EXPERIMENTAL AND LIFE-CYCLE INVESTIGATION OF NONSTEROIDAL ANTI-INFLAMMATORY DRUG REMOVAL IN SOURCE SEPARATED URINE

By

KELLY ANN LANDRY

A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT

OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY

UNIVERSITY OF FLORIDA

2017

© 2017 Kelly Ann Landry

To my husband, for everything

4

ACKNOWLEDGMENTS

I would like to thank Dr. Treavor Boyer for his mentorship, encouragement, and

support throughout my undergraduate and graduate career, and inspiring excellent

dinner conversation related to all things urine. I also thank my committee members: Dr.

Paul Chadik for his inspiring lectures throughout my undergraduate studies which have

contributed to my passion in environmental engineering, and for his continued support

throughout my graduate studies, Dr. Robert Ries for providing valuable insight and

expertise on Life Cycle Assessment, Dr. Nancy Denslow for supporting my endeavors

into ecotoxicology and providing laboratory access to conduct my experiments, and Dr.

Guenther Hochhaus for generously providing access to his laboratory space and

analytical instruments.

I also extend my gratitude to several groups and individuals for their help: Dr.

Ching-Hua Huang and Dr. Peizhe Sun at the Georgia Institute of Technology, the UF

Physical Plant Department and UF Water Reclamation Facility, Kevin Kroll, and Dr.

Hochhaus’ Research Group. This material is based upon work supported by the

National Science Foundation Graduate Fellowship under Grant No. DGE-1315138, the

National Science Foundation CAREER grant under Grant No. CBET-1150790, and the

UF Graduate School Fellowship.

My graduate career would not have been as successful if it weren’t for the

encouragement and support of my colleagues, friends, and family. I thank all of the

wonderful Boyer Research Group members for providing endless entertainment,

commiseration, and lifelong friendship. I would not have maintained my sanity in my

pursuit of a PhD if it weren’t for my closest friends who provided laughter over many

shared bottles of wine and cheese. I also am grateful for my mom, dad, Nanna, and

5

Rob for their endless love and encouragement. I thank Chewie for his unyielding

support and snuggles. I am most thankful for my husband, Tyler, for providing

emotional, intellectual, and nutritional support throughout my graduate studies. I look

forward to this next chapter in life with you by my side. I am forever grateful for your

unconditional love, and I love you.

6

TABLE OF CONTENTS

page

ACKNOWLEDGMENTS .................................................................................................. 4

LIST OF TABLES ............................................................................................................ 9

LIST OF FIGURES ........................................................................................................ 12

LIST OF OBJECTS ....................................................................................................... 15

LIST OF ABBREVIATIONS ........................................................................................... 16

ABSTRACT ................................................................................................................... 18

CHAPTER

1 INTRODUCTION .................................................................................................... 20

Pharmaceuticals and Nutrients in the Environment ................................................ 20 Urine Source Separation ......................................................................................... 23 Nonsteroidal Anti-Inflammatory Drugs .................................................................... 25

Organization of Dissertation .................................................................................... 26

2 ION-EXCHANGE SELECTIVITY OF DICLOFENAC, IBUPROFEN, KETOPROFEN, AND NAPROXEN IN UREOLYZED HUMAN URINE ................... 29

Application of Sorption Processes for Pharmaceutical Removal ............................ 29

Experimental Methods ............................................................................................ 32 Synthetic Human Urine .................................................................................... 32 Pharmaceuticals in Urine .................................................................................. 33

Anion Exchange Resin ..................................................................................... 33 Batch Equilibrium Tests .................................................................................... 34

Column Tests ................................................................................................... 34 Analytical Methods ........................................................................................... 35 Data Analysis ................................................................................................... 36

Isotherm Models ............................................................................................... 36 Results and Discussion........................................................................................... 37

Ion-Exchange of Individual Pharmaceuticals at Realistic Concentrations ........ 37

Effect of Pharmaceutical Properties ................................................................. 41

Effect of Urine Composition .............................................................................. 46 Effect of Multiple Pharmaceuticals ................................................................... 47 Column Studies ................................................................................................ 49 Practical Application and Future Work ............................................................. 52

Concluding Remarks............................................................................................... 53

7

3 FIXED BED MODELING OF NONSTEROIDAL ANTI-INFLAMMATORY DRUG REMOVAL BY ION-EXCHANGE IN SOURCE SEPARATED URINE: MASS REMOVAL OR TOXICITY REDUCTION? .............................................................. 60

Application of Bioassays and Modeling to Assess Pharmaceutical Ecotoxicity ...... 60 Materials and Methods............................................................................................ 64

Pharmaceutical and Pharmaceutical Metabolites ............................................. 64 Synthetic and Real Urine .................................................................................. 64 Anion Exchange Resin ..................................................................................... 65

Pharmaceutical Concentrations in Urine .......................................................... 65 Toxicity Bioassays ............................................................................................ 65 Batch Kinetic and Equilibrium Tests ................................................................. 66 Fixed-Bed Column Modeling ............................................................................ 67

Sample Preparation .......................................................................................... 67 Analytical Methods ........................................................................................... 67

Data Analysis ................................................................................................... 68 Results and Discussion........................................................................................... 68

COX-1 Inhibition for Individual Compounds ...................................................... 68 COX-1 Inhibition Mixture Effects ...................................................................... 71 Comparison of Urine Matrices .......................................................................... 75

Concluding Remarks............................................................................................... 79

4 LIFE CYCLE ASSESSMENT AND COSTING OF URINE SOURCE SEPARATION: FOCUS ON NONSTEROIDAL ANTI-INFLAMMATORY DRUG REMOVAL .............................................................................................................. 87

Application of Life Cycle Assessment for Pharmaceutical Treatment ..................... 87 Life Cycle Model ..................................................................................................... 89

Scope of the Study ........................................................................................... 89

Life Cycle Inventory .......................................................................................... 92 Life Cycle Costing ............................................................................................ 93

Life Cycle Impact Assessment ......................................................................... 93 Sensitivity and Uncertainty Analysis ................................................................. 94

Results and discussion ........................................................................................... 95

Overall Comparison of Scenarios ..................................................................... 95 Urine Source Separation .................................................................................. 98 Pharmaceutical Toxicity.................................................................................. 101 Model Sensitivity ............................................................................................ 104

Concluding Remarks............................................................................................. 107

5 CONCLUSIONS ................................................................................................... 113

APPENDIX

A SUPPLEMENTARY INFORMATION FOR CHAPTER 2 ....................................... 118

B SUPPLEMENTARY INFORMATION FOR CHAPTER 3 ....................................... 140

8

C SUPPLEMENTARY INFORMATION FOR CHAPTER 4 ....................................... 170

LIST OF REFERENCES ............................................................................................. 218

BIOGRAPHICAL SKETCH .......................................................................................... 238

9

LIST OF TABLES

Table page 2-1 Composition of synthetic fresh and ureolyzed urine used in ion-exchange

experiments. ....................................................................................................... 55

2-2 Continuous-flow column ion-exchange of DCF, IBP, KTP, and NPX onto Dowex 22 AER followed by in-column regeneration over three treatment–regeneration cycles. ........................................................................................... 56

3-1 Estimated active ingredient (AI) and metabolite concentrations in urine and fraction excreted in urine. ................................................................................... 81

4-1 Capital and operation and management (O&M) costs, net present value (NPV) for each urine treatment scenario .......................................................... 109

A-1 Properties of pharmaceuticals used in ion-exchange experiments. .................. 121

A-2 Estimated and measured pharmaceutical concentrations in urine based on previous literature. ............................................................................................ 122

A-3 Properties of strong-base, anion exchange polymer resins. ............................. 123

A-4 Linear form of isotherm models and plots to determine estimated initial values for non-linear isotherm modeling parameters. ....................................... 124

A-5 Individual equilibrium experiment isotherm parameters for Dowex 22 AER sorption ............................................................................................................. 125

A-6 Isotherm parameters, sum of squares errors (SSE), correlation coefficients (R2), and average relative errors (ARE) of the Freundlich, Langmuir, Dubinin-Astakhov, and Dubinin-Radushkevich models determined by nonlinear regression for the different ion-exchange resins used to remove diclofenac (Co = 0.2 mmol/L) in ureolyzed urine. .............................................. 126

A-7 Estimated physicochemical parameters of the four major diclofenac metabolites. ...................................................................................................... 127

A-8 Equilibrium experiment isotherm parameters for Dowex 22 AER sorption of ibuprofen (C0 = 0.2 mmol/L) present in fresh urine ........................................... 128

A-9 Combined equilibrium experiment isotherm parameters for Dowex 22 AER sorption. ............................................................................................................ 129

A-10 Analysis of covariance (ANOCOVA) test results to determine whether there was a significant difference at the 95% confidence interval (α = 0.05)

10

between DCF, IBP, KTP, and NPX ion-exchange when present individually or combined in synthetic ureolyzed urine ......................................................... 130

B-1 Active ingredient and metabolite structure and chemical properties. ................ 145

B-2 Synthetic ureolyzed urine composition adapted from Landry et al. (2015). ...... 146

B-3 Estimated and measured pharmaceutical concentrations in urine from literature. .......................................................................................................... 147

B-4 Pharmaceutical dose-response concentrations used to evaluate COX-1 inhibition of single compounds.......................................................................... 148

B-5 Pharmaceutical dose-response concentrations used to evaluate COX-1 inhibition of the pharmaceutical mixture. .......................................................... 149

B-6 Nomenclature used to calculate liquid-phase mass transfer coefficient, liquid-phase diffusion coefficient, and surface diffusion coefficient. ........................... 150

B-7 Urine properties assumed to be equivalent to water at 25°C. ........................... 151

B-8 Molar volume (Vb), liquid diffusivity (DL), and liquid-phase mass transfer coefficients (kL). ................................................................................................ 152

B-9 Surface diffusion coefficient (Ds). ...................................................................... 153

B-10 Column operational parameters. ...................................................................... 154

B-11 Resin properties. .............................................................................................. 155

B-12 Freundlich isotherm parameters. ...................................................................... 156

B-13 Hill model parameters from the COX-1 inhibition bioassays ............................. 157

B-14 Alternative Hill model parameters from the COX-1 inhibition bioassays ........... 158

B-15 In vivo chronic toxicity data for organisms exposed to diclofenac, ibuprofen, naproxen, and ketoprofen. ................................................................................ 159

C-1 Average urination volumes and frequency for asymptomatic men and women. ............................................................................................................. 190

C-2 Total number of weekdays during the fall, spring, and summer semesters, excluding major holidays .................................................................................. 191

C-3 Estimated urine production for entire UF campus over different time periods. . 192

C-4 Daily refuse route distance (km) traveled during fall, spring, and summer semesters. ........................................................................................................ 193

11

C-5 Mass of diclofenac, ibuprofen, ketoprofen, and naproxen sorbed onto AER (mg) and desorbed from AER using a 5% NaCl, 50% methanol regeneration solution. ............................................................................................................ 194

C-6 Inventory data for ion-exchange vessel components. ....................................... 195

C-7 Inventory data for incineration of a regeneration brine at a cement kiln plant. .. 196

C-8 Recommended defined daily dose (DDD), fraction of dose excreted in urine as the parent compound (Fex), and estimated pharmaceutical concentrations in urine. ............................................................................................................. 197

C-9 Unit cost of inventory items. ............................................................................. 198

C-10 USEtox characterization factors (human toxicity in cases/kg and ecotoxicity in PAF·m3·day/kg) for diclofenac, ibuprofen, ketoprofen, and naproxen. ......... 199

C-11 Baseline, minimum, and maximum values used for various input parameter assumptions ..................................................................................................... 200

C-12 Baseline, minimum, and maximum values used for various cost assumptions 201

12

LIST OF FIGURES

Figure page 1-1 Visual representation of the urban life-cycle of pharmaceuticals and nutrients .. 28

2-1 Experimental equilibrium data and isotherm models determined by nonlinear regression ........................................................................................................... 57

2-2 Comparison of pharmaceutical removal when present individually or combined in ureolyzed urine ............................................................................... 58

2-3 Column saturation curves of Dowex 22 AER by pharmaceutical mixture ........... 59

3-1 Predicted column breakthrough curves as a function of mass removal and COX-1 inhibition ................................................................................................. 82

3-2 Cyclooxygenase subtype-1 inhibition curve for a pharmaceutical mixture containing diclofenac, ketoprofen, ketoprofen glucuronide, naproxen, and o-desmethylnaproxen. ........................................................................................... 83

3-3 Predicted column breakthrough curves as a function of mass removal and COX-1 inhibition for a pharmaceutical mixture ................................................... 84

3-4 Ion-exchange removal in real urine and synthetic urine with and without metabolites ......................................................................................................... 85

3-5 Mass of endogenous metabolites (TOC) removed (mg C) during equilibrium experiments for synthetic urine with metabolites and real urine ......................... 86

4-1 Treatment schematic for scenarios A–H (light gray horizontal arrows) and contributing processes ...................................................................................... 110

4-2 Normalized TRACI impact score for all scenarios ............................................ 111

4-3 Comparison of ecotoxicity impact (CTUe = PAF·m3·day) ................................. 112

5-1 Visual representation of the systematic approach for evaluating sorption materials to remove pharmaceuticals in source separated urine ...................... 117

A-1 Individual experimental data and sorption isotherms determined by nonlinear regression of paracetamol (PCM) using Dowex 22 anion exchange resin. ...... 131

A-2 Experimental data and isotherm models for naproxen and ketoprofen ............. 132

A-3 Experimental data and ion-exchange isotherms of diclofenac removal by various resins ................................................................................................... 133

13

A-4 Mole fraction distribution of the neutral and ionized species present in the octanol and water phase .................................................................................. 134

A-5 Combined pharmaceutical experimental data and sorption isotherms determined by nonlinear regression ................................................................. 135

A-6 Sorption by Dowex 22 anion exchange resin over three treatment cycles using fresh resin (Cycle 1) and regenerated resin (Cycles 2 and 3) in a continuous-flow mini-column ............................................................................ 136

A-7 Simultaneous column regeneration curves ....................................................... 137

B-1 Fixed bed ion-exchange removal of diclofenac by Dowex Marathon 11 fit to the homogenous surface diffusion model (HSDM). .......................................... 160

B-2 Fixed bed ion-exchange removal of diclofenac, ketoprofen, and naproxen in synthetic ureolyzed urine using Dowex 22 fit to the homogenous surface diffusion model (HSDM) ................................................................................... 161

B-3 Cyclooxygenase subtype-1 inhbition curves for diclofenac, ketoprofen, naproxen, and O-desmethylnaproxen .............................................................. 162

B-4 Alternative cyclooxygenase subtype-1 inhbition curves for naproxen, and O-desmethylnaproxen .......................................................................................... 163

B-5 Alternative predicted COX-1 inhibition as a function of bed volumes treated by fixed bed ion-exchange of naproxen, and O-desmethylnaproxen ................ 164

B-6 Cyclooxygenase subtype-1 inhbition curves for ibuprofen, OH-ibuprofen, 4’OH-diclofenac, and ketoprofen glucuronide ................................................... 165

B-7 ToxCast database in vitro bioassays for various endpoints plotted as a function of the concentration that induces 50% activity (AC50) ......................... 166

B-8 Predicted column breakthrough curves as a function of mass removal and COX-1 inhibition for diclofenac ion-exchange in real urine ............................... 167

B-9 Isotherm data for ion-exchange removal of diclofenac, ibuprofen, ketoprofen, naproxen, and O-desmethylnaproxen in synthetic urine with and without metabolites and real human urine (DCF only). ................................................. 168

B-10 Kinetic data for ion-exchange removal of diclofenac, ibuprofen, ketoprofen, and naproxen and O-desmethylnaproxen in synthetic urine with and without metabolites and real human urine (DCF only) .................................................. 169

C-1 Bench scale column results for removal of diclofenac, ibuprofen, ketoprofen, and naproxen by anion-exchange resin. ........................................................... 202

14

C-2 Manufacturer data and resulting linear regressions of fiberglass water softener tank ..................................................................................................... 203

C-3 Manufacturer data and resulting linear regressions of centrifugal pump power specifications .................................................................................................... 204

C-4 Relative frequency diagram of ibuprofen concentrations in urine for a community where 1–100% of the population is consuming ibuprofen (Fc) for 1–100% of the collection time (Fd). ................................................................... 205

C-5 Normalized TRACI impact score for centralized wastewater treatment and urine source separation. ................................................................................... 206

C-6 Comparison of ozone depletion impacts (kg CFC-11 eq.) ................................ 207

C-7 Comparison of global warming impacts (kg CO2 eq.) ...................................... 208

C-8 Comparison of smog impacts (kg O3 eq.) ........................................................ 209

C-9 Comparison of acidification impacts (kg SO2 eq.) ............................................ 210

C-10 Comparison of eutrophication impacts (kg N eq.). ............................................ 211

C-11 Comparison of carcinogenic impacts (CTUh). .................................................. 212

C-12 Comparison of respiratory effects impacts (kg PM2.5 eq.) ............................... 213

C-13 Comparison of fossil fuel depletion impacts (MJ surplus) ................................. 214

C-14 Impact assessment results for methanol, sodium chloride, and potable water production used in the regeneration process (positive (+) percent contributions), compared to CO2 and NOx emission offsets, heavy fuel offsets, and hard coal offsets from incineration of the regeneration brine at a cement kiln plant (negative (–) percent contributions) ...................................... 215

C-15 Normalized TRACI impact score (PE) of vacuum truck collection compared to the vacuum sewer collection as a function of vacuum sewer pipe length or distance traveled by vacuum truck (km). .......................................................... 216

C-16 Comparison of non-carcinogenic human toxicity impact (CTUh = number of disease cases). ................................................................................................. 217

15

LIST OF OBJECTS

Object page 4-1 Environmental impact and economic costing sensitivity analysis results .......... 107

16

LIST OF ABBREVIATIONS

AER Anion exchange resin

ANOCOVA Analysis of covariance

BV Bed volume

COX Cyclooxygenase

D-A Dubinin-Astakhov

DCF Diclofenac

D-R Dubinin-Radushkevich

EBCT Empty bed contact time

HSDM Homogenous surface diffusion model

IBP Ibuprofen

IC50 Concentration corresponding to 50% COX-1 inhibition

IC10 Concentration corresponding to 10% COX-1 inhibition

KTP Ketoprofen

KTP-gluc Ketoprofen glucuronide

LCA Life cycle assessment

LDF Linear driving force

N Nitrogen

NPX Naproxen

NSAID Nonsteroidal anti-inflammatory drug

Odm-NPX O-desmethylnaproxen

OH-DCF 4’-OH-diclofenac

OH-IBP Hydroxy ibuprofen

P Phosphorus

PCM Paracetamol

17

TN Total nitrogen

TP Total phosphorus

TRACI Tool for the Reduction and Assessment of Chemical and Other Environmental Impacts

WWTP Wastewater treatment plant

18

Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy

EXPERIMENTAL AND LIFE-CYCLE INVESTIGATION OF NONSTEROIDAL ANTI-

INFLAMMATORY DRUG REMOVAL IN SOURCE SEPARATED URINE

By

Kelly Ann Landry

May 2017

Chair: Treavor H. Boyer Major: Environmental Engineering Sciences

Treatment of source separated urine is one proposed method to effectively and

efficiently remove pharmaceuticals excreted in urine, such as nonsteroidal anti-

inflammatory drugs (NSAIDs), to reduce environmental loading. Furthermore, high

nitrogen and phosphorus content makes urine a valuable fertilizer alternative, thus it is

imperative that potential contaminants are removed prior to reuse. Ion-exchange has

the potential to selectively remove NSAIDs with minimal co-removal of nutrients.

Realizing the benefits of an emerging treatment process depends on understanding the

mechanisms of removal, process sustainability, and the ability to protect human and

environmental health.

The work presented here focuses on a systematic approach to evaluate sorption

processes (i.e., ion-exchange and adsorption) to remove pharmaceuticals in source

separated urine. Specifically, the removal of NSAIDs using anion-exchange resin

(AER). Ion-exchange selectivity and mechanisms of removal were elucidated to better

understand NSAID removal. The reduction in ecotoxicity potential was evaluated by

applying in vitro bioassays to the predicted fixed-bed removal. Lastly, life cycle

19

environmental impacts and economic costs of implementing urine source separation

and pharmaceutical removal in a university community were evaluated.

Results suggest that the ion-exchange selectivity of NSAIDs is influenced by

concerted electrostatic and van der Waals interactions between the acidic

pharmaceuticals and the AER. Pharmaceutical hydrophobicity may vary under fresh

and ureolyzed urine conditions, thereby influencing ion-exchange selectivity. The

homogenous surface diffusion model predicted diclofenac, ketoprofen, naproxen, and

O-desmethylnaproxen fixed-bed breakthrough performance. Dose-response

cyclooxygenase inhibition of diclofenac, ketoprofen, ketoprofen glucuronide, naproxen,

and O-desmethylnaproxen followed the generalized concentration addition model for

mixture toxicity. Evaluation of cyclooxygenase inhibition as a function of bed volume

found that complete mass removal may not be necessary to achieve a reduction in

toxicity potential. Furthermore, endogenous metabolites in urine competed for ion-

exchange sites on the resin suggesting that a resin with higher selectivity and/or

capacity may improve pharmaceutical removal in urine. Major benefits of urine source

separation at the community-scale include flush water savings, reduced electricity use

for wastewater treatment (WWT), and reduced nutrient loading. Building-level urine

treatment or collection by vacuum truck for centralized treatment had negligible cost

difference compared with WWT.

20

CHAPTER 1 INTRODUCTION

Historically, nutrient and water management have been viewed as linear

processes with the “take, make, waste” approach growing increasingly unsustainable.

Perspectives on the urban water cycle are shifting as we recognize the limitations of

conventional drinking water and wastewater management to address water stress,

resource consumption, water scarcity, and water quality. Similarly, growing population

concerns regarding global food security, and the environmental consequences of poor

nutrient management are motivating communities to pursue alternative nutrient

management strategies. As the water industry moves towards more sustainable water

management, an issue that is frequently discussed is the presence of emerging

contaminants. Specifically, pharmaceuticals as they relate to environmental and human

health and source water protection. Furthermore, pharmaceuticals may act as a barrier

for some nutrient recovery efforts. The work presented in this dissertation pertains to the

evaluation of a novel treatment process to help address the global issue of

pharmaceutical and nutrient pollution, and enhance nutrient recovery efforts.

Pharmaceuticals and Nutrients in the Environment

Figure 1-1 provides a visual representation of pharmaceuticals (red arrows) and

nutrients (green arrows) in the urban water cycle and the challenges they present for

sustainable water management practices. After pharmaceuticals are ingested, they are

metabolized and excreted in urine and feces as either the parent compound or

metabolites (Lienert et al. 2007b). This waste is then combined with greywater and

conveyed to the centralized wastewater treatment plant. As demonstrated by process A

in Figure 1-1, conventional wastewater treatment processes are generally ineffective

21

and/or inconsistent at removing these constituents, and they are ultimately discharged

to the environment (Blair et al. 2015, Verlicchi et al. 2012). Subsequently, wastewater

effluent has been designated as one of the major point sources of pharmaceutical

pollution in the environment (Daughton and Ternes 1999, Neale et al. 2017, Subedi and

Loganathan 2016). Numerous studies have documented the adverse effects of

pharmaceuticals on aquatic life (Wilkinson et al. 2016). Furthermore, as a result of de

facto reuse (i.e., unplanned reuse) in drinking water systems, pharmaceuticals have

been detected in source water and finished water (Benotti et al. 2008, Furlong et al.

2017, Rice and Westerhoff 2015). Recognizing the risks from unregulated

contaminants, such as pharmaceuticals, the U.S. Environmental Protection Agency has

identified the need to strengthen source water protection (U.S. EPA 2016a). For water

scarce locations direct potable reuse (DPR) (i.e., the use of wastewater as a drinking

water source) has become a necessary option for diversifying water supply. As

demonstrated by process B in Figure 1-1, the presence of pharmaceuticals remains an

issue for DPR systems with respect to source control, and treatment often includes high

energy processes to remove and/or destroy these compounds such as reverse osmosis

and advanced oxidation (WRRF 2015).

In addition to the “take, make, waste” approach to water management, a similar

approach for nutrient management has led to stress on resource consumption,

wastewater treatment, and environmental water quality. Two of the primary nutrients

utilized in fertilizer is phosphorus and nitrogen. Nitrogen fertilizers are created through

fixation of atmospheric nitrogen using the Haber-Bosch process, however, this process

is limited by the cost and availability of fossil fuels (Maurer et al. 2003). Phosphate rock

22

mining, the primary source of phosphorus, is a non-renewable resource whose global

reserves are being depleted at a rapid rate with an expected lifetime of 61 years to 400

years (Cordell et al. 2009, Desmidt et al. 2015). Coupled with growing population rate

and the geo-political challenges associated with the global distribution of phosphate

reserves, sustainable fertilizer resources are necessary to ensure global food security

(Desmidt et al. 2015). Furthermore, nutrient loading in the environment induces

significant water quality issue due to eutrophication (U.S. EPA 2016b). This has led

regulatory agencies to establish more stringent treatment criteria, such as the Numeric

Nutrient Criteria, to reduce nutrient loading to surface water bodies (FDEP 2015b). As

shown in Figure 1-1, the green arrows demonstrate the fate of nutrients in wastewater.

Wastewater effluent has been identified as a major point source of nutrient discharge

and more stringent regulations have made the technical and economic feasibility of

municipal wastewater treatment plants to meet effluent standards difficult (Stone and

Reardon 2011). In the National Water Program Research Strategy, the EPA has

identified the importance of addressing nutrient pollution using a multi-barrier approach

including source reduction, best management practices, sustainable treatment

technologies, and resource recovery (U.S. EPA 2015). Due to the numerous issues

associated with fertilizer production and the high nutrient content in wastewater,

treatment efforts have shifted to recover nutrients from wastewater for reuse as

agricultural fertilizer which can reduce the costs associated with extensive wastewater

treatment and reduce dependence on commercial fertilizers. However, the presence of

pharmaceuticals in wastewater remains a barrier to nutrient recovery efforts. Advanced

treatment of wastewater (i.e., advanced oxidation) for reuse is often employed for

23

pharmaceutical destruction (Gomes et al. in press, Snyder et al. 2014). Furthermore,

adsorption of pharmaceuticals to activated sludge is a barrier to land application due to

potential desorption from sludge, plant uptake, and risk to animal and human exposure

(Taylor-Smith 2015).

Urine Source Separation

A common management strategy among pharmaceutical and nutrient pollution is

source water protection and source reduction. One potential process that addresses this

issue is urine source separation. Urine source separation is the process by which urine

is diverted at the source (i.e., toilet or urinal), instead of being combined with black

water and greywater (Larsen and Gujer 1996a). The motivation for urine source

separation is that urine contributes 1% of the volumetric flow to combined wastewater

but >80% of the nitrogen load and >50% of the phosphorus load (Larsen and Gujer

1996b). In addition, approximately 64% of ingested pharmaceuticals intended for human

use are excreted in urine as the parent compound or metabolites (Lienert et al. 2007a).

As demonstrated by process C in Figure 1-1, urine source separation intercepts the

major sources of nutrient and pharmaceutical loading in wastewater.

In addition to pharmaceuticals, urine is rich in nitrogen and phosphorus which

may be utilized as an alternative nutrient source in agriculture (Kirchmann and

Pettersson 1995). Treating undiluted urine as a separate waste stream reduces nutrient

loading at the wastewater treatment plant and subsequent receiving waters, reduces

energy requirements associated with advanced nutrient removal, and has significant

potable water savings (Ishii and Boyer 2015, Maurer et al. 2003). Nutrient recovery

efforts in urine span several options, ranging from direct land application of urine, and

advanced treatment of urine such as struvite precipitation, adsorption, and ammonia

24

stripping (Maurer et al. 2006). Removing and/or destroying pharmaceuticals in undiluted

urine as opposed to municipal wastewater is expected to be more efficient because

pharmaceuticals are present at much greater concentrations in urine (Lamichhane and

Babcock 2012). To enhance nutrient recovery efforts, either through direct application of

urine or advanced nutrient recovery options, preliminary treatment of urine to separate

pharmaceuticals from nutrients is necessary to ensure a contaminant free nutrient

product (Maurer et al. 2006).

Various advanced treatment processes to remove or destroy pharmaceuticals in

source separated urine have been investigated with varying degrees of success.

Nanofiltration, for example, was effective at rejecting >90% of pharmaceuticals (Pronk

et al. 2006b). However, it was not effective at separating pharmaceuticals from nutrients

as indicated by 100% and >50% rejection of phosphate and ammonia, respectively.

Furthermore, ozonation of ureolyzed urine required very high ozone doses to oxidize

pharmaceuticals due to ozone scavenging by ammonia and other reactive matrix

constituents in ureolyzed urine (Dodd et al. 2008), and electrodialysis detected high

permeation of ibuprofen in the concentrate (Pronk et al. 2006a). The major limitations of

nanofiltration, electrodialysis, and ozonation of source separated urine is that these

treatment methods do not effectively remove pharmaceuticals from urine and separate

them from nutrients to create a contaminant free nutrient product. Previous work by

Landry and Boyer (2013) investigated the removal of diclofenac, an acidic

pharmaceutical, from urine using anion exchange resins (AER). Greater than 90%

removal of diclofenac was achieved under both fresh and ureolyzed urine conditions

with <20% co-removal of phosphate, thereby effectively separating diclofenac and

25

ketoprofen from nutrients. Considering the results from previous work, sorption

processes appear to be an effective method to selectively remove pharmaceuticals from

urine with minimal co-sorption of valuable nutrients. Furthermore, sorption processes

are attractive for pharmaceutical removal because they are scalable and low energy

(Crittenden et al. 2012).

Although AER was shown to be effective at separating pharmaceuticals from

nutrients and urine, the specific pharmaceutical–urine and pharmaceutical–AER

interactions of structurally similar pharmaceuticals at realistic concentrations in urine is

unknown. In addition, evaluating the reduction in ecotoxicity after pharmaceutical

removal in source separated urine provides perspective for the ecotoxicological

implications of the treatment process. Furthermore, conducting a life cycle assessment

of pharmaceutical removal by sorption processes in source separated urine is

necessary to understand the cradle-to-grave environmental impacts of the overall

treatment process.

Nonsteroidal Anti-Inflammatory Drugs

The focus of this dissertation is on the removal of nonsteroidal anti-inflammatory

drugs (NSAIDs); specifically, diclofenac (DCF), ibuprofen (IBP), ketoprofen (KTP), and

naproxen (NPX). This pharmaceutical class is widely consumed globally in large

quantities; non-narcotic analgesics, which includes NSAIDs, was ranked 15 out of 20 for

global therapeutic drug sales with $12.3 billion sold in 2011 (IMS Health 2011).

Approximately 50–100% of an ingested NSAID dose is excreted in urine as the parent

compound or metabolites (Lienert et al. 2007b). Due to the high excretion rates in urine,

urine source separation has been proposed as an effective method to reduce NSAID

loading into the environment (Lienert et al. 2007a). Furthermore, the removal of these

26

compounds in conventional wastewater treatment range from <50% for DCF, 20–50–

80% for KTP and NPX, and >80% for IBP (Petrie et al. 2015). In a review of

pharmaceutical and personal care products in the freshwater environment, Ebele et al.

(2016) found that 17 surface water studies in 13 different countries detected at least one

nonsteroidal anti-inflammatory drug (NSAID) (i.e., diclofenac, (DCF), ibuprofen (IBP),

ketoprofen (KTP), and naproxen (NPX)) ranging from 10 ng/L to >10 µg/L. Furthermore,

in an ecotoxicological risk model, ibuprofen and diclofenac were identified as having the

greatest ecotoxicological risk among pharmaceuticals studied (Lienert et al. 2007b). The

mode of action of NSAIDs is inhibition of the COX enzyme. Cyclooxygenase enzymes

are classified into two subtypes, COX-1 and COX-2, which catalyze prostaglandin (PG)

biosynthesis (Blobaum and Marnett 2007). The COX-2 enzyme produces PGs under

acute inflammatory conditions, and is the target enzyme for the anti-inflammatory

effects of NSAIDs (Blobaum and Marnett 2007). The COX-1 enzyme is associated with

normal cellular homeostasis, and inhibition has been attributed to gastrointestinal

toxicity in humans, gastrulation arrest and defective vascular tube formation in zebrafish

(Cha et al. 2005, Warner et al. 1999). Furthermore, chronic exposure of Japanese

medaka exposure to DCF resulted in decreased hatching success and delay in hatching

(Lee et al. 2011). Prostaglandin E2 (PGE2) was shown to be involved in estrogen

biosynthesis in mice, however it is unknown whether a similar mechanism of COX

enzyme applies to aquatic vertebrates (Lee et al. 2011).

Organization of Dissertation

The goal of this doctoral research was three-fold: (1) to improve the

understanding of pharmaceutical removal by sorption processes in source separated

urine at realistic concentrations in urine, (2) to elucidate the efficacy of ion-exchange

27

resins to reduce the ecotoxicity potential of pharmaceuticals and pharmaceutical

metabolites, and (3) to evaluate the environmental and economic implications of

pharmaceutical removal by ion-exchange in source separated urine. Within individual

chapters, the focus is on one of the specified goals. The following chapter, Chapter 2,

pertains to evaluating the ion-exchange selectivity and competitive sorption of the

NSAIDs, DCF, IBP, KTP, and NPX. Chapter 2 is the topic of a manuscript published in

Water Research. Chapter 3 pertains to the comparison of predicted fixed-bed column

removal of NSAIDs and NSAID metabolites and the corresponding reduction in

ecotoxicity quantified by their ability to inhibit the cyclooxygenase enzyme. The target

journal for findings discussed in Chapter 3 is Environmental Science & Technology and

submission will take place in 2017. The environmental and economic life cycle impacts

of implementing urine source separation with ion-exchange removal at the University of

Florida is the topic of Chapter 4. The system boundaries include potable water

production, urine treatment (i.e., separation, storage disinfection, pharmaceutical

removal, and struvite precipitation), and centralized wastewater treatment with or

without ozone for pharmaceutical destruction. The work presented in Chapter 4 is the

topic of a manuscript published in Water Research. Lastly, the Conclusions chapter

highlights the interconnectedness of the three main chapters’ systematic approach for

evaluating a new process to address pharmaceutical loading in the environment, as well

as address future inquiries for research.

28

Figure 1-1. Visual representation of the urban life-cycle of pharmaceuticals and nutrients in (A) conventional wastewater

treatment (WWT), (B) direct potable reuse (DPR), and (C) urine source separation (USS). In WWT, pharmaceuticals and nutrients discharged to receiving waters result in water quality issues (e.g., eutrophication) and ecotoxicological risk (Corcoran et al. 2010, Smith et al. 1999), and de facto reuse of wastewater effluent for drinking water purposes results in detected pharmaceuticals in finished water (Benotti et al. 2008). To address pharmaceutical concerns in DPR, high energy advanced treatment is often utilized (WRRF 2015). In USS, nutrients and pharmaceuticals are diverted from the general waste stream for more effective and efficient pharmaceutical removal and nutrient recovery (Larsen and Gujer 1996a, Lienert et al. 2007a).

29

CHAPTER 2 ION-EXCHANGE SELECTIVITY OF DICLOFENAC, IBUPROFEN, KETOPROFEN,

AND NAPROXEN IN UREOLYZED HUMAN URINE*

Application of Sorption Processes for Pharmaceutical Removal

Human urine is the major contributor of pharmaceuticals to wastewater treatment

plants, which are not designed to effectively remove pharmaceuticals by conventional

biological treatment (Joss et al. 2005, Salgado et al. 2012). As a result, the

pharmaceuticals are discharged to surface water where they pose an ecotoxicological

risk to aquatic organisms (Lienert et al. 2007b). Non-steroidal anti-inflammatory drugs

(NSAIDs), such as diclofenac, ibuprofen, naproxen, and ketoprofen, pose a high

ecotoxicological risk to species in the aquatic food chain when exposed to

environmentally relevant concentrations (Hernando et al. 2006). Approximately 70% of

ingested pharmaceuticals intended for human use are excreted in urine as either the

parent compound or its metabolites (Lienert et al. 2007a). For this reason, urine source-

separation and treatment is a proposed method to reduce pharmaceutical loading to the

environment by diverting undiluted urine from domestic wastewater (Lamichhane and

Babcock 2012).

Urine source-separation and treatment is also of interest as an alternative

approach to address excess nitrogen and phosphorus loading to aquatic systems

(Larsen et al. 2009). The high nutrient content in urine can be recovered to produce

fertilizer, which in turn can offset the raw materials and energy required to produce

synthetic fertilizer for agriculture (Kirchmann and Pettersson 1995). However, for

*Reproduced with permission from Landry, K.H., Sun, P., Huang, C.H., Boyer, T.H. 2015. Ion-exchange selectivity of diclofenac, ibuprofen, ketoprofen, and naproxen in ureolyzed human urine. Water Research

68, 510–521, DOI: http://dx.doi.org/10.1016/j.watres.2014.09.056. Copyright 2014 Elsevier Ltd.

30

nutrient recovery from source-separated urine to be considered a viable fertilizer

alternative, it is necessary to separate pharmaceuticals from nutrients to produce a

contaminant-free product.

When human urine leaves the body it is known as fresh urine (pH 6), and is

composed of urea, inorganic anions (Cl–, SO42–, PO4

3–), inorganic cations (Na+, K+,

Ca2+, Mg2+), and natural organic metabolites (Saude and Sykes 2007, Udert et al.

2003a). After a period of time, urease active bacteria, which are assumed ubiquitous in

wastewater collection systems, hydrolyze urea to form ammonia and bicarbonate and

increase the pH from 6 to 9 (Udert et al. 2003a). Prevention of urea hydrolysis is an

active area of research, and would require the addition of urease inhibitors to the urine

collection system to prevent pipe blockages in plumbing due to precipitation (Hellström

et al. 1999, Krajewska 2009). Therefore, it is considered more practical to implement

urine treatment technologies that effectively separate pharmaceuticals from nutrients in

ureolyzed urine without the added step of preventing urea hydrolysis.

Advanced treatment processes that have been applied to source-separated urine

for pharmaceutical removal or destruction include nanofiltration, ozonation,

electrodialysis, and anion exchange. Nanofiltration rejected >90% of diclofenac and

ibuprofen in urine, but also rejected 100% of phosphate and >50% of ammonia (Pronk

et al. 2006b). Ozonation of ureolyzed urine was inefficient at pharmaceutical destruction

due to ozone scavenging by ammonia (Dodd et al. 2008). Electrodialysis of urine was

partially effective at separating nutrients from pharmaceuticals, but high permeation of

ibuprofen was detected in the concentrate (Pronk et al. 2006a). Struvite precipitation in

urine can produce a fertilizer product with low pharmaceutical contamination

31

(Kemacheevakul et al. 2012, Ronteltap et al. 2007), but does not prevent

pharmaceuticals from entering the environment. Previous research investigated the use

of anion exchange resin (AER) to remove diclofenac and ketoprofen from synthetic

fresh and ureolyzed urine with high pharmaceutical removal of >90% (Landry and Boyer

2013). Additionally, the AER investigated was not selective for phosphate with negligible

removal in ureolyzed urine, thereby effectively separating pharmaceuticals from

nutrients. The primary mechanism of removal was electrostatic (i.e., Coulombic)

interactions between the carboxylate functional group of the pharmaceutical and the

quaternary ammonium functional group of the resin (Landry and Boyer 2013).

Furthermore, high pharmaceutical removal by AER required van der Waals interactions

between the benzene rings of the pharmaceutical and the polystyrene resin matrix.

Complete regeneration of the AER was achieved using a 5% NaCl, equal-volume

water–methanol solution due to the disruption of the Coulombic interactions between

the functional group of the resin and carboxylate functional group of the pharmaceutical

and van der Waals interactions between the resin matrix and benzene rings of the

pharmaceutical (Landry and Boyer 2013).

Considering previous research on urine treatment, anion exchange appears to be

an effective method to separate acidic pharmaceuticals from nutrients in urine. Although

the previous work by the authors provided new information regarding the use of AERs

to selectively remove diclofenac and ketoprofen from urine, it is unknown how other

pharmaceuticals with structurally similar properties may be removed. Other NSAIDs,

such as ibuprofen and naproxen, contain benzene rings and carboxylic acid functional

groups that deprotonate under fresh and ureolyzed urine conditions allowing for

32

Coulombic and van der Waals interactions with AER. Additionally, the previous research

was conducted at pharmaceutical concentrations much higher than what would be

realistically present in urine (Winker et al. 2008b). Isotherm modeling is often used to

investigate the underlying mechanisms of sorption processes, selectivity of sorbates to

sorbents, and resin capacity (Delle Site 2001). Previous research has incorporated

linearized isotherm modeling when studying the ion-exchange of charged

micropollutants in water using AER (Bäuerlein et al. 2012). However, the use of

linearized isotherm models can lead to errors when estimating model parameters (Foo

and Hameed 2010).

The goal of this research was to generate new experimental data on the ion-

exchange removal of diclofenac (DCF), ibuprofen (IBP), ketoprofen (KTP), naproxen

(NPX), and paracetamol (PCM) by AER in synthetic ureolyzed urine when present at

realistic concentrations in urine. The pharmaceuticals were selected based on

widespread use and high potential for ecotoxicity (Hernando et al. 2006, Li 2014). The

specific objectives of this work were to (i) elucidate the underlying mechanisms that

dictate the selectivity of AER for structurally similar pharmaceuticals, (ii) evaluate the

ion-exchange removal of pharmaceuticals when present individually or combined as a

mixture in synthetic ureolyzed urine, and (iii) evaluate the ion-exchange behavior of

pharmaceuticals under continuous-flow conditions.

Experimental Methods

Synthetic Human Urine

Synthetic ureolyzed urine was used for most experiments and synthetic fresh

urine was used for one experiment. The urine composition is given in Table 2-1 and was

33

based on previous work (Landry and Boyer 2013), with adjustment to maintain nitrogen

and inorganic carbon mass balance in fresh and ureolyzed urine (Boyer et al. 2014).

