ann M. Richard, Alessandra Roncaglioni, Patricia Ruiz, Chetan

Rupakheti, Sugunadevi Sakkiah, Alessandro Sangion, Karl-Werner Schramm,

Chandrabose Selvaraj, Imran Shah, Sulev Sild, Lixia Sun, Olivier Taboureau, Yun

Tang, Igor V. Tetko, Roberto Todeschini, Weida Tong, Daniela Trisciuzzi, Alexander

Tropsha, George Van Den Driessche, Alexandre Varnek, Zhongyu Wang, Eva B.

Wedebye, Antony J. Williams, Hongbin Xie, Alexey V. Zakharov, Ziye Zheng, and

Richard S. Judson. "CoMPARA: Collaborative Modeling Project for Androgen

Receptor Activity." Environmental Health Perspectives 128.2 (2020): 027002.

III. Ziye Zheng, Gregory Peters, Hans Peter Arp, and Patrik L. Andersson. "Combining

in Silico Tools with Multicriteria Analysis for Alternatives Assessment of Hazardous

Chemicals: A Case Study of Decabromodiphenyl Ether Alternatives." Environmental

Science & Technology 53.11 (2019): 6341-6351.

IV. Ziye Zheng, Hans Peter Arp, Gregory Peters, and Patrik L. Andersson. "Combining

in silico Tools with Multicriteria Analysis for Alternatives Assessment of Hazardous

Chemicals: Accounting for the Transformation Products of decaBDE and Its

Alternatives." Environmental Science & Technology 55.2 (2021): 1088-1098.

V. Linhong Xiao, Ziye Zheng, Knut Irgum, Patrik L. Andersson. "Studies of Emission

Processes of Polymer Additives into Water Using Quartz Crystal Microbalance—A

Case Study on Organophosphate Esters." Environmental Science & Technology 54.8

(2020): 4876-4885.

Reprints were made with permission from the publishers (American Chemical Society

Publications and Environmental Health Perspectives)

xii

Author’s contributions

Paper I

I was responsible for the QSAR model development and interpretation of QSAR

results, and involved in writing the corresponding section of the paper.

Paper II

I was responsible for developing one set of the QSAR models presented in the

paper, determining the applicability domains for the models and using the

models for prediction (the developed model set was presented in the paper as

“UMEA”).

Paper III

I was involved in planning of the study. I was responsible for constructing the

work flow; data collection, analysis and interpretation; as well as writing the

paper.

Paper IV

I was involved in planning of the study. I was responsible for constructing the

work flow; data collection, analysis and interpretation; as well as writing the

paper.

Paper V

I was involved in planning of the study. I participated in the establishment of

experimental set-up and data collection. I was responsible for data analysis,

empirical modelling, studying potential mechanism of emission based on model

results, as well as writing the corresponding section of the paper.

xiii

I also contributed to the following publications during my doctoral education:

1. Ge Yin, Yihui Zhou, Anna Strid, Ziye Zheng, Anders Bignert, Taowu Ma,

Ioannis Athanassiadis, Yanling Qiu. "Spatial distribution and

bioaccumulation of polychlorinated biphenyls (PCBs) and polybrominated

diphenyl ethers (PBDEs) in snails (Bellamya aeruginosa) and sediments

from Taihu Lake area, China." Environmental Science and Pollution

Research 24.8 (2017): 7740-7751.

2. Xinyu Du, Bo Yuan, Yihui Zhou, Ziye Zheng, Yan Wu, Yanling Qiu, Jianfu

Zhao, Ge Yin. "Tissue-specific accumulation, sexual difference, and maternal

transfer of chlorinated paraffins in black-spotted frogs." Environmental

Science & Technology 53.9 (2019): 4739-4746.

3. Xinyu Du, Bo Yuan, Yihui Zhou, Cynthia A de Wit, Ziye Zheng, Ge Yin.

"Chlorinated paraffins in two snake species from the Yangtze River Delta:

tissue distribution and biomagnification." Environmental Science &

Technology 54.5 (2020): 2753-2762.

4. Xinyu Du, Yihui Zhou, Jun Li, Yan Wu, Ziye Zheng, Ge Yin, Yanling Qiu,

Jianfu Zhao, Guoli Yuan. "Evaluating oral and inhalation bioaccessibility of

indoor dust-borne short-and median-chain chlorinated paraffins using in

vitro Tenax-assisted physiologically based method." Journal of Hazardous

Materials 402 (2020): 123449.

1

1 Introduction

1.1 Chemical substitution and alternatives assessment

With increasing concerns about the health and environmental safety of synthetic

chemicals used in products, a number of chemicals have been restricted or

banned by increasingly stringent chemical regulations.1-4 When hazardous

chemicals are regulated, or are facing increasing scrutiny, industries have to find

alternative substances with similar technical functions.

Alternatives assessment frameworks assess chemical substances with the same or

similar uses to minimize the risk of regrettable substitution, which refers to the

situation where a hazardous chemical is substituted with a chemical of similar or

higher risk.5, 6 Several examples of regrettable substitution have been reported in

the literature. One example is the substitution of bisphenol A with other

bisphenols (e.g. bisphenol S, bisphenol AF, fluorene-9-bisphenol etc.), of which

many have also been identified as endocrine disruptors.7, 8 Compared to a typical

chemical risk assessment process which evaluates the hazard and exposure of a

chemical, a thorough alternatives assessment framework should additionally

consider multiple aspects of target alternatives including life cycle impacts from

the production phase to the end-of-life stage, material or product performance

and economic costs (Figure 1).6, 9, 10 Since the 1990s, several alternatives

assessment frameworks have been developed and implemented.10 Most of the

frameworks are focused on hazards to consumers, although some frameworks

also consider occupational hazards (Table 1).

2

Figure 1. Six major components of a thorough alternatives assessment framework

Table 1. Well-known alternatives assessment frameworks focusing on consumer hazards

or occupational hazards

Consumer hazards Occupational hazards

Cleaner Technologies Substitutes Assessment(CTSA) by Environmental Protection Agency of the United States (USEPA)11

European Commission’s Directorate General for Employment, Social Affairs and Inclusion (DGE)12

UNEP Persistent Organic Pollutants (POPs) Review Committee’s General Guidance on Alternatives13

Pollution Prevention-Occupational Safety and Health (P2OSH)14

BizNGO chemicals alternatives assessment protocol including GreenScreen®15

Work Environment Impact Assessment (WEIA)16

National Academy of Science (NAS) Guideline17

3

1.2 Case chemicals – flame retardants

Plastics are widely used both in industry and in our daily life, with a global

production volume reaching 348 million metric tonnes in 2017.18 Most polymers

are first compounded with additives to improve the performance and

functionality of the final products, such as plasticizers, flame retardants and light

stabilizers. These additives are usually not covalently bound to the polymer,19 and

emission of additive compounds from plastic products occurs in all phases of the

products’ life cycle,20-24 posing risks to the environment and human health.25

Among the various kinds of polymer additives, flame retardants are a commonly

used group for prohibiting or slowing down the progress of fire. Brominated

flame retardants (BFRs) used to be the most heavily applied group of flame

retardants, while in recent years a number of them have been banned or restricted

for certain applications due to their hazardous effects.26, 27 One review claimed

that for certain BFRs, the adverse environmental and human health impacts

might even be greater than their fire safety benefits.28 A well-known example is

decabromodiphenyl ether (decaBDE) that has been used in different products

and materials including electronics, vehicles, furniture and textiles since 1970s,

but was later found to be persistent in the environment29, bioaccumulative30, and

to induce adverse effects in both humans and other species.1, 31-34 DecaBDE has

been banned in the EU since 2008 for use in electronics and electrical

applications35-37 and listed by the Stockholm Convention on Persistent Organic

Pollutants (Annex A)38. In the US, both producers and importers announced in

2010 that they would voluntarily phase out decaBDE by the end of 2013.39

Since some hazardous BFRs were restricted, other novel BFRs have been

commonly used as replacements. For example, decabromodiphenyl ethane

(DBDPE) has been widely used to replace decaBDE. However, this substance was

later found to also be persistent and bioaccumulative40-43, and potentially toxic27,

and could be considered as an example of a regrettable substitution. The focus

was then turned to halogen-free flame retardants including the

organophosphorus flame retardants (OPFRs). Although some studies have

shown that these OPFRs are generally safer than BFRs27, many studies also point

out that some OPFRs are persistent, have the potential for long-range transport44,

and can be neurotoxic.45, 46 As a result, it is essential to adjust the strategy of

selecting alternatives and conduct thorough alternatives assessments for the

selection of best alternatives.

1.3 Hazard assessment and in silico tools

Hazard is the most well-studied aspect of alternatives assessment for avoiding

regrettable substitutions. Almost all existing alternatives assessment frameworks

consider both human health and environmental effects, and divide hazard

4

endpoints into three major categories: persistency (P), bioaccumulation potential

(B), and toxicity (T),9 as these are hazard properties applied in chemical

regulations like for example Registration, Evaluation, Authorisation and

Restriction of Chemicals (REACH) 1 and the Plant Protection Product legislation2

for the European Union (EU), as well as Toxic Substance Control Act (TSCA)3 and

the Frank R. Lautenberg Chemical Safety for the 21st Century Act4 for the United

States. More recently in the field of chemical hazard assessment, arguments have

been presented that mobility (M) in water is an additional important hazard

property to be considered.47-51 Chemical risk assessment has traditionally

focussed more on nonpolar hydrophobic compounds in particular due to their

bioaccumulation potential, while polar organic chemicals have a lower potential

to sorb to organic matter in soil and sediment and are thus more mobile in aquatic

environments. Compounds that are persistent and mobile can reach aquatic biota

in surface waters, and even reach drinking water resources, potentially leading to

chronic human exposure. Therefore, a PMT assessment has been suggested in

addition to the widely-used PBT assessment and will be included in REACH as

part of the EU's Chemical Strategy for Sustainability.52 However, this concept has

yet not been much discussed for alternatives assessments.

When estimating chemical hazards, many alternatives assessment frameworks

recommend the use of experimental hazard data.10 However, such data is often

lacking as some alternatives are newly designed or just introduced to the market

with low production volumes, and hazard evaluation for alternatives assessments

are often incomplete with large data gaps.6 In such cases, in silico tools such as

“read-across” approaches or quantitative structure-activity relationship (QSAR)

models could be useful tools to fill in data gaps, and in many occasions these in

silico data might be the only available data. Chemicals with similar structures can

be “grouped” together assuming that they share physico-chemical or toxicological

properties. Consequently “read-across” approaches can be applied to predict

properties within the group or for structurally similar chemicals.53 The “read-

across” may rely on relatively sparse data which do not support complicated

predictive modelling, while a QSAR, which can be considered a subset of “read-

across” approaches, is a regression model where chemical descriptors

representing structural information or key properties of chemicals are correlated

with a biological or toxicological activity, or other chemical properties. QSAR

models can be either quantitative or qualitative depending on the type of training

data as well as selected modelling algorithm used for the model development. In

silico data are generally considered with higher uncertainties compared to

experimental data, but if used properly following certain protocols1, 54, 55, such

data can be useful especially at the screening level as the in silico tools have the

advantage of providing large amounts of data with much higher economic and

time efficiency.

5

1.4 Exposure and emission of plastic additives

Failure to consider diverse exposure pathways is one underlying cause of

regrettable substitution. An example of this was the choice of n-hexane to replace

chlorinated solvents for automotive cleaning industry,56 as n-hexane is

considered overall a safer solvent, which is generally true if the major exposure

route is inhalation. However, n-hexane has been shown to be neurotoxic after

direct dermal exposure,57 which is a typical exposure scenario for automotive

mechanics, making this a regrettable substitution. To avoid this type of

regrettable substitution, the differences between usage rates required by different

alternatives to reach similar technical performance need to be considered, and

also the differences in physico-chemical properties between different alternatives

as these can lead to different exposure levels and routes.

To better understand the human and environmental exposure of polymer

additives, it is of great importance to increase the understanding of the emission

processes of polymer additives. It has been reported that the rate and extent of

emission of additives can be affected both by the properties of polymers and

additives, as well as environmental factors like temperature and environmental

pH.22, 58 In most studies of emission processes, the experiments usually take a

long period from few weeks to hundreds of days to collect sufficient data. 22, 59-61

Long experimental periods may increase the measurement uncertainties and

prevent the collection of large amounts of data. Therefore, an efficient and

accurate method to measure the emissions from polymers is desired for

understanding the emission mechanisms and to find alternatives with lower

emission and exposure potential.

1.5 Hazard data and making decisions

There are different ways to present the results of an alternatives assessment of

chemical hazards or other outputs. In some studies62, 63, the results are presented

for each alternative with classification labels for each of the assessed criteria,

without making a decision regarding “the best choice”. However, stakeholders,

decision makers or users from industry of the alternatives assessment

frameworks often demand a conclusive decision strategy. In such cases,

Multicriteria Decision Analysis (MCDA) methods, which are drawn from

operations research for decision making based on conflicting criteria64, can be

useful tools. MCDA was first developed for business sectors, but it has also been

proven useful for several environmental management applications65-71. The two

major kinds of MCDA methods are full aggregation methods and outranking

methods (Table 2).72-74 In general, there is no “best” MCDA method, but the most

suitable method should be chosen based on data availability, the decision purpose

and the role stakeholders will play in the analysis.64

6

Table 2. Differences between the two main kinds of aggregating MCDA methods applied

in this thesis

Full aggregation methods Outranking methods

Basic method

All criteria converted into comparable scales (utility functions) and aggregated with trade-off weights

Pair-wise comparison of alternatives with respect to each criterion based on set significant levels (thresholds)

Algorithm Relatively simple Relatively complex

Typical example Multi-attribute utility theory (MAUT)

Elimination Et Choix Traduisant la REalité (ELECTRE III)

1.6 The role of chemical transformation products

The importance of including transformation products has been pointed out in

alternatives assessment frameworks62, 75, 76, as the chemical hazards can be

underestimated when only parent compounds are considered. Examples for

chemicals degrading into more hazardous compounds are quite common.

