ABSTRACT TAN, ZHUO. Analysis of An

ABSTRACT

TAN, ZHUO. Analysis of An Antimicrobial Dual-metal Implant System Activated by Low Intensity Direct Current. (Under the direction of Dr. Rohan A. Shirwaiker).

Orthopaedic implant infections pose serious threats to patients in terms of morbidity, mortality and medical costs. Because drug resistance can prevent systemic administration of antibiotics from providing effective treatment for the infections, new technologies are oriented towards the increasing needs of non-antibiotics. Silver-based system activated by electric current has raised the interest of many investigators because of its broad-spectrum antimicrobial activity.

However, so far there has been limited research on the parameter characterization of the system in terms of antimicrobial efficacy and biocompatibility. Furthermore, the control mechanism of the system, which is the key issue for clinical applications, needs to be investigated.

The goal of this dissertation was to understand, model and analyze the antimicrobial efficacy and the biocompatibility of the electrically-activated dual-metal implant system (DIS), and to incorporate the closed loop control mechanism into the system. The specific objectives of this research were as follows:

1. Analyzing and modeling the effects of primary system design parameters on the in vitro

antimicrobial efficacy of the DIS in a simulated environment.

2. Evaluating the in vitro cytotoxic effects of the DIS with critical system design based on

results from Objective 1.

3. Developing a closed loop control on the output current of the DIS, and evaluating the

antimicrobial performance of the closed loop controlled DIS.

This research provides a systematic understanding of the functionality of the proposed implant system. The testing models and mathematical analysis developed in this research give a generalized approach of quantitative assessment of antimicrobial efficacy and cytotoxicity of similar systems. In addition, the design of closed loop control mechanism serves as a foundation of future development of antimicrobial and biocompatible implant system. Overall the characterization of the system provides references for future standards and regulations of electrically-activated silver-based antimicrobial prophylaxis.

Analysis of An Antimicrobial Dual-metal Implant System

Activated by Low Intensity Direct Current

by Zhuo Tan

A dissertation submitted to the Graduate Faculty of North Carolina State University in partial fulfillment of the requirements for the degree of Doctor of Philosophy

Industrial Engineering

Raleigh, North Carolina

2015

APPROVED BY:

______Rohan A. Shirwaiker Richard A. Wysk Committee Chair

______Ola L. A. Harrysson Paul E. Orndorff

BIOGRAPHY

Zhuo Tan, George, received his B.S. degree in Statistics from Communication University of

China, Beijing, in 2009. He worked as an Assistant Systems Engineer at China Film Crest

Digit Co.,Ltd., a China-US joint venture in Beijing, from 2009 to 2010. George joined the

Ph.D. program in Industrial & Systems Engineering at North Carolina State University in Fall

2010. During his Ph.D. study, he married his wife Xiaomei and they had their first baby, Eliana.

ACKNOWLEDGEMENTS

I first want to give thanks and praise to my God who has been blessing me with strength, wisdom, health and endurance in Christ through these five years at North Carolina State

University. He has also blessed me by placing many special and important people in my life to guide me along the way. I must thank my beloved wife Xiaomei. Without her love and support,

I could never have completed this research and doctoral program. A special thanks to my parents who always espoused the virtues of education and supported me with all their hearts.

I would like to graciously acknowledge Dr. Rohan Shirwaiker, my committee chair, for his guidance, support, patience and understanding to maintain the high standards of this research, and for his every moment devoted to mentoring me to be an interdisciplinary industrial engineer. It has been a privilege to be his first PhD student. I am thankful for the time and support given to me by Dr. Richard Wysk, who was my advisor in the first year of my graduate study. His insights and mentoring was extremely helpful in structuring and communicating my research ideas. I express my gratitude to Dr. Paul Orndorff for his assistance and guidance. I thoroughly enjoyed my interactions with him in conducting biological experiments. I also appreciate the support and feedback from Dr. Ola Harrysson.

I am truly grateful for having such wonderful colleagues who made my journey to this achievement full of love, inspiration and courage. They are Dr. Edward Havell, Ms. Patty

Spears, Ms. Mitsu Suyemoto and Dr. Meghan Samberg. I cherish every moment that they offered my selfless supports when I felt frustrated in my research. God has showed me His grace and love through these friends.

iii

I give my sincere thanks to Mr. Edward P. Fitts for his generous endowment. I am just one of the many students who have benefited from the Scholarship created by him. I would like to thank Dr. Paul Cohen and Ms. Cecilia Chen for their constant advice, counsel and help throughout my stay at NC State. Last but not least, I appreciate the friendship and assistance my research group gave me during my time here: Ms. Priyanka Sheshadri, Mr. Anirudh

Ganapathy and Mr. Lokesh Karthik Narayanan. My best wishes to them all in their future pursuits.

TABLE OF CONTENTS

LIST OF TABLES ...... viii

LIST OF FIGURES ...... ix

CHAPTER 1 INTRODUCTION ...... 1

1.1 Background ...... 1 1.2 Motivation ...... 6 1.3 Research Objectives ...... 8 1.4 Research Contributions ...... 8 1.5 Dissertation Outline ...... 9 1.6 Chapter Summary ...... 10

CHAPTER 2 LITERATURE REVIEW ...... 11

2.1 Overview of Orthopaedic Implant Infections ...... 11 2.2 Causes of Orthopaedic Implant Infections ...... 13 2.3 Conventional Treatment Options ...... 15 2.4 Silver-based Implants ...... 19 2.5 Potential Side Effects of Silver ...... 22 2.6 Silver Oligodynamic Iontophoresis ...... 25 2.7 Chapter Summary ...... 31

CHAPTER 3 SYSTEM DESIGN AND IN VITRO EMPIRICAL MODELING OF ANTIMICROBIAL EFFICACY ...... 33

3.1 Introduction ...... 33 3.2 System Design ...... 34 3.3 Antimicrobial Efficacy Study-1...... 36 3.3.1 Materials and Methods ...... 37 3.3.2 Results and Discussion ...... 42 3.4 Antimicrobial Efficacy Study-2...... 50 3.4.1 Materials and Methods ...... 51

3.4.2 Results and Discussion ...... 59 3.5 Antimicrobial Efficacy Study-3...... 67 3.5.1 Model Development ...... 68 3.5.2 Data Analysis...... 75 3.5.3 Results and Discussion ...... 76 3.6 Chapter Summary ...... 95

CHAPTER 4 IN VITRO CYTOTOXICITY OF THE SYSTEM ...... 98

4.1 Introduction ...... 98 4.2 Cytotoxicity Study Phase-1 ...... 100 4.2.1 Materials and Methods ...... 100 4.2.2 Results and Discussion ...... 105 4.3 Cytotoxicity Study Phase-2 ...... 109 4.3.1 Materials and Methods ...... 109 4.3.2 Results and Discussion ...... 113 4.4 Cytotoxicity Study Phase-3 ...... 121 4.4.1 Materials and Methods ...... 121 4.4.2 Results and Discussion ...... 121 4.5 Chapter Summary ...... 124

CHAPTER 5 CLOSED-LOOP CONTROL ON OUTPUT CURRENT ...... 126

5.1 Introduction ...... 126 5.2 Materials and Methods ...... 127 5.2.1 Electronic design ...... 127 5.2.2 Electric performance test ...... 130 5.2.2 Antimicrobial test ...... 130 5.2.3 Design of Experiments ...... 132 5.3 Results and Discussion ...... 132 5.4 Chapter Summary ...... 136

CHAPTER 6 CONCLUSIONS AND FUTURE WORK ...... 137

6.1 Introduction ...... 137 6.2 Research Summary ...... 137 6.3 Reserch Contributions ...... 140 6.4 Future Research ...... 141

REFERENCES ...... 144

APPENDICES ...... 158

Appendix A: VB Codes in MS Excel® ...... 159 Appendix B: Table of Abbreviations...... 163

vii

LIST OF TABLES

Table 1.1 Distribution of microorganisms causing orthopaedic implant infections ...... 3

Table 2.1 Characteristics and comparisons between three groups of staphylococcal orthopedic implant infections...... 14

Table 2.2 A summary of recent animal studies focusing on silver-based infection control ... 22

Table 3.1 Levels of current intensity in 2D agar test model ...... 41

Table 3.2 Mean values and standard deviation of IZ area in 2D agar study ...... 45

Table 3.3 Results of verification experiments for the regression model ...... 49

Table 3.4 List of materials for the electrode settings ...... 52

Table 3.5 Experimental design for important factors ...... 57

Table 3.6 AICs for different polynomial models ...... 62

Table 3.7 Longitudinal analysis results for the test of CM ...... 63

Table 3.8 AEP and MAE for different levels of CM ...... 79

Table 3.9 AEP and MAE for different levels of electrode separation distance ...... 82

Table 3.10 AEP and MAE for different levels of anode surface area ...... 85

Table 3.11 AEP and MAE for different levels of current intensity ...... 91

Table 3.12 AEP and MAE for different levels of current frequency ...... 95

Table 4.1 Experimental design of cytotoxicity test on MG-63 cells ...... 113

Table 5.1 Actual output current with different voltages and resistances ...... 133

Table 5.2 Statistical analysis of the antimicrobial efficacy test ...... 134

viii

LIST OF FIGURES

Figure 1.1 Annual growth rate of primary TJA procedures performed in the US (including THA and TKA), 1996 - 2010 ...... 2

Figure 1.2 Basic concept of LIDC activated silver antimicrobial system ...... 5

Figure 2.1 Periosteal elevation and calcification caused by the infection of tibial intramedullary nail detected by calcium binding fluorophores (green light) ...... 12

Figure 2.2 A hip prosthesis coated with antibiotic-loaded bone cement ...... 17

Figure 2.3 The inhibition zone (kill zone) in the P. aeruginosa-inoculated agar (a) a LIDC activated silver hip implant prototype (b) a rat tibia inserted with a LIDC activated silver wire ...... 31

Figure 3.1 Framework of the in vitro antimicrobial efficacy studies ...... 34

Figure 3.2 Key components of a basic DIS ...... 35

Figure 3.3 DIS adaptation for a hip implant ...... 36

Figure 3.4 A schematic of the 2D agar test model setup ...... 39

Figure 3.5 An actual experimental setup of the 2D agar test model ...... 39

Figure 3.6 Sample agar plates from experiments on E. coli ...... 43

Figure 3.7 Sample agar plates from experiments on S. aureus ...... 44

Figure 3.8 Histogram of IZ areas in 2D agar study ...... 45

Figure 3.9 ANOVA results of 2D agar test...... 46

Figure 3.10 Results of regression model selection for 2D agar test ...... 47

Figure 3.11 Surface approximation graph for 2D agar test ...... 48

Figure 3.12 A schematic diagram of the DIS prototype for the 3D broth test model...... 52

Figure 3.13 The schematic of the 3D broth test setup ...... 54

Figure 3.14 Teensy® USB development board 3.0 for 3D broth test model ...... 54 ix

Figure 3.15 Time-kill curves from the experiments to test the effects of cathode material ... 60

Figure 3.16 Time-kill curves from the experiments to test the effects of electrode separation distance ...... 60

Figure 3.17 Time-kill curves from the experiments to test the effects of anode surface area 61

Figure 3.18 Time-kill curves from the experiments to test the effects of current frequency . 61

Figure 3.19 Time-kill curves from the experiments to test the effects of current intensity .... 62

Figure 3.20 Illustration of the modified PK/PD model...... 69

Figure 3.21 Natural growth of S. aureus for 192 hr...... 70

Figure 3.22 Natural growth of S. aureus for 6 hr...... 71

Figure 3.23 Algorithms for determining (a) the optimal AEP and (b) the MAE...... 76

Figure 3.24 Plots of the differential models (a) with same x0 but different k0 and δ (b) with same k0 and δ but different x0...... 77

Figure 3.25 (a) Empirical bacterial concentration data and b) simulated time-kill curves for cathode material test ...... 79

Figure 3.26 Empirical bacterial concentration data and b) simulated time-kill curves for electrode separation distance test ...... 81

Figure 3.27 (a) Empirical bacterial concentration data and b) simulated time-kill curves for anode surface area test ...... 84

Figure 3.28 Surface oxidation of silver wires after 12 hr ...... 87

Figure 3.29 A silver oxide layer sample observed by Hirox® KH-7700 digital microscope with magnification of 3500 ...... 88

Figure 3.30 (a) Empirical bacterial concentration data and b) simulated time-kill curves for current intensity ...... 90

Figure 3.31 The relationships between the current intensity and the AEPs ...... 92

Figure 3.32 Simulated time-kill curves for current intensity of 5 µA and 6 µA ...... 92 x

Figure 3.33 (a) Empirical bacterial concentration data and b) simulated time-kill curves for current frequency ...... 94

Figure 4.1 Framework of the in vitro cytotoxicity studies...... 100

Figure 4.2 The schematic of the cytotoxicity test model ...... 103

Figure 4.3 The actual experimental setup of the cytotoxicity test model ...... 103

Figure 4.4 Means of viable Mode-K cell number ...... 106

Figure 4.5 Normality test results of cytotoxicity test on Mode-K cells ...... 106

Figure 4.6 Normal probability plot for residuals, cytotoxicity test on Mode-K cells ...... 107

Figure 4.7 t-test results of preliminary cytotoxicity test on Mode-K cells ...... 107

Figure 4.8 Flow chart of preliminary tests for experimental design of Phase-2 Study ...... 112

Figure 4.9 Cell viability results of preliminary Phase-2 study a) with current of 14 µA, and b) with current of 3 µA...... 114

Figure 4.10 The t-test results for preliminary Phase-2 study ...... 115

Figure 4.11 Sample plates of MG-63 cells after 48 treatment in phase-2 study (a) no treatment, (b) titanium with 3 µA, (c) titanium with 6 µA, (d) passive DIS, (e) DIS with 3 µA, and (f) DIS with 6 µA...... 116

Figure 4.12 MG-63 cells after 48 treatment (a) at the periphery of a passive DIS, (b) 1 cm away from a passive DIS, (c) at the periphery of a DIS with 3 µA, (d) 1 cm away from a DIS with 3 µA, (e) at the periphery of a DIS with 6 µA, and (f) 1 cm away from a DIS with 6 µA...... 117

Figure 4.13 Number of MG-63 cells after 48-hr treatment a). original data, and b). data in logarithmic scale...... 118

Figure 4.14 Results of the ANOVA for Phase-2 study ...... 119

Figure 4.15 Number of NHOst cells after 48-hr treatment a). original data, and b). data in logarithmic scale...... 122

Figure 4.16 Results of the T-test for phase-3 study ...... 123

Figure 5.1 LM334 adjustable current sources...... 127

Figure 5.2 Application of LM334 as a basic 2-terminal current source ...... 128

Figure 5.3 Performance characteristics of LM334, Ratio of ISET to IBIAS ...... 129

Figure 5.4 Circuit configuration with incorporated LM 334 ...... 130

Figure 5.5 The actual setup of the antimicrobial efficacy experiment of the closed loop controlled DIS ...... 131

Figure 5.6 Time-kill curves of the closed-loop controlled DIS ...... 135

Figure 6.1 MG-63 cells growing into the inhibition zone 96 hours after removing the DIS and refreshing the medium...... 144

xii

CHAPTER 1 INTRODUCTION

1.1 Background

Total joint arthroplasty (TJA), or total joint replacement as it is also known, is a common procedure that is used to treat debilitating conditions of the joints in both human and veterinary medicine [1]. It entails the use of orthopaedic implants to replace or support missing or damaged bones or bone joints [2]. Nowadays, a wide variety of orthopaedic implants are available to replace or support various bones and joints in the body. According to 2010 Centers for Disease Control and Prevention statistics, there were approximately 332,000 total hip arthroplasties (THA) and 719,000 total knee arthroplasties (TKA) in the US alone [3]. As shown in Figure 1.1, the number of TJA in 2010 has increased by 200% since 1996 and is projected to be 4.05 million in 2030 [4].

One of the most threatening complications associated with orthopaedic implantation is infection. Orthopaedic implant infections can be particularly devastating and are often accompanied by high risk and significant treatment costs. The infection rate is 1-2.5 % for primary TJA and 2.1-5.8 % for revision surgeries [5]. For patients who receive cancer or nutritional therapy, the risk of the implant infections can be as high as 60% [6]. Given the large number of patients who undergo TJA every year, these relatively low infection rates still translate to a substantial number of actual infection incidences. Treatment for the infection is usually accompanied by pain and discomfort. It may require surgical procedures including implant removal, debridement of infected tissue, implant replacement, and 6-12 weeks of antimicrobial therapy [7]. In addition, these procedures are very expensive. Revision TKA 1

caused by infection have an average cost of $109,805, as opposed to the $55,911 for revision

TKA performed for other causes [8]. The annual cost of mitigating infected fracture ﬁxation implants was estimated to be $1.5 billion, in 2004 [9]. Infection is therefore among the most critical risk factors associated with orthopaedic implant applications.

3 TJA

2.5 THA

TKA 2

1.5

% % Increase rate in relative 1996to 0.5

0 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010

Figure 1.1 Annual growth rate of primary TJA procedures performed in the US (including THA and TKA), 1996 - 2010 [4].

The pathogenesis of implant infection involves interactions between the microorganisms

(primarily bacteria), the implant and the host [10]. A list of common pathogens causing the infections is shown in Table 1.1. Staphylococcus spp. and coagulase-negative Staphylococci spp. alone count for 50-60% of TJA infections. The source of infections is usually complicated.

It can be exogenous such as inappropriate surgical operation or aseptic technique, or endogenous, especially in cases of late infection of TJA, because bacteria may spread via the

bloodstream from a distant focus [11]. Thus, standard sterile surgery procedures alone cannot guarantee the absence of late infection [12].

Table 0.1 Distribution of microorganisms causing orthopaedic implant infections [10]

Pathogen Percent of infections Staphylococcus aureus 33-43% Coagulase-negative Staphylococci 17-21% Streptococci 11-12% Gram-negative bacilli 5-14% Enterococci 3-7% Anaerobes 2-5% Polymicrobial 5-13% Other 5-6%

The increasing demand for treatment against implant-associated infection is leading to the development of new clinical approaches. Studies have shown that local antibiotic delivery methods are more effective than pre-operative systemic antibiotic administration for inhibiting bacteria growth at wound site in short post-operative periods [13]. This is because the local antibiotic concentration within the wound cavity can be maintained at a high level while the systemic levels remain relatively low. One method to realize the local administration is creating orthopaedic implants with inherently antimicrobial materials. One of the most common clinical practices in the prevention and treatment of implant-associated infections is the use of antibiotic loaded bone cements (ALBC), which are supposed to attain high local drug levels while maintaining low systemic levels [14]. Commercially prepared ALBC has been used in the US since 2003, but its routine application in infection prophylaxis is still

controversial due to its inconsistent antimicrobial efficacy [15]. Furthermore, the overuse of antibiotics has resulted in the emergence of resistant species of pathogens, such as methicillin- resistant S. aureus (MRSA), which are extremely difficult to treat [9]. Therefore, there is an emerging need for non-antibiotic technologies to prevent or to treat implant infections.

Among many choices of alternative materials for infection treatment or prevention, silver captures much attention due to its broad-spectrum antimicrobial activity and lower propensity to induce bacterial resistance than antibiotics [16]. Silver and its compounds have been used in medicine for decades as components of wound dressings, external skin treatments, debridement agents and eye medicines [17]. The use of silver in coatings currently spans from central venous catheters to urinary tract catheters and coated orthopaedic implants [18] [19].

While the exact mechanism by which silver demonstrates its efficacy is still under investigation, it is widely accepted that the antimicrobial effect is due to its ionic form [20]. In most existing silver-based antimicrobial products, the silver ions are passively diffused from the incorporated surface layer [21]. It has been found through in vivo animal models and clinical experiences that medical prostheses coated with silver alone do not necessarily demonstrate a discernible antimicrobial effect [22]. Thus the slow silver ion release rate becomes the major bottleneck of antimicrobial efficacy for most existing silver medical implants.

Indicated by its relatively low positive standard aqueous electrode potential, silver, like many other heavy metals, does not readily ionize unless an electrical potential is applied. The release of antimicrobial silver ions can be accelerated via low intensity direct current (LIDC). The

basic concept is described in Figure 1.2. The silver, which is connected to a DC power source, provides antimicrobial activities surrounding the device by releasing silver ions. This process is also known as oligodynamic iontophoresis [23]. The effectiveness of this basic concept has been tested in both in vitro [24] [25] and in vivo studies [22] [26]. With significant antimicrobial efficacy against almost all common harmful bacteria (including MRSA) and fungi [27] [28] [29], Electrically-activated silver device has thus raised great interest both in theory and experiment.

Figure 1.2 Basic concept of LIDC activated silver antimicrobial system

When this dissertation research was initiated, all known previous studies with such systems had always focused on a configuration with silver at both electrodes. Based on the working principle of the system – oligodynamic iontophoresis – the release of the antimicrobial silver ion (Ag+) only occurs at the anode. Our research group first investigated the effects of alternative cathode materials on the in vitro antimicrobial efficacy of the system [30]. The system with silver cathode was compared to those with titanium cathode and stainless steel cathode. Based on a modified Kirby-Bauer diffusion test with a broad spectrum of pathogens

(methicillin-resistant S. aureus, Escherichia coli, Streptococcus agalactiae and Aspergillus flavus), the results showed the lack of statistical difference between efficacies of the three

cathode material configurations against all pathogens tested. The study demonstrated that the silver cathode in the LIDC activated system can be substituted with another conductive cathode material without compromising on the system’s antimicrobial efficacy. In order to develop the electrically-activated system for orthopaedic applications, this dissertation focuses on a dual- metal configuration that has evolved from this basic concept, and the motivation for this research is discussed below.

1.2 Motivation

The electrically-activated silver system possesses more intensive antimicrobial efficacy than traditional silver-coated products, but there are potential health safety concerns when it comes to clinical applications. One of the major issues for an orthopaedic implant is its biocompatibility, which refers to the ability of the implant to perform with an appropriate host response [31]. Although silver-coated devices show little genotoxic or cytotoxic effects in many studies [32] [33] [34], little evidence has been found to indicate a good osteoconduction of silver, meaning that bone can grow on its surface [35]. In order to allow the clinical applications of the electrically-activated system, there is a necessity to improve the biocompatibility of the system.

One strategy to improve the biocompatibility is to minimize the quantity of silver in the system by partially substituting it with a biocompatible material like titanium without compromising on the antimicrobial efficacy. Titanium and its alloys are widely used in current orthopaedic implants due to their favorable mechanical properties and biocompatibility. In vitro and in vivo studies have shown that surface treated titanium surfaces promote osseointegration by

stimulating bone formation at the cellular and molecular levels [36] [37]. Furthermore, this metal possesses good corrosion resistance and compressive yield strength which are critical to implants [38] [39].

The motivation of this dissertation is that so far there has been limited research on the parameter characterization of the electrically-activated silver system in terms of antimicrobial efficacy and cytotoxicity. Although several research groups have investigated the effects of some critical design parameters on the antimicrobial efficacy, very few have established truly quantitative models. It is widely known that inappropriate use of silver, in vivo, may lead to a myriad of health and safety problems such as argyria, cytotoxicity or even liver and kidney damage [40]. An overdose of silver will also reduce the amount of probiotics in the body, potentially compromising the mammalian immune system [41]. In order to optimize the efficacy of the proposed system by maintaining a systemically safe level of local silver concentration, reliable quantitative models must first be established to describe the effects of the critical system design parameters on the antimicrobial efficacy as well as the cytotoxicity.

From an application perspective, it is also of great importance to incorporate a control mechanism in the electrical-activation process to guarantee the consistency of the actual efficacy with theoretical approximations.

This dissertation focuses on the characterization of an electrically-activated dual-metal implant system (DIS). The DIS uses titanium as cathode to increase the biocompatibility of the system without compromising the antimicrobial efficacy. As a part of the research, a closed loop control is also incorporated in this system for improving the consistency of silver ion release

from the system independently of the operating environment variations. The research objectives and scope are discussed in the following section.

1.3 Research Objectives

The broader objective of this research was to understand, model and analyze the antimicrobial efficacy and biocompatibility of the electrically-activated DIS and to improve its reliability.