Pharmaceuticals in Urine

The chemical characteristics of the pharmaceuticals investigated in this work are

listed in Table A-1. Diclofenac sodium (CAS 15307-79-6, MP Biomedicals), ibuprofen

sodium (CAS 31121-93-4, Fluka Analytical), ketoprofen (CAS 22071-15-4, Sigma-

Aldrich), and naproxen sodium (CAS 26159-54-2, Sigma-Aldrich) are all weakly acidic

pharmaceuticals from the NSAID class. Paracetamol (CAS 103-90-2, Sigma-Aldrich) is

a weakly acidic pharmaceutical from the analgesic pharmaceutical class. Stock

solutions (1000 mg/L) of each pharmaceutical were made using equal-volume water–

methanol. Published data was reviewed to estimate realistic pharmaceutical

concentrations in urine as described in Table A-2 in Appendix A (Joss et al. 2005,

Salgado et al. 2012, Ternes 1998, Winker et al. 2008b). Based on this analysis, it was

estimated that pharmaceutical concentrations in urine range from 2–1,600 µg/L. The

stock solutions were used to spike the synthetic ureolyzed urine at an initial

pharmaceutical concentration of 2,000 µg/L (0.006–0.013 mmol/L). The solvent content

in the synthetic urine was 0.1% (v/v) for the individual pharmaceutical equilibrium

experiments and 0.4% (v/v) for the NSAID mixture equilibrium and column experiments.

One equilibrium experiment was performed with 0.2 mmol/L ibuprofen in synthetic fresh

urine.

Anion Exchange Resin

Dowex 22, a strong-base, polystyrene AER was used in all isotherm and column

experiments. A complete description of the AER is described in Appendix A (Table A-3).

34

Batch Equilibrium Tests

Batch equilibrium tests were performed in triplicate to investigate the ion-

exchange behavior of each pharmaceutical individually and as a combination of DCF,

IBP, NPX, and KTP. Ureolyzed urine was measured at 125 mL and added to 125 mL

Erlenmeyer flasks. Varying amounts of dried Dowex 22 AER (average density = 0.366 g

mL–1) was added at corresponding wet doses of 0.16, 2.12, 4.08, 6.04, and 8 mL/L. The

resin doses were selected to span a wide range of removal. Samples were mixed on a

shaker table at 325 rpm for an equilibrium time of 24 h and filtered using a 0.45 µm

PVDF syringe filter before being analyzed for pH and temperature. Filtered samples

were stored in 2 mL low-adsorption LC/MS vials and kept refrigerated until analyzed for

pharmaceutical concentrations. Batch equilibrium tests performed with 0.2 mmol/L IBP

in synthetic fresh urine followed the same method but with varying amounts of dried

Dowex 22 AER corresponding to wet doses of 1, 2, 4, 8, and 16 mL/L, which were the

same doses used in previous work (Landry and Boyer 2013), and analyzed using UV-

absorbance.

Column Tests

Fixed-bed column runs were conducted in a glass column (0.7854 cm inner

diameter) packed with 1 mL of Dowex 22 AER to obtain a height:diameter ratio of at

least 2 (Edzwald 2011). All column tests were performed under the same conditions by

maintaining an empty bed contact time (EBCT) and flow rate of 2 min and 0.5 mL min–1,

respectively. The synthetic ureolyzed urine was spiked with a mixture of DCF, IBP, KTP,

and NPX at an initial concentration of 2000 µg/L (0.006–0.013 mmol/L), each. For the

first column run, 14,300 bed volumes (BVs) of synthetic ureolyzed urine were treated.

Effluent composite samples were collected every 12 h and influent control samples

35

were collected every 24 h. After treatment, the column was rinsed with DI water to

displace the synthetic ureolyzed urine in the column. Regeneration of the AER in the

column was conducted using a regeneration solution that contained 5% (m/m) NaCl in

an equal-volume mixture of water–methanol. Column regeneration was completed with

24 BV of regeneration solution at an EBCT and flow rate of 4 min and 0.25 mL min–1,

respectively. Regeneration effluent samples were collected every 8 min resulting in 2

mL samples which were further diluted 187.5× prior to analysis by LC/MS. The

regenerated AER was used to treat 5,950 BVs of synthetic ureolyzed urine spiked with

the pharmaceutical mixture under the same conditions. The column tests were

conducted for three treatment and regeneration cycles.

Analytical Methods

The synthetic urine was filtered before each test using 0.45 µm membrane filter

(Millipore Durapore) to separate particulate impurities from urine, and after each test

using a 0.45 µm PVDF membrane syringe filter (Millipore Durapore) to separate the

AER from urine. Preliminary experiments showed negligible adsorption of the studied

pharmaceuticals to the PVDF filter (results not shown). Pharmaceutical concentrations

for the equilibrium and column experiments were measured using an Agilent 1100

Series LC/MSD system (Agilent Technologies, Palo Alto, CA) equipped with a reversed-

phase column (2.1 × 150 mm, 3 μm Ascentis RP-amide column; Supelco, Bellefonte,

PA). The mobile phase consisted of (A) a mixture of HPLC grade water and formic acid

(99.9/0.1 v/v), and (B) HPLC grade methanol and acetonitrile (50/50 v/v). The mass

spectrometer was set at positive electron-spray ionization (ESI+) with select ion

monitoring (SIM) mode. A five-point calibration curve (100, 200, 500, 1,000, 2,000 µg/L)

was created by serial dilution of the stock standards. The coefficient of determination

36

(R2) was ≥ 0.992 for all pharmaceuticals except for ibuprofen (0.932 ≤ R2 ≤ 0.958). The

analytical method had a detection limit of around 7–9 µg/L for IBP, KTP, NPX and PCM,

and around 40 µg/L for DCF. Ibuprofen concentrations in the fresh urine experiment (C0

= 0.2 mmol/L) were measured using UV absorbance on a U-2900 UV–visible

spectrophotometer (Hitachi High Technologies) and 1 cm quartz cuvette at a

wavelength of 222 nm. A more detailed explanation of the method used can be found

elsewhere (Landry and Boyer 2013). All samples were measured for pH and

temperature at the end of each experiment using an Accumet AB-15 + pH meter and

pH/ATC probe. The pH meter was calibrated prior to each use with 4, 7, and 10 buffer

solutions.

Data Analysis

Data from the equilibrium tests were the mean value of triplicate samples.

Analysis of covariance (ANOCOVA) was conducted using MATLAB (8.2.0.701 R2013b)

(MathWorks 2013) to determine if there was a significant difference (α = 0.05) between

the slopes of the log-log transformed ion-exchange isotherms. The null hypothesis

states that there was not a significant difference between slopes (p > 0.05) and the

alternative hypothesis states that there was a significant difference between the slopes

(p < 0.05). Integration of the column sorption and regeneration curves was conducted

using trapezoidal numerical integration method in MATLAB.

Isotherm Models

Nonlinear isotherm modeling of the equilibrium experiments was performed using

MATLAB (8.2.0.701 R2013b) following the nonlinear least squares method. The

experimental data were fit to the Freundlich, Langmuir, Dubinin-Astakhov (D-A), and

Dubinin-Radushkevich (D-R) isotherm models; a detailed description of the theory

37

behind each model is given in Appendix A. Linear regression of each isotherm model

was conducted to establish initial values for the nonlinear model parameters. The linear

form and plot of each isotherm are shown in Appendix A (Table A-4). To determine the

best fitting isotherm model, the experimental data were evaluated using the correlation

coefficient (R2), the sum of squares error (SSE), and the average relative error (ARE).

The SSE was used to compare the fit of the four isotherm models to the experimental

data from one equilibrium experiment. The R2 and ARE were used to compare the fit of

one isotherm model to multiple equilibrium experiments. The pharmaceutical

concentration on the AER (qe, mmol/g) was calculated as the difference between initial

and equilibrium aqueous concentrations divided by the dose of AER.

Results and Discussion

Ion-Exchange of Individual Pharmaceuticals at Realistic Concentrations

Batch equilibrium tests were performed to investigate the removal of individual

pharmaceuticals in ureolyzed urine. Some pharmaceuticals were not completely soluble

at the spiked concentration, possibly due to the high ionic strength of the synthetic

ureolyzed urine. For all analysis and discussion of the experimental data, the measured

concentration of the control sample (C0) was used as follows: diclofenac (DCF, C0 =

2.96×10–3 mmol/L), ibuprofen (IBP, C0 = 3.65×10–3 mmol/L), ketoprofen (KTP, C0 =

7.80×10–3 mmol/L), naproxen (NPX, C0 = 7.51×10–3 mmol/L), and paracetamol (PCM,

C0 = 1.38×10–2 mmol/L). High removal was observed for DCF (95%), IBP (93%), KTP

(86%), and NPX (94%) at the highest AER dose of 8 mL/L. High removal of DCF, IBP,

KTP, and NPX was due to the combination of electrostatic (i.e., Coulombic) interactions

between the ionized carboxylic acid functional group of the pharmaceutical and the

quaternary ammonium functional group of the AER as well as the concurrent non-

38

electrostatic (i.e., van der Waals) interactions between the benzene rings of the

pharmaceutical and the polystyrene matrix of the AER (Landry and Boyer 2013). Under

ureolyzed urine conditions, at pH 9, the carboxylic acid functional group of DCF, IBP,

KTP, and NPX are all nearly 100% deprotonated. Low removal was observed for PCM

(14%) due to the lack of Coulombic interactions, only 40% of PCM was present in its

ionized form.

Nonlinear isotherm modeling was conducted to elucidate the selectivity and

capacity of the AER for each pharmaceutical. The selectivity was defined as the

equilibrium ratio of the solid-phase concentration (qe) to the liquid-phase pharmaceutical

concentration (Ce) where a higher selectivity indicates more pharmaceutical is present

on the solid-phase than in solution (Saikia and Dutta 2008). High selectivity of AER for

the pharmaceutical is beneficial because large amounts of pharmaceutical can be

sorbed when present at low concentrations or by using a small amount of AER, as well

as exhibit preferential ion-exchange over competing compounds. The capacity was

defined as the maximum amount of pharmaceutical that can be sorbed to the AER

before it is considered saturated.

Figure 2-1 shows the experimental data and nonlinear isotherm models for DCF,

IBP, KTP, and NPX in ureolyzed urine. The experimental data were fit to the Freundlich,

Langmuir, D-A, and D-R isotherm models; the isotherm parameters and goodness-of-fit

data are listed in Table A-5 in Appendix A. The DCF, KTP, and NPX experimental data

showed more favorable ion-exchange behavior as depicted by the steep slope and

concave-down shape, which allows for higher pharmaceutical loading on the AER at

lower concentrations. The IBP and PCM ion-exchange systems followed an unfavorable

39

ion-exchange trend as shown by a concave-up shape, where high removal was only

achieved at high AER dose (see Figure 2-1 and Figure A-1). Due to the very low

removal of PCM, none of the isotherm models fit the experimental data well (–0.297 <

R2 < 0.142; 82% < ARE < 277%) and therefore PCM was excluded from the remainder

of the discussion.

All isotherm models fit the data well for DCF, IBP, and KTP. However, the

isotherm models exhibited a poor fit to the NPX ion-exchange data (–0.538 ≤ R2 ≤

0.061; 67 ≤ ARE ≤ 91%). The poor fit of NPX to the isotherm models was likely due to

error when determining the amount exchanged onto the resin at the lowest AER dose

(0.16 mL/L). Very low removal occurred at the lowest resin dose, which may have led to

error in determining the amount exchanged onto the resin. In the individual equilibrium

experiments with NPX, only 5.72×10–4–5.55×10–3 mmol/g was exchanged onto the resin

at the lowest resin dose. However, as described later in the discussion, 1.06×10–2–

2.02×10–2 mmol g –1of NPX was exchanged onto the resin during the combined

equilibrium experiments at the same resin dose. Excluding the lowest measured AER

dose from the experimental data improved the fit to the isotherm models (0.961 < R2 <

0.989; 4 ≤ ARE ≤ 20%). For brevity, the discussion will focus on the NPX isotherm

model with the lowest measured AER dose excluded from the experimental data.

However, the isotherm models with the entire range of experimental data may be found

in Appendix A (Table A-5, Figure A-2).

The Langmuir model presented the best fit for DCF, IBP, and KTP ion-exchange

systems (0.751 < R2 < 0.960; 11% < ARE < 34%). Negative parameter values obtained

for the IBP ion-exchange system indicate that the Langmuir model does not provide a

40

good description of the ion-exchange process because these parameters signify the

surface binding energy and monolayer coverage of the AER (Fungaro et al. 2009). A

separation factor (RL) was calculated from the Langmuir constant (KL) and initial solute

concentration (C0) to indicate whether the ion-exchange process was favorable (RL < 1)

or unfavorable (RL > 1), a more detailed description can be found in Appendix A (Foo

and Hameed 2010). For the IBP equilibrium experiment, an RL of 1.84 indicated that ion-

exchange by the AER was unfavorable. However, favorable ion-exchange was

observed for DCF, KTP, and NPX, as shown by RL < 1. In addition, the adsorption

energy of a solute on a sorbent may also be expressed by the change in Gibbs free

energy (ΔG°) calculated from the Langmuir constant, KL, a more detailed description is

presented in Appendix A. The ΔG° values suggested an order of decreasing selectivity

of DCF > NPX > KTP > IBP. The Freundlich isotherm fit the data well for all ion-

exchange systems (0.750 ≤ R2 ≤ 0.988; 10% < ARE < 37%). The selectivity of the AER

for the pharmaceuticals was determined from the Freundlich parameter, 1/nF, and

followed the order of decreasing selectivity of NPX > DCF > KTP > IBP. Favorable ion-

exchange (1/nF < 1) was observed for NPX, DCF, and KTP and unfavorable ion-

exchange (1/nF > 1) was observed for IBP.

Similar to the Langmuir and Freundlich model, the D-A and D-R models fit the

DCF, IBP, and KTP ion-exchange systems fairly well (0.749 < R2 < 0.955; 10 < ARE% <

40%), and the D-A model fit the NPX ion-exchange system best. The mean free energy

of sorption (E) determined from the D-A and D-R isotherms may be used to estimate the

type of sorption and is defined as the free energy change when 1 mole of ion is

transferred to the surface of a solid (Dominguez et al. 2011, Mahramanlioglu et al.

41

2002). Values of 8 kJ mol–1 < E < 16 kJ mol–1 indicate pure ion-exchange and values of

E < 8 kJ mol–1 indicate van der Waals interactions (Mahramanlioglu et al. 2002). The E

for the D-R model were 5.5–9.1 kJ mol–1, suggesting that the sorption mechanism was

not pure ion-exchange. This is consistent with the conclusion from the authors’ previous

work that a combination of Coulombic and van der Waals interactions were necessary

to selectively remove DCF using strong-base, polystyrene AER (Landry and Boyer

2013). The E for the D-A model suggested an order of decreasing selectivity of IBP >

KTP > DCF > NPX, which was not consistent with the ΔG° values deduced from the

Langmuir isotherm. However, the E determined for the D-R model suggested an order

of decreasing selectivity of NPX > DCF > KTP > IBP, which was closely aligned with the

order of selectivity determined by the Freundlich and Langmuir isotherms.

Effect of Pharmaceutical Properties

The NSAIDs investigated in this work all possessed the necessary Coulombic

interactions to participate in ion-exchange. Previous work by the authors found that

although a stoichiometric release of the counter-ion indicated ion-exchange as the main

mechanism of removal, van der Waals interactions were necessary to increase

selectivity between the pharmaceutical and AER (Landry and Boyer 2013). As a result,

the van der Waals interactions between the benzene rings of the pharmaceutical and

the polystyrene matrix of the AER appear to be the underlying reason for the order of

ion-exchange selectivity. Li and SenGupta (2004) found that if the charge of

hydrophobic ionizable organic compounds are identical then the non-polar portion will

influence the ion-exchange selectivity, where larger non-polar domains exhibit higher

selectivity. All of the NSAIDs investigated herein have identical charge (i.e., one

deprotonated carboxylic acid); DCF, NPX, and KTP contain two benzene rings whereas

42

IBP contains one benzene ring. The variation in the number of benzene rings implies

that the polarizability, and the dispersive interactions of DCF, NPX, and KTP are greater

than IBP, similar to that of pyrene, naphthalene, and benzene (Schwarzenbach et al.

2002). The molar refractivity is a measure of the total polarizability of a compound and

can also be used as a measure of the strength of the van der Waals forces between the

sorbate and sorbent (Ghose and Crippen 1987). The molar refractivity of the four

NSAIDs was predicted using the ChemAxon Calculator Plugin in Marvin (v6.3.0,

(ChemAxon 2013)) and followed an order of decreasing magnitude (i.e., decreasing van

der Waals forces) of DCF (75.46 cm3 mol–1) > KTP (72.52 cm3 mol–1) > NPX (64.85 cm3

mol–1) > IBP (60.73 cm3 mol–1). This order suggests that the van der Waals interactions

between DCF, KTP, and NPX and the AER are stronger than the interaction between

IBP and the AER. To further elucidate the pharmaceutical–AER interactions, isotherm

modeling of DCF ion-exchange by three AERs—Dowex 22, A520E, and Dowex 11—

was performed. The varying AER properties are given in Appendix A (Table A-3), and

the respective isotherm figures and parameters are given in Appendix A (Figure A-3 and

Table A-6). In general, the Freundlich, Langmuir, and D-R isotherms suggested an

order of affinity of Dowex 22 > A520E > Dowex 11. It was speculated that the Dowex 22

AER exhibited the greatest selectivity for DCF due to additional hydrogen bonding

between the carboxylic acid functional group or secondary amine of DCF and the

dimethylethanol amine functional group of the AER. Hydrogen bonding between DCF

and A520E or Dowex 11 was not possible because the functional groups of the AERs

cannot form hydrogen bonds. Recent work by Zhang et al. (2014) observed a similar

relationship between the selectivity of AER for anionic organic compounds and

43

hydrogen bonding abilities. Furthermore, steric hindrance may play a role in the AER

selectivity of the investigated pharmaceuticals. The hydrodynamic radii follows a

decreasing order of IBP (0.680 nm) > DCF (0.458 nm) > NPX (0.377 nm) (Bester-Rogac

2009); the hydrodynamic radius of KTP could not be determined from the literature. It is

speculated that the larger hydrated size of IBP in urine may inhibit effective ion-

exchange.

Previous research has established a correlation between the hydrophobicity of

an organic compound and ion-exchange behavior where the more hydrophobic

compounds demonstrate better ion-exchange due the favorable partitioning to the

surface of the sorbent from the bulk aqueous phase (Hand and Williams 1987, Li and

SenGupta 1998, Schwarzenbach et al. 2002). Anion exchange resins, such as Dowex

22, can be viewed as a particle containing a matrix of aromatic hydrocarbons with

charged functional groups, similar to that of particulate organic matter (Schwarzenbach

et al. 2002). The adsorption of hydrophobic organic compounds onto organic sediments

has been described as a partitioning process between water and the lipophilic solid

phase that may be simulated by the octanol–water partitioning coefficient (Kow) (Gawlik

et al. 1997), and linear free energy relationships have been established to demonstrate

the correlation between the Kow and the adsorption of organic compounds onto

particulate organic matter (Gawlik et al. 1997, Schwarzenbach et al. 2002). For

ionizable organic compounds, similar relationships may be developed to estimate ion-

exchange onto sediments using the pH-dependent octanol–water distribution coefficient

(D) as a predictor (Kah and Brown 2007). Dominguez et al. (2011) illustrated that the

maximum sorption of various pharmaceuticals onto a polymeric adsorbent was

44

dependent on the log D of the pharmaceutical in solution, where the more hydrophobic

species (i.e., increasing log D) exhibited greater sorption. The purpose of using the log

D is to account for the change in hydrophobicity at varying pH. Previous studies have

determined that fully ionized hydrophobic organic compounds, such as the ones

investigated here, may partition into octanol in appreciable amounts at high pH, as

shown in Eq. 2-1 (Jafvert et al. 1990, Strathmann and Jafvert 1998):

𝐷 =[𝐶𝑛]𝑜+[𝐶𝑖]𝑜

[𝐶𝑛]𝑤+[𝐶𝑖]𝑤 (2-1)

where [Cn]o is the concentration of the neutral species present in the octanol phase, [Ci]o

is the concentration of the ionized species present in the octanol phase, [Cn]w is the

concentration of the neutral species in the water phase, and [Ci]w is the concentration of

the ionized species in the water phase.

A simple equation (Eq. 2-2) may be used to calculate D of acidic ionized organic

compounds over the entire pH range (Kah and Brown 2008):

𝐷 = 𝐾𝑜𝑤 (1

1+10𝑝𝐻−𝑐𝑝𝐾𝑎) + 𝐾𝑜𝑤

′ (1 −1

1+10𝑝𝐻−𝑐𝑝𝐾𝑎) (2-2)

where (Kow) is the octanol–water partitioning coefficient of the neutral species, K′ow is

the octanol–water partitioning coefficient of the fully ionized species, the pH of

ureolyzed urine, and the conditional acid dissociation constant (cpKa) of the organic

compound which was corrected for the ionic strength of ureolyzed urine. The K′ow

values were determined previously by Scott and Clymer (2002) using a nonlinear least

squares best fit of Eq. 2-2 using experimental data. At pH 9, the hydrophobicity of DCF,

IBP, KTP, and NPX decreased considerably, as indicated by a reduction in the log D

value from that of the neutral species (Table A-1). The mole fraction distributions of the

neutral and ionized species present in the octanol and water phases were determined

45

by using Eq. 2-1 and the Henderson-Hasselbach equation. As illustrated in Figure A-4

at pH 9, 61–85% of the ionized molar fraction of DCF, KTP, and NPX was present in the

octanol phase, whereas only 14–38% of the ionized molar fraction was present in the

water phase. The partitioning behavior of IBP was much different than the other

NSAIDs, where 33% and 66% of ionized IBP was present in the octanol and water

phases, respectively. This suggests that IBP was more hydrophilic in ureolyzed urine

than DCF, KTP, and NPX, and therefore exhibited unfavorable ion-exchange to the

AER due to preferential partitioning in the aqueous phase. However, in fresh urine at pH

6, 92–99% of the molar fractions for all four NSAIDs were present in the octanol phase.

Therefore, it was hypothesized that IBP may be more hydrophobic under fresh urine

conditions and exhibit a greater ion-exchange selectivity for the AER.

Following the logic in the previous paragraph, a qualitative estimate can be made

on the ion-exchange behavior of pharmaceutical metabolites. This is an important

consideration because the majority of the pharmaceuticals found in urine are likely

present in the metabolized form (Lienert et al. 2007a). For example, 6% of diclofenac is

present in urine as the parent compound and 60% as its metabolites (Zhang et al.

2008). Through hydroxylation and conjugation, hydroxyl and carboxyl groups are added

to the diclofenac parent compound, altering the acidity and hydrophobicity. The pKa, log

Kow, and log D for the four major diclofenac metabolites were estimated using the

PALLAS computational program (v3.8.1.2, pKalc, PrologP, and PrologD prediction

programs for Windows, (CompuDrug 2006)); a description of the estimation procedure

is given elsewhere (Parang et al. 1997). The estimated chemical properties of the four

major diclofenac metabolites are listed in Table A-7. The estimated pKa was slightly

46

higher than diclofenac and the addition of phenol groups during hydroxylation added a

second acid dissociation constant. The log Kow decreased in comparison to the parent

compound, with the exception of 4'-hydroxydiclofenac and the log D at pH 6 and 9 were

much lower than diclofenac. Thus, it is speculated that the metabolites may be more

hydrophilic than diclofenac and have a lower selectivity for AER.

Effect of Urine Composition

To investigate the hypothesis that the selectivity of the AER for IBP would

increase in fresh urine, an equilibrium experiment was conducted with IBP (C0 = 0.256

mmol/L) in synthetic fresh urine. The initial concentration of IBP was higher than

previous experiments due to difficulty in measuring the pharmaceutical concentrations

by LC/MS in the synthetic fresh urine. Removal of IBP ranged from 15% at the lowest

resin dose (1 mL/L) to 80% at the highest resin dose (16 mL/L). As shown in Table A-8,

the experimental data was fit to the Freundlich, Langmuir, D-A, and D-R isotherms. Ion-

exchange of IBP was favorable according to the Freundlich parameter, 1/nF = 0.727,

and a RL of 0.562 determined from the Langmuir isotherm. The ion-exchange energy

(E) of 4.22 kJ mol–1 determined from the D-R isotherm indicate that ion-exchange was

not the only sorption mechanism and was coupled with additional interactions such as

van der Waals and hydrogen bonding. These results suggest that the selectivity of the

AER for IBP increased under fresh urine conditions due to the more hydrophobic nature

of IBP at pH 6. Therefore, it may be more advantageous to treat fresh urine to achieve

greater selectivity of IBP. Conversely, pharmaceuticals that are hydrophobic over the

entire pH range, such as DCF and KTP, will exhibit the same ion-exchange behavior in

fresh and ureolyzed urine (Landry and Boyer 2013).

47

Effect of Multiple Pharmaceuticals

Realistically, source separated urine would contain a mixture of various

pharmaceuticals that may compete for ion-exchange sites on the AER, interact in

solution, or in some cases, aid in the ion-exchange process. A previous study by Bui

and Choi (2009) showed an increase in removal due to non-competitive multilayer co-

adsorption of multiple pharmaceuticals onto mesoporous silica. To determine the effect

of multiple pharmaceuticals on ion-exchange, a mixture of DCF (C0 = 3.53×10–3

mmol/L), IBP (C0 = 4.70×10–3 mmol/L), KTP (C0 = 7.33×10–3 mmol/L), and NPX (C0 =

7.45×10–3 mmol/L) was spiked in ureolyzed urine. An equilibrium experiment was

conducted following the same approach as the individual pharmaceutical experiments.

High removal was observed for DCF (96%), IBP (80%), KTP (84%), and NPX (95%) at

the highest AER dose of 8 mL/L. A slight increase in removal was observed at all AER

doses for DCF and most AER doses for NPX compared to the individual ion-exchange

experiments (Figure 2-2), and there was a decrease in removal at all AER doses for IBP

and KTP.

For the pharmaceutical mixture equilibrium experiments, nonlinear isotherm

modeling was conducted individually for each pharmaceutical present (Table A-9,

Figure A-5). All isotherm models fit the data well for DCF, IBP, and NPX (0.761 ≤ R2 ≤

0.989; 15% ≤ ARE ≤ 48%). However, the isotherm models exhibited a poor fit to the

KTP ion-exchange data, possibly due to error when determining the amount exchanged

onto the resin at the lowest AER dose (0.16 mL/L) (Figure A-2). At the lowest resin

dose, 0–5.12×10–3 mmol/g KTP was exchanged onto the resin during the

pharmaceutical mixture equilibrium experiment; however, 7.48×10–3–1.78×10–2 mmol/g

KTP was exchanged onto the resin during the individual equilibrium experiment.

48

Excluding the lowest measured AER dose from the experimental data improved the fit

to the isotherm models (0.981 < R2 < 0.988; 3 ≤ ARE ≤ 37%). Again, the discussion will

focus on the KTP isotherm model with the lowest measured AER dose excluded from

the experimental data. The Langmuir model presented the best fit for the ion-exchange

systems excluding IBP where favorable ion-exchange was observed for DCF, KTP, and

NPX (0 < RL < 1). The Langmuir model exhibited a poor fit to the IBP ion-exchange data

as indicated by negative KL and q0 values and RL > 1. The ΔG° values indicated an

order of decreasing selectivity of DCF > KTP > NPX > IBP. Overall, there was a

decrease in the ion-exchange capacity for each pharmaceutical, suggesting that there

was competition for ion-exchange sites at increasing concentrations.

According to the Freundlich parameter, 1/nF, ion-exchange of the

pharmaceuticals followed the order of decreasing selectivity of NPX > KTP > DCF > IBP

where NPX, KTP, and DCF exhibited favorable ion-exchange and IBP exhibited

unfavorable ion-exchange (Table A-9). This trend varied slightly from the order of

selectivity established by the Freundlich model for the individual pharmaceutical ion-

exchange experiments. It can be observed that the presence of multiple

pharmaceuticals decreased the ion-exchange selectivity, as indicated by an increase in

1/nF. However, an increase in KF for all pharmaceuticals signified an increase in the ion-

exchange capacity of the AER, which is consistent with previous research that studied

the effect of co-sorption of multiple pharmaceuticals onto a polymeric adsorbent

(Dominguez et al. 2011).

The E calculated for the D-A model followed the order of decreasing selectivity of

IBP > DCF > NPX > KTP, which was inconsistent with the order of selectivity

49

established by the Langmuir and Freundlich isotherms. However, the E determined for

the D-R model suggested an order of decreasing selectivity of DCF > KTP > NPX > IBP.

Again, E values of 4–7.84 kJ mol–1 indicated that pure ion-exchange was not the only

mechanism for removal and other interactions, such as van der Waals or hydrogen

bonding, was occurring between the pharmaceuticals and AER.

An analysis of covariance (ANOCOVA) was conducted to determine if there was

a significant difference (α = 0.05) in the slopes of the isotherm models derived for

pharmaceuticals present individually and as a mixture. The x and y data points, Ce and

qe, respectively, were log-transformed to obtain a linear equation for the ANOCOVA

analysis. The prediction plots and ANOCOVA table are given in Table A-10. For all ion-

exchange systems, there was not a significant difference between the slopes for the

individual and pharmaceutical mixture experiments. However, when comparing all of the

data points for NPX or KTP (i.e. including the lowest measured AER dose (0.16 mL/L)

there was a significant difference in removal when KTP or NPX was present individually

or as a mixture. This reinforces the earlier conclusion that the poor ion-exchange

exhibited by the lowest AER dose may be due to experimental error in the NPX

individual experimental data set and the KTP pharmaceutical mixture data set.

Column Studies

Continuous-flow column studies were performed using a mixture of DCF, IBP,

KTP, and NPX in synthetic ureolyzed urine. The process of treatment and regeneration

was completed for three cycles. Figure 2-3 shows the effluent pharmaceutical

concentration (Ce) normalized by the influent pharmaceutical concentration (C0), and

Figure A-6 shows the absolute pharmaceutical influent and effluent concentrations

during each treatment cycle. Effluent sample concentrations that measured greater than

50

the control sample or less than zero were set equal to the concentration of the control

sample or zero, respectively. Elution curves for regeneration cycles 1–3 are shown in

Figure A-7 and a mass balance of the pharmaceuticals sorbed and desorbed on the

resin is shown in Table 2-2.

The equations used for determining the mass balance in Table 2-2 are outlined in

Appendix A, the following mass balance for DCF during cycles 1–3 is provided as an

example calculation. In cycle 1, the mass of DCF removed from urine was determined

by trapezoidal numerical integration of the column sorption curve (Figure A-6) using

MATLAB. A total mass of 23.2 µmol DCF removed from urine is equivalent to the mass

sorbed onto the AER in cycle 1 because the AER is considered “fresh” (i.e., no

contaminant was initially present). The amount desorbed from the AER in cycle 1 (i.e.,

21.7 µmol DCF) was also determined by trapezoidal numerical integration of the elution

curve (Figure A-7). The amount remaining on the resin was determined by taking the

difference between the mass of DCF sorbed and mass of DCF desorbed (e.g., 1.5 µmol

DCF) and the % regeneration was determined by dividing the amount desorbed by the

amount sorbed (e.g., 93%). In cycle 2, the amount removed from urine (e.g., 9.2 µmol

DCF) was added to the amount remaining on the AER after cycle 1 regeneration to

determine the total amount sorbed onto the AER (e.g., 10.7 µmol DCF). If <100%

regeneration was achieved in the previous cycle then the amount sorbed onto the AER

may be greater than the amount removed from urine, as is the case for DCF in cycle 2.

The amount desorbed from the AER in cycle 2 was greater than the amount sorbed

onto the resin, which theoretically is not possible. Error may have been introduced when

51

diluting the regeneration sample prior to analysis and/or when integrating the sorption

and regeneration curves using the trapezoidal numerical integration method.

For cycle 1, the column treated 14,300 BVs of ureolyzed urine until complete

resin saturation of pharmaceuticals was achieved (i.e., Ce/C0 ≈ 1), and subsequently

regenerated using a 5% (m/m) NaCl, equal-volume water–methanol solution. The

column reached saturation of IBP first after 2,190 BVs followed by KTP and NPX after

5,160 BVs. The column did not reach saturation of DCF after the treated volume of

14,300 BVs. The order of decreasing AER capacity was DCF > NPX > KTP > IBP,

which was the same order observed for the Langmuir isotherm parameter, q0, from the

pharmaceutical mixture equilibrium experiments (Table A-9). DCF had the greatest

amount sorbed onto the resin, followed by NPX and KTP. Complete regeneration was

achieved for NPX in cycles 1–3. Regeneration efficiency for DCF and KTP was 97%

and 74%, respectively. IBP exhibited the lowest amount sorbed onto the AER due to the

unfavorable selectivity of the AER, as well as the lowest regeneration efficiency at 64%.

For cycles 2 and 3, 5,950 BVs of ureolyzed urine was treated. There was a marked

decrease in the total amount sorbed onto the AER because of the lower number of BVs

treated compared to cycle 1. Complete regeneration for DCF was achieved in cycles 2

and 3. Regeneration for KTP decreased in cycles 2 and 3 to 41% and 23%,

respectively. IBP continually decreased in regeneration efficiency from 22% in cycle 2 to

2% in cycle 3. It was expected that complete desorption of IBP would have occurred

during regeneration because the AER had the lowest selectivity for IBP. The low

amount of IBP desorbed from the AER was due to the low amount sorbed onto the

AER. The maximum ion-exchange capacity determined by the Langmuir isotherm from

52

the pharmaceutical mixture equilibrium experiments tended to underestimate the

maximum ion-exchange capacity for each pharmaceutical (cycle 1, Table 2-2) with a

relative error of 61–331%, except for NPX which had an 11% relative error. The D-R

isotherm parameters overestimated the maximum ion-exchange capacity with a relative

error of 36–100% for each pharmaceutical. It should be noted that DCF did not reach

saturation in the column, therefore the estimated capacity of the resin of 0.106 mmol/g

determined from the D-R isotherm (Table A-9) may be an accurate estimation of column

capacity for DCF. However, for the remaining pharmaceuticals, the isotherm models do

not accurately estimate sorption capacities of the AER under continuous flow

conditions.

Practical Application and Future Work

An ion-exchange column for pharmaceutical removal would ideally precede

nutrient removal and/or recovery in source separated urine to produce a contaminant

free nutrient product. For perspective, a 2 L column of AER could potentially treat up to

4,380 L or 28,600 L of ureolyzed urine to fully saturate the AER column with IBP or

DCF, respectively. However, the AER would not treat urine until saturation but would

rather reach a predetermined operating capacity followed by regeneration. The

operating capacity should be established based on a correlation between % removal

and % reduction in ecotoxicity. Future work will need to be conducted to evaluate the %

reduction in ecotoxicity after treatment. Furthermore, brine disposal also poses an

issue, therefore, advanced oxidation of the regeneration brine is being investigated to

further destroy the pharmaceuticals as well as potentially produce a reusable

regeneration solution.

53

Concluding Remarks

The Langmuir and Freundlich isotherm models indicated that the selectivity of

Dowex 22 AER followed the order DCF > NPX > KTP > IBP > PCM and NPX > DCF >

KTP > IBP > PCM, respectively. Favorable ion-exchange was observed for DCF, KTP,

and NPX and unfavorable ion-exchange was observed for IBP and PCM. The D-R

isotherm suggested that the sorption interactions between the AER and

pharmaceuticals were not purely ion-exchange. The ion-exchange selectivity was

governed by van der Waals interactions between the acidic pharmaceuticals and AER.

Based on experimental results, it is predicted that AER will be less selective for the

pharmaceutical metabolites than the parent compound because of more hydrophilic

character of the metabolites. These conclusions are expected to apply generally to

strong-base, polystyrene AER. The selectivity of the AER for IBP was greater in fresh

urine due to increasing hydrophobicity of the pharmaceutical. This result suggests that

more efficient separation of IBP from urine may be achieved in fresh urine as opposed

to ureolyzed urine. Urine chemistry should be considered during treatment design to

achieve greater selectivity of IBP, particularly in demographic areas where IBP may be

consumed in large quantities. The ion-exchange behavior of the NSAIDs was not

significantly different when present individually or as a mixture in solution. Continuous-

flow column experiments provide valuable insight on the practical application of AER to

separate pharmaceuticals from ureolyzed urine. Because the pharmaceuticals

investigated in this work reached saturation at varying bed volumes, the size of the AER

bed may need to vary according to the pharmaceutical present either at the highest

concentration or greatest ecotoxicological risk. Regeneration of the column using a 5%

54

(m/m) NaCl, equal-volume water–methanol solution allowed for repeated use of the

AER.

55

Table 2-1. Composition of synthetic fresh and ureolyzed urine used in ion-exchange experiments.

Chemical (mmol/L) Fresh urine Ureolyzed urine

Urea as N 500 – NaCl 44 60 Na2SO4 15 15 KCl 40 40 NH4OH – 250 NaH2PO4 20 14 NH4HCO3 – 250 MgCl2·6H2O 4 – CaCl2·2H2O 4 – pH 6 9 Ionic strength (mol/L)a 0.15 0.47 a Calculated using Visual MINTEQ, version 3.0

56

Table 2-2. Continuous-flow column ion-exchange of DCF, IBP, KTP, and NPX onto Dowex 22 AER followed by in-column regeneration over three treatment–regeneration cycles.

Removed from urine (µmol)

Sorbed onto resin (µmol)

Desorbed from resin (µmol)

Remaining on resin (µmol) % Regeneration

Diclofenac cycle 1 23.2 23.2 21.7 1.5 93% cycle 2 9.2 10.7b 11.5a 0.0 100% cycle 3 6.8 6.8c 9.3a 0.0 100% Ibuprofen cycle 1 3.8 3.8 2.4 1.4 64% cycle 2 1.7 3.0b 0.7 2.4 22% cycle 3 0.5 2.9c 0.1 2.8 2% Ketoprofen cycle 1 8.1 8.1 6.0 2.1 74% cycle 2 4.7 6.8b 2.8 4.0 41% cycle 3 3.2 7.2c 1.7 5.5 23% Naproxen cycle 1 9.2 9.2 9.9a 0.0 100% cycle 2 3.0 3.0b 4.4a 0.0 100% cycle 3 2.4 2.4c 2.5a 0.0 100% a Analyzed sample measured greater than amount exchanged onto the AER, assumed complete regeneration of pharmaceutical b Amount exchanged on the AER for cycle 2 is the summation of the amount removed from urine in cycle 2 and the amount remaining on the AER after regeneration in cycle 1 c Amount exchanged on the AER for cycle 3 is the summation of the amount removed from urine in cycle 3 and the amount remaining on the AER after regeneration in cycle 2

57

Figure 2-1. Experimental equilibrium data and isotherm models determined by

nonlinear regression of (a) diclofenac (DCF) (C0 = 3.0 µmol/L), (b) ibuprofen (IBP) (C0 = 3.6 µmol/L), (c) ketoprofen (KTP) (C0 = 7.8 µmol/L), and (d) naproxen (NPX) (C0 = 7.5 µmol/L) using Dowex 22 AER. Figure (d) *Naproxen illustrates the plotted experimental isotherms excluding the lowest resin dose of 0.16 mL/L) (i.e. excluding the data point with the highest Ce and corresponding nonlinear isotherm models (Freundlich, Langmuir, Dubinin-Astakhov (D-A), and Dubinin-Radushkevich (D-R)).

0

0.005

0.01

0.015

0.02

0 0.002 0.004 0.006 0.008

qe

, m

mo

l/g

Ce, mmol/L

(a) Diclofenac

DCFLangmuirFreundlichD-AD-R

0

0.005

0.01

0.015

0.02

0 0.002 0.004 0.006 0.008

qe

, m

mo

l/g

Ce, mmol/L

(b) Ibuprofen

IBPLangmuirFreundlichD-AD-R

0

0.005

0.01

0.015

0.02

0 0.002 0.004 0.006 0.008

qe

, m

mo

l/g

Ce, mmol/L

(c) Ketoprofen

KTPLangmuirFreundlichD-AD-R

0

0.005

0.01

0.015

0.02

0 0.002 0.004 0.006 0.008

qe

, m

mo

l/g

Ce, mmol/L

(d) *Naproxen

NPXLangmuirFreundlichD-AD-R

58

Figure 2-2. Comparison of pharmaceutical removal when present individually or

combined in ureolyzed urine for (a) diclofenac, (b) ibuprofen, (c) ketoprofen, and (d) naproxen ion-exchange by Dowex 22 AER.

0%

20%

40%

60%

80%

100%

0.16 2.12 4.08 6.04 8

% R

em

oval

Resin dose, mL/L

(a) Diclofenac IndividualCombinedIndividual, C0 = 3.0 µmol/LCombined, C0 = 3.5 µmol/L

0%

20%

40%

60%

80%

100%

0.16 2.12 4.08 6.04 8

% R

em

oval

Resin dose, mL/L

(b) Ibuprofen IndividualCombinedIndividual, C0 = 3.6 µmol/LCombined, C0 = 4.7 µmol/L

0%

20%

40%

60%

80%

100%

0.16 2.12 4.08 6.04 8

% R

em

oval

Resin dose, mL/L

(c) Ketoprofen IndividualCombinedIndividual, C0 = 7.8 µmol/L Combined, C0 = 7.3 µmol/L

0%

20%

40%

60%

80%

100%

0.16 2.12 4.08 6.04 8

% R

em

oval

Resin dose, mL/L

(d) Naproxen IndividualCombinedIndividual, C0 = 7.5 µmol/L Combined, C0 = 7.4 µmol/L

59

Figure 2-3. Column saturation curves of Dowex 22 AER by pharmaceutical mixture of

(a) diclofenac (DCF), (b) ibuprofen (IBP), (c) ketoprofen (KTP), and (d) naproxen (NPX) over three treatment–regeneration cycles with fresh AER (cycle 1) and regenerated AER (cycles 2 and 3).