Perchloroethylene, a dry cleaning solvent widely used until 1990s, can degrade

into the more toxic compound vinyl chloride77; bronopol (2-bromo-2-nitro-1,3-

propanediol), a preservative compound, can aquatically transform into two more

persistent and toxic products: 2-bromo-2-nitroethanol and

bromonitromethane78; carbamazepine, an anti-epileptic pharmaceutical agent,

turned out to photodegrade into the more toxic products: acridine and acridone79;

and a rubber antioxidant widely used for car tires, N-(1,3-dimethylbutyl)-N'-

phenyl-p-phenylenediamine), can be transformed into 2-anilino-5-[(4-

methylpentan-2-yl)amino]cyclohexa-2,5-diene-1,4-dione with high aquatic

toxicity80. As for the case of flame retardants, decaBDE is considered to have low

bioaccumulation potential and low toxicity but classified under REACH as a PBT

substance81, 82 for its potential to transform into dozens of other

polybromodiphenyl ethers (PBDEs) that are more bioaccumulative and toxic

than decaBDE. Therefore, it is important to include transformation products in

any alternatives assessment process.

However, this remains a difficult task as many of the chemicals considered in

alternative assessments are novel compounds that have been subjected to few

experimental investigations themselves, let along their transformation products.

In such cases, in silico tools can be useful to predict transformation products as

7

well as their hazard profiles. One example is Martin et al.62 from 2017 which

concerns decaBDE as well as three organophosphate-based flame retardants that

are proposed alternatives. In that study, transformation products were predicted

by the Chemical Transformation Simulator (CTS).83, 84 GreenScreen75 suggests

the use of models provided by OECD QSAR Toolbox85 for predicting chemical

transformation when there is no experimental data available. There is a range of

other in silico tools available for the prediction of transformations products,

including commercial tools such as Metasite86, Stardrop87, Zeneth88 and Meteor

Nexus89, as well as open source tools such as BioTransformer90 and the EAWAG-

biocatalysis/biodegradation database pathway prediction system (EAWAG-

BBD/PPS)91.

1.7 Study aims

The overall goal of this PhD project was to explore different aspects of hazard

assessment within alternatives assessment frameworks and develop practical

tools for making decisions across diverse hazard aspects of organic plastic

additives. A major challenge in alternatives assessments is the lack of

experimental data. An approach to address this is using in silico tools, which is

rare in current alternatives assessment. Here QSAR models were developed

based on experimental data with the aim of estimating hazard data where such

was lacking. The first focus was on developing QSAR models for the hazard

category of endocrine disruption (Paper I and II). The project further focused

on exploring the use of in silico methods in alternatives assessments, including

the development of a hazard ranking tool for alternatives using a combination of

in silico derived hazard data and MCDA methods (Paper III). This was then

expanded to include hazards of the transformation products predicted by in silico

tools (Paper IV). Finally, emission characteristics of plastic additives from

polymers were explored by developing a new experimental approach, and the

important chemical properties that influenced the emissions were discussed

using empirical models (Paper V).

8

2 Materials and Methods

2.1 Case chemicals

For the development of a hazard ranking tool in Paper III, a set of target

chemicals was identified from the literature6, 36, 37, 63 to make sure they have been

or could be used as alternatives to decaBDE. Inorganic compounds and polymers

were excluded due to the lack of in silico tools for such compounds. The identified

least hazardous alternatives in Paper III (DBDPE, BEH-TEBP and melamine)

were further studied with decaBDE and two OPFRs (TPHP and TTBNPP) to

develop the hazard ranking framework to include chemical transformations

(Paper IV).

In Paper V, in order to study the emission process and underlying mechanisms,

nine OPFRs with various physicochemical properties (Table S2 in Paper V) were

studied using three polymers with different properties (Table S1 in Paper V).

2.2 Data collection

2.2.1 Experimental hazard data

In Paper I and II, hazard data were collected to establish QSAR models. In

Paper I, the binding affinities of polybrominated diphenyl ethers (PBDEs) and

their hydroxylated metabolites (OH-PBDEs) with estrogen-related receptor γ

(ERRγ) were measured and converted into dissociation constants (Kd). ERRγ is

a nuclear receptor that regulates endocrine and metabolic genes. The measured

Kd values for 12 PBDEs and 18 OH-PBDEs were provided by the co-authors in

Paper I. In Paper II, experimental androgen receptor pathway data were

provided by USEPA for QSAR model establishment. The provided training sets

contain data for 1662 chemicals for the binding model, 1659 chemicals for the

agonist model and 1525 chemicals for the antagonist model.

In order to determine the most suitable data source for the development of hazard

ranking tools, experimental hazard data were collected for Paper III using the

OECD QSAR Toolbox85. The reason for choosing OECD QSAR Toolbox is that it

contains several experimental databases (five databases for physicochemical

properties, 12 databases for environmental fate and transport parameters, six

databases for ecotoxicity and 39 databases for human toxicity data) and supports

batch screening for a large number of chemicals. Among the databases provided

in OECD QSAR Toolbox, the ECHA database of the REACH registration

dossiers92 is one of the most important database as it provides large amount of

9

data for substances registered under REACH provided by registrants who

manufacture or import these substances within EU.93

2.2.2 In silico hazard data from open source platforms

In order to make the hazard ranking tools developed in Paper III and IV

available for use by different users, the in silico based hazard data in Paper III

and IV were derived using open source software packages or platforms,

including EPISUITE94, VEGA95, TEST96 and OECD QSAR Toolbox85. Results

from the USEPA organized Collaborative Estrogen Receptor Activity Prediction

Project (CERAPP)97 and Collaborative Modelling Project for Androgen Receptor

Activity (CoMPARA, paper II) were also included. By these, in silico data were

derived for 20 different hazard properties (Table 1 in Paper III).

2.2.3 Chemical transformation data

To be able to predict various transformation pathways and to compare the

prediction results of different in silico tools, several open source software

including OECD QSAR Toolbox85, CTS83, 84, BioTransformer90 and EAWAG-

BBD/PPS91, as well as one commercial software package, Meteor Nexus89, was

included in Paper IV. In order to validate the model prediction results,

experimental data for the transformation of six target FRs in Paper IV were

derived from literature.98-114

2.3 Establishment of QSAR models

In paper I, a partial least-squares (PLS) model was developed in SIMCA115 to

understand the relationship between Kd values and physico-chemical properties

of PBDEs and OH-PBDEs. The chemical descriptors were calculated using the

software Molecular Operating Environment (MOE)116 to reach 65 descriptors117

used in a previous study and 148 additional descriptors for 3D structures.

In paper II, the Online Chemical Database (OCHEM) platform118, a web-based

platform for establishing QSAR models using machine-learning methods, was

used for the model development. All fifteen machine learning methods available

in OCHEM were applied in combination with the nine available descriptor sets in

search for the most statistically significant models, and the final selected models

were based on associative neural networks (ASNN)119 with Dragon descriptors120.

ASNN is a machine-learning algorithm that combines the neural network method

with the k-nearest neighbour method. Dragon descriptors cover 4885 different

chemical descriptors. The established models were used for the prediction of

androgen receptor activity for 55450 organic chemicals.

10

In Paper III and IV, simple QSAR models were established by PLS to fill in a

few remaining data gaps for the in silico data calculated from open source

platforms.

In Paper V, the distance between the Hansen solubility parameters (DHSP) of

OPFRs and polymers were used to study the solubility of OPFRs in different

polymers. Most of the nine studied OPFRs do not have experimental data for

Hansen solubility parameters (HSP) and, therefore, a PLS model was established

to predict the HSP based on measured HSP values from literature121.

2.4 Establishment of hazard ranking tools

2.4.1 Work flow of the studies

In Paper III, a hazard ranking framework was established following the work

flow in Figure 2a, and in Paper IV, the work flow was further developed for the

inclusion of transformation products (Figure 2b).

11

a)

b)

Figure 2. Work flow of the developed hazard ranking framework: (a) without inclusion of

the transformation product (Paper III); (b) with inclusion of transformation products

(Paper IV)

2.4.2 Decision approaches with MCDA methods

The obtained human health and environmental hazard parameters were used as

assessment criteria within three MCDA strategies: heat mapping, MAUT and

ELECTRE III. For the heat map, the range of each criterion was divided into four

intervals, and each interval was colour-coded (red, orange, yellow or green – from

hazardous to benign) to aid visual interpretation. Thresholds of the assessment

criteria were based on regulation/literature values (Table S3 in Paper III).

12

For the MAUT approach, each criterion was scaled from 0 (worst) to 1 (best)

based on the average result of all models for that criterion. After the scaling,

partial scores were calculated by considering each criterion relevant to an aspect

(P, B, M, or T) with the same weight. Final scores were calculated by treating

either P, B and T aspects as equally important (marked as PBT scores); or by P,

M and T (marked as PMT scores) as equally important; or by P, B, M and T

(marked as PBMT scores) as equally important, in Papers III and IV.

For ELECTRE III, the calculation was done in a manner consistent with other

publications.72, 122, 123 In brief, three important thresholds have to be set for each

criterion: indifference thresholds (q), preference thresholds (p), and veto

thresholds (v). In paper III and IV, q and p were set based on standard deviation

(SD) of different model results to reflect data uncertainties (Table 3 in Paper III).

Pair-wise comparisons were then conducted on each assessment criterion, and

final scores were also presented as PBT, PMT or PBMT scores.

2.5 Study of OPFRs emission from polymer films

Quartz crystal microbalance (QCM) is a technique that can record mass changes

in real-time on the surface of quartz crystal. The mass changes (Δm, g) were

measured by the frequency change (Δf, Hz) based on the Sauerbrey

relationship124:

where f0 (c.a. 106 Hz) is the resonant frequency of an uncoated quartz crystal, A

(0.2826 cm‒2) is the gold-coated electrode-covered area of the quartz crystal, ρq

(2.648 g·cm‒3) is the density of quartz and µq (2.947×1011 g·cm‒1·s‒2) is the shear

modulus of quartz crystal. In paper V, emissions of OPFRs were studied from

polymer films following the work flow in Figure 3.

2

02

q q

ff m

A

13

Figure 3. Work flow for the study of the plastic additive emission process using QCM

measurement

The experimental set-up for the study is shown in Figure 4. Different empirical

models were fitted with the measured emission data, and the best model was

selected with the consideration of not only the fitting performance but also

minimizing the risk of overparameterization. The fitted parameters of the

selected empirical model were then studied to understand the emission

mechanism by correlating with different physico-chemical properties of the

studies OPFRs and polymers.

Figure 4. Experimental set-up for the measurement of plastic additives emission process

using QCM

14

3 Results and Discussion

3.1 Hazard ranking in alternatives assessment

3.1.1 Hazard properties for alternatives assessment

A comparison of hazard properties included in some well-known alternatives

assessment frameworks was completed and compared to those included in the

hazard ranking tools developed in paper III and IV (Table 7).

Table 7. Hazard properties included in well-known alternatives assessment frameworks

compared to those in paper III and IV

Persisten

ce

Bio

accu

mu

latio

n

Mo

bility

in a

qu

atic en

viro

nm

ent

Hu

ma

n

tox

icity

Eco

tox

icity Tra

nsfo

rma

tion

pro

du

ct ha

zard

s U

ncerta

inty

of h

aza

rd co

nclu

sion

s

Mu

tag

enicity

Ca

rcino

gen

icity

Dev

elop

me

nta

l/repro

du

ctive to

xicity

En

do

crine

disru

ptio

n

Sk

in se

nsitiza

tion

Sk

in irrita

tion

Ey

e irrita

tion

Acu

te o

ral to

xicity

Neu

roto

xicity

Resp

irato

ry sen

sitivity

Aq

ua

tic tox

icity

Te

rrestrial eco

tox

icity

USEPA CTSA11

UNEP POPs Review Committee’s General Guidance on Alternatives13

BizNGO chemicals alternatives assessment protocol including

GreenScreen®15

NAS guideline17

European Commission DGE12

Hazard ranking tools in Paper III and IV

15

It is evident that most of the well-known alternatives assessment frameworks

include the hazard properties of P, B and T. The hazard ranking tools developed

in Paper III and IV are the first to include M for alternatives assessment.