The specific objectives of this research were as follows:

1. Developing quantitative testing models for analyzing in vitro antimicrobial efficacy and

modeling the effects of primary system design parameters on the antimicrobial efficacy of

the DIS in a nutrient-rich fluid environment.

2. Establishing in vitro cytotoxicity testing models and analyzing the cytotoxic effects of the

DIS to mammalian cells with critical system design based on results from Objective 1.

3. Incorporating a closed loop control mechanism on the output current of the DIS to improve

its reliability and evaluating the in vitro antimicrobial performance of the closed loop

controlled DIS.

1.4 Research Contributions

This dissertation contributes to the development of technologies utilizing antimicrobial properties of silver ions. It provides a systematic understanding of the design and functionality of the proposed electrically-activated DIS. The quantitative modeling of the antimicrobial efficacy of the system aids in enhancing the predictability of the system performance with different design parameters. The cytotoxicity analysis of the system extends the knowledge of

the biological properties of the LIDC-activated silver devices. It also provides operating guidelines of DIS application which can potentially serve as a reference for future standards and regulations. The closed loop control design of DIS is expected to enhance the stability of the existing system and can be applied to analogous systems utilizing other LIDC-activated antimicrobial agents. In addition, the research methodologies and in vitro test models developed in this dissertation can be extended to investigations of antimicrobial efficacy and cytotoxicity for other medical devices. In a broader sense, the research will provide impetus to future interdisciplinary research to develop novel technologies with practical applications in improving human health.

1.5 Dissertation Outline

In Chapter 2, literature review related to orthopaedic implant infections, conventional prophylaxis options, clinical applications of silver, and silver oligodynamic iontophoresis is presented. Chapter 3 focuses on Objective 1 of this dissertation, evaluating the in vitro antimicrobial efficacy of the DIS through two different testing models and developing a mathematical model for the effects of critical design parameters. Chapter 4 focuses on

Objective 2, investigating the in vitro cytotoxicity of the DIS to human cells. In Chapter 5, a closed loop controlled DIS was developed and analyzed to satisfy Objective 3. The dissertation concludes in Chapter 6 with discussion about the contributions and directions for future research.

1.6 Chapter Summary

Due to the increasing incidences of orthopaedic implant surgeries and the high costs associated with implant infection treatment, development of antimicrobial implant technologies has become critical. The electrically-activated silver antimicrobial system has raised great interest due to its broad-spectrum antimicrobial efficacy. However, the system lacks comprehensive study on the parameter characterization for its antimicrobial performance and cytotoxicity.

This research provides a systematic analysis on the electrically-activated DIS as a foundation for developing a prophylactic biocompatible orthopaedic implant system for clinical applications. The broader objectives include: 1) characterizing the antimicrobial efficacy of the DIS, 2) analyzing the cytotoxicity of the DIS, and 3) developing a closed loop control on the output current of the DIS.

CHAPTER 2 LITERATURE REVIEW

2.1 Overview of Orthopaedic Implant Infections

Orthopedic implants are widely used as a treatment for bone fractures and diseases to repair or replace the function of affected joints, fractured bone segments and impaired limbs. The global market for orthopedic devices is estimated to increase from $21.1 billion in 2007 to $46.5 billion in 2017 [42]. A major complication associated with orthopaedic implants is the risk of infection which is one of the leading causes of arthroplasty failure. Across the board, an average of 5% of all primary internal fixation devices become infected [43]. Considering the large number of patients that receive orthopaedic implants, even a low risk of infection translates to a very high number of absolute incidences, and can lead to devastating consequences in a large population. In fact, during the first two years following TKA interventions, infections have been reported as the second main cause of revisions just after instability [44]. It poses serious threats to patients in terms of morbidity, mortality and medical costs.

In case of implant infections (example Figure 2.1), it is first necessary to eradicate the pathogen through a series of procedures including removing the prosthesis, debriding the joint, implanting antibiotic-coated spacers and administering systemic antibiotics, before the revision arthroplasty can be performed [45]. That results in remarkably high cost for each single episode of treatment, and tremendous suffering for the patient. Revision TKA caused by infection have an average cost of $109,805, which is approximately 5.23 times more compared to a non-infected revision [46]. Furthermore, the infection rate after revision surgery 11

is considerably higher than for the primary arthroplasty (up to 10%) [47]. These infections have a major impact on the psychological morbidity of patients, impairing joint function and quality of life, and sometimes requiring arthrodesis, permanent removal of the prosthesis, or even amputation [48].

Figure 2.1 Periosteal elevation and calcification caused by the infection of tibial intramedullary nail detected by calcium binding fluorophores (green light) [49].

Implant infections can occur by direct contamination during surgery (intraoperative infections), bacteremia secondary to a remote site of infection (hematogenous infections) or by contact with an adjacent site of infection or an open wound (contiguous infection) [50].

Generally, the relative risk of infection is associated with the time for which the device remains within the body. Infections are classified as early if they occur within 3 months of the initial implantation (within 2 weeks for osteosynthesis), and are usually the result of direct exogenous

contamination by a highly virulent microorganism either on the implant or into the surgical site [51]. Under these conditions, pathogen clot formation and dissolution would be detected prior to stem cell commitment and osteoblast colonization of the implant. Subacute infections occur 3 to 24 months postoperatively (2 to 10 weeks for osteosynthesis) and are caused by low- virulence pathogens contaminating the implant during surgery. Early loosening of the prosthetic device is the leading sign. These infections are often labeled as low-grade infections, usually accompanied with mild fever but no pus production [52]. Late infections may arise more than 2 years after surgery (more than 10 weeks for osteosynthesis) and usually result from blood-borne dissemination such as transient bacteremia experienced with urinary tract infections. These infections are very hard to differentiate from mechanical loosening because they progress insidiously and produce very similar symptoms [53] [54].

2.2 Causes of Orthopaedic Implant Infections

The most common causative organisms of orthopaedic implant infections include methicillin- sensitive Staphylococci aureus (MSSA), methicillin-resistant Staphylococci aureus (MRSA),

Coagulase-negative staphylococci (CoNS), Streptococci and Gram-negative bacilli such as

Escherichia coli (E. coli), out of which Staphylococci comprises of up to two-thirds [55]. Table

2.1 summarizes the differences in clinical presentation among the three groups of

Staphylococci spp. for all types of implant infections recorded at Geneva University Hospitals

(Geneva, Switzerland) based on 5374 surgical procedures from January 1996 to June 2008

[18].

Table 0.1 Characteristics and comparisons between three groups of staphylococcal orthopedic implant infections. [56]

All types of implant infections MRSA MSSA CoNS (N = 163) (n = 44) (n = 58) (n = 61) Female sex 23 (52%) 23 (40%) 29 (48%) Median age 74 years 60 years 71 years Chronic immunosuppression 11 (25%) 14 (24%) 17 (28%) Median time delay between 21 days 125 days 129 days previous implantation and infection onset Median duration of antibiotics 10 weeks 6 weeks 9 weeks Removal of infected implant 29 (66%) 46 (79%) 46 (75%) Re-implantation of a new implant 9 (20%) 13 (22%) 27 (44%) Recurrence of infection 5 (11%) 9 (16%) 6 (10%) Cure 25 (57%) 42 (72%) 50 (82%) Median length of hospital stay 57 days 29 days 39 days

The pathogenesis of implant infections is more complicated than that of other surgical site infections due to the presence of the “foreign” implants. Experimental models have shown that the critical dose of contaminating microorganisms required to produce infection is much lower when a foreign material is present at the surgical site [57]. The interstitial milieu surrounding orthopaedic implants is known to represent a region of local immune depression which is susceptible to microbial colonization, and favorable to the initiation of infections [58].

Furthermore, in orthopaedics, the movements of the implants inserted in hard tissues and the detrimental release of wear debris, as in the case of TJA, can damage the tissues surrounding the implant, creating conditions where the immune defenses are mostly depleted [59].

The implant infections are even more difficult to treat when the organisms grow into biofilms and adhere to the implant surface in a highly hydrated extracellular matrix. The microorganisms within biofilms form complex, highly organized and metabolically inactive communities which are extremely resistant to antibiotics [60]. Once implant colonization has occurred, continuing infection is propagated not only through the bacteria disseminated from the biofilm on the implant, but through adherence and colonization of contiguous tissues.

Additionally, pathogen colonization of implant can be enhanced by the host response to implantation [61]. The serum proteins coated on implanted materials by the host to promote cell recruitment and tissue repair, at the same time, are used by pathogens for adhesion and virulence. [62] In other words, continuous seeding of surrounding tissues increases the probability of further infection.

To sum up, orthopaedic implants, like many other foreign materials, create a site of weakness for the immune system where even bacteria with a low level of virulence can easily cause catastrophic consequences. For this reason, the most critical goal for prevention of implant infections is to eradicate pathogens before they can populate the implant/tissue environment.

2.3 Conventional Treatment Options

With significant advancements in medical treatment, especially the use of antibiotics and novel operating room techniques, the infection rate of orthopaedic implantation has decreased markedly in recent decades. However, challenges still remain for better preventive and therapeutic regimens as discussions above. Given that most existing prosthetic bulk materials

(e.g., titanium and stainless steel alloys) already have a track record of appropriate

biomechanical and biocompatible performances, current policy for the prevention of implant infections largely relies on the strategies at the level of biomaterial-tissue interface. It is widely believed that the most efficient measure to interfere with the early phases of microorganism adhesion lies in the improvement of the biomedical properties of the implant surface so that the device can be both cell-friendly and antimicrobial [63] [64].

The most common clinical solution to prevent bacteria adhesion is the use of local antibiotics in the outer layer of implant, which allows for a much higher concentration (up to 1000 fold) than that achievable with systemic antibiotics [65]. The prevailing commercial products based on this technology include gentamicin polymethylmethacrylate (PMMA) beads [66], antibiotic loaded bone cement (ALBC) used in arthroplasty [67], and gentamicin-collagen sponge used in trauma surgery [68]. PMMA was first implemented in orthopedic surgery as a hip implant in 1945 by Scales and Herschell to treat coxarthrosis [69]. In 1970, Buchholz and Engelbrecht reported the incorporation of antibiotics into PMMA bone cement to reduce the infection rate in arthroplasty [70]. The gentamicin sulphate is gradually released by the cement into the surgical site to maintain a high concentration at local level, which could not be reached by venous administration without general complications and toxicity. Since 1990s, ALBC

(Figure 2.2) became widely used in arthroplasty as a treatment of established implant infections. Most commercially available bone cements currently marketed have two components: PMMA and liquid methyl methacrylate (MMA). Apart from these main components, other substances are required to achieve a controlled polymerization at body temperature, such as N,N-dimethyl-p-toluidine and hydroquinone [71]. Besides gentamicin,

other antibiotics have also been used as an additive to bone cement, such as tobramycin, vancomycin [72], clindamycin, fusidic acid [73] and combinations of these antibiotics [74].

Figure 2.2 A hip prosthesis coated with antibiotic-loaded bone cement [75].

In vitro studies indicate that ALBC prevents haematogenous infection within hours after implantation [76] but fails to demonstrate a complete eradication of biomaterial-adherent bacteria [77]. Antibiotics are released from bone cement in a typical bi-phasic fashion. At first there is a peak release followed by a long, tail of low release that continues for days to months

[78]. After months of implantation, the bone cement was found not to release further antibiotics, unless the samples were sawed or broken. Despite the ongoing slow release, there is still almost 80% of the antibiotic locked in the bone cement after many years [79].

The routine clinical use of ALBC, however, still remains controversial. Many in vivo studies support the use of low-dose ALBC as one of the effective means in preventing infection in 17

primary TJAs in both high-risk and non-risk patients. Extensive data from Norwegian and

Swedish hip registries and limited US data indicate that the use of low-dose (less than 2 g of antibiotic per 40 g of cement) ALBC is safe and effective in primary TJAs [75]. However, there are many published variations, and currently no standardized protocol of ALBC application [80]. A statistical analysis of primary TKA from May 2003 to March 2007 was conducted in the US using a community-based total joint registry. Out of 22,889 primary TKA, the rate of infection was significantly higher (1.4%; 28 out of 2030 cases]) in TKA performed with ALBC than with regular cement (0.7%; 154 out of 20,869 cases) (p = .002)

[81]. Differences in outcomes of various prophylactic measures are difficult to determine, especially due to the low infection rates. Generally, the lowest risk of revision for implant infection was found if the antibiotic prophylaxis was given both systemically and locally in the cement [82].

The use of ALBC carries the risk of inducing antibiotic resistance, which can make the management of infected TKA even more difficult. Studies suggest that up to 8% of the antibiotic in ALBC is quickly released after surgery, and the low-dose release thereafter may not be effective in fighting infection; in fact, it may contribute to the problem of antibiotic resistance [83]. Some in vitro studies have observed bacterial growth on ALBC [84] and in the interfacial gap between cement and the bone [85]. Subsequent infections were more likely to be with resistant organisms in cases where ALBC had been used earlier. Hope et al. reported on 34 cases where Gentamicin-loaded cement was used for primary surgery. The biopsy and culture testing from the infection site revealed that 30 of the patients (88%) were infected with at least one gentamicin-resistant strain [86]. Orthopaedic infections caused by antibiotic-

resistant bacteria cause further morbidity in patients and result in additional treatment costs

[87]. Aggressive management of such a resistant infection often includes minimization of patient contact and attempts to resolve the infection/colonization through the use of a greatly limited choice of therapeutic antibiotics [88]. Furthermore, the prolonged low release of antibiotics from ALBC may also bring about side effects, such as damage to kidneys and allergic reactions [89]. The problems of ALBC therefore raise the discussion of non-antibiotic alternatives.

2.4 Silver-based Implants

Silver possess a good inhibitory efficacy against various types of pathogens such as bacteria, viruses, yeasts and fungal species [24]. Since 20th century, silver-containing compounds are widely used in both standard and alternative medical instruments and products, such as urinary catheters [90], drainage catheters [91], wound and burn bandages [92], prosthetic heart valve

[93], sutures [94], and fracture fixation devices [95]. Many of such silver coated medical prostheses have proved to reduce the infection rate and, despite their higher price, may cut overall treatment costs as well [96].

Bone cement loaded with metallic silver nanoparticles has shown good antimicrobial activity against Methicillin-resistant Staphylococcus epidermidis (MRSE) and MRSA, which

Gentamicin cement was not effective against due to the high-level gentamicin resistance [34].

Silver-loaded coatings on titanium surface, titanium oxide surface and hydroxyapatite (HAp) film have demonstrated antimicrobial performance against S. aureus, S. epidermidis, E. coli and Pseudomonas aeruginosa while retaining good biocompatibility at the same time, in vitro

[97] [98] [99] [100] [101] [102]. Silver was also incorporated in polypropylene meshes or even titanium nanotubes for medical applications and demonstrated significant in vivo antimicrobial efficacy and excellent biocompatibility [103] [104] [105]. Gosheger et al. conducted a comprehensive in vivo studies to examine the antimicrobial efficacy and possible side effects of a silver coated implant using a rabbit model [106]. Measurements of the C-reactive-protein

(CRP), neutrophilic leukocytes, rectal temperature and body weight showed significant (p <

0.05) lower signs of inflammation in the rabbits with silver implants compared to the ones with titanium implants. Histological samples were obtained from the brain, heart, thymus, liver, kidney, spleen, ovary and testes. The analysis of silver concentration in blood (median of 1.883 ppb) and the organs showed elevated silver concentrations (0.798 - 86.002 ppb) without pathological changes in leukocytes, neutrophilic granulocytes, lymphocytes and CRP values, nor histological changes in organs.

A German company, implantcast GmbH, has developed a silver coated orthopaedic implant,

MUTARS®Silver (Figure 2.3), which is being marketed as a salvage and prophylaxis against implant infections [107]. The CE certified MUTARS®Silver components are now considered standard implants for the treatment of tumor patients in several major tumor centers throughout

Europe. So far more than 2,000 implantations have been performed successfully. In a clinical study, the infection rate in 51 patients with sarcoma who underwent placement of a silver coated implant was assessed prospectively over a 5-year period compared to 74 patients with uncoated titanium implants [108]. The results showed a substantially lower infection rate of

5.9% in the silver group compared to 17.6% in the titanium control group. 38.5% of patients

in the titanium group ultimately had to undergo amputation. This study demonstrated the effectiveness of silver coated implants to reduce infection rates in the medium term.

Figure 2.3 MUTARS®Silver (Modular Universal Tumour and Revision System coated with

silver) by implantcast GmbH [107]

However, Fordham et al evaluated the in vitro bactericidal performance of silver coating for biomedical implants and suggested that pure silver is an inadequate bactericide for orthopedic implants inserted in the medullary canal but may be useful as a coating to prevent bacterial adhesion or colonization [109]. In their study, the silver films only demonstrated bactericidal properties for bacteria that landed on the surface of the film, but failed to create a zone of inhibition in a nearby area. The antimicrobial efficacy of medical devices with passive silver thin films is still under discussion. Investigating mechanisms to enhance the diffusion of silver

ions will be of more practical significance and clinical value. Some recent animal studies examining silver-based treatments are summarized in Table 2.2.

Table 0.2 A summary of recent animal studies focusing on silver-based infection control

Authors Model Experimental group Results compared to control group(s) Stinner et al Goat Silver dressings + negative  25% decrease in wound [110] pressure wound therapy bacterial load Tran et al [111] Goat Silver-coated intramedullary  Decreased bacterial load nail  Increased mechanical strength of femur after five weeks Secinti et al Rabbit Electrically-activated  Decreased bacterial counts in [112]. silver-coated titanium screws bone surrounding experimental group screws Gosheger et al Rabbit Silver-coated  Significantly lower infection [106] megaendoprosthesis rate in experimental group (7%) compared to control group (47%)

2.5 Potential Side Effects of Silver

A major concern with the usage silver-incorporated medical devices is the potential adverse health effects. When exposed to light in the near-skin regions, silver ions are easily photoreduced to metallic silver and deposited, causing the bluish-gray appearance associated with argyria [113]. In 2010, it was reported that 33 % of patients (4 out of 12) classified as high-risk types with chronic infections and major bone loss developed localized cutaneous argyria after receiving silver coated implants [114]. Inhibition of the proliferation of 22

keratinocytes and fibroblasts after treatment with silver-sulfadiazine has also been reported

[115]. Furthermore, accumulation of silver may have possible negative influence on fatty degeneration of the liver, kidneys and heart [116].

The toxicity of silver is dependent on the dose and the exposure time, and may be related to damage at the cell membrane, oxidative stress, or/and interactions of silver ions with proteins and enzymes. Bosetti et al investigated the in vitro toxicity and biocompatibility of silver coated material for external fixation devices [32]. No genotoxicity nor cytotoxicity to osteoblast cells was detected after a 4-day incubation with passive silver coated stainless steel pins. In addition, cells cultured on the silver-coated material evidenced good cell spreading and a higher cell count with respect to the uncoated material. Some studies, in vitro [34] and in vivo [108] [106], also suggested that silver-coated implant allowed a release of silver without showing any local or systemic side-effects. However, other studies showed that silver was highly toxic to human cells [117] [118] and implantation of silver or silver-plated devices was not recommendable [119]. The decreased viability of cells caused by silver ions was primarily detected as necrosis [120] [121] [122]. Silver also caused a significant reduction of mitochondrial membrane potential and a significant depletion (p < 0.05) of Glutathione levels, a ubiquitous sulfhydryl-containing molecule in cells which was responsible for maintaining cellular oxidation–reduction homeostasis [123]. In addition, AshaRani et al suggested that silver involves in disruption of the mitochondrial respiratory chain leading to production of

ROS and interruption of adenosine triphosphate synthesis, which in turn cause DNA damage.

It was anticipated that DNA damage is augmented by deposition, followed by interactions of silver nanoparticles to the DNA leading to cell cycle arrest in the G(2)/M phase [124].

In the literature, there has no consistent data concerning the necessary dose of silver to cause apparent adverse effects. Some studies suggested that 1 – 8 g of inhaled silver dust may cause an argyria [125]. Other reports indicated that silver levels exceeding 300 ppb in the blood would lead to leukopenia and liver damage [126] [127]. Hussaine et al suggested that the threshold dosage of silver ions which might cause a significant cytotoxicity was from 5 to 50

μg/ml. In an in vitro study with rat liver cells, the level of reactive oxygen species (ROS) in cells increased in a silver concentration-dependent manner and was statistically increased from

10 μg/ml concentration, which resulted in significant damage to cell structures.

Although no regulatory guidance on silver implant has been published so far in the US, the

Environmental Protection Agency (EPA) has published an oral reference dose (RfD) of 5 µg / kg / day for silver oral consumption [128]. This dose is an estimate of the maximum amount of silver a human can be exposed to on a daily basis without experiencing any appreciable adverse medical effects over a lifetime. According to the secondary maximum contaminant levels (SMCL), the “unenforceable” maximum allowable amount of silver for water is 0.1 mg

/ L [129]. The average water consumption for an adult is assumed to be 2 L [130]. With an estimation of 0.25 mg of silver per day from the food and water in total, for a 75 kg adult, the conservative number for safe intake of silver from other sources can be estimated as 0.125 mg

/ day.

The absence of control in the ion diffusion mechanism impedes further development of safe local administration of antimicrobial silver ions. On one side, products with low silver concentration may not demonstrate significant local antimicrobial efficacy. On the other, long exposure periods of implants with high silver concentration will increase the risk of system 24

hazard and induce silver-resistance species. A possible solution to this dilemma is the application of oligodynamic iontophoresis, an active control mechanism which utilizes direct electric current to regulate the ion release.

2.6 Silver Oligodynamic Iontophoresis

The oligodynamic effect was discovered in 1893 by the Swiss, Karl Wilhelm von Nägeli, as a toxic effect of heavy metal ions on living cells even in relatively low concentrations [131].

Subsequent studies suggest that silver ions penetrate the cell wall and subsequently denature enzymes of the target cell by binding to reactive groups, resulting in their precipitation and inactivation. At the same time, DNA molecules of cells also become condensed as a reaction against the denaturation effects of silver ions and thus lose their replication abilities [132]. All these phenomena together lead to the damage or even the death of bacteria cells (Figure 2.4). Hence, any silver-based antimicrobial products have to release the silver ions to a pathogenic environment in order to be effective. Medical devices with passive silver thin films or compounds may not be efficacious until they can effectively produce and diffuse enough silver ions [109]. Hence, multiple techniques including the use of direct current activation are being investigated in order to enhance the antimicrobial efficacy of silver-based medical devices.

Figure 2.4 Internal structure of S. aureus cells exposed to silver. (a) An electron-light region (red arrow) in the center of the cell with condensed and concentrated DNA molecules in its center (white arrow). (b) Cytoplasm membrane detached from the cell wall (white arrow) [132].

To accelerate the oligodynamic process, Spadaro et al. connected silver to the positive electrode (anode) of a power source? and activated the release of silver ions by low intensity direct current (LIDC, 0.02 – 20 µA) [133]. This process is known as oligodynamic iontophoresis [134].A series of in vitro experiments confirmed the high efficacy of this configuration in inhibiting bacterial growth. The inhibitory and antimicrobial concentrations of LIDC generated silver ions were 10 to 100 times lower than those for silver sulfadiazine.

Meanwhile, cell growth tests showed no obvious detrimental effects such as cell aggregation, distortion, lysis, or pH changes as compared with the controls [135]. This finding led to

explorations of silver medical devices based on LIDC activated silver ion action. The ion release process is essentially based on electrolysis, which is the passage of a direct electric current through metallic electrodes in a suitable solvent, resulting in ion exchange between the anode and cathode. According to Faraday’s first law of electrolysis, by controlling the current intensity and duration, the concentration and the release rate of silver ions can be determined

[24]. The silver dosage can be regulated accordingly to be below the health safety upper bound.