0

0.2

0.4

0.6

0.8

1

1.2

0 5000 10000 15000

C/C

0

Bed Volume

(a) Diclofenac

Cycle 1Cycle 2Cycle 3

0

0.2

0.4

0.6

0.8

1

1.2

0 5000 10000 15000

C/C

0

Bed Volume

(b) Ibuprofen

Cycle 1Cycle 2Cycle 3

0

0.2

0.4

0.6

0.8

1

1.2

0 5000 10000 15000

C/C

0

Bed Volume

(c) Ketoprofen

Cycle 1Cycle 2Cycle 3

0

0.2

0.4

0.6

0.8

1

1.2

0 5000 10000 15000

C/C

0

Bed Volume

(d) Naproxen

Cycle 1Cycle 2Cycle 3

60

CHAPTER 3 FIXED BED MODELING OF NONSTEROIDAL ANTI-INFLAMMATORY DRUG

REMOVAL BY ION-EXCHANGE IN SOURCE SEPARATED URINE: MASS REMOVAL OR TOXICITY REDUCTION?

Application of Bioassays and Modeling to Assess Pharmaceutical Ecotoxicity

Approximately 50–100% of a consumed dose of nonsteroidal anti-inflammatory

drugs (NSAIDs) are excreted in urine as the parent compound and metabolites

(Houghton et al. 1984, Lienert et al. 2007b, Sawchuk et al. 1995, Sugawara et al. 1978).

Conventional wastewater treatment is ineffective at removing these compounds, and is

considered a major point source of pharmaceutical discharge in the environment (Petrie

et al. 2015, Verlicchi et al. 2012). Furthermore, ibuprofen, diclofenac, and their

metabolites have been identified as having the highest potential ecotoxicological risk out

of 42 pharmaceuticals from 27 therapeutic groups (Lienert et al. 2007b). Urine source

separation has been proposed as an effective method to target these compounds for

more efficient removal, as opposed to centralized wastewater treatment where urine is

diluted by a factor of 100 (Lamichhane 2013, Larsen and Gujer 1996b). In addition to

pharmaceuticals, urine is high in nitrogen and phosphorus which may be utilized as an

alternative fertilizer (Kirchmann and Pettersson 1995). Therefore, effective separation of

pharmaceuticals from nutrients is necessary to produce a “contaminant free” fertilizer

product. From previous research, ion-exchange treatment of source separated urine is

an effective method to selectively remove NSAIDs with no co-removal of nutrients

(Landry and Boyer 2013, Landry et al. 2015). However, the work by Landry and Boyer

(2013) and Landry et al. (2015) primarily focused on ion-exchange of parent

compounds, and no research has been done evaluating pharmaceutical metabolite

removal. This is important because pharmaceuticals are primarily excreted as

61

metabolites, some of which may induce a response, or may be converted back to the

parent compound (Moser et al. 1990, Upton et al. 1980). Practical operation of sorption

processes is usually performed under continuous-flow conditions where concentration

profiles vary in space and time (Alberti et al. 2012). Although isotherm modeling

provides information describing how pollutants interact with sorbent materials (e.g.,

sorption mechanisms, surface properties, selectivity), these experiments are performed

under batch conditions at equilibrium (Foo and Hameed 2010). Column modeling is

commonly used to describe breakthrough curves which are influenced by equilibrium

isotherms, and individual transport processes in the column and sorbent (Alberti et al.

2012). However, mass removal alone is inadequate at evaluating pharmaceutical risk

and the efficacy of using sorption processes to reduce ecotoxicity potential is unknown.

Recently, there has been a paradigm shift in toxicity testing towards in vitro cell-based

and cell-free bioassays to rapidly assess efficacy of water quality treatment processes

(Escher et al. 2013).

Sorption processes are an attractive treatment method for pharmaceutical

removal in urine because it is low energy and has low environmental impact compared

to conventional wastewater treatment (Landry and Boyer 2016). Common

configurations include continuous flow batch reactors and fixed-bed columns

(Crittenden et al. 2012). Most sorption studies include kinetic and equilibrium batch

data, as well as fixed-bed column studies. The equilibrium and kinetic data obtained

from batch tests and fixed-bed configurations are the same, therefore these intrinsic

properties (e.g., surface and film diffusion coefficients) may be used to predict sorption

behavior under both conditions (Chu 2010). Several models have been developed to

62

predict sorption behavior (Xu et al. 2013). One such model is the homogenous surface

diffusion model (HSDM), which requires liquid-phase and intraparticle-phase mass

transfer coefficients, and isotherm parameters which may be determined from kinetic

and equilibrium batch data (Xu et al. 2013, Zhang et al. 2009). This type of predictive

modeling is useful for evaluating fixed-bed behavior under varying conditions, such as

empty bed contact time. From a practical standpoint, batch kinetic and equilibrium tests

are rapid and require limited materials. This is especially useful when evaluating

removal of emerging contaminants, such as pharmaceuticals, which are present at low

concentrations. Conducting long-term column experiments would require large volumes

of urine, synthetic or real, and high material costs particularly for pharmaceutical

metabolites.

Several researchers have developed various batteries of assays to evaluate the

efficacy of treatment methods to reduce pharmaceutical ecotoxicity in water and source

separated urine. For example, Escher et al. (2006) evaluated pharmaceutical ecotoxicity

in source separated urine using bioassays to detect baseline toxicity (i.e., chlorophyll

fluorescence test), estrogenic endocrine disruption (i.e., yeast estrogen screen), and

genotoxicity (i.e., umu test) after urine was treated using various advanced processes.

To our knowledge, no research has been done evaluating pharmaceutical toxicity

reduction using ion-exchange in source separated urine. Furthermore, Escher et al.

(2013) evaluated 103 in vitro bioassays to benchmark organic micropollutants in water,

wastewater, and reclaimed water and found that xenobiotic metabolism, hormone-

related modes of action, genotoxicity, and adaptive stress response were the most

responsive health-related endpoints. However, COX inhibition was not included in this

63

study. Nishi et al. (2010) evaluated NSAID ecotoxicity of surface water and wastewater

using an in vitro cyclooxygenase (COX) inhibition bioassay, which is the primary mode

of action of NSAIDs, and a dose-response relationship was observed between COX

inhibition and NSAID distribution. The cyclooxygenase enzyme has two subtypes, COX-

1 and COX-2. Inhibition of the COX-2 enzyme is attributed to the anti-inflammatory

effects of NSAIDs (Blobaum and Marnett 2007). Inhibition of the COX-1 enzyme, which

is associated with normal cellular homeostasis, has been attributed to aquatic toxicity

including gastrulation arrest and defective vascular tube formation in zebrafish, and

reproductive issues in Japanese medaka (Cha et al. 2005, Lee et al. 2011). For this

reason, inhibition of COX-1 was the mode of action evaluated in this study. The benefit

of using cell-based bioassays is that they evaluate the potential for adverse effect.

Cellular response is one aspect of taking a systems-level approach to assess whole

organism and population response (Julia and Portier 2007).

This study combined predictive column modeling with in vitro bioassays to

provide a preliminary assessment of fixed-bed NSAID ion-exchange removal to reduce

toxic potential. The goal of this research was to develop a systematic approach to

evaluate the ion-exchange removal of pharmaceutical parent compounds and

pharmaceutical metabolites in urine and evaluate the corresponding reduction in

ecotoxicity utilizing the entire dose-response curve through three main objectives: (1)

compare COX-1 inhibition and mass removal for individual compounds, (2) compare

COX-1 inhibition and mass removal for a pharmaceutical mixture, and (3) compare the

effect of urine matrices on pharmaceutical ion-exchange removal.

64

Materials and Methods

Pharmaceutical and Pharmaceutical Metabolites

The chemical characteristics of the pharmaceutical parent compounds and

respective metabolites investigated in this work are listed in Table B-1. Diclofenac

sodium (DCF) (CAS 15307-79-6), ibuprofen sodium (IBP) (CAS 31121-93-4),

ketoprofen (KTP) (CAS 22071-15-4), and naproxen sodium (NPX) (CAS 26159-54-2)

are all weakly acidic pharmaceuticals from the NSAID class. A primary metabolite of

each parent compound were also investigated. 4’-OH-diclofenac (OH-DCF) (CAS

64118-84-9), hydroxy ibuprofen (OH-IBP) (CAS 53949-53-4), ketoprofen acyl

glucuronide (KTP-gluc) (CAS 76690-94-3), and O-desmethylnaproxen (Odm-NPX)

(CAS 52079-10-4). All metabolites were purchased from Toronto Research Chemicals

and all pharmaceutical parent compounds were purchased from Sigma Aldrich.

Separate stock solutions were made by dissolving each compound in methanol.

Synthetic and Real Urine

Synthetic ureolyzed (i.e., aged) urine was made according to a previously

described method and adjusted to include the six major endogenous metabolites found

in human urine (Table B-2) (Landry et al. 2015). Pharmaceutical parent compounds and

metabolites were spiked individually in ureolyzed urine at an initial concentration of

1,000 µg/L. The same bulk solution of ureolyzed urine was used for both the kinetic test

and equilibrium test of the respective compounds. Real ureolyzed urine was collected

from one male and one female. The total organic carbon (TOC) concentration and

conductivity are shown in Table B-2.

65

Anion Exchange Resin

Dowex 22 strong-base, polymeric anion exchange resin (AER) was used for all

batch kinetic and equilibrium experiments. This resin is a macroporous AER

functionalized with dimethylethanolamine functional groups. The AER was

preconditioned using NaCl, and dried following a previously described method (Landry

and Boyer 2013).

Pharmaceutical Concentrations in Urine

Pharmaceutical parent compound and metabolite concentrations in urine were

estimated in urine following a previously described method (Landry et al. 2015).

Detailed methodology may be found in Appendix B. Table 3-1 lists the excretion rates

and estimated parent compound and metabolite concentrations in urine.

Toxicity Bioassays

Cyclooxygenase (COX) inhibiting activity was measured using a COX

Colorimetric Inhibitor Screening Assay Kit (Cayman Chemical Co.) according to the

protocol provided by Cayman Chemical Co. COX subtype 1 (COX-1) was the only

enzyme evaluated for inhibiting activity. COX-1 enzyme was incubated with each

inhibitor for 30 min prior to plate development. Each compound was evaluated for COX-

1 inhibition at five concentration points and performed in triplicate. To evaluate COX-1

inhibition from the pharmaceutical parent compounds and pharmaceutical metabolites

only, and to avoid interference from the high concentrations of nutrients, salts, and

endogenous metabolites in synthetic urine, pharmaceutical stock solutions were diluted

in methanol for the bioassays. The concentration points were made by serial dilution

and corresponded to a 10-log concentration factor (i.e., 0.01×, 0.1×, 1×, 10×, 100×),

where 1× corresponds to the realistic concentration found in urine (Table 3-1). Effect

66

concentrations for single compound dose-response curves is listed in Table B-4. Dose-

response curves were modeled to the classic Hill equation (Eq. 3-1) using a 3-

parametric logistic regression developed by Cardillo (2012) in MATLAB (8.2.0.701

R2013b) (MathWorks 2013).

𝐼 = 𝐼0 +(𝐼𝑚𝑎𝑥−𝐼0)

1+(𝐼𝐶50

𝐶)

𝐻 (3-1)

Where I is the observed inhibition, I0 is the minimum observed inhibition, Imax is

the maximum observed inhibition, IC50 is concentration at which 50% of the COX-1

enzyme is inhibited (µmol/L), C is the inhibitor concentration (µmol/L), and H is the Hill

slope. One experiment was conducted as a mixture of DCF, KTP, KTP-gluc, NPX, and

Odm-NPX. Mixture toxicity was evaluated using the generalized concentration addition

model (Eq. 3-2) (Howard and Webster 2009).

𝐼𝑚𝑖𝑥 =𝐼𝑚𝑎𝑥𝐴𝐶𝐴 𝐼𝐶50𝐴⁄ +𝐼𝑚𝑎𝑥𝐵𝐶𝐵 𝐸𝐶50𝐵⁄ +⋯

1+𝐶𝐴 𝐸𝐶50𝐴⁄ +𝐶𝐵 𝐸𝐶50𝐵⁄ +⋯ (3-2)

Where Imix is the effect of the mixture at a specific concentration, ImaxA is the

maximum inhibition of chemical A, IC50A is the IC50 of chemical A, and CA is the

concentration of chemical A in the mixture, and so-forth for chemical B, etc. Inhibition

concentrations for the pharmaceutical mixture dose-response curves is listed in Table

B-5.

Batch Kinetic and Equilibrium Tests

Batch kinetic and equilibrium tests were performed following a previously

described method using ureolyzed urine at an initial pharmaceutical parent compound

or metabolite concentration of 1,000 µg/L (Landry and Boyer 2013). Details regarding

the experimental method are provided in Appendix B.

67

Fixed-Bed Column Modeling

The unsteady-state adsorption of pharmaceutical parent compounds and

metabolites in a fixed-bed column were predicted by the homogenous surface diffusion

model (HSDM) using the Fixed-bed Adsorption Simulation Tool (Fast 2.1beta) (Sperlich

et al. 2008). Details regarding the HSDM may be found in Appendix B.

Sample Preparation

Pharmaceutical samples from the column experiments were separated from the

urine matrix using a solid phase extraction (SPE) vacuum station (Supelco Visiprep)

and phenyl SPE columns (SiliaPrep, SiliCycle), evaporated, and reconstituted following

a previously described method (Magiera et al., 2014). The dry residue of DCF, KTP,

and NPX samples were dissolved in 1 mL of acetonitrile:10 mM K2HPO4 (pH 3) (10:90;

v/v) mobile phase and 100 µL was injected into the HPLC-UV system (Hewlett Packard

1050 series detector and Agilent 1100 series auto sampler). The dry residue of Odm-

NPX was dissolved in 1 mL of 25 mM KH2PO4 (pH 3) mobile phase and 25 µL was

injected into the HPLC-UV system.

Analytical Methods

The COX Colorimetric Inhibitor Screening Assay Kit was analyzed using

microplate reader (SpectraMax Plus 384) at 590 nm. Pharmaceutical concentrations for

the column experiments were measured using HPLC-UV (Hewlett Packard 1050 series

detector and Agilent 1100 series auto sampler) at 230 nm, equipped with a reversed-

phase column (2.1 × 150 mm, 3 μm Ascentis RP-amide column; Supelco, Bellefonte,

PA). For DCF, KTP, and NPX analysis, the mobile phase consisted of a mixture of

acetonitrile and 10 mM K2HPO4 (pH 3) (55:45 v/v). For Odm-NPX analysis, the mobile

phase consisted of a mixture of acetonitrile and 25 mM KH2PO4 (pH 3) (40:60 v/v). A

68

seven-point calibration curve (0, 50, 100, 500, 1,000, 5,000, and 10,000 µg/L) was

created by serial dilution of the stock standards. The limit of detection (LOD) was 50

μg/L. Pharmaceutical concentrations were set to the LOD if the effluent concentration

fell below the LOD. Endogenous metabolite concentrations in synthetic and real urine

for equilibrium experiments with DCF were analyzed by measuring the TOC

concentration using a Shimadzu TOC-VCPH analyzer equipped with an ASI-V

autosampler (Apell and Boyer 2010). The relative difference between all duplicate

samples was <5%. Several samples had final TOC concentrations greater than the

initial TOC concentrations, in these cases, the final concentration was set equal to the

initial concentration and yielded 0% removal. Conductivity was analyzed using an Orion

Star A212 conductivity meter, and was calibrated prior to use using three conductivity

standards (14, 50, and 100 mS/cm).

Data Analysis

Data from the toxicity bioassays and equilibrium tests were the mean value of

triplicate samples. Data from the kinetic tests were the mean value of duplicate

samples. Analysis of covariance (ANOCOVA) was conducted using MATLAB (8.2.0.701

R2013b) to determine if there was a significant difference (α = 0.05) between the slopes

of the log-log transformed isotherms (MathWorks 2013). The null hypothesis states that

there was not a significant difference between slopes (p > 0.05) and the alternative

hypothesis states that there was a significant difference between the slopes (p < 0.05).

Results and Discussion

COX-1 Inhibition for Individual Compounds

The HSDM was selected to predict fixed-bed performance of ion-exchange

removal of DCF, KTP, NPX, and Odm-NPX (Figure 3-1). The Freundlich isotherm

69

parameters used for model calibration are listed in Table B-12. The HSDM model was

also fit to existing fixed-bed column data to confirm model validity. The R2 and sum of

squares error (SSE) was 0.98 and 1.22, respectively, for the column data shown in

Figure B-1. For the column data in Figure B-2, the SSE was 38, 57, and 7.5 for DCF,

KTP, and NPX, respectively. Furthermore, the R2 was 0.48 for KTP, and 0.88 for DCF

and NPX. Broad tailing in the experimental data, particularly for DCF, may be attributed

to flow non-idealities such as column channeling (Chu 2004). Nevertheless, the HSDM

was deemed appropriate to pursue the objective of coupling toxicity reduction with ion-

exchange removal. The Freundlich isotherm parameter 1/n and Biot number, which is

the ratio of the external mass transfer rate to the intraparticle mass transfer rate, are

indicators of the controlling phase for mass transfer (Hand et al. 1984a). As the 1/n

approaches 1 and the Biot number increases, external mass transfer and intraparticle

mass transfer contribute equally to the rate of adsorption. The Biot number for DCF,

NPX, Odm-NPX were 30, 139, and 20, and 1/n values were 1.05, 0.74, 0.86,

respectively, which indicated both external and intraparticle mass transfer rates

contributed to the rate of adsorption. For irreversible isotherms, such as KTP, where 1/n

= 0, the rate of adsorption is controlled by intraparticle mass transfer (Hand et al.

1984b). The mass breakthrough curves in Figure 3-1 exhibit a broad trailing edge

possibly due to slow intraparticle diffusion within the AER pore space (Chu 2004).

Furthermore, the Freundlich isotherm parameters influence the breakthrough curve

profile (Hand et al. 1984a). In general, increasing selectivity (i.e., decreasing 1/n) or

increasing AER capacity (i.e., KF) increases the volume treated until breakthrough, and

decreases the intraparticle mass transfer rate (DS) resulting in a broad trailing edge.

70

Conversely, decreasing selectivity (i.e., increasing 1/n) or decreasing AER capacity (i.e.,

decreasing KF) decreases the volume treated until breakthrough, and increases the

intraparticle mass transfer rate (DS) resulting in a sharper trailing edge. The benefit of

predicting fixed-bed column performance is that column parameters may be optimized,

and material requirements and costs may be estimated prior to pilot or full-scale

implementation (Crittenden et al. 1987).

As stated previously, the premise of this research is that both mass removal and

toxicity potential are needed to evaluate pharmaceutical risk. To address this, an

alternative approach to evaluating the fixed-bed breakthrough was taken by converting

the commonly depicted normalized effluent concentration (i.e., C/C0) to percent COX-1

inhibition. By evaluating treatment performance as function of COX-1 inhibition, ion-

exchange performance may be compared to the entire dose-response curve and used

as a decision tool to establish treatment objectives. The absolute effluent concentrations

(i.e., µmol/L) from the breakthrough curves for DCF, KTP, NPX, and Odm-NPX were

transformed to COX-1 inhibition using the Hill parameters from the dose-response

curves (Table B-13, Figure B-3). Figure 3-1 shows the simultaneous mass removal

predicted from the HSDM and COX-1 inhibition as a function of treated bed volumes

(BV) of urine. The expected COX-1 inhibition of untreated urine, based on the predicted

pharmaceutical concentrations in urine (Table 3-1), followed a decreasing trend of DCF

(74%) > KTP (51%) > NPX (26%) > Odm-NPX (20%) (Figure 3-1). Using the IC10 (i.e.,

pharmaceutical concentration corresponding to 10% COX-1 inhibition) as the treatment

criteria (i.e., breakthrough), 616 and 209 BV of synthetic urine containing DCF and KTP,

respectively, may be treated before reaching breakthrough. Although DCF was more

71

active than KTP (see IC50 values in Table B-13), the AER had a greater capacity for

DCF compared with KTP so a larger volume of urine may be treated before COX-1

inhibition by DCF is detected in the effluent. This demonstrates that although a

pharmaceutical may not be as active, less effective mass removal may induce greater

ecotoxicity potential. Furthermore, the IC10 breakthrough point corresponded to 96%

DCF mass removal and 89% KTP mass removal suggesting that stringent treatment

objectives (i.e., complete mass removal) may not be necessary to achieve effective

reduction in ecotoxicity potential. Although complete removal (i.e., C/C0 ≈ 0) was

achieved for NPX and Odm-NPX only, COX-1 inhibition was only reduced from 26%

and 20% in untreated urine to 20% and 13%, respectively. This may be attributed to the

dose-response curves which had I0 values of 20% and 13%, respectively, which

suggests that targeting these compounds for removal may not significantly improve

urine quality with respect to COX-1 inhibition. However, the maximum response for NPX

and Odm-NPX did not reach 100% for either compound. When the Hill model was

adjusted to force the minimum and maximum response to 0% and 100%, respectively,

the IC50 was 132 µmol/L and 416 µmol/L for NPX and Odm-NPX, respectively (see

Table B-14 and Figure B-4), which was more consistent with literature (Davies and

Anderson 1997). For the alternate breakthrough curves (Figure B-5), 60 and 550 BV of

urine may be treated before NPX and Odm-NPX reach breakthrough, respectively.

COX-1 Inhibition Mixture Effects

Realistically, NSAIDs are present in urine as a mixture and at varying

concentrations. The generalized concentration addition (GCA) model was used to

predict mixture effects. The benefit of using the GCA model is that individual dose-

response curves may be used to predict mixture response for multiple pharmaceuticals

72

(Howard and Webster 2009). As shown in Figure B-3 and Figure B-5, DCF, KTP, KTP-

gluc, NPX, and Odm-NPX inhibited the COX-1 enzyme to different extents. However,

Ibuprofen, 4-OH diclofenac, and OH-ibuprofen did not inhibit COX-1 enzyme at any

pharmaceutical dose (Figure B-6). For brevity, results and discussion of this paper will

focus on pharmaceuticals that inhibit COX-1 enzyme. The IC50 values for investigated

pharmaceuticals followed the order of increasing magnitude of DCF (0.24 µmol/L) <

KTP (1.30 µmol/L) < Odm-NPX (4.13 µmol/L) < NPX (16.8 µmol/L) < KTP-gluc (73.1

µmol/L) (Table B-13). This trend is consistent with other literature, where IC50 for NPX is

two orders of magnitude greater than DCF and KTP (Cryer and Feldman 1998). The

NPX metabolite, Odm-NPX was more active than the parent compound, based on the

IC50 value.

The Hill parameters for individual COX-1 inhibition curves (Table B-13) were

used to evaluate COX-1 inhibition for a pharmaceutical mixture containing DCF, KTP,

KTP-gluc, NPX, and Odm-NPX (Figure 3-2). The estimated COX-1 inhibition of

untreated urine for the pharmaceutical mixture was 63%. The GCA model adequately

predicted pharmaceutical mixture toxicity, with an R2 of 0.98, although it slightly

overestimated the expected COX-1 inhibition in urine to be 75%. Total excretion for

NSAIDs, including parent compounds and metabolites, range from 50%–100%. As

many as five metabolites may be excreted, however only one metabolite with the

highest excretion was evaluated in this study. For example, only 6.4% of KTP is

excreted in urine unchanged and 52.8% is excreted as the KTP glucuronic acid

conjugate (KTP-gluc) (Table 3-1) (Houghton et al. 1984). However, glucuronic acid

conjugates have been shown to be highly unstable in urine and rapidly hydrolyze back

73

to the parent compound (Upton et al. 1980). This suggests that the concentration of

KTP in urine may be much greater than what was estimated in urine based on

excretion. The expected COX-1 inhibition of KTP-gluc in urine was 2%. However, if

KTP-gluc was completely hydrolyzed back to KTP in ureolyzed urine, the predicted

COX-1 inhibition due to KTP would increase from 51% to 83%, and increase COX-1

inhibition for the pharmaceutical mixture from 75% to 91%. This is demonstrated by the

shift in the GCA model in Figure 3-2.

The influent concentration of NSAIDs at varying concentrations will influence

both fixed-bed performance and expected COX-1 inhibition. Furthermore, the effluent

concentration of each NSAID constantly changes as a function of bed volume until the

resin is fully saturated. The GCA model was used to predict COX-1 inhibition as a

function of bed volume for a pharmaceutical mixture containing DCF, KTP, NPX, and

Odm-NPX (Figure 3-3). Approximately 210 BV of urine may be treated before reaching

breakthrough. Ketoprofen was the greatest contributor to COX-1 inhibition for the

pharmaceutical mixture at breakthrough. Furthermore, if KTP-gluc hydrolyzed back to

KTP, increasing the initial concentration of KTP in urine, breakthrough would decrease

to <25 BV (Figure 3-3). However, a resin with higher capacity for KTP, would allow a

larger volume of urine to be treated before breakthrough. The instability of acyl

glucuronide metabolites provides insight into the practical application of urine source

separation. Pharmaceutical removal under fresh urine conditions may be less effective

at removing acyl glucuronides due to their hydrophilic nature, and the remaining acyl

glucuronides in treated urine may hydrolyze back to the parent compound. This

suggests that pharmaceutical removal may be more effective under ureolyzed urine

74

conditions after acyl glucuronide metabolites hydrolyze back to the parent compound.

Breakthrough of a pharmaceutical mixture may be used to estimate the operation

requirements (e.g., resin volume) and costs to effectively reduce COX-1 inhibition.

Evaluating treatment efficacy in terms of COX-1 inhibition for the pharmaceutical

mixture holistically synthesizes the concurrent relationships between varying

pharmaceutical concentrations in urine, pharmaceutical mixture toxicity, and resin-

pharmaceutical interactions.

Evaluating toxic response of NSAIDs may not be limited to only COX-1 inhibition.

The ToxCast database developed by the EPA evaluated >800 in vitro endpoints for

>2,000 chemicals. For example, DCF and IBP induced a response in 48 and 17

bioassays, respectively, with biological endpoints ranging from cell death, regulation of

gene expression, and receptor binding, to name a few (U.S. EPA 2016c). Figure B-7 is

a graphical depiction of the AC50 (i.e., concentration that induces 50% activity) of the in

vitro bioassays with various endpoints that induce a response from exposure to DCF

and IBP. For DCF, the COX-1 bioassay may be considered a protective assay because

it is more sensitive than other endpoints evaluated. On the other hand, IBP did not

induce COX inhibition, however alternative in vitro bioassays such as the nuclear

receptor assay ATG_ERE_CIS_up may be utilized to evaluate the estrogen response

(U.S. EPA 2016c).

Linking in vitro assays to long term in vivo outcomes is difficult due to the

complex molecular, cellular, and tissue changes from the biological target to adverse

outcomes (Liu et al. 2015). Table B-15 lists the EC50 values for in vivo chronic

ecotoxicity studies from literature. In general, the COX-1 bioassay was more sensitive

75

than the in vivo studies, with the exception of M. galloprovincialis larvae development

when exposed to DCF (Fabbri et al. 2014). This suggests that although the in vitro

bioassay may detect COX-1 inhibition activity, it may not elicit a toxic response in

aquatic life due to repair and defense mechanisms that may prevent toxicity (Escher et

al. 2013). The development of adverse outcomes pathways (AOPs) is a framework that

links molecular-level changes in an organism with adverse outcomes such as survival,

growth, and reproduction (Schroeder et al. 2016). For example, one AOP of

cyclooxygenase inhibition is decreased ovulation and reduced reproductive success

leading to a decline in population (AOPWiki 2016). Efforts to use high-throughput

assays to predict in vivo response is an active area of research. When comparing

estrogenic activity of wastewater using the in vitro yeast estrogen screen assay and in

vivo vitellogenin assay, Huggett et al. (2003) found that the in vivo assay had 10-fold

greater estrogenic activity than the in vitro assay. Furthermore, researchers have

utilized ToxCast, a database containing 1,057 chemicals and >800 in vitro endpoints,

and the Toxicity Reference Database containing in vivo chronic toxicity data to develop

predictive toxicity models including rat reproductive toxicity, hepatotoxicity, estrogenic

activity, and prenatal developmental toxicity (Liu et al. 2015, Martin et al. 2011, Rotroff

et al. 2014, Sipes et al. 2011).

Comparison of Urine Matrices

As shown in Figure B-8, approximately 3.4× more synthetic urine than real urine

may be treated before DCF reaches breakthrough, this suggests that the presence of

endogenous metabolites in real urine may be interfering with DCF ion-exchange.

Equilibrium pharmaceutical removal was conducted in synthetic ureolyzed urine

containing six endogenous metabolites present at the greatest concentrations in urine.

76

This data was compared to previously conducted equilibrium experiments in synthetic

urine in the absence of endogenous metabolites (Landry et al. 2015). Furthermore, an

equilibrium experiment was performed using real ureolyzed human urine spiked with

DCF. As shown in Figure 3-4 and Figure B-9, the presence of endogenous metabolites

in synthetic urine reduced the ion-exchange capacity and removal efficiency. At a resin

dose of 2 mL/L, DCF removal decreased from 89% in synthetic urine without

metabolites to 74% in synthetic urine with metabolites, and further decreased to 32% in

real urine. A similar trend was observed at the 4 mL/L resin dose, however at the 8

mL/L resin dose, diclofenac removal was 95%, 91%, and 97% in synthetic urine with

and without metabolites, and real urine, respectively. A reduction in color and

discoloration of the AER was observed visually with increasing AER dose. It was

hypothesized that endogenous metabolites were competing for ion-exchange sites on

the AER. To confirm this, DCF samples from experiments using synthetic urine with

metabolites and real human urine were analyzed for total organic carbon (TOC) to

estimate endogenous metabolite removal. As shown in Figure 3-5, the mass of TOC

(mg as C) removed from synthetic and real urine increased with increasing resin dose.

Furthermore, the TOC content due to endogenous metabolites was 3,200× and 27,000×

greater than the pharmaceutical content in synthetic and real urine, respectively. The

metabolites present in real urine was 2.7× greater than the concentration (mg C/L) in

synthetic urine (Table B-2). Similar competition was observed for micropollutant

adsorption in the presence of natural organic matter during drinking water treatment,

which is present at much higher concentrations than micropollutants (Worch 2012). Ion-

exchange removal of NSAIDs in urine is due to the electrostatic interactions between

77

the negatively charged functional group of the pharmaceutical and positively charged

quaternary ammonium functional group of the AER, and van der Waals interactions

between the aromatic ring structure between the pharmaceutical and AER (Landry et al.

2015). In addition to being primarily negatively charged or neutral, endogenous

metabolites have an aliphatic or aromatic organic structure (Bouatra et al. 2013). Thus,

it is reasonable to expect that negatively charged endogenous metabolites with an

aromatic ring structure would compete with pharmaceuticals for ion-exchange sites on

the resin due to favorable van de Waals interactions between the metabolites and AER.

However, removal of positively charged pharmaceuticals by a cation exchange resin,

such as citalopram, may experience less competition for ion-exchange sites on the resin

due to unfavorable electrostatic interactions with negatively charged endogenous

metabolites (Solanki and Boyer 2017).

Synthetic urine has been used in several urine source separation studies for

nutrient recovery and pharmaceutical removal. Tarpeh et al. (2017) observed no

significant difference between ammonium adsorption by cliniptilolite zeolite, a

polyacrylic cation exchange resin, or a polystyrene cation exchange resin in synthetic

and real urine. Minimal impact between urine compositions may be because most

endogenous metabolites are negatively charged or neutral in ureolyzed urine, thus lack

the necessary electrostatic interactions for cation exchange removal (Bouatra et al.

2013). Precipitation processes, such as struvite, are driven by supersaturation of the

respective inorganic compounds (e.g., Mg+2, PO4–3, and NH4

+) which is dependent on

their concentration in urine (Udert et al. 2003b). During the nucleation step, organic

compounds can adsorb to the crystals and inhibit further precipitation (Lin et al. 2005,

78

Sindelar et al. 2015). The presence of endogenous metabolites in urine slightly reduced

the amount of struvite precipitated but decreased the rate of precipitation by a factor of

4 (Udert et al. 2003a, Udert et al. 2003b). Conversely, Pronk et al. (2006b) found an

increase in pharmaceutical retention during nanofiltration of real urine compared with

synthetic due to complexation of pharmaceuticals with endogenous metabolites,

changes in surface charge and/or membrane fouling due to endogenous metabolites.

suggests that synthetic urine may or may not be an adequate proxy for evaluating urine

source separation processes. In general, the presence of endogenous metabolites

appears to least impact nitrogen cation exchange and slightly impact struvite

precipitation and pharmaceutical removal by membrane processes. However, favorable

interactions between endogenous metabolites and AER significantly impacts removal of

negatively charged pharmaceuticals.

The competitive effects of organic metabolites on ion-exchange removal of

NSAIDs highlights the need to evaluate alternative sorbents that have higher selectivity

or capacity. The AER used in this study is a commercially available material, however,

sorbent material designed to selectively remove target compounds may improve

pharmaceutical removal in urine. For example, molecularly imprinted polymers (MIPs)

have been used extensively as extraction methods for sample analysis (Beltran et al.

2010), including selective extraction of naproxen in urine (Caro et al. 2004). Studies

have also shown that MIP adsorption may be used to selectively remove >90% of

NSAIDs from surface water (Dai et al. 2012). Alternatively, an adsorbent with much

higher capacity and similar selectivity would increase pharmaceutical removal in the

presence of endogenous metabolites.

79

Concluding Remarks

This study utilized a high-throughput in vitro bioassay to evaluate the treatment

efficacy of ion-exchange resins to remove pharmaceuticals in source separated urine

and reduce ecotoxicity potential. Evaluating breakthrough curves as a function of

toxicity as opposed to mass removal provides a better understanding of treatment

objectives for emerging contaminants, such as pharmaceuticals. For example,

increasing mass removal of naproxen and O-desmethylnaproxen did not necessarily

reduce ecotoxicity potential due to the dose-response behavior. Mass removal of 89%

for KTP and 96% for DCF corresponded with 90% reduction in COX-1 inhibition. This

demonstrates that complete removal (i.e., 0% mass breakthrough) may not be

necessary to achieve an effective reduction in ecotoxicity potential. Furthermore, KTP

was less active than DCF but because the AER had a lower capacity for KTP, it

reached breakthrough (i.e., IC10) sooner than DCF. The generalized concentration

addition model may be used to predict COX-1 inhibition as a function of bed volumes

treated for a pharmaceutical mixture with varying concentrations and mass removal

efficacy. Due to a lack of regulatory framework for pharmaceutical treatment guidelines,

treatment efficacy for emerging contaminants should include toxicity reduction as well

as mass removal. Furthermore, in vitro dose-response curves provide a unique

opportunity to evaluate treatment performance to various pharmaceuticals and toxicity

endpoints. However, linking in vitro bioassays to in vivo effects is a growing research

area. Utilizing kinetic and equilibrium tests to predict fixed-bed breakthrough is a rapid

way to generate data which will provide insights on the process, and compare

pharmaceutical sorption performance under varying conditions. Lastly, human urine

contains a complex mixture of heterogeneous endogenous metabolites that may

80

compete for ion-exchange sites on the resin. More selective or higher capacity resins

may improve the efficacy of using sorption technologies to remove pharmaceuticals

from urine.

81

Table 3-1. Estimated active ingredient (AI) and metabolite concentrations in urine and fraction excreted in urine.

Compound Concentration in urine, µg/L (µmol/L) Fraction of dose excreted in urine

Diclofenac 174 (0.547)a 0.06 b

4’-OH-diclofenac 456 (1.46) 0.16 b

Ibuprofen 2,409 (10.6)a 0.07 c

Hydroxy ibuprofen 5,697 (25.6) 0.17 c

Ketoprofen 342 (1.35)a 0.064 d Ketoprofen acyl glucuronide 4,777 (11.1) 0.528 d

Naproxen 758 (3.01)a 0.013 e

O-Desmethylnaproxen 300 (1.39) 0.006 e

a Average concentration from Table B-3 b Sawchuk et al. (1995) c Lienert et al. (2007b) d Houghton et al. (1984) e Sugawara et al. (1978)

82

Figure 3-1. Predicted column breakthrough curves as a function of mass removal and

COX-1 inhibition for (a) diclofenac (C0 = 0.55 µmol/L), (b) ketoprofen (C0 = 1.3 µmol/L), (c) naproxen (C0 = 3.0 µmol/L), and (d) O-desmethylnaproxen (C0 = 1.4 µmol/L).

0%

20%

40%

60%

80%

100%

0 1000 2000 3000 4000Bed Volume

(a)

C/C0×100%

% Inhibition

0%

20%

40%

60%

80%

100%

0 200 400 600 800 1000

Bed Volume

(b)

C/C0×100%% Inhibition

0%

20%

40%

60%

80%

100%

0 200 400 600 800 1000Bed Volume

(c)

C/C0×100%

% Inhibition

0%

20%

40%

60%

80%

100%

0 1000 2000 3000Bed Volume

(d)

C/C0×100%

% Inhibition

83

Figure 3-2. Cyclooxygenase subtype-1 inhibition curve for a pharmaceutical mixture containing diclofenac, ketoprofen, ketoprofen glucuronide, naproxen, and o-desmethylnaproxen. The dashed line represents the GCA model for the pharmaceutical mixture, and the dotted line represents the GCA model for the pharmaceutical mixture assuming ketoprofen glucuronide completely hydrolyzed back to the parent compound. The symbols are the mean triplicate samples with error bars showing one standard deviation.

-20%

0%

20%

40%

60%

80%

100%

% C

OX

-1 Inhib

itio

n

Concentration, µmol/L

MixHillGCAGCA, KTP-gluc hydrolyzed

10–2 100 102 104

84

Figure 3-3. Predicted column breakthrough curves as a function of mass removal and

COX-1 inhibition for a pharmaceutical mixture containing (a) diclofenac (C0 = 0.55 µmol/L), ketoprofen (C0 = 1.3 µmol/L), naproxen (C0 = 3.0 µmol/L), and O-desmethylnaproxen (C0 = 1.4 µmol/L), and (b) diclofenac (C0 = 0.55 µmol/L), ketoprofen (C0 = 12.4 µmol/L), naproxen (C0 = 3.0 µmol/L), and O-desmethylnaproxen (C0 = 1.4 µmol/L). In figure (b), ketoprofen glucuronide was assumed to be hydrolyzed back to ketoprofen. The mass removal curve is a summation of the molar mass removal normalized by the total concentration.

0%

20%

40%

60%

80%

100%

0 1000 2000 3000

Bed Volumes

(a)

% Inhibition

C/C0×100%

0%

20%

40%

60%

80%

100%

0 1000 2000 3000

Bed Volumes

(b)

% Inhibition

C/C0×100%

85

Figure 3-4. Ion-exchange removal in real urine and synthetic urine with and without

metabolites of (a) diclofenac, (b) ibuprofen, (c) ketoprofen, and (d) naproxen. Data without metabolites reproduced from Landry et al. (2015). The symbols are the mean triplicate samples with error bars showing one standard deviation.

0

0.2

0.4

0.6

0.8

1

0 2 4 6 8 10

C/C

0

Resin dose, mL/L

(a) No metabolites, C0 = 3.0 µMMetabolites, C0 = 2.2 µMReal urine, C0 = 0.71 µM

0

0.2

0.4

0.6

0.8

1

0 2 4 6 8 10

C/C

0

Resin dose, mL/L

(b) No metabolites, C0 = 3.6 µMMetabolites, C0 = 3.2 µM

0

0.2

0.4

0.6

0.8

1

0 2 4 6 8 10

C/C

0

Resin dose, mL/L

(c) No metabolites, C0 = 7.8 µM

Metabolites, C0 = 3.0 µM

0

0.2

0.4

0.6

0.8

1

0 2 4 6 8 10

C/C

0

Resin dose, mL/L

(d) No metabolites, C0 = 7.5 µMMetabolites, C0 = 3.7 µM

86

Figure 3-5. Mass of endogenous metabolites (TOC) removed (mg C) during equilibrium experiments for synthetic urine with metabolites and real urine. The sample volume was 125 mL. The bars are the mean triplicate samples with error bars showing one standard deviation.

87

CHAPTER 4

LIFE CYCLE ASSESSMENT AND COSTING OF URINE SOURCE SEPARATION: FOCUS ON NONSTEROIDAL ANTI-INFLAMMATORY DRUG REMOVAL‡

Application of Life Cycle Assessment for Pharmaceutical Treatment

Approximately 64% of ingested pharmaceuticals intended for human use are

excreted in urine as the parent compound or metabolites (Lienert et al. 2007a). Human

urine is the primary contributor of pharmaceuticals in municipal wastewater but only

constitutes 1% of the total volumetric flow (Joss et al. 2005, Larsen and Gujer 1996a).

Urine source separation has been proposed as a more efficient method to remove

and/or destroy pharmaceuticals as opposed to centralized biological wastewater

treatment because pharmaceuticals are present at much higher concentrations in

undiluted urine (Lamichhane and Babcock 2012). In addition, human urine contributes

80% of the nitrogen (N) and 50% of the phosphorus (P), indicating separate treatment

of urine may have significant impacts on centralized wastewater treatment (Larsen and

Gujer 1996a). Furthermore, human urine may be utilized as an alternative fertilizer

source because N and P are essential nutrients used in agriculture (Kirchmann and

Pettersson 1995). Therefore, effective separation of pharmaceuticals from nutrients is

necessary to obtain a “contaminant free” nutrient product. From previous research,

sorption processes are an effective method to selectively remove nonsteroidal anti-

inflammatory drugs (NSAIDs) from urine with minimal co-sorption of nutrients, and may

be effectively regenerated using a 5% sodium chloride, 50% methanol solution (Landry

‡ Reproduced with permission from Landry, K.H., Boyer, T.H. 2016. Life cycle assessment and costing of urine source separation: Focus on nonsteroidal anti-inflammatory drug removal. Water Research 105,

487–495, DOI: http://dx.doi.org/10.1016/j.watres.2016.09.024. Copyright 2016 Elsevier Ltd.

88

and Boyer 2013, Landry et al. 2015). Furthermore, sorption is attractive for

pharmaceutical removal because it is low energy and has different treatment

configurations such as fixed-bed or mixed reactors, continuous flow or batch, and

sorbent regeneration or disposal (Crittenden et al. 2012). The basis of this research is

that removing pharmaceuticals from undiluted urine would be more effective and

efficient than in centralized wastewater, particularly for pharmaceuticals primarily

excreted in urine (Lienert et al. 2007a, Winker et al. 2008a), however, the environmental

impacts of using sorption processes to remove pharmaceuticals in urine is unknown.