Though the NAS guideline17 as well as some other alternatives assessment studies

include water solubility as a hazard property10, this is not consistent with the most

recent criteria of M49, 125. The selection of toxicity criteria varied between different

frameworks but are generally similar, except for the UNEP POPs Review

Committee’s General Guidance on Alternatives which includes relatively few

toxicity properties, as it focuses on hazard properties covered in the Stockholm

Convention. Most of the frameworks use aquatic toxicity to assess ecotoxicity,

while the NAS guideline17 also pointed out the need of including terrestrial

toxicity for e.g. plants and non-mammal animals. Although these alternatives

assessment frameworks usually include abundant hazard properties for the

assessment (Table 7), the issue of data availability has been a continuing point of

discussion, with varying opinions on how to best collect data for these various

hazard properties and how to fill the large data gaps when trustworthy

experimental data are lacking.10 The tools developed in Paper III and IV

included the hazard properties with available data for the target chemicals, which

is further discussed here in sections 3.1.2 and section 3.2.

As discussed in the introduction, chemical transformation can play an important

role in alternatives assessment. As presented in table 7, there are a few

alternatives assessment frameworks that stress the importance of considering

chemical transformations (USEPA CTSA, BizNGO and NAS). However, there is

no clear guideline regarding how this should be done in these frameworks. In

Paper IV, a novel approach was developed to include transformation products

as part of the hazard ranking, which is further discussed in section 3.3. Another

important issue widely discussed in alternatives assessment frameworks is data

reliability and uncertainties. The reliability of collected experimental data and in

silico data for studied flame retardants in Paper III and IV is further discussed

in section 3.1.2 and 3.2.3 respectively; and the approaches developed to include

the consideration of uncertainties in MCDA is discussed in section 3.4 and 3.5,

3.1.2 Availability of experimental data

In Paper III, experimental data were collected for the 17 flame retardants to

better understand the problem of data availability. A total of 2866 experimental

data points were collected from the OECD QSAR Toolbox for the 17 flame

retardants. However, most of these data points were only available for well-

studied chemicals such as decaBDE, TPP and melamine. Five of the flame

retardants studied in this thesis have no relevant experimental data. Also, data

for some hazard properties like neurotoxicity or reproductive toxicity are

extremely scarce. To manually enlarge the collection of experimental data by

16

using other databases than the QSAR Toolbox can to some extent yield more data,

but such efforts are very time consuming and were shown to be insufficient to

close data gaps. Further in Paper IV when transformation products are included,

the data gap problem becomes more severe as experimental hazard data are even

more rare for most of the transformation products compared to the parent

compounds. Besides data gaps, previous reports have also criticized the reliability

problems of these experimental data for both flame retardants126 and other

chemicals127, 128, making their use for hazard comparison questionable.

3.2 QSAR models for alternatives assessment

3.2.1 Availability of open source QSAR models for alternatives

assessment

Regarding the issue of incomplete experimental hazard data for alternatives

assessment, some frameworks suggest the use of QSAR modelling data, including

USEPA Design for Environment framework (DfE)129 and GreenScreen®15. In a

recent study that screened 189 alternatives assessment reports under REACH,

only 24 of these reports were found that used QSAR data.130 A potential reason

for the use of QSAR models to be still relatively limited is that proper regulatory

guidelines for using QSAR tools in alternatives assessment is lacking, particularly

regarding issues of data uncertainty and chemical applicability domains.

In Paper III, QSAR models were found for 20 hazard criteria from open source

platforms, which covered most of the hazard criteria concerned by REACH1 as

well as the most well-known alternatives assessment frameworks (Table 7). The

largest number of QSAR models was found for aquatic toxicity (including fish and

daphnia magna acute toxicity), while persistence criteria (half-lives in different

environmental media and biota) as well as some hazard criteria (skin

sensitization, skin and eye irritation) had only one model for each criterion. No

models were found for neurotoxicity, respiratory sensitivity and terrestrial

ecotoxicity, that are included in some alternatives assessment frameworks (Table

7). The lack of models could be due to lack of sufficient experimental data to

establish high quality QSAR models. Considering the data availability and

consistency, the collected in silico data appeared more suitable than experimental

data for the hazard ranking of studied case chemicals, and was thus used for

hazard ranking in Paper III and IV.

3.2.2 Development of QSAR models

Besides existing QSAR models in open source platforms, it is also possible to

establish new QSAR models for alternatives assessment. In Paper II, three

17

QSAR models were established for androgen receptor activity by the ASNN

method using the OCHEM platform. As shown in Table 8, the performances of

developed models are generally good, and with similar levels to the models

developed by other researchers on the same training sets (Figure 2 of Paper II).

The level of specificity is higher than sensitivity for all three models, which is due

to the highly imbalanced training sets (the ratio of inactives versus actives equals

to 7.4, 38 and 8.6 for binding, agonist and antagonist training sets respectively,

Table 1 of Paper II). The developed binding model was further used in

developing the hazard ranking tools in Paper III and IV. In Paper III, there

were a few data gaps for the data from open source QSAR platforms, and these

were filled by a combination of experimental data and “read-across” and grouping

or simple regression based QSAR models.

Table 8. Statistical results of the three androgen receptor activity models

Model Accuracy Balanced Accuracy Sensitivity Specificity

Binding 84.6% 81.7% 77.7% 85.6%

Agonist 92.2% 85.9% 79.2% 92.6%

Antagonist 84.3% 80.3% 75.3% 85.4%

*Accuracy is the rate of correct predictions, sensitivity is the rate of correct positive

predictions (true positive), specificity is the rate of correct negative predictions (true

negative), and balanced accuracy is the average of sensitivity and specificity.

QSAR models can also be used to predict hazard properties that have not yet been

included in existing alternatives assessment frameworks and these could be

included in future developments. For example, the latest version of the VEGA

platform95 has included QSAR models for several new hazard properties

including hepatotoxicity, micronucleus activity and skin permeation. Another

example is that in paper I, a QSAR model was developed to study the ERRγ

binding potency of PBDEs and OH-PBDEs. A PLS model was established with

three significant components (eigenvalues above 1.5), and the model provided a

good fit between the predicted log Kd and the observed values (Figure 2 in Paper

I). The model was able to identify the most significant chemical properties that

influence the ERRγ binding potency of PBDEs and OH-PBDEs, and can be used

to predict other PBDEs/OH-PBDEs. ERRγ binding potency has not yet been

included in any alternatives assessment framework, but could be included in the

future. The QSAR model developed in Paper I could be further developed for

alternatives assessment with a larger training set that include compounds other

than PBDEs/OH-PBDEs.

18

3.2.3 Reliability of QSAR data

One of the important issues for using QSAR models to predict effects of other

chemicals is the applicability domain. In Paper I, mild outliers searched using

Hotellings T2 test (95%) indicated that all studied chemicals were within the

model domain. However, the domain was not evaluated for chemicals other than

PBDEs/OH-PHBEs as there were no other chemicals in the training set. In Paper

II for the 55450 chemicals in the prediction set, 35065 (63%), 13738 (25%) and

28415 (51%) chemicals were predicted as out of domain for androgen binding,

agonistic, and antagonistic interactions, respectively, based on their distance to

the model provided by the OCHEM platform. For the open source tools used in

Paper III and IV, information regarding the applicability domains are

sometimes unclear or even lacking, which was also pointed out in a previous

study131.

Besides the issue of the chemical applicability domain, the quality of these in

silico data, at large, varies depending on e.g. experimental data used in training

the models, the choice of the used parameters and their representation of the

underlying mechanisms, and complexity of models.126, 131 The studies in Paper

III and IV generally agree with a recent review study130, that although the

uncertainty in QSAR model data is generally higher than the experimental data

used to train the models, for certain case chemicals with few and unreliable

experimental data, the use of prediction results from QSAR models might yield

lower uncertainties. For chemicals falling well within the applicability domain of

a well, calibrated QSAR, the QSAR can even be used to verify an experimental

result. However, a general limitation of QSARs is they work generally best for

compounds that they were calibrated for, which comprises the types of chemicals

that are most frequent in their chemical applicability domain. If QSARs give

reliable predictions for chemicals outside their applicability domain is always

uncertain, until tested. In such cases, a decision would be more reliable if it were

based on several QSAR models compared to using single model results only. In

Paper II, a consensus model was developed based on single models provided by

different model developers. The use of consensus models to reduce uncertainties

have been discussed for other hazard properties based on various QSAR

models.132 In Paper III and IV, average results were taken for any hazard

property with more than one model, and the model uncertainties were further

considered in the decision making step which is discussed in section 3.4.

19

3.3 Developed approach for transformation products

3.3.1 Transformation products predicted by in silico tools

The in silico tools used in Paper IV were able to predict a number of different

transformation mechanisms, among which oxidation, hydroxylation and

debromination, which are the most frequently observed reactions for mammalian

metabolisms and abiotic degradations, while more complex reactions such as the

degradation of cyanuric acid into biuret and allophanate were more observed in

microbial metabolisms. For the six studied flame retardants, the selected tools

gave similar results for some and gave disparate results for the others (Figure 2

and Table S4 in Paper IV). If all predicted transformation products were taken

into consideration for further hazard assessment, the number of considered

compounds would be large. The recent literature indicates that some in silico

tools tend to over-predict the possibilities of degradation reactions133. Thus, it is

important to prioritize transformation products with high occurrence potential

(Step 3 in Figure 2b).

3.3.2 Identify and validate transformation products with high

occurrence potential

There are some previous studies that discuss the procedure of selecting

transformation products with the most exposure and hazard potential.134-138 With

the inspiration of these studies, in Paper IV transformation products were

considered having a high occurrence potential if they meet both conditions: (i)

they are predicted to have a high occurrence frequency across various models;

and (ii) they are stable compounds in the environment with a high predicted

persistency. Five to nine transformation products with high occurrence potential

were identified for each of the case chemicals. It is worth pointing out that none

of the transformation products with high occurrence frequency were ruled out by

persistence, in other words persistent compounds are likely to degrade into other

persistent compounds.

In order to validate if the prioritization of transformation products is reasonable,

the prioritized transformation products were compared with those identified in

the literature98-110. Most of the transformation products reported in the literature

for the case chemicals were successfully predicted and identified as with high

occurrence potential with a few exceptions (Paper IV), partly because of the lack

of predictive tools for some pathways like photodegradation.

20

3.4 Developed approach for MCDAs

The strengths and weaknesses of three different MCDA methods were discussed

in Paper III. In brief, heat mapping was conducted by marking each hazard

criterion with four color codes based on regulation/literature based thresholds

(Figure 4, from Figure 2 in Paper III). This is similar to the hazard ranking

methods in many alternatives assessment frameworks, which rank or categorize

the hazard of different alternatives by labelling each hazard criterion based on –

for instance – the Globally Harmonized System of Classification and Labelling of

Chemicals (GHS) or similar approaches.10 Some alternatives assessment

frameworks just present the labelled results like a heat map62, 63, while others (e.g.

GreenScreen®15) make further ranking based on the labelled results.

Figure 4. Heat map of the 17 flame retardants (Figure 2 in Paper III)

21

The heat mapping is the most apparent method to show the advantages and

disadvantages for choosing each alternative. However, simply counting the colour

codes on a heat map is not the most sophisticated way for picking out the best

alternative. Another issue discovered in the heat mapping of Paper III is that

setting thresholds based on regulation or literature values can sometimes make

the differences between chemicals less clear. For example, all 17 chemicals were

marked red for their half-lives in sediment without showing the variation of

almost a factor of five, which is a common limitation for alternatives assessment

frameworks with fixed thresholds for all chemicals, such as the USEPA DfE76 and

GreenScreen®15.

Considering this, a different strategy was developed with MAUT, which considers

both regulation/literature values and relative hazard level differences between

target chemicals for each criterion. Weighing factors were assigned to all criteria

to make P, B, T and M equally important. One problem for MAUT is that this

method requires aggregation independence,139 which is problematic for some

criteria that are correlated to similar physicochemical properties. Another

problem is that the MAUT score is based on average values taken from different

models, neglecting the model uncertainties. In order to take these model

uncertainties into consideration, thresholds for ELECTRE III were set based on

the variation of different model results (Paper III and IV).

Current alternatives assessment frameworks use various final decision making

approaches, from general narrative guides to quantitative approaches like MCDA

(Table 9).10 The MCDA methods used in Paper III and IV were applied for the

ranking of hazards, but it can readily be expanded to include other aspects of

alternatives assessment (e.g. exposure, technical performance and economic

costs). A recent study140 shows that the use of MCDA methods has become

increasingly popular in the field of alternatives assessment, and reported also that

many other alternatives assessments used more than one MCDA method for a

more trustworthy conclusion. It should also be pointed out that besides

identification of the best alternatives, MCDA methods have other benefits for

alternatives assessment, such as for identifying the most critical criteria. MCDA

methods need not be difficult to implement, such as MAUT and ELECTRE III in

Paper III and IV, which was conducted using simple spreadsheet. Stakeholders

and researchers can easily adapt the method either for hazard ranking or for final

decision making of alternatives by setting new thresholds and weighting factors.