The first human clinical study using LIDC activated silver devices as a treatment for orthopaedic infection was reported by Becker et al. in 1978 [136]. In this six-week study, a pure silver wire was inserted directly into infected non-unions in which the wound was to be surgically closed (Figure 2.5). The 8 cm long electrodes were insulated with Teflon except for the terminals. The electrode was first made LIDC positive with a current of 1 µA for twenty- four hours after surgery. Then the current was reduced to 0.1 µA and the electrode polarity was reversed without interruption until 1.5 to 2.5 J of total energy had been delivered. The treatment resulted in control of the infection in 12 of the 15 patients after follow-up ranging from 3 to 36 months. In the other three cases which failed to eradicate the infection, two remarkably reduced bacterial flora in the wound. No detectable toxic effect on local tissues was reported. In fact, deposition of substantial amounts of new bone was reported with the silver-Teflon anodes.

Figure 2.5 Application of silver-Teflon sheet to treat infected tibial non-union. (a) The wound as seen through window in cast. (b) Saline-wetted silver-Teflon sheet placed in contact with all wound surfaces below skin level with a “tail’ left out to establish electrical contact. (c) A gauze stent moistened with normal saline packed over the silver-Teflon. (d) Dry sterile gauze placed over entire wound with tail protruding. (e) Constant voltage source with positive lead connected to the silver-Teflon [136].

In a later clinical study conducted by Webster et al, 25 patients with active, chronic osteomyelitis, resistant to conventional management were treated with surgical debridement and daily application of LIDC activated silver nylon dressings [137]. The dressing was placed in close contact with all exposed bone and soft tissue with the exception of tendons and ligaments. The tail portion of the silver nylon was connected to the positive terminal of a DC source unit. A skin pad (negative terminal) was taped firmly directly opposite to the wound.

During the actual treatment, the driving voltage of the DC source was 0.8 - 0.6 V with a corresponding current between 100 - 250 µA throughout the treatment period. This corresponded to a nominal current density of about 0.7 to 1.6 µA / cm2 in the actual silver nylon dressing. The current at 24 hours was observed to be 10% to 50% lower than the initial value as impedance of the electrode/tissue/electrode circuit increased. The initial current level was restored after fresh electrodes (silver-nylon and skin pad) were replaced, each day. After

4 weeks of treatment, 16 cases (64%) resulted in closed, stable, pain-free wounds. No side effects of anodic silver administration, toxic or otherwise, were noted in this study.

Fuller et al. improved the LIDC activated silver device design and investigated different configurations and operational parameters (Figure 2.6(a)) [24]. The re-engineered system demonstrated significant inhibition efficacy against seven different microbial species – S. aureus, Enterococcus faecalis (E. faecalis), MRSA, E. coli, Pseudomonas aeruginosa (P. aeruginosa), Proteus mirabilis (P. mirabilis) and Candida albicans (C. albicans). Five levels of currents (0.15 µA, 1.5 µA, 15 µA, 20 µA and 1.5 A) and three levels of anode surface area

(varied by changing anode length: 6.75 mm, 12.7 mm and 25.4 mm) were tested. Agar-based semi-quantitative analysis found that the performance of this system was highly dependent on

the current and surface area of the anode. This system design was further developed into an orthopaedic implant prototype and tested in vitro and in vivo [26]. The prototype device was inserted in the medulary cavity of an excised rat tibia embedded in Mueller Hinton agar and tested against P. aeruginosa. An inhibition zone formed around the rat tibia indicated that the silver ions released from the system could penetrate through the bone (Figure 2.6(b)). Further, a total of 37 surgeries were performed by implanting live rats with the LIDC activated silver wires. The results showed a very high confidence level in the effectiveness of LIDC activated silver implant.

To sum up, a series of in vitro and in vivo studies have demonstrated the efficacy of LIDC activated silver implants, and such an approach may have advantages over traditional antibiotics in terms of antimicrobial efficacy. By controlling the current and silver surface area, the amount of silver ions can be set to a level which is effective against pathogens but relatively safe to human cells. Therefore the antimicrobial silver implant system activated by LIDC can be developed not only for prophylaxis, but also as a potentially reliable treatment for orthopaedic implant infection.

Figure 2.6 The inhibition zone (kill zone) in the P. aeruginosa-inoculated agar (a) a LIDC activated silver hip implant prototype [24] (b) a rat tibia inserted with a LIDC activated silver wire [26].

2.7 Chapter Summary

Prevention and treatment of orthopaedic implant infections are becoming increasingly important due to the continued increase in the number of patients receiving such implants. It is widely believed that the most efficient measure to prevent implant infections is through improvements in the biomedical surface properties of the implant devices so that they can be both cell-friendly and antimicrobial. Due to the growing prevalence of antibiotic-resistant

species, silver draws a wide attention as an alternative antimicrobial material for clinical device applications. Silver ions attack bacteria cells in a multimodal manner, making it much more unlikely for bacteria to develop resistant properties. A major concern with the usage silver- incorporated medical devices, however, is their potential adverse health effects. One of the possible remedies to achieve remarkable efficacy under safe silver dosage is by applying an active control mechanism which utilizes electric current to regulate the ion release process. In vitro and in vivo studies have suggested the advantages of electrically-activated silver implant over traditional antibiotics in terms of bacterial resistance and clinical effectiveness. Some preliminary studies have suggested the possibility of sustaining silver ion concentration at levels that are detrimental to bacteria without inducing human cellular toxicity.

CHAPTER 3 SYSTEM DESIGN AND IN VITRO EMPIRICAL MODELING OF

ANTIMICROBIAL EFFICACY

3.1 Introduction

This dissertation focuses on a dual-metal (silver-titanium) implant system, in which a silver electrode serves as the anode and a titanium electrode serves as the cathode. As summarized in Chapter 2, the silver-based system activated by low intensity direct current (LIDC) holds great potential for antimicrobial orthopaedic implant applications. Meanwhile, titanium is widely used in current orthopaedic implants due to its favorable mechanical properties and biocompatibility. The basic hypothesis of this dissertation is that the dual-metal implant system

(DIS) can minimize the potential cytotoxicity and material costs, and increase the biocompatible surface area of the system for potential osseointegration without diminishing its antimicrobial efficacy. Therefore the LIDC system would be biologically more optimum by adopting the dual-metal configuration than the original comprehensive silver-based configuration.

This chapter focuses on Objective 1 of this dissertation: developing quantitative testing models for analyzing in vitro antimicrobial efficacy and modeling the effects of primary system design parameters on the antimicrobial efficacy of the DIS in a nutrient-rich fluid environment. Three studies have been conducted to achieve this objective (Figure 3.1). In Study-1, the effects of electric current intensity and duration were investigated using a 2D agar test model, which was an adaptation of the standard Kirby-Bauer disc diffusion method for evaluating antibiotics

[138]. In Study-2, the effects of several critical design parameters were investigated using a 33

3D broth test model, which simulated potential in vivo fluid environment. In Study-3 a differential model was derived based on classical in vitro Pharmacokinetic/Pharmacodynamic

(PK/PD) model in order to establish a theoretical framework for analyzing the mechanism of the antimicrobial efficacy of the DIS.

The system design is described in Section 3.2. The Study-1 is presented in Section 3.3. The

Study-2 is presented in Section 3.4. The Study-3 is presented in Section 3.5. The chapter concludes in Section 3.6 with a summary of the results, and motivation for the next chapter.

Figure 3.1 Framework of the in vitro antimicrobial efficacy studies

3.2 System Design

There are four central components in the system: power source, silver anode, titanium cathode, and insulating electrode divider (Figure 3.2). The silver is connected to the DC power source 34

(batteries) as the anode to provide antimicrobial activities surrounding the device by releasing controlled amounts of Ag+. Despite the potential difference between the electrodes, the system is in a passive state (no system current) before it is implanted due to the insulating divider between electrodes. On implantation, the electrically conductive tissue and bodily fluids complete the circuit between the electrodes resulting in the active release of Ag+ from the anode. In this configuration, the rate of ion release may vary over time due to the electricity fluctuations caused by the batteries. In order to ensure consistency in the actual Ag+ release rate, with theoretical approximations, this research will also explore the feasibility of a closed loop feedback mechanism in the system by adopting an adjustable current source to regulate the current output. The detailed design of closed loop control will be presented in Chapter 5.

Figure 3.2 Key components of a basic DIS

Orthopaedic implants can vary substantially in geometry and structure based on their area of application and desired functionality. Instead of investigating the DIS for a specific implant 35

application, this dissertation focuses on a fundamental configuration that can be adapted in scale and geometry to satisfy functional requirements for specific orthopaedic implant applications. An example of the basic DIS’s adaptation for a hip implant is shown in Figure

3.3.

Figure 3.3 DIS adaptation for a hip implant

3.3 Antimicrobial Efficacy Study-1

The specific objective of antimicrobial efficacy Study-1 was to investigate the effects of electric current intensity and duration on the short-term antimicrobial efficacy of the DIS.

Electric current has been studied the most in similar LIDC activated silver-based systems since it is the driving force for release of the ions. Previous studies have shown a positive association between the current intensity and the antimicrobial efficacy of the system [24] [9]. This finding is consistent with the Faraday’s law that the amount of substance liberated at the anode is

directly proportional to the electric charge passing through the system [139]. However, it is not clear whether the antimicrobial efficacy relies solely on the amount of silver ions regardless of the manner in which they are released. In an LIDC activated system, the electric charge is the product of the current intensity and the duration. Hence, the interactions of these two parameters are the effects of main interest in the 2D agar testing study.

3.3.1 Materials and Methods

Test Protocols

In this test model, the DIS design was simplified to two separate metal wire electrodes connecting to an external DC source so that the testing was easy to operate. The electrodes were fully embedded within the Mueller Hinton (MH) agar without direct contact, thus allowing the need for a separate insulating electrode divider in the test model. The top surface of agar was inoculated with bacteria. If the bacteria were sensitive to the ions released, an inhibition zone (IZ), seen as a clear ring, would form around the anode while the rest of the top surface area of the agar would be covered by bacterial colonies. The degree of antimicrobial efficacy could be interpreted from the IZ area.

The first step of 2D agar test model was the preparation of the MH agar plates. 39g of MH agar powder was mixed with 1000 ml of deionized water and autoclaved at 121oC / 15 pps for 30 min. On completion of the autoclave process, the bottle was cooled in a water bath at 60°C for

30 min, and 25 ml molten agar was poured into each petri dish (100×15 mm; Fisher Scientific®,

Pittsburg, PA). The agar plates were allowed to sit undisturbed for 30 min to cool and solidify.

Next, two small holes were punched on diametrically opposite locations on the side wall of the petri dish using a heated needle (Ø 0.6 mm), about halfway along the agar thickness. Ag and

Ti electrode wires (Ø 0.5 mm) were inserted into the agar along an orthogonal line and connected to the external breadboard using copper connecting wires. The resistors were inserted on the breadboard and connected with the Ag wire in series. The resistance of the MH agar was estimated to be 200 kΩ. Through the breadboard, the Ag wire was connected to the positive lead of the battery, and Ti wire to the negative lead. A detailed schematic of the setup was presented in Figure 3.4.

To begin the actual test, 100 µl of the bacteria-inoculated PBS was pipetted onto the surface of agar and smeared uniformly over the surface in 3 directions using a cotton swab. By weighing the used swabs before and after agar inoculation, the average volume of the PBS on agar was estimated to be 50 µl. The electrical circuit was turned on to activate the DIS, and the entire setup (Figure 3.5) was incubated at 37°C for the specified test interval and without electricity for overnight thereafter. The area of the IZ was estimated by multiplying its averaged width by length.

Figure 3.4 A schematic of the 2D agar test model setup

Figure 3.5 An actual experimental setup of the 2D agar test model

Bacteria Inoculant Preparation

Escherichia coli (E. coli, MG 1655) and Staphylococcus aureus (S. aureus, ATCC 25213) were the two primary bacteria species tested in this dissertation. Isolated bacteria colonies were grown overnight at 37ºC from refrigerated samples on an MH agar plate. The desired starting inoculant concentration values for the 2D agar test was 107 CFU / m. The inoculant was prepared in 0.7% phosphate buffered saline (PBS). Two 5 ml tubes and two 1 ml cubic vials were used. One of the 5 ml tubes was marked “original” and the other as “dilute”, while the cubic vials were marked “control” and “test”. 2 ml of PBS was pipetted into the original tube, and 1 ml into the control vial. A needle loop was used to scrape off 1-2 bacterial colonies from the culture plate that was then mixed into the PBS in the original tube and vortexed for 5 sec.

Next, 1 ml of bacterial solution was pipetted into the test vial, and both, the control and the test vials were assessed for their optical density (OD 600 spectrophotometer, Shimadzu Inc,

Norcross, GA; λ = 600 nm). The control vial was used for calibration. Using the tasks described above, the spectrophotometer reading would ideally be ~0.1-0.3 indicating a concentration of

1-5×108 CFU / ml. 0.1 ml of the original bacterial solution was then pipetted into the 0.9 ml

PBS in dilute tube. For the 2D agar test, the inoculant was then applied to the agar plates using a cotton swab.

Design of Experiments

A randomized block experimental design was adopted in which bacteria type (E. coli and S. aureus) served as the blocking factor (a source of variability that is not of primary interest).

Three levels were tested for each main factors – 1.2 µA, 5 µA and 20 µA for current intensity,

and 15 hr, 1 hr and 4 hr for duration. The objective of adopting such low levels was to determine the lower bound of the parameter settings to achieve a significant antimicrobial efficacy. The current intensity was determined by proper combination of the resistors (Table 3.1). Three replicates were used for each of the nine combinations in every block resulting in a total of 54 agar plate-based antimicrobial efficacy experiments.

A regression model was built based on the empirical data and a set of validation experiments was carried out to evaluate the reliability of the regression model for the antimicrobial efficacy.

The levels of the factors were set to be at the quarter points of the original experimental design.

There were two samples in each group.

Table 0.1 Levels of current intensity in 2D agar test model

Set values of Actual value of Actual values of Resistor combinations current intensity battery voltage current intensity 1.2 µA Two 10 MΩ resistors in parallel 1.19 µA 5 µA 6.2 V One 1 MΩ resistor 5.16 µA 20 µA One 1 kΩ resistor 20.67 µA

Statistical Models

The effects of the investigated factors, current intensity and duration, were modeled using a two-way ANOVA model with the response variable y representing the IZ area, B representing the blocking effect, α and β representing the effects of current intensity and current duration respectively. Statistical significance (p < 0.05) was determined using the PROC GLM

Procedure (SAS 9.2 for Windows).

푦푖푗푘 = 휇 + 퐵푖 + 훼푗 + 훽푘 + 훼훽푗푘 + 휀푖푗푘 (Eq.3.1) where i = 1, 2 (index for blocks)

j = 1, 2, 3 (index for current intensity) k = 1, 2, 3 (index for current duration)

2 휀푖푗푘 ~ iid N (0, σ )

A regression model was established to estimate the relationships among variables using the maximum improvement R2 (MAXR) method [140]. The blocking factor (bacterial species) was transferred into a numeric form with E. coli = 1 and S. aureus = 2. The regression model was first established by fitting current intensity and duration with original values (1.2, 5, 20 and 0.25, 1, 4). A second model was then built with these two variables in common logarithmic form (0.08, 0.7, 1.3 and -0.6, 0, 0.6). The analysis was conducted using the PROC REG

Procedure (SAS 9.2 for Windows).

3.3.2 Results and Discussion

Sample images of agar plates from the experiments on E. coli and S. aureus are presented in

Figure 3.6 and 3.7 respectively. The rounded rectangle shape of IZ indicates that the diffusion of Ag+ in the agar is almost uniform along the length of the silver anode.

Figure 3.6 Sample agar plates from experiments on E. coli

Figure 3.7 Sample agar plates from experiments on S. aureus

Figure 3.8 gives the histogram of the IZ area. The positive effects of current intensity and duration on the antimicrobial efficacy can be observed from the graph. Furthermore, E. coli appears to be more sensitive to the DIS than S. aureus. For E. coli, treatment of the DIS with

1.2 µA for 0.25 hr leads to a discernible IZ (151.7 mm2). However, for S. aureus, treatment at this level can hardly result in a detectable IZ. The lower bounds of duration to achieve remarkable antimicrobial efficacy (IZ > 100 mm2) at 1.2 µA, 5 µA and 20 µA are 4 hr, 1 hr and 0.25 hr, respectively. The mean values of IZ area as well as the standard deviations are summarized in Table 3.2.

Figure 3.8 Histogram of IZ areas in 2D agar study

Table 0.2 Mean values and standard deviation of IZ area in 2D agar study

Factors Levels # of samples IZ Mean (mm2) IZ Std Dev (mm2) E. coli 27 247.4 53.8 Bacteria S. aureus 27 119.4 67.6 1.2 18 164.8 111.0 Intensity (µA) 5 18 184.8 80.6 20 18 200.6 70.3 0.25 18 125.3 79.1 Duration (hr) 1 18 192.1 92.4 4 18 232.7 58.7

The ANOVA results are summarized in Figure 3.9. The goodness-of-fit of the statistical model is validated by a high R2 of 0.8506. The p-values of all factors, including the interaction effect, are less than 0.05. This indicates that the antimicrobial efficacy of the system (IZ area) was significantly influenced by the bacterial specie, the current intensity, the current duration and the interaction effect between intensity and duration.

Dependent Variable: area Source DF Squares Mean Square F Value Pr > F Model 9 353529.1343 39281.0149 27.84 <.0001 Error 44 62071.3333 1410.7121 Corrected Total 53 415600.4676

R-Square Coeff Var Root MSE area Mean 0.850647 20.48180 37.55945 183.3796

Source DF Type I SS Mean Square F Value Pr > F

bacteria 1 221376.0417 221376.0417 156.93 <.0001 intensity 2 11626.9537 5813.4769 4.12 0.0229 duration 2 105755.5648 52877.7824 37.48 <.0001 intensity*duration 4 14770.5741 3692.6435 2.62 0.0477

Source DF Type III SS Mean Square F Value Pr > F

bacteria 1 221376.0417 221376.0417 156.93 <.0001 intensity 2 11626.9537 5813.4769 4.12 0.0229 duration 2 105755.5648 52877.7824 37.48 <.0001 intensity*duration 4 14770.5741 3692.6435 2.62 0.0477

Figure 3.9 ANOVA results of 2D agar test

Figure 3.10 summarizes the results of regression analysis. The model with logarithmic variables possesses better goodness-of-fit (R2 = 0.8411) compared to the one with original variables (R2 = 0.7840). The regression model for 2D agar test is given as follows:

퐼푍 = 355.66 − 128.24 ∙ 퐵푆 + 58.25 ∙ log10 (퐷) + 12.71 ∙ log10 (퐼) − 12.23 (Eq.3.2) ∙ log10 (퐷) log10 (퐼) where BS = bacterial species

D = current duration

I = current intensity

original model: R-Square = 0.7840 and C(p) = 5.000

Parameter Standard Variable Estimate Error Type II SS F Value Pr > F Intercept 301.82873 21.58486 359733 195.53 <.0001 bacteria -128.24074 11.67378 222017 120.68 <.0001 duration 34.05837 5.29204 76201 41.42 <.0001 current 3.62065 1.05856 21523 11.70 0.0013 interact -1.11931 0.44387 11699 6.36 0.0150 Log model : R-Square = 0.8411 and C(p) = 5.0000

Parameter Standard Variable Estimate Error Type II SS F Value Pr > F Intercept 355.65645 17.28814 572537 423.22 <.0001 bacteria -128.24074 10.01042 222017 164.11 <.0001 logD 58.25642 7.56954 80128 59.23 <.0001 logI 12.71312 4.35762 11514 8.51 0.0053 logInter -12.23442 3.84981 13662 10.10 0.0026

Figure 3.10 Results of regression model selection for 2D agar test

The coefficients of log variables in the regression model indicate that the antimicrobial efficacy of the system is more sensitive to the current duration (58.25) than the intensity (12.71). In

Figure 3.11, the surface approximation graph visualizes the unbalanced positive influences of duration and intensity. In addition, the logarithmic relationship is an implication of the

diminishing marginal effectiveness of these two factors. It indicated that the sustained release of silver ions has superior antimicrobial efficacy compared to the fast release. It also implied that the incremental antimicrobial efficacy of the system decreases as the current intensity incrementally increases while all other factors remains constant.

Figure 3.11 Surface approximation graph for 2D agar test

The results of the validation experiments are summarized in Table 3.3. It is observed that the theoretical values for all groups show an accuracy of over 80%. The regression model is particularly reliable (error = 0.01) for describing the antimicrobial efficacy of the DIS against

S. aureus.

Table 0.3 Results of verification experiments for the regression model

Intensity (µA) and Theoretical IZ Empirical IZ Bacteria % error duration (hr) means (mm2) means (mm2) E. coli 2.5, 0.5 279.75 237 18% E. coli 10, 2 350.84 293.25 20% S. aureus 2.5, 0.5 151.51 150 1% S. aureus 10, 2 222.6 220 1%

As discussed in Chapter 2, previous studies have shown that the antimicrobial efficacy of the

LIDC activated silver device largely depends on the quantity of Ag+ released from the anode, which is proportional to the total quantity of electric charge (QE) transferred at the anode.

However, the empirical results of the 2D agar study reveal an unbalanced interaction effect of the current intensity and the duration on the size of the IZ area, which indicates that the antimicrobial efficacy of the DIS is not only dependent on the quantity of the delivered Ag+, but also the manner of delivering. With the same QE, long-term current activation at lower intensity may result in better antimicrobial efficacy than short-term current activation at higher intensity. The logarithmic linear relation shows that the marginal effects of both factors decreases, further suggesting the lack of necessity of applying high current intensity in the DIS.

According to Faraday’s law, in order to restrict the amount of silver released from the DIS under the daily safety upper-bound (Section 2.5), the highest level of QE generated from the

DIS per day can be calculated as

푚퐹푧 0.125 ∙ 10−3 ∙ 96485 ∙ 1 (Eq.3.3) 푄 = = ≈ 111.8 푚푐 퐸 푀 107.87

where m = the mass of the substance liberated at an electrode

F = 96485 C / mol

M = the molar mass of silver z = the valency number of Ag+

In the 2D agar test model, the maximum of 푄퐸 is 18.6 mc (from the group of 20 µA × 4 hr), which is below the safety upper-bound. It should be noted that in clinical cases, with a well- controlled diet environment and corresponding medical care, the daily safety upper-bound of

QE for short treatment can be higher than the value proposed above.

3.4 Antimicrobial Efficacy Study-2

2D agar test is essentially a semi-quantitative assessment of the inhibition zones on the surface of a static agar plate. This testing method does not provide accurate information on the antimicrobial activity over time, nor can it describe the interactions between the bacteria strains and Ag+ in a 3D space. In order to conduct an accurate quantitative analysis and model the antimicrobial efficacy of the system over longer operating intervals, a 3D broth test model has been also been developed as part of this dissertation. In phase-1 of the study, we investigated the effect of cathode material on the antimicrobial efficacy of the dual metal (silver-titanium) design. Based on the results of the phase-1, phase-2 of the study was performed to investigate the effect of four other parameters, namely electrode separation distance, anode surface area, current frequency and current intensity focusing on the silver-titanium configuration only.

From the perspective of product design, the factors listed above are among the most critical

parameters in the DIS. Investigating the impacts of these factors is therefore the prerequisite of setting guidelines for future development and scale-up.

3.4.1 Materials and Methods

Prototype Preparation

In preparing the DIS prototype, 23 mm titanium wire (Ø 1 mm) was cut and a mark is made at the 20 mm interval. The excess 3 mm was used for electrical connections. Similarly, the required length of anodic silver wire (Ø 1 mm) was measured, marked, and cut. Proper length of Polytetrafluoroethylene (PTFE) heat shrinkable tubing (ZEUS®, Orangeburg, SC) was cut for each pair of electrodes according to test requirements. The edges of the PTFE tubing aligned with the markings (excess 3 mm) on the electrodes so that the correct surface area of each electrode was exposed. Next, approximately 100 mm of enameled copper connecting wire (Ø 0.2 mm) was cut and the insulation coating at both ends was removed by exposing to the flame of a Bunsen burner. The length of exposed wire did not exceed 10 mm. One exposed end of each wire was wound tightly around the 3 mm section of the DIS electrodes and the other ends were left loose for the time being. It was important to ensure that the wound end of the connecting wire did not protrude past the edge of the PTFE tubing beyond the 3 mm marking. Following that, the assembled electrodes with the PTFE tubing were subject to a hot air stream from a heat gun for up to 1 min until the tubing shrank to hold the electrodes in place. The prototype was then sterilized with 70% ethanol and rinsed by DI water three times.