Life cycle assessment (LCA) applied to urine source separation is an emerging

research area with only one study considering removal of pharmaceuticals. The primary

focus of several papers included the source separation system (i.e., urine piping,

collection, and storage), fertilizer offsets, wastewater treatment offsets, and potable

water offsets (Ishii and Boyer 2015, Lam et al. 2015, Maurer et al. 2003, Remy 2010).

Remy (2010) conducted an LCA that included an ozonation process for pharmaceutical

destruction in source separated urine, however they did not evaluate the specific toxicity

of pharmaceuticals in the model. Previous LCA studies have evaluated the

environmental impacts of pharmaceuticals in wastewater effluent. A study by Muñoz et

al. (2008) concluded that pharmaceuticals were a significant contributor to the toxicity of

the studied wastewater. Conversely, it was found that pharmaceuticals in decentralized

hospital wastewater exhibited negligible environmental impact compared with the

impacts generated by wastewater treatment (Igos et al. 2013, Igos et al. 2012).

Advanced treatment of decentralized hospital wastewater would not decrease

pharmaceutical toxicity in total wastewater effluent because the contribution of

89

pharmaceuticals from hospital wastewater was much smaller than other sources (e.g.,

pharmaceutical usage at homes, office buildings, etc.) (Igos et al. 2012). Ort et al.

(2010) estimated that hospital wastewater contributed 15% of pharmaceuticals to

centralized wastewater. This suggests that an alternative approach to treating municipal

wastewater at the community-level, such as urine source separation, could be more

effective at reducing pharmaceutical loading to the environment. To date, there is not a

published LCA study on urine source separation considering removal of

pharmaceuticals by sorption process and corresponding reduction in toxicity. The goal

of this research was to compare the overall environmental and economic impacts of

pharmaceutical removal from urine generated in a university community by centralized

wastewater treatment, advanced treatment of centralized wastewater, and centralized

and decentralized treatment of source separated urine. The pharmaceuticals

investigated in this study were from the NSAID pharmaceutical class and included

diclofenac (DCF), ibuprofen (IBP), ketoprofen (KTP), and naproxen (NPX). They were

selected because of high ecotoxicity potential, prevalence, and variable removal rates in

biological wastewater treatment (Hernando et al. 2006, Lienert et al. 2007b, Verlicchi et

al. 2012). For the reasons given above, NSAIDs have been the focus of ion-exchange

removal studies in urine. This study utilizes lab-scale experimental data to build a robust

framework and conduct a baseline assessment that may be augmented with new

pharmaceutical adsorption data as it becomes available.

Life Cycle Model

Scope of the Study

The functional unit for this study was the conveyance, storage, pharmaceutical

management (i.e., ion-exchange treatment), and nutrient management (i.e., struvite

90

precipitation) of 11,184 m3 of urine, which is equivalent to the estimated annual

production of urine at the University of Florida (UF) campus in Gainesville, Florida, USA.

This study builds upon the LCA model developed by Ishii and Boyer (2015) by

expanding the system boundary from residence halls to include the entire UF campus,

and pharmaceutical management. Detailed calculations used to determine the

functional unit are provided in Appendix C.

Figure 4-1 shows the wastewater management scenarios considered in this LCA.

Scenario AWWT serves as the baseline scenario and included combined collection of

urine, feces, and greywater, and biological treatment at the UF Water Reclamation

Facility. The upstream system boundary includes operational inputs for potable flush

water production at the nearby drinking water treatment facility. The construction and

decommission phase of the wastewater treatment plant was not included in this

assessment because they were assumed to be equal across all scenarios, thus

negating the contributions of these phases to the environmental assessment. Scenario

BWWT,O3 is a hypothetical scenario which included combined collection of urine, feces,

and greywater, and biological treatment at the UF Water Reclamation Facility upgraded

with an ozonation process for pharmaceutical destruction (Ternes et al. 2003). The

construction phase for the ozone contactor and operational phase of the ozone process

were included in this assessment. Decommission of the ozone system was not taken

into consideration. It was assumed that no nutrients were recovered for reuse as

fertilizer in AWWT and BWWT,O3. Land application of biosolids was excluded from the

system boundary because the local utility ceased land application and currently

disposes of biosolids in a landfill. Furthermore, the effect that urine source separation

91

has on the composition of biosolids at the centralized wastewater treatment plant is

unknown. Modeling the composition of biosolids at the wastewater treatment plant after

urine source separation was considered outside the scope of this model.

Scenarios C–H were the hypothetical urine source separation scenarios. The

system boundary includes the processes related to potable water production, urine

source separation, and treatment (i.e., storage disinfection, pharmaceutical removal by

ion-exchange, and struvite precipitation), centralized wastewater treatment, and

wastewater discharge to surface water and landscape irrigation. For scenarios

Ctruck,landfill and Dtruck,regen, urine was collected by a vacuum truck and transported to

a central location for processing. In scenarios Esewer,landfill and Fsewer,regen, urine was

conveyed by vacuum sewer to a central location for processing. In scenarios

Gdecen,landfill and Hdecen,regen, urine was collected and processed at the building level

for decentralized treatment. For scenarios Ctruck,landfill, Esewer,landfill, and Gdecen,landfill, it

was assumed spent anion exchange resin (AER) was transported and disposed of in a

landfill. For scenarios Dtruck,regen, Fsewer,regen, and Hdecen,regen, it was assumed spent

AER was regenerated using 5% NaCl, 50% methanol, and the brine was transported

and incinerated at a cement kiln plant for energy recovery. The system boundaries do

not include redistribution of struvite to agriculture. It was assumed that struvite fertilizer

would replace commercial fertilizers used in AWWT and BWWT,O3and that struvite

fertilizer granules were comparable to commercial fertilizers, allowing the use of

commercial fertilizer spreading equipment (Forrest et al. 2008). Furthermore, the

ammonia, nitrous oxide, and phosphate emissions for struvite and commercial fertilizer

92

(i.e., monoammonium phosphate (MAP)) were assumed to be equivalent due to

identical emission factors and nutrient content (Nemecek and Kägi 2007). However, the

cadmium content of struvite and commercial fertilizers were considered an emission to

land (i.e., 0.39 mg Cd/kg P2O5 in struvite and 97.5 mg Cd/kgP2O5 in MAP) (Lugon-

Moulin et al. 2006, Ronteltap et al. 2007). Infrastructure for the urine source separation

system (e.g., vacuum sewer, urine piping, and ion-exchange system) and operation

(e.g., road transport, energy and chemical requirements) were included within the

system boundary. Decommission of the urine diversion and treatment system was not

taken into consideration.

Life Cycle Inventory

The data sources and design parameters used to develop foreground processes

for each treatment scenario are provided in detail in Appendix C. The life cycle inventory

included potable flush water production, centralized wastewater treatment, ozonation of

wastewater, urine source separation infrastructure, urine collection by vacuum truck or

vacuum sewer, ion-exchange infrastructure and treatment, struvite precipitation for

nutrient recovery, and estimated pharmaceutical concentrations in urine. Background

inventory data for each scenario were designed using existing components in two

databases, the Ecoinvent unit processes (version 2.2) and the U.S. Life Cycle Inventory

Database (USLCI) (Ecoinvent Centre 2015, NREL 2012). Data from the Ecoinvent

database is based on either European, Swiss, or North American technologies

published between 2007–2009. Data from the USLCI database is based on North

American technologies or processes published between 2003–2008. European based

data was adopted without any modification for this study.

93

Life Cycle Costing

The economic analysis included infrastructure and operational costs for

pharmaceutical removal in urine by the alternative treatment scenarios. Net present

value (NPV) was estimated using an interest rate of 3% (National Center for

Environmental Economics 2010). The sources and assumptions underlying all cost

estimates are given in the relevant life cycle inventory sections in Appendix C and are

listed in Table C-9. All infrastructure costs were updated to 2016 based on inflation.

Labor costs were excluded in the cost analysis.

Life Cycle Impact Assessment

The LCA model for all scenarios was constructed using SimaPro 8.0.3.14

software (PRé Consultants 2014). The TRACI impact assessment method was used to

evaluate the contributions of processes, generated, and avoided impacts to ten midpoint

impact categories (U.S. EPA 2014). This method was selected because the study

pertains to wastewater treatment in a U.S. community and TRACI was developed by the

U.S. Environmental Protection Agency. The ten midpoint impact categories (e.g., ozone

depletion, global warming, smog, etc.) were evaluated with respect to a reference unit

(e.g., kg CFC-11 eq, kg CO2 eq, kg O3 eq, etc.) and normalized to obtain a single impact

score, expressed in Person-Equivalent (PE). Normalization is a conversion step that

compares the magnitude of impacts relative to a common reference. For TRACI 2.1,

results were normalized to the average annual impact of a U.S. citizen using 2008 as

the reference year (Ryberg et al. 2014).

The UNEP-SETAC toxicity model USEtox is the basis for the TRACI impact

categories for human health non-carcinogenic, and ecotoxicity and are expressed in

comparative toxic units (U.S. EPA 2014). Non-carcinogenic human toxicity (CTUh) is

94

characterized by the estimated morbidity increase in the total population per unit mass

of emitted chemical (disease cases/kg) and ecotoxicity (CTUe) is an estimate of the

potentially affected fraction of species over time and volume per unit mass of chemical

emitted (PAF·m3·day/kg). Characterization factors for DCF, IBP, and NPX were

obtained from literature (Alfonsín et al. 2014). The USEtox model was used to

determine an ecotoxicity characterization factor for KTP using the model’s substance

database and ecotoxicity data from literature (Andersson et al. 2007, Hauschild et al.

2015, Morais 2014). A characterization factor for human toxicity was not determined for

KTP due to a lack of data. The characterization factors are listed in Table C-10.

Sensitivity and Uncertainty Analysis

The uncertainty of input parameters on the impact assessment results for each

scenario was evaluated using the integrated Monte Carlo module in SimaPro. In each

Monte Carlo analysis, 3000 iterations were conducted. Table C-11 lists all of the input

parameters, range of variation, justification and assumed distribution considered in the

uncertainty analyses. Variability of unit costs were also included to evaluate the

uncertainty of assumed input operational costs (Table C-12). Cost variability of

magnesium oxide and liquid oxygen were excluded due to a lack of data. Infrastructure

costs were assumed to remain constant. Additional sensitivity analyses were conducted

to evaluate the effect of assumed model inputs (Table C-11) and unit costs (Table C-12)

on the environmental impacts for each scenario. The sensitivity analysis was conducted

by varying each parameter individually between the minimum and maximum values. A

parameter was considered sensitive if results varied from the baseline ±10%.

95

Results and discussion

Overall Comparison of Scenarios

Figure 4-2 provides a comparison of the environmental impacts of the treatment

scenarios, subdivided into the contributing processes (e.g., potable water, WWTP

electricity, urine transport, etc.), generated impacts (e.g., nutrient and pharmaceutical

emissions), and avoided impacts (e.g., fertilizer offsets). Alternatively, Figure C-5 shows

the same total impact score of each scenario subdivided into the contributing mid-point

impact categories. Together, these two figures provide a holistic view of the major

contributing processes and impact categories to the total environmental impact. Non-

normalized results for individual TRACI impact categories, excluding ecotoxicity and

non-carcinogenic impacts are provided in Figures B-6–B-13).

The order of decreasing total environmental impact was BWWT,O3 > AWWT >

Esewer,landfill > Fsewer,regen > Ctruck,landfill > Gdecen,landfill > Dtruck,regen > Hdecen,regen. The

trend suggests that centralized wastewater treatment had greater environmental

impacts than the source separation scenarios, primarily due to the potable water

requirements for flushing, electricity for wastewater treatment, and nutrient emissions.

Furthermore, struvite precipitation of source separated urine reduces nutrient

emissions, offsets commercial fertilizer production, and reduces cadmium emissions

due to commercial fertilizers. These results are similar to other LCA studies that found

that potable water savings, electricity savings, reduction in nutrient loading, and reduced

cadmium emissions from commercial fertilizers in the environment are major benefits

gained from urine source separation (Berndtsson 2006, Ishii and Boyer 2015, Lam et al.

2015, Lamichhane and Babcock 2012, Ronteltap et al. 2007). Results of the Monte

96

Carlo simulation indicate that uncertainty does not affect the overall trends for the total

environmental impact, where AWWT and BWWT,O3 have greater observed environmental

impact compared with the urine source separation and the vacuum sewer scenarios had

the largest impact among the source separation scenarios. However, there was

uncertainty between scenarios Ctruck,landfill, Dtruck,regen, Gdecen,landfill, and Hdecen,regen at

the 97.5% confidence interval suggesting that the difference in environmental impact of

urine collection by vacuum truck or decentralized treatment is not significant. However,

the total environmental impact for the resin disposal scenarios was less than the

scenarios where resin was disposed of in a landfill (e.g., Ctruck,landfill > Dtruck,regen). Within

impact categories, AWWT and BWWT,O3 maintained the greatest impact for all categories

except ozone depletion. However, some uncertainty was observed within impact

categories for the source separation scenarios, with the exception of the eutrophication

impact category.

Replacing conventional fixtures with urine diverting flush toilets and waterless

urinals would conserve 2.6×105 m3 of potable flush water and $231,000 annual potable

water savings. The implications of potable water savings would be of particular

importance in areas that face water scarcity and quality issues (Ishii and Boyer 2015).

Accounting for the reduction in potable flush water, influent flow at the wastewater

treatment plant would decrease by 17%. This reduction in influent flow could reduce the

electricity requirements for wastewater treatment. A limitation of this study is that

quantifying electricity use at the plant simplifies the impact urine source separation can

have on centralized wastewater treatment. Jimenez et al. (2015) found that urine source

separation can reduce influent N and P loading and potentially eliminate the need for

97

nitrification, reduce sludge age, and reduce chemical requirements for chemical P

removal.

Table 4-1 provides a summary of the estimated economic impacts associated

with infrastructure, energy, potable flush water, chemicals, and urine-based fertilizer

revenue for each treatment scenario. The NPV and EAC of scenarios Ctruck,landfill,

Dtruck,regen, Gdecen,landfill, and Hdecen,regen vary from AWWT by only ±2–13%. These

scenarios could be considered comparable to AWWT due to the preliminary nature and

uncertainty of the economic evaluation. However, the economic costs of BWWT,O3,

Esewer,landfill, and Fsewer,regen was 21–45% greater than AWWT. Due to the uncertainty of

the input parameters and unit costs, Monte Carlo analysis show that the cost of each

scenario increases and decreases at the 2.5 % confidence interval (CI) and 97.5% CI,

respectively. Compared across scenarios, cost savings for urine source separation may

be even greater (i.e., 48%–69% less than AWWT) at the 2.5% confidence interval (CI).

However, at the 97.5% CI, observed trends were the same as the baseline values. This

suggests that scenarios Ctruck,landfill, Dtruck,regen, Gdececn,landfill, and Hdecen,regen have lower

environmental impact and similar or lower economic cost compared to AWWT. This result

is similar to Ishii and Boyer (2015) which concluded that urine source separation and

struvite precipitation for maximum P recovery had significantly lower environmental

impact but negligible cost differences, compared with centralized wastewater treatment.

This suggests that the cost of additional pharmaceutical treatment of source separated

urine would not limit implementation.

98

Urine Source Separation

The urine source separation scenarios had a lower impact for all impact

categories compared with AWWT and BWWT,O3, with the exception of the ozone depletion

impact category (Figure 4-2). Anion exchange resin is the major contributing process to

the ozone depletion impact category because of the trichloromethane solvent used to

add quaternary amine functional groups to the polymer backbone for a type I AER

(Figure C-6) (Althaus et al. 2007). This result differs from Choe et al. (2013) who found

that ion-exchange resins dominate all impact categories except for ozone depletion,

however ion-exchange resin was modeled as a general polystyrene and did not include

the additional functionalization step in resin manufacturing. Eutrophication is the

greatest contributor to the total environmental impact in the source separation

scenarios, which is primarily due to the N remaining in urine after struvite precipitation

for P recovery (Figure C-5 and Figure C-10). However, Ishii and Boyer (2015) found that

struvite precipitation for maximum P and N recovery had a greater environmental impact

than struvite precipitation for maximum P recovery due to the chemical inputs. This

suggests that alternative N removal or recovery technologies should be explored.

Alternatively, direct application of stored liquid urine could serve as a complete nutrient

source (Kirchmann and Pettersson 1995), with the added benefit of reducing both N and

P loading at the wastewater treatment plant and subsequent receiving waters. Coupled

with the fact that MgO and struvite storage requirements had the second greatest

environmental impact in Ctruck,landfill, Dtruck,regen, Gdececn,landfill, and Hdecen,regen, application

of liquid urine may reduce the total environmental impact. However, the social

implications of applying liquid urine compared to a urine-derived solid fertilizer should be

99

considered. In general, user acceptance surveys found that >79% of respondents

approved of urine source separation technology in residence halls, public buildings, or

workplaces (Ishii and Boyer 2016, Lienert and Larsen 2010), and 85% approved of

urine-derived fertilizers (Lienert and Larsen 2010). However, only 50% of farmers

regarded urine fertilizer as a good idea primarily due to fear of liability (Lienert and

Larsen 2010). Farmers in Switzerland preferred a grainy and odorless ammonium

nitrate fertilizer, but they were willing to use odorous urine fertilizer in fields (Lienert et

al. 2003). This suggests that a mineral fertilizer (e.g., struvite) may be more appropriate

for application in urban areas compared with liquid urine, which could be applied in rural

and agricultural settings.

Scenarios Esewer,landfill and Fsewer,regen have the largest environmental impact

among the source separation scenarios due to the vacuum sewer infrastructure and

operation. This indicates that the method of urine collection and handling (e.g., vacuum

sewer vs. vacuum truck vs. decentralized treatment) is a critical consideration during the

design phase. The material and process inputs for the vacuum sewer system is largely

dependent on the geographical size of the collection area. Conversely, the process

inputs for collection by vacuum truck is dependent on both distance traveled (size of

collection area) and volume of urine produced. As shown in Figure C-15, the total

normalized impact of the vacuum sewer exceeds vacuum truck collection when plotted

as a function of total pipe length of the sewer system or distance traveled (km). This

suggests that a vacuum sewer system would have a greater environmental impact than

vacuum truck collection, regardless of the size of the collection area. Overall,

centralized urine treatment collected by vacuum truck and decentralized urine treatment

100

had the lowest environmental impact. However, the feasibility of implementing multiple

decentralized systems on a large scale must be considered. Facilities and maintenance

staff maintain all building services and operations (e.g., janitorial and maintenance) on

campus. It is expected that maintenance staff would maintain the ion-exchange system

and struvite precipitation operations. This would be a significant new task for

maintenance staff to undertake and would likely require hiring personnel to handle these

tasks or subcontracting to a private firm. The labor requirements and additional costs

were not included in this model but could be considered in future work (Ramos et al.

2014). With respect to other types of communities, decentralized treatment may be

more appropriate in rural areas. For example, Wood et al. (2015) found that urine

diversion coupled with conventional septic systems for greywater management

exhibited the lowest economic cost and highest cost effectiveness for N mitigation for

rural households.

The scenarios where spent resin was disposed of in a landfill had a slightly

greater environmental impact than the scenarios where resin was regenerated and the

waste regeneration solution (i.e., 5% NaCl, 50% methanol) was incinerated at a cement

kiln plant. Although the production of methanol, salt, and potable water used for

regeneration generates environmental impact, greater environmental offsets were

achieved from incinerating the methanol-containing brine for energy recovery as

opposed to fossil fuels used at the cement kiln plant (Figure C-14). A limitation of the

Ecosolvent model used to generate the life cycle inventory for brine incineration is that it

represents Swiss technology, however plants may vary according to the kiln and flue

gas treatment technology (Seyler et al. 2005). Furthermore, the fuel mix assumed in the

101

Ecosolvent model is different compared to the U.S. cement kiln industry (Hanle 2004).

Choe et al. (2013) and Maul et al. (2014) found that salt requirements for the ion-

exchange was the major contributor to the environmental impact of an ion-exchange

process. This indicates that improving the sustainability of ion-exchange regeneration

(e.g., brine incineration for energy recovery or brine recycling) can make regeneration

more favorable than disposing of resin in a landfill. The potential benefits from

incinerating the regeneration brine are two-fold: environmental offsets due to reduced

fossil fuel consumption and ultimate destruction of the pharmaceuticals to prevent

release to the environment. An additional driver for resin regeneration and brine

incineration, as opposed to landfill disposal, is the potential for pharmaceuticals to end

up in landfill leachate (Lu et al. in press). However, occurrence of pharmaceuticals in

landfill leachate was not included within the LCA framework. Alternatively, a semi-

closed loop system may be developed by destroying pharmaceuticals in the

regeneration brine by advanced oxidation processes to allow brine recycling (Zhang et

al. 2015).

Pharmaceutical Toxicity

Figure 4-3a shows the TRACI impact results for ecotoxicity (CTUe) subdivided

into the contributing processes, generated impacts, and avoided impact, and Figure 4-

3b shows the ecotoxicity impact only due to pharmaceutical emissions in wastewater

effluent discharged to surface water and reclaimed water. A similar figure for non-

carcinogenic human toxicity (CTUh) is given in Figure C-16. For brevity, this discussion

focuses on ecotoxicity because the same general trends were observed for human

toxicity. Overall, the order of decreasing total ecotoxicity was BWWT,O3 > AWWT >

102

Esewer,landfill > Fsewer,regen > Ctruck,landfill > Gdecen,landfill > Dtruck,regen > Hdecen,regen (Figure 4-

3a). Ecotoxicity due to pharmaceutical emissions followed a decreasing trend of AWWT >

Ctruck,landfill = Dtruck,regen = Esewer,landfill = Fsewer,regen = Gdecen,landfill = Hdecen,regen >

BWWT,O3 (Figure 4-3b).

As expected, AWWT had the greatest ecotoxicity due to pharmaceutical emissions

because biological treatment only achieves 28–87% pharmaceutical removal and can

vary for individual pharmaceuticals (Fernandez-Fontaina et al. 2012, Hollender et al.

2009, Joss et al. 2005, Lindqvist et al. 2005, Rivera-Utrilla et al. 2013, Rosal et al. 2010,

Salgado et al. 2012, Ternes 1998). The fate of pharmaceuticals in raw wastewater is

removal by adsorption to sludge and/or biotransformation, or discharged in the effluent

(Cook et al. 2012). Furthermore, biological wastewater treatment does not maintain

consistent treatment efficiencies as observed by negative mass balances in wastewater

effluent (Blair et al. 2015). Although biosolids disposal was not included within the

framework of the LCA, land application of biosolids is a common emission source for

pharmaceuticals in wastewater treatment. It was estimated that 210–250 tonnes/year of

72 pharmaceuticals and personal care products are land applied to U.S. soils from

biosolids recycling, nationwide (McClellan and Halden 2010). The pharmaceuticals

investigated in this study, on average, were estimated to be removed 5–7% by sludge

adsorption based on their sludge adsorption coefficient (Kd) (Alvarino et al. 2014, Blair

et al. 2015, Carballa et al. 2008, Jelic et al. 2011, Jones et al. 2002, Joss et al. 2005,

Radjenović et al. 2009, Sipma et al. 2010). High variability of overall removal suggests

that biological wastewater treatment cannot consistently achieve effective removal of

pharmaceuticals.

103

Scenario BWWT,O3 had the lowest pharmaceutical ecotoxicity due to high

pharmaceutical destruction (53–98%) (Hollender et al. 2009, Margot et al. 2013, Rosal

et al. 2010, Ternes et al. 2003), however it exhibits the greatest total ecotoxicity due to

the additional ozone process (e.g., ozone contactor infrastructure, electricity, liquid

oxygen, water, and transport requirements). One of the limitations of this model is that

only the ecotoxicity of the parent compound is considered and reduction in ecotoxicity is

directly related to its removal. However, ecotoxicity studies have shown that more toxic

byproducts may be formed after ozonated wastewater compared to the pre-ozonated

water and would require an extended contact time, or an additional treatment step such

as sand filtration or activated carbon, to remove the oxidation byproduct (Magdeburg et

al. 2012, Sánchez-Polo et al. 2008, Stalter et al. 2010).

For the urine source separation scenarios, the total ecotoxicity impact is 90%

less than AWWT and BWWT,O3, primarily due to the reduction in potable water use and

electricity at the wastewater treatment plant (Figure 4-3a). Indirect toxicity of producing

auxiliary materials and energy for potable water production and electricity use at the

wastewater treatment plant is originated by several substances emitted to water during

electricity production and lime sludge disposal (for potable water production only). As

shown in Figure 4-3b, the toxicity due to pharmaceutical emissions is the same for all

urine source separation scenarios because equivalent pharmaceutical removal was

achieved by ion-exchange. In general, pharmaceutical ecotoxicity followed a decreasing

order of IBP > DCF > KTP > NPX. The fact that DCF removal was highest in these

treatment scenarios (98% removal) but remains the second most toxic pharmaceutical

highlights the importance of evaluating the reduction in toxicity of each pharmaceutical

104

as opposed to the average mass removal. Considering the majority of pharmaceuticals

in centralized wastewater come from human urine, separation and treatment of this

unique waste stream may be the most effective pharmaceutical management strategy.

An expressed limitation of LCA regarding toxicity include not being able to accurately

evaluate mixture toxicity (Muñoz et al. 2009). Toxicity studies have observed

antagonistic or synergistic toxicity effects in pharmaceutical mixtures (Pomati et al.

2008). However, a recent study by Watanabe et al. (2016) found that concentration

addition and independent action are accurate at predicting chronic mixture toxicity of

pharmaceuticals at environmentally relevant concentrations. In addition, LCA does not

evaluate endocrine disruption potential due to limited information and lack of an

epidemiological framework (Finkbeiner et al. 2014). Another limitation of this model is

that the pharmaceuticals evaluated in this study (i.e., non-steroidal anti-inflammatory

drugs) are not comprehensive of all pharmaceutical compounds with respect to toxicity

and removal efficiency.

Model Sensitivity

Object 4-1 is an Excel spreadsheet that shows the results of the sensitivity

analysis as a percent change in each urine treatment scenario’s impact within an impact

category, total impact, and total cost, relative to the baseline assumption; impact

changes within varying ranges (i.e., 10–19%, 20–49%, and ≥50%) are highlighted in

color. Overall, the most sensitive assumptions to the model were pharmaceutical

concentrations in urine, TN and TP in urine, WWTP energy, storage time, and resin

capacity. Similar to Ishii and Boyer (2015), the various impact categories within the

treatment scenarios were sensitive to the assumed concentration of P in urine, the

assumption for electricity use at the wastewater treatment plant, and storage time. In

105

general, AWWT and BWWT,O3 maintained the greatest total impact throughout the entire

sensitivity analysis. Although the model was sensitive to the assumed WWTP energy

requirements, and TN and TP concentrations in urine, the observed trend for total

environmental impact for the scenarios remained the same.

The source separation scenarios were sensitive to the assumed pharmaceutical

concentrations in urine for three out of ten impact categories (i.e., ozone depletion,

acidification, and respiratory effects). A decrease or increase in impact was observed

when pharmaceutical concentrations were minimized or maximized, respectively. This

was attributed to the decrease or increase in resin and chemical requirements (i.e.,

NaCl and methanol for regeneration), and size of the ion-exchange vessel. An inverse

relationship (e.g., a decrease in impact when pharmaceutical concentrations were

maximized) was observed for Dtruck,regen, Fsewer,regen, and Hdecen,regen. This was due to

the greater methanol requirements for regeneration and subsequent fossil fuel offsets

due to brine incineration at the cement kiln plant. Uncertainty regarding the estimated

pharmaceutical concentrations in urine may be improved with increased sampling

campaigns, improved understanding of pharmaceutical consumption, and modeling

procedures. A model developed by Winker et al. (2008b) to predict pharmaceutical

concentrations in urine had a strong correlation but only accounted for prescribed

pharmaceuticals, however a significant amount of pharmaceuticals may be purchased

over-the-counter. Furthermore, there is a lack of data regarding the amount of over-the-

counter pharmaceuticals actually consumed. A similar trend was observed when the

column was sized to achieve maximum IBP removal compared to the baseline, which

was sized to achieve maximum DCF removal. Due to the low capacity of the AER for

106

IBP compared to DCF resin, chemical requirements and corresponding fossil fuel

offsets from brine incineration increased, resulting in a decrease in impact for the

regeneration scenarios, increase in impact for the landfill scenarios, and an increase in

total cost.

The results of the economic sensitivity analysis are shown in the second tab of

Object 4-1. The second table in Object 4-1 shows the percent change from the baseline

NPV values for each scenario, respectively. The third table in Object 4-1 shows the

percent difference in NPV compared with Scenario A. The cost of urine diverting flush

toilets and waterless urinals was the most sensitive cost for the urine source separation

scenario and also the largest single economic cost. Urine-diverting flush toilets are not

widely used compared with conventional toilets. Considering the material inputs of these

fixtures do not differ from conventional toilets, it is reasonable to expect that increasing

demand would decrease market price. If market value of these fixtures cost the same as

conventional fixtures, urine source separation would cost 18–54% less than AWWT. Total

cost was also sensitive to utility rates for potable water and electricity, particularly for

AWWT and BWWT,O3. For example, when the cost of potable water was minimized, the

cost of urine source separation was 24–75% greater than AWWT. However, when the

cost was maximized, urine source separation cost 63–74% less than AWWT. This

suggests that potable water savings may be a driver for or against implementation of

urine source separation, depending on the community. The cost of electricity also varied

between communities. In communities with a high cost of electricity, implementation of

urine source separation may result in appreciable cost savings compared to AWWT.

Finally, the total cost was sensitive to the assumed interest rate. In general, as interest

107

rate increased, the cost of BWWT,O3 increased compared to AWWT, and the cost of urine

source separation decreased compared to AWWT.

Object 4-1. Environmental impact and economic costing sensitivity analysis results (.xlsx file 43.7 KB)

Concluding Remarks

There are numerous environmental benefits associated with urine source

separation (e.g., potable flush water savings, electricity savings at the wastewater

treatment plant, and nutrient offsets), Compared with centralized wastewater treatment,

ozonation of wastewater had a higher environmental impact and economic cost, urine

collected by vacuum sewer had lower environmental impact and higher economic cost,

and urine collected by vacuum truck collection or treated at decentralized locations had

lower environmental impact and similar economic cost. Urine source separation can

achieve a high reduction of pharmaceutical toxicity and comparatively low total toxicity

from the treatment process compared with BWWT,O3. Additional sorption studies are

needed to evaluate the removal of various pharmaceutical compounds from diverse

therapeutic classes and chemical structure, the results of which could be incorporated

into a future version of this LCA framework. The benefit of this LCA framework is that

the environmental impact and economic cost of alternative sorbents can easily be

evaluated. Although the AER used for this study may not be the most appropriate to

remove all pharmaceuticals, multiple sorbents may be utilized which have a higher

selectivity and capacity for the pharmaceuticals of concern. Using a more selective and

higher capacity resin would decrease the resin requirements and subsequent costs. In

conclusion, the framework created and tested herein estimates the environmental and

108

economic impacts of alternative treatment technologies that remove pharmaceuticals

and recover nutrients in source separated urine in a community setting.

109

Table 4-1. Capital and operation and management (O&M) costs, net present value (NPV) for each urine treatment scenario; positive values indicate cost, negative values indicate revenue.

Capital Costs AWWT BWWT,O3 Ctruck,landfill Dtruck,regen Esewer,landfill Fsewer,regen Gdecen,landfill Hdecen,regen

Fixturesa $1,590,000 $1,590,000 $4,880,000 $4,880,000 $4,880,000 $4,880,000 $4,880,000 $4,880,000

Vacuum sewer $0 $0 $0 $0 $651,500 $651,500 $0 $0

Urine piping $0 $0 $171,000 $171,000 $171,000 $171,000 $171,000 $171,000

Urine storageb $0 $0 $1,300,000 $1,300,000 $1,300,000 $1,300,100 $69,000 $69,000

Ozone system $0 $2,810,000 $0 $0 $0 $0 $0 $0

Fiberglass IX vessel $0 $0 $8,600 $8,600 $8,600 $8,600 $8,600 $8,600

Struvite storage $0 $0 $400 $400 $400 $400 $400 $400

O&M Costs

Diesel fuelc $0 $160 $2,900 $3,000 $4 $100 $12 $100

Potable flush water $234,000 $234,000 $1,300 $1,300 $1,300 $1,300 $1,300 $1,300

Electricity at WWTPd $36,300 $36,300 $1,700 $1,700 $1,700 $1,700 $1,700 $1,700

Ozone operatione $0 $19,900 $0 $0 $0 $0 $0 $0

Vacuum sewer $0 $0 $0 $0 $36,900 $36,900 $0 $0

IX resin $0 $0 $52,800 $52,800 $52,800 $52,800 $52,800 $52,800

IX operationf $0 $0 $1,500 $9,000 $1,500 $9,000 $1,500 $9,000

Struvite revenueg $0 $0 –$12,100 –$12,100 –$12,100 –$12,100 –$12,100 –$12,100

NPV ($M)h $10.1 $14.6 $10.5 $10.7 $12.2 $12.4 $8.83 $9.05

2.5% CI $36.2 $41.5 $16.0 $16.7 $18.3 $18.9 $11.2 $11.8

97.5% CI $7.47 $11.9 $7.60 $7.68 $9.28 $9.40 $7.04 $7.12 a Cost of conventional toilets and urinals (Scenarios A and B) or urine diverting flush toilets and waterless urinals (Scenarios C–F) b Includes centralized (Scenarios C and D) and decentralized (Scenarios C–F) urine storage c Cost of diesel for all unit processes (e.g., ozonation process, urine, resin disposal to landfill or brine disposal to cement kiln plant, and/or struvite collection) d Only pertains to electricity use based on influent flow at wastewater treatment e Includes liquid oxygen, potable water, diesel, and energy requirements

f Includes potable water, chemical (e.g., NaCl, methanol), and energy requirements g Net balance of MgO costs for struvite precipitation and value of struvite h 60 year planning horizon and 3% interest rate

110

Figure 4-1. Treatment schematic for scenarios A–H (light gray horizontal arrows) and contributing processes. Single black lines represent urine flow, grey lines represent ion-exchange resin flow, and double black lines represent struvite flow. The solid lines indicate transport through pipes and dashed lines represent road transport by truck.

111

Figure 4-2. Normalized TRACI impact score for all scenarios (a) centralized wastewater

treatment scenarios (AWWT, BWWT,O3), and (b) urine source separation

scenarios (C–H). Each colored bar represents input processes (e.g., potable water, electricity use at the WWTP, urine source separation (USS) infrastructure), avoided impacts (e.g., fertilizer offsets, brine incineration), and generated impacts (e.g., nutrient discharge, pharmaceutical discharge). The brackets around each error bar represent the 95% confidence interval resulting from Ecoinvent database distributions from the Monte Carlo uncertainty analysis.

112

Figure 4-3. Comparison of ecotoxicity impact (CTUe = PAF·m3·day) due to (a) contributing processes (e.g., flush water, urine transport) and generated emissions (e.g., nutrient discharge, pharmaceutical discharge) and avoided impacts (e.g., fertilizer offsets, brine incineration) in each scenario and (b) pharmaceutical emissions only. The brackets around each error bar represent the 95% confidence interval resulting from Ecoinvent database distributions from the Monte Carlo uncertainty analysis.

113

CHAPTER 5 CONCLUSIONS

Urine source separation is a process that can help advance two paradigm shifts

for sustainable water and nutrient management: resource recovery of valuable nutrients,

and holistic management of contaminants of emerging concern. The significant

contribution of urine to NSAID loading in the environment makes urine source

separation an attractive process to address pharmaceutical pollution. Furthermore, the

beneficial reuse of nutrients in urine makes it necessary to employ a treatment process

that selectively removes NSAIDs without co-removal of nutrients. The work presented

here provides insights into the effectiveness and efficiency of using an AER to

selectively remove NSAIDs from urine. Although this work focused on the removal of

NSAIDs by one AER, this framework may be utilized to evaluate various sorbents and

pharmaceuticals and hormones. Figure 5-1 depicts the systematic approach to evaluate

sorption processes to remove pharmaceuticals in urine, and how future work may be

incorporated into the framework. Generated experimental isotherm and kinetic data for

diverse sorbents and pharmaceuticals may be used to predict fixed-bed breakthrough

performance and compared with bench scale column test. The breakthrough curve may

be compared to in vitro bioassay dose-response curves, and potential in vivo effects to

evaluate the reduction in ecotoxicity potential and to establish a treatment objective

(e.g., breakthrough is when effluent reaches EC10). The treatment objective determines

the reactor size, resin requirements, and regeneration schedule which may be

incorporated into the LCA framework to evaluate the overall environmental impacts and

economic costs of the treatment process. The framework may be expanded to include

alternative sorbents or nutrient recovery technologies such as biochar and ammonia

114

stripping, respectively. The results of the LCA may be used to determine areas of

improvement and the process repeated to optimize the system.

Results from the batch equilibrium tests of NSAIDs at realistic concentrations in

urine highlight the primary interactions that dictate ion-exchange selectivity, and the

impact of urine composition on removal. Understanding the mechanisms of interaction

is important for selecting the appropriate material for successful removal. The removal

of acidic organic compounds (i.e., NSAIDs) was governed by both the electrostatic

interaction between the (i) carboxylic acid functional group of the pharmaceutical and

quaternary ammonium functional group of the resin, and (ii) the non-electrostatic

interactions between the aromatic ring structure of the pharmaceutical and aromatic ring

structure of the AER. Alternatively, carbamazepine, which is a neutral pharmaceutical

containing three aromatic ring structures, may be selectively removed by a polymeric

adsorbent. Furthermore, the hydrophobicity and charge of a pharmaceutical may vary

under fresh (pH 6) and ureolyzed urine (pH 9) conditions, and impact removal efficacy.

For example, the hydrophobicity of the NSAIDs studied decreased with increasing pH

suggesting that sorption was less selective under ureolyzed urine conditions compared

with fresh urine. Although the NSAIDs studied were negatively charged over the entire

pH range, the charge of other pharmaceuticals of interest may vary with pH. Depending

on the point of implementation of a sorption column, whether at the toilet or a central

location, the hydrophobicity and charge of the pharmaceutical compound should be

considered when selecting an appropriate sorbent material.

Predicting fixed-bed breakthrough performance using kinetic and equilibrium

batch data is a rapid way to evaluate pharmaceutical removal performance.

115

Furthermore, this method allows the user to optimize the treatment process by

evaluating treatment performance for varying conditions prior to bench-scale or pilot

column testing, for example, whether treatment would vary significantly under fresh or

ureolyzed urine conditions. Furthermore, the presence of endogenous metabolites in

urine competed for ion-exchange sites on the resin, reducing the capacity of the resin

for pharmaceuticals. Evaluation of more selective sorbents, such as molecularly

imprinted polymers, or sorbents with higher capacity, such as activated carbon, may

improve pharmaceutical removal in the presence of organic endogenous metabolites.

Evaluation of fixed-bed performance as a function of toxicity rather than mass removal

provides a better understanding of treatment efficacy. Evaluating pharmaceutical activity

using high-throughput in vitro bioassays provides a rapid assessment of the potential

effect of pharmaceuticals. The development of high-throughput assay toxicity databases

provides the opportunity to evaluate treatment efficacy with respect to various cellular

response pathways.

The results of the LCA demonstrate that urine source separation has significant

benefits with respect to water conservation, energy savings, and reduced nutrient

loading compared to conventional wastewater treatment. The economic benefits

associated with the water and energy savings gained from implementing urine source

separation is dependent on utility costs for potable water and electricity. However, if the

cost of urine-source separating fixtures (e.g., urine diverting flush toilets and waterless

urinals) was equivalent to conventional fixtures, urine source separation may become

more economically feasible. Furthermore, the scale at which urine is treated (e.g.,

building level or collected by vacuum truck for centralized treatment) have similar

116

environmental impacts and economic costs which provides flexibility for communities’

decision making when considering implementation. The LCA model serves as a

framework to evaluate the environmental impact and economic costs to remove a

variety of pharmaceuticals using alternative sorbents. For example, pharmaceutical

removal by biochar may be evaluated and the framework expanded to include biochar

manufacturing.

Urine source separation is one solution that may help to address pharmaceutical

loading in the environment. However, urine source separation is limited to addressing

pharmaceuticals primarily excreted in urine. Therefore, it may be one of multiple

solutions to address the growing issue of pharmaceuticals, personal care products, and

endocrine disrupting compounds in the environment. The additional benefit of urine

source separation is recovery of valuable nutrients for beneficial reuse, thus requiring a

treatment process, such as ion-exchange, to effectively separate pharmaceuticals from

nutrients. However, early evaluation of a sorption technology or process requires a

multi-faceted, systems-level evaluation to ensure treatment efficacy with respect to both

mass removal and toxicity reduction while minimizing the upstream environmental

impact and economic costs associated with treatment.

117

Figure 5-1. Visual representation of the systematic approach for evaluating sorption materials to remove pharmaceuticals

in source separated urine. Dashed boxes indicate opportunities for future work. Generation of experimental isotherm and kinetic data for various sorbent materials may be utilized to model column breakthrough curves. Breakthrough curves may be compared to dose-response toxicity curves to establish a treatment objective (e.g., IC10). The treatment objective determines the capital (e.g., reactor size) and operation requirements (e.g., resin volume, regeneration) which may be included within the LCA framework to evaluate the overall environmental and economic costs. The resulting LCA may be used to identify areas of improvement to further optimize the sorption process.