A previous case study has shown an example of using two MCDA methods to

make decision based on different hazard impacts as well as technical and

economic performance, and was able to present a transparent result of the

rankings of different alternatives, as well as the impact of each different aspect.68

22

Table 9. Decision approaches used in selected alternatives assessment frameworks

Alternatives assessment framework Decision approaches

USEPA CTSA11 Narrative guide for decision maker

UNEP POPs Review Committee’s General Guidance on Alternatives13

Not specified

BizNGO chemicals alternatives assessment protocol including GreenScreen®15

Structured ranking system

NAS guideline17 Analytical decision tools including MCDA

European Commission DGE12 Analytical decision tools based on cost-benefit analysis

Paper III and IV (for hazard ranking)

In silico hazard data coupled with MCDA

3.5 Overall hazard ranking results of alternatives for

decaBDE

From the heat map in Paper III (Figure 4), it can be concluded that none of the

studied flame retardants is “hazard free”, which agrees with a previous screening

of flame retardants.27 The ranking from the two MCDA methods generally

correlated with each other (R2=0.6) (Figure 5). As the thresholds for ELECTRE

III were set based on the uncertainties in in silico models, the disagreement of

the two methods for some case chemicals indicates that data uncertainties have

an impact on the decision results; while the overall agreement indicates that for

the studied case chemicals these uncertainties are generally smaller than the

hazard level differences between the studied flame retardants. Based on MAUT

and ELECTRE III results, studied OPFRs are less hazardous than decaBDE but

do not appear among the best alternatives. The two MCDA methods yield similar

results on potential regrettable substitutes (e.g. EBPEBPI, TBBPA-BDBPE,

BTBPE and DP) as well as the less hazardous alternatives DBDPE, BEH-TEBP

and melamine.

Several alternatives assessment frameworks also address the issue of data

uncertainties in hazard assessment (Table 7), but usually in a broad and

qualitative way. For example, CTSA lists probable sources of uncertainties but

discuss them in a qualitative way.11 DGE generally concludes that there is “no

clear answer” for how to handle data uncertainties as it depends on several

different factors, and only suggests that all uncertainties need to be recorded and

an “educated guess” might be needed to assess the overall reliability based on the

recorded uncertainties.12 GreenScreen ranks the confidence level of hazard data

23

based on data source and data quality and marks the confidence level in the heat

map141, while for the case of open source QSAR tools used in Paper III and IV,

the data quality are usually not clearly reported. In such cases, GreenScreen

suggests expert judgement to assess data quality.141 The NAS guideline suggests

multiple ways to deal with data uncertainties, including propagate uncertainties

to the aggregated results using trade-off functions in MCDA methods.17 Paper

III and IV demonstrate a more sophisticated way to take uncertainty into

account by pairwise comparisons in MCDA (ELECTRE III) using in silico data to

assess hazard. Using these methods provided conclusive results for the studied

flame retardants as a case study.

Figure 5. Ranking results from the two MCDA methods (PBMT scoring)

The heat map presented in paper IV (Figure 3 in Paper IV) showed that most

of the selected transformation products have similar environmental persistency

but higher mobility compared to their parent compounds, in agreement with

previous studies50, 142. BFRs also appear to more likely transform into more

bioaccumulative forms. As for toxicity, burden shifting was widely observed that

the transformation products are worse than their parent compound for some

toxicity criteria, and better for others, in agreement with a previous study.62

After including transformation products in Paper IV, MAUT and ELECTRE III

results both showed that decaBDE is clearly more hazardous compared to the

alternative flame retardants, while the differences among the other alternatives

become less significant. Based on both parent and transformation compounds,

BEH-TEBP and melamine still remained the least-hazardous alternatives for

decaBDE, while DBDPE as an alternative became less preferable if

transformation products were considered (Figure 4 in Paper IV). Again, the

24

general agreement between MAUT and ELECTRE III indicated that the data

uncertainty in QSAR models had low impact on the absolute ranking of the

studied alternatives. In Paper III and IV, each model for the same hazard

property was assigned with the same weight, regardless of varying uncertainties

for the individual models. A proper next step to further reduce the impact of

uncertainties would be to better evaluate the applicability domains and

prediction accuracies of the applied QSAR models, and assign weighting factors

based on that.

For BEH-TEBP, one concern is that it’s usually used in technical mixtures

together with other flame retardants like EH-TBB, which is a more hazardous

chemical according to the analysis presented in Paper III. For melamine, both

itself and its transformation products generally appears to be less toxic, while

Paper III and IV suggested that the high persistency and high mobility in the

aquatic environment is of concern for both melamine and its transformation

products. In addition, melamine has been heavily used in many products both as

a flame retardant and as additive for other purposes, thus the general exposure

level may be higher compared to other flame retardants. Taken together this

suggests that although these two compounds have been identified as less

hazardous by the hazard ranking tools developed, whether they are truly safer in

any certain product needs to be evaluated by a higher tier alternatives assessment

approach, which considers both hazard and exposure.

3.6 Measurement of OPFRs emission and modelling

approaches

3.6.1 QCM method development and OPFRs emission measurement

Although there are alternatives assessment frameworks that focus on hazard

ranking with less consideration of exposure, most frameworks take both hazard

and exposure into account, either in parallel or include an exposure assessment

after the hazard assessment.10 When assessing exposure, most of the existing

alternatives assessment frameworks do not use exposure data or models due to

their complexity; they rather use indirect parameters like physicochemical

properties, use characteristics, emission and environmental fate data.10 In

comparing alternatives, it is important to be able to describe the emission process,

or understand the emission mechanism (driven by physicochemical properties)

to assess variation in risk of exposure to alternatives.

In Paper V, an emission measurement method was developed using QCM to

assess exposure of OPFR from water. TCEP and TPP emissions from PS films

were measured with different initial concentrations (20−50 wt.-%, considering

25

the typical OPFR loads in polymer products are 3−70 wt.-%19, 44). Both the release

pattern (Figure 1 of Paper V) and scanning electron microscope results (Figure

2 of Paper V) suggest that a high loading (40 and 50 wt.-% of OPFRs) created

syneresis, meaning that a significant fraction of the OPFRs was concentrated on

the film surface. In order to reach homogeneous polymer films that can better

represent the situation in real life, experiments were conducted using 30 wt.-% of

OPFRs in the polymers. The repeatability of the developed QCM method was

examined by TPP emission yielding a low standard error (± 2.4% with 95%

confidence interval); the accuracy of the developed QCM was validated by GC-MS,

and proved to have high accuracy and low variation.

The developed method was then applied to nine OPFRs and their emissions were

studied from PS films (Figure 3 of Paper V), and a selected subset of OPFRs

(TBP, TDCPP, and TPP) from two other polymers (PVC and PMMA, Figure 4 of

Paper V). TBP, TDCPP, and TPP were selected because of their similar

hydrophobicity (as quantified by octanol-water partition coefficient (KOW) values)

despite different molecular structures.

3.6.2 Modelling approaches to describe the emission pattern and

understand the emission mechanism

Empirical models have been used to describe the emission processes of small

molecules from polymers into different media (air, water or biota tissues)143-147,

and several of these models fit well with the emission data measured in Paper V

(Table S5 in Paper V). However, a previous study pointed out that many of the

empirical models have the risk of overparameterization that may lead to faulty

conclusions.145 To avoid this, a sensitivity analysis was conducted and results

indicated that all models with three or more parameters have the risk of

overfitting, while the one-compartment Weibull distribution model145, 148 can also

fit well (R2=0.90-0.99, Figure 6) and with lowest risk of overfitting. Previous

studies usually considered the emission process as a combination of two phases.

The first phase is a fast releasing phase, which is dominated by the partitioning

of additives between the polymer surface and water.149 This is followed by a

diffusion dominated phase with a lower emission rate.143, 150-152 The Weibull

model has a scale parameter α and a shape parameter β. The sensitivity analysis

showed that the parameter α has more impact on the initial fast releasing phase,

whereas β is more related to the latter slower phase (Figure 6).

26

Figure 6. Illustrative example of Weibull model describing the two-phase emission,

where m0 is the initial mass of OPFR in the measured film and mt is the mass of OPFR

remaining in the film at time point t; α and β are Weibull model parameters obtained from

fitting the experimental data

In order to understand the emission mechanisms, a correlation between Weibull

model parameters and physico-chemical properties of OPFRs was conducted for

the emission of OPFRs from PS films. The results showed that log α was linearly

correlated to the water solubility of OPFRs and their hydrophobicity (KOW)

(Figure S13 in Paper V), which agrees with previous studies regarding

partitioning152-156. The diffusion phase on the other hand, is typically believed to

be driven by molecular size of additives or strength of interactions between

polymer and additive.151, 157-159 For the nine studied OPFRs, the β parameter

showed some correlation with molecular volume and with the solubility of OPFRs

in PS reflected by DHSP (Figure S13 in Paper V). One possible reason for the lower

correlation for β is that there might be other factors influencing the diffusion,

such as the other structure differences among the OPFRs. Another reason might

be that the QSAR predicted DHSP values for the nine OPFRs showed a low

variation that can increase the uncertainty in correlation.

Correlation between Weibull model parameters and properties of both polymers

and additives were also studied for the emission of TBP, TDCPP, and TPP from

three different polymers, while no clear relationship could be discovered. Besides

potential measurement uncertainties, the most likely reason is that both the

partition and diffusion phases are not driven by any single property of either

polymers or additives, but more likely driven by a combination of different

27

properties. This can be further studied by multivariate models, which require a

larger amount of data from polymers and additives with a wider range of

physicochemical properties. Also, the study noticed a lack of parameters to

describe the interactions between polymers and additives, for example the

crystallinity of polymers can change with different concentration of plasticizing

additives160.

Overall, in paper V, an efficient QCM method was developed to measure the

emission process of polymer additives. The measured OPFRs emitted from the

studied polymers and the Weibull model indicated that hydrophobicity and

molecular size are important drivers of emission and thus can be used as

indicators for exposure via water in alternatives assessment. The study also

demonstrated the complexity of the emission process of additives from for

materials, which justifies the need for further studies.

28

4 Conclusions

Alternatives assessment requires consideration of multiple aspects (Figure 1),

and data gaps are the biggest challenge. This thesis explored the use of in silico

tools for the prediction of hazard properties and the formation of chemical

transformation products within the frame of alternatives assessment. Different

MCDA methods were compared for hazard ranking and a new measurement

method was developed to assess emissions of organic additives from polymers to

water that was also studied using various modelling approaches.

Paper I and II showed how QSAR models can be established to fill data gaps in

hazard profiles with the example of endocrine disruption related endpoints. In

paper I QSARs were developed to increase the understanding of the underlying

structural drivers for modelled features, and Paper II showed how QSAR models

can be established and validated to predict hazard properties for a large number

of chemicals. In paper III such QSARs were used to develop a hazard ranking

tool for alternatives assessment. This study demonstrated that there are currently

sufficient open access QSAR models available for many hazard properties

prioritized in most of the existing alternatives assessment frameworks (Table 7).

These results are promising for other cases in which in silico models are the only

available hazard property data source for novel alternatives, particularly when

the alternatives fall within the chemical applicability domain of the QSARs. In

paper IV a tool was developed to include transformation products, and showed

how in silico tools can be used for the prediction of such products. An approach

was also developed to prioritize the transformation products with the highest

occurrence potential from the prediction results of multiple in silico tools, and

the prioritized transformation products were further included in the comparison

of chemical hazards within the overarching alternatives assessment. The use of

MCDA tools for decision making in alternatives assessment was explored in

paper III and IV, and advantages and disadvantages of different MCDA

methods were discussed. These studies showed that not only levels set as

regulatory criteria, but also relative hazard levels between the studied alternatives,

should be considered when conducting MCDA in hazard ranking, Further, data

uncertainties should also be included in MCDA, specifically in the context of using

QSARs and read-across or grouping approaches. None of the studied flame

retardants is hazard-free; however, melamine and BEH-TEBP appear to be the

two least hazardous alternatives for decaBDE using the estimated, in silico data

and applied MCDA methods.

Since these two least hazardous alternatives are not hazard-free, the choice of

flame retardant for certain product needs to be further evaluated with

consideration of other aspects like exposure and life cycle impacts. In Paper V,

29

an efficient QCM based method was developed to measure the emission of OPFRs

into water, and the Weibull model was chosen to describe the emission process

to avoid the risk of over-fitting. Many alternatives assessment frameworks use

physico-chemical properties that influence the emission process to assess the

exposure aspect. In Paper V, hydrophobicity and molecular size of plastic

additives were identified as the most important physico-chemical properties for

the emission process, but also pointed out that there might be other properties

and interactions between polymers and additives that remain to be studied. Since

the OPFRs with smaller molecule size and lower hydrophobicity are generally less

hazardous (Paper III) but with higher emission potential into aquatic

environments (Paper V), selection of the most sustainable OPFR requires a

higher-tier quantitative risk assessment for specific uses.

In summary, this thesis shows that in silico tools can be useful for different

aspects of alternatives assessment, especially for narrowing down the range of

potential safer alternatives early in the design or chemical alternative selection

process. Such an approach could be integrated into the concept of Safe and

Sustainable-by-Design, as recently promoted by the EU Chemicals Strategy for

Sustainability.161

30

5 Future perspectives

The research presented in this thesis focused on the hazard and exposure

components of alternatives assessment. The research can be further expanded

into alternatives assessment frameworks that include other components that

need further development, such as technical performance, economic costs and

life cycle impacts for specific chemical uses.

5.1 Improving in silico tools for hazard assessment

In Paper III and IV, various open source QSAR tools were used for hazard

ranking. One identified problem of these tools is that chemical applicability

domain and prediction quality are not clearly defined and transparent in these

QSAR models, and comparisons conducted in Paper III between outputs of

different models for the same hazard criteria as well as modelling outputs with

experimental data suggest there can be substantial prediction uncertainties. Also,

for some hazard criteria, the number of available QSAR tools is limited, and high

quality QSAR models are desired. For example, the environmental half-lives

predicted by EPISuite are the most widely-used quantitative model for

environmental persistency prediction, while the predicted half-lives for water,

soil and sediment are proportional to each other as they are all derived from the

same model (BIOWIN) within EPISuite. All these problems point out a need to

develop new QSAR models with clear and transparent information. In turn, these

will need new, high-quality experimental data for establishing these models.