A schematic diagram of the prototype was shown in Figure 3.12. The loose ends of the connecting copper wires were later connected to the electrical circuit during the actual broth

test to complete the DIS. Table 3.4 listed the functions, specifications and suppliers of materials for the electrode settings.

Figure 3.12 A schematic diagram of the DIS prototype for the 3D broth test model.

Table 0.4 List of materials for the electrode settings

Material Function Specifications Manufacturer/Supplier

Anode, 99.99% purity Advent Research Materials Ag wire cathode temper annealed Ø 1 mm Ltd, Oxford, UK

99.6% purity Advent Research Materials Ti wire Cathode temper annealed Ø 1 mm Ltd, Oxford, UK

PTFE heat Insulating AWG size No. 24 (I.D.) Zeus Inc, Orangeburg, SC shrink tubing material

Power Batteries 1.5 V Duracell RadioShack, Fort Worth, TX source

Rowan Cable Products Ltd, Copper wire Conduction Enameled, Ø 0.2 mm Hertfordshire, UK

Power Breadboard 2 1/8” Modular RadioShack, Fort Worth, TX circuit

Test Protocols

A volume of 1000 ml of MH broth was prepared by mixing 21 g of MH broth powder with

1000 ml DI water. The broth was autoclaved for 30 min and then cooled at room temperature.

9 ml MH broth was poured in a 15 ml centrifuge tube (Fisher Scientific®, Pittsburg, PA). The

DIS prototype was perpendicularly inserted into the centrifuge tube through a hole in the tube cap, with the silver anode at the top (Figure 3.13). The electrodes were then connected to the breadboard as described in Section 3.3.1. Four 1.5 V AA batteries (RadioShack®, Cary, NC) were connected in series serving as the power source. Teensy® USB development board 3.0

(PJRC, Sherwood, Oregon) was incorporated in the electric circuit to provide pulsating current with a frequency of 1 Hz. Current control program was pre-edited in C and uploaded into the development board. The development board and the circuit configuration were presented in

Figure 3.14. This board was driven by a 32 bit ARM Cortex-M4 48 MHz processor with 3.3

V output.

Figure 3.13 The schematic of the 3D broth test setup

Figure 3.14 Teensy® USB development board 3.0 for 3D broth test model 54

The experiment was performed using S. aureus as the target pathogen due to its wide presence in implant-associated infections. In addition, given that Gram-positive bacteria such as S. aureus are comparatively more difficult to inhibit than Gram-negative ones due to the presence of a thicker peptidoglycan layer on the cell wall, it is safe to assume that testing the performance of the DIS against Gram-positive species is the most critical [141]. 1 ml of the bacterial suspension (1 – 2 × 105 CFU /ml) is added into the tube. In order to have homogenized bacteria distribution in the broth and to avoid accumulation of dead cells or any filtrate, the entire setup tubes is incubated in an incu-shaker at 37oC with shaking speed of 100 rpm. The preliminary tests have been performed for 48 hours with the bacterial concentration measured at eight successive time points, namely 0 h, 3 h, 6 h, 9 h, 12 h, 24 h, 36 h and 48 h.

The bacterial concentration was measured through the plate counting method. At each time point, the solution was diluted to the level of 300-1000 CFU / ml through a series of one tenth log dilutions performed by adding 0.1 ml bacterial solution into 0.9 ml PBS each time.

Segmented X-dishes (VWR International®, Radnor, PA) were used as the agar plates. The process of preparing these MH agar plates was the same as described in Section 3.2.2. For each segmented X-dish, two sections were marked for the dilution level, one for 10 times higher and one for 10 times lower. After the bacterial solution was dry on the agar. The agar plates were incubated in 37°C overnight for colonies to form. By multiplying the count of plates by the total dilution factor of the solution, the total number of CFU in the original sample can be estimated.

N = Mean [CN × DF] (Eq.3.4)

where N = the concentration of the original bacterial suspension

CN = the colony number

DF = the dilution factor (marked on the plate)

Design of Experiments

The DOE was conducted in two phases. In phase-1 of the study, the effect of cathode material on the antimicrobial efficacy of the dual metal (silver-titanium) design was investigated. Based on the results of the phase-1, phase-2 of the study was performed to investigate the effect of four other parameters, namely electrode separation distance, anode surface area, current frequency and current intensity focusing on the silver-titanium configuration only. The effects of these four parameters were evaluated independently as the main factors through one-way factorial experiments. The experimental design is presented in Table 3.5. In phase-1, the current was set to be constant at 14 µA, which had been found to be effective against multiple pathogens in prior studies with the silver-silver implant configuration [24]. For each factorial level, three replicates were performed with four samples in each replicate (n=12). The level of electrode separation distance were changed by varying the length of the PTFE tubing. The levels of anode surface area were changed by controlling the effective length of the anode. The bacteria inoculated broth without the device served as the control group in all tests.

Table 0.5 Experimental design for important factors

Design Parameters Levels Controlled parameters

Anode length = 20 mm

- Silver Cathode length = 20 mm Cathode material Electrode separation distance = 10 mm Phase Titanium Current = constant 14 uA Cathode material = Ti 10 Electrode separation Anode length = 20 mm distance (mm) Cathode length = 20 mm 20 Current = constant 14 uA 31.4 Cathode material = Ti Cathode length = 20 mm Anode surface area (mm2) 62.8 Electrode separation distance = 10 mm

2 - 94.2 Current = constant 14 uA Cathode material = Ti

Phase 0 Anode length = 20 mm Current frequency (Hz) Cathode length = 20 mm 1 Electrode separation distance = 10 mm 1 Cathode material = Ti Anode length = 20 mm Current intensity (µA) 7 Cathode length = 20 mm 14 Electrode separation distance = 10 mm

Statistical Analysis

The bacterial concentrations in the broth were plotted as time-kill curves and were analyzed through the longitudinal model (Eq.3.2), which has a polynomial structure with a single independent variable: time, denoted as t. The effects of factor and time × factor are considered as fixed effects.

2 3 푦푖푗푘 = 훽0푖 + 훽1푖푡푖푗 + 훽2푖푡푖푗 + 훽3푖푡푖푗 + ⋯ + 휀푖푗푘 (Eq.3.5) where i = 1, 2 or i = 1, 2, 3 (index for factor level)

j = 1, 2, ... , 8 (index for time point) k = 1, 2,…, 12 (index for sample) 57

푦푖푗푘 = bacterial concentration (in log10) at time 푡푖푗 for factor level i in sample k

푡푖푗 = time point j for factor level i

훽0푖 = intercept for factor level i

훽n푖 = parameter of 푡푖푗 with power of n for factor level i

2 휀푖푗푘~𝑖𝑖푑 푁(0, 𝜎 )

The statistical analysis was conducted in two phases – model determination (phase-1) and significance analysis (phase-2). In phase-1, we tested four different polynomial structures

(linear to quartic polynomial) to fit the empirical data in order to determine the most proper model structure for the significance analysis. The Akaike information criterion (AIC) was used as the standard measurement of the relative quality of goodness-of-fit as well as complexity of the model [142]. The general form of AIC is given by:

퐴퐼퐶 = 2푘 − 2ln (퐿) (Eq.3.6) where k = the number of parameters in the statistical model

L = the maximized value of the likelihood function for the estimated model

AIC not only rewards goodness-of-fit, but also includes a penalty that is an increasing function of the number of estimated parameters [143]. Given a set of candidate models for the data, the preferred model is the one with the minimum AIC value.

The empirical data were modelled with a spatial power covariance structure based on the assumption that the covariance of two data points is related to the time interval. The spatial

2 푑푖푗 power covariance has the form of 𝜎푖푗 = 𝜎 𝜌 in which 푑푖푗 is the interval between time i and j, and ρ is a correlation parameter that measures the linear association between the errors 58

measured by hr. In phase-2, with the selected model structure, five independent significance analyses were performed, one for each design parameter, to determine whether or not there is a statistically significant difference (α = 0.05) between parameter levels in terms of antimicrobial efficacy. The analyses in both phases were performed using the PROC MIXED procedure in SAS (Version 9.2, SAS Institute Inc. Cary, NC).

3.4.2 Results and Discussion

The time-kill curves for S. aureus from the phase-1 and phase-2 testing are shown in (Figures

3.15-3.19). In the first 6 hours, there was an increase in bacterial concentration in all test and control groups in the first three hours, but the slopes from test groups were noticeably lower compared to the control groups. In this period of time, Ag+ are assumed to impede the growth rate by interfering with the cell replication process. After 6 hours of treatment, the bacteria concentration in test groups began to decrease. Except for the small group in the anode surface area test (Figure 3.17), the curves of concentration for all other test groups showed a continuous diminishing trend from 12 hours to 48 hours, forming a long-tail shape without signs of recovery indicating that the S. aureus cells in the broth were completely killed by the

DIS. However, for the small group in the anode surface area test, the bacterial concentration slowly increased after 12 hours of treatment until it reached the level of 108 CFU / ml. In each experiment, the curves of bacterial concentration for test groups suggested a firstly-increase- then-decrease fashion of the antimicrobial efficacy. This observation is the empirical foundation for establishing the quantitative model of the antimicrobial efficacy.

10 9 8 7 6 5 4 Ag 3 Ti 2 Control

1 Bacterial concentration(cfu/ml) Bacterial concentration(cfu/ml) in log10 0 0 10 20 30 40 50 Time (hr)

Figure 3.15 Time-kill curves from the experiments to test the effects of cathode material

10 9 8 7 6 5 4 10 mm 3 20 mm 2 control

1 Bacterial concentration(cfu/ml) Bacterial concentration(cfu/ml) in log10 0 0 10 20 30 40 50 Time (hr)

Figure 3.16 Time-kill curves from the experiments to test the effects of electrode separation

distance

10 9 8 7 6 5 7.85 mm2 4 15.7 mm2 3 23.55 mm2 2 control 1

Bacterial concentration(cfu/ml) Bacterial concentration(cfu/ml) in log10 0 0 10 20 30 40 50 Time (hr)

Figure 3.17 Time-kill curves from the experiments to test the effects of anode surface area

10 9 8 7 6 5 4 0 hz 3 1 hz 2 Control

1 Bacterial concentration(cfu/ml) Bacterial concentration(cfu/ml) in log10 0 0 10 20 30 40 50 Time (hr)

Figure 3.18 Time-kill curves from the experiments to test the effects of current frequency

10 9 8 7 6 5 14 µA 4 7 µA 3 1 µA 2 Control

1 Bacterial concentration(cfu/ml) Bacterial concentration(cfu/ml) in log10 0 0 10 20 30 40 50 Time (hr)

Figure 3.19 Time-kill curves from the experiments to test the effects of current intensity

In phase-1 of the statistical analysis, the AIC increased when more parameters were added into the model (Table 3.6). The linear model had the lowest AIC and therefore was chosen for phase-2 of the analysis.

Table 0.6 AICs for different polynomial models

Order dimension Model structure AIC means

st 1 (linear) 푦 = 훽0 + 훽1푡 + 휀 205.5

nd 2 2 (quadratic) 푦 = 훽0 + 훽1푡 + 훽2푡 + 휀 211.4

rd 2 3 3 푦 = 훽0 + 훽1푡 + 훽2푡 + 훽3푡 + 휀 237.1

th 2 3 4 4 푦 = 훽0 + 훽1푡 + 훽2푡 + 훽3푡 + 훽4푡 + 휀 282.1

The results of the significance analysis in phase-2 are summarized in Table 3.7 grouped by the criterion parameters.

Table 0.7 Longitudinal analysis results for the test of CM

Criterion Parameters Comparisons ChiSq F Value Pr > F

Silver vs. Control 127.78 63.89 <.0001

- Cathode material Titanium vs. Control 129.35 64.67 <.0001

Phase Silver vs. Titanium 0.01 0.01 0.9946 10 mm vs. Control 176.7 88.35 <.0001 Electrode separation 20 mm vs. Control 178.04 89.02 <.0001 distance 10 mm vs. 20 mm 0.01 0.01 0.9926 31.4 mm2 vs. Control 3.39 1.7 0.1897 62.8 mm2 vs. Control 36.52 18.26 <.0001 94.2 mm2 vs. Control 112.66 56.33 <.0001 Anode surface area 31.4 mm2 vs. 62.8 mm2 54.5 27.25 <.0001 31.4 mm2 vs. 94.2 mm2 41.35 20.67 <.0001

2 - 62.8 mm2 vs. 94.2 mm2 4.59 2.29 0.1075 0 Hz vs. Control 46.14 23.07 <.0001

Phase Current frequency 1 Hz vs. Control 44.68 22.34 <.0001 0 Hz vs. 1 Hz 0.01 0.01 0.9956 1 µA vs. Control 7.72 3.86 0.0264 7 µA vs. Control 93.81 46.9 <.0001 14 µA vs. Control 65.93 32.97 <.0001 Current intensity 1 µA vs. 7 µA 50.29 25.15 <.0001 1 µA vs. 14 µA 77.62 38.81 <.0001 7 µA vs. 14 µA 0.92 0.46 0.6334

Phase-1: The differences in the time-kill curves between the devices with silver cathode and the control group, and between the devices with titanium cathode and the control group, was

significant (p<0.0001). However, no significant difference was detected in the comparison of devices with silver and titanium cathodes (p=0.9946).

Phase-2: The differences in the time-kill curves between the devices with electrode separation distance of 10 mm and the control group, and between the devices with electrode separation distance of 20 mm and the control group, was significant (p<0.0001). However, no significant difference was detected between the two test groups (p=0.9926).

The differences in the time-kill curves between the devices with anode surface area of 62.8 mm2 and the control group, and between the devices with anode surface area of 94.2 mm2 and the control group, were significant (p<0.0001). In addition, the differences between the devices with anode surface area of 31.4 mm2 and 62.8 mm2, and between the devices with anode surface area of 31.4 mm2 and 94.2 mm2, were also significant (p<0.0001). However, no significant difference was detected between the devices with anode surface area of 31.4 mm2 and the control group (p=0.1897), or between the devices with the anode surface area of 62.8 mm2 and 94.2 mm2 (p=0.1075).

The differences in the time-kill curves between the test groups and the control group were significant (p<0.0001). However, no significant difference was detected between the devices with constant direct current (0 Hz) and square wave pulsating current (1 Hz) (p=0.9956).

The differences in the time-kill curves between the test groups and the control group were significant (p<0.05). The differences between the devices with 1 µA and 7 µA, and between the devices with 1 µA and 14 µA, were also significant (p<0.0001). However, no significant difference was detected between the devices with 7 µA and 14 µA (p=0.6335). 64

Due to the good antimicrobial efficacy against various types of pathogens such as bacteria, viruses, and fungal species, the electrically activated Ag mechanism has raised general interest as a non-antibiotic alternative for implant-associated infection prophylaxis and treatment [97]

[98] [144]. With the objective of minimizing Ag-usage in the system due to its potential cytotoxicity, this study adopted a dual metal implant configuration with silver and titanium.

The lack of statistical significance in the comparison of cathode material supports the hypothesis that the Ag cathode of the LIDC activated system can be substituted with Ti without undermining the system’s antimicrobial efficacy. This result is consistent with a recent semi- quantitative in vitro study. It can be explained by the fact that the primary antimicrobial activity of the system is attributed to the Ag+ which are released only from the anode. As long as the anode is Ag and the cathode material is conductive to enable the electrolysis, the antimicrobial efficacy of the system can be guaranteed. The efficacy of the system is also suggested to be independent of electrode separation distance, which refers to the positions of the Ag anode and the Ti cathode in the implant design. The variation in electric field caused by the changes in electrode separation within a certain range will not affect the Ag+ kinetics in the fluid. The proposed system can be hence reconfigured to some extent in scale and geometry to satisfy functional requirements for specific orthopaedic implant applications without compromising on the antimicrobial performance. Current frequency also appeared to be an insignificant parameter in this study. It suggested that the system can be driven by pulsating current, which consumes less energy, to achieve the same level of performance as using constant current.

Since this study only adopted 1 hz square wave, further studies with more levels of frequency

and wave shape are needed to characterize the relationship between these parameters and the antimicrobial efficacy of the system.

The antimicrobial efficacy of the system is sensitive to current intensity and anode surface area and the impacts of both parameters on the efficacy are nonlinear. The current intensity determines the reaction rate of the electrolysis, which will affect the antimicrobial efficacy of the system in two aspects. First, it is directly correlated to the Ag+ release from the anode surface, which inhibit the bacteria proliferation. Second, it is related to the oxidation process of the anode surface which is the main cause of the declines in the efficacy. Ion exchanges in the medium broth result in anodic oxide films, the product of the interactions between Ag and nutrition ingredients such as proteins, which reduce the device conductivity and thus the ion release rate. Therefore, given a certain anode surface area level, variation of the current intensity becomes a tradeoff between the initial Ag+ releasing rate and the anode surface oxidation rate. On the other side, the variation of anode surface area changes the ratio of the current to the anode surface over which that current is applied, which is correlated with the anode surface oxidation rate in the energized status. Hence it is theoretically feasible to reduce the anode surface oxidation rate by increasing the anode surface area. The nonlinear relationships between the current intensity, the anode surface area and the antimicrobial efficacy indicate that there will an effective range of parameter settings within which the system can demonstrate significant efficacy. Optimal antimicrobial efficacy could be achieved through appropriate combination of these two factors.

The objective of the statistical analysis is to test whether or not the selected design parameters in the study have significant impacts on the system’s antimicrobial performance. However, 66

because the linear structure is not capable to fit the true curve of the bacterial density, a simple regression model cannot provide a manageable way of understanding how the antimicrobial efficacy changes according to parameter variations. More proper mathematical models based on dynamic process simulation are needed to further analyze the relationships between the design parameters and the antimicrobial efficacy of the system. In addition to evaluating the in vitro antimicrobial performance, it is equally critical to assess the toxicity of the DIS before extending it to in vivo studies and applications.

3.5 Antimicrobial Efficacy Study-3

Due to the limitations of the traditional testing models, this dissertation introduced a broth- based in vitro testing model in which the system is immersed within a bacterial broth suspension and incubated in a shaker apparatus at 37°C. The experimental protocols and the time-kill curves were presented in Study-2. The well-controlled homogenous environment and quantitative nature of this test model provides an empirical foundation to build a mathematical model for the system. By adopting the broth-based testing model, it was found that the anode surface area and the current intensity would influence the antimicrobial efficacy of the system in a non-linear and interactive manner. However, the statistical model fails to provide a manageable way of understanding how the system impacts the bacterial density over time and how the intensity of the antimicrobial efficacy changes according to parameter variations, because the linear structure is not capable to fit or to predict the true curve of the bacterial density. In order to establish a theoretical framework for analyzing the antimicrobial efficacy of the DIS in the simulated fluid environment, a differential model is derived in this section based on classical in vitro Pharmacokinetic/Pharmacodynamic (PK/PD) model. Study-3 67

focuses on modelling the effects of the critical design parameters on the antimicrobial efficacy of the DIS.

3.5.1 Model Development

There are two basic compartments in a classic PK/PD model. The first compartment describes the microorganism growth and the effect of antimicrobial drugs (the pharmacokinetic submodel). The second compartment describes the changes of drug concentrations (the pharmacokinetic submodel) [145]. As illustrated in Figure 3.20, the antimicrobial activities of the DIS in a static environment can be modelled as an interaction of two processes: 1) silver- susceptible growing bacteria with first-order rate constants for growth 푟0 and 2) the effect of released silver ions on bacterial killing with a variable rate of k. The equations below describe the function of the bacterial concentration in the broth treated with the implant system:

푑푥 (Eq.3.7) = 푟 × 퐷(푥) − 푘 × 퐷(푥) {푑푡 0 푥(0) = 푥0 x = bacterial concentration at time t

푟0 = initial growth rate of bacterial cells D(x) = degree of the saturation concentration

푥 (Eq.3.8) 퐷(푥) = 푥 (1 − ) 푥푚푎푥

Firstly, equations of a common model of capacity-limited bacterial growth were derived based on S. aurues. Secondly, an antimicrobial submodel of silver ions were discussed based on empirical time-kill curves. 68

Figure 3.20 Illustration of the modified PK/PD model

In an ideal environment without any interference factors, the bacterial growth can be modeled with four different phases: lag phase, exponential phase, stationary phase, and decline phase

[146]. In order to characterize the natural growth of S. aureus, six samples of 10 ml S. aureus inoculated MH broth were prepared with the initial density of 103 – 104 CFU / ml. These samples were incubated at 37oC and monitored for 192 hr (eight days). A plot of the mean values of bacterial densities over time is presented in Figure 3.21. In the graph, the lag phase is indiscernible and therefore can be neglected in the model. The exponential phase lasted for

12 hr followed by the stationary phase. No sharp decrease or dramatic change was detected afterwards. Since the duration of the 3D broth test model is 48 hr, it is appropriate to not consider the decline phase in the model.

10.00

8.89 9.00 8.66 8.71 8.80 8.83 8.32 8.43 8.55 8.50

8.00 7.34 7.00

6.00

5.00

4.30

4.00 Bacterial density log10) in /ml(cfu density Bacterial 3.00 0 24 48 72 96 120 144 168 192 216 Time (hr)

Figure 3.21 Natural growth of S. aureus for 192 hr

In the exponential phase, the growth rate of S. aureus, denoted as r, is constant. The growth function is given by:

푟푡 푥(푡) = 푥0푒 (Eq.3.9)

퐿(푡) = 푙표푔10푥(푡) = 퐿0 + 푟푡(푙표푔10푒) (Eq.3.10) where 푥(푡) =the bacterial density at time t (CFU / ml)

퐿(푡) = the logarithmic bacterial density at time t (CFU / ml)

푥0 = the initial bacterial density

푟 = the logarithmic growth rate of S. aureus

Figure 3.22 gives the log plot of bacterial density for the first 6 hr. The parameters of Eq.3.10 are obtained using the least square method with R2 = 0.9949.

퐿(푡) = 푙표푔10푥(푡) = 3.6054 + 0.5874 ∙ 푡 (Eq.3.11)

푟 = 0.5874/(푙표푔10푒) = 1.3525 (Eq.3.12)

8 7 6 5

4 log10) 3 2

1 Bacterial concentration (cfu /ml in 0 0 1 2 3 4 5 6 7 Time (hr) .

Figure 3.22 Natural growth of S. aureus for 6 hr

The growth rate declines with the bacterial density and approaches to zero when the density reaches its maximum. The dynamics of the bacterial density is described by the model as below:

푑푥 푥 (Eq.3.13) = 푟푥 (1 − ) 푑푡 푥푚

푥(0) = 푥0

Where 푥푚 = the maximum of bacterial density in the broth.

Eq.3.13 be solved as

푥푚 (Eq.3.14) 푥(푡) = 푥 1 + ( 푚 − 1)푒−푟푡 푥0

8.89 From Figure 3.22, 푥푚 is estimated to be 10 .

From Figure 3.15-3.19 (kill-curve figures in Section 3.4), it can be observed that the DIS firstly reduced the growth rate of the bacteria and then reduced the bacterial concentration. This phenomenon was probably due to the antimicrobial mechanisms of Ag+. The other aspect was that the decrease rate of the density slowed down after 12 hr, indicating the antimicrobial efficacy of the DIS declined over time. Based on these observations, the basic assumptions for the modified PK/PD model are summarized below:

1) All batches of bacteria with the initial density within the range of 104-105 CFU / ml, which is the experimental setting, have an initial natural growth rate r0.

2) All batches of bacteria have the same maximal density of 108.89 CFU / ml, which is determined through the preliminary experiments.

3) The intensity of the antimicrobial efficacy exerted by the DIS, namely 푘, can be described as first order derivative of the bacterial concentration: 푘 = 푑푥/푑푡 . Obviously k < 0.

4) The antimicrobial effect of the system decays over time. The fatigue characteristic of k obeys power exponential law.

5) As long as the bacterial density reaches zero in log10 scale, which indicates that the absolute majority of bacteria have been killed, it stays at zero.