118

APPENDIX A SUPPLEMENTARY INFORMATION FOR CHAPTER 2

Estimation of Realistic Pharmaceutical Concentrations in Urine

Data from previous publications was reviewed to estimate realistic

pharmaceutical concentrations in urine as shown in Table A-1. Studies conducted by

Ternes (1998) and Salgado et al. (2012) measured pharmaceutical loading (g/d) in raw

wastewater. For simplicity, it was assumed that all pharmaceuticals originated from

excretion in human urine therefore concentrations in urine were estimated based on

total treatment volume in population equivalents (p.e.) and a daily urine void volume of

1.6 L/person (FitzGerald et al. 2002, Latini et al. 2004). A study by Joss et al. (2005)

measured pharmaceutical concentrations (ng/L) in raw wastewater; an average

volumetric flow rate of the wastewater treatment plant was obtained from the 2005

Annual Report of the wastewater treatment plant studied (Kloten/Opfikon 2005). Lastly,

a study by Winker et al. (2008b) directly measured pharmaceutical concentrations in

human urine as well as theoretically calculated concentrations.

Isotherm Models

Freundlich Isotherm

The Freundlich isotherm is an empirical model that does not imply maximum

adsorption capacity of the sorbent. Adsorption is non-ideal, reversible, and is not

restricted to monolayer adsorption. The amount adsorbed is the summation of

adsorption on all sites, with stronger binding sites occupied first, until adsorption energy

is exponentially decreased. It is based on the following equation,

𝑞𝑒 = 𝐾𝐹𝐶𝑒1/𝑛𝐹 (A-1)

119

where KF (mmol1-1/nF L1/nF/g) and nF (dimensionless) are the Freundlich isotherm

constants determined from nonlinear regression. KF is an approximate indicator of the

adsorption capacity and the nF parameter represents the adsorption intensity and

surface heterogeneity, where 0 < 1/nF < 1 indicates favorable adsorption and 1/nF = 1

indicates linear adsorption (Delle Site 2001).

Langmuir Isotherm

The Langmuir isotherm assumes monolayer adsorption on a homogenous

surface (Foo and Hameed 2010). Graphically, this is characterized by a plateau where

the saturation point is reached and no additional adsorption can take place and is based

on Eq. A-2,

𝑞𝑒 =𝐾𝐿𝑞0𝐶𝑒

1+𝐾𝐿𝐶𝑒 (A-2)

where KL (L/mol) and q0 (mmol/g) are determined from nonlinear regression. The

adsorption energy of a solute on a sorbent can be determined from the Langmuir

isotherm parameter, KL, as a change in Gibbs free energy as shown in Eq. A-3

(Ghodbane and Hamdaoui 2008),

Δ𝐺° = −𝑅𝑇ln (𝐾𝐿) (A-3)

where R is the ideal gas constant, 8.314 J/mol∙K, and T is the temperature (K).

Values of ΔG° < 0 suggest favorable and spontaneous sorption of the solutes

(Ghodbane and Hamdaoui 2008). A separation factor, RL, may also be used to describe

the sorption behavior and can be calculated following Eq. A-4,

𝑅𝐿 =1

1+𝐾𝐿𝐶0 (A-4)

120

where KL is the Langmuir constant, and C0 is the initial concentration of the

solute (mmol/L). Values of 0 < RL < 1 indicate favorable sorption and RL > 1 indicates

unfavorable sorption (Foo and Hameed 2010).

D–A and D–R Isotherm

The Dubinin–Astakhov (D–A) and Dubinin–Radushkevich (D–R) are based on

the Polyani adsorption potential theory where the adsorption potential ɛ (A-5) is related

to the average free energy change of a substance from the liquid to the resin phase

(Foo and Hameed 2010). These models imply a micropore volume filling adsorption

process, which is in contrast to the layer-by-layer and monolayer adsorption modeled by

the Freundlich and Langmuir isotherms (Foo and Hameed 2010), and are temperature

dependent,

𝜀 = 𝑅𝑇 ln (1 +1

𝐶𝑒) (A-5)

The amount adsorbed is quantified by a function of the adsorption potential (Eq.

A-6),

𝑞𝑒 = 𝑞0 exp (− (𝜀

√2𝐸)

𝑛𝐷

) (A-6)

where E is the adsorption energy (J/mol), nD is the heterogeneity factor

(dimensionless), and q0 (mmol/g) is the maximum adsorption capacity of the sorbent. In

the case of the D-A model, q0 is assumed to be limited by a maximum and matches the

manufacturer’s capacity of the AER, while E and nD were determined from nonlinear

regression. For the D-R model nD = 2, and similar to the Freundlich model q0 is not

limited by a maximum adsorption capacity. Along with E, q0 was determined by

nonlinear regression (Eq. A-7),

𝑞𝑒 = 𝑞0 exp (− (𝜀

√2𝐸)

2

) (A-7)

121

Table A-1. Properties of pharmaceuticals used in ion-exchange experiments.

Pharmaceutical, CAS number, molecular weight

Structure pKa log Kow log D, pH 6

log D, pH 9

m/za

Diclofenac sodium 15307-79-6 318.1

4.24d 4.51f 2.49c 0.77c 296

Ibuprofen sodium 31121-93-4 228.26

4.38d 3.97f 2.09c –0.30c 207

Ketoprofen 22071-15-4 254.28

4.07d 3.12e 1.12c 0.67c 255

Naproxen sodium 26159-34-2 252.24

4.15f 3.18f 1.12c 0.20c 231

Paracetamol 103-90-2 151.16

9.38f 0.46f 0.53b 0.41b 151

a Observed under method’s conditions

b Estimated using the PALLAS PrologD prediction program (CompuDrug 2006) c Estimated using Eq. 2-2 d Meloun et al. (2007) e Sangster (2014) f Hazardous Substances Data Bank (TOXNET 2012)

122

Table A-2. Estimated and measured pharmaceutical concentrations in urine based on previous literature.

Reference Matrix Treatment volume Compound Concentration in wastewater, µg/L (µmol/L)

Load in wastewater, g/d (mmol/d)

Concentration in urine, µg/L (µmol/L)

Ternes (1998) Raw wastewater

312,000 p.e.

Diclofenac – 100 (310) 210 (0.67)

Ibuprofen – 250 (1100) 530 (2.3)

Naproxen – 80 (320) 170 (0.68)

Salgado et al. (2012) Raw wastewater

32,700 p.e. Diclofenac – 35 (110) 710 (2.9) Ibuprofen – 46 (200) 940 (4.1) Ketoprofen – 83 (330) 1,700 (6.6)

Joss et al. (2005) Raw wastewater

55,000 p.e., 1.5 × 106 L/d c

Diclofenac 1.1 (3.5) – 210 (0.67)

Ibuprofen 2.0 (8.8) – 390 (1.7)

Naproxen 1.1 (4.4) – 210 (0.84)

Winker et al. (2008b) Urine –

Diclofenac – – 21a/12b

(6.7×10–2/3.7×10–2)

Ibuprofen – – 496a/678b (2.2/3.0)

Ketoprofen – – 2b (7.1×10–3) a Average concentration of two sampling locations b Theoretical concentration c Average volumetric flow rate of sampling site based on wastewater treatment plant annual report (Kloten/Opfikon 2005)

123

Table A-3. Properties of strong-base, anion exchange polymer resins. Resin Pore structure Polymer Functional group Capacityb

(meq/mL) Density (g/mL)

Dowex 22a macroporous styrene R–N+(CH3)2(CH2OH), dimethylethanolamine

1.2 0.317c

Purolite A520E macroporous styrene R–N+(C2H5)3, triethylamine 0.9 0.323d

Dowex Marathon 11

gel styrene R–N+(CH3)3, trimethylamine 1.3 0.322d

a Primary resin investigated b Manufacturer data c Determined experimentally (this study) d Determined experimentally (Landry and Boyer 2013)

124

Table A-4. Linear form of isotherm models and plots to determine estimated initial values for non-linear isotherm modeling parameters.

Isotherm Linear form Plot

Freundlich log(𝑞𝑒) = log(𝐾𝐹) +1

𝑛log(𝐶𝑒) log(𝑞𝑒) 𝑣𝑠 log(𝐶𝑒)

Langmuir 1

𝑞𝑒=

1

𝑞0+

1

𝑘𝐿𝑞0𝐶𝑒

1

𝑞𝑒 𝑣𝑠

1

𝐶𝑒

Dubinin-Astakhov ln(𝑞0) − ln(𝑞𝑒) = (𝜖

√2𝐸)

𝑛𝐷 ln(𝑞0) − ln(𝑞𝑒) 𝑣𝑠 𝜖

Dubinin-Radushkevich ln(𝑞𝑒) = ln(𝑞0) − (𝜖

√2𝐸)

2

ln(𝑞𝑒) 𝑣𝑠 𝜖2

125

Table A-5. Individual equilibrium experiment isotherm parameters for Dowex 22 AER sorption of diclofenac (C0 = 3.0 µmol/L), ibuprofen (C0 = 3.6 µmol/L), ketoprofen (C0 = 7.8 µmol/L), naproxen (C0 = 7.5 µmol/L), and paracetamol (C0 = 14 µmol/L) in ureolyzed urine. Isotherm parameters and goodness of fit statistics (sum of squares errors (SSE), correlation coefficients (R2), and average relative errors (ARE)) for the Langmuir, Freundlich, Dubinin-Astakhov, and Dubinin-Radushkevich models were determined by nonlinear regression.

Langmuir

Pharmaceutical KL (L/mmol) q0 (mmol/g) RL ΔG° (J/mol) SSE (mmol2/g2) R2 ARE (%)

Diclofenac 989 1.41×10–2 0.254 –15,600 6.90×10–6 0.960 32 Ibuprofen –125 –1.16×10–2 1.84 –b 7.53×10–6 0.919 34 Ketoprofen 46.5 4.85×10–2 0.734 –8,710 6.33×10–5 0.751 11 *Naproxenc 610 9.35×10–3 0.179 –14,600 8.62×10–7 0.961 7 Naproxen 1230 5.61×10–3 9.78×10–2 –16,100 3.35×10–5 0.061 70 Paracetamol –77.6 –5.31×10–5 –14.6 –b 9.14×10–4 –0.297a 100

Freundlich

Pharmaceutical KF (mmol1–1/nF L1/nF/g) 1/nF SSE (mmol2/g2) R2 ARE (%)

Diclofenac 0.481 0.639 7.62×10–6 0.956 32 Ibuprofen 9.79 1.25 9.01×10–6 0.903 37 Ketoprofen 0.874 0.864 6.37×10–5 0.750 10 *Naproxenc 0.118 0.510 2.63×10–7 0.988 4 Naproxen 1.15×10–2 0.183 3.41×10–5 0.044 67 Paracetamol 1.00×1089 48.3 7.98×10–4 –0.132a 82

Dubinin-Astakhov

Pharmaceutical E (kJ/mol) nD SSE (mmol2/g2) R2 ARE (%)

Diclofenac 0.810 0.692 7.87×10–6 0.955 32 Ibuprofen 1.97 1.120 9.18×10–6 0.901 37 Ketoprofen 0.903 0.772 6.38×10–5 0.749 10 *Naproxenc 0.228 0.490 2.39×10–7 0.989 20 Naproxen 0.009 0.274 4.47×10–5 –0.253a 76 Paracetamol 6.28 9.20 6.05×10–4 0.142 277

Dubinin-Radushkevich

Pharmaceutical E (kJ/mol) q0 (mmol/g) SSE (mmol2/g2) R2 ARE (%)

Diclofenac 7.94 6.12×10–2 8.05×10–6 0.954 25 Ibuprofen 5.53 0.209 1.02×10–5 0.890 40 Ketoprofen 6.31 8.05×10–2 6.35×10–5 0.751 12 *Naproxenc 9.12 2.08×10–2 3.94×10–7 0.982 4 Naproxen 9.16 1.37×10–2 5.49×10–5 –0.538a 91 Paracetamol 0.862 1.00×1032 7.31×10–4 –0.037a 113 a R2 is a proportion of variance explained by the fit, it the fit is worse than fitting a horizontal line, then R2 is negative and cannot be interpreted as the square of a correlation (MathWorks 2013) b Could not be determined due to negative KL value c * Denotes isotherm modeling of naproxen experimental data excluding data for lowest resin dose

126

Table A-6. Isotherm parameters, sum of squares errors (SSE), correlation coefficients (R2), and average relative errors (ARE) of the Freundlich, Langmuir, Dubinin-Astakhov, and Dubinin-Radushkevich models determined by nonlinear regression for the different ion-exchange resins used to remove diclofenac (Co = 0.2 mmol/L) in ureolyzed urine.

Freundlich

Resin KF (mmol1–1/nF L1/nF/ g) 1/nF SSE (mmol2/g2) R2 ARE (%)

A520E 0.993 0.646 1.96×10–3 0.978 10 Dowex22 1.29 0.551 3.51×10–3 0.982 15 Dowex Marathon 11 5.72 0.871 8.20×10–3 0.976 25 Langmuir

Resin KL (L/mmol) q0 (mmol/g) RL ΔG° (J/mol) SSE (mmol2/g2) R2 ARE (%)

A520E 10.8 0.436 0.316 –5400 1.30×10–3 0.986 6 Dowex22 24.9 0.484 0.163 –7290 2.21×10–3 0.988 8 Dowex Marathon 11 6.30 1.76 0.444 –4280 6.44×10–3 0.981 23 Dubinin-Astakhov

Resin E (kJ/mol) nD SSE (mmol2/g2) R2 ARE (%)

A520E 1.13 0.703 2.19×10–3 0.976 11 Dowex22 1.18 0.670 4.42×10–3 0.977 16 Dowex Marathon 11 2.68 1.18 7.12×10–3 0.979 23 Dubinin-Radushkevich

Resin E (kJ/mol) q0 (mmol/g) SSE (mmol2/g2) R2 ARE (%)

A520E 5.29 0.424 1.30×10–3 0.986 7 Dowex22 6.18 0.554 1.74×10–3 0.991 9 Dowex Marathon 11 5.04 1.30 4.77×10–3 0.986 20

127

Table A-7. Estimated physicochemical parameters of the four major diclofenac metabolites, 3’-hydroxydiclofenac, 4’-hydroxydiclofenac, 5’-hydroxydiclofenac, and 3’hydroxy-4’-methoxydiclofenac.

Metabolite pKaa LogKowb LogD, pH 6c LogD, pH 9c

3’-hydroxydiclofenac 4.50, 8.05 4.00 2.48 –1.00 4’-hydroxydiclofenac 4.50, 8.82 4.86 2.34 –0.55 5’-hydroxydiclofenac 4.40, 10.43 3.88 2.26 –0.18 3’hydroxy-4’-methoxydiclofenac

4.52, 8.05 3.83 2.34 –1.16

a Acid dissociation constant calculated using the PALLAS pKalc prediction program, v.3.8.1.2 (CompuDrug 2006) b Octanol–water partitioning coefficient calculated using the PALLAS PrologP prediction program, v.3.8.1.2 (CompuDrug 2006) c pH dependent distribution coefficient calculated using the PALLAS PrologD prediction program, v.3.8.1.2 (CompuDrug 2006)

128

Table A-8. Equilibrium experiment isotherm parameters for Dowex 22 AER sorption of ibuprofen (C0 = 0.2 mmol/L) present in fresh urine. Isotherm parameters and goodness of fit statistics (sum of squares errors (SSE), correlation coefficients (R2), and average relative errors (ARE)) for the Langmuir, Freundlich, Dubinin-Astakhov, and Dubinin-Radushkevich models were determined by nonlinear regression.

Langmuir KL (L/mmol) q0 (mmol/g) RL ΔG° (J/mol) SSE (mmol2/g2) R2 ARE (%)

3.45 0.236 0.562 -2810 2.80E-04 0.966 3

Freundlich KF (mmol1–1/nF L1/nF/ g) 1/nF SSE (mmol2/g2) R2 ARE (%)

0.317 0.727 3.28E-04 0.960 4

Dubinin-Astakhov E (kJ/mol) nD SSE (mmol2/g2) R2 ARE (%)

0.187 0.459 3.59E-04 0.956 5

Dubinin-Radushkevich E (kJ/mol) q0 (mmol/g) SSE (mmol2/g2) R2 ARE (%)

4.22 0.166 2.73E-04 0.967 4

129

Table A-9. Combined equilibrium experiment isotherm parameters for Dowex 22 AER sorption of diclofenac (C0 = 3.5 µmol/L), ibuprofen (C0 = 4.7 µmol/L), ketoprofen (C0 = 7.3 µmol/L), and naproxen (C0 = 7.4 µmol/L) all present in ureolyzed urine. Isotherm parameters and goodness of fit statistics (sum of squares errors (SSE), correlation coefficients (R2), and average relative errors (ARE)) for the Langmuir, Freundlich, Dubinin-Astakhov, and Dubinin-Radushkevich models were determined by nonlinear regression.

Langmuir

Pharmaceutical KL (L/mmol) q0 (mmol/g) RL ΔG° (J/mol) SSE (mmol2/g2) R2 ARE (%)

Diclofenac 354 3.78×10–2 0.445 –13300 6.48×10–6 0.989 32 Ibuprofen –173 –4.65×10–3 5.32 –b 2.54×10–5 0.885 31 *Ketoprofenc 182 1.24×10–2 0.429 –11800 1.58×10–7 0.988 3 Ketoprofen 595 5.12×10–3 0.187 –14500 2.89×10–5 –0.046a 65535 Naproxen 123 2.99×10–2 0.521 –10900 7.70×10–5 0.763 20

Freundlich Pharmaceutical KF (mmol1–1/nF L1/nF/ g) 1/nF SSE (mmol2/g2) R2 ARE (%)

Diclofenac 1.87 0.777 6.99×10–6 0.988 34 Ibuprofen 1130 2.10 3.02×10–5 0.863 44 *Ketoprofenc 0.306 0.733 2.39×10–7 0.981 3 Ketoprofen 0.176 0.302 3.02×10–5 –0.091a 65535 Naproxen 0.467 0.707 7.61×10–5 0.765 16

Dubinin-Astakhov Pharmaceutical E (kJ/mol) nD SSE (mmol2/g2) R2 ARE (%)

Diclofenac 1.59 0.889 7.04×10–6 0.988 34 Ibuprofen 3.83 1.91 3.16×10–5 0.857 47 *Ketoprofenc 0.472 0.623 2.49×10–7 0.981 4 Ketoprofen 0.054 0.375 3.79×10–5 –0.371a 65535 Naproxen 0.65 0.664 7.60×10–5 0.766 15

Dubinin-Radushkevich Pharmaceutical E (kJ/mol) q0 (mmol/g) SSE (mmol2/g2) R2 ARE (%)

Diclofenac 7.84 0.106 7.86×10–6 0.987 38 Ibuprofen 4.00 3.65 3.17×10–5 0.857 48 *Ketoprofenc 7.25 3.12×10–2 1.82×10–7 0.986 37 Ketoprofen 10.0 8.86×10–3 3.09×10–5 -0.119a 65535 Naproxen 7.03 6.45×10–2 7.74×10–5 0.761 20 a R2 is a proportion of variance explained by the fit, it the fit is worse than fitting a horizontal line, then R2 is negative and cannot be interpreted as the square of a correlation (MathWorks 2013) b Could not be determined due to negative KL value c * Denotes isotherm modeling of ketoprofen experimental data excluding data for lowest resin dose

130

Table A-10. Analysis of covariance (ANOCOVA) test results to determine whether there was a significant difference at the 95% confidence interval (α = 0.05) between DCF, IBP, KTP, and NPX ion-exchange when present individually or combined in synthetic ureolyzed urine. The null hypothesis states that there was not a significant difference between slopes (p > 0.05) and the alternative hypothesis states that there was a significant difference between the slopes (p < 0.05).

Pharmaceutical F Statistic Probability > F

Diclofenac 0.602 0.446 Ibuprofen 1.08 0.309 Ketoprofen 11.8 0.002 Ketoprofena 0.602 0.446 Naproxen 9.04 0.006 Naproxena 0.163 0.690 a Excluding lowest resin dose (0.16 mL/L)

131

Figure A-1. Individual experimental data and sorption isotherms determined by nonlinear regression of paracetamol (PCM) (C0 = 14 µmol/L) using Dowex 22 anion exchange resin.

0

0.005

0.01

0.015

0.02

0.025

0 0.005 0.01 0.015

qe

, m

mo

l/g

Ce, mmol/L

Paracetamol

PCMLangmuirFreundlichD-AD-R

132

Figure A-2. Experimental data and isotherm models for naproxen and ketoprofen

when (a) naproxen (NPX) present individually in ureolyzed urine, and (b) ketoprofen (KTP) present as a mixture in ureolyzed urine. Both figures illustrate the plotted experimental isotherms including the lowest resin dose of 0.16 mL/L (i.e. including the data point with the highest Ce) and corresponding nonlinear isotherm models (Freundlich, Langmuir, Dubinin-Astakhov (D-A), and Dubinin-Radushkevich (D-R)).

0

0.005

0.01

0.015

0.02

0 0.002 0.004 0.006 0.008

qe

, m

mo

l/g

Ce, mmol/L

(a) Naproxen

NPXLangmuirFreundlichD-AD-R

0

0.005

0.01

0.015

0.02

0 0.002 0.004 0.006 0.008

qe

, m

mo

l/g

Ce, mmol/L

(b) KetoprofenKTPLangmuirFreundlichD-AD-R

133

Figure A-3. Experimental data and ion-exchange isotherms of diclofenac removal by various resins (a) A520E, (b) Dowex 22, and (c) Dowex Marathon 11 AER; C0 = 0.2 mmol/L.

0

0.1

0.2

0.3

0.4

0.5

0 0.02 0.04 0.06 0.08 0.1 0.12 0.14

qe

, m

mo

l/g

Ce, mmol/L

(a) A520E

A520EFreundlichLangmuirD-AD-R

0

0.1

0.2

0.3

0.4

0.5

0 0.02 0.04 0.06 0.08 0.1 0.12 0.14

qe

, m

mo

l/g

Ce, mmol/L

(b) Dowex 22

Dowex 22FreundlichLangmuirD-AD-R

0

0.1

0.2

0.3

0.4

0.5

0 0.02 0.04 0.06 0.08 0.1 0.12 0.14

qe

, m

mo

l/g

Ce, mmol/L

(c) Dowex Marathon 11

Dowex 11FreundlichLangmuirD-AD-R

134

Figure A-4. Mole fraction distribution of the neutral and ionized species present in the

octanol and water phase for (a) diclofenac (DCF), (b) ibuprofen (IBP), (c) ketoprofen (KTP), and (d) naproxen (NPX). Dashed red lines indicate the mole fraction of the ionized species present in the octanol phase at pH 6 (i.e., fresh urine) and the solid red lines indicate the mole fraction of the ionized species present in the octanol phase at pH 9 (i.e., ureolyzed urine).

0

0.2

0.4

0.6

0.8

1

1 2 3 4 5 6 7 8 9 10 11 12 13 14

Mole

Fra

ction

pH

(a)

[DCF]o

[DCF–]o

[DCF]w

[DCF–]w

0

0.2

0.4

0.6

0.8

1

1 2 3 4 5 6 7 8 9 10 11 12 13 14

Mole

Fra

ction

pH

(b)

[IBP]o[IBP–]o[IBP]w[IBP–]w

0

0.2

0.4

0.6

0.8

1

1 2 3 4 5 6 7 8 9 10 11 12 13 14

Mole

Fra

ction

pH

(c)

[KTP]o[KTP–]o[KTP]w[KTP–]w

0

0.2

0.4

0.6

0.8

1

1 2 3 4 5 6 7 8 9 10 11 12 13 14

Mole

Fra

ction

pH

(d) [NPX]o[NPX–]o[NPX]w[NPX–]w

135

Figure A-5. Combined pharmaceutical experimental data and sorption isotherms

determined by nonlinear regression of (a) diclofenac (DCF) (C0=3.5 µmol/L), (b) ibuprofen (IBP) (C0=4.7 µmol/L), (c) ketoprofen (KTP) (C0=7.3 µmol/L), and (d) naproxen (NPX) (C0=7.4 µmol/L) using Dowex 22 anion exchange resin. Figure (c) *Ketoprofen illustrates the plotted experimental isotherms excluding the lowest resin dose of 0.16 mL/L (i.e. excluding the data point with the highest Ce and corresponding nonlinear isotherm models (Freundlich, Langmuir, Dubinin-Astakhov (D-A), and Dubinin-Radushkevich (D-R)).

0

0.005

0.01

0.015

0.02

0 0.002 0.004 0.006 0.008

qe

, m

mo

l/g

Ce, mmol/L

(a) Diclofenac

DCFLangmuirFreundlichD-AD-R

0

0.005

0.01

0.015

0.02

0 0.002 0.004 0.006 0.008

qe

, m

mo

l/g

Ce, mmol/L

(b) Ibuprofen

IBPLangmuirFreundlichD-AD-R

0

0.005

0.01

0.015

0.02

0 0.002 0.004 0.006 0.008

qe

, m

mo

l/g

Ce, mmol/L

(c) *KetoprofenKTPLangmuirFreundlichD-AD-R

0

0.005

0.01

0.015

0.02

0 0.002 0.004 0.006 0.008

qe

, m

mo

l/g

Ce, mmol/L

(d) Naproxen

NPXLangmuirFreundlichD-AD-R

136

C0, Cycle 1 Ce, Cycle 1 C0, Cycle 2 Ce, Cycle 2

C0, Cycle 3 Ce, Cycle 3

Figure A-6. Sorption by Dowex 22 anion exchange resin over three treatment cycles using fresh resin (Cycle 1) and regenerated resin (Cycles 2 and 3) in a continuous-flow mini-column. Influent (solid shapes) and effluent (open shapes) concentrations (mmol/L) of (a) diclofenac (DCF), (b) ibuprofen (IBP), (c) ketoprofen (KTP), and (d) naproxen (NPX) following sorption by Dowex 22 anion exchange resin over three treatment cycles.

0

2

4

6

8

0 5000 10000 15000

Ce,

mm

ol/L

Bed Volume

(a) Diclofenac

0

2

4

6

8

0 5000 10000 15000

Ce,

mm

ol/L

Bed Volume

(b) Ibuprofen

0

2

4

6

8

0 5000 10000 15000

Ce,

mm

ol/L

Bed Volume

(c) Ketoprofen

0

2

4

6

8

0 5000 10000 15000

Ce,

mm

ol/L

Bed Volume

(d) Naproxen

137

Figure A-7. Simultaneous column regeneration curves of (a) diclofenac, (b) ibuprofen,

(c) ketoprofen, and (d) naproxen using a 5% NaCl, equal-volume water–methanol solution. Regeneration was performed after the column was saturated with each pharmaceutical (i.e. influent = effluent) and reused for a total of three cycles.

0.0

1.0

2.0

3.0

4.0

0 4 8 12 16 20 24

Ce

, m

mo

l/L

Bed Volume

(a) DiclofenacCycle 1

Cycle 2

Cycle 3

0.00

0.25

0.50

0.75

0 4 8 12 16 20 24

Ce

, m

mo

l/L

Bed Volume

(b) IbuprofenCycle 1

Cycle 2

Cycle 3

0.0

0.5

1.0

1.5

0 4 8 12 16 20 24

Ce

, m

mo

l/L

Bed Volume

(c) Ketoprofen Cycle 1

Cycle 2

Cycle 3

0.0

0.5

1.0

1.5

2.0

0 4 8 12 16 20 24

Ce,

mm

ol/L

Bed Volume

(d) Naproxen Cycle 1

Cycle 2

Cycle 3

138

Column Mass Balance

The following equations summarize the calculations used for the pharmaceutical

mass balance. The mass of pharmaceuticals removed from urine (Mrem) and desorbed

(Mdesorb) from the AER column over three treatment and regeneration cycles was

performed by trapezoidal numerical integration of the column saturation and

regeneration curves using MATLAB (8.2.0.701 R2013b) software. For cycle 1, the mass

of pharmaceutical removed from urine (Mrem) was equivalent to the mass of

pharmaceutical sorbed onto the fresh resin (Msorb). After regeneration, the amount

remaining on the resin (Mfoul) was determined by subtracting Msorb from Mdesorb. For

cycles 2 and 3, the Msorb was the summation of the Mrem during each cycle and Mfoul

from the previous cycle.

C0 = influent concentration in urine (µmol/L)

Ce = effluent concentration in urine (µmol/L)

Cr = effluent concentration in regeneration solution (µmol/L)

Mrem = mass removed from urine (µmol)

Msorb = mass sorbed onto resin (µmol)

Mdesorb = mass desorbed from resin (µmol)

Mfoul = mass remaining on resin (µmol)

Vu = volume of urine treated (L)

Vr = volume of regeneration solution (L)

139

Cycle 1:

𝑀𝑟𝑒𝑚1= 𝑀𝑠𝑜𝑟𝑏1

= ∫ (𝐶0 − 𝐶𝑒)𝑑𝑉𝑉𝑢

𝑉0 (A-8)

𝑀𝑑𝑒𝑠𝑜𝑟𝑏1= ∫ 𝐶𝑟𝑑𝑉

𝑉𝑟

𝑉0 (A-9)

𝑀𝑓𝑜𝑢𝑙1= 𝑀𝑠𝑜𝑟𝑏1

− 𝑀𝑑𝑒𝑠𝑜𝑟𝑏1 (A-10)

Cycle 2:

𝑀𝑟𝑒𝑚2= ∫ (𝐶0 − 𝐶𝑒)𝑑𝑉

𝑉𝑢

𝑉0 (A-11)

𝑀𝑠𝑜𝑟𝑏2= 𝑀𝑟𝑒𝑚2

+ 𝑀𝑓𝑜𝑢𝑙1 (A-12)

𝑀𝑑𝑒𝑠𝑜𝑟𝑏2= ∫ 𝐶𝑟𝑑𝑉

𝑉𝑟

𝑉0 (A-13)

𝑀𝑓𝑜𝑢𝑙2= 𝑀𝑠𝑜𝑟𝑏2

− 𝑀𝑑𝑒𝑠𝑜𝑟𝑏2 (A-14)

Cycle 3

𝑀𝑟𝑒𝑚3= ∫ (𝐶0 − 𝐶𝑒)𝑑𝑉

𝑉𝑢

𝑉0 (A-15)

𝑀𝑠𝑜𝑟𝑏3= 𝑀𝑟𝑒𝑚3

+ 𝑀𝑓𝑜𝑢𝑙2 (A-16)

𝑀𝑑𝑒𝑠𝑜𝑟𝑏3= ∫ 𝐶𝑟𝑑𝑉

𝑉𝑟

𝑉0 (A-17)

𝑀𝑓𝑜𝑢𝑙3= 𝑀𝑠𝑜𝑟𝑏3

− 𝑀𝑑𝑒𝑠𝑜𝑟𝑏3 (A-18)

140

APPENDIX B SUPPLEMENTARY INFORMATION FOR CHAPTER 3

Estimation of Pharmaceutical Concentrations in Urine

A literature review was conducted to estimate pharmaceutical concentrations in

urine based on total treatment volume in population equivalents (p.e.) or flow rate (L/d),

and pharmaceutical influent loading. Several papers directly measured the mass

loading of pharmaceuticals (g/d) (Guerra et al. 2014, Salgado et al. 2012, Ternes 1998).

Other studies measured pharmaceutical concentrations in the influent wastewater

(Chen et al. 2015, Clara et al. 2005, Ferrando-Climent et al. 2012, Joss et al. 2005,

Lindqvist et al. 2005, Margot et al. 2013, Reungoat et al. 2010, Rosal et al. 2010,

Santos et al. 2007, Zorita et al. 2009). One paper directly measured pharmaceuticals in

urine (Winker et al. 2008b). It was assumed that all pharmaceuticals originated from

excretion in human urine, therefore the theoretical undiluted concentration in urine was

based on a daily urine void volume of 1.6 L/person (FitzGerald et al. 2002, Latini et al.

2004). Excretion rates of the parent compounds and metabolites were obtained from

literature to estimate the concentration of metabolites in urine.

Experimental Methods for Batch Tests

Batch kinetic and equilibrium test results were based on a measured wet weight

density of 0.8183 g/mL. Kinetic and equilibrium tests were only performed for DCF, IBP,

KTP, NPX, and Odm-NPX. A kinetic and equilibrium test was not conducted for KTP-

gluc due to its high cost and expected instability of the metabolite. For the kinetic tests,

flasks were placed on a shaker table and mixed at 350 rpm and removed at pre-

determined contact times (5 min, 30 min, 1 h, 2 h, 6 h, and 24 h). For the equilibrium

tests, samples were dosed with 0.25, 1, 2, 4, and 8 mL/L AER and mixed on a shaker

141

table at 350 rpm for 24 h. For one equilibrium experiment, real human urine was

collected and stored until hydrolysis (pH 9) and spiked with 1,000 µg/L of DCF. An

aliquot of treated urine was taken prior to sample preparation and analysis.

Homogenous Surface Diffusion Model

Briefly, the HSDM describes the mass transport through the filter bed (Eq. B-1)

and intraparticle mass transport through the adsorbent grains described by Fick’s

second law (Eq. B-2).

𝜖𝐵𝜕𝐶

𝜕𝑡+

𝑣𝐹𝜕𝐶

𝜕𝑧+ 3(1 − 𝜖𝐵) (

𝑘𝐿

𝑟𝑝) (𝑐 − 𝑐∗) = 0 (B-1)

𝜕𝑞

𝜕𝑡= 𝐷𝑆 (

𝜕2𝑞

𝜕𝑞2 +2𝜕𝑞

𝑟𝜕𝑟) (B-2)

Where εB is the bed porosity; q is the solid phase concentration (M/M); t is the

time coordinate (t); vF is the superficial velocity (L/t); kL is the liquid phase mass transfer

coefficient (L/t); rp is the AER particle radius (L); c and c* (M/L3) are the liquid phase

concentration in the bulk solution and exterior adsorbent surface, respectively; Ds is the

surface diffusion coefficient (L2/t); and r is the radial coordinate (L). The HSDM assumes

plug flow conditions in the fixed bed, liquid-phase and solid phase mass transfer,

constant flow, and constant diffusion coefficients, adsorbent grains are assumed to be

spherical and the adsorption equilibrium can be described by the Freundlich isotherm

(Eq. B-3).

𝑞𝑒 = 𝐾𝐹𝐶𝑒1/𝑛

(B-3)

A more detailed explanation of the model used in Fast 2.1beta may be found in

Sperlich et al. (2008). The input parameters for Fast 2.1beta included liquid mass

transfer and surface diffusion coefficients (Table B-8 and Table B-9), column operation

parameters (Table B-10) and AER properties (Table B-11). To validate the models,

142

experimental data from previously published research was fit to the HSDM (Landry and

Boyer 2016). An additional simplified fixed-bed column experiment was conducted by

dosing 0.2 mmol/L of DCF in synthetic ureolyzed urine without endogenous metabolites

following the same conditions previously described in Landry and Boyer (2016), using

Dowex Marathon 11 AER.

The Gnielinski correlation for packed beds was used to estimate the liquid-phase

mass transfer coefficient (Eq. B-4 ) (Crittenden et al. 2012). Liquid-phase mass transfer

is a function of the liquid-phase diffusion coefficient (Dl), Reynolds number (Re) and

Schmidt number (Sc), which may be calculated using Eq. B-5 through B-7, respectively.

The liquid diffusivity was calculated using the Hayduk–Laudie correlation (Crittenden et

al. 2012). The molar volume (Vb) of the solutes was calculated using the group

contribution method (Fedors 1974). A table of nomenclature used in Eq. C4–S7 is listed

in Table B-1. The fluid density and viscosity of urine were assumed to be the same as

water at 25°C (Table B-2). The values for Vb, Dl, and kl used in column modeling are

listed in Table B-3.

𝑘𝐿 =[1+1.5(1−𝜖𝐵)]𝐷𝐿

𝑑𝑝(2 + 0.644𝑅𝑒

1

2𝑆𝑐1

3) (B-4)

𝐷𝐿 =13.26×10−9

𝜇𝑙1.14𝑉𝑏

0.589 (B-5)

𝑅𝑒 =(𝜌𝑙)(𝜙)(𝑑𝑃)(𝑣𝑙)

(𝜖𝐵)(𝜇𝑙) (B-6)

𝑆𝑐 =𝜇𝑙

(𝜌𝑙)(𝐷𝐿) (B-7)

Batch kinetic data was used to estimate the NPX, and Odm-NPX surface

diffusion coefficients (Ds) (Table B-4) using the method developed by (Zhang et al.

143

2009). Kinetic data was used to calculate the dimensionless concentration 𝐶�̅�𝑎𝑡𝑎 at

time, t using Eq. B-8,

𝐶�̅�𝑎𝑡𝑎 =𝐶𝑡−𝐶𝑒

𝐶0−𝐶𝑒 (B-8)

Where C is the liquid-phase concentration at time t, C0 is the initial liquid-phase

concentration, and Ce is the equilibrium concentration determined by the Freundlich

isotherm. The values for 𝐶�̅�𝑜𝑑𝑒𝑙 were calculated using Eq. B-9,

𝐶�̅�𝑜𝑑𝑒𝑙 = 𝐴0 + 𝐴1(ln 𝑡̅) + 𝐴2(ln 𝑡̅)2 + 𝐴3(ln 𝑡̅)3 (B-9)

where the coefficients A1, A2, and A3 are parameters specific to the Freundlich

parameter 1/n, and Ce/C0 and may be found in (Zhang et al. 2009); and 𝑡̅ is the

dimensionless time (Dst/rp2) where rp is the particle radius. The experimental data was fit

to the model by optimizing Ds/rp2 by minimizing the objective function (Eq. B-10) using

the Solver function in Excel.

𝑂𝐹 = √∑ (𝐶̅𝑑𝑎𝑡𝑎,𝑖−�̅�𝑚𝑜𝑑𝑒𝑙,𝑖)2𝑛

𝑖=1

𝑛−1 (B-10)

This method for determining Ds is limited to data with Freundlich parameters 0.1

< 1/n < 0.9. For this reason, Ds for DCF, KTP and DCF in real urine was estimated

using the correlation developed by Crittenden et al. (1987), which relates the surface

diffusion flux to the pore diffusion flux (Eq. B-11 and Eq. B-12).

𝐷𝑠 = (𝑆𝑃𝐷𝐹𝑅)(𝑃𝐷𝐹𝐶) (B-11)

𝑃𝐷𝐹𝐶 =(𝜖𝑃)(𝐶0)(𝐷𝐿)

(𝜌𝑃)(𝐾𝐹𝐶0

1𝑛)(𝜏𝑃)

(B-12)

Where SPDFR is the surface-to-pore diffusion flux ratio, assumed to be 0.4 due

to the presence of DOC; PDFC is the pore diffusion flux; εp is the particle porosity; C0 is

the initial phase of the solute in the liquid phase; ρp is the particle density; KF is the

144

Freundlich isotherm parameter; and τp is the particle tortuosity, assumed to be 1

because its effects are accounted for in the SPDFR.

Fast2.1beta software predicts fixed-bed column breakthrough using the

homogenous surface diffusion model (HSDM) but is limited for case where the

Freundlich parameter 0.01 < 1/n < 1.05. For irreversible isotherms (e.g., 1/n = 0), Wicke

(1939) provided a solution to the HSDM in Eq. B-13

𝐶̅(𝑧̅ = 1, 𝑇) = 1 −6

𝜋2∑

1

𝑘2 exp (−𝑘2 {𝜋2𝐸𝑑 [𝑇(𝐷𝑔+1)−1

𝐷𝑔− 1] + 0.64}) …∞

𝑘=1 (B-13)

Where 𝐶̅ is the is the reduced fluid-phase concentration as a function of mass

throughput (T) calculated in Eq. B-14 and reduced axial position (𝑧̅), Dg is the solute

distribution parameter, calculated by Eq. B-15, Ed is the diffusivity modular, calculated

by Eq. B-16, and k is an integer constant.

𝑇 =𝑡

𝐸𝐵𝐶𝑇∗𝜖𝐵(𝐷𝑔+1) (B-14)

𝐷𝑔 =𝜌𝐵𝑞𝑒(1−𝜖𝐵)

𝜖𝐵𝐶0 (B-15)

𝐸𝑑 =𝐷𝑠𝐷𝑔𝐸𝐵𝐶𝑇(𝜖𝐵)

(𝑑𝑝

2)

2 (B-16)

145

Table B-1. Active ingredient and metabolite structure and chemical properties. Compound, CAS, molecular weight Structure pKa logKow Diclofenac sodium 15307-79-6 318.1 ≥98% purity

4.15a 4.51a

4’-Hydroxy diclofenac 64118-84-9 312.2 98% purity

3.76b 3.96b

Ibuprofen sodium 31121-93-4 228.26 ≥98% purity

4.9a 3.97a

Hydroxy ibuprofen 53949-53-4 222.3 98% purity

4.55b 2.69b

Ketoprofen 22071-15-4 254.28 ≥98% purity

3.88b 3.61b

Ketoprofen acyl glucuronide 76690-94-3 430.4 98% purity

3.24b 1.67 b

Naproxen 26159-34-2 252.24 98%–102% purity

4.15a 3.18a

O-Desmethylnaproxen 52079-10-4 216.23 98% purity

4.31 b 2.84 b

aTOXNET (2016)

bChemAxon (2016)

146

Table B-2. Synthetic ureolyzed urine composition adapted from Landry et al. (2015). Chemical Synthetic urine Real urine

NaCl, mmol/L 60 n.a.b

Na2SO4, mmol/L 15 n.a. KCl, mmol/L 40 n.a. NH4OH, mmol/L 250 n.a. NaH2PO4, mmol/L 14 n.a. NH4HCO3, mmol/L 250 n.a. Citrate, mmol/L 2.49 n.a. Creatinine, mmol/L 0.56 n.a. Glycine, mmol/L 1.24 n.a. Hippurate, mmol/L 2.80 n.a. L-Cysteine, mmol/L 0.81 n.a. Taurine, mmol/L 0.99 n.a. TOC, mg C/La 1,280 3,220 Conductivity, mS/cma 39.7 20.8 a Measured b Not analyzed

147

Table B-3. Estimated and measured pharmaceutical concentrations in urine from literature.

Reference Matrix Treatment volume

Compound Concentration in wastewater, μg/L

Load in wastewater, g/d

Concentration in urine, μg/L

Ternes (1998) Raw wastewater

312,000 p.e.

Diclofenac – 100 200

Ibuprofen – 250 501

Naproxen – 80 160

Salgado et al. (2012)

Raw wastewater

32,700 p.e.