Another problem identified in Paper III and IV for hazard assessment is the

potential underestimation of bioaccumulation. In REACH, the B parameter is

based on the estimated bioconcentration factor (BCF) in fish with a limit value of

2000,1 and most of the available QSAR models are developed for BCF predictions.

However, BCF only considers the partitioning between the organism and the

surrounding environment (water), and does not account for biomagnification via

food intake, which can be a more critical pathway especially for very hydrophobic

compounds. In such cases, the B parameter assessed by BCF are likely to be

underestimated. For example, the measured bioaccumulation factor (log BAF) for

DBDPE can be up to 7.1 considering food and diet,162 while the average predicted

log BCF value (Paper III) is 0.82, and this is quite common for many other

similar compounds.163 In Paper IV, the use of BAF predicted by EPISuite is

discussed to assess B, and the paper pointed out the problem that using this single

model has higher uncertainty compared to the use of BCF based on several

models. This points out the need for both experimental measurements and model

establishment for other B related measures like BAF, which are currently

lacking.163

31

Paper IV highlighted that some of the QSAR tools used in Paper III and IV

are only suitable for neutral compounds while some alternatives or their

transformation products might be ionized under environmental conditions, thus

research is needed to advance the in silico tools to enable hazard assessment of

ionizable compounds. Another research gap is models that cover inorganics and

polymers. These alternatives were excluded from Paper III and IV as the QSAR

methodologies are not able to calculate their hazard properties. There are

alternative in silico approaches that can be applied and further developed, for

example the rule-based estimations for both inorganic chemicals164 and

polymers165, 166, and recently provided QSAR tools for the prediction of polymer

toxicities167.

5.2 Measurement of emission process and model description

In Paper V, QCM measurements were conducted for nine OPFRs emitted from

three different polymers. For future studies of emission mechanisms of plastic

additives using QCM, measurements are needed for a larger range of plastic

additives with different physico-chemical properties, such as crystallinity, as well

as different environmental conditions (e.g. temperature, pH etc.). Further

method developments are needed to increase the sensitivity and reduce

uncertainties, so studies can be applied to polymer films with lower additive

loadings. Model approaches would also require improvement to increase the

accuracy of the description of the emission process under different use scenarios.

The experimental method could also be developed to capture emissions from

polymers to air.

To better model and describe emission processes of organic chemicals from

polymers, more informative parameters describing polymer properties are

required. For organic additives (small organic molecules), there are now

thousands of available chemical descriptors to describe property differences,

while both calculated and experimental parameters are lacking to describe

polymer properties such as crystallinity and molecular diffusion in polymers.

5.3 Improving prediction for chemical transformation

An important limitation of in silico tools used in Paper IV for the prediction of

transformation products is that there is no clear information regarding the

applicability domain of the models, and it is hard to evaluate the quality of

predictions. Also, transformation rates and the fractions of formation of different

transformation products under certain environmental conditions are not

predicted by any of the tools. To address these weaknesses, a strategy was used in

Paper IV to prioritize predicted transformation products that have the highest

occurrence potential. This strategy can be improved by further validation and

32

improvement of the in silico tools, which requires more high-quality target and

non-target screening data to unravel chemical and biological transformation

reactions.

5.4 Life cycle aspects in alternatives assessment

In order to avoid regrettable substitutions, it is important to consider the impacts

of products including the whole life cycle, from the extraction and manufacturing

of the crude materials to the final disposal of the waste product (cradle to grave).

An example of regrettable substitution is N-vinyl formamide that was chosen to

replace the neurotoxic acrylamide in manufacturing polymers for water

treatment.6 Although N-vinyl formamide itself is a safer compound, the synthesis

of N-vinyl formamide requires toxic hydrogen cyanide during synthesis. Many

existing alternatives assessment frameworks include different levels of

consideration of life cycle impacts.10 For example, the National Research Council

gives guidance on a qualitative assessment of shifts in toxicity impacts caused by

different production procedures of alternatives.6 Another option is that after a

preferrable alternative has been identified based on a single aspect (for example

the hazard ranking using the tools developed in Paper III and IV), a thorough

life cycle assessment can be conducted to evaluate potential impacts in all life

cycle stages for all impact categories for the final selection of a sustainable

alternative.10

Data availability is the most apparent issue for the inclusion of life cycle impacts

in alternatives assessment. For example, production data of novel flame

retardants like BEH-TEBP is lacking in LCA databases like GaBi168, and even for

flame retardants like decaBDE and melamine that have production data, the

emission of these compounds during the production processes are lacking in

these databases. In such cases, assumptions can be made based on in silico

methods as discussed in this thesis. For example, the lack of emission factors can

be estimated by QSAR models built on chemicals with similar structure, there are

QSAR based tools like USETox169 for toxicity characterization,170 and studies that

combine USETox with other in silico tools have also been presented.171 The MCDA

tools developed in Paper III and IV can be further developed for the decision

making stage of LCA, as there are different life cycle impacts that need to be

balanced by proper trade-off methods to select a more safe and sustainable

alternative.

33

6 Acknowledgements

First of all, I would like to express my deepest gratitude to my supervisor Patrik

Andersson for providing me this opportunity to work on this challenging but

interesting project. I am greatly appreciated for your inspiring help and patient

support all the time, guiding me since the beginning when I had little experience

in this field. Thanks for your guidance I have grown my interest in the field of

environmental chemistry, and you are also a good example for me as well as the

other students in your group on how to be a good scientist, teacher, project leader

and group leader.

I would like to express my great appreciation for my co-supervisors, Gregory

Peters from Chalmers University of Technology and Hans Peter Arp from

Norwegian Geotechnical Institute. Greg gave me a lot of help in the field of

decision analysis and life cycle assessment which I had no experience before, all

the way from course recommendation to the use of software. Hans Peter gave me

plenty of insight in the field of environmental chemistry including PMT

assessment and chemical transformations. I also appreciate your valuable

guidance and great contribution to the manuscripts and this thesis.

I would like to thank all the other co-authors of my papers. For the QCM project,

big thanks to Linhong Xiao for leading the project and conducting all the

experimental work, and to Knut Irgum for your constructive and enlightening

contributions. To Linying Cao, Xiaomin Ren and Lianghong Guo from Chinese

Academy of Sciences, it’s my pleasure to cooperate with you in the ERRγ project.

For the CoMPARA project, although I haven’t got much chance to communicate

with most of the co-authors, it was really a great project and it was an honor to be

able to participate in such a project in the beginning of my PhD studies.

Thanks to the Department of Chemistry, Umeå University for providing a good

studying, teaching and working environment. Thanks to all the friendly and

helpful colleagues. Especially, thanks to my office mates, Marcus Östman, Majid

Mustafa and Dong Wang. I felt very comfortable working with you. Thanks should

also go to my annual evaluation committee members, Erik Chorell and Venkata

Krishna Kumar Upadhyayula, for your interesting suggestions to my research.

To the former and current members of our research group; Jin Zhang, Qiuju Gao,

Kristin Blum, Linhong Xiao, Ioana Chelcea, Elena Dracheva, Jenny Lexén and

Maria Sapounidou, I have gotten many help from all of you, and the discussions

in our journal club and group meetings have always been illuminating. The pizza

and Chinese hot pot party with you were enjoyable and relaxing. Also, thanks to

all the other current and former members of the Environmental Chemistry group.

I appreciate the harmonic atmosphere for conducting research here, and I really

enjoy the Friday Fika.

34

I would also like to thank the funding source of this thesis. This thesis has

emerged through financial support from Faculty of Science and Technology,

Umeå University and a grant from the Swedish Research Council for the

Environment, Agricultural Sciences and Spatial Planning (Formas) project No.

942-2015-672 “Ett verktyg för att identifiera hållbara plastkemikalier (A tool for

identifying sustainable plastic chemicals)”.

Sincere gratitude also goes to my previous teachers from Tongji University,

Stockholm University and Umeå University, as well as the teachers in different

courses I took during my PhD studies from various universities of Sweden and

Finland. In particular, I would like to thank Professor Yanling Qiu from Tongji

University for being my guide to the field of environmental chemistry and offering

me the opportunity to study environmental chemistry in Sweden, and Anna Strid

from Stockholm University who gave me a lot of help when supervising my master

project. I’m lucky to have your guidance in the beginning phase of my research

journey.

I would like to express my very profound gratitude to my friends; Shengyang

Chen and Min Yao, my close friends since childhood, thank you for your help

when I applied for this PhD position and your warm hospitality when I visited the

UK. Xiaojun Liu, our daily chatting and gaming nights have been an important

part of my PhD life, and thanks for your encouragement all the time. Xinyu Du, I

really enjoy our gaming weekends, and it’s my honor to participate in many of

your research projects. Ge Yin and Yihui Zhou, thanks for offering me a lot of help

both with my study and with my life when I first came to Sweden. Also, I would

like to thank those friends who were willing to choose me as your travel mates,

including Xiaojun Liu, Tingting Ji, Yumeng An and Mengjiao Zhang. Together

with you I have explored many fascinating countries in Europe during the past

years. To Xiaojun Liu, Xinyu Du, Chang Liu, Anni Sun and Tingting Ji, I truly love

our annual summer get-together in Shanghai. Hopefully the pandemic situation

will end soon and we can meet again next summer.

Finally, and most importantly, my deepest gratitude to my parents and family.

This thesis would never be accomplished without your support and

encouragement. 最后也是最重要的，特别感谢我的父母与家人。我能够顺利完成

博士学业离不开你们的鼓励与支持。

The just passed 2020 have been a tough year for everyone in the world due to the

pandemic attack of Covid-19. I am grateful for all help and support from all

friends and colleagues to be able to complete this interesting work, and I truly

wish that with the support from each other, we can all overcome the current

situation and welcome our beautiful life back soon.

35

7 References

1. EU Regulation (EC) No 1907/2006 of the European Parliament and of the

Council of 18 December 2006 concerning the Registration, Evaluation,

Authorisation and Restriction of Chemicals (REACH), establishing a European

Chemicals Agency, amending Directive 1999/45/EC and repealing Council

Regulation (EEC) No 793/93 and Commission Regulation (EC) No 1488/94 as

well as Council Directive 76/769/EEC and Commission Directives 91/155/EEC,

93/67/EEC, 93/105/EC and 2000/21/EC. Off. J. Eur. Communities: Legis.

2006, L396, 1-849.

2. Regulation, E., No 1107/2009 of the European Parliament and of the Council

of 21 October 2009 concerning the placing of plant protection products on the

market and repealing Council Directives 79/117/EEC and 91/414/EEC. EU,

Brussels 2009.

3. Toxic Substances Control Act, Pub. L. No. 94–469, 90 Stat. 2003 (October

11, 1976).

4. Frank R. Lautenberg Chemical Safety for the 21st Century Act, H.R. 2576,

2016, 114th Congress, Washington, DC, USA.

5. Report of the Persistent Organic Pollutants Review Committee on the work

of its fifth meeting, addendum: General guidance on considerations related to

alternatives and substitutes for POPs; Persistent Organic Pollutants Review

Committee: 2009.

6. A framework to guide selection of chemical alternatives. National

Academies Press: Washingto DC, 2014.

7. Zimmerman, J. B.; Anastas, P. T., Toward substitution with no regrets.

Science 2015, 347, (6227), 1198-1199.

8. Horan, T. S.; Pulcastro, H.; Lawson, C.; Gerona, R.; Martin, S.; Gieske, M. C.;

Sartain, C. V.; Hunt, P. A., Replacement Bisphenols Adversely Affect Mouse

Gametogenesis with Consequences for Subsequent Generations. Curr. Biol. 2018,

28, (18), 2948-2954.e3.

9. Tickner, J. A.; Schifano, J. N.; Blake, A.; Rudisill, C.; Mulvihill, M. J.,

Advancing safer alternatives through functional substitution. Environ. Sci.

Technol. 2015, 49, (2), 742-749.

36

10. Jacobs, M. M.; Malloy, T. F.; Tickner, J. A.; Edwards, S., Alternatives

assessment frameworks: research needs for the informed substitution of

hazardous chemicals. Environ. Health Perspect. 2016, 124, (3), 265.

11. Kincaid, L.; Meline, J.; Davis, G. Cleaner technologies substitutes

assessment (CTSA) methodology and resource guide; EPA744: 1996.

12. Gilbert, Y.; Pessala, P.; Aho, J.; Lehti, R.; Vehviläinen, I.; Hjelt, M.,

Minimising Chemical Risk to Workers’ Health and Safety through Substitution.

Publications Office of the European Union: Luxembourg, 2012.

13. Stockholm Convention on Persistent Organic Pollutants Review Committee.

Report of the persistent organic pollutants review committee on the work of its

first meeting. UNEP/POPS/POPRC.1/10. Geneva, 7 - 11, November 2005.

14. Quinn, M. M.; Fuller, T. P.; Bello, A.; Galligan, C. J., Pollution prevention—

occupational safety and health in hospitals: alternatives and interventions. J.