The antimicrobial effect of the system decays over time. The fatigue characteristic of k obeys power exponential law. With the assumptions above, a symbolic regression model, which

involves the breeding of simple computer programs or functions that are a good fit to a given set of data [1], is proposed to fit the shape of the experimental time-kill curves.

Let k(t) represents the antimicrobial efficacy of the system at time t, k(t) is given by

푡 푘(푡) = 푘0(1 − 훿) 푡, (Eq.3.15)

푘(0) = 푘0, 0 < 훿 < 1

훿 represents the accumulated fatigue rate of the antimicrobial efficacy.

The changing rate of bacterial concentration is given by

푡 푟(푡) = 푟0 − 푘(푡) = 푟0 − 푘0(1 − 훿) 푡 (Eq.3.16)

Eq.3.14 is therefore modified as below.

푑푥 푥 푡 (Eq.3.17) = 푥 (1 − ) [푟0 − 푘0(1 − 훿) 푡] 푑푡 푥푚

푥(0) = 푥0

Eq.3.17 can be solved by the method of separation of variables.

1 푥 (Eq.3.18) ∫ 푥 푑푥 = ln ( ) = ∫ 푟(푡) 푑푡 푥 (1 − ) 푥푚 − 푥 푥푚

Given that 푥(0) = 푥0,

푥푚 (Eq.3.19) 푥(푡) = 푡 푥푚 − ∫ 푟(푠)푑푠 1 + ( − 1)푒 0 푥0

The integration of 푟(푡) from 0 to time t is given by

푡 푡 푡 푡 (Eq.3.20) 푥 푥 ∫ 푟(푥)푑푥 = ∫ (푟0 − 푘0(1 − 훿) 푥)푑푥 = ∫ 푟0푑푥 − 푘0 ∫ (1 − 훿) 푥푑푥 0 0 0 0

푘 (1 − 훿)푡 − 1 = 푟 푡 − 0 [푡(1 − 훿)푡 − ] 0 ln (1 − 훿) ln (1 − 훿)

Take Eq.3.20 back into Eq.3.19 and it becomes

푥푚 (Eq.3.21) 푥(푡) = 푡 푥푚 푘0 푡 (1 − 훿) − 1 {1 + ( − 1) exp (푟0푡 − [푡(1 − 훿) − ])} 푥0 ln(1 − 훿) ln(1 − 훿)

Take the logarithm of Eq.3.21 and it becomes

푡 푥푚 푘0 푡 (1 − 훿) − 1 퐿(푡) = 퐿푚 − 퐿표푔10 {1 + ( − 1) exp [−푟0푡 + [푡(1 − 훿) − ]]} 푥0 ln(1 − 훿) ln (1 − 훿)

(Eq.3.22)

where 퐿푚 = 퐿표푔10푥푚 = 8.89

Under the assumption (5), the bacterial density under the effect of DIS is in a discrete fashion.

For an arithmetic progression TN = (t1, t2, … … , tN−1, tN), the bacterial density at time 푡푛 is given by

max {0, 퐿(푡푛)}, 퐷(푡푛−1) > 0 (Eq.3.23) 퐷(푡푛) = { 0, 퐷(푡푛−1) = 0

For 푡푛 ∈ 푇푁

3.5.2 Data Analysis

With a fixed initial bacterial density x0, the concentration plot is determined by k0 and δ, which are known as the antimicrobial efficacy parameters (AEP). For each level of the factors, the

AEP which result in the maximal coefficient of determination (R2) will be given to fit the empirical data. The maximal R2 is searched using the enumeration method, which is to calculate the R2 of plots with all combinations of AEP in a finite set of reasonable values.

Because there are eight time points in the experiment, only the corresponding eight simulated data points will be used in calculating the R2. The means of the AEP in all three replicates were taken into the final differential model.

The maximal antimicrobial efficacy (MAE) for a given set of AEP is defined as the maximal

th x0 with which the simulated bacterial density at the 48 hr is 0. The practical meaning of the

MAE is the maximal level of initial bacterial density in the 10 ml MH broth in which the DIS with certain parameter settings can kill all of the bacteria. Similarly, the MAE will be acquired via enumerating the values of x0. The analysis will be performed in MS Excel 2010® with MS

Visual Basic for Application 7.0. The algorithms (pseudo codes) for determining the optimal

AEP and MAE are shown in Figure 3.23.

Step 1. Determine the optimal AEP: Datapoint(1-8) = the empirical data points D(1-8) = the simulated data points calculated according to the differential model meanData = mean(Datapoint(1),Datapoint(2),…, Datapoint (8)) 8 2 Rtt = ∑푖=1(Datapoint(i) − meanData) For k = 0.3 to 1.3, delta = 0.01 to 0.3 8 2 Rer = ∑푖=1(Datapoint(i) − D(i)) If Rer < min_Rer Then min_Rer = Rer k_op = k delta_op = delta End if delta = delta + 0.001 Loop k = k + 0.01 Loop R_square = 1 – min_Rer/Rtt

Step 2. Determine the MAE For x0 = 2 to 8.89 Bacterial density (0) = x0 For t = 1 to 48 Calculate D(t) with k_op and delta_op If D(48) = 0 Then max_x1 = x1 End If x1 = x1 + 0.01 Loop

Figure 3.23 Algorithms for determining (a) the optimal AEP and (b) the MAE

3.5.3 Results and Discussion

Model with simulated data

Figure 3.24 shows the simulated time-kill curves generated by the modified PK/PD model given in Eq.4.17 with different parameters. It can be seen that some of the simulated plots are very close to the empirical plots presented in Figure 3.15-3.19. On the other hand, with fixed

AEP, different x0 also resulted in variations in the density plots (Figure 3.24(b)). For example,

5 6 the MAE for the setting of k0 = 0.7 and δ = 0.1 is between 10 and 10 .

10 9 8 7 6 k0=0 5 k0=0.3, δ=0.1 4 k0=0.3, δ=0.05 3 k0=0.7, δ=0.1 2 Bacterial density Bacterial(cfu /ml in density log10) k0=0.7, δ=0.15 1 0 0 10 20 30 40 50 60 Time (hr) a

9 k0 = 0.7, δ = 0.1 8

3 log(x0)=4 2 log(x0)=5 1 Bacterial concentration (cfu/ml Bacterial concentration (cfu/ml in log10) log(x0)=6 0 0 10 20 30 40 50 b Time (hr)

Figure 3.24 Plots of the differential models (a) with same x0 but different k0 and δ (b) with

same k0 and δ but different x0. 77

Cathode material

The simulated time-kill curves and the analysis results of the cathode material test are summarized in Figure 3.25 and Table 3.8 respectively. The model demonstrated excellent reliability with high level of goodness-of-fit. The R2 values are greater than 90% for all groups.

The initial hourly antimicrobial rates by using silver cathode and titanium cathode are 47.97% and 45.48% respectively. The simulation results confirm that the cathode material does not have a practically significant impact (difference of 0.68%) on the antimicrobial efficacy of the system. Given that no statistical difference is detected in the longitudinal analysis, the simulation results further suggest that the MAE of the DIS are within the range of 104.74 CFU

/ ml to 105.38 CFU / ml. It should be noted that these values only apply to the 48-hr testing in

10 ml MH broth inoculated with S. aureus.

Silver 2 Titanium

1 Bacterial concentration (cfu/ml Bacterial concentration (cfu/ml in log10) 0 0 10 20 30 40 50 a Time (hr)

2 Silver Titanium

1 Bacterial concentration (cfu/ml Bacterial concentration (cfu/ml in log10)

0 0 10 20 30 40 50 b Time (hr)

Figure 3.25 (a) Empirical bacterial concentration data and b) simulated time-kill curves for cathode material test

Table 0.8 AEP and MAE for different levels of CM

2 Cathode Replicates k0 δ RSS TSS R Rep1 0.64 0.098 1.276 29.004 95.60% Silver Rep2 0.75 0.106 0.044 17.395 99.75% Rep3 0.58 0.092 0.816 27.143 96.99% Rep1 0.56 0.088 0.319 33.687 99.05% Titanium Rep2 0.69 0.094 0.096 17.947 99.47% Rep3 0.58 0.092 1.511 20.817 92.74% Mean Initial hourly Cathode k0 δ MAE antimicrobial rate Silver 0.66 0.099 48.14% 104.74 CFU / ml Titanium 0.61 0.091 45.66% 105.38 CFU / ml

Electrode separation distance

The simulated time-kill curves and the analysis results of the electrode separation distance test are summarized in Figure 3.26 and Table 3.9 respectively.

10 mm 2 20 mm

1 Bacterial concentration (cfu/ml Bacterial concentration (cfu/ml in log10) 0 0 10 20 30 40 50 60 b Time (hr)

10 mm 2 20 mm

1 Bacterial concentration (cfu/ml in log10) 0 0 10 20 30 40 50 60 Time (hr) a

Figure 3.26 Empirical bacterial concentration data and b) simulated time-kill curves for electrode separation distance test

Table 0.9 AEP and MAE for different levels of electrode separation distance

electrode separation 2 distance Replicates k0 δ RSS TSS R Rep1 0.78 0.112 0.170 31.67 0.995 10 mm Rep2 0.59 0.074 0.041 35.55 0.999 Rep3 0.6 0.082 0.113 30.09 0.996 Rep1 0.77 0.106 0.290 31.09 0.991 20 mm Rep2 0.72 0.106 0.734 30.08 0.976 Rep3 0.64 0.08 0.115 29.51 0.996 Mean electrode separation Initial hourly k0 δ MAE distance antimicrobial rate 10 mm 0.66 0.089 48.14% 107.54 CFU / ml 20 mm 0.71 0.097 50.84% 106.65 CFU / ml

The initial antimicrobial rates of two levels are 48.14% and 50.14%, with δ of 9% and 10% respectively. In the simulated data graph (Figure 4.7(b)), there is no discernable difference in the time-kill curves for the two groups. The simulation results reveal that the electrode separation distance does not have a practically significant impact on the antimicrobial efficacy of the system. It is worth noting that the estimated MAE in the electrode separation distance test are 106.65 CFU / ml to 107.54 CFU / ml, which are higher than the MAE estimated in the cathode material test. Since all controlled variables in these two tests remained the same, this change is probably attributed to the improvement of the author’s proficiency in the experimental operations.

Electrode separation distance did not have a statistically or practically significant impact on the antimicrobial efficacy of the DIS. Since the antimicrobial efficacy is independent of electrode separation distance, which is one of the key design parameters, the proposed system can be reconfigured to some extent in scale and geometry to satisfy functional requirements for specific orthopaedic implant applications without compromising on the antimicrobial performance.

Anode surface area

The simulated time-kill curves and the analysis results of the anode surface area test are summarized in Figure 3.27 and Table 3.10 respectively.

10 9 8 7 6 5 4 31.4 mm2 3 62.8 mm2 2 94.2 mm2

Bacterial concentration (cfu/ml Bacterial concentration (cfu/ml in log10) 1 0 0 10 20 30 40 50 Time (hr) a

4 31.4 mm2 3 62.8 mm2 2 94.2 mm2 Bacterial concentration (cfu/ml Bacterial concentration (cfu/ml in log10) 1

0 0 10 20 30 40 50 b Time (hr) Figure 3.27 (a) Empirical bacterial concentration data and b) simulated time-kill curves for anode surface area test

Table 0.10 AEP and MAE for different levels of anode surface area

Anode 2 Replicates k0 δ RSS TSS R surface area Rep1 0.74 0.12 0.630 28.33 0.978 31.4 mm2 Rep2 0.76 0.238 0.819 20.92 0.961 Rep3 0.91 0.18 0.762 36.68 0.979 Rep1 0.81 0.116 0.264 26.81 0.990 62.8 mm2 Rep2 0.63 0.098 0.291 26.72 0.989 Rep3 0.69 0.104 0.314 28.99 0.989 Rep1 0.78 0.106 0.391 27.57 0.986 94.2 mm2 Rep2 0.71 0.088 0.233 28.03 0.992 Rep3 0.81 0.116 0.471 26.71 0.982 Mean Anode Initial hourly k0 δ MAE surface area antimicrobial rate 31.4 mm2 0.80 0.179 55.22% -- 62.8 mm2 0.71 0.106 50.84% 104.42 CFU / ml 94.2 mm2 0.77 0.103 53.54% 106.52 CFU / ml

The estimated initial antimicrobial rates of small (31.4 mm2), medium (62.8 mm2) and large

(94.2 mm2) anode surface area are 55.22%, 50.84% and 53.54%, respectively. The difference is almost negligible. But the δ for small anode surface area is 0.179, which is about 69% greater than that of medium (0.106) and large group (0.103). Obviously, it is the increase in fatigue rate that weakens the overall antimicrobial efficacy. The reason for this phenomenon may be explained by the following considerations and observations. Since the QE transferred between electrodes is constant, the variation of anode surface area essentially changes the ratio of current to the anode surface over which that current is applied, namely surface current density

(SCD) (µA / mm2), which reflect the electrochemical reaction rate of the silver anode in the energized status.

Q /푡 퐼 (Eq.4.18) 푆퐶퐷 = E = 푆 푆 where t = energizing time (s)

S = ASA (mm2)

I = current intensity (µA)

It was observed after the experiments that the anode surface was covered by a thin layer of black silver compounds, which was supposed to be the product of the interactions between silver and nutrition ingredients in MH broth such as proteins. The silver anodes of the small group became completely black after 12 hr while the anodes of the other two groups were only partially black (Figure 3.28). The silver compound coating created considerably high resistance which inhibited efficient oligodynamic iontophoresis. In other words, the system became less effective. Given that the kinetic Ag+ in the growth media would easily bind with other negative charges and lose the antimicrobial activities, the accumulative effect of the antimicrobial efficacy of the system was limited. Therefore the bacterial concentration would rise again in the small anode surface area group. For the groups with medium and large anode surface area, lower SCD resulted in slower reaction rate of silver oxidation. Therefore the samples in these groups had longer effective time, or smaller accumulative fatigue rate, during the 48-hr test interval. It can be concluded that the upper threshold of the SCD for a sustaining effectiveness is between 0.22 µA / mm2 and 0.45 µA / mm2.

Small Medium Large =m

Figure 3.28 Surface oxidation of silver wires after 12 hr

The black silver compounds were examined using Hirox® KH-7700 digital microscope system

(Hirox-USA, Hackensack, NJ). The silver compounds formed a cliff-shape raised platform at the boundary between the area exposed to broth media and the hidden area protected by PTEF tubing. The thickness of the silver compounds was found to be within the range of 2.9 – 5.2

µm. A sample picture of the silver compounds is presented in Figure 3.29.

Figure 3.29 A silver oxide layer sample observed by Hirox® KH-7700 digital microscope with magnification of 3500

As discussed above, the fatigue of the antimicrobial efficacy of the DIS is attributed to the oxidation of the silver anode in the MH broth. The fatigue rate of the anode is associated with the SCD, the current intensity per unit anode surface. Below a certain level of current intensity, the SCD is proportional to the anode surface area. Therefore, it is possible to reduce the fatigue rate of antimicrobial efficacy by increasing the anode surface area.

Current Intensity

The simulated time-kill curves and the analysis results of the anode surface area test are summarized in Figure 3.30 and Table 3.11 respectively.

The estimated k0 for 1 µA, 7 µA and 14 µA are 0.32, 0.63 and 0.66 respectively. These values translate to initial antimicrobial rates of 27.63%, 46.92% and 48.26%. The system with current intensity of 1 µA caused a growth delay but failed to diminish the bacterial density. With 7

µA, the system was able to eradicate the bacteria completely after 48 hr. These results imply that the lower bound of the current intensity to completely eliminate the bacteria cells with initial density of 104 CFU / ml lies between 1 µA and 7 µA.

The estimated δ for 1 µA, 7 µA and 14 µA are 0.073, 0.97 and 0.1 respectively. Lower current intensity reduces the accumulated fatigue rate in a nonlinear manner. Decreasing the current from 14 µA to 7 µA only caused 3% decline in δ. However, decreasing the current from 7 µA to 1 µA resulted in 27% decline in δ. The relationship between the current intensity and the AEPs are shown in Figure 3.31. Logarithmic trend lines demonstrated high goodness-of-fit to describe the nonlinear relationships.

10 9 8 7 6 5 1 uA 4 7 uA 3 2 14 uA

1 Bacterial concentration (cfu/ml Bacterial concentration (cfu/ml in log10) 0 0 10 20 30 40 50 Time (hr) a

10 9 8 7 6 5 4 1 µA 3 7 µA 2 14 µA

bacterial concentration (cfu/ml bacterial concentration (cfu/ml in log10) 1 0 0 10 20 30 40 50 b Time (hr)

Figure 3.30 (a) Empirical bacterial concentration data and b) simulated time-kill curves for current intensity

Table 0.11 AEP and MAE for different levels of current intensity

2 Current intensity Replicates k0 δ RSS TSS R Rep1 0.2 0.05 16.020 23.90 0.330 1 µA Rep2 0.47 0.1 1.048 16.81 0.938 Rep3 0.3 0.07 8.706 22.08 0.606 Rep1 0.56 0.09 7.728 12.96 0.404 7 µA Rep2 0.79 0.114 0.123 28.28 0.996 Rep3 0.55 0.088 4.633 15.37 0.699 Rep1 0.53 0.086 4.777 15.49 0.690 14 µA Rep2 0.56 0.088 0.319 33.69 0.991 Rep3 0.78 0.112 0.170 31.67 0.995 Mean Initial hourly Current intensity k0 δ MAE antimicrobial rate 1 µA 0.32 0.073 27.63% -- 4.06 7 µA 0.63 0.097 46.92% 10 CFU / ml 4.47 14 µA 0.66 0.100 48.26% 10 CFU / ml

With the relationships established in Figure 3.31. It was estimated that the minimum current intensity for the DIS to control the simulated infection in this study is 6 µA. Two simulated time-kill curves were given in Figure 3.32 to show the difference.

0.8 K0 Delta Log. (K0) Log. (Delta) 0.7 0.6 y = 0.1341ln(x) + 0.3335 0.5 R² = 0.9652 0.4 0.3 0.2 y = 0.0093ln(x) + 0.0746 R² = 0.8995 0.1 0 0 2 4 6 8 10 12 14 16 Current intensity (µA)

Figure 3.31 The relationships between the current intensity and the AEPs

8 5 µA 7 6 µA 6

1 Bacterial concentration (cfu/ml Bacterial concentration (cfu/ml in log10) 0 0 10 20 30 40 50 Time (hr)

Figure 3.32 Simulated time-kill curves for current intensity of 5 µA and 6 µA

Current Frequency

The simulated time-kill curves and the analysis results of the anode surface area test are summarized in Figure 3.33 and Table 3.12 respectively. The estimated k0 for 0 hz and 1 hz are 0.70 and 0.66 respectively. These values translate to initial antimicrobial rates of 50.51%, and 48.14%. The two groups have an accumulated fatigue rate of approximately 10%. The

MAE for 48 hr is within the range of 104.56 CFU / ml to 104.85 CFU / ml. No practically significant difference among groups is found according to the differential models.

Based on the analysis above, it can be concluded that the effect of changing the constant direct current to square wave pulsating current with is negligible on the antimicrobial efficacy of the system. The variation of frequency of direct current essentially changed the total electric quantity delivered through the system. Theoretically, applying 1 hz square wave pulsating current with intensity of 14 µA should have the same effect with a constant direct current of 7

µA. The empirical results were consistent with this hypothesis.

0 hz 2 1 hz

1 Bacterial concentration (cfu/ml Bacterial concentration (cfu/ml in log10)

0 0 10 20 30 40 50 a Time (hr)

3 0 hz 2 1 hz

1 Bacterial concentration (cfu/ml Bacterial concentration (cfu/ml in log10) 0 0 10 20 30 40 50 Time (hr) b

Figure 3.33 (a) Empirical bacterial concentration data and b) simulated time-kill curves for current frequency

Table 0.12 AEP and MAE for different levels of current frequency

2 CF Replicates k0 δ RSS TSS R Rep1 0.81 0.112 0.172 29.95 0.994 0 hz (constant Rep2 0.75 0.11 0.305 27.82 0.989 DC) Rep3 0.55 0.088 3.178 18.39 0.827 Rep1 0.74 0.108 0.250 29.70 0.992 1 hz (pulsating Rep2 0.71 0.106 0.211 29.69 0.993 current) Rep3 0.52 0.084 7.258 14.10 0.485 Mean Initial hourly CF k0 δ MAE antimicrobial rate 0 hz 0.70 0.103 50.51% 104.85 CFU / ml 1 hz 0.66 0.099 48.14% 104.56 CFU / ml

3.6 Chapter Summary

In this chapter, three studies were conducted to analyze and model the effects of several design parameters on the antimicrobial efficacy of the DIS, which was the first objective of this dissertation. Study-1 adopted a 2D agar test model as a semi-quantitative method to evaluate the effects of electric activation duration and current intensity on the short-term antimicrobial efficacy of the DIS. It was found that the antimicrobial performance of the DIS was more sensitive to current duration than current intensity, and the marginal antimicrobial efficacy of the DIS decreased as the current intensity increased. Therefore the low-intensity activation is optimal for the DIS activated by direct current.

Study-2 identified the critical design parameters which influence the antimicrobial efficacy of the DIS through a two-phase in vitro quantitative analysis. In phase-1, the silver-titanium configuration demonstrated equivalent antimicrobial efficacy compared to the silver-silver configuration. In phase-2, electric current intensity and anode surface area were demonstrated to be the dominant parameters influencing the antimicrobial efficacy of the silver-titanium configuration. The interactive effects of these two parameters influence not only the release of silver ions but also the formation of anodic oxide films. Electrode separation distance and the current frequency had an insignificant effect on the antimicrobial efficacy of the system, indicating that the geometry and the driving power of the system can be reconfigured to some degree without compromising its performance.

In study-3, a modified PK/PD model was established to provide a quantified association between the variation of the critical system parameters and the degree of antimicrobial efficacy. Based on the analysis of anode surface area and current intensity, it can be concluded that the change of current intensity would affect the antimicrobial efficacy of the DIS in two aspects. First, it directly determines the reaction rate of the oligodynamic iontophoresis, which is positively correlated to the initial killing rate (k0). Second, it is also one of the two determinants of the surface current density which is associated with the fatigue rate of the antimicrobial efficacy. Given a certain level of anode surface area, variation of current intensity becomes a tradeoff between the initial killing rate (k0) and the accumulative fatigue rate (δ).

The nonlinear relationships established by the model indicates that for a given level of anode surface area, there is an effective range of current intensity within which the system has significant antimicrobial efficacy.

The Objective 1 has been achieved in Chapter 3. Chapter 4 will focus on the Objective 2, which is to investigate the in vitro cytotoxicity of the DIS to human cells.

CHAPTER 4 IN VITRO CYTOTOXICITY OF THE SYSTEM

4.1 Introduction

In addition to evaluating the in vitro antimicrobial performance, it is equally critical to assess the in vitro toxicity of the DIS before extending it to in vivo studies and applications. Therefore,

Objective 2 of this dissertation was to assess the cytotoxicity of the DIS to mammalian cells and determine the optimal ranges of operating parameters to achieve antimicrobial efficacy without inducing discernible cytotoxicity. The cytotoxicity testing study is conducted at specific current and anode surface area levels within ranges which have been determined to demonstrate significant antimicrobial efficacy in Chapter 3. The cytotoxicity test model described here was developed specifically for the DIS, but can be potentially extended to other medical products.

The cytotoxicity studies were conducted in three phases in this dissertation (Figure 4.1). In

Phase-1, a mouse small intestine epithelial cell line (Mode-K, from Institut Pasteur de Lyon,

France) which was derived from C3H/HeJ mice, was used as the response cells [147]. The objective of Phase-1 was to develop an in vitro cytotoxicity test model for DIS on mammalian cells and to evaluate the feasibility of the proposed test model, especially, the possibility of cross contamination during testing. Mode-K cells were selected for this study because their fast growth rate is conducive to an efficiency model-proofing process and the material costs are lower.