Diclofenac – 35 670

Ibuprofen – 46 879

Ketoprofen – 83 1,586

Joss et al. (2005)

Raw wastewater

55,000 p.e. 15,992 m3/d

Diclofenac 1.10 – 200

Ibuprofen 2.00 – 363

Naproxen 1.10 – 200

Winker et al. (2008b)

Urine – Diclofenac – – 21b

– Ibuprofen – – 496b

Chen et al. (2015)

Raw wastewater

388,333 p.e. 224,333 m3/d

Diclofenac 1.42 – 513

Ibuprofen 14.20 – 5,127

Naproxen 8.44 – 3,047

Ferrando-Climent et al. (2012)

Raw wastewater

175,000 p.e. 35,000 m3/d

Ibuprofen 10.73 – 1,341

Guerra et al. (2014)

Raw wastewater

– Ibuprofen 8.60 2,300 1,075

Naproxen 6.28 1,600 785

Lindqvist et al. (2005)

Raw wastewater

167714 p.e.a 50476 m3/da

Ibuprofen 13.10 – 2,108b

Naproxen 4.99 – 842b

Ketoprofen 2.21 – 364b

Diclofenac 0.50 – 87b

Margot et al. (2013)

Raw wastewater

220,000 p.e. 95,000 m3/d

Diclofenac 0.48 – 323

Ibuprofen 1.20 – 1,107

Ketoprofen 4.10 – 302

Naproxen 1.12 – 188

Reungoat et al. (2010)

Raw wastewater

40,000 p.e. 10,000 m3/d

Diclofenac 0.20 – 31

Ibuprofen 0.09 – 14

Naproxen 0.29 – 45

Rosal et al. (2010)

Raw wastewater

72,000 m3/d

Diclofenac 0.23 – 29

Ibuprofen 2.69 – 336

Ketoprofen 0.44 – 55

Naproxen 2.36 – 295

Santos et al. (2007)

Raw wastewater

425,000 p.e.a 72,638 m3/da

Ibuprofen 99.53 – 9,917b

Ketoprofen 0.67 – 62b

Naproxen 6.36 – 670b

Clara et al. (2005)

Raw wastewater

461,610 p.e.a Ibuprofen 2.31 – 289b

Diclofenac 2.12 – 265b

Zorita et al. (2009)

Raw wastewater

55,000 p.e. 20,000 m3/d

Ibuprofen 6.90 – 1,568

Naproxen 4.90 – 1,114

Diclofenac 0.23 – 52 a Average treatment volume of multiple facilities b Average concentration of multiple sampling locations

148

Table B-4. Pharmaceutical dose-response concentrations used to evaluate COX-1 inhibition of single compounds.

Compound 100× (µmol/L) 10× (µmol/L) 1× (µmol/L) 0.1× (µmol/L) 0.01× (µmol/L)

Diclofenac 5.53×101 5.53 5.53×10–1 5.53×10–2 5.53×10–3

4-OH Diclofenac 1.47×102 1.47×101 1.47 1.47×10–1 1.47×10–2

Ibuprofen 1.05×103 1.05×102 1.05×101 1.05 1.05×10–1

OH-Ibuprofen 2.56×103 2.56×102 2.56×101 2.56 2.56×10–1

Ketoprofen 1.34×102 1.34×101 1.34 1.34×10–1 1.34×10–2

Ketoprofen glucuronide 1.12×103 1.12×102 1.12×101 1.12 1.12×10–1

Naproxen 3.01×102 3.01×101 3.01 3.01×10–1 3.01×10–2

O-Desmethylnaproxen 1.39×102 1.39×101 1.39 1.39×10–1 1.39×10–2

149

Table B-5. Pharmaceutical dose-response concentrations used to evaluate COX-1

inhibition of the pharmaceutical mixture. Compound 100× (µmol/L) 10× (µmol/L) 1× (µmol/L) 0.1× (µmol/L) 0.01× (µmol/L)

Diclofenac 5.06×101 5.06 5.06×10–1 5.06×10–2 5.06×10–3

Ketoprofen 9.78×101 9.78 9.78×10–1 9.78×10–2 9.78×10–3

Ketoprofen glucuronide 1.24×103 1.24×102 1.24×101 1.24 1.24×10–1

Naproxen 2.19×102 2.19×101 2.19 2.19×10–1 2.19×10–2

O-Desmethylnaproxen 8.63×101 8.63 8.63×10–1 8.63×10–2 8.63×10–3

Total 1.69×103 1.69×102 1.69×101 1.69 1.69×10–1

150

Table B-6. Nomenclature used to calculate liquid-phase mass transfer coefficient, liquid-phase diffusion coefficient, and surface diffusion coefficient.

Nomenclature Definition Unit

kL Liquid-phase mass transfer coefficient m/s DL Liquid-phase diffusion coefficient m2/s εB Bed porosity Dimensionless dp Particle diameter m Re Reynolds number Dimensionless Sc Schmidt number Dimensionless µl Fluid-phase viscositya kg/m-s, cP Vb Molar volume of solute cm3/mol ρl Fluid-phase density kg/m3

Φ Sphericity Dimensionless υl Superficial liquid velocity m/s Ds Surface diffusion coefficient m2/s εp Particle porosity Dimensionless C0 Initial solute concentration mg/L KF Freundlich isotherm parameter mg/g(L/mg)1/n

ρB Bed density g/L τP Resin tortuosity dimensionless a Fluid-phase viscosity is in units of centipoise (cP) and in units of kg/m-s when calculating the liquid-phase diffusion coefficient and liquid phase mass transfer coefficient, respectively

151

Table B-7. Urine properties assumed to be equivalent to water at 25°C.

Viscosity (µl), cP 0.89 Density (ρl), kg/m3 997

152

Table B-8. Molar volume (Vb), liquid diffusivity (DL), and liquid-phase mass transfer coefficients (kL).

Pharmaceutical Vb, cm3/mol DL, m2/s kL, m/s

Diclofenac 186.9 6.95×10–10 1.84×10–5 Ketoprofen 195.6 6.77×10–10 1.81×10–5 Naproxen 165.1 7.48×10–10 1.95×10–5 O-Desmethylnaproxen 137.8 8.32×10–10 2.12×10–5

153

Table B-9. Surface diffusion coefficient (Ds). Pharmaceutical Ds, m2/s

Diclofenac 6.97×10–14 Ketoprofen 4.43×10–13 Naproxen 8.78×10–14 O-Desmethylnaproxen 1.76×10–13

Diclofenac, real urine 7.18×10–13

154

Table B-10. Column operational parameters. Empty bed contact time, mina 8.3 Flow rate, mL/mina 0.72 Cross sectional area, cm2 0.7854 Superficial velocity, m/s 1.53×10–4

a Assumed b Determined experimentally c Manufacturer data

155

Table B-11. Resin properties. Wet resin bed density (ρB), g/mLb 0.8183 Particle density (ρp), g/mLc 1.1 Particle diameter (dp), µmc 450 Bed porosity (εB)

0.27 Particle porosity (εP) d 0.59 Particle sphericity (Φ) a 0.9 a Assumed b Determined experimentally c Manufacturer data d Particle porosity is calculated by the equation 1–ρP/ ρs, where ρs is the density of graphite (2.2 g/mL) e Assumed from literature

156

Table B-12. Freundlich isotherm parameters. Pharmaceutical KF, (µmol/g)(L/g)1/nF 1/nF SSE R2

Diclofenac 1.871 1.050 1.476 0.90 Ibuprofen 0.310 0.313 0.424 0.01 Ketoprofen 0.374 1.72×10–8 0.013 0.00 Naproxen 0.436 0.738 0.288 0.63 O-Desmethylnaproxen 1.250 0.865 1.343 0.88 Diclofenac, real urine 0.134 0.061 0.009 0.11

157

Table B-13. Hill model parameters from the COX-1 inhibition bioassays, including 95% confidence intervals, and goodness of fit measures for Diclofenac (DCF), Ketoprofen (KTP), Naproxen (NPX), Ketoprofen glucuronide (KTP gluc), O-desmethylnaproxen (Odm-NPX), and a pharmaceutical mixture (Mix).

DCF KTP NPX KTP gluc Odm-NPX Mix

I0 1% 0% 20% 0% 13% 0%

Imax 105% 100% 58% 100% 44% 100%

H 1.12 0.69 0.99 2.00 1.03 1.00

IC50, µmol/L 0.24 (0.161, 0.341)

1.30 (0.487, 2.12)

16.8 (–6.71, 40.2)

73.1 (52, 119)

4.13 (–0.486, 2.55)

9.73 (2.90, 16.8)

SSE 0.021 0.163 0.054 0.079 0.034 0.094

R2 0.987 0.926 0.851 0.967 0.827 0.955

dfe 9 13 11 13 8 10

Adjusted R2 0.984 0.921 0.810 0.964 0.762 0.951

RMSE 0.049 0.112 0.070 0.078 0.065 0.097

158

Table B-14. Alternative Hill model parameters from the COX-1 inhibition bioassays, including 95% confidence intervals, and goodness of fit measures for Naproxen (NPX), and O-desmethylnaproxen (Odm-NPX) where I0 and Imax

bounds extrapolated to 0% and 100%, respectively. NPX Odm-NPX

I0 0% 0%

Imax 100% 100%

H 0.21 0.21

IC50, µmol/L 132 (–40, 305) 416 (–493, 1235)

SSE 0.076 0.041

R2 0.787 0.794

dfe 13 10

Adjusted R2 0.771 0.773

RMSE 0.077 0.064

159

Table B-15. In vivo chronic toxicity data for organisms exposed to diclofenac, ibuprofen, naproxen, and ketoprofen. Compound Organism End point Exposure time EC50, μmol/L NOEC, μmol/L LOEC, μmol/L Reference

Diclofenac M. galloprovincialis Larvae development 48 h

3.14×10–6 3.14×10–5 (Fabbri et al. 2014)

Rainbow trout Histopathology 95 d

1.01

(Memmert et al. 2013)

D. magna Reproduction 21 d

26.1

(Lee et al. 2011)

O. latipes Survival 30 dph

31.4

(Lee et al. 2011)

O. latipes Survival 77 dph

31.4

(Lee et al. 2011)

M. macrocopa Reproduction 7 d

52.5

(Lee et al. 2011)

D. magna Survival 21 d

78.6

(Lee et al. 2011)

M. macrocopa Survival 7 d

157

(Lee et al. 2011)

Ibuprofen O. latipes Adult survival 120 dph

4.38×10–4

(Han et al. 2010)

O. latipes Adult survival 90 dph

4.38×10–3

(Han et al. 2010)

O. latipes Reproduction 120 dph

4.38×10–3

(Han et al. 2010)

M. galloprovincialis Larvae development 48 h

4.38×10–2 0.438 (Fabbri et al. 2014)

O. latipes Juvenile survival 30 dph

0.438

(Han et al. 2010)

Hydra attenuata Morphology 96 h 7.23 4.38 0.438 (Quinn et al. 2008)

D. magna Reproduction 21 d

5.39

(Han et al. 2010)

Hydra attenuata Feeding 96 h 16.9

(Quinn et al. 2008)

Naproxen Hydra attenuata Morphology 96 h 10.4 21.9 4.38 (Quinn et al. 2008)

C. dubia Growth/reproduction inhibition 7 d 1.31

(Isidori et al. 2005)

B. calyciflorus Growth/reproduction inhibition 48 h 2.22

(Isidori et al. 2005)

Hydra attenuata Feeding 96 h 10.6

(Quinn et al. 2008)

P. subcapitata Growth/reproduction inhibition 72 h 126

(Isidori et al. 2005)

L. peronii Tactile responsiveness 72 h 132

(Melvin et al. 2014)

L. peronii Tactile responsiveness 96 h 132

(Melvin et al. 2014)

L. peronii Tactile responsiveness 24 h 149

(Melvin et al. 2014) L. peronii Tactile responsiveness 48 h 149

(Melvin et al. 2014)

Ketoprofen P. subcapitata Growth 72 h 96.7 39.1 (Watanabe et al. 2016)

C. dubia Reproduction 6 d 116 88.5 (Watanabe et al. 2016)

Zebrafish Reproduction 9 d 58.6 24.6 (Watanabe et al. 2016)

160

Figure B-1. Fixed bed ion-exchange removal of diclofenac (DCF) by Dowex Marathon

11 fit to the homogenous surface diffusion model (HSDM).

0

0.2

0.4

0.6

0.8

1

0 2000 4000 6000

C/C

0

Bed Volume

DCF, 0.2 mmol/LHSDMR2 = 0.98SSE = 1.22

161

Figure B-2. Fixed bed ion-exchange removal of diclofenac (DCF), ketoprofen (KTP),

and naproxen (NPX) in synthetic ureolyzed urine using Dowex 22 fit to the homogenous surface diffusion model (HSDM). Data reproduced from (Landry and Boyer 2016).

0.00

0.20

0.40

0.60

0.80

1.00

0 1000 2000 3000 4000

C/C

0

Bed Volume

DCF, C0 = 4.5 μM DCF HSDMKTP, C0 = 5.9 μM KTP HSDMNPX, C0 = 5.1 μM NPX HSDM

162

Figure B-3. Cyclooxygenase subtype-1 inhbition curves for (a) diclofenac (DCF), (b)

ketoprofen (KTP), (c) naproxen (NPX), and (d) O-desmethylnaproxen (Odm-NPX). The symbols are the mean triplicate samples with error bars showing one standard deviation.

0%

20%

40%

60%

80%

100%%

CO

X-1

Inhib

itio

n

Concentration, µmol/L

(a)

DCF

Hill model

10–4 10–2 100 102 104 -20%

0%

20%

40%

60%

80%

100%

% C

OX

-1 Inhib

itio

n

Concentration, µmol/L

(b)

KTP

Hill model

10–3 10–1 101 103 105

0%

20%

40%

60%

80%

100%

% C

OX

-1 Inhib

itio

n

Concentration, µmol/L

(c) NPX

Hill model

10–4 10–2 100 102 104

-20%

0%

20%

40%

60%

80%

100%

% C

OX

-1 Inhib

itio

n

Concentration, µmol/L

(d) Odm-NPX

Hill model

10–3 10–1 101 103 105

163

Figure B-4. Alternative cyclooxygenase subtype-1 inhbition curves for (a) naproxen

(NPX), and (b) O-desmethylnaproxen (Odm-NPX) with I0 and Imax extrapolated to 0% and 100%, respectively. The symbols are the mean triplicate samples with error bars showing one standard deviation.

0%

20%

40%

60%

80%

100%%

CO

X-1

Inhib

itio

n

Concentration, µmol/L

(a) NPX

Hill model

10–5 10–3 10–1 101 103 105

0%

20%

40%

60%

80%

100%

% C

OX

-1 Inhib

itio

n

Concentration, µmol/L

(b) Odm-NPX

Hill model

10–5 10–3 10–1 101 103 105 107

164

Figure B-5. Alternative predicted COX-1 inhibition as a function of bed volumes treated

by fixed bed ion-exchange of (a) naproxen (NPX) (C0 = 3.0 µmol/L), and (b) O-desmethylnaproxen (Odm-NPX) (C0 = 1.4 µmol/L) for dose-response with I0 and Imax extrapolated to 0% and 100%, respectively.

0%

20%

40%

60%

80%

100%

0 200 400 600 800 1000Bed Volume

(a)

C/C0×100%

% Inhibition

0%

20%

40%

60%

80%

100%

0 500 1000 1500 2000 2500 3000Bed Volume

(b)

C/C0×100%

% Inhibition

165

Figure B-6. Cyclooxygenase subtype-1 inhbition curves for (a) ibuprofen (IBP), (b) OH-

ibuprofen (OH-IBP), (c) 4’OH-diclofenac (OH-DCF), and (d) ketoprofen glucuronide (KTP gluc). The symbols are the mean triplicate samples with error bars showing one standard deviation.

-50%

-25%

0%

25%

50%

75%

100%%

CO

X-1

Inhib

itio

n

Concentration, µmol/L

(a)

100 101 102 103 104 -25%

0%

25%

50%

75%

100%

% C

OX

-1 Inhib

itio

n

Concentration, µmol/L

(b)

100 101 102 103 104

-25%

0%

25%

50%

75%

100%

% C

OX

-1 Inhib

itio

n

Concentration, µmol/L

(c)

10–1 100 101 102 103 -20%

0%

20%

40%

60%

80%

100%

% C

OX

-1 Inhib

itio

n

Concentration, µmol/L

(d)

KTP gluc

Hill model

10–2 100 102 104 106

166

Figure B-7. ToxCast database in vitro bioassays for various endpoints plotted as a function of the concentration that induces 50% activity (AC50) for (a) diclofenac and (b) ibuprofen. For figure (a), the data points within the circle are the results from COX-1 and COX-2 inhbition bioassays from ToxCast and this study.

167

Figure B-8. Predicted column breakthrough curves as a function of mass removal and

COX-1 inhibition for diclofenac ion-exchange in real urine (C0 = 0.55 µmol/L).

0%

20%

40%

60%

80%

100%

0 200 400 600 800 1000

Bed Volume

C/C0×100%

% Inhibition

168

Figure B-9. Isotherm data for ion-exchange removal of (a) diclofenac, (b) ibuprofen, (c)

ketoprofen, (d) naproxen and (e) O-desmethylnaproxen in synthetic urine with and without metabolites and real human urine (DCF only). Isotherm data without metabolites reproduced from Landry et al. (2015). The symbols are the mean triplicate samples with error bars showing one standard deviation.

0

2

4

6

0 1 2 3

qe,

µm

ol/g

Ce, µmol/L

(a) Diclofenac

No metabolites

Metabolites

Real urine

0

1

2

3

4

0 1 2 3 4

qe,

µm

ol/g

Ce, µmol/L

(b) Ibuprofen

No metabolitesMetabolites

0

1

2

3

4

0 1 2 3 4

qe,

µm

ol/g

Ce, µmol/L

(c) Ketoprofen

No metabolites

Metabolites

0

1

2

3

4

0 1 2 3 4

qe,

µm

ol/g

Ce, µmol/L

(d) Naproxen

No metabolites

Metabolites

0

1

2

3

4

0 1 2 3 4

qe,

µm

ol/g

Ce, µmol/L

(e) O-Desmethylnaproxen

Metabolites

169

Figure B-10. Kinetic data for ion-exchange removal of (a) diclofenac, (b) ibuprofen, (c)

ketoprofen, and (d) naproxen and (e) O-desmethylnaproxen in synthetic urine with and without metabolites and real human urine (DCF only). The symbols are the mean duplicate samples.

0

0.2

0.4

0.6

0.8

1

0 500 1000 1500

C/C

0

Time, min

(a) DiclofenacC0 = 2.2 µmol/L

0

0.2

0.4

0.6

0.8

1

0 500 1000 1500

C/C

0

Time, min

(b) IbuprofenC0 = 3.2 µmol/L

0

0.2

0.4

0.6

0.8

1

0 500 1000 1500

C/C

0

Time, min

(c) KetoprofenC0 = 3.0 µmol/L

0

0.2

0.4

0.6

0.8

1

0 500 1000 1500

C/C

0

Time, min

(d) NaproxenC0 = 3.7 µmol/L

170

APPENDIX C SUPPLEMENTARY INFORMATION FOR CHAPTER 4

Determination of Functional Unit

Urine production at the University of Florida (UF) was estimated from solid waste

production according to UF annual refuse routes (personal communication with UF

Physical Plant Department). The refuse routes included collection locations, collection

days (e.g., Monday–Thursday, Monday and Thursday only), and dumpster volumes.

Some buildings shared dumpsters, which are labeled according to the nearest building.

If one building or cluster of buildings (e.g., dormitories) had multiple dumpsters, they

were consolidated into one cumulative-volume dumpster. This reduced a total of 188

urine-producing buildings to 125 decentralized collection areas. UF student, faculty, and

staff produce 1.46 lb waste/person∙d (0.662 kg waste/person∙d) and 0.76 lb

waste/person∙d (0.345 kg waste/person∙d) of that waste is landfilled (Townsend et al.

2015). The approximate density of landfilled waste at UF is 75.4 lb/yd3 (44.7 kg/m3)

(Townsend et al. 2015). It was assumed that at the time of collection, the dumpsters

were filled to capacity (personal communication with UF Physical Plant Division), and

non-collection days the dumpsters were assumed to be empty. Building-level waste

production and per capita waste production were used to estimate daily building

occupancy (Eq. C-1).

𝐵𝑢𝑖𝑙𝑑𝑖𝑛𝑔 𝑜𝑐𝑐𝑢𝑝𝑎𝑛𝑐𝑦 = (𝐷𝑢𝑚𝑝𝑠𝑡𝑒𝑟 𝑐𝑎𝑝𝑎𝑐𝑖𝑡𝑦, 𝑦𝑑3) (75.4𝑙𝑏

𝑦𝑑3) (𝑝𝑒𝑟𝑠𝑜𝑛−𝑑𝑎𝑦

0.76 𝑙𝑏) (C-1)

Daily urination events at the building-level were estimated based on urination

frequency and number of hours a building was assumed to be occupied (Eq. C-2).

According to a 7-day sleep log of 237 people, college-aged students on average sleep

6.40 h per night resulting in 17.6 waking h/d It was assumed that residence halls were

171

occupied for 9.14 h/d (Ishii and Boyer 2015), and all other campus buildings were

occupied for 8 h/d. Outlined in Table C-1 are the average urination volumes and

frequency for asymptomatic men and women (FitzGerald et al. 2002, Latini et al. 2004).

𝑈𝑟𝑖𝑛𝑎𝑡𝑖𝑜𝑛 𝑒𝑣𝑒𝑛𝑡𝑠 = (𝐵𝑢𝑖𝑙𝑑𝑖𝑛𝑔 𝑜𝑐𝑐𝑢𝑝𝑎𝑛𝑐𝑦)(𝐻𝑜𝑢𝑟𝑠 𝑜𝑐𝑐𝑢𝑝𝑖𝑒𝑑, ℎ) ((0.4𝑣𝑜𝑖𝑑𝑠

ℎ) (0.46 ) + (0.45

𝑣𝑜𝑖𝑑𝑠

ℎ) (0.54))

𝑈𝑟𝑖𝑛𝑎𝑡𝑖𝑜𝑛 𝑒𝑣𝑒𝑛𝑡𝑠 = (𝐵𝑢𝑖𝑙𝑑𝑖𝑛𝑔 𝑜𝑐𝑐𝑢𝑝𝑎𝑛𝑐𝑦)(𝐻𝑜𝑢𝑟𝑠 𝑜𝑐𝑐𝑢𝑝𝑖𝑒𝑑, ℎ) (0.428𝑣𝑜𝑖𝑑𝑠

ℎ) (C-2)

Based on enrollment and employment data, UF population is composed of 46%

males and 54% females (UF 2014b, 2015). The volume of urine produced daily was

estimated using Eq. C-3.

𝑉𝑜𝑙𝑢𝑚𝑒 𝑢𝑟𝑖𝑛𝑒 𝑝𝑟𝑜𝑑𝑢𝑐𝑒𝑑,𝐿

𝑑𝑎𝑦= (𝑈𝑟𝑖𝑛𝑎𝑡𝑖𝑜𝑛 𝑒𝑣𝑒𝑛𝑡𝑠) ((0.46) (0.237

𝐿

𝑣𝑜𝑖𝑑) + (0.54) (0.204

𝐿

𝑣𝑜𝑖𝑑))

𝑉𝑜𝑙𝑢𝑚𝑒 𝑢𝑟𝑖𝑛𝑒 𝑝𝑟𝑜𝑑𝑢𝑐𝑒𝑑,𝐿

𝑑𝑎𝑦= (𝑈𝑟𝑖𝑛𝑎𝑡𝑖𝑜𝑛 𝑒𝑣𝑒𝑛𝑡𝑠) (0.219

𝐿

𝑣𝑜𝑖𝑑) (C-3)

Discrete values for urine production was estimated for each day of the week

(Monday–Sunday), therefore the annual urine production was estimated by multiplying

daily urine production by the number of days that campus was assumed to be occupied

during the academic year (Table C-2). The 2014–2015 UF academic calendar was used

to estimate the number of days that students, faculty, and staff were present on campus

(UF 2014a). It was assumed campus was closed over major holidays (e.g.,

Thanksgiving, Christmas, spring break, and breaks between semesters), resulting in no

urine production (Table C-2). Table C-3 lists the estimated daily, 60-day, and annual

urine production for all of University of Florida campus.

Life Cycle Inventory

The following section describes the data sources and design parameters used to

assess the various treatment scenarios.

172

Flush Water

Estimated potable flush water requirements were based on flush water

specifications for conventional toilets (6 L/flush), conventional urinals (3.8 L/flush), and

urine diverting flush toilets (0.05 L/small flush) and the estimated number of urination

events over the course of a year at UF (Ishii and Boyer 2015, Zinckgraf et al. 2014).

Operational phase inputs (e.g., energy, chemicals, raw groundwater) for producing

potable water were included within the life cycle boundary (Ishii and Boyer 2015).

Operational costs for potable water and electricity were based on local utility rates (Ishii

and Boyer 2015).

Centralized Wastewater Treatment

At the WRF, 15% and 32% of wastewater effluent is reused as cooling water at a

cogeneration plant and landscape irrigation across campus, respectively (FDEP 2015a).

The remaining 51% is discharged by deep well injection. Deep well injection of

municipal wastewater is conducted primarily in Florida (U.S. EPA 2012). In the United

States, 14,651 out of 15,837 wastewater treatment plants discharge to surface waters

(Rice and Westerhoff 2015). To make this study transferable across communities in the

U.S. and elsewhere, it was assumed that non-reclaimed wastewater was discharged to

surface water. Only the reclaimed water and surface discharge effluent were considered

within the system boundary of the LCA; water sent to the cogeneration plant was

assumed to have negligible impact on the environment with respect to pharmaceuticals.

Waste sludge collected during secondary clarification is transported off campus to the

city’s wastewater treatment plant for further processing and land application. The

environmental impact of biosolids was considered outside the scope of this study

173

because the treatment facility ceased land application of biosolids in February 2016 and

currently disposes of biosolids in a landfill (personal communication with J.H. Hope,

June 26, 2016). Several alternative disposal options are currently under review. The

most cost effective recommended option is waste-to-energy disposal (GRU 2011).

Furthermore, the effect urine source separation has on the composition of biosolids at

centralized wastewater treatment is unknown. Jimenez et al. (2015) modeled the effect

of urine source separation on biological wastewater treatment but not necessarily how

the composition of biosolids would change, with respect to N and P content. Due to the

complexity of wastewater modeling, the N and P content of biosolids was considered

outside the scope of this model. N and P were assumed to be partially removed by

biological treatment (Ishii and Boyer 2015).

The electricity and cost requirements for urine treatment at the centralized

wastewater treatment plant were based on the influent volumetric flow of urine and urine

flush water and the flow normalized electricity use at the plant. (Ishii and Boyer 2015).

Costs were based on local utility rates (Ishii and Boyer 2015). For scenarios AWWT and

BWWT,O3, the impact of centralized wastewater treatment pertained only to inputs related

to the functional unit, i.e., the influent flow was attributed to the total volume of urine

(11,184 m3) and associated urine flush water from conventional toilets and urinals, and

did not account for additional wastewater inputs (e.g., greywater). Similarly, for

scenarios C–F, the influent flow was attributed to the total volume of urine and urine

flush water from urine diverting flush toilets. At UF, one central department (i.e.,

Physical Plant Division (PPD)) maintains all operations on campus, including irrigation

and grounds maintenance (e.g., fertilization with commercial fertilizers). For all

174

scenarios it was assumed that 84.3% and 49.8% of influent N and P mass loads were

removed during centralized treatment with no nutrient recovery (Ishii and Boyer 2015).

Pharmaceutical removal for each compound was estimated using the average

pharmaceutical removal by biological wastewater treatment in literature (Fernandez-

Fontaina et al. 2012, Hollender et al. 2009, Joss et al. 2005, Lindqvist et al. 2005,

Margot et al. 2013, Rivera-Utrilla et al. 2013, Rosal et al. 2010, Salgado et al. 2012,

Santos et al. 2007, Ternes 1998). The N and P that remain in the fraction of treated

wastewater effluent discharged to surface water was considered an emission. The N

and P in the fraction of treated wastewater used as reclaimed water in landscape

irrigation was assumed to be completely taken up by turf grass. Reclaimed water

containing 9 mg/L N may be applied at a rate of 2 cm/week without N leaching

(Hochmuth et al. 2013). This corresponds to ~8900 kg N/year that may be applied to

UF’s 235 acres of active irrigation on campus. It was estimated that 571 kg N and 558

kg N was applied to landscape irrigated with reclaimed water in scenarios A–B and C–

H, respectively. The mass of pharmaceutical remaining in wastewater effluent was

considered an emission to surface water or an emission to non-industrial, urban land for

the respective fractions discharged to surface water or used as landscape irrigation.

Ozonation of Wastewater

For scenario BWWT,O3, an additional ozonation step was added to the centralized

wastewater treatment plant in scenario AWWT to treat the influent urine and urine flush

water. Pharmaceutical destruction for each compound was estimated using the average

pharmaceutical destruction by ozonation of secondary wastewater in literature

(Hollender et al. 2009, Huber et al. 2003, Margot et al. 2013, Rosal et al. 2010, Ternes

175

1998). The system boundary included the infrastructure requirements for the ozone

contactor, production of oxygen, electricity, transport, and cooling water for ozone

production. The material inputs for infrastructure did not include the ozone generator,

due to a lack of data. The ozone contactor was sized to treat the entire influent flow at

the wastewater treatment plant (i.e., urine, flush water, feces, and greywater). The

ozone contactor was assumed to have an HRT of 5 min and was designed to meet

specifications outlined by Snyder et al. (2014). The ozone contactor was assumed to

have 4 cells (1.2 m/cell), 5.8 m of submergence, and 1.5 m of freeboard. The length and

width of the contactor were 5 m and 0.64 m, respectively (personal communication with

Mike Witwer, 2016). Material inputs for the ozone contactor only included concrete

requirements. The infrastructure costs included the total ozonation system (e.g., ozone

contactor, ozone generator, installation costs, yard piping, landscaping, electrical and

construction, and labor) (Snyder et al. 2014). Inventory data for the operational phase

(e.g., electricity, oxygen, water, and transport) were estimated on the basis of treating 1

m3 of wastewater at a full-scale plant according to Muñoz et al. (2009).

Urine Source Separation Infrastructure

There are 5,666 toilets and 1,237 urinals in 189 buildings on UF campus whose

wastewater is conveyed to the UF WRF. For scenarios C–H it was assumed that the

conventional toilets and urinals were replaced with urine diverting toilets and waterless

urinals. Conventional fixtures were replaced to make a fair economic comparison with

other scenarios that use waterless urinals and urine-diverting flush toilets. Costs for

replacing toilets and urinals (conventional and alternative fixtures) were based on

market prices (Ishii and Boyer 2015, Kohler 2016, U.S. EPA 2016d). It was assumed

176

that the urine diverting toilets had an 80% separation efficiency (Vinnerås 2001), and

that these were used exclusively by women. Waterless urinals were assumed to have

100% separation efficiency. The manufacturing and installation of conventional

(scenarios AWWT and BWWT,O3) and urine diverting fixtures (scenarios C–H) were

assumed to be equal, thus negating these fixtures in the environmental assessment.

Material and formation processes and associated costs required for pipes and storage

tanks were included in this assessment with an expected pipe lifetime of 50 years (Ishii

and Boyer 2015). A separate urine collection piping system was added to divert urine

and urine flush water (generated by urine-diverting flush toilets only) from the general

waste stream and collected in decentralized HDPE storage tanks located at 125

collections areas on campus. Pipe requirements for urine diverting toilets were based

on the requirements for a model apartment in Remy (2010) and the requirements for

urinals were assumed to be equivalent.

Urine was assumed to be stored for 60 days to inactivate potential pathogens

and/or fecal contamination (Nordin et al. 2009, Vinnerås et al. 2008). Decentralized

HDPE urine storage tanks were sized according to the estimated volume of urine

produced at each of the decentralized treatment areas on campus. For scenarios

Ctruck,landfill and Dtruck,regen, one HDPE tank was located at each decentralized collection

area and sized to hold the estimated daily maximum volume of urine produced before

being collected and transported to a central location for treatment on campus. The

material and formation inputs for the HDPE urine storage tanks per functional unit was

estimated using previous research and an expected tank lifetime of 40 years (Ishii and

Boyer 2015). Decentralized HDPE storage tanks were estimated using a linear

177

regression for tank costs as a function of storage volume (Ishii and Boyer 2015). The

centralized treatment area was equipped with two bolted steel and polyurethane lined

storage tanks, where one tank collects new urine and urine flush water while the other

stores previously collected urine and urine flush water for stabilization and disinfection.

The steel tanks were assumed to meet AWWA D-103 steel tank specifications and lined

with polyurethane to protect the steel from corrosion (AWWA 2009, Richardson 1999,

STI/SPFA 2016). Centralized urine storage tank costs were estimated using a cost

analysis tool for AWWA D-103 steel water storage tanks (STI/SPFA 2016). For

scenarios Gdecen,landfill and Hdecen,regen, urine was collected, stored, and treated at the

building level. In these scenarios, each collection area required two HDPE tanks for

simultaneous collection and storage disinfection of urine and urine flush water.

Urine Transport

For scenarios Ctruck,landfill and Dtruck,regen, urine was collected following the same refuse

routes established by UF for municipal solid waste. In SimaPro, transportation (kg∙km)

is quantified by the emissions and diesel fuel consumption for a truck that has an

efficiency of 1.72×104 kg∙km/L diesel (PRé Consultants 2014). UF refuse routes are

subdivided into north, central, and south campus routes. The roundtrip distance for each

route was estimated by plotting the dumpster locations (i.e., decentralized collection

areas) on Google Earth and using the “path” function to best guess the route and

estimated distance traveled for every day of the week, as shown in Table C-4.

For ease of calculation, decentralized areas within each route (i.e., north, central,

and south campus) were assumed to be equidistant. For example, the north route is

approximately 12 km and the 27 decentralized areas within that route were assumed to

178

be 0.43 km apart. To account for the incremental increase in weight with the addition of

urine in the vacuum truck at each pickup location, the daily urine transport for the north

(tn), central (tc), and south (ts) campus routes was estimated using Eq. C-4,

𝑡𝑛,𝑐,𝑠 = ∑ 𝑑𝑖𝑚𝑖 + 𝑑𝑖+1(𝑚𝑖 + 𝑚𝑖+1) + ⋯ + 𝑑𝑛(𝑚𝑖 + ⋯ + 𝑚𝑛)𝑛𝑖=1 (C-4)

where dn is the incremental distance between each decentralized location (km)

and mn is the mass of urine (kg) collected at each location. Discrete values for urine

transport was estimated at each decentralized area for every day of the week (Monday–

Sunday). The maximum capacity of the vacuum truck was assumed to be 4,000 gal

(15,142 L). If the cumulative daily volume at each route exceeded the maximum

capacity, it was assumed the truck stopped collecting urine and returned to the

centralized location to unload the urine before completing the route. Annual transport

was estimated by multiplying daily transport by the number of days that urine was

assumed to be collected during the academic year (Table C-2). The cost of urine

transport was estimated based on market price of diesel fuel (U.S. EIA 2015).

Vacuum Sewer System

For scenarios Esewer,landfill and Fsewer,regen, a vacuum sewer system was assumed

to be installed to convey source separated urine and urine flush water to a centralized

location on campus for further treatment. The wastewater planning model for

decentralized systems (version 1.0) (Buchanan et al. n.d.) was used to estimate cost, 4”

(102 mm) PVC pipe requirements, energy for the vacuum and wastewater transfer

pumps, and pump station to implement a vacuum sewer system servicing 188 buildings

on UF campus. It was assumed that 100% of the collection system was vacuum based

and 152.4 m was the typical distance between each source. The vacuum sewer system

179

was assumed to have a lifetime of 60 years. In SimaPro, a gravity pump station

inventory item was substituted for the vacuum pump station.

Ion-Exchange Treatment and Disposal

Bench-scale column experiments were used to estimate full-scale column design

for pharmaceutical removal. Full-scale columns were designed to achieve maximum

DCF removal, which was the pharmaceutical most selective for the resin. The operating

capacity was calculated as the mass of DCF sorbed onto the resin before removal of

DCF fell below the maximum achievable level (i.e., mass of DCF sorbed onto the resin

when DCF removal <98% after 1266 BV of treatment). The results of the treatment and

regeneration experiments are shown in Figure C-1 and Table C-5, respectively.

Columns were scaled to treat the entire volume of urine and urine flush water

collected by the source separation system, with one preconditioning cycle at the

beginning (scenarios C–H) and one regeneration cycle (scenarios Dtruck,regen,

Fsewer,regen, and Hdecen,regen) at the end of the year. Energy, water, and chemical

requirements were included for 10 BV of resin preconditioning using 5% NaCl and 10

BV of regeneration solution using 5% NaCl and 50% methanol. The column was

designed to maintain an EBCT of 8.3 min and minimum HLR of 10 m/h (Taute et al.

2013). For scenarios C–F, one large column was used to treat the entire volume of

urine, and for scenarios Gdecen,landfill and Hdecen,regen, one column was scaled to treat

urine produced annually at each decentralized location. Market values for fiberglass ion-

exchange vessels of varying sizes were used to generate a linear regression for vessel

cost as a function of volume (Fresh Water Systems 2016, Water Softeners & Filters

180

2016). The material input and cost of the fiberglass column vessel were estimated using

linear regressions shown in Figure C-2 (Choe et al. 2013).

Additional components (e.g., valves, pressure indicator, etc.) were estimated

from a pilot scale ion-exchange vessel (personal communication with a representative

at Tonka Water). A description of the components included in each ion-exchange

column and list of materials and respective masses are provided in Table C-6. Pump

power requirements (kW) was estimated as a function of flow rate using a linear

regression (Figure C-3) developed from various centrifugal pump specifications.

Disposal of the spent resin or regeneration brine was included within the system

boundary because they were considered integral to the overall life cycle impacts of the

treatment process. For scenarios Gdecen,landfill and Hdecen,regen, the fiberglass ion-

exchange columns at each decentralized location were collected and transported to a

central location for further processing. For scenarios Ctruck,landfill, Esewer,landfill, and

Gdecen,landfill, spent resin was transported and disposed of in a local Class I landfill.

Disposal of the resin was modeled as polystyrene because it constitutes the backbone

of the ion-exchange resin (Choe et al. 2013). In scenarios Dtruck,regen, Fsewer,regen, and

Hdecen,regen, the regeneration brine was transported to a local cement kiln plant where it

was incinerated. The use of waste solvents as fuel in cement production reduces the

need for fossil fuels. Ecosolvent 1.0.1 life cycle assessment tool was used to generate

the life cycle inventory data for solvent combustion as a function of the elemental

solvent composition (e.g., 5% NaCl and 50% methanol) and technology used (e.g.,

incineration at a cement kiln) (Weber et al. 2006). The resulting inventory included the

amount of fossil fuels substituted by waste solvents and changes in the atmospheric

181

emissions (Table C-7); changes to infrastructure at the cement kiln plant were not

considered. The Ecosolvent model for solvent incineration in a cement kiln was based

on average technology used in Switzerland (Seyler et al. 2005). The cost of chemicals

(i.e., methanol and NaCl) and potable water used during the preconditioning and

regeneration process were based on market price (Ishii and Boyer 2015, methanex

2015, USGS 2015).

Nutrient Recovery

For scenarios C–H, struvite (MgNH4PO4∙6H2O) precipitation of urine, after ion-

exchange treatment, was conducted to recover maximum P in urine. It was assumed

that all magnesium and calcium in collected urine was lost in the collection system due

to spontaneous precipitation of struvite and hydroxyapatite resulting in some nutrient

loss (Udert et al. 2003a). Magnesium oxide (MgO) was dosed to stored urine to achieve

a molar Mg:P ratio of 1.2:1 to achieve maximum P recovery as struvite. The cost of

magnesium oxide was based on market price (Ishii and Boyer 2015). The value of

struvite fertilizer was estimated using a regression model of common fertilizers and

costs of contributing nutrients (Ishii and Boyer 2015). In scenarios Gdecen,landfill and

Hdecen,regen, struvite precipitation was conducted every 60 days after storage

disinfection, collected and centrally stored in a HDPE tank. The recovered struvite can

be used directly as a slow-release fertilizer in place of conventional fertilizers (Johnston

and Richards 2003), and thus contributes to the environmental and cost benefits for

scenarios C–H. For these scenarios, conventional fertilizers were considered an

“avoided product” in SimaPro, represented as “monoammonium phosphate, as P2O5 at

regional storehouse/RER U” and “monoammonium phosphate, as N, at regional

182

storehouse/RER U” (NH4H2PO4) because it is a commonly used fertilizer product

containing both N and P. It was assumed that the quality and size of precipitated

struvite was comparable to commercial granular fertilizers and that identical commercial

spreading machines were used for struvite or commercial fertilizers (Forrest et al. 2008).

This assumption negates the power requirements for spreading struvite or

monoammonium phosphate. Furthermore, emissions (e.g., ammonia, nitrous oxide, and

phosphorus) for commercial fertilizers and struvite fertilizers were assumed to be

equivalent. Ammonia emissions for multi-nutrient fertilizers (e.g., struvite, and

monoammonium phosphate) are quantified by an identical emission factor of 4%

(Nemecek and Kägi 2007). Furthermore, nitrous oxide and phosphorus emissions are

quantified as a function of N and P content in the fertilizer (Nemecek and Kägi 2007).

Those emissions were also negated because struvite fertilizer is assumed to offset the

equivalent amount of N and P in monoammonium phosphate. One of the benefits of

struvite precipitation from urine is the low heavy metal content compared to commercial

fertilizers. However, studies have shown that heavy metals in urine (e.g.,) cadmium may

be incorporated into the final struvite product, although to a much lesser extent than

what is found in commercial fertilizers (Lugon-Moulin et al. 2006, Ronteltap et al. 2007).

Therefore, cadmium emissions from struvite and monoammonium phosphate was

included within the LCA boundary. Cadmium content of struvite and monoammonium

phosphate was assumed to be 0.397 mg Cd/kg P2O5 and 97.5 mg Cd/kg P2O5,

respectively (Lugon-Moulin et al. 2006, Ronteltap et al. 2007).