Occup. Env. Hyg. 2006, 3, (4), 182-193.

15. Rossi, M.; Peele, C.; Thorpe, B. BizNGO chemicals alternatives assessment

protocol: How to select safer alternatives to chemicals of concern to human health

or the environment.

https://www.bizngo.org/static/ee_images/uploads/resources/BizNGOChemica

lAltsAssessmentProtocol_V1.1_04_12_12-1.pdf

16. Rosenberg, B. J.; Barbeau, E. M.; Moure‐Eraso, R.; Levenstein, C., The work

environment impact assessment: A methodologic framework for evaluating

health‐based interventions. Am. J. Ind. Med. 2001, 39, (2), 218-226.

17. Council, N. R., A Framework to Guide Selection of Chemical Alternatives.

2014.

18. Europe, P., Plastics-The facts 2018: An Analysis of European Plastics

Production, Demand and Waste Data. 2018.

19. Hahladakis, J. N.; Velis, C. A.; Weber, R.; Iacovidou, E.; Purnell, P., An

overview of chemical additives present in plastics: Migration, release, fate and

environmental impact during their use, disposal and recycling. J Hazard Mater

2018, 344, 179-199.

20. Xu, Y.; Liu, Z.; Park, J.; Clausen, P. A.; Benning, J. L.; Little, J. C., Measuring

and Predicting the Emission Rate of Phthalate Plasticizer from Vinyl Flooring in

37

a Specially-Designed Chamber. Environ. Sci. Technol. 2012, 46, (22), 12534-

12541.

21. Liang, Y. R.; Liu, X. Y.; Allen, M. R., Measurements of Parameters

Controlling the Emissions of Organophosphate Flame Retardants in Indoor

Environments. Environ. Sci. Technol. 2018, 52, (10), 5821-5829.

22. Paluselli, A.; Fauvelle, V.; Galgani, F.; Sempere, R., Phthalate Release from

Plastic Fragments and Degradation in Seawater. Environ. Sci. Technol. 2019, 53,

(1), 166-175.

23. Li, J.; Yu, N. Y.; Zhang, B. B.; Jin, L.; Li, M. Y.; Hu, M. Y.; Zhang, X. W.; Wei,

S.; Yu, H. X., Occurrence of organophosphate flame retardants in drinking water

from China. Water Res 2014, 54, 53-61.

24. McDonough, C. A.; De Silva, A. O.; Sun, C. X.; Cabrerizo, A.; Adelman, D.;

Soltwedel, T.; Bauerfeind, E.; Muir, D. C. G.; Lohmann, R., Dissolved

Organophosphate Esters and Polybrominated Diphenyl Ethers in Remote Marine

Environments: Arctic Surface Water Distributions and Net Transport through

Fram Strait. Environ. Sci. Technol. 2018, 52, (11), 6208-6216.

25. Lithner, D.; Larsson, A.; Dave, G., Environmental and health hazard ranking

and assessment of plastic polymers based on chemical composition. Sci Total

Environ 2011, 409, (18), 3309-3324.

26. Birnbaum, L. S.; Staskal, D. F., Brominated flame retardants: cause for

concern? Environ. Health Perspect. 2004, 112, (1), 9.

27. Gramatica, P.; Cassani, S.; Sangion, A., Are some “safer alternatives”

hazardous as PBTs? The case study of new flame retardants. J. Hazard. Mater.

2016, 306, 237-246.

28. Shaw, S., Halogenated flame retardants: do the fire safety benefits justify the

risks? Rev. Environ. Health 2010, 25, (4), 261-306.

29. Robrock, K. R.; Korytár, P.; Alvarez-Cohen, L., Pathways for the anaerobic

microbial debromination of polybrominated diphenyl ethers. Environ. Sci.

Technol. 2008, 42, (8), 2845-2852.

30. Boon, J. P.; Lewis, W. E.; Tjoen-A-Choy, M. R.; Allchin, C. R.; Law, R. J.; de

Boer, J.; ten Hallers-Tjabbes, C. C.; Zegers, B. N., Levels of polybrominated

diphenyl ether (PBDE) flame retardants in animals representing different trophic

38

levels of the North Sea food web. Environ. Sci. Technol. 2002, 36, (19), 4025-

4032.

31. Darnerud, P. O., Toxic effects of brominated flame retardants in man and in

wildlife. Environ. Int. 2003, 29, (6), 841-853.

32. Herbstman, J. B.; Sjödin, A.; Kurzon, M.; Lederman, S. A.; Jones, R. S.; Rauh,

V.; Needham, L. L.; Tang, D.; Niedzwiecki, M.; Wang, R. Y., Prenatal exposure to

PBDEs and neurodevelopment. Environ. Health Perspect. 2010, 118, (5), 712.

33. Schreiber, T.; Gassmann, K.; Götz, C.; Hübenthal, U.; Moors, M.; Krause, G.;

Merk, H. F.; Nguyen, N.-H.; Scanlan, T. S.; Abel, J., Polybrominated diphenyl

ethers induce developmental neurotoxicity in a human in vitro model: evidence

for endocrine disruption. Environ. Health Perspect. 2010, 118, (4), 572.

34. Chao, H.-R.; Wang, S.-L.; Lee, W.-J.; Wang, Y.-F.; Päpke, O., Levels of

polybrominated diphenyl ethers (PBDEs) in breast milk from central Taiwan and

their relation to infant birth outcome and maternal menstruation effects. Environ.

Int. 2007, 33, (2), 239-245.

35. H Türk, A., Joined Cases C-14/06 & C-295/06, European Parliament and

Denmark v. Commission, judgment of the Court (Grand Chamber) of 1 April

2008. Common Market Law Review 2009, 46, (4), 1293-1303.

36. Lassen, C.; Havelund, S.; Leisewitz, A.; Maxson, P., DecaBDE and

alternatives in electrical and electronic equipment. Danish Ministry of the

Environment: Copenhagen, Denmark, 2006.

37. Pakalin, S.; Cole, T.; Steinkellner, J.; Nicolas, R.; Tissier, C.; Munn, S.;

Eisenreich, S. Review on production processes of decabromodiphenyl ether

(decaBDE) used in polymeric applications in electrical and electronic equipment,

and assessment of the availability of potential alternatives to decaBDE;

European Commission, Directorate General Joint Research Center, European

Chemicals Bureau: 2007.

38. All POPs listed in the Stockholm Convention; United Nations Environment

Programme: 2017.

39. DecaBDE phase-out initiative; US Environmental Protection Agency: 2010.

40. Betts, K., Glut of data on “new” flame retardant documents its presence all

over the world. Environ. Sci. Technol. 2009, 43, (2), 236-237.

39

41. Ricklund, N.; Kierkegaard, A.; McLachlan, M. S., Levels and Potential

Sources of Decabromodiphenyl Ethane (DBDPE) and Decabromodiphenyl Ether

(DecaBDE) in Lake and Marine Sediments in Sweden. Environ. Sci. Technol.

2010, 44, (6), 1987-1991.

42. Barón, E.; Santín, G.; Eljarrat, E.; Barceló, D., Occurrence of classic and

emerging halogenated flame retardants in sediment and sludge from Ebro and

Llobregat river basins (Spain). J. Hazard. Mater. 2014, 265, 288-295.

43. Stapleton, H. M.; Allen, J. G.; Kelly, S. M.; Konstantinov, A.; Klosterhaus, S.;

Watkins, D.; McClean, M. D.; Webster, T. F., Alternate and New Brominated

Flame Retardants Detected in U.S. House Dust. Environ. Sci. Technol. 2008, 42,

(18), 6910-6916.

44. Wei, G.-L.; Li, D.-Q.; Zhuo, M.-N.; Liao, Y.-S.; Xie, Z.-Y.; Guo, T.-L.; Li, J.-J.;

Zhang, S.-Y.; Liang, Z.-Q., Organophosphorus flame retardants and plasticizers:

sources, occurrence, toxicity and human exposure. Environ. Pollut. 2015, 196,

29-46.

45. Sun, L.; Xu, W.; Peng, T.; Chen, H.; Ren, L.; Tan, H.; Xiao, D.; Qian, H.; Fu,

Z., Developmental exposure of zebrafish larvae to organophosphate flame

retardants causes neurotoxicity. Neurotoxicol. Teratol. 2016, 55, 16-22.

46. Hendriks, H. S.; Westerink, R. H., Neurotoxicity and risk assessment of

brominated and alternative flame retardants. Neurotoxicol. Teratol. 2015, 52,

248-269.

47. Urs Berger, N. O., Daniel Sättler, Ivo Schliebner, Ralph Kühne, Gerrit

Schüürmann, Michael Neumann, Thorsten Reemtsma, Assessment of

persistence, mobility and toxicity (PMT) of 167 REACH registered substances.

2018.

48. Fritz Kalberlah, J. O., Markus Schwarz, Guidance for assessing persistence,

mobility and toxicity of chemicals registered under REACH and their metabolites

to protect raw water destined for drinking water: A Practical Guide. 2014.

49. Neumann, M.; Schliebner, I., Protecting the sources of our drinking water−

A revised proposal for implementing criteria and an assessment procedure to

identify persistent, mobile and toxic (PMT) and very persistent, very mobile

(vPvM) substances registered under REACH; position, October 2017. In German

Environment Agency: 2017; p 16.

40

50. Arp, H.; Brown, T.; Berger, U.; Hale, S., Ranking REACH registered neutral,

ionizable and ionic organic chemicals based on their aquatic persistency and

mobility. Environ. Sci. Process. Impacts 2017, 19, (7), 939-955.

51. Schulze, S.; Sättler, D.; Neumann, M.; Arp, H. P. H.; Reemtsma, T.; Berger,

U., Using REACH registration data to rank the environmental emission potential

of persistent and mobile organic chemicals. Sci. Total Environ. 2018, 625, 1122-

1128.

52. Commission, E. E., Communication from the Commission to the European

Parliament, the Council, the Economic and Social Committee and the Committee

of the Regions. Our life insurance, our natural capital: an EU biodiversity

strategy to 2020, 1-14.

53. Patlewicz, G.; Roberts, D. W.; Aptula, A.; Blackburn, K.; Hubesch, B.,

Workshop: use of “read-across” for chemical safety assessment under REACH.

Regul. Toxicol. Pharm. 2013, 65, (2), 226-228.

54. Gramatica, P., Principles of QSAR models validation: internal and external.

QSAR & combinatorial science 2007, 26, (5), 694-701.

55. OECD principles for the validation, for regularoty purposes, of (quantitative)

structure-activity relationship models.

http://www.oecd.org/chemicalsafety/risk-assessment/37849783.pdf

56. Arnold, S. M.; Greggs, B.; Goyak, K. O.; Landenberger, B. D.; Mason, A. M.;

Howard, B.; Zaleski, R. T., A quantitative screening‐level approach to incorporate

chemical exposure and risk into alternative assessment evaluations. Integr.

Environ. Assess. Manag. 2017, 13, (6), 1007-1022.

57. Control, C. f. D.; Prevention, n-Hexane-related peripheral neuropathy

among automotive technicians--California, 1999-2000. MMWR. Morbidity and

mortality weekly report 2001, 50, (45), 1011.

58. Sun, B. B.; Hu, Y. N.; Cheng, H. F.; Tao, S., Kinetics of Brominated Flame

Retardant (BFR) Releases from Granules of Waste Plastics. Environ. Sci. Technol.

2016, 50, (24), 13419-13427.

59. Sun, B. B.; Hu, Y. A.; Cheng, H. F.; Tao, S., Releases of brominated flame

retardants (BFRs) from microplastics in aqueous medium: Kinetics and

molecular-size dependence of diffusion. Water Res 2019, 151, 215-225.

41

60. Rani, M.; Shim, W. J.; Jang, M.; Han, G. M.; Hong, S. H., Releasing of

hexabromocyclododecanes from expanded polystyrenes in seawater -field and

laboratory experiments. Chemosphere 2017, 185, 798-805.

61. Paul-Pont, I.; Tallec, K.; Gonzalez-Fernandez, C.; Lambert, C.; Vincent, D.;

Mazurais, D.; Zambonino-Infante, J. L.; Brotons, G.; Lagarde, F.; Fabioux, C.;

Soudant, P.; Huvet, A., Constraints and Priorities for Conducting Experimental

Exposures of Marine Organisms to Microplastics. Front Mar Sci 2018, 5, 22.

62. Martin, T. M., A framework for an alternatives assessment dashboard for

evaluating chemical alternatives applied to flame retardants for electronic

applications. Clean technologies and environmental policy 2017, 19, (4), 1067-

1086.

63. An Alternatives Assessment For the Flame Retardant Decabromodiphenyl

Ether (DecaBDE); US Environmental Protection Agency: 2014.

64. Ishizaka, A.; Nemery, P., Multi-criteria decision analysis: methods and

software. John Wiley & Sons: 2013.

65. Cinelli, M.; Coles, S. R.; Kirwan, K., Analysis of the potentials of multi criteria

decision analysis methods to conduct sustainability assessment. Ecol. Indicators

2014, 46, 138-148.

66. Janssen, R., On the use of multi‐criteria analysis in environmental impact

assessment in The Netherlands. J. Mult. Crit. Anal. 2001, 10, (2), 101-109.