Once the basic test model and protocols were established, Phase-2 was performed focusing on evaluating the toxicity of the DIS to human osteosarcoma cells (MG-63, ATCC® CRL-1427) at different current levels. MG-63 is a type of human osteosarcoma cell line derived from malignant bone tumors. It is commonly used as an osteoblastic model due to its excellent interferon production and potentials and growth capacity. Because it is an immortalized cell line which has evaded normal cellular senescence and instead can keep undergoing division, it is suitable for a design of experiment which requires multiple groups and a large sample size.

In addition, the mean doubling time of MG-63 cells is 2 to 3 fold faster than osteoblasts and the saturation density is 15 to 20-fold higher [148]. Therefore the total operating costs of using

MG-63 cells are lower than osteoblast cells. Finally, a validation study on normal human osteoblast cells (NHOst, CC-2538, Lonza®, Basel) was conducted in Phase-3 with the operating parameters confirmed by Phase-2 study. The cells have the ability to differentiate into mature osteoblasts expressing the normal osteoblast phenotype thus provide a homogenous, rapidly proliferating model system for studying cytotoxicity on human cells

[149].

Figure 4.1 Framework of the in vitro cytotoxicity studies

The Phase-1 is presented in Section 4.2. The Phase-2 is presented in Section 4.3. The Phase-3 is presented in Section 4.4. The Chapter concludes with research summary in Section 4.5.

4.2 Cytotoxicity Study Phase-1

The objectives of the cytotoxicity study Phase-1 are to develop an in vitro cytotoxicity test model of the DIS on mammalian cells and to evaluate the cytotoxicity of the DIS to mouse cells. The prototype configuration of the DIS used for cytotoxicity tests was similar to the one used in 3D broth antimicrobial efficacy test – the dual-metal electrodes separated by the PTFE heat shrinkable tubing, and connected by enameled copper wires to the external electrical circuit activated by LIDC. The detailed experimental procedures are summarized below.

4.2.1 Materials and Methods 100

Cell Preparation

An ampule of Mode-K cells, which were stored in liquid nitrogen (-180°C), were quick thawed in a 37°C water bath and then seeded in a 75 cm2 tissue culture flask (T-75 flask, Fisher

Scientific®, Pittsburg, PA) with 15 ml Dulbecco's Modified Eagle Medium (DMEM,

GIBCO®, Grand Island, NY) which contains 10% heat-inactivated fetal bovine serum. The

o cells were incubated at 37 C (5% CO2). Cells were examined under microscope every 24 hr to ensure there was no cross contamination. The medium was refreshed every three days. Cells in the T-75 flask were transferred into ten 60 mm tissue culture plates (T-60 plates, Greiner

Bio-One®, Frickenhausen, Germany) when they reached 90% confluency.

Before trypsinizing, the cells in T-75 flasks were rinsed with 5 ml Hank's balanced salt solution

(HBSS, no calcium no magnesium, GIBCO®, Grand Island, NY). After aspirating the HBSS,

5 ml Trypsin-EDTA (GIBCO®, Grand Island, NY) was added into each T-75 flask and the cells were incubated for 3-5 min. The cell suspension was then transferred from the T-75 flask into a 50 ml conical tube and centrifuged for 1 min at 800 G. After centrifugation, the cell pellet was resuspended in 10 ml DMEM by pipetting in and out gently.

A volume of 100 µl of cell suspension was mixed with 100 µl of Trypan Blue solution

(GIBCO®, Grand Island, NY) in a 5 ml snap cap tube. 10 µl of the mixed suspension was applied to the edge of the coverslip of the Fuchs–Rosenthal hemacytometer to be sucked into the void by capillary action which completely filled the chamber with the sample. The cells within the 4 corner grids of the chamber were counted under the microscope. The cell concentration per milliliter was calculated as

101

N = AC × 10,000 × DF (Eq.4.1) where N = total cell number in the plate

AC = average number of cells in the 4 corner grids of the chamber

DF = dilution factor

Based on the total cell number in the T-75 flask, appropriate amount of DMEM was added to cell suspension so that 4 ml of mammalian cells could be seeded in each of the eight T-60 plates at a concentration of ~1 × 106 cells / ml. Cells were further incubated for 3 days to reach

80 % confluency.

Cytotoxicity Test Protocol

The DIS prototypes were autoclaved for 30 min in a 250 ml glass bottle. For each T-60 plate in the test group, the lid was drilled with a small hole (Ø 1 mm) at the center so that the prototype could be suspended through the lid into the plate with the copper connecting wires passing through the hole. The height of suspension was set to be 1 mm above the plate bottom so that the prototype could be submerged into the liquid media without contacting the cells.

The prototype suspension was held in place using transparent tape. The electrical circuit and connections were similar to the 3D broth test model. Figure 4.2 gives the schematic of the testing setup. The entire setup including the electric circuit was then sterilized under UV light in the biosafety cabinet for 30 min. After checking the actual current, the electrodes were connected to the electric circuit. The current intensity was set to be 14 µA. All plates were then incubated at 37ºC (5% CO2). In experimental duration of Phase-1 is 24 hrs. The actual

102

experimental setup is presented in Figure 4.3. At the end of the testing interval, the electrodes were disconnected and removed before cell counting.

Figure 4.2 The schematic of the cytotoxicity test model

Figure 4.3 The actual experimental setup of the cytotoxicity test model

103

The dye exclusion test with Trypan blue was used to determine the number of viable cells present in a cell suspension after the testing interval. It was based on the principle that live cells would remain clear due to the intact cell membranes that exclude Trypan blue while the dead cells would be dyed blue. In this test, the growth medium in each tissue culture dish was aspirated after 24 hr of exposure to activated DIS and the cells were washed with 2 ml HBSS.

Cells were trypsinized from subconfluent cultures by adding 2 ml of Trypsin-EDTA solution followed by 5 min incubation at 37°C with regular gentle shaking. The cell suspension was then transferred to a 15 ml tube. 2 ml of HBSS was added to the empty plate to thoroughly rinse the plate and to collect the residual cells to the 15 ml tube. 100 µl of cell suspension was then added into 100 µl Trypan blue and thoroughly mixed. The counting procedures were the same as previously described. 20 µl of the dyed cell suspension was pipeted into the side of the hemacytometer under the slide. The total number of living cells in each plate was calculated as:

4 푁푡표푡푎푙 = 푚푒푎푛(푐표푢푛푡𝑖푛푔 푓푟표푚 ℎ푒푚푎푐푦푡표푚푒푡푒푟) × 8 × 10 (Eq.4.2)

Design of Experiment and Statistical Analysis

Each T-60 plate would have either a DIS as the test group or remained devoid of electrodes and served as a control. Four samples of each group (test and control) were used during each replicate. Only one replicate had been performed at each current intensity in the preliminary tests. The cells treated with and without the DIS served as the test group and the control group, respectively. There were four samples in each group. The viable cell number was analyzed as

104

the response variable using t-test with the null hypothesis that the means of test group and control group were equal.

Folded-F method was adopted to test the equality of variance between test and control groups.

If the variances of two groups were equal, Pooled method would be used to conduct the t-test on the means. Otherwise, Satterthwaite method would be used for the t-test. The normality test was performed to check the prerequisite assumption that the data for each group were normally distributed. Four different statistics were adopted, namely Kolmogorov-Smirnov statistic,

Anderson-Darling statistic, and Cramér-von Mises statistic which were based on the empirical distribution function (EDF). The statistical analysis was performed through PRC TTEST in

SAS 9.4.

4.2.2 Results and Discussion

No contamination was detected through the experiment and the morphology of the cells appeared normal before treated by DIS. Therefore the feasibility of the test model was confirmed. The means of the viable cell number of both groups are presented in Figure 4.4

105

6.91 7.13 7.00

6.00

5.00

4.00

3.00

2.00

Cell number Cell number scale onof Log10 1.00

0.00 Test Control

Figure 4.4 Means of viable Mode-K cell number

The results of the normality test are presented in Figure 4.5. With p-values of all tests greater than 0.05, the normality of the data was confirmed. Figure 4.6 shows the normal probability plot of the standardized residuals. It can be seen that the standardized residuals fit the linear pattern representing the idealized normally distributed data.

Fitted Normal Distribution for number

Parameters for Normal Distribution Parameter Symbol Estimate Mean Mu 6.980113 Std Dev Sigma 0.250628

Goodness-of-Fit Tests for Normal Distribution

Test ----Statistic------p Value------Kolmogorov-Smirnov D 0.17163565 Pr > D >0.150 Cramer-von Mises W-Sq 0.03604848 Pr > W-Sq >0.250 Anderson-Darling A-Sq 0.22811053 Pr > A-Sq >0.250

Figure 4.5 Normality test results of cytotoxicity test on Mode-K cells

106

Figure 4.6 Normal probability plot for residuals, cytotoxicity test on Mode-K cells

The t-test results are summarized in Figure 4.7. The assumption of equal variances is given in the last line of the output labeled “Equality of Variances”. The number after Prob>F gives the p-value for the test of equal variances. For the preliminary cytotoxicity test, this p-value is

0.5042. It is thus acceptable to assume equal variances between the two groups.

The TTEST Procedure Variable: cell number

Group Method Mean 95% CL Mean Std Dev 95% CL Std Dev Control 7.1083 6.8291 7.3876 0.1755 0.0994 0.6543 Test 6.8519 6.4251 7.2787 0.2682 0.1519 1.0000 Diff Pooled 0.2564 -0.1357 0.6486 0.2266 0.1460 0.4991 Diff Satterthwaite 0.2564 -0.1515 0.6643

Method Variances DF t Value Pr > |t| Pooled Equal 6 1.60 0.1607 Satterthwaite Unequal 5.171 1.60 0.1686

Equality of Variances Method Num DF Den DF F Value Pr > F Folded F 3 3 2.34 0.5042

Figure 4.7 t-test results of preliminary cytotoxicity test on Mode-K cells 107

The 95% confidence intervals of means for control group and test group were estimated to be

[6.8291, 7.3876] and [6.4251, 7.2787] respectively. Note that there are two types of t-tests, each of which could be used based on whether the null-hypothesis of equality of variances was rejected or not; Pooled method was selected in this study because the null-hypothesis of equal variances was accepted above. The t-test statistic was given as 1.60 and the column of

“Prob>|T|” gave the p-value of 0.1607. The null hypothesis that the group means were equal was thus accepted at the 5% significance level. It was concluded that the mean value of the viable cell number from test group was not significantly different from the mean value from control group. The system showed no signs of cytotoxicity on Mode-K cells in 24 hr treatment.

The phase-1 study demonstrated the feasibility of the cytotoxicity test model on Mode-K cells.

However, in consideration of efficiency and precision, the cell counting procedures can be further improved in the following two aspects. First, in the original protocol, the trypsinized cells were centrifuged and resuspended into 2 ml HBSS before cell counting in order to avoid long time exposure to trypsin which may cause damages to the integrity of the cells. In the actual experiment, the cell counting for four T-60 plates can be completed within only 15 minutes. In addition, resuspending the cells increases the system errors and the total experiment time. Therefore the trypsinized cells can be used for counting directly without being centrifuged and resuspended. Second, because dead cells will detach from the plate and become afloat, the absolutely majority of dead cells will be washed away by aspirating the medium and washing the cells with 2 ml HBSS before adding the trypsin. All the trypsinized cells can be considered as viable cells. Therefore it is not necessary to use Trypan blue to indicate the viability of the cells.

108

4.3 Cytotoxicity Study Phase-2

In phase-1 of the cytotoxicity study, an in vitro test model was established using mouse intestine cells (Mode-K). The feasibility of the model has been confirmed in Section 4.2. In this section, phase-2 of the cytotoxicity study is presented focusing on investigating the effects of current intensity and silver surface on the toxicity of the system using human osteosarcoma cells (MG-63, ATCC® CRL-1427).

4.3.1 Materials and Methods

Cell preparation

An ampule of frozen cells from liquid nitrogen was quick thawed in a 37°C water bath. The cell suspension was then transferred into a 5 ml conical tube and centrifuged for 10 seconds at

800 G. The supernatant was aspirated, and the cell pellet was resuspended into 2 ml of corresponding medium pipetting in and out gently. The cell suspension was transferred into a

75 cm2 tissue culture flask (T-75 flask, Fisher Scientific®, Pittsburg, PA) with 6 ml medium.

The medium for MG-63 cells include 500 ml Eagle's Minimum Essential Medium (EMEM)

(ATCC® 30-2003™) and 50 ml heat-inactivated fetal bovine serum. The cells were then

o incubated at 37 C (5% CO2) for seven days. Cells were examined under microscope every 24 hr to ensure there was no cross contamination. The medium was refreshed every three days.

Seven days after the incubation, the cells in the T-75 flask reached 100% confluency. The medium was aspirated and the cells were rinsed with 5 ml HBSS. A volume of 4 ml Trypsin-

EDTA (GIBCO®, Grand Island, NY) was added into the T-75 flask and the cells were

109

incubated for 3-5 min. The cell number was estimated according to direct counting method described in Section 4.2.1.

The cell suspension was resuspended into a 50 ml conical tube with 46 ml complete medium by pipetting and out gently. The cell suspension was then transferred into ten 60 mm tissue culture plates (T-60 plates, Greiner Bio-One®, Frickenhausen, Germany) with 4 ml in each plate, and one T-75 flask with 6 ml in the flask. The cells in the flask were prepared for the

o next replicate of experiment. The plates and the flask were incubated at 37 C (5% CO2) until the cells reach 80% confluency.

Cytotoxicity test protocol

As described in Section 4.2.1, the implant prototypes were assembled with silver or titanium anodes, which had been sterilized by UV light for 30 minutes. All plates were then incubated at 37ºC (5% CO2) with or without current for 48 hr. At the end of the testing interval, the electrodes were removed. The growth medium was aspirated and the cells were washed with

2 ml HBSS. Cells were trypsinized by adding 2 ml of Trypsin-EDTA solution followed by 5 min incubation at 37°C with regular gentle shaking. The trypsinized cell suspension was then directly used for cell counting.

An extra replicate of experiment was carried out for cell staining. After the treatment, the cells were washed twice with 0.9% Phosphate-buffered saline (PBS). A volume of 2 ml of 100% methanol was added into each plate and removed immediately. The cells were washed with

PBS again and 2 ml of Harris Hematoxylin (Protocol®, Fisher Scientific, Hampton, NH) was

110

added into each plate. After a three-minute incubation at room temperature, the Hematoxylin was removed. The cells were washed in tap water and dehydrated at room temperature.

Design of Experiment

Two factors were investigated in phase-2 experiment: silver anode surface and electric current.

The duration of the electric activation and the levels of the current intensity were determined by a series of single-replicate preliminary tests. In the preliminary tests, the cells treated with active DIS (test group) were compared to the cells with no treatment (control group) using the same test protocols described in Section 4.2. There were four samples in each group. The statistical significance was determined by PRC TTEST in SAS 9.3 with α=0.05. A flow chart

(Figure 4.8) shows the process of determining the experimental design.

111

Figure 4.8 Flow chart of preliminary tests for experimental design of Phase-2 Study

According to the results of the preliminary tests, the duration of electric activation was set to be 48 hours, and the current intensity was between 3 µA to 6 µA. The experimental design for phase-2 study is given in Table 4.1. The experiment were performed in five replicates with two samples in each replicate, resulting a total sample size of ten (n = 10 per treatment group).

A two-way ANOVA was performed to analyze the data in phase-2 study through PRC GLM in SAS 9.3.

112

Table 0.1 Experimental design of cytotoxicity test on MG-63 cells

Factors Levels Silver anode surface (mm2) 0 62.8 Electric current (µA) 0 3 6

4.3.2 Results and Discussion

The plots of cell viability in the preliminary tests are presented in Figure 4.9 and the t-test results are summarized in Figure 4.10. In the test with 14 µA, there was a significant difference

(p = 0.0281) between the test group and the control group in the viable cell number. In the test with 3 µA, however, there was no significant difference (p = 0.3155). According to the decision process given in Figure 4.8, the experimental design was determined as Table 4.1.

113

6.5 Test Control

6.3

6.1

5.9 Cell number in log10 5.7

5.5 0 5 10 15 20 25 a Time (hr)

6.5 Test Control

6.3

6.1

5.9 Cell number in Log10 5.7

5.5 0 5 10 15 20 25 b Time (hr)

Figure 4.9 Cell viability results of preliminary Phase-2 study a) with current of 14 µA, and b) with current of 3 µA.

114

Test with 14 µA

Tests for Normality

Test --Statistic------p Value------Shapiro-Wilk W 0.872191 Pr < W 0.1583 Kolmogorov-Smirnov D 0.24007 Pr > D >0.1500 Cramer-von Mises W-Sq 0.091428 Pr > W-Sq 0.1264 group Method Mean 95% CL Mean Std Dev 95% CL Std Dev c 6.3111 6.2681 6.3541 0.0270 0.0153 0.1007 t 6.0795 5.9074 6.2516 0.1082 0.0613 0.4033 Diff (1-2) Pooled 0.2316 0.0952 0.3680 0.0788 0.0508 0.1736 Diff (1-2) Satterthwaite 0.2316 0.0648 0.3984

Equality of Variances

Method Num DF Den DF F Value Pr > F Folded F 3 3 38.34 0.0137

The TTEST Procedure

Method Variances DF t Value Pr > |t| Pooled Equal 6 3.86 0.0084 Satterthwaite Unequal 3.1564 3.86 0.0281

Test with 3 µA

Tests for Normality Test --Statistic------p Value------Shapiro-Wilk W 0.96194 Pr < W 0.8284 Kolmogorov-Smirnov D 0.19208 Pr > D >0.1500 Cramer-von Mises W-Sq 0.045116 Pr > W-Sq >0.2500

Variable: number group Method Mean 95% CL Mean Std Dev 95% CL Std Dev C 6.3380 6.1944 6.4816 0.0902 0.0511 0.3364 T 6.2686 6.1270 6.4103 0.0890 0.0504 0.3319 Diff (1-2) Pooled 0.0694 -0.0857 0.2245 0.0896 0.0578 0.1974 Diff (1-2) Satterthwaite 0.0694 -0.0857 0.2245

Equality of Variances Method Num DF Den DF F Value Pr > F Folded F 3 3 1.03 0.9828

The TTEST Procedure Method Variances DF t Value Pr > |t| Pooled Equal 6 1.09 0.3155 Satterthwaite Unequal 5.9989 1.09 0.3155

Figure 4.10 The t-test results for preliminary Phase-2 study 115

Some representative plates in Phase-2 study are presented in Figure 4.11. Inhibition zones (IZ) were observed around the DIS with or without electrical current after 48 hr treatment. For the group of passive DIS, the IZs were around the silver electrodes. For the group of electrically- activated DIS, the IZs were around the whole implant prototype. No IZ was observed in the control group or in the titanium implant group. The cells at the periphery of the IZ were damaged. Structural signs that indicated irreversible cell injury and the progression of necrosis included the loss of cell membrane integrity and the release of cell’s content. However, the structure and appearance of the viable cells outside the IZs appeared to be normal under the microscope. The cells treated with the DIS were showed in Figure 4.12.

116

117

The total cell number of all testing groups after the treatment in phase-2 study were shown in

Figure 4.13. The mean values of the DIS groups were lower than the control group, but were at the same level of the starting number.

3500000

3000000

2500000

2000000

1500000

1000000

Cell number Cell number after treatement 500000

0 No treatment Passive DIS Ti with 3 µA DIS with 3 Ti with 6 µA DIS with 6 a µA µA

6.6 No treatment Passive DIS Ti with 3 µA 6.5 DIS with 3 µA Ti with 6 µA DIS with 6 µA

6.4

6.3

6.2

6.1 Cell number Cell number in 10log

5.9 0 10 20 30 40 50 b Time（hr）

Figure 4.13 Number of MG-63 cells after 48-hr treatment a). original data, and b). data in logarithmic scale. 118

The results of ANOVA were presented in Figure 4.14. The effect of silver surface on the cell number was significant (p=0.0045), while the effect of electric current (p=0.1889) as well as the interaction of silver and current were not significant (p=0.7071). In addition, the effect of replicate was not significant (p=3412) indicating a good consistency of experimental operation.

The SAS System The GLM Procedure Dependent Variable: cell number Source DF Sum of Squares Mean Square F Value Pr > F Model 9 0.54511184 0.06056798 1.97 0.0624 Error 50 1.53380946 0.03067619 Corrected Total 59 2.07892130

R-Square Coeff Var Root MSE cell Mean 0.262209 2.785086 0.175146 6.288718

Source DF Type III SS Mean Square F Value Pr > F replicate 4 0.14192234 0.03548059 1.16 0.3412 silver 1 0.27197615 0.27197615 8.87 0.0045 current 2 0.10573885 0.05286943 1.72 0.1889 silver*current 2 0.02140949 0.01070475 0.35 0.7071

Figure 4.14 Results of the ANOVA for Phase-2 study

The results of the statistical analysis suggested that the toxicity of the DIS was solely resulted from the silver anode. Low intensity direct current (<6 µA) did not show any detectable cytotoxicity, nor did it enhance the cytotoxicity of the silver anode. Due to the limitation of

119

research resources, we were unable to directly measure the silver ion concentration in the medium. Because the cytotoxicity of the silver is dosage dependent, the inhibition zone around the active DIS indicated a heterogeneous distribution of silver ions in the plate. Although the theoretical amount of silver ions released from the active DIS was remarkably higher than that from the passive DIS, the diffusion of the silver ions in plates with active DIS, according to the inhibition zone area, was still limited in the peripheral area around the DIS. This phenomenon was probably due to the nature of the electrolysis, in which the electric field constrained the silver ion diffusion in the pathway area between electrodes therefore controlled the silver ion concentration at the distal area.

With the same design parameters, the DIS did not have the same cytotoxic effects on prokaryotic (bacteria) and eukaryotic cells. This is probably due to the different cellular structure. Eukaryotic cells are usually larger than prokaryotic cells and exhibit a far bigger target for attacking silver ions. Eukaryotic cells also show higher structural and functional redundancy compared to prokaryotic cells and, therefore, higher silver ion concentrations are required to achieve comparable toxic effects than for bacterial cells. This difference provides a “therapeutic window” in which bacterial cells are successfully attacked, whereas harmful effects on eukaryotic cells cannot yet be observed [34].

The LIDC-activated DIS demonstrated significantly superior antimicrobial efficacy compared to passive DIS in previous studies. Meanwhile, it did not increase the cytotoxicity of the silver in this in vitro study. The cytotoxicity of the DIS was validated in Phase-3 study with normal human osteoblast cells.

120

4.4 Cytotoxicity Study Phase-3

In Phase-2, the source of the cytotoxicity of the DIS has been determined. The objective of

Phase-3 study is to validate the cytotoxicity of the DIS with normal human osteoblast cells

(NHOst).

4.4.1 Materials and Methods

The medium for NHOst cells include 500 ml Osteoblast Basal Medium (OBM™, Lonza®,

Basel) and OGM™ SingleQuots™ containing the following growth supplements: 50 ml of fetal bovine serum, 0.5 ml of Ascorbic Acid and 0.5 ml of Gentamicin (GA-1000). The experimental protocols were the same as Phase-2 study.

In Phase-3, the cells treated with active DIS (test group) for 48 hours were compared to the cells with no treatment (control group). The current intensity of the DIS was set to be 6 µA.

There were four samples in each of the five replicates, resulting in a total sample size of 20 in each group. A T-Test was performed using PROC TTest in SAS 9.4.

4.4.2 Results and Discussion

The total cell number of all testing groups after the treatment in phase-3 study are shown in

Figure 4.15. The mean value of the cell number in the test group, which adopted the DIS with

6 µA, was approximately 30% lower than that in the control group.

121

400000

350000

300000

250000

200000

150000

Cell number after the treatment 100000

50000

0 a control test

6 control

test 5.5

Cell number Cell number in log10 4.5

4 0 10 20 30 40 50 b Time (hr)

Figure 4.15 Number of NHOst cells after 48-hr treatment a). original data, and b). data in logarithmic scale.