Estimation of Pharmaceuticals in Urine

183

Pharmaceutical concentrations in urine can vary as a function of fraction of

population use, duration of use, and rate of urine collection. A system of equations was

used to estimate pharmaceutical concentrations in urine as a function of these

variables. All calculations were based on urinary excretion rates of the pharmaceutical

active ingredient in urine, pharmaceutical metabolites were not considered in this study.

Sample calculations are provided using ibuprofen as the example pharmaceutical.

According to Khan and Nicell (2010) the pharmaceutical concentration in urine can be

estimated as a function of the percentage of the population that is using a particular

pharmaceutical and the maximum pharmaceutical concentration that may be present if

100% of the population consumed pharmaceuticals, as shown in Eq. C-5 (Khan and

Nicell 2010).

𝐹𝑐 =𝐶

𝐶100 (C-5)

Where Fc is the fraction of the population that is currently using a pharmaceutical,

C is the pharmaceutical concentration in urine and C100 is the pharmaceutical

concentration in urine if 100% of the population consumed a pharmaceutical. This

equation can be rearranged so that the concentration in urine can be determined by Eq.

C-6:

𝐶 = 𝐹𝑐 × 𝐶100 (C-6)

C100 may be estimated by Eq. C-7:

𝐶100 =𝐷𝐷𝐷×𝐹𝑒𝑥

𝑈 (C-7)

Where DDD (mg) is the defined daily dose of a pharmaceutical (Table C-8). In

this example, the defined daily dose was based on the World Health Organization

recommendation (Table C-8) (Holloway and Green 2003), Fex is the fraction of the

184

consumed dose excreted in urine as the pharmaceutical active ingredient based on

pharmacokinetics in literature (Table C-8), and U is the average urine excretion volume

per person per day which is estimated to be 1.5 L/p/d (FitzGerald et al. 2002, Latini et

al. 2004).

The pharmaceutical concentration can be diluted by two additional factors:

duration of pharmaceutical use and urine storage collection time. Duration of

pharmaceutical use is expressed as a fraction of the urine storage collection time. For

example, if the urine storage collection time was one week and the maximum daily dose

of a pharmaceutical was consumed for one week, then the pharmaceutical was

consumed for 100% of the collection time. Conversely, if the pharmaceutical was

consumed 1 day out of the 7-day collection time, the pharmaceutical was consumed for

14.3% of the collection time. The percent duration of use can be determined by Eq. C-8:

𝐹𝑑 =𝑑𝑢

𝑑𝑐 (C-8)

Where Fd is the fraction of duration of use compared to collection time, du is the

number of days a pharmaceutical was consumed, and dc is the storage collection time

in days. The total concentration in urine over the entire collection period may be

calculated by Eq. C-9:

𝐶𝑇 = 𝐶 × 𝐹𝑑 = 𝐶100 × 𝐹𝑐 × 𝐹𝑑 (C-9)

Where CT is the total pharmaceutical concentration in collected urine.

Ex. What is the concentration of ibuprofen in urine if 100 out of 100 students in a dorm consumed ibuprofen for 7 days and the urine collection time was 7 days? What if 25% of the students consumed ibuprofen for 4 days?

𝐹𝑢 =100

100= 1

𝐶 = 56𝑚𝑔

𝐿× 1 = 56

𝑚𝑔

𝐿

185

𝐹𝑑 =7

7= 1

𝐶𝑇 = 56𝑚𝑔

𝐿× 1 = 56

𝑚𝑔

𝐿

Intuitively, this value makes sense because if the entire population of students consumes ibuprofen for the entire collection period, the pharmaceutical concentration would not be diluted and would simply equal C100. Ex. What if 25% of the students consumed ibuprofen for 4 days? 𝐹𝑢 = 0.25

𝐶 = 56𝑚𝑔

𝐿× 0.25 = 14

𝑚𝑔

𝐿

𝐹𝑑 =4

7= 0.57

𝐶𝑇 = 14𝑚𝑔

𝐿× 0.57 = 7.98

𝑚𝑔

𝐿

One final variable that could impact pharmaceutical concentrations in urine is if

different fractions of the population consume pharmaceuticals for varying durations of

time. For example, if 25% of the population consumed a pharmaceutical for 10% of the

collection time and then 15% of the population consumed the same pharmaceutical for

25% of the collection time. The concentration of pharmaceuticals in urine is simply the

summation of the total pharmaceutical concentration in urine according to each

pharmaceutical consumption scenario, as outlined in Eq. C-10:

CTn= CT1

+ CT2+ ⋯ CTn = 𝐶100 ∑ 𝐹𝑐𝑛

𝐹𝑑𝑛

𝑛𝑖=1 (C-10)

Ex. What is the ibuprofen concentration in urine if 25 of 100 students in the dorm population consumed the pharmaceutical for 2 out of 7 days of collection time and a few days later 15 students consumed ibuprofen for 4 out of 7 days of the collection time?

𝐶𝑇 = (56𝑚𝑔

𝐿×

25

100×

2

7) + (56

𝑚𝑔

𝐿×

15

100×

4

7) = 4

𝑚𝑔

𝐿+ 4.8

𝑚𝑔

𝐿

𝐶𝑇 = 8.8𝑚𝑔

𝐿

This value can be confirmed by individually calculating the mass pharmaceutical load (Mpharm) for each population fraction and determining the concentration of pharmaceuticals for the entire storage collection period.

186

𝑀𝑝ℎ𝑎𝑟𝑚1= (25 𝑝𝑒𝑜𝑝𝑙𝑒)(2 𝑑𝑎𝑦𝑠) (1.5

𝐿

𝑝 ∙ 𝑑 ) (56

𝑚𝑔

𝐿) = 4,200 𝑚𝑔

𝑀𝑝ℎ𝑎𝑟𝑚2= (15 𝑝𝑒𝑜𝑝𝑙𝑒)(4 𝑑𝑎𝑦𝑠) (1.5

𝐿

𝑝 ∙ 𝑑) (56

𝑚𝑔

𝐿) = 5,040 𝑚𝑔

Total pharmaceutical mass load: 𝑀𝑇 = 4,200 𝑚𝑔 + 5,040 𝑚𝑔 = 9,240 𝑚𝑔 Total urine production over 7 days:

(100 𝑠𝑡𝑢𝑑𝑒𝑛𝑡𝑠) (1.5𝐿

𝑝 ∙ 𝑑) (7 𝑑𝑎𝑦𝑠) = 1,050 𝐿

Total pharmaceutical concentration in urine:

𝐶𝑇 =9,240 𝑚𝑔

1,050 𝐿= 8.8

𝑚𝑔

𝐿

Figure C-4 is a frequency diagram of all the possible ibuprofen concentrations in

urine for a community for Fc and Fd range from 1% to 100%. The relative frequency is

skewed to the right, resulting in a non-normal distribution. For a non-normal distribution,

the central tendency is best measured by the median value in the dataset (Ott and

Longnecker 2004), or 10.7 mg/L ibuprofen. For this model, it was assumed that DCF,

IBP, KTP, and NPX were present in urine at concentrations of 767, 10,735, 4,792, and

831 μg/L, respectively (Table C-8). A lognormal distribution was assumed for the data

with a standard deviation of 1.31. The minimum and maximum concentrations in urine

were based on the 95% confidence interval of the lognormal distribution and is defined

by dividing or multiplying the median pharmaceutical concentration in urine with the

squared standard deviation (i.e., 1.71).

Materials and Methods for Bench Scale Ion-Exchange Column Experiments

Synthetic Human Urine

Synthetic ureolyzed human urine was used for all experiments. The urine

composition was based on previous work (Landry et al. 2015), with adjustment to

187

include the six endogenous metabolites present at the greatest concentrations in urine

(Bouatra et al. 2013).

Pharmaceutical Compounds

Four pharmaceuticals were investigated for this study; the chemical

characteristics were described previously (Landry et al. 2015). Diclofenac sodium (CAS

15307-79-6, MP Biomedicals), ibuprofen (CAS 311-21-95-4, Fluka Analytical), naproxen

(CAS 26159-54-2, Sigma-Aldrich), and ketoprofen (CAS 22071-15-4, Sigma-Aldrich)

are all acidic pharmaceuticals from the non-steroidal anti-inflammatory drugs (NSAIDs)

pharmaceutical class. A stock solution containing 1,000 mg/L of each solution was

made by diluting the pharmaceutical salts in methanol.

Ion-Exchange Resin

A strong-base, polymeric anion exchange resins (AER), Dowex 22, was used in

all experiments. Dowex 22 is a strong-base, macroporous polystyrene AER

functionalized with dimethyl ethanol functional groups with a manufacturer’s total

capacity of 1.2 meq/mL. The AER was pre-conditioned using NaCl, following a method

described elsewhere (Landry and Boyer 2013).

Column Tests

Fixed bed column runs were conducted in a glass column (0.7854 cm diameter)

packed with 6 mL of Dowex 22 AER to obtain a height: diameter of at least 2:1

(Edzwald 2011). All column runs were performed under the same conditions by

maintaining an empty bed contact time (EBCT) and flow rate of 8.3 min and 0.72

mL/min, respectively. Synthetic ureolyzed urine was spiked with the pharmaceutical

stock solution at an initial concentration of 1,000 µg/L. 100 mL of sample was collected

188

every 12 h using an IS-95 interval sampler (Spectra/Chrom). Control samples were

collected at the beginning and end of the column experiment. At the end of the run, the

column was regenerated using 10 BV of regeneration solution containing 5% NaCl, 50%

methanol by maintaining an EBCT and flow rate of 16.7 min and 0.36 mL/min,

respectively.

Sample Preparation

Pharmaceutical samples from the column experiments were separated from the

urine matrix using a solid phase extraction (SPE) vacuum station (Supelco Visiprep)

and phenyl SPE columns (SiliaPrep, SiliCycle), evaporated, and reconstituted following

a previously described method (Magiera et al. 2014). The dry residue was dissolved in 1

mL of mobile phase (acetonitrile:25 mM KH2PO4 (pH 3) (40:60; v/v)) and 100 µL was

injected into the HPLC-UV system (Hewlett Packard 1050 series detector and Agilent

1100 series auto sampler).

Analytical Methods

Pharmaceutical concentrations for the batch regeneration experiments were

measured using UV absorbance (Hitachi U-2900) following a method described

elsewhere (Landry and Boyer 2013). Pharmaceutical concentrations for the column

experiments were measured using HPLC-UV (Hewlett Packard 1050 series detector

and Agilent 1100 series auto sampler) at 230 nm, equipped with a reversed-phase

column (2.1 × 150 mm, 3 μm Ascentis RP-amide column; Supelco, Bellefonte, PA). The

mobile phase consisted of (A) a mixture of acetonitrile and 25 mM KH2PO4 (pH 3)

(40:60 v/v), and (B) HPLC grade acetonitrile. Elution was performed by increasing

mobile phase B from 0% (5 min) to 50% (20 min), hold for 1 min, and decrease to 0%

189

(21.5 min) to re-equilibrate the baseline for 9.5 min. A seven-point calibration curve (0,

10, 50, 100, 500, 1,000, 5,000 µg/L) was created by serial dilution of the stock

standards previously mentioned in mobile phase A. The minimum limit of detection

(LOD) was 50 μg/L for DCF and IBP, 10 μg/L for KTP and NPX. Pharmaceutical

concentrations were set to the LOD if the effluent concentration fell below the LOD. One

DCF sample in the treatment cycle fell below the LOD. Eight KTP samples, two NPX

samples, seven IBP samples, and two DCF samples fell below the LOD in the

regeneration cycle.

190

Table C-1. Average urination volumes and frequency for asymptomatic men and women.

Men Women

Volume, L/d 1.65a 1.62b

Total daytime voids 7a 8b

Total nighttime voids 0a 0b

Mean voided volume, mL/void 237a 204b

Average hours awake, h c 17.6 17.6 Urination frequency, void/h d 0.40 0.45 a Latini et al. (2004) b FitzGerald et al. (2002) c Tsai and Li (2004) d Calculated by dividing total daytime voids by average hours awake

191

Table C-2. Total number of weekdays during the fall, spring, and summer semesters, excluding major holidays; data from the University of Florida 2014–2015 academic calendar.

Monday Tuesday Wednesday Thursday Friday Saturday Sunday

Fall 17 17 15 16 15 15 15 Spring 15 16 16 16 16 14 14 Summer 12 12 12 12 12 11 10

192

Table C-3. Estimated urine production for entire UF campus over different time periods. Time Urine production, m3

Dailya 39.7 60-dayb 2,381 Annual 11,184 a Average daily urine production over the week b Average 60-day urine production

193

Table C-4. Daily refuse route distance (km) traveled during fall, spring, and summer semesters.

Monday Tuesday Wednesday Thursday Friday Saturday Sunday

Fall & Spring

North Campus 12 12 12 12 12 10 0

Central Campus 14 16 14 10 16 8 0

South Campus 14 18 15 20 13 10 0

Summer

North Campus 12 8 12 9 12 10 0

Central Campus 13 12 14 10 14 8 0

South Campus 11 15 14 20 11 10 0

194

Table C-5. Mass of diclofenac (DCF), ibuprofen (IBP), ketoprofen (KTP), and naproxen (NPX) sorbed onto AER (mg) and desorbed from AER using a 5% NaCl, 50% methanol regeneration solution.

Pharmaceutical Mass sorbed (mg) Mass desorbed (mg) % Regeneration

DCF 25.6 3.63 14% IBP 2.4 0.331 14% KTP 18.7 0.15 1% NPX 4.6 1.346 30%

195

Table C-6. Inventory data for ion-exchange vessel components; data provided by Tonka Water (personal communication).

Component Material Mass, kgd

Ball valves with lever Brass 1.34 Air release valve PVC 0.195 Pressure indicatorsa Steel 0.029 Aluminum 0.029 Bronze 0.029 Brass 0.029 1” Tee connectors to tank HDPE 3.63×10–3

1” ID × 1/4” OD Tubingb PVDF 0.6 1/4” ID × 3/8” ODc Tubingc PVDF 0.490 a Pressure indicator composed of multiple materials, total mass of pressure indicator equally distributed among components b Tubing used for ion-exchange vessel in scenarios C–F only c Tubing used for ion-exchange vessels in scenarios G and H only d Non-normalized mass of components; mass of ion-exchange vessel and components normalized by 40 year lifetime in LCA inventory

196

Table C-7. Inventory data for incineration of a regeneration brine at a cement kiln plant. kg per m3 of total regeneration solution volume (e.g., water + methanol)

Regeneration solution 5% NaCl, 50% Methanola

Hard coal –148b

Heavy fuel oil –54.5 Carbon dioxide –76.3 Carbon dioxide fuel –621

Nitrogen oxides –1.67 Nickel –1.11×10–7 Copper –6.59×10–7 Zinc –1.10×10–4 Metals unspecified –1.43×10–5 Arsenic –3.09×10–7 Cadmium –4.45×10–5 Chromium –9.39×10–7 Mercury –8.86×10–5 Lead –1.01×10–4 a Inventory data obtained from Ecosolvent 1.0.0 software (Weber et al. 2006) b Negative values indicate an avoided impact (i.e., offset)

197

Table C-8. Recommended defined daily dose (DDD), fraction of dose excreted in urine as the parent compound (Fex), and estimated pharmaceutical concentrations in urine.

Pharmaceutical DDD, mg Fex Median (minimum, maximum), μg/L

Diclofenac (DCF) 100a 0.06b 767 (450, 1,308) Ibuprofen (IBP) 1,200a 0.07c 10,735 (6,294, 18,309) Ketoprofen (KTP) 300a 0.06d 1,230 (721, 2,098) Naproxen (NPX) 500a 0.013e 831 (487, 1,417) a WHOCC (2013) b Sawchuk et al. (1995) c Lienert et al. (2007b) d Houghton et al. (1984) e Sugawara et al. (1978)

198

Table C-9. Unit cost of inventory items.

Input Unit Unit pricef Year of original cost data

Justification Impacted Scenarios

Infrastructure

Ozone system (30 years)a $ unit–1 $2,808,853 2011 Cost regression curve, as a function of plant capacity B

Conventional toilet (25 years)b $ fixture–1 $203 2015 Ishii and Boyer (2015) A, B

Conventional urinal (25 yearsb $ fixture–1 $355 2015 U.S. market price, www.us.kohler.com A, B

Urine diverting flush toilet (25 years)b $ fixture–1 $700 2016 Ishii and Boyer (2015) C–H

Waterless urinal (25 years)b $ fixture–1 $740 2016 U.S. market price, www.us.kohler.com C–H

Urine piping (50 years)b $ fixture–1 $25 2015 Ishii and Boyer (2015) C–H

Steel and HDPE lined urine central storage tanks (40 years)c

$ m–3 $272 2008 Cost estimation tool (STI/SPFA 2016) C–F

HDPE storage tanks (40 years)b $ m–3 $29 2015

Linear regression of vessel cost as a function of volume (Ishii and Boyer 2015)

C–H

Fiberglass ion-exchange vessel (10 years)d

$ m–3 $1,955 2015

Linear regression of vessel cost as a function of volume (Figure C-2)

C–H

4” PVC vacuum sewer pipe (60 years)e $ m–1 $10 2009 Wastewater planning model (Buchanan et al. n.d.) E, F

Vacuum sewer station (60 years)e $ unit–1 $503,928 2009 Wastewater planning model (Buchanan et al. n.d.) E, F

Operation

Potable water $ m–3 $0.92 2015 Local utility rates (Ishii and Boyer 2015) A–H

Liquid oxygen $ kg–1 $0.12 2012 U.S. market price (Carollo Engineers 2012) B

Vacuum sewer annual maintenance $ yr–1 $29,677 2009 Wastewater planning model (Buchanan et al. n.d.) E, F

Anion exchange resin $ L–1 $12 2016 U.S. market price, www.apswater.com C–H

Electricity $ kWh–1 $0.10 2015 Local utility rates (Ishii and Boyer 2015) A–H

Sodium chloride $ kg–1 $0.20 2014 U.S. market price (USGS 2015) C–H

Methanol $ m3 $322 2015 U.S. market price, www.methanex.com D, F, H

Diesel fuel $ L–1 $0.73 2015 U.S. market price, (U.S. EIA 2015) A–H

Magnesium oxide $ kg–1 $0.21 2015 Ishii and Boyer (2015) C–H

Struvite profit $ kg–1 $0.57 2013 Ishii and Boyer (2015) C–H a Carollo Engineers (2012) b Ishii and Boyer (2015) c Guishard (n.d.) d Choe et al. (2013) e Buchanan et al. (n.d.) f Infrastructure costs and vacuum sewer operation costs adjusted to 2016 based on inflation (www.bls.gov)

199

Table C-10. USEtox characterization factors (human toxicity in cases/kg and ecotoxicity in PAF·m3·day/kg) for diclofenac, ibuprofen, ketoprofen, and naproxen.

Pharmaceutical

Emissions to freshwater Emissions to soil

Reference FAETP HTP-NC FAETP HTP-NC

Diclofenac 2,670 1.22×10–6 105 1.24×10–6 Alfonsín et al. (2014) Ibuprofen 209 3.71×10-7 3.67 1.51×10–7 Alfonsín et al. (2014) Ketoprofena 113 – 6.92 – Andersson et al. (2007), Morais (2014)b

Naproxen 218 2.95×10–7 4.86 4.26×10–8 Alfonsín et al. (2014) a Characterization factors calculated using USEtox 2.0 (Hauschild et al. 2015) b Toxicity data references

200

Table C-11. Baseline, minimum, and maximum values used for various input parameter assumptions. Minimum and maximum values used to conduct a sensitivity analysis of TRACI impact assessment results and uncertainty analysis with assumed distribution (e.g., uniform, normal, lognormal).

Assumption Baseline Minimum Maximum Basis for assumption

Pharmaceutical concentrations in urine, μg/L

767 (DCF) 10,735 IBP) 1,230 KTP) 831 (NPX)

450 (DCF) 6,294 IBP) 721 (KTP) 487 (NPX)

1,308 (DCF) 18,309(IBP) 2,098 (KTP) 1,417 (NPX)

Range of all possible concentrations in urine estimated based on DDD, urinary excretion rates, and theoretical fraction of the population consuming the pharmaceutical for a theoretical length of time. Data is positively skewed; baseline is median value. Lognormal distribution assumed with minimum and maximum values determined by dividing or multiplying baseline with the squared standard deviation.

Pharmaceutical removal by biological treatment, %

27.5 (DCF) 87.3 (IBP) 54.9 (KTP) 71.1 (NPX)

5 (DCF) 40 (IBP) 10 (KTP) 0 (NPX)

90 (DCF) 100 (IBP) 98 (KTP) 98 (NPX)

Uniform distribution of pharmaceutical removal by biological wastewater treatment in literature. Baseline average of literature values.

Pharmaceutical removal by ozonation, %

97.8 (DCF) 53.1 (IBP) 76.7 (KTP) 79.5 (NPX)

94 (DCF) 32 (IBP) 63 (KTP) 50 (NPX)

100 (DCF) 77 (IBP) 98 (KTP) 90 (NPX)

Uniform distribution of pharmaceutical removal by ozonation of wastewater in literature. Baseline average of literature values.

Pharmaceutical removal by ion-exchange, %a

98.4 (DCF) 17.1 (IBP) 45.9 (KTP) 36.2 (NPX)

98.4 (DCF) 17.1 (IBP) 45.9 (KTP) 36.2 (NPX)

98.4 (DCF) 98.4 (IBP) 98.4 (KTP) 98.4 (NPX)

Arbitrary; baseline from experimental column results and maximum based on the assumption that an AER may be developed to achieve equivalent removal as diclofenac for all pharmaceuticals

Capacity of resina 5.52×10–3

meq/mL DCF 3.07×10–4 meq/mL IBP

5.52×10–3 meq/mL DCF

Baseline capacity of resin based on maximum diclofenac removal, minimum capacity of resin based on maximum ibuprofen removal

Urine storage time, days 60 14 180 Uniform distribution; min based on optimal storage conditions, max based on WHO recommendation (Ishii and Boyer 2015).

N content in urine, kg/m3 6.9 4.89 12.07 Uniform distribution of urine nitrogen concentrations in literature (Ishii and Boyer 2015). Baseline average of literature values.

P content in urine, kg/m3 0.559 0.37 0.80 Uniform distribution of urine phosphorus concentrations in literature (Ishii and Boyer 2015). Baseline average of literature values.

Electricity use at drinking water treatment plant to produce potable flush water (kWh/m3)

0.558 0.533 0.583 Normal distribution of data provided by drinking water treatment plant, Min = Mean – 2 St. Dev., Max = Mean + 2 St. Dev (Ishii and Boyer 2015).

Electricity use at wastewater treatment plant to treat influent urine and flush water (kWh/m3)

1.366 0.777 1.955 Normal distribution of data provided by wastewater treatment plant, Min = Mean – 2 St. Dev., Max = Mean + 2 St. Dev (Ishii and Boyer 2015).

a Capacity of resin and pharmaceutical removal by ion-exchange excluded from the uncertainty analysis due to a lack of data and arbitrarily assumed values.

201

Table C-12. Baseline, minimum, and maximum values used for various cost assumptions. Minimum and maximum values used to conduct a sensitivity analysis of the economic costs and uncertainty analysis assuming a uniform distribution.

Input Unit Baseline Min Max Justification

Interest rate 3% 3% 7% Interest rates of 3%, 5%, and 7% were evaluated (National Center for Environmental Economics 2010)

Infrastructure

Urine diverting flush toileta $ fixture–1 $700 $203 $700

Minimum price assumes demand for urine diverting flush toilets increases, driving costs down to meet cost of conventional toilets.

Waterless urinala $ fixture–1 $740 $355 $740 Minimum price assumes demand for waterless urinals increases, driving costs down to meet cost of conventional urinals

Operation

Potable water $ m–3 $0.92 $0.51 $4.17 Range of U.S. water rates by city based on 2014 data (Walton 2014). Baseline based on local utility rates (Ishii and Boyer 2015).

Anion exchange resin

$ L–1 $12 $7 $18 Range of U.S. market prices (www.apswater.com). Baseline average of market values.

Electricity $ kWh–1 $0.10 $0.07 $0.33

Range of U.S. energy rates by state based on 2014 data (U.S. EIA 2016). Baseline based on local utility rates (Ishii and Boyer 2015).

Sodium chloride $ kg–1 $0.20 $0.19 $0.20 Range of U.S. market prices for vacuum and open pan salt based on 2010-2014 data (USGS 2015).

Methanol $ m–3 $322 $95 $660 Range of U.S. methanol market prices based on 2001-2016 data (www.methanex.com).

Diesel fuel $ L–1 $0.73 $0.53 $1.24 Range of U.S. diesel market prices based on 2007-2016 data (U.S. EIA 2015).

Struvite profit $ kg–1 $0.57 $0.00 $1.35 95% confidence interval of linear regression model (Ishii and Boyer 2015).

a Cost of fixtures excluded from the uncertainty analysis, only included in sensitivity analysis to evaluate effect of decreasing fixture cost

202

Figure C-1. Bench scale column results for removal of diclofenac (DCF), ibuprofen (IBP), ketoprofen (KTP), and naproxen (NPX) by anion-exchange resin.

0%

25%

50%

75%

100%

0 1000 2000 3000 4000

% R

em

oval

Bed Volumes

KTP, C0=1472 µg/L NPX, C0=1256 µg/LIBP, C0=1120 µg/L DCF, C0=1409 µg/L

203

Figure C-2. Manufacturer data and resulting linear regressions of fiberglass water softener tank (a) empty weight (kg) as a function of volume (m3) and (b) cost ($) as a function of volume (m3); data provided by waterpurification.pentair.com, reskem.com, freshwater.

y = 140.29xR² = 0.9294

0

200

400

600

800

1000

0 2 4 6 8

Weig

ht, k

g

Volume, m3

(a)

y = 1928.8xR² = 0.6983

0

200

400

600

800

1000

1200

0 0.2 0.4 0.6 0.8 1

Cost, $

Volume, m3

(b)

204

Figure C-3. Manufacturer data and resulting linear regressions of centrifugal pump power specifications; data provided by grainger.com and northerntool.com.

y = 0.8461xR² = 0.778

0

1

2

3

4

0 1 2 3 4 5

kW

m3/h

205

Figure C-4. Relative frequency diagram of ibuprofen concentrations in urine for a community where 1–100% of the population is consuming ibuprofen (Fc) for 1–100% of the collection time (Fd).

0

0.02

0.04

0.06

0.08

1 8

15

22

29

36

43

50

Rela

tive F

requency

Ibuprofen, mg/L

206

Figure C-5. Normalized TRACI impact score for centralized wastewater treatment and

urine source separation. Each bar represents TRACI impact categories (e.g., fossil fuel depletion, respiratory effects, carcinogenics). The brackets around each error bar represent the 95% confidence interval resulting from Ecoinvent database distributions from the Monte Carlo uncertainty analysis.

207

Figure C-6. Comparison of ozone depletion impacts (kg CFC-11 eq.) due to

contributing processes (e.g., flush water, urine transport), generated emissions (e.g., nutrient discharge, pharmaceutical discharge), and avoided impacts (e.g., fertilizer offsets, brine incineration) in each scenario. The brackets around each error bar represent the 95% confidence interval resulting from Ecoinvent database distributions from the Monte Carlo uncertainty analysis.

208

Figure C-7. Comparison of global warming impacts (kg CO2 eq.) due to contributing

processes (e.g., flush water, urine transport), generated emissions (e.g., nutrient discharge, pharmaceutical discharge), and avoided impacts (e.g., fertilizer offsets, brine incineration) in each scenario. The brackets around each error bar represent the 95% confidence interval resulting from Ecoinvent database distributions from the Monte Carlo uncertainty analysis.

209

Figure C-8. Comparison of smog impacts (kg O3 eq.) due to contributing processes

(e.g., flush water, urine transport), generated emissions (e.g., nutrient discharge, pharmaceutical discharge), and avoided impacts (e.g., fertilizer offsets, brine incineration) in each scenario. The brackets around each error bar represent the 95% confidence interval resulting from Ecoinvent database distributions from the Monte Carlo uncertainty analysis.

210

Figure C-9. Comparison of acidification impacts (kg SO2 eq.) due to contributing

processes (e.g., flush water, urine transport), generated emissions (e.g., nutrient discharge, pharmaceutical discharge), and avoided impacts (e.g., fertilizer offsets, brine incineration) in each scenario. The brackets around each error bar represent the 95% confidence interval resulting from Ecoinvent database distributions from the Monte Carlo uncertainty analysis.

211

Figure C-10. Comparison of eutrophication impacts (kg N eq.) due to contributing

processes (e.g., flush water, urine transport), generated emissions (e.g., nutrient discharge, pharmaceutical discharge), and avoided impacts (e.g., fertilizer offsets, brine incineration) in each scenario. The brackets around each error bar represent the 95% confidence interval resulting from Ecoinvent database distributions from the Monte Carlo uncertainty analysis.

212

Figure C-11. Comparison of carcinogenic impacts (CTUh) due to contributing

processes (e.g., flush water, urine transport), generated emissions (e.g., nutrient discharge, pharmaceutical discharge), and avoided impacts (e.g., fertilizer offsets, brine incineration) in each scenario. The brackets around each error bar represent the 95% confidence interval resulting from Ecoinvent database distributions from the Monte Carlo uncertainty analysis.

213

Figure C-12. Comparison of respiratory effects impacts (kg PM2.5 eq.) due to

contributing processes (e.g., flush water, urine transport), generated emissions (e.g., nutrient discharge, pharmaceutical discharge), and avoided impacts (e.g., fertilizer offsets, brine incineration) in each scenario. The brackets around each error bar represent the 95% confidence interval resulting from Ecoinvent database distributions from the Monte Carlo uncertainty analysis.

214

Figure C-13. Comparison of fossil fuel depletion impacts (MJ surplus) due to

contributing processes (e.g., flush water, urine transport), generated emissions (e.g., nutrient discharge, pharmaceutical discharge), and avoided impacts (e.g., fertilizer offsets, brine incineration) in each scenario. The brackets around each error bar represent the 95% confidence interval resulting from Ecoinvent database distributions from the Monte Carlo uncertainty analysis.

215

Figure C-14. Impact assessment results for methanol, sodium chloride, and potable

water production used in the regeneration process (positive (+) percent contributions), compared to CO2 and NOx emission offsets, heavy fuel offsets, and hard coal offsets from incineration of the regeneration brine at a cement kiln plant (negative (–) percent contributions). The total bar length is equal to 100% of the impact within an impact category.

216

Figure C-15. Normalized TRACI impact score (PE) of vacuum truck collection compared to the vacuum sewer collection as a function of vacuum sewer pipe length or distance traveled by vacuum truck (km).

0

100

200

300

400

500

600

0 25000 50000 75000 100000

TR

AC

I score

, P

E

Pipe length or distance traveled, km

Vacuum sewerVacuum truck

217

Figure C-16. Comparison of non-carcinogenic human toxicity impact (CTUh = number

of disease cases) due to (a) contributing processes (e.g., flush water, urine transport) and generated emissions (e.g., nutrients, pharmaceuticals) and avoided impacts (e.g., P offsets, N offsets) in each scenario and (b) pharmaceutical emissions only. The brackets around each error bar represent the 95% confidence interval resulting from Ecoinvent database distributions from the Monte Carlo uncertainty analysis.

(a)

218

LIST OF REFERENCES

Alberti, G., Amendola, V., Pesavento, M. and Biesuz, R. (2012) Beyond the synthesis of novel solid phases: Review on modelling of sorption phenomena. Coordination Chemistry Reviews 256(1–2), 28-45.

Alfonsín, C., Hospido, A., Omil, F., Moreira, M.T. and Feijoo, G. (2014) PPCPs in wastewater – Update and calculation of characterization factors for their inclusion in LCA studies. Journal of Cleaner Production 83, 245-255.

Althaus, H.J., Hischier, R. and Osses, M. (2007) Life Cycle Inventories of Chemicals, Data v2.0.

Alvarino, T., Suarez, S., Lema, J.M. and Omil, F. (2014) Understanding the removal mechanisms of PPCPs and the influence of main technological parameters in anaerobic UASB and aerobic CAS reactors. Journal of Hazardous Materials 278, 506-513.

Andersson, J., EKHEDEN, Y., KAJ, L., REMBERGER, M. and WOLDEGIORGIS, A. (2007) Results from the Swedish National Screening Programme, 2005. Subreport 1: Antibiotics, Anti-Inflammatory Substances, and Hormones.

AOPWiki (2016), Cyclooxygenase inhibition leading to reproductive dysfunction. https://aopwiki.org/aops/63 (accessed 01/2017)

Apell, J.N. and Boyer, T.H. (2010) Combined ion exchange treatment for removal of dissolved organic matter and hardness. Water Research 44(8), 2419-2430.

AWWA (2009) AWWA D103-09 Factory-Coated Bolted Carbon Steel Tanks for Water Storage.

Bäuerlein, P.S., ter Laak, T.L., Hofman-Caris, R.C.H.M., de Voogt, P. and Droge, S.T.J. (2012) Removal of charged micropollutants from water by ion-exchange polymers – Effects of competing electrolytes. Water Research 46(16), 5009-5018.

Beltran, A., Borrull, F., Marcé, R.M. and Cormack, P.A.G. (2010) Molecularly-imprinted polymers: useful sorbents for selective extractions. TrAC Trends in Analytical Chemistry 29(11), 1363-1375.

Benotti, M.J., Trenholm, R.A., Vanderford, B.J., Holady, J.C., Stanford, B.D. and Snyder, S.A. (2008) Pharmaceuticals and Endocrine Disrupting Compounds in U.S. Drinking Water. Environmental Science & Technology 43(3), 597-603.

Berndtsson, J.C. (2006) Experiences from the implementation of a urine separation system: Goals, planning, reality. Building and Environment 41(4), 427-437.

219

Bester-Rogac, M. (2009) Nonsteroidal Anti-Inflammatory Drugs Ion Mobility : A Conductometric Study of Salicylate, Naproxen, Diclofenac and Ibuprofen Dilute Aqueous Solutions. Slovenian Chemical Society 56(1), 8.

Blair, B., Nikolaus, A., Hedman, C., Klaper, R. and Grundl, T. (2015) Evaluating the degradation, sorption, and negative mass balances of pharmaceuticals and personal care products during wastewater treatment. Chemosphere 134, 395-401.

Blobaum, A.L. and Marnett, L.J. (2007) Structural and Functional Basis of Cyclooxygenase Inhibition. Journal of Medicinal Chemistry 50(7), 1425-1441.

Bouatra, S., Aziat, F., Mandal, R., Guo, A.C., Wilson, M.R., Knox, C., Bjorndahl, T.C., Krishnamurthy, R., Saleem, F., Liu, P., Dame, Z.T., Poelzer, J., Huynh, J., Yallou, F.S., Psychogios, N., Dong, E., Bogumil, R., Roehring, C. and Wishart, D.S. (2013) The Human Urine Metabolome. PLoS ONE 8(9), e73076.

Boyer, T.H., Taylor, K., Reed, A. and Smith, D. (2014) Ion-exchange softening of human urine to control precipitation. Environmental Progress & Sustainable Energy 33(2), 564-571.

Buchanan, J.R., Deal, N.E., Lindbo, D.L., Hanson, A.T., Gustafson, D. and Miles, R.J. (n.d.) Cost of Individual and Small Community Wastewater Management Systems: Wastewater Planning Model, version 1.0. Foundation, W.E.R. (ed).

Bui, T.X. and Choi, H. (2009) Adsorptive removal of selected pharmaceuticals by mesoporous silica SBA-15. Journal of Hazardous Materials 168(2–3), 602-608.

Carballa, M., Fink, G., Omil, F., Lema, J.M. and Ternes, T. (2008) Determination of the solid–water distribution coefficient (Kd) for pharmaceuticals, estrogens and musk fragrances in digested sludge. Water Research 42(1–2), 287-295.

Cardillo, G. (2012) Four parameters logistic regression - There and back again.

Caro, E., Marcé, R.M., Cormack, P.A.G., Sherrington, D.C. and Borrull, F. (2004) A new molecularly imprinted polymer for the selective extraction of naproxen from urine samples by solid-phase extraction. Journal of Chromatography B 813(1–2), 137-143.

Carollo Engineers (2012), Appedix P - Ozone Sizing and Cost Estimate for TORC Removal. http://www.cityofpaloalto.org/civicax/filebank/documents/32098 (accessed 01/2016)

Cha, Y.I., Kim, S.-H., Solnica-Krezel, L. and DuBois, R.N. (2005) Cyclooxygenase-1 signaling is required for vascular tube formation during development. Developmental Biology 282(1), 274-283.

ChemAxon (2013) Marvin, ChemAxon, Ltd.

220

ChemAxon (2016), chemicalize.org. http://www.chemicalize.org/ (accessed 01/2014)

Chen, M., Cooper, V.I., Deng, J., Amatya, P.L., Ambrus, D., Dong, S., Stalker, N., Nadeau-Bonilla, C. and Patel, J. (2015) Occurrence of Pharmaceuticals in Calgary's Wastewater and Related Surface Water. Water Environ Res 87(5), 414-424.

Choe, J.K., Mehnert, M.H., Guest, J.S., Strathmann, T.J. and Werth, C.J. (2013) Comparative Assessment of the Environmental Sustainability of Existing and Emerging Perchlorate Treatment Technologies for Drinking Water. Environmental Science & Technology 47(9), 4644-4652.

Chu, K.H. (2004) Improved fixed bed models for metal biosorption. Chemical Engineering Journal 97(2–3), 233-239.

Chu, K.H. (2010) Fixed bed sorption: Setting the record straight on the Bohart–Adams and Thomas models. Journal of Hazardous Materials 177(1–3), 1006-1012.

Clara, M., Kreuzinger, N., Strenn, B., Gans, O. and Kroiss, H. (2005) The solids retention time—a suitable design parameter to evaluate the capacity of wastewater treatment plants to remove micropollutants. Water Research 39(1), 97-106.

CompuDrug (2006) PALLAS, CompuDrug Chemistry Ltd.

Cook, S.M., VanDuinen, B.J., Love, N.G. and Skerlos, S.J. (2012) Life Cycle Comparison of Environmental Emissions from Three Disposal Options for Unused Pharmaceuticals. Environmental Science & Technology 46(10), 5535-5541.

Corcoran, J., Winter, M.J. and Tyler, C.R. (2010) Pharmaceuticals in the aquatic environment: A critical review of the evidence for health effects in fish. Critical Reviews in Toxicology 40(4), 287-304.

Cordell, D., Drangert, J.-O. and White, S. (2009) The story of phosphorus: Global food security and food for thought. Global Environmental Change 19(2), 292-305.

Crittenden, J.C., Hand, D.W., Arora, H. and Lykins, B.W. (1987) Design Considerations for GAC Treatment of Organic Chemicals. Journal (American Water Works Association) 79(1), 74-82.

Crittenden, J.C., Trussell, R.R., Hand, D.W., Howe, K.J. and Tchobanoglous, G. (2012) MWH's Water Treatment: Principles and Design.

Cryer, B. and Feldman, M. (1998) Cyclooxygenase-1 and cyclooxygenase-2 selectivity of widely used nonsteroidal anti-inflammatory drugs. Am J Med 104(5), 413-421.

221

Dai, C.-M., Zhang, J., Zhang, Y.-L., Zhou, X.-F., Duan, Y.-P. and Liu, S.-G. (2012) Selective removal of acidic pharmaceuticals from contaminated lake water using multi-templates molecularly imprinted polymer. Chemical Engineering Journal 211-212, 302-309.

Daughton, C.G. and Ternes, T.A. (1999) Pharmaceuticals and personal care products in the environment: agents of subtle change? Environ Health Perspect 107(Suppl 6), 907-938.

Davies, N.M. and Anderson, K.E. (1997) Clinical Pharmacokinetics of Naproxen. Clin Pharmacokinet 32(4), 268-293.

Delle Site, A. (2001) Factors Affecting Sorption of Organic Compounds in Natural Sorbent/Water Systems and Sorption Coefficients for Selected Pollutants. A Review. Journal of Physical and Chemical Reference Data 30(1), 187-439.

Desmidt, E., Ghyselbrecht, K., Zhang, Y., Pinoy, L., Van der Bruggen, B., Verstraete, W., Rabaey, K. and Meesschaert, B. (2015) Global Phosphorus Scarcity and Full-Scale P-Recovery Techniques: A Review. Critical Reviews in Environmental Science and Technology 45(4), 336-384.

Dodd, M., Zuleeg, S., Von Gunten, U. and Pronk, W. (2008) Ozonation of Source-Separated Urine for Resource Recovery and Waste Minimization: Process Modeling, Reaction Chemistry, and Operational Considerations. Environmental Science & Technology 42(24), 9329-9337.

Dominguez, J., Gonzalez, T., Palo, P. and Cuerda-Correa, E. (2011) Removal of common pharmaceuticals present in surface waters by Amberlite XAD-7 acrylic-ester-resin: Influence of pH and presence of other drugs. Desalination 269(1-3), 231-238.

Ebele, A.J., Abou-Elwafa Abdallah, M. and Harrad, S. (2016) Pharmaceuticals and personal care products (PPCPs) in the freshwater aquatic environment. Emerging Contaminants.

Ecoinvent Centre (2015), http://www.ecoinvent.org/database/database.html (accessed 01/2016)

Edzwald, J.K. (2011) Water quality & treatment: a handbook on drinking water, McGraw-Hill New York.

Escher, B.I., Allinson, M., Altenburger, R., Bain, P.A., Balaguer, P., Busch, W., Crago, J., Denslow, N.D., Dopp, E., Hilscherova, K., Humpage, A.R., Kumar, A., Grimaldi, M., Jayasinghe, B.S., Jarosova, B., Jia, A., Makarov, S., Maruya, K.A., Medvedev, A., Mehinto, A.C., Mendez, J.E., Poulsen, A., Prochazka, E., Richard, J., Schifferli, A., Schlenk, D., Scholz, S., Shiraishi, F., Snyder, S., Su, G., Tang, J.Y.M., Burg, B.v.d., Linden, S.C.v.d., Werner, I., Westerheide, S.D., Wong, C.K.C., Yang, M., Yeung, B.H.Y., Zhang, X. and Leusch, F.D.L. (2013)

222

Benchmarking Organic Micropollutants in Wastewater, Recycled Water and Drinking Water with In Vitro Bioassays. Environmental Science & Technology 48(3), 1940-1956.