67. Hermann, B.; Kroeze, C.; Jawjit, W., Assessing environmental performance

by combining life cycle assessment, multi-criteria analysis and environmental

performance indicators. J. Cleaner Prod. 2007, 15, (18), 1787-1796.

68. Malloy, T. F.; Sinsheimer, P. J.; Blake, A.; Linkov, I., Use of multi‐criteria

decision analysis in regulatory alternatives analysis: A case study of lead free

solder. Integr. Environ. Assess. Manag. 2013, 9, (4), 652-664.

69. Linkov, I.; Satterstrom, F. K.; Steevens, J.; Ferguson, E.; Pleus, R. C., Multi-

criteria decision analysis and environmental risk assessment for nanomaterials.

J. Nanopart. Res. 2007, 9, (4), 543-554.

70. Linkov, I.; Satterstrom, F. K.; Kiker, G.; Batchelor, C.; Bridges, T.; Ferguson,

E., From comparative risk assessment to multi-criteria decision analysis and

adaptive management: Recent developments and applications. Environ. Int.

2006, 32, (8), 1072-1093.

42

71. Kiker, G. A.; Bridges, T. S.; Varghese, A.; Seager, T. P.; Linkov, I., Application

of multicriteria decision analysis in environmental decision making. Integr.

Environ. Assess. Manag. 2005, 1, (2), 95-108.

72. Rowley, H. V.; Peters, G. M.; Lundie, S.; Moore, S. J., Aggregating

sustainability indicators: beyond the weighted sum. J. Environ. Manage. 2012,

111, 24-33.

73. Von Winterfeldt, D.; Fischer, G. W., Multi-attribute utility theory: Models

and assessment procedures. In Utility, probability, and human decision making,

Springer: 1975; pp 47-85.

74. Roy, B., ELECTRE III: Un algorithme de classement fondé sur une

représentation floue des préférences en présence de critères multiples. Cahiers

du CERO 1978, 20, (1), 3-24.

75. Clean Production Action. GreenScreen Transformation Product Worksheet

& Resources.

https://www.greenscreenchemicals.org/images/ee_images/uploads/resources/

GreeScreen1.4-Annex9-1.18.pdf

76. Lavoie, E. T.; Heine, L. G.; Holder, H.; Rossi, M. S.; Lee, R. E.; Connor, E. A.;

Vrabel, M. A.; DiFiore, D. M.; Davies, C. L., Chemical Alternatives Assessment:

Enabling Substitution to Safer Chemicals. Environ. Sci. Technol. 2010, 44, (24),

9244-9249.

77. Bradley, P. M.; Chapelle, F. H., Anaerobic Mineralization of Vinyl Chloride

in Fe(III)-Reducing, Aquifer Sediments. Environ. Sci. Technol. 1996, 30, (6),

2084-2086.

78. Cui, N.; Zhang, X.; Xie, Q.; Wang, S.; Chen, J.; Huang, L.; Qiao, X.; Li, X.;

Cai, X., Toxicity profile of labile preservative bronopol in water: the role of more

persistent and toxic transformation products. Environ. Pollut. 2011, 159, (2),

609-615.

79. Donner, E.; Kosjek, T.; Qualmann, S.; Kusk, K. O.; Heath, E.; Revitt, D. M.;

Ledin, A.; Andersen, H. R., Ecotoxicity of carbamazepine and its UV photolysis

transformation products. Sci. Total Environ. 2013, 443, 870-876.

80. Tian, Z.; Zhao, H.; Peter, K. T.; Gonzalez, M.; Wetzel, J.; Wu, C.; Hu, X.; Prat,

J.; Mudrock, E.; Hettinger, R., A ubiquitous tire rubber–derived chemical

induces acute mortality in coho salmon. Science 2020.

43

81. Pakalin, S.; Cole, T.; Steinkellner, J.; Nicolas, R.; Tissier, C.; Munn, S.;

Eisenreich, S., Review on production processes of decabromodiphenyl ether

(decaBDE) used in polymeric applications in electrical and electronic equipment,

and assessment of the availability of potential alternatives to decaBDE. European

Commission, Directorate General Joint Research Center, European Chemicals

Bureau 2007.

82. ECHA Registered substances of the REACH registration dossiers,

Bis(pentabromophenyl) ether, PBT assessment.

https://echa.europa.eu/registration-dossier/-/registered-dossier/14217/2/3

83. Wolfe, K.; Pope, N.; Parmar, R.; Galvin, M.; Stevens, C.; Weber, E.; Flaishans,

J.; Purucker, T., Chemical transformation system: cloud based cheminformatic

services to support integrated environmental modeling. 2016.

84. CTS: Chemical Transformation Simulator. https://qed.epacdx.net/cts/

85. Dimitrov, S.; Diderich, R.; Sobanski, T.; Pavlov, T.; Chankov, G.; Chapkanov,

A.; Karakolev, Y.; Temelkov, S.; Vasilev, R.; Gerova, K., QSAR Toolbox–Workflow

and major functionalities. SAR QSAR Environ. Res. 2016, 27, (3), 203-219.

86. Cruciani, G.; Carosati, E.; De Boeck, B.; Ethirajulu, K.; Mackie, C.; Howe, T.;

Vianello, R., MetaSite:  Understanding Metabolism in Human Cytochromes from

the Perspective of the Chemist. J. Med. Chem. 2005, 48, (22), 6970-6979.

87. Stardrop, Optibrium Ltd., Cambridge, UK. In.

88. Ott, M. A.; Toy, R. L.; Button, W. G., An Approach toward the Prediction of

Chemical Degradation Pathways. Lhasa Limited, 22-23.

89. Marchant, C. A.; Briggs, K. A.; Long, A., In silico tools for sharing data and

knowledge on toxicity and metabolism: Derek for windows, meteor, and vitic.

Toxicol. Mech. Methods 2008, 18, (2-3), 177-187.

90. Djoumbou-Feunang, Y.; Fiamoncini, J.; Gil-de-la-Fuente, A.; Greiner, R.;

Manach, C.; Wishart, D. S., BioTransformer: a comprehensive computational tool

for small molecule metabolism prediction and metabolite identification. J.

Cheminform. 2019, 11, (1), 2.

91. Gao, J.; Ellis, L. B. M.; Wackett, L. P., The University of Minnesota

Biocatalysis/Biodegradation Database: improving public access. Nucleic Acids

Research 2009, 38, (suppl_1), D488-D491.

44

92. ECHA Registered substances of the REACH registration dossiers.

https://echa.europa.eu/information-on-chemicals/registered-substances

93. Regulation, E., No 1907/2006 of the European Parliament and of the Council

of 18 December 2006 concerning the Registration. Evaluation, Authorisation

and Restriction of Chemicals (REACH), establishing a European Chemicals

Agency, amending Directive 1999, 45, 1-849.

94. Estimation Programs Interface Suite™ for Microsoft® Windows, v 4.11;

United States Environmental Protection Agency, Washington, DC, USA: 2012.

95. Benfenati, E.; Manganaro, A.; Gini, G. C. VEGA-QSAR: AI Inside a Platform

for Predictive Toxicology, Proceedings of Workshop PAI: Turin Italy, 2013.

96. Martin, T. Toxicity Estimation Software Tool (TEST), U.S, Environmental

Protection Agency: Washington D.C., 2016.

97. Mansouri, K.; Abdelaziz, A.; Rybacka, A.; Roncaglioni, A.; Tropsha, A.;

Varnek, A.; Zakharov, A.; Worth, A.; Richard, A. M.; Grulke, C. M., CERAPP:

Collaborative estrogen receptor activity prediction project. Environ. Health

Perspect. 2016, 124, (7), 1023.

98. Sandholm, A.; Emanuelsson, B.-M.; Wehler, E. K., Bioavailability and half-

life of decabromodiphenyl ether (BDE-209) in rat. Xenobiotica 2003, 33, (11),

1149-1158.

99. Van der Ven, L. T.; van de Kuil, T.; Leonards, P. E.; Slob, W.; Cantón, R. F.;

Germer, S.; Visser, T. J.; Litens, S.; Håkansson, H.; Schrenk, D., A 28-day oral

dose toxicity study in Wistar rats enhanced to detect endocrine effects of

decabromodiphenyl ether (decaBDE). Toxicol. Lett. 2008, 179, (1), 6-14.

100. Mörck, A.; Hakk, H.; Örn, U.; Wehler, E. K., Decabromodiphenyl ether in the

rat: absorption, distribution, metabolism, and excretion. Drug Metab.

Disposition 2003, 31, (7), 900-907.

101. Deng, D.; Liu, J.; Xu, M.; Zheng, G.; Guo, J.; Sun, G., Uptake, translocation

and metabolism of decabromodiphenyl ether (BDE-209) in seven aquatic plants.

Chemosphere 2016, 152, 360-368.

102. Draft Screening Assessment of Certain Organic Flame Retardants

Substance Grouping: 1,3,5-Triazine-2,4,6-triamine (Melamine); Environment

and Climate Change Canada, Health Canada: Canada, 2016.

45

103. Shelton, D. R.; Karns, J. S.; McCarty, G. W.; Durham, D. R., Metabolism of

Melamine by Klebsiella terragena. Appl. Environ. Microbiol. 1997, 63, (7), 2832-

2835.

104. Crews, G. M.; Ripperger, W.; Kersebohm, D. B.; Güthner, T.; Mertschenk, B.,

Melamine and guanamines. Ullmann's Encyclopedia of Industrial Chemistry

2000.

105. Bai, X.-y.; Lu, S.-y.; Xie, L.; Zhang, B.; Song, S.-m.; He, Y.; Ouyang, J.-p.;

Zhang, T., A pilot study of metabolites of organophosphorus flame retardants in

paired maternal urine and amniotic fluid samples: potential exposure risks of

tributyl phosphate to pregnant women. Environ. Sci. Process. Impacts 2019, 21,

(1), 124-132.

106. Zhang, B.; Lu, S.; Huang, M.; Zhou, M.; Zhou, Z.; Zheng, H.; Jiang, Y.; Bai,

X.; Zhang, T., Urinary metabolites of organophosphate flame retardants in 0–5-

year-old children: Potential exposure risk for inpatients and home-stay infants.

Environ. Pollut. 2018, 243, 318-325.

107. Carignan, C. C.; Mínguez-Alarcón, L.; Butt, C. M.; Williams, P. L.; Meeker, J.

D.; Stapleton, H. M.; Toth, T. L.; Ford, J. B.; Hauser, R.; Team, E. S., Urinary

concentrations of organophosphate flame retardant metabolites and pregnancy

outcomes among women undergoing in vitro fertilization. Environ. Health

Perspect. 2017, 125, (8), 087018.

108. Wang, G.; Du, Z.; Chen, H.; Su, Y.; Gao, S.; Mao, L., Tissue-specific

accumulation, depuration, and transformation of triphenyl phosphate (TPHP) in

adult zebrafish (Danio rerio). Environ. Sci. Technol. 2016, 50, (24), 13555-13564.

109. Sasaki, K.; Suzuki, T.; Takeda, M.; Uchiyama, M., Metabolism of phosphoric

acid triesters by rat liver homogenate. Bull. Environ. Contam. Toxicol. 1984, 33,

(1), 281-288.

110. Springer, C.; Dere, E.; Hall, S. J.; McDonnell, E. V.; Roberts, S. C.; Butt, C.

M.; Stapleton, H. M.; Watkins, D. J.; McClean, M. D.; Webster, T. F., Rodent

thyroid, liver, and fetal testis toxicity of the monoester metabolite of bis-(2-

ethylhexyl) tetrabromophthalate (TBPH), a novel brominated flame retardant

present in indoor dust. Environ. Health Perspect. 2012, 120, (12), 1711-1719.

111. Kajiwara, N.; Noma, Y.; Takigami, H., Photolysis studies of technical

decabromodiphenyl ether (DecaBDE) and ethane (DeBDethane) in plastics under

natural sunlight. Environ. Sci. Technol. 2008, 42, (12), 4404-4409.

46

112. Ecological State of the Science Report on Decabromodiphenyl Ether

(decaBDE): Bioaccumulation and Transformation; Environment Canada 2010.

113. Mas, S.; de Juan, A.; Lacorte, S.; Tauler, R., Photodegradation study of

decabromodiphenyl ether by UV spectrophotometry and a hybrid hard-and soft-

modelling approach. Anal. Chim. Acta 2008, 618, (1), 18-28.

114. Söderström, G.; Sellström, U.; de Wit, C. A.; Tysklind, M., Photolytic

debromination of decabromodiphenyl ether (BDE 209). Environ. Sci. Technol.

2004, 38, (1), 127-132.

115. Wu, Z.; Li, D.; Meng, J.; Wang, H., Introduction to SIMCA-P and Its

Application. In Handbook of Partial Least Squares: Concepts, Methods and

Applications, Esposito Vinzi, V.; Chin, W. W.; Henseler, J.; Wang, H., Eds.

Springer Berlin Heidelberg: Berlin, Heidelberg, 2010; pp 757-774.

116. Molecular Operating Environment (MOE), C. C. G. U., 1010 Sherbooke St.

West, Suite #910, Montreal, QC, Canada, H3A 2R7, 2018.

117. Rannar, S.; Andersson, P. L., A novel approach using hierarchical clustering

to select industrial chemicals for environmental impact assessment. J. Chem. Inf.

Model. 2010, 50, (1), 30-36.