The results of the T-test were presented in Figure 4.16. A high p-value of 0.9828 in the test of equality of variances indicated that the variances in two groups are equal, therefore the

Pooled method was adopted for the t-test. There was a significant difference (p = 0.006) between the test group and the control group in the viable cell number. 122

The SAS System The TTEST Procedure Variable: cell number Group N Mean Std Dev Std Err Minimum Maximum 0 20 262625 96164.9 21503.1 112500 467500 1 20 182125 77639.0 17360.6 70000.0 312500 Diff (1-2) 80500.0 87394.2 27636.5

group Method Mean 95% CL Mean Std Dev 95% CL Std Dev 0 262625 217618 307632 96164.9 73132.5 140456 1 182125 145789 218461 77639.0 59043.7 113397 Diff (1-2) Pooled 80500.0 24552.9 136447 87394.2 71422.6 112632 Diff (1-2) Satterthwaite 80500.0 24471.1 136529

Method Variances DF t Value Pr > |t| Pooled Equal 38 2.91 0.0060 Satterthwaite Unequal 36.383 2.91 0.0061

Equality of Variances Method Num DF Den DF F Value Pr > F Folded F 19 19 1.53 0.3591

Figure 4.16 Results of the T-test for phase-3 study

The safety use of silver-based medical devices, especially the in vitro toxicity and biocompatibility of silver coated material for orthopaedic implant, has been discussed for decades. It is still in controversial that whether silver is genotoxic or cytotoxic to osteoblast cells. Silver-coated implant has been approved for clinical use in Europe since 1999. However, some studies showed that silver was highly toxic to human cells and did not recommend the use of implantation of silver or silver-plated devices. In this study, silver showed toxicity to osteoblast cells. The IZs appeared only around the silver implants indicating that the toxicity of silver was dosage dependent. It was a limitation of methods and equipment in this study that

123

the silver ion concentration in the medium could not be monitored, nor the exact toxicity mechanism of silver ions could be investigated.

It should be noted that although the electrically-activated silver implant resulted in larger IZs, the statistical analysis on viable cell number suggested the electrical activation did not cause a higher cell death rate compared to passive silver implant. The current intensity in the 48-hr treatment was as high as 6 µA, at which the DIS has demonstrated antimicrobial efficacy in previous studies. The study indicated that the control of silver ions diffusion might be critical to limit the toxicity. Because of the cytotoxicity to osteoblast cells, the LIDC-activated DIS is not recommended for long-term implant at this stage. However, it can be configured for use in temporary articulating spacers as an alternative treatment for orthopaedic infection. Further studies on the cytotoxicity of DIS with more advanced models to evaluate cell morphology, metabolic activity and oxidative stress will be of interest.

4.5 Chapter Summary

In this chapter, an in vitro cytotoxicity analyses were conducted in three phases to complete the Objective 2 of this dissertation. In Phase-1, the basic test model and protocols for in vitro cytotoxicity evaluation of the DIS were established using mouse small intestine cells. The preliminary test results showed no cytotoxicity of the DIS to mouse cells after a 24-hr treatment with current intensity of 14 µA. At this current level, the DIS had demonstrated significant antimicrobial efficacy to reduce the bacterial concentration in vitro. The test model was further revised for Phase-2 and Phase-3 studies to improve the efficiency of experimental operations.

124

In Phase-2, the cytotoxicity of the DIS was evaluated with human osteosarcoma cells at different current intensity levels. The experimental design was determined through a series of preliminary tests. It was showed that after a 48 hr treatment, both passive and active DIS caused inhibition zones (IZ) around the DIS and significant decreases in cell viability. The LIDC- activated DIS with 3 µA and 6 µA caused an average decrease in cell viability of 30 % and

28% respectively. It should be noted that there was no significant difference in cell viability between groups with passive DIS and active DIS. This result suggested that the cytotoxicity of the DIS was solely attributed to the presence of silver and the electric activation did not increase the cytotoxicity of silver.

In Phase-3, the cytotoxicity of the active DIS was validated with normal human osteoblast cells. The 48-hr treatment caused a significant decrease in cell number by ~30%, which was consistent with Phase-2. It was noted that with the same design parameters, the DIS demonstrated more intensive cytotoxic effects on prokaryotic cells (in Chapter 3) than eukaryotic cells (in Chapter 4). Based on the in vitro analysis, LIDC-activated DIS at current stage was not recommended for long-term orthopaedic implant. Instead, new mechanisms need to be investigated for safer and more reliable medical applications of DIS as an alternative treatment for orthopaedic infections.

The Objective 2 has been achieved in Chapter 4. Chapter 5 will focus on the Objective 3, which is to develop the closed-loop controlled DIS and to evaluate the in vitro antimicrobial efficacy of the reengineered DIS.

125

CHAPTER 5 CLOSED-LOOP CONTROL ON OUTPUT CURRENT

5.1 Introduction

It has been shown in Chapter 4 that low intensity electrical activation did not increase the cytotoxicity of the DIS compared to the passive silver-based implant. Therefore it is of great importance to guarantee a stable antimicrobial efficacy of the DIS. With the consideration of the feasibility of clinical application, standard batteries are used in this research as power sources instead of galvanostat or potentiostat. However, using batteries can cause significant fluctuations in the current level [148]. Inaccurately controlled oligodynamic iontophoresis may potentially result in either inefficient antimicrobial activity or high silver concentrations in serum and organs. On the other hand, as battery drains the voltage will decrease. It is necessary to maximize the effective duration of the system with a given battery capacity level. Hence there is a fundamentally need for a constant output current from the DIS regardless of the inherent instability of its voltage. The issue of output current control has to be solved before applying the DIS to actual implant devices. A closed loop feedback control for the electrical system is of great importance to sustain stable performance of the system over a longer time interval.

This chapter focuses on Objective 3 of the dissertation, which is to design a closed-loop controlled DIS which is able to automatically adjust the output current to be constant according to the input voltage variation so that a stable antimicrobial efficacy can be achieved. The closed-loop control was realized by incorporating an adjustable current source, namely

126

LM334, into the electric circuit board. The broth-based testing model was adopted to investigate the lower bound of current intensity for a significant antimicrobial efficacy.

5.2 Materials and Methods

5.2.1 Electronic design

As shown in Figure 5.1, the LM334 (Texas Instruments, Dallas, TX) was a 3-terminal adjustable current source featuring a range of 1 to 104 µA operating current under dynamic voltage range of 1 to 40 V. It was incorporated with an operational amplifier in a closed loop circuit which could adjust internal voltage according to external voltage and resistance variation caused by environmental changes. Stable current intensity could be established with one external resistor and no other parts were required.

Figure 5.1 LM334 adjustable current sources

127

The total current through the LM334 (ISET) was the sum of the current passing through the SET resistor (IR) and the LM334’s bias current (IBIAS). A graph showing the ratio of these two currents was supplied under ratio of ISET to IBIAS in the Figure 5.2.

Figure 5.2 Application of LM334 as a basic 2-terminal current source

As shown in Eq.5.1, the current flowing through the regulation resistor (RSET) was determined

o by the corresponding voltage (VR), which was approximately 214 µV / K.

푉푅 (Eq.5.1) 퐼푆퐸푇 = 퐼푅 + 퐼퐵퐼퐴푆 = + 퐼퐵퐼퐴푆 푅푆퐸푇

Since IBIAS was simply a percentage of ISET for a given set current, the equation could be rewritten as follows.

푉푅 푛 (Eq.5.2) 퐼푆퐸푇 = ( ) 푅푆퐸푇 푛 − 1

128

In Eq.5.2, n was the ratio of ISET to IBIAS as specified in Figure 5.3. Since n was typically 16

o o for 퐼푆퐸푇 ≤ 10 휇퐴, for most set currents at body temperature of 37 C (~310 K), the equation could be further simplified to

푉푅 67.776 푚푉 (Eq.5.3) 퐼푆퐸푇 = (1.067) = 푅푆퐸푇 푅푆퐸푇

Figure 5.3 Performance characteristics of LM334, Ratio of ISET to IBIAS

The improved circuit was packaged as simple modules and incorporated into the existing DIS prototype. The circuit diagram of the closed-loop design was presented in Figure 5.4.

129

Figure 5.4 Circuit configuration with incorporated LM 334

5.2.2 Electric performance test

The actual output current of the closed loop controlled DIS was evaluated with different combinations of voltages and resistances. The output voltage ranges from 1.5 to 12 V and the regulating resistance ranges from 100 Ω to 100 kΩ. The electrical circuit board was configured using packages of 1.5 V batteries (1.5V x 4, Radioshack, Fort Worth, TX) and appropriate resistors to generate a constant current.

5.2.2 Antimicrobial test

In order to have results comparable to previous cytotoxicity study, the in vitro antimicrobial performance of the closed loop controlled system was evaluated in a similar model developed in Chapter 4. The implant prototypes were assembled with 20 mm in length and 1 mm in 130

diameter for both electrodes. After the sterilization, the assembled electrodes were placed in a

60-mm tissue culture plate (Greiner Bio-One®, Frickenhausen, Germany) with the connection wires passing through a small hole on the lid. The electrodes were connected to the current source LM 334 assembled in a development board according to Figure 5.4. The actual setup of the antimicrobial efficacy experiment was presented in Figure 5.5.

Figure 5.5 The actual setup of the antimicrobial efficacy experiment of the closed loop controlled DIS

Methicillin-resistant Staphylococcus aureus (MRSA, ATCC 43300) was used as the target pathogen in this study because it was difficult to treat infections caused by this strain through regular antibiotic-based treatment. MRSA was grown in Mueller Hinton broth medium at 37oC

131

overnight using the same protocol described in Chapter 3. The initial bacterial concentration of the suspension was estimated by the spectrophometer and the suspension was then serially diluted to 106 CFU / ml. A volume of 100 µl bacterial suspension was added into 4 ml Mueller

Hinton broth medium in each plate. After an inspection of the actual electric current, the implant system was connected to a 1.5 v batteries. The entire configuration was incubated at

37oC for 48 hrs and the bacterial concentration was measured at 24 hr and 48 hr.

5.2.3 Design of Experiments

The objective of the antimicrobial test is to investigate the threshold of current intensity with which the closed loop controlled dual metal implant system can reduce the bacterial concentration in a simulated infection environment. Based on the simulation results from the modified PK/PD model developed in Chapter 3, the lower bound of current intensity for a significant antimicrobial efficacy is below 6 µA. Therefore the study was modeled as a one- way factorial design of experiments with three levels, namely 0 µA (control group), 3 µA and

6 µA. Each experimental level has a sample size of eight. The data were assessed using the non-parametric Kruskal-Wallis Test with a significance level α = 0.05 (SAS® 9.3, Cary, NC).

5.3 Results and Discussion

The electric performance of the current source LM 334 incorporated circuit is shown in Table

5.1. For regulating resistance > 1 kΩ, the output current remained stable (standard deviation =

0.1% - 0.7%) when the voltage varied between 1.5 v to 12 v. The errors between the actual current values and the theoretical values were within 5%. For regulating resistance = 100 Ω, the output current remained stable (standard deviation = 0.1%) when the voltage varied 132

between 3 v to 12 v. However, when the voltage decreased to 1.5 v, the output current dropped from 659.8 µA to 286.5 µA due to the internal impedance of the current source. It indicated that the internal impedance of LM 334 was approximately 4 – 5 kΩ.

Table 0.1 Actual output current with different voltages and resistances

100 Ω 1 kΩ 10 kΩ 100 kΩ

1.5 v 286.5 µA 69.4 µA 6.89 µA 0.72 µA

3 v 659.8 µA 69.6 µA 6.9 µA 0.72 µA

6 v 658.1 µA 69.6 µA 6.9 µA 0.72 µA

12 v 659.9 µA 69.3 µA 6.91 µA 0.71 µA

Mean 566.08 µA 69.48 µA 6.9 µA 0.72 µA

Standard Deviation 186.385 µA 0.15 µA 0.008 µA 0.005 µA

Theoretical Values 677.76 µA 67.78 µA 6.78 µA 0.68 µA

The time-kill curves of the antimicrobial tests were presented in Figure 5.6 and the results of the statistical analysis were summarized in Table 5.2. The differences in the bacterial concentration between groups were significant at 24 hr and 48 hr (p < 0.0001). Compared to the passive dual metal implant prototypes, the closed loop controlled prototypes with current of 3 µA inhibited the bacterial growth for over two orders of magnitude at 24 hr and one order of magnitude at 48 hr. The prototypes with 6 µA reduced the bacterial concentration for over seven orders of magnitude at 24 hr and 48 hr. The empirical results indicated that the threshold

133

of effective current intensity was between 3 µA and 6 µA, which was consistent with previous hypothesis.

Table 0.2 Statistical analysis of the antimicrobial efficacy test

24 th hour Wilcoxon Scores

current N Sum of scores Expected under H0 Std Dev under H0 Mean score

0 µA 8 36.0 100.0 16.219509 4.5000

3 µA 8 103.0 100.0 16.219509 12.8750

6 µA 8 161.0 100.0 16.219509 20.1250

Kruskal-Wallis Test

Chi-Square 19.8323

DF 2

Pr > Chi-Square <.0001

48 th hour Wilcoxon Scores

current N Sum of scores Expected under H0 Std Dev under H0 Mean score

0 µA 8 36.0 100.0 16.151457 4.50

3 µA 8 100.0 100.0 16.151457 12.50

6 µA 8 164.0 100.0 16.151457 20.50

Kruskal-Wallis Test

Chi-Square 20.9351

DF 2

Pr > Chi-Square <.0001

134

10 0 µA 3 µA 6 µA 9 8 7 6 5 4 3 2

1 Bacterial concentration (cfu ml / in log10) 0 0 10 20 30 40 50 Time (hr)

Figure 5.6 Time-kill curves of the closed-loop controlled DIS

With the functioning current level in scale of microampere, the voltage-sensitive DIS has made treatment processes more vulnerable to supply voltage deviations. Such voltage deviations, commonly in the form of voltage sags, can cause process disruptions and result in loss of antimicrobial activities. Therefore, it is of great practical significance to seek cost-effective solutions which can help sensitive loads momentary power supply disturbances. The current source successfully maintained the output current of the DIS under the dramatic variations of voltage and thus stabilized the release of silver ions. There will be wide applications for closed loop controlled system in prevention and treatment for implant-associated infections. In the first phase, it can be utilized as the temporary bone replacement/fixation implants due to the

135

potential cytotoxicity of overdosed silver. In future product development, wireless control technology can be incorporated into the system to regulate the electric current in real time.

After the long-term antimicrobial efficacy, biocompatibility and mechanical properties are validated through in vivo studies, the system can be further developed not only for orthopaedic applications but also as a general solution for infection control.

5.4 Chapter Summary

A new closed loop control scheme has been developed for the LIDC-activated DIS. In addition to the existing open-loop circuit, a current source which utilizes an internal operational amplifier has been incorporated. Through analysis on the electrical performance and the antimicrobial efficacy of the new system, it is shown that the proposed scheme is superior compared to the existing scheme in terms of stability under variable voltages and significant antimicrobial efficacy with low vantage power supply. The study also shows that the lower bound of current intensity for a bactericidal effect was indicated to be between 3 µA to 6 µA.

This study suggested the possibility of further developing the DIS to temporary implant devices as an alternative treatment for orthopaedic infections.

136

CHAPTER 6 CONCLUSIONS AND FUTURE WORK

6.1 Introduction

This chapter summarizes the main results and specific contributions of this dissertation.

Directions for future research based on the knowledge gained through this dissertation research are also presented.

6.2 Research Summary

In this dissertation, an LIDC-activated dual metal implant system (DIS) was presented as an antimicrobial device for orthopaedic applications. The system was designed to induce local administration of silver ions from silver anode via low intensity direct current. The motivation of this dissertation was the lack of understanding of quantitative relations between the system parameters and the antimicrobial efficacy, as well as the cytotoxicity of the system. In addition, previous systems lacked a feedback-control mechanism on the output current which was the prerequisite for a stable antimicrobial performance. Therefore this dissertation addressed three fundamental problems associated with the functionality of the DIS: the antimicrobial efficacy, the cytotoxicity and the closed loop control of electric current.

In Chapter 2, a comprehensive literature review was been presented with regard to orthopaedic implant infections, conventional treatment options, clinical applications of silver, and silver oligodynamic iontophoresis. Recent statistics have shown that there is growing concern about the implant-associated infections. Due to the increasing presence of antibiotic-resistant species, silver has been widely used as an alternative antimicrobial material against infection because 137

of its multimodal bactericidal manner. LIDC-activated silver system which utilizes electric current to regulate the ion release process has demonstrated superior antimicrobial efficacy compared to passive silver devices both in vitro and in vivo. However, the major concern with the usage silver-incorporated medical devices is their potential adverse health effects. Some preliminary studies have suggested the possibility of sustaining silver ion concentration at levels that are detrimental to bacteria without inducing human cellular toxicity.

Chapter 3 focused on the system design and the first objective of this research, which was to characterize the antimicrobial efficacy of the DIS. The prototype proposed in this dissertation consists of a silver anode and a titanium cathode activated by a direct current source. Despite the potential difference between the electrodes, the system is in a passive state before it is placed into a conductive environment due to the insulating divider between electrodes. On implantation, the electrically conductive tissue and bodily fluids complete the circuit between the electrodes resulting in the active release of silver ions from the anode. Three studies were conducted to characterize the in vitro antimicrobial efficacy of the system. First, a semi- quantitative agar-based test model was developed to evaluate the interactive effects of electric current intensity and duration on the short-term antimicrobial efficacy of the DIS. The regression model based on the empirical data indicated that the antimicrobial performance of the DIS was more sensitive to activation duration than current intensity. Second, a quantitative broth-based test model was developed to identify the critical design parameters of the system.

Through a longitudinal analysis, electric current intensity and anode surface area were discovered to be the dominant parameters influencing the antimicrobial efficacy of the DIS.

Electrode separation distance and the current frequency had no significant effect on the

138

antimicrobial efficacy of the system, indicating that the geometry and the driving power of the system can be reconfigured to some degree without compromising its performance. Third, a modified PK/PD model was developed to model the relationships between the parameters

(electric current and silver anode surface) and the antimicrobial efficacy of the system, and to investigate the fatigue mechanism of the system. It was found that the electric current and the silver anode surface influenced both the release of silver ions and the formation of anodic oxide in a non-linear manner. The lower bound of these two parameters for the system to achieve a significant bactericidal effect were estimated by the modified PK/PD model.

Chapter 4 focused on the second objective of this research, which was to evaluate the cytotoxic effects of the DIS. First, the basic test model and protocols for in vitro cytotoxicity evaluation of the DIS were established and the preliminary test showed no cytotoxicity of the DIS on mouse intestine cells. Second, the effects of current intensity and silver surface on the toxicity of the system were investigated using human osteosarcoma cells. Silver showed toxicity to osteoblast cells after a 48 hr treatment, however, the electrical activation did not cause a higher cell death rate compared to passive silver implant. At last, the cytotoxicity of silver ions was validated with normal human osteoblast cells.

Chapter 5 focused on the third objective of this research, which was to incorporate a closed loop control mechanism into the DIS. The closed-loop control was achieved by incorporating an adjustable current source into the electric circuit board so that the output current could maintain stable regardless of voltage variation. The electrical performance of the closed loop controlled DIS demonstrated high reliability. The antimicrobial efficacy of the closed loop

139

controlled DIS was validated by MRSA. It was found that the lower bound of the current intensity for the DIS to exert a significant bactericidal effect was between 3 µA and 6 µA.

6.3 Reserch Contributions

The dual metal (silver-titanium) configuration of the implant system proposed in this research provides a conceptual solution for infection treatment which balances the antimicrobial efficacy and the biocompatibility. This research provides a systematic understanding of the functionality of the proposed antimicrobial implant system. A summary of the contributions identified from this research are presented as follows:

 By accomplishing Objective 1, this research provides substantive insights to the

interactive effects of electric current intensity and anode surface area on the

antimicrobial efficacy, as well as the fatigue rate of the DIS. It extends the current

knowledge of the fatigue mechanism of LIDC–activated silver system and quantifies

the associations between the variations of design parameters and the antimicrobial

performance. The fundamental concepts of the modified PK/PD model can be applied

to dynamic system simulation for future product development and industrial

assessment.

 By accomplishing Objective 2, this research contributes to the literature of LIDC-

activated silver, a quantitative assessment of in vitro cytotoxicity to mammalian cells.

It identified the source of the cytotoxicity of the DIS and validated that the LIDC

activation has insignificant effect on the cytotoxicity of silver. The testing models

140

developed in this research also provide a generalized approach of quantitative analysis

cytotoxicity of medical micro-devices.

 By accomplishing Objective 3, this research demonstrates the feasibility of stabilizing

the output current by incorporating a closed-loop controlled mechanism into the DIS.

It serves as a foundation of future development of silver-based system which is able to

maintain constant antimicrobial activities at local microenvironment regardless of the

fluctuation of power source.

Overall, this research has contributed to the growing literature on the efficacy analysis of non- antibiotic agent, and complemented the literature in both engineering design and microbiological analysis that links information to the biomedical engineering. The systematic characterization of the functionality and mechanisms of the DIS has provided guidelines of product development, which can be used as reference of future standards and regulations for

LIDC-activated silver-based medical devices.

6.4 Future Research

The major limitations of the LIDC-activated DIS proposed in this dissertation include the short effective duration, the significant cytotoxicity to human bone cells, and the rudimental electric configuration. The limitations of our research lead to several potential directions for future research.

In terms of antimicrobial performance, the system can be reengineered in future by adopting alternating current as the power source to prevent the oxidation on the anode so that the effective duration of the system can be prolonged. It will be important to characterize the 141

effects of current intensity and frequency on the efficacy of the system and to optimize the parameter settings to achieve optimal performance. In addition, the antimicrobial efficacy analysis can be improved by adopting dynamic testing system in which the medium is continuously flowing with or without periodic renewal. The modified PK/PD model can be improved by including more factors such as medium volume, medium flow rate, dilution rate, etc. Therefore the in vitro study can better reflect the in vivo conditions of the silver ions.

In regard to toxicity studies, more advanced models are required to investigate the adverse effects on cell viability, structural integrity as well as functions caused by silver ions. In our cytotoxicity study on human osteosarcoma cells (MG-63 cells, Phase-2 study), we performed an additional pilot test trying to investigate the functionality of the viable cells after the treatment. The cells were treated with the active DIS in growth medium for 48 hours as described in Section 4.3. After the treatment, the DIS was removed and the cells were

o continuously incubated with refreshed growth medium at 37 C (5% CO2) for another 96 hours.

The cells were then stained by Harris Hematoxylin and observed under microscope. Figure

6.1 shows that the cells which survived after the treatment of active DIS were able to grow into the inhibition zone, indicating that 1) these viable cells maintained the reproduction function, and 2) the silver ions did not cause a permanent damage to the cells outside the inhibition zone.

In future, it will be critical to analyze the time-course concentration of silver ions in the medium with different design parameters and to establish a quantitative model describing the relationship between silver concentration and the corresponding cytotoxicity. In vivo studies on the toxicity of the DIS will be needed before further development for clinical applications.

142

Figure 6.1 MG-63 cells growing into the inhibition zone 96 hours after removing the DIS and refreshing the medium.

With respect to electrical design, future research work may include establishing an electrical equivalent model characterizing both attachment and growth of the biofilm on implant surfaces. It is of practical significance to discuss the feasibility of detecting and quantifying the implant-associated infection using biosensors based on electrochemical impedance measurement. The ultimate goal of the research field is to develop wireless controlled antimicrobial devices by incorporating medical sensor, vibration-based electromechanical power generator and wireless technologies for batteryless real time monitoring, analysis and control of local infection.

143

REFERENCES

[1] C. A. Jones and L. C. Li, "Total Joint Arthroplasty," Clinics in Geriatric Medicine, vol. 28, no. 3, pp. xi-xii, 2012.

[2] R. M. Moore, S. Hamburger, L. L. Jeng and P. M. Hamilton, "Orthopedic implant devices: prevalence and sociodemographic findings from the 1988 National Health Interview Survey," Journal of Applied Biomaterials, vol. 2, no. 2, pp. 127-131, 1991.

[3] Centers for Disease Control and Prevention,, "Number of all-listed procedures from discharges from short-stay hospitals by procedure category and age: United States," 2010. [Online]. Available: http://www.cdc.gov/nchs/fastats/inpatient-surgery.htm. [Accessed 7 1 2015].