Escher, B.I., Pronk, W., Suter, M.J.F. and Maurer, M. (2006) Monitoring the Removal Efficiency of Pharmaceuticals and Hormones in Different Treatment Processes of Source-Separated Urine with Bioassays. Environmental Science & Technology 40(16), 5095-5101.

Fabbri, R., Montagna, M., Balbi, T., Raffo, E., Palumbo, F. and Canesi, L. (2014) Adaptation of the bivalve embryotoxicity assay for the high throughput screening of emerging contaminants in Mytilus galloprovincialis. Marine Environmental Research 99, 1-8.

FDEP (2015a) 2014 Reuse Inventory: Appendix D - Reclaimed Water Utilization.

FDEP (2015b), Numeric Nutrient Standards for Florida Waters. http://www.dep.state.fl.us/water/wqssp/nutrients/ (accessed 02/2017)

Fedors, R.F. (1974) A method for estimating both the solubility parameters and molar volumes of liquids. Polymer Engineering & Science 14(2), 147-154.

Fernandez-Fontaina, E., Omil, F., Lema, J.M. and Carballa, M. (2012) Influence of nitrifying conditions on the biodegradation and sorption of emerging micropollutants. Water Research 46(16), 5434-5444.

Ferrando-Climent, L., Collado, N., Buttiglieri, G., Gros, M., Rodriguez-Roda, I., Rodriguez-Mozaz, S. and Barceló, D. (2012) Comprehensive study of ibuprofen and its metabolites in activated sludge batch experiments and aquatic environment. Science of The Total Environment 438, 404-413.

Finkbeiner, M., Ackermann, R., Bach, V., Berger, M., Brankatschk, G., Chang, Y.-J., Grinberg, M., Lehmann, A., Martínez-Blanco, J., Minkov, N., Neugebauer, S., Scheumann, R., Schneider, L. and Wolf, K. (2014) Background and Future Prospects in Life Cycle Assessment. Klöpffer, W. (ed), pp. 207-258, Springer Netherlands, Dordrecht.

FitzGerald, M.P., Stablein, U. and Brubaker, L. (2002) Urinary habits among asymptomatic women. American Journal of Obstetrics and Gynecology 187(5), 1384-1388.

Foo, K.Y. and Hameed, B.H. (2010) Insights into the modeling of adsorption isotherm systems. Chemical Engineering Journal 156(1), 2-10.

Forrest, A., Fattah, K., Mavinic, D. and Koch, F. (2008) Optimizing Struvite Production for Phosphate Recovery in WWTP. Journal of Environmental Engineering 134(5), 395-402.

223

Fresh Water Systems (2016), Composite Well Tanks. https://www.freshwatersystems.com/c-35-composite-well-tanks.aspx (accessed 01/2016)

Fungaro, D.A., Bruno, M. and Grosche, L.C. (2009) Adsorption and kinetic studies of methylene blue on zeolite synthesized from fly ash. Desalination and Water Treatment 2(1-3), 231-239.

Furlong, E.T., Batt, A.L., Glassmeyer, S.T., Noriega, M.C., Kolpin, D.W., Mash, H. and Schenck, K.M. (2017) Nationwide reconnaissance of contaminants of emerging concern in source and treated drinking waters of the United States: Pharmaceuticals. Science of The Total Environment 579, 1629-1642.

Gawlik, B.M., Sotiriou, N., Feicht, E.A., Schulte-Hostede, S. and Kettrup, A. (1997) Alternatives for the determination of the soil adsorption coefficient, KOC, of non-ionicorganic compounds — a review. Chemosphere 34(12), 2525-2551.

Ghodbane, I. and Hamdaoui, O. (2008) Removal of mercury(II) from aqueous media using eucalyptus bark: Kinetic and equilibrium studies. Journal of Hazardous Materials 160(2–3), 301-309.

Ghose, A.K. and Crippen, G.M. (1987) Atomic physicochemical parameters for three-dimensional-structure-directed quantitative structure-activity relationships. 2. Modeling dispersive and hydrophobic interactions. Journal of Chemical Information and Computer Sciences 27(1), 21-35.

Gomes, J., Costa, R., Quinta-Ferreira, R.M. and Martins, R.C. (in press) Application of ozonation for pharmaceuticals and personal care products removal from water. Science of The Total Environment.

GRU (2011) Biosolids Management Plan Update.

Guerra, P., Kim, M., Shah, A., Alaee, M. and Smyth, S.A. (2014) Occurrence and fate of antibiotic, analgesic/anti-inflammatory, and antifungal compounds in five wastewater treatment processes. Science of The Total Environment 473–474, 235-243.

Guishard, T. (n.d.), Comparisons of Steel Water Storage Tanks. http://timswatersolutions.com/pdf/2.pdf (accessed 01/2016)

Han, S., Choi, K., Kim, J., Ji, K., Kim, S., Ahn, B., Yun, J., Choi, K., Khim, J.S., Zhang, X. and Giesy, J.P. (2010) Endocrine disruption and consequences of chronic exposure to ibuprofen in Japanese medaka (Oryzias latipes) and freshwater cladocerans Daphnia magna and Moina macrocopa. Aquatic Toxicology 98(3), 256-264.

Hand, D.W., Crittenden, J.C. and Thacker, W.E. (1984a) Simplified Models for Design of Fixed Bed Adsorption Systems. Journal of Environmental Engineering 110(2).

224

Hand, D.W., Crittenden, J.C. and Thacker, W.E. (1984b) Simplified Models for Design of Fixed&#x2010;Bed Adsorption Systems. Journal of Environmental Engineering 110(2).

Hand, V.C. and Williams, G.K. (1987) Structure-activity relationships for sorption of linear alkylbenzenesulfonates. Environmental Science & Technology 21(4), 370-373.

Hanle, L.J. (2004), CO2 emissions profile of the U.S. cement industry. Washington D.C., 2004. https://www3.epa.gov/ttnchie1/conference/ei13/ghg/hanle.pdf (accessed

Hauschild, M., McKone, T., Van de Meent, D., Huijbregts, M., Margni, M., Rosenbaum, R.K., Jolliet, O. and Fantke, P. (2015) USEtox 2.0.

Hellström, D., Johansson, E. and Grennberg, K. (1999) Storage of human urine: acidification as a method to inhibit decomposition of urea. Ecological Engineering 12(3–4), 253-269.

Hernando, M.D., Mezcua, M., Fernández-Alba, A.R. and Barceló, D. (2006) Environmental risk assessment of pharmaceutical residues in wastewater effluents, surface waters and sediments. Talanta 69(2), 334-342.

Hochmuth, G., Fan, J.H., Kruse, J. and Sartain, J. (2013), SL389/SS591: Using Reclaimed Water to Irrigate Turfgrass - Lessons Learned from Research with Nitrogen. University of Florida, https://edis.ifas.ufl.edu/ss591 (accessed

Hollender, J., Zimmermann, S., Koepke, S., Krauss, M., McArdell, C., Ort, C., Singer, H., von Gunten, U. and Siegrist, H. (2009) Elimination of Organic Micropollutants in a Municipal Wastewater Treatment Plant Upgraded with a Full-Scale Post-Ozonation Followed by Sand Filtration. Environmental Science & Technology 43(20), 7862-7869.

Holloway, K. and Green, T. (2003) Drug and Therapeutics Committees - A Practical Guide, World Health Organization.

Houghton, G.W., Dennis, M.J., Rigler, E.D. and Parsons, R.L. (1984) Urinary pharmacokinetics of orally administered Ketoprofen in man. European Journal of Drug Metabolism and Pharmacokinetics 9(3), 201-204.

Howard, G.J. and Webster, T.F. (2009) Generalized concentration addition: A method for examining mixtures containing partial agonists. Journal of Theoretical Biology 259(3), 469-477.

Huber, M.M., Canonica, S., Park, G.-Y. and von Gunten, U. (2003) Oxidation of Pharmaceuticals during Ozonation and Advanced Oxidation Processes. Environmental Science & Technology 37(5), 1016-1024.

225

Huggett, D.B., Foran, C.M., Brooks, B.W., Weston, J., Peterson, B., Marsh, K.E., La Point, T.W. and Schlenk, D. (2003) Comparison of in Vitro and in Vivo Bioassays for Estrogenicity in Effluent from North American Municipal Wastewater Facilities. Toxicological Sciences 72(1), 77-83.

Igos, E., Benetto, E., Venditti, S., Kohler, C. and Cornelissen, A. (2013) Comparative and integrative environmental assessment of advanced wastewater treatment processes based on an average removal of pharmaceuticals. Water Science and Technology 67(2).

Igos, E., Benetto, E., Venditti, S., Kohler, C., Cornelissen, A., Moeller, R. and Biwer, A. (2012) Is it better to remove pharmaceuticals in decentralized or conventional wastewater treatment plants? A life cycle assessment comparison. Science of The Total Environment 438(0), 533-540.

IMS Health (2011), Top 20 Therapeutic Classes, 2011. https://www.imshealth.com/en/about-us/news/top-line-market-data (accessed

Ishii, S.K.L. and Boyer, T.H. (2015) Life cycle comparison of centralized wastewater treatment and urine source separation with struvite precipitation: Focus on urine nutrient management. Water Research 79, 88-103.

Ishii, S.K.L. and Boyer, T.H. (2016) Student support and perceptions of urine source separation in a university community. Water Research 100, 146-156.

Isidori, M., Lavorgna, M., Nardelli, A., Parrella, A., Previtera, L. and Rubino, M. (2005) Ecotoxicity of naproxen and its phototransformation products. Science of The Total Environment 348(1–3), 93-101.

Jafvert, C.T., Westall, J.C., Grieder, E. and Schwarzenbach, R.P. (1990) Distribution of hydrophobic ionogenic organic compounds between octanol and water: organic acids. Environmental Science & Technology 24(12), 1795-1803.

Jelic, A., Gros, M., Ginebreda, A., Cespedes-Sánchez, R., Ventura, F., Petrovic, M. and Barcelo, D. (2011) Occurrence, partition and removal of pharmaceuticals in sewage water and sludge during wastewater treatment. Water Research 45(3), 1165-1176.

Jimenez, J., Bott, C., Love, N. and Bratby, J. (2015) Source Separation of Urine as an Alternative Solution to Nutrient Management in Biological Nutrient Removal Treatment Plants. Water Environ Res 87(12), 2120-2129.

Johnston, A.E. and Richards, I.R. (2003) Effectiveness of different precipitated phosphates as phosphorus sources for plants. Soil Use and Management 19(1), 45-49.

226

Jones, O.A.H., Voulvoulis, N. and Lester, J.N. (2002) Aquatic environmental assessment of the top 25 English prescription pharmaceuticals. Water Research 36(20), 5013-5022.

Joss, A., Keller, E., Alder, A., Gobel, A., McArdell, C., Ternes, T. and Siegrist, H. (2005) Removal of pharmaceuticals and fragrances in biological wastewater treatment. Water Research 39(14), 3139-3152.

Julia, M.G. and Portier, C.J. (2007) The Forest for the Trees: A Systems Approach to Human Health Research. Environ Health Perspect 115(9), 1261-1263.

Kah, M. and Brown, C.D. (2007) Prediction of the adsorption of ionizable pesticides in soils. J Agric Food Chem 55(6), 2312-2322.

Kah, M. and Brown, C.D. (2008) LogD: Lipophilicity for ionisable compounds. Chemosphere 72(10), 1401-1408.

Kemacheevakul, P., Otani, S., Matsuda, T. and Shimizu, Y. (2012) Occurrence of micro-organic pollutants on phosphorus recovery from urine. Water Science and Technology 66(10), 2194-2201.

Khan, U. and Nicell, J.A. (2010) Assessing the risk of exogenously consumed pharmaceuticals in land-applied human urine. Water Science & Technology 62(6).

Kirchmann, H. and Pettersson, S. (1995) Human Urine - Chemical-Composition and Fertilizer Use Efficiency. Fertilizer Research 40(2), 149-154.

Kloten/Opfikon, K. (2005) Geschäftsbericht 2005 Management Review, p. 16.

Kohler (2016), Steward waterless urinal. http://www.us.kohler.com/ (accessed 01/2016)

Krajewska, B. (2009) Ureases I. Functional, catalytic and kinetic properties: A review. Journal of Molecular Catalysis B: Enzymatic 59(1–3), 9-21.

Lam, L., Kurisu, K. and Hanaki, K. (2015) Comparative environmental impacts of source-separation systems for domestic wastewater management in rural China. Journal of Cleaner Production 104, 185-198.

Lamichhane, K. and Babcock, R. (2012) An economic appraisal of using source separation of human urine to contain and treat endocrine disrupters in the USA. Journal of Environmental Monitoring 14(10), 2557-2565.

Lamichhane, K.M. (2013) Environmental consequences of adopting source separated sanitation system: First and third world perspectives. Ph.D., University of Hawai'i at Manoa, Ann Arbor.

227

Landry, K.A. and Boyer, T.H. (2013) Diclofenac removal in urine using strong-base anion exchange polymer resins. Water Research 47(17), 6432-6444.

Landry, K.A. and Boyer, T.H. (2016) Life cycle assessment and costing of urine source separation: Focus on nonsteroidal anti-inflammatory drug removal. Water Research 105, 487-495.

Landry, K.A., Sun, P., Huang, C.-H. and Boyer, T.H. (2015) Ion-exchange selectivity of diclofenac, ibuprofen, ketoprofen, and naproxen in ureolyzed human urine. Water Research 68(0), 510-521.

Larsen, T. and Gujer, W. (1996a) Separate management of anthropogenic nutrient solutions (human urine). Water Science and Technology 34(3-4), 87-94.

Larsen, T.A., Alder, A.C., Eggen, R.I.L., Maurer, M. and Lienert, J. (2009) Source Separation: Will We See a Paradigm Shift in Wastewater Handling? Environmental Science & Technology 43(16), 6121-6125.

Larsen, T.A. and Gujer, W. (1996b) Separate management of anthropogenic nutrient solutions (human urine). Water Science and Technology 34(3–4), 87-94.

Latini, J.M., Mueller, E., Lux, M.M., Fitzgerald, M.P. and Kreder, K.J. (2004) VOIDING FREQUENCY IN A SAMPLE OF ASYMPTOMATIC AMERICAN MEN. The Journal of Urology 172(3), 980-984.

Lee, J., Ji, K., Lim Kho, Y., Kim, P. and Choi, K. (2011) Chronic exposure to diclofenac on two freshwater cladocerans and Japanese medaka. Ecotoxicol Environ Saf 74(5), 1216-1225.

Li, P. and SenGupta, A. (2004) Sorption of hydrophobic ionizable organic compounds (HIOCs) onto polymeric ion exchangers. Reactive & Functional Polymers 60, 27-39.

Li, P. and SenGupta, A.K. (1998) Genesis of Selectivity and Reversibility for Sorption of Synthetic Aromatic Anions onto Polymeric Sorbents. Environmental Science & Technology 32(23), 3756-3766.

Li, W.C. (2014) Occurrence, sources, and fate of pharmaceuticals in aquatic environment and soil. Environmental Pollution 187(0), 193-201.

Lienert, J., Burki, T. and Escher, B. (2007a) Reducing micropollutants with source control: substance flow analysis of 212 pharmaceuticals in faeces and urine. Water Science and Technology 56(5), 87-96.

Lienert, J., Gudel, K. and Escher, B. (2007b) Screening method for ecotoxicological hazard assessment of 42 pharmaceuticals considering human metabolism and excretory routes. Environmental Science & Technology 41(12), 4471-4478.

228

Lienert, J., Haller, M., Berner, A., Stauffacher, M. and Larsen, T.A. (2003) How farmers in Switzerland perceive fertilizers from recycled anthropogenic nutrients (urine). Water Sci Technol 48(1), 47-56.

Lienert, J. and Larsen, T.A. (2010) High Acceptance of Urine Source Separation in Seven European Countries: A Review. Environmental Science & Technology 44(2), 556-566.

Lin, Y.-P., Singer, P.C. and Aiken, G.R. (2005) Inhibition of Calcite Precipitation by Natural Organic Material:  Kinetics, Mechanism, and Thermodynamics. Environmental Science & Technology 39(17), 6420-6428.

Lindqvist, N., Tuhkanen, T. and Kronberg, L. (2005) Occurrence of acidic pharmaceuticals in raw and treated sewages and in receiving waters. Water Research 39(11), 2219-2228.

Liu, J., Mansouri, K., Judson, R.S., Martin, M.T., Hong, H., Chen, M., Xu, X., Thomas, R.S. and Shah, I. (2015) Predicting hepatotoxicity using ToxCast in vitro bioactivity and chemical structure. Chem Res Toxicol 28(4), 738-751.

Lu, M.-C., Chen, Y.Y., Chiou, M.-R., Chen, M.Y. and Fan, H.-J. (in press) Occurrence and treatment efficiency of pharmaceuticals in landfill leachates. Waste Management.

Lugon-Moulin, N., Ryan, L., Donini, P. and Rossi, L. (2006) Cadmium content of phosphate fertilizers used for tobacco production. Agron. Sustain. Dev. 26(3), 151-155.

Magdeburg, A., Stalter, D. and Oehlmann, J. (2012) Whole effluent toxicity assessment at a wastewater treatment plant upgraded with a full-scale post-ozonation using aquatic key species. Chemosphere 88(8), 1008-1014.

Magiera, S., Hejniak, J. and Baranowski, J. (2014) Comparison of different sorbent materials for solid-phase extraction of selected drugs in human urine analyzed by UHPLC–UV. Journal of Chromatography B 958(0), 22-28.

Mahramanlioglu, M., Kizilcikli, I. and Bicer, I.O. (2002) Adsorption of fluoride from aqueous solution by acid treated spent bleaching earth. Journal of Fluorine Chemistry 115(1), 41-47.

Margot, J., Kienle, C., Magnet, A., Weil, M., Rossi, L., de Alencastro, L.F., Abegglen, C., Thonney, D., Chèvre, N., Schärer, M. and Barry, D.A. (2013) Treatment of micropollutants in municipal wastewater: Ozone or powdered activated carbon? Science of The Total Environment 461–462, 480-498.

Martin, M.T., Knudsen, T.B., Reif, D.M., Houck, K.A., Judson, R.S., Kavlock, R.J. and Dix, D.J. (2011) Predictive model of rat reproductive toxicity from ToxCast high throughput screening. Biol Reprod 85(2), 327-339.

229

MathWorks (2013) MATLAB and Statistics Toolbox Release 2013b, The MathWorks, Inc., Natick, Massachusetts.

Maul, G.A., Kim, Y., Amini, A., Zhang, Q. and Boyer, T.H. (2014) Efficiency and life cycle environmental impacts of ion-exchange regeneration using sodium, potassium, chloride, and bicarbonate salts. Chemical Engineering Journal 254(0), 198-209.

Maurer, M., Pronk, W. and Larsen, T.A. (2006) Treatment processes for source-separated urine. Water Research 40(17), 3151-3166.

Maurer, M., Schwegler, P. and Larsen, T.A. (2003) Nutrients in urine: energetic aspects of removal and recovery. Water Science and Technology 48(1), 37-46.

McClellan, K. and Halden, R.U. (2010) Pharmaceuticals and Personal Care Products in Archived U.S. Biosolids from the 2001 EPA National Sewage Sludge Survey. Water Research 44(2), 658-668.

Meloun, M., Bordovska, S. and Galla, L. (2007) The thermodynamic dissociation constants of four non-steroidal anti-inflammatory drugs by the least-squares nonlinear regression of multiwavelength spectrophotometric pH-titration data. Journal of Pharmaceutical and Biomedical Analysis 45(4), 552-564.

Melvin, S.D., Cameron, M.C. and Lanctot, C.M. (2014) Individual and mixture toxicity of pharmaceuticals naproxen, carbamazepine, and sulfamethoxazole to Australian striped marsh frog tadpoles (Limnodynastes peronii). J Toxicol Environ Health A 77(6), 337-345.

Memmert, U., Peither, A., Burri, R., Weber, K., Schmidt, T., Sumpter, J.P. and Hartmann, A. (2013) Diclofenac: New data on chronic toxicity and bioconcentration in fish. Environ Toxicol Chem 32(2), 442-452.

methanex (2015), Methanex posts regional contract methanol prices for North America, Europe, and Asia. https://www.methanex.com/our-business/pricing (accessed 01/2015)

Morais, S.A. (2014) Multimedia fate modelling and impact of pharmaceutical compounds on freshwater ecosystems.

Moser, P., Sallmann, A. and Wiesenberg, I. (1990) Synthesis and quantitative structure-activity relationships of diclofenac analogs. Journal of Medicinal Chemistry 33(9), 2358-2368.

Muñoz, I., José Gómez, M., Molina-Díaz, A., Huijbregts, M.A.J., Fernández-Alba, A.R. and García-Calvo, E. (2008) Ranking potential impacts of priority and emerging pollutants in urban wastewater through life cycle impact assessment. Chemosphere 74(1), 37-44.

230

Muñoz, I., Rodríguez, A., Rosal, R. and Fernández-Alba, A.R. (2009) Life Cycle Assessment of urban wastewater reuse with ozonation as tertiary treatment: A focus on toxicity-related impacts. Science of The Total Environment 407(4), 1245-1256.

National Center for Environmental Economics (2010), Guidelines for preparing economic analyses. U.S. Environmental Protection Agency, https://yosemite.epa.gov/ee/epa/eerm.nsf/vwAN/EE-0568-50.pdf/$file/EE-0568-50.pdf (accessed

Neale, P.A., Munz, N.A., Aїt-Aїssa, S., Altenburger, R., Brion, F., Busch, W., Escher, B.I., Hilscherová, K., Kienle, C., Novák, J., Seiler, T.-B., Shao, Y., Stamm, C. and Hollender, J. (2017) Integrating chemical analysis and bioanalysis to evaluate the contribution of wastewater effluent on the micropollutant burden in small streams. Science of The Total Environment 576, 785-795.

Nemecek, T. and Kägi, T. (2007) Life cycle inventories of agricultural production systems, Data v2.0, Ecoinvent Centre, Zurich and Dubendorf.

Nishi, I., Komuro, T., Kawakami, T. and Onodera, S. (2010) In vitro cyclooxygenase inhibition assay for evaluating ecotoxicity of the surface water and domestic wastewater in the Tone Canal, Japan. Arch Environ Contam Toxicol 58(3), 535-542.

Nordin, A., Nyberg, K. and Vinnerås, B. (2009) Inactivation of Ascaris Eggs in Source-Separated Urine and Feces by Ammonia at Ambient Temperatures. Applied and Environmental Microbiology 75(3), 662-667.

NREL (2012), U.S. Life Cycle Inventory Database. https://www.lcacommons.gov/nrel/search (accessed 01/2015)

Ort, C., Lawrence, M.G., Reungoat, J., Eaglesham, G., Carter, S. and Keller, J. (2010) Determining the fraction of pharmaceutical residues in wastewater originating from a hospital. Water Research 44(2), 605-615.

Ott, L. and Longnecker, M. (2004) A first course in statistical methods, Thomson-Brooks/Cole, Belmont, CA.

Parang, K., Wiebe, L.I., Knaus, E.E., Huang, J.-S., Tyrrell, D.L. and Csizmadia, F. (1997) In vitro antiviral activities of myristic acid analogs against human immunodeficiency and hepatitis B viruses. Antiviral Research 34(3), 75-90.

Petrie, B., Barden, R. and Kasprzyk-Hordern, B. (2015) A review on emerging contaminants in wastewaters and the environment: current knowledge, understudied areas and recommendations for future monitoring. Water Research 72(0), 3-27.

231

Pomati, F., Orlandi, C., Clerici, M., Luciani, F. and Zuccato, E. (2008) Effects and Interactions in an Environmentally Relevant Mixture of Pharmaceuticals. Toxicological Sciences 102(1), 129-137.

PRé Consultants (2014), http://www.pre-sustainability.com/ (accessed 01/2015)

Pronk, W., Biebow, M. and Boller, M. (2006a) Electrodialysis for Recovering Salts from a Urine Solution Containing Micropollutants. Environmental Science & Technology 40(7), 2414-2420.

Pronk, W., Palmquist, H., Biebow, M. and Boller, M. (2006b) Nanofiltration for the separation of pharmaceuticals from nutrients in source-separated urine. Water Research 40(7), 1405-1412.

Quinn, B., Gagne, F. and Blaise, C. (2008) An investigation into the acute and chronic toxicity of eleven pharmaceuticals (and their solvents) found in wastewater effluent on the cnidarian, Hydra attenuata. Sci Total Environ 389(2-3), 306-314.

Radjenović, J., Petrović, M. and Barceló, D. (2009) Fate and distribution of pharmaceuticals in wastewater and sewage sludge of the conventional activated sludge (CAS) and advanced membrane bioreactor (MBR) treatment. Water Research 43(3), 831-841.

Ramos, T.R.P., Gomes, M.I. and Barbosa-Póvoa, A.P. (2014) Planning a sustainable reverse logistics system: Balancing costs with environmental and social concerns. Omega 48, 60-74.

Remy, C. (2010) Life Cycle Assessment of conventional and source-separation systems for urban wastewater management. Ph.D. Dissertation, Technische Universität Berlin.

Reungoat, J., Macova, M., Escher, B.I., Carswell, S., Mueller, J.F. and Keller, J. (2010) Removal of micropollutants and reduction of biological activity in a full scale reclamation plant using ozonation and activated carbon filtration. Water Research 44(2), 625-637.

Rice, J. and Westerhoff, P. (2015) Spatial and Temporal Variation in De Facto Wastewater Reuse in Drinking Water Systems across the U.S.A. Environmental Science & Technology 49(2), 982-989.

Richardson, G. (1999) Bolted steel tanks, HDPE geomembrane linings.

Rivera-Utrilla, J., Sánchez-Polo, M., Ferro-García, M.Á., Prados-Joya, G. and Ocampo-Pérez, R. (2013) Pharmaceuticals as emerging contaminants and their removal from water. A review. Chemosphere 93(7), 1268-1287.

232

Ronteltap, M., Maurer, M. and Gujer, W. (2007) The behaviour of pharmaceuticals and heavy metals during struvite precipitation in urine. Water Research 41(9), 1859-1868.

Rosal, R., Rodríguez, A., Perdigón-Melón, J.A., Petre, A., García-Calvo, E., Gómez, M.J., Agüera, A. and Fernández-Alba, A.R. (2010) Occurrence of emerging pollutants in urban wastewater and their removal through biological treatment followed by ozonation. Water Research 44(2), 578-588.

Rotroff, D.M., Martin, M.T., Dix, D.J., Filer, D.L., Houck, K.A., Knudsen, T.B., Sipes, N.S., Reif, D.M., Xia, M., Huang, R. and Judson, R.S. (2014) Predictive Endocrine Testing in the 21st Century Using in Vitro Assays of Estrogen Receptor Signaling Responses. Environmental Science & Technology 48(15), 8706-8716.

Ryberg, M., Vieira, M.M., Zgola, M., Bare, J. and Rosenbaum, R. (2014) Updated US and Canadian normalization factors for TRACI 2.1. Clean Technologies and Environmental Policy 16(2), 329-339.

Saikia, M.D. and Dutta, N.N. (2008) Adsorption affinity of certain biomolecules onto polymeric resins: Effect of solute chemical nature. Reactive and Functional Polymers 68(1), 33-38.

Salgado, R., Marques, R., Noronha, J., Carvalho, G., Oehmen, A. and Reis, M. (2012) Assessing the removal of pharmaceuticals and personal care products in a full-scale activated sludge plant. Environmental Science and Pollution Research 19(5), 1818-1827.

Sánchez-Polo, M., Rivera-Utrilla, J., Prados-Joya, G., Ferro-García, M.A. and Bautista-Toledo, I. (2008) Removal of pharmaceutical compounds, nitroimidazoles, from waters by using the ozone/carbon system. Water Research 42(15), 4163-4171.

Sangster, J. (2014) LOGKOW: A databank of evaluated octanol-water partition coefficients (Log P), Sangster Research Laboratories, Montreal, Quebec, Canada.

Santos, J.L., Aparicio, I. and Alonso, E. (2007) Occurrence and risk assessment of pharmaceutically active compounds in wastewater treatment plants. A case study: Seville city (Spain). Environment International 33(4), 596-601.

Saude, E. and Sykes, B. (2007) Urine stability for metabolomic studies: effects of preparation and storage. Metabolomics 3(1), 19-27.

Sawchuk, R., Maloney, J., Cartier, L., Rackley, R., Chan, K.H. and Lau, H.L. (1995) Analysis of Diclofenac and Four of Its Metabolites in Human Urine by HPLC. Pharmaceutical Research 12(5), 756-762.

233

Schroeder, A.L., Ankley, G.T., Houck, K.A. and Villeneuve, D.L. (2016) Environmental surveillance and monitoring--The next frontiers for high-throughput toxicology. Environ Toxicol Chem 35(3), 513-525.

Schwarzenbach, R.P., Gschwend, P.M. and Imboden, D.M. (2002) Environmental Organic Chemistry, 2nd Edition.

Scott, D.C., Clymer, J.W. (2002) Estimation of Distribution Coefficients from the Partition Coefficient and pKa. Pharmaceutical Technology 26(11), 30.

Seyler, C., Hellweg, S., Monteil, M. and Hungerbühler, K. (2005) Life Cycle Inventory for Use of Waste Solvent as Fuel Substitute in the Cement Industry - A Multi-Input Allocation Model (11 pp). The International Journal of Life Cycle Assessment 10(2), 120-130.

Sindelar, H.R., Brown, M.T. and Boyer, T.H. (2015) Effects of natural organic matter on calcium and phosphorus co-precipitation. Chemosphere 138, 218-224.

Sipes, N.S., Martin, M.T., Reif, D.M., Kleinstreuer, N.C., Judson, R.S., Singh, A.V., Chandler, K.J., Dix, D.J., Kavlock, R.J. and Knudsen, T.B. (2011) Predictive models of prenatal developmental toxicity from ToxCast high-throughput screening data. Toxicol Sci 124(1), 109-127.

Sipma, J., Osuna, B., Collado, N., Monclús, H., Ferrero, G., Comas, J. and Rodriguez-Roda, I. (2010) Comparison of removal of pharmaceuticals in MBR and activated sludge systems. Desalination 250(2), 653-659.

Smith, V.H., Tilman, G.D. and Nekola, J.C. (1999) Eutrophication: impacts of excess nutrient inputs on freshwater, marine, and terrestrial ecosystems. Environmental Pollution 100(1–3), 179-196.

Snyder, S.A., Gunten, U.V., Amy, G., Debroux, J. and Gerrity, D. (2014) Use of Ozone in Water Reclamation for Contaminant Oxidation, pp. 373-376.

Solanki, A. and Boyer, T.H. (2017) Pharmaceutical removal in synthetic human urine using biochar. Environmental Science: Water Research & Technology.

Sperlich, A., Schimmelpfennig, S., Baumgarten, B., Genz, A., Amy, G., Worch, E. and Jekel, M. (2008) Predicting anion breakthrough in granular ferric hydroxide (GFH) adsorption filters. Water Research 42(8–9), 2073-2082.

Stalter, D., Magdeburg, A., Weil, M., Knacker, T. and Oehlmann, J. (2010) Toxication or detoxication? In vivo toxicity assessment of ozonation as advanced wastewater treatment with the rainbow trout. Water Research 44(2), 439-448.

STI/SPFA (2016), Water Storage Tank Total Cost of Ownership. http://www.steeltank.com/FabricatedSteelProducts/FieldErectedWaterTanks/TotalCostofOwnershipfreeonlinetool/tabid/474/Default.aspx (accessed 01/2015)

234

Stone, E. and Reardon, R. (2011) Technologies to meet numeric nutrient criteria. Florida Water Resources Journal.

Strathmann, T.J. and Jafvert, C.T. (1998) Ion-pair association of substituted phenolates with K+ in octanol. Environmental Toxicology and Chemistry 17(3), 369-376.

Subedi, B. and Loganathan, B. (2016) Persistent Organic Chemicals in the Environment: Status and Trends in the Pacific Basin Countries II Temporal Trends, pp. 181-202, American Chemical Society.

Sugawara, Y., Fujihara, M., Miura, Y., Hayashida, K. and Takahashi, T. (1978) Studies on the fate of naproxen. II. Metabolic fate in various animals and man. Chem Pharm Bull (Tokyo) 26(11), 3312-3321.

Tarpeh, W.A., Udert, K.M. and Nelson, K.L. (2017) Comparing Ion Exchange Adsorbents for Nitrogen Recovery from Source-Separated Urine. Environmental Science & Technology.

Taute, J.J., Sole, K.C. and Hardwick, E. (2013) Zinc removal from a base metal solution by ion exchange: process design to full-scale operation, The Southern African Institute of Mining and Metallurgy, White River, Mpumalanga, South Africa.

Taylor-Smith, A. (2015) Pharmaceutical Compounds in Land-Applied Sludge and Plant Uptake: A Review, North Carolina State University, Raleigh, North Carolina.

Ternes, T.A. (1998) Occurrence of drugs in German sewage treatment plants and rivers. Water Research 32(11), 3245-3260.

Ternes, T.A., Stüber, J., Herrmann, N., McDowell, D., Ried, A., Kampmann, M. and Teiser, B. (2003) Ozonation: a tool for removal of pharmaceuticals, contrast media and musk fragrances from wastewater? Water Research 37(8), 1976-1982.

Townsend, T., Gustitus, S.A., Walsh, A. and Krause, M.J. (2015) Municipal Solid Waste Composition Study at the University of Florida 2014, University of Florida.

TOXNET (2012), Hazardous Substances Data Bank. National Library of Medicine, Bethesda, Maryland. http://toxnet.nlm.nih.gov/cgi-bin/sis/htmlgen?HSDB (accessed July)

TOXNET (2016), Hazardous Substances Data Bank. http://www.ncbi.nlm.nih.gov/pubmed/ (accessed 01/2014)

Tsai, L.L. and Li, S.P. (2004) Sleep patterns in college students: gender and grade differences. Journal of Psychosomatic Research 56(2), 231-237.

U.S. EIA (2015) U.S. Gasoline and Diesel Retail Prices.

235

U.S. EIA (2016), State Electricity Profiles. http://www.eia.gov/electricity/state/ (accessed 01/2016)

U.S. EPA (2012), Industrial & Municipal Waste Disposal Wells (Class I). http://water.epa.gov/type/groundwater/uic/wells_class1.cfm (accessed 01/2016)

U.S. EPA (2014), Tool for the Reduction and Assessment of Chemical and Other Environmental Impacts. http://www.epa.gov/nrmrl/std/traci/traci.html (accessed 01/2015)

U.S. EPA (2015) 21st Century Science Challenges for EPA’s National Water Program: An Update to the National Water Program Research Strategy.

U.S. EPA (2016a) Drinking Water Action Plan.

U.S. EPA (2016b) National Rivers and Stream Assessment 2008-2009: A Collaborative Survey, Washington, D.C.

U.S. EPA (2016c), Toxicity Forecaster (ToxCast). http://epa.gov/ncct/toxcast/ (accessed 01/2017)

U.S. EPA (2016d), WaterSense Labeled Flushing Urinals. https://www3.epa.gov/watersense/pubs/faq_lfu.html (accessed 01/2016)

Udert, K., Larsen, T., Biebow, M. and Gujer, W. (2003a) Urea hydrolysis and precipitation dynamics in a urine-collecting system. Water Research 37(11), 2571-2582.

Udert, K.M., Larsen, T.A. and Gujer, W. (2003b) Estimating the precipitation potential in urine-collecting systems. Water Research 37(11), 2667-2677.

UF (2014a), Approved Calendar: 2014-2015 Academic Year. https://catalog.ufl.edu/ugrad/current/Pages/calendar1415.pdf (accessed 01/2016)

UF (2014b) Human Resources 2013-2014. Florida, U.o. (ed), Integrated Postsecondary Education Data System.

UF (2015) Common Data Set 2014-2015, University of Florida.

Upton, R.A., Buskin, J.N., Williams, R.L., Holford, N.H. and Riegelman, S. (1980) Negligible excretion of unchanged ketoprofen, naproxen, and probenecid in urine. J Pharm Sci 69(11), 1254-1257.

USGS (2015) Mineral Commodity Summaries 2015, pp. 134-135.

Verlicchi, P., Al Aukidy, M. and Zambello, E. (2012) Occurrence of pharmaceutical compounds in urban wastewater: Removal, mass load and environmental risk

236

after a secondary treatment—A review. Science of The Total Environment 429(0), 123-155.

Vinnerås, B. (2001) Feacal separation and urine diversion for nutrient management of household biodegradable waste and wastewater.

Vinnerås, B., Nordin, A., Niwagaba, C. and Nyberg, K. (2008) Inactivation of bacteria and viruses in human urine depending on temperature and dilution rate. Water Research 42(15), 4067-4074.

Walton, B. (2014), Price of Water 2014: Up 6 Percent in 30 Major U.S. Cities; 33 Percent Rise Since 2010 - Circle of Blue. Circle of Blue, http://www.circleofblue.org/2014/world/price-water-2014-6-percent-30-major-u-s-cities-33-percent-rise-since-2010/ (accessed 01/2016)

Warner, T.D., Giuliano, F., Vojnovic, I., Bukasa, A., Mitchell, J.A. and Vane, J.R. (1999) Nonsteroid drug selectivities for cyclo-oxygenase-1 rather than cyclo-oxygenase-2 are associated with human gastrointestinal toxicity: a full in vitro analysis. Proc Natl Acad Sci U S A 96(13), 7563-7568.

Watanabe, H., Tamura, I., Abe, R., Takanobu, H., Nakamura, A., Suzuki, T., Hirose, A., Nishimura, T. and Tatarazako, N. (2016) Chronic toxicity of an environmentally relevant mixture of pharmaceuticals to three aquatic organisms (alga, daphnid, and fish). Environ Toxicol Chem 35(4), 996-1006.

Water Softeners & Filters (2016), Water Softener Tanks. http://www.water-softeners-filters.com/water-softeners/softener-tanks (accessed 01/2016)

Weber, D., Capello, C., Seyler, C., Kohler, A., Hellweg, S. and Hungerbuhler, K. (2006), Ecosolvent 1.0.1. ETH Zurich, http://www.sust-chem.ethz.ch/tools/ecosolvent (accessed 01/2015)

WHOCC (2013), ATC/DDD Index 2015. http://www.whocc.no/atc_ddd_index/ (accessed 01/2015)

Wicke, E. (1939) Empirische und theoretische Untersuchungen der Sorptionsgeschivindigkeit von Gasen an porosen Stoffen I and II. Kolloid-Z. 86, 295.

Wilkinson, J.L., Hooda, P.S., Barker, J., Barton, S. and Swinden, J. (2016) Ecotoxic pharmaceuticals, personal care products, and other emerging contaminants: A review of environmental, receptor-mediated, developmental, and epigenetic toxicity with discussion of proposed toxicity to humans. Critical Reviews in Environmental Science and Technology 46(4), 336-381.

Winker, M., Faika, D., Gulyas, H. and Otterpohl, R. (2008a) A comparison of human pharmaceutical concentrations in raw municipal wastewater and yellowwater. Science of The Total Environment 399(1–3), 96-104.

237

Winker, M., Tettenborn, F., Faika, D., Gulyas, H. and Otterpohl, R. (2008b) Comparison of analytical and theoretical pharmaceutical concentrations in human urine in Germany. Water Research 42(14), 3633-3640.

Wood, A., Blackhurst, M., Hawkins, T., Xue, X., Ashbolt, N. and Garland, J. (2015) Cost-effectiveness of nitrogen mitigation by alternative household wastewater management technologies. Journal of Environmental Management 150, 344-354.

Worch, E. (2012) Adsorption Technology in Water Treatment, Fundamentals, Processes, and Modeling.

WRRF (2015) Framework for Direct Potable Reuse.

Xu, Z., Cai, J.-g. and Pan, B.-c. (2013) Mathematically modeling fixed-bed adsorption in aqueous systems. Journal of Zhejiang University SCIENCE A 14(3), 155-176.

Zhang, H., Shields, A.J., Jadbabaei, N., Nelson, M., Pan, B. and Suri, R.P.S. (2014) Understanding and Modeling Removal of Anionic Organic Contaminants (AOCs) by Anion Exchange Resins. Environmental Science & Technology 48(13), 7494-7502.

Zhang, Q., Crittenden, J., Hristovski, K., Hand, D. and Westerhoff, P. (2009) User-oriented batch reactor solutions to the homogeneous surface diffusion model for different activated carbon dosages. Water Research 43(7), 1859-1866.

Zhang, R., Sun, P., Boyer, T.H., Zhao, L. and Huang, C.-H. (2015) Degradation of Pharmaceuticals and Metabolite in Synthetic Human Urine by UV, UV/H2O2, and UV/PDS. Environmental Science & Technology 49(5), 3056-3066.

Zhang, Y., Geissen, S. and Gal, C. (2008) Carbamazepine and diclofenac: Removal in wastewater treatment plants and occurrence in water bodies. Chemosphere 73(8), 1151-1161.

Zinckgraf, B., Liebnitzky, C., Kruse, J. and Boyer, T.H. (2014) Nutrient recovery potential from stadium wastewater for use as turfgrass fertilizer. University of Florida Journal of Undergraduate Research 15(2), 4.

Zorita, S., Mårtensson, L. and Mathiasson, L. (2009) Occurrence and removal of pharmaceuticals in a municipal sewage treatment system in the south of Sweden. Science of The Total Environment 407(8), 2760-2770.

238

BIOGRAPHICAL SKETCH

Kelly Landry began her academic career as a Gator in 2010 and has since

earned her Bachelor of Science degree (2013), and Doctor of Philosophy degree

(2017), both within the Department of Environmental Engineering Sciences at the

University of Florida. Kelly’s interest for water and wastewater treatment began during

her undergraduate career and has been strengthened through coursework and research

as well as involvement with the American Water Works Association. She looks forward

to a career dedicated to furthering the efforts of the one-water paradigm.