118. Sushko, I.; Novotarskyi, S.; Körner, R.; Pandey, A. K.; Rupp, M.; Teetz, W.;

Brandmaier, S.; Abdelaziz, A.; Prokopenko, V. V.; Tanchuk, V. Y.; Todeschini, R.;

Varnek, A.; Marcou, G.; Ertl, P.; Potemkin, V.; Grishina, M.; Gasteiger, J.; Schwab,

C.; Baskin, I. I.; Palyulin, V. A.; Radchenko, E. V.; Welsh, W. J.; Kholodovych, V.;

Chekmarev, D.; Cherkasov, A.; Aires-de-Sousa, J.; Zhang, Q.-Y.; Bender, A.;

Nigsch, F.; Patiny, L.; Williams, A.; Tkachenko, V.; Tetko, I. V., Online chemical

modeling environment (OCHEM): web platform for data storage, model

development and publishing of chemical information. J. Comput. Aided Mol. Des.

2011, 25, (6), 533-554.

119. Tetko, I. V., Neural network studies. 4. Introduction to associative neural

networks. J. Chem. Inf. Comput. Sci. 2002, 42, (3), 717-728.

120. srl, T., Dragon (Software for Molecular Descriptor Calculation) Version 6.0.

2015.

121. Mathieu, D., Pencil and Paper Estimation of Hansen Solubility Parameters.

ACS omega 2018, 3, (12), 17049-17056.

47

122. El-Zein, A.; Tonmoy, F. N., Assessment of vulnerability to climate change

using a multi-criteria outranking approach with application to heat stress in

Sydney. Ecol. Indicators 2015, 48, 207-217.

123. Figueira, J.; Greco, S.; Ehrgott, M., Multiple criteria decision analysis: state

of the art, International Series in Operations Research Management Science,

vol. 78. Springer Verlag: Boston, 2005.

124. Buttry, D. A.; Ward, M. D., Measurement of interfacial processes at electrode

surfaces with the electrochemical quartz crystal microbalance. Chemical Reviews

1992, 92, (6), 1355-1379.

125. Arp, H. P. H. & Hale, S. E. REACH: Improvement of guidance methods for

the identification and evaluation of PM/PMT substances. UBA TEXTE 126/2019.

German Environment Agency (UBA), Dessau-Roßlau, Germany. ISBN: 1862-

4804. 130 pages.

126. Stieger, G.; Scheringer, M.; Ng, C. A.; Hungerbühler, K., Assessing the

persistence, bioaccumulation potential and toxicity of brominated flame

retardants: Data availability and quality for 36 alternative brominated flame

retardants. Chemosphere 2014, 116, 118-123.

127. Ylä-Mononen, L. ECHA’s New Strategy to Ensure Data Quality in REACH

Registrations. https://www.bfr.bund.de/cm/349/echas-new-strategy-to-

ensure-data-quality-in-reach-registrations.pdf

128. REACH data quality comes under renewed fire.

https://chemicalwatch.com/57000/reach-data-quality-comes-under-renewed-

fire

129. USEPA Design for the Environment Alternatives Assessments.

https://www.epa.gov/saferchoice/design-environment-alternatives-

assessments

130. Chinen, K.; Malloy, T., QSAR Use in REACH Analyses of Alternatives to

Predict Human Health and Environmental Toxicity of Alternative Chemical

Substances. Integr. Environ. Assess. Manag. 2020, 16, (5), 745-760.

131. Cappelli, C.; Cassano, A.; Golbamaki, A.; Moggio, Y.; Lombardo, A.;

Colafranceschi, M.; Benfenati, E., Assessment of in silico models for acute aquatic

toxicity towards fish under REACH regulation. SAR QSAR Environ. Res. 2015,

26, (12), 977-999.

48

132. Lunghini, F.; Marcou, G.; Azam, P.; Enrici, M.; Van Miert, E.; Varnek, A.,

Consensus QSAR models estimating acute toxicity to aquatic organisms from

different trophic levels: algae, Daphnia and fish. SAR QSAR Environ. Res. 2020,

31, (9), 655-675.

133. Fenner, K.; Gao, J.; Kramer, S.; Ellis, L.; Wackett, L., Data-driven extraction

of relative reasoning rules to limit combinatorial explosion in biodegradation

pathway prediction. Bioinformatics 2008, 24, (18), 2079-2085.

134. Escher, B. I.; Fenner, K., Recent advances in environmental risk assessment

of transformation products. Environ. Sci. Technol. 2011, 45, (9), 3835-3847.

135. Ng, C. A.; Scheringer, M.; Fenner, K.; Hungerbuhler, K., A Framework for

Evaluating the Contribution of Transformation Products to Chemical Persistence

in the Environment. Environ. Sci. Technol. 2011, 45, (1), 111-117.

136. Fenner, K.; Scheringer, M.; Hungerbühler, K., Joint persistence of

transformation products in chemicals assessment: case studies and uncertainty

analysis. Risk Analysis: AN INTERNATIONAL JOURNAL 2003, 23, (1), 35-53.

137. Fenner, K.; Schenker, U.; Scheringer, M., Modelling environmental exposure

to transformation products of organic chemicals. 2008.

138. Wicker, J.; Fenner, K.; Ellis, L.; Wackett, L.; Kramer, S., Predicting

biodegradation products and pathways: a hybrid knowledge-and machine

learning-based approach. Bioinformatics 2010, 26, (6), 814-821.

139. Clemen, R. T.; Reilly, T., Making hard decisions with DecisionTools.

Cengage Learning: 2013.

140. Beaudrie, C.; Corbett, C. J.; Lewandowski, T. A.; Malloy, T.; Zhou, X.,

Evaluating the Application of Decision Analysis Methods in Simulated

Alternatives Assessment Case Studies: Potential Benefits and Challenges of Using

MCDA. Integr. Environ. Assess. Manag. 2020.

141. Action, C. P., GreenScreen for safer chemicals hazard assessment guidance.

In Version: 2016.

142. Boxall, A. B.; Sinclair, C. J.; Fenner, K.; Kolpin, D.; Maund, S. J., Peer

Reviewed: When Synthetic Chemicals Degrade in the Environment. Environ. Sci.

Technol. 2004, 38, (19), 368A-375A.

49

143. Mohamed Nor, N. H.; Koelmans, A. A., Transfer of PCBs from microplastics

under simulated gut fluid conditions is biphasic and reversible. Environ. Sci.

Technol. 2019, 53, (4), 1874-1883.

144. Wells, M.; Wick, L. Y.; Harms, H., Model polymer release system study of

PAH bioaccessibility: the relationship between “rapid” release and

bioaccessibility. Environ. Sci. Technol. 2005, 39, (4), 1055-1063.

145. Wells, M.; Wick, L. Y.; Harms, H., Perspectives on modeling the release of

hydrophobic organic contaminants drawn from model polymer release systems.

J. Mater. Chem. 2004, 14, (15), 2461-2472.

146. Sarmah, A. K.; Rohan, M., Evaluation of four mathematical models to

describe dissipation kinetics of 4-n-nonylphenol and bisphenol-A in

groundwater–aquifer material slurry. J. Environ. Monit. 2011, 13, (1), 157-166.

147. Tavares, J. K.; de Souza, A. A. U.; de Oliveira, J. V.; Priamo, W. L.; de Souza,

S. M. G. U., Modeling of the controlled release of betacarotene into anhydrous

ethanol from microcapsules. OpenNano 2016, 1, 25-35.

148. van Boekel, M. A., On the use of the Weibull model to describe thermal

inactivation of microbial vegetative cells. Int. J. Food Microbiol. 2002, 74, (1-2),

139-159.

149. Lee, H.; Byun, D.-E.; Kim, J. M.; Kwon, J.-H., Desorption modeling of

hydrophobic organic chemicals from plastic sheets using experimentally

determined diffusion coefficients in plastics. Mar. Pollut. Bull. 2018, 126, 312-

317.

150. Adams, W. A.; Xu, Y.; Little, J. C.; Fristachi, A. F.; Rice, G. E.; Impellitteri, C.

A., Predicting the migration rate of dialkyl organotins from PVC pipe into water.

Environ. Sci. Technol. 2011, 45, (16), 6902-6907.

151. Sun, B.; Hu, Y.; Cheng, H.; Tao, S., Kinetics of brominated flame retardant

(BFR) releases from granules of waste plastics. Environ. Sci. Technol. 2016, 50,

(24), 13419-13427.

152. Lao, W.; Hong, Y.; Tsukada, D.; Maruya, K. A.; Gan, J., A new film-based

passive sampler for moderately hydrophobic organic compounds. Environ. Sci.

Technol. 2016, 50, (24), 13470-13476.

153. Kerr, B. A.; McDonough, C.; Lohmann, R., Determination and estimation of

polyethylene-water equilibrium partition coefficients for organophosphate flame

50

retardants and current-use pesticides. SURFO Technical Report No. 15-01 2015,

43.

154. Pitt, C. G.; Bao, Y. T.; Andrady, A. L.; Samuel, P. N., The correlation of

polymer-water and octanol-water partition coefficients: estimation of drug

solubilities in polymers. Int. J. Pharm. 1988, 45, (1-2), 1-11.

155. Smedes, F.; Geertsma, R. W.; Zande, T. v. d.; Booij, K., Polymer− water

partition coefficients of hydrophobic compounds for passive sampling:

Application of cosolvent models for validation. Environ. Sci. Technol. 2009, 43,

(18), 7047-7054.

156. Fischer, F. C.; Cirpka, O. A.; Goss, K.-U.; Henneberger, L.; Escher, B. I.,

Application of experimental polystyrene partition constants and diffusion

coefficients to predict the sorption of neutral organic chemicals to multiwell

plates in in vivo and in vitro bioassays. Environ. Sci. Technol. 2018, 52, (22),

13511-13522.

157. Pintado‐Herrera, M. G.; Lara‐Martín, P. A.; González‐Mazo, E.; Allan, I. J.,

Determination of silicone rubber and low‐density polyethylene diffusion and

polymer/water partition coefficients for emerging contaminants. Environ.

Toxicol. Chem. 2016, 35, (9), 2162-2172.

158. Liu, X.; Allen, M. R.; Roache, N. F., Characterization of organophosphorus

flame retardants’ sorption on building materials and consumer products. Atmos.

Environ. 2016, 140, 333-341.

159. Seidensticker, S.; Zarfl, C.; Cirpka, O. A.; Fellenberg, G.; Grathwohl, P., Shift

in mass transfer of wastewater contaminants from microplastics in the presence

of dissolved substances. Environ. Sci. Technol. 2017, 51, (21), 12254-12263.

160. Hale, S. E.; Cornelissen, G.; Arp, H. P. H., Comment on “Partition

coefficients of organic contaminants with carbohydrates”. Environ. Sci. Technol.

2011, 45, (3), 1158-1158.

161. The EU's chemicals strategy for sustainability towards a toxic-free

environment. https://ec.europa.eu/environment/strategy/chemicals-

strategy_en

162. He, M.-J.; Luo, X.-J.; Chen, M.-Y.; Sun, Y.-X.; Chen, S.-J.; Mai, B.-X.,

Bioaccumulation of polybrominated diphenyl ethers and decabromodiphenyl

ethane in fish from a river system in a highly industrialized area, South China. Sci.

Total Environ. 2012, 419, 109-115.

51

163. Arnot, J. A.; Gobas, F. A., A review of bioconcentration factor (BCF) and

bioaccumulation factor (BAF) assessments for organic chemicals in aquatic

organisms. Environ. Rev. 2006, 14, (4), 257-297.

164. Verougstraete, V., Risk Management of Complex Inorganic Materials: A

Practical Guide. Academic Press: 2017.

165. USEPA Interpretive Assistance Document for Assessment of Polymers;

2013.

166. Australian Gorvernment, D. o. H. Polymer assessment methodology.

https://www.nicnas.gov.au/chemical-information/imap-assessments/how-

chemicals-are-assessed/polymer-assessment-methodology

167. Sanderson, H.; Khan, K.; Hansen, A. M. B.; Connors, K.; Lam, M. W.; Roy,

K.; Belanger, S., Environmental Toxicity (Q) SARs for Polymers as an Emerging

Class of Materials in Regulatory Frameworks, with a Focus on Challenges and

Possibilities Regarding Cationic Polymers. In Ecotoxicological QSARs, Springer:

2020; pp 681-705.

168. GaBi, 10.0; Sphera Solutions GmbH.

169. Rosenbaum, R. K.; Bachmann, T. M.; Gold, L. S.; Huijbregts, M. A.; Jolliet,

O.; Juraske, R.; Koehler, A.; Larsen, H. F.; MacLeod, M.; Margni, M., USEtox—

the UNEP-SETAC toxicity model: recommended characterisation factors for

human toxicity and freshwater ecotoxicity in life cycle impact assessment. The

International Journal of Life Cycle Assessment 2008, 13, (7), 532.

170. Roos, S.; Peters, G. M., Three methods for strategic product toxicity

assessment—the case of the cotton T-shirt. The International Journal of Life

Cycle Assessment 2015, 20, (7), 903-912.

171. Roos, S.; Holmquist, H.; Jönsson, C.; Arvidsson, R., USEtox characterisation

factors for textile chemicals based on a transparent data source selection strategy.

The International Journal of Life Cycle Assessment 2018, 23, (4), 890-903.