[4] S. Kurtz, K. Ong, E. M. F. Lau and M. Halpern, "Projections of primary and revision hip and knee arthroplasty in the United States from 2005 to 2030," J Bone Joint Surg Am, vol. 89, no. 4, pp. 780-785, 2007.

[5] J. R. Lentino, "Prosthetic Joint Infections: Bane of Orthopedists,," Clinical Practice, vol. 36, no. 1, pp. 1157-1161, 2003.

[6] J. S. Groeger, A. B. Lucas, H. T. Thaler, H. Friedlander-Klar, A. E. Brown, T. E. Kiehn and D. Armstrong, "Infectious Morbidity Associated with Long-Term Use of Venous Access Devices in Patients with Cancer," Annals of Internal Medicine, vol. 119, no. 12, p. 1168, 1993.

[7] A. Trampuz and W. Zimmerli, "Diagnosis and treatment of infections associated with fracture-fixation devices," Injury, vol. 37, no. 2, pp. S59-66, 2006.

[8] C. Lavernia, D. J. Lee and V. H. Hernandez, "The Increasing Financial Burden of Knee Revision Surgery in the United States," Clinical Orthopaedics & Related Research, vol. 446, pp. 221-226, 2006.

[9] R. A. Shirwaiker, M. E. Samberg, P. H. Cohen, R. A. Wysk and N. A. Monteiro- Riviere, "Nanomaterials and synergistic low-intensity direct current (LIDC) stimulation technology for orthopedic implantable medical devices," WIREs Nanomedicine and Nanobiotechnology, vol. 5, p. 191–204, 2013.

[10] A. Trampuz and W. Zimmerli, "Antimicrobial Agents in Orthopaedic Surgery: Prophylaxis and Treatment," Drugs, vol. 66, no. 8, p. 1089, 2006.

144

[11] E. M. Downes, "Late infection after total hip replacement," The Journal of bone and joint surgery, vol. 59, no. 1, p. 42, 1977.

[12] A. Widmer, M. Rotter, A. Voss, P. Nthumba, B. Allegranzi, J. Boyce and D. Pittet, "Surgical hand preparation: state-of-the-art," Journal of Hospital Infection, vol. 74, no. 2, pp. 112-122, 2010.

[13] D. L. Cavanaugh, J. Berry, S. R. Yarboro and L. E. Dahners, "Better Prophylaxis Against Surgical Site Infection with Local as Well as Systemic Antibiotics," The Journal of Bone and Joint Surgery, vol. 91, pp. 1907-1912, 2009.

[14] L. Hernandez, M. E. Muñoz, I. Goñi and M. Gurruchaga, "New injectable and radiopaque antibiotic loaded acrylic bone cements," Journal of Biomedical Materials Research Part B: Applied Biomaterials, vol. 87B, no. 2, pp. 312-320, 2008.

[15] R. Namba, E. Paxton, D. Fithian and T. Slipchenko, "Increased Infection Rates with Primary TKA Implanted with Antibiotic Loaded Cement," The Journal of Arthroplasty, vol. 24, no. 2, p. e29, 2009.

[16] S. A. Jones, P. G. Bowler, M. Walker and D. Parsons, "Controlling wound bioburden with a novel silver-containing Hydrofiber® dressing," Wound Repair and Regeneration, vol. 12, no. 3, pp. 288-294, 2004.

[17] G. Pugliese and M. S. Favero, "Efficacy of silver-coating central venous catheters," Infection Control and Hospital Epidemiology, vol. 20, no. 6, p. 416, 1999.

[18] M. Pollini, F. Paladini, M. Catalano, A. Taurino, A. Licciulli, A. Maffezzoli and A. Sannino, "Antibacterial coatings on haemodialysis catheters by photochemical deposition of silver nanoparticles," Journal of Materials Science: Materials in Medicine, vol. 22, no. 9, pp. 2005-2012, 2011.

[19] J. C. Odekerken, T. J. Welting, J. J. Arts, G. H. Walenkamp and P. J. Emans, "Modern Orthopaedic Implant Coatings — Their Pro’s, Con’s and Evaluation Methods," in Modern Surface Engineering Treatments, InTech, 2013.

[20] M. Ip, S. Lui, V. Poon, I. Lung and A. Burd, "Antimicrobial activities of silver dressings: an in vitro comparison," Journal of Medical Microbiology, vol. 55, no. 1, pp. 59-63, 2006.

[21] J. J. Castellano, F. K. Susan M Shafii, G. Donate, T. E. Wright, R. J. Mannari, W. G. Payne, D. J. Smith and M. C. Robson, "Comparative evaluation of silver-containing antimicrobial dressings and drugs," International Wound Journal, vol. 4, no. 2, pp. 114-122, 2007.

145

[22] R. O. Darouiche, "Anti-Infective Efficacy of Silver-Coated Medical Prostheses," Clinical Infectious Diseases, vol. 29, no. 6, pp. 1371-1377, 1999.

[23] A. Chakravarti, S. Gangodawila, M. J. Long, N. S. Morris, A. R. Blacklock and D. J. Stickler, "An electrified catheter to resist encrustation by Proteus mirabilis biofilm," The Journal of urology, vol. 174, no. 3, pp. 1129-1132, 2005.

[24] T. A. Fuller, R. A. Wysk, C. Charumani, M. Kennett, W. J. Sebastiennelli, R. Abrahams, R. A. Shirwaiker, R. C. Voigt and P. Royer, "Developing an engineered antimicrobial/prophylactic system using electrically activated bactericidal metals," Journal of Materials Science: Materials in Medicine, vol. 21, no. 7, pp. 2103-2114, 2010.

[25] R. A. Shirwaiker, R. A. Wysk, S. Kariyawasam, H. Carrion and R. C. Voigt, "Micro- scale fabrication and characterization of a silver-polymer-based electrically activated antibacterial surface," Biofabrication, vol. 3, no. 1, p. 015003, 2011.

[26] R. A. Wysk, W. J. Sebastianelli, R. A. Shirwaiker, G. M. Bailey, C. Charumani, M. Kennett, A. Kaucher, R. Abrahams, T. A. Fuller, P. Royer, R. C. Voigt and P. H. Cohen, "Prophylactic Bactericidal Orthopedic Implants - Animal Testing Study," Journal of Biomedical Science and Engineering, vol. 3, pp. 917-926, 2010.

[27] A. Nanda and M. Saravanan, "Biosynthesis of silver nanoparticles from Staphylococcus aureus and its antimicrobial activity against MRSA and MRSE," Nanomedicine: Nanotechnology, Biology and Medicine, vol. 5, no. 4, p. 452–456, 2009.

[28] R. Strohala, M. Schellingb, M. Takacsc, W. Jureckac, U. Gruberd and F. Offner, "Nanocrystalline silver dressings as an efficient anti-MRSA barrier: a new solution to an increasing problem," Journal of Hospital Infection, vol. 60, no. 3, pp. 226-230, 2005.

[29] J. V. Loh, S. L. Percival, E. J. Woods, N. J. Williams and C. A. Cochrane, "Silver resistance in MRSA isolated from wound and nasal sources in humans and animals," International Wound Journal, vol. 6, no. 1, pp. 32-38, 2009.

[30] Z. Tan, A. Ganapathy, P. E. Orndorff and R. A. Shirwaiker, "Effects of cathode design parameters on in vitro antimicrobial efficacy of electrically-activated silver- based iontophoretic system.," Journal of Materials Science: Materials in Medicine, vol. 26, no. 1, pp. 1-10, 2015.

[31] D. Williams, "Revisiting the Definition of Biocompatibility," Medical Device Technology, vol. 14, no. 8, p. 10, 2003.

146

[32] M. Bosettia, A. Masse, E. Tobin and M. Cannas, "Silver coated materials for external fixation devices: in vitro biocompatibility and genotoxicity," Biomaterials, vol. 23, no. 3, pp. 887 - 892, 2002.

[33] Y. Zhou, J. Yang, T. He, H. Shi, X. Cheng and Y. Lu, "Highly stable and dispersive silver nanoparticle-graphene composites by a simple and low-energy-consuming approach and their antimicrobial activity," Small, vol. 9, no. 20, pp. 3445-3454, 2013.

[34] V. Alt, T. Bechert, P. Steinrücke, M. Wagener, P. Seidel, E. Dingeldein, E. Domann and R. Schnettler, "An in vitro assessment of the antibacterial properties and cytotoxicity of nanoparticulate silver bone cement," Biomaterials, vol. 25, no. 18, p. 4383–4391, 2004.

[35] T. Albrektsson and C. Johansson, "Osteoinduction, osteoconduction and osseointegration," European Spine Journal, vol. Suppl 2, pp. S96-101, 2001.

[36] L. Carlsson, Röstlund, T., B. Albrektsson, T. Albrektsson and P. Brånemark, "Osseointegration of titanium implants.," Acta Orthop Scand , vol. 57, no. 4, p. 285, 1986.

[37] L. Le Guéhennec, A. Soueidan, P. Layrolle and Y. Amouriq, "Surface treatments of titanium dental implants for rapid osseointegration," Dental Material , vol. 23, no. 7, pp. 844-854, 2007.

[38] J. Emsley, Nature's Building Blocks: An A-Z Guide to the Elements, Oxford: Oxford University Press, 2001.

[39] V. V. Tsvetkov, "Corrosion-resistant titanium alloys," Pharmaceutical Chemistry Journal, vol. 29, no. 8, pp. 564 - 566, 1995.

[40] P. L. Drake and K. J. Hazelwood, "Exposure-Related Health Effects of Silver and Silver Compounds: A Review," The Annals of Occupational Hygiene, vol. 49, no. 7, pp. 575-585, 2005.

[41] J. Rungby, "An experimental study on silver in the nervous system and on aspects of its general cellular toxicity.," Danish Medical Bulletin, vol. 37, no. 5, pp. 442-449, 1990.

[42] Harris Williams & Co., "Orthopedic Implants—A Global Market Overview," Industry Experts, Hyderabad, India, 2011.

147

[43] A. Virk and D. Osmon, "Prosthetic joint infection," Current Treatment Options in Infectious Diseases, vol. 3, p. 287–300, 2001.

[44] A. Widmer, "New developments in diagnosis and treatment of infection in orthopedic implants," Clinical Infectious Diseases, vol. 33, no. Suppl. 2, p. S94–S106, 2001.

[45] C. Cyteval and A. Bourdon, "Imaging orthopedic implant infections," Diagnostic and Interventional Imaging, vol. 93, no. 6, p. 547–557, 2012.

[46] E. Berbari, A. Hanssen, M. Duffy, J. Steckelberg, D. Ilstrup, W. Harmsen and D. Osmon, "Risk factors for prosthetic joint infection: case-control study," Clinical Infectious Diseases, vol. 27, no. 5, pp. 1247-1254, 1998.

[47] D. Campoccia, L. Montanaro and C. R. Arciola, "The significance of infection related to orthopedic devices and issues of antibiotic resistance," Biomaterials, vol. 27, no. 11, p. 2331–2339, 2006.

[48] J. Wiedel, "Salvage of infected total knee fusion: the last option," Clinical Orthopedic Related Research, vol. 404, p. 139–142, 2002.

[49] M. Aliofkhazraei, "Modern Orthopaedic Implant Coatings — Their Pro’s, Con’s and Evaluation Methods," in Modern Surface Engineering Treatments, Amazon Digital Services, Inc, 2013.

[50] R. Darouiche, "Treatment of infections associated with surgical implants," The New England Journal of Medicine , vol. 350, pp. 1422-1429, 2004.

[51] K. Ong, S. Kurtz, E. Lau, K. Bozic, D. Berry and J. Parvizi, "Prosthetic joint infection risk after total hip arthroplasty in the Medicare population," The Journal of Arthroplasty, vol. 24, no. 6, p. 105–109, 2009.

[52] W. Zimmerli and P. Ochsner, "Management of Infection Associated with Prosthetic Joints," Infection, vol. 31, no. 2, pp. 99-108, 2003.

[53] H. Goitz, R. Goitz, J. Watson, J. S. 2nd. and H. Roth, "Orthopedic implants: a guide to radiographic analysis," Current problems in diagnostic radiology, vol. 25, no. 4, p. 109–168, 1996.

[54] E. Toubes and J. Segreti, "Treatment options for orthopedic device-related infections," Current Infectious Disease Reports, vol. 4, no. 5, pp. 433-438, 2002.

[55] E. Moran, S. Masters, A. Berendt, P. McLardy-Smith, I. Byren and B. Atkins, "Guiding empirical antibiotic therapy in orthopaedics: the microbiology of prosthetic

148

joint infection managed by debridement, irrigation and prosthesis retention," Journal of Infection, vol. 55, pp. 1-7, 2007.

[56] D. Teterycz, T. Ferry, D. Lew, R. Stern, M. Assal, P. Hoffmeyer, L. Bernard and I. Uçkay, "Outcome of orthopedic implant infections due to different staphylococci," International Journal of Infectious Diseases, vol. 14, no. 10, p. e913–e918, 2010.

[57] J. Cordero, Munuera, L. and M. Folgueira, "Inﬂuence of bacterial strains on bone infection," Journal of orthopaedic research, vol. 14, no. 4, pp. 663-667, 1996.

[58] Y. An and R. Friedmann, "Concise review of mechanisms of bacterial adhesion to biomaterial surfaces," Journal of Biomedical Material Research, vol. 43, p. 338–348, 1998.

[59] A. Gristina, "Implant failure and the immuno-incompetent ﬁbro-inﬂammatory zone," Clinical Orthopedic Related Research, vol. 298, p. 106–118, 1994.

[60] C. Broekhuizen, M. Schultz, A. van der Wal, L. Boszhard, L. de Boer, C. Vandenbroucke-Grauls and S. Zaat, "Tissue around catheters is a niche for bacteria associated with medical device infection," Critical Care Medicine, vol. 36, no. 8, pp. 2395-2402, 2008.

[61] D. Neut, H. van der Mei, S. Bulstra and H. Busscher, "The role of small-colony variants in failure to diagnose and treat biofilm infections in orthopedics," Acta Orthopaedica, vol. 78, no. 3, pp. 299-308, 2007.

[62] S. Ahmed, S. Meghji, R. J. Williams, B. Henderson, J. H. Brock and S. P. Nair, "Staphylococcus aureus fibronectin binding proteins are essential for internalization by osteoblasts but do not account for differences in intracellular levels of bacteria," Infection and immunity, vol. 69, no. 5, p. 2872–2877, 2001.

[63] A. Gristina, "Biomaterial-centered infection: microbial adhesion versus tissue integration.," Science, vol. 237, no. 4822, pp. 1588-1595, 1987.

[64] H. J. Busscher, H. C. Van Der Mei, G. Subbiahdoss, P. C. Jutte, J. van den Dungen, S. Zaat, M. Schultz and D. Grainger, "Biomaterial-associated infection: locating the finish line in the race for the surface," Science Translational Medicine, vol. 4, no. 153, p. 153rv10, 2012.

[65] M. Diefenbeck, T. Mückley and G. O. Hofmann, "Prophylaxis and treatment of implant-related infections by local application of antibiotics," Injury-International Journal of the Care of the Injured, vol. 37, p. S95—S104, 2006.

149

[66] C. Zalavras, M. Patzakis and P. Holtom, "Local antibiotic therapy in the treatment of open fractures and osteomyelitis," Clinical Orthopaedics and Related Research, vol. 427, p. 88−93, 2004.

[67] M. Wannske and H. Tscherne, "Results of prophylactic use of Refobacin-Palacos in implantation of endoprosteses of the hip joint in Hannover," Aktuelle Probleme in Chirurgie und Orthopädie, vol. 12, p. 201−208, 1979.

[68] T. Sorensen, A. Sorensen and S. Merser, "Rapid release of gentamicin from collagen sponge," Acta Orthopaedica Scandinavica, vol. 61, no. 4, p. 353–356, 1990.

[69] J. Scales and W. Herschell, "Perspex (Methylmethacrylate) in orthopaedics," British Medical Journal, vol. 2, p. 245−249, 1945.

[70] H. Buchholz and H. Engelbrecht, "Depot effects of various antibiotics mixed with Palacos resins," Chirurg, vol. 41, no. 11, p. 511−515, 1970.

[71] M. Puska, A. J. Aho and P. Vallittu, "Polymer Composites for Bone Reconstruction," in Advances in Composite Materials - Analysis of Natural and Man-Made Materials, InTech, 2011, pp. 55-68.

[72] M. Chohfi, F. Langlais, J. Fourastier, J. Minet, H. Thomazeau and M. Cormier, "Pharmacokinetics, uses, and limitations of vancomycin-loaded bone cement," International Orthopaedics, vol. 22, p. 171–177, 1998.

[73] B. Atkins and T. Gottlieb, "Fusidic acid in bone and joint infections," International Journal of Antimicrobial Agents, vol. 12, p. S79–S93, 1999.

[74] D. Cerretani, P. F. G. Giorgi, L. Bocchi, L. Neri, R. Ceffa, F. Ghisellini and M. Ritter, "The in vitro elution characteristics of vancomycin combined with imipenem- cilastatin in acrylic bone-cementsa pharmacokinetic study," The Journal of Arthroplasty, vol. 17, p. 619–626, 2002.

[75] J. Hendriks, J. van Horn, H. van der Mei and H. Busscher, "Backgrounds of antibiotic-loaded bone cement and prosthesis-related infection," Biomaterials, vol. 25, p. 545–556, 2004.

[76] B. Picknell, L. Mizen and R. Sutherland, "Antibacterial activity of antibiotics in acrylic bone cement," Journal of Bone and Joint Surgery, British Volume, vol. 59, pp. 302-307, 1977.

150

[77] H. Van de Belt, D. Neut, W. Schenk, J. Van Horn, H. Van der Mei and H. Busscher, "Staphylococcus aureus biofilm formation on different gentamicin-loaded polymethylmethacrylate bone cements," Biomaterials, vol. 22, p. 1607–1611, 2001.

[78] M. Virto, P. Frutos, S. Torrado and G. Frutos, "Gentamicin release from modified acrylic bone cements with lactose and hydroxypropylmethylcellulose," Biomaterials, vol. 24, pp. 79-87, 2003.

[79] A. Baker and L. Greenham, "Release of gentamicin from acrylic bone cement. Elution and diffusion studies," Journal of Bone and Joint Surgery, American Volume, vol. 70, pp. 1551-1557, 1988.

[80] V. Krebs, E. S. Yerger, W. K. Barsoum, T. W. Bauer and L. S. Borden, "Treatment of the Infected Total Hip Arthroplasty," Seminars in Arthroplasty, vol. 16, no. 2, pp. 153 - 160, 2006.

[81] R. S. Namba, Y. Chen, E. W. Paxton, T. Slipchenko and D. C. Fithian, "Outcomes of Routine Use of Antibiotic-Loaded Cement in Primary Total Knee Arthroplasty," The Journal of Arthroplasty, vol. 24, no. 6, p. 44–47, 2009.

[82] L. Engesæter, S. A. Lie, B. Espehaug, O. Furnes, S. E. Vollset and L. I. Havelin, "Antibiotic prophylaxis in total hip arthroplasty Effects of antibiotic prophylaxis systemically and in bone cement on the revision rate of 22,170 primary hip replacements followed 0-14 years in the Norwegian Arthroplasty Register," Acta Orthopaedica, vol. 74, no. 6, pp. 644-651, 2003.

[83] N. Dunne, J. Hill, P. Mcafee, K. Todd, R. Kirkpatrick, M. Tunney and S. Patrick, "In vitro study of the efficacy of acrylic bone cement loaded with supplementary amounts of gentamicin: effect on mechanical properties, antibiotic release, and biofilm formation," Acta orthopaedica, vol. 78, no. 6, pp. 774-785, 2007.

[84] R. Kendall, C. Duncan and J. Smith, "Persistence of bacteria on antibiotic loaded acrylic depots: A reason for caution," Clinical Orthopaedics, vol. 329, p. 273−280, 1996.

[85] J. Hendriks, D. Neut, J. van Horn, H. van der Mei and H. Busscher, "Bacterial survival in the interfacial gap in gentamicin-loaded acrylic bone cements," Journal of Bone and Joint Surgery, vol. 87, no. 2, p. 272−276, 2005.

[86] P. Hope, K. Kristinsson and P. Norman, "Deep infection of cemented total hip arthroplasties caused by coagulase-negative staphylococci," Journal of Bone and Joint Surgery - British Volume, vol. 71, p. 851, 1989.

151

[87] E. Phillips and T. Young, "Methicillin-resistant Staphylococcus aureus and wound management," British Journal of Nursing, vol. 4, p. 1345–1349, 1995.

[88] J. Wright, K. Lam and R. E. Burrell, "Wound management in an era of increasing bacterial antibiotic resistance: A role for topical silver treatment," American Journal of Infection Control, vol. 26, no. 6, p. 572–577, 1998.

[89] S. Trippel, "Antibiotic-impregnated cement in total joint arthroplasty," Journal of Bone and Joint Surgery, American Volume, vol. 68, p. 1297–1302, 1986.

[90] M. E. Rupp, T. Fitzgerald, N. Marion, V. Helget, S. Puumala, J. R. Anderson and P. D. Fey, "Effect of silver-coated urinary catheters: Efficacy, cost-effectiveness, and antimicrobial resistance," American Journal of Infection Control, vol. 32, no. 8, p. 445–450, 2004.

[91] J. Fichtner, E. Güresir, V. Seifert and A. Raabe, "Efficacy of silver-bearing external ventricular drainage catheters: a retrospective analysis: Clinical article," Journal of Neurosurgery, vol. 12, no. 4, pp. 840-846, 2010.

[92] J. Wright, K. Lam, D. Hansen and R. Burrell, "Efficacy of topical silver against fungal burn wound pathogens," American Journal of Infection Control, vol. 27, no. 4, pp. 344-350, 1999.

[93] J. Butany, R. L. Leask, N. D. Desai, A. Jegatheeswaran, C. Silversides, H. E. Scully and C. Feindel, "Pathologic analysis of 19 heart valves with silver-coated sewing rings," Journal of cardiac surgery, vol. 21, no. 6, pp. 530-538, 2006.

[94] S. T. Dubas, S. Wacharanad and P. Potiyaraj, "Tunning of the antimicrobial activity of surgical sutures coated with silver nanoparticles," Colloids and Surfaces A: Physicochemical and Engineering Aspects, vol. 380, no. 1, pp. 25-28, 2011.

[95] P. Khalilpour, K. Lampe, M. Wagener, B. Stigler, C. Heiss, M. S. Ullrich, E. Domann, R. Schnettler and V. Alt, "Ag/SiO(x)C(y) plasma polymer coating for antimicrobial protection of fracture fixation devices," Journal of biomedical materials research. Part B, Applied biomaterials, vol. 94, no. 1, p. 196, 2010.

[96] M. McCarthy, "Silver-alloy urinary catheters cut infections costs," The Lancet, vol. 356, no. 9236, p. 1171, 2000.

[97] K. Das, S. Bose, A. Bandyopadhyay, B. Karandikar and B. L. Gibbins, "Implants, Surface coatings for improvement of bone cell materials and antimicrobial activities of Ti," Journal of Biomedical Materials Research Part B: Applied Biomaterials, vol. 87B, no. 2, p. 455–460, 2008.

152

[98] E. D. Giglio, D. Cafagna, S. Cometa, A. Allegretta, A. Pedico, L. C. Giannossa, L. Sabbatini, M. Mattioli-Belmonte and R. Iatta, "An innovative, easily fabricated, silver nanoparticle-based titanium implant coating: development and analytical characterization," Analytical and Bioanalytical Chemistry, vol. 405, no. 2-3, pp. 805- 816, 2013.

[99] H. Cao, X. Liu, F. Meng and P. K. Chu, "Biological actions of silver nanoparticles embedded in titanium controlled by micro-galvanic effects," Biomaterials, vol. 32, no. 3, pp. 693-705, 2011.

[100] H. Qin, H. Cao, Y. Zhao, C. Zhu, T. Cheng, Q. Wang and P. K. Chu, "Anti-biofilm effects of silver nanoparticles immobilized on titanium," Biomaterials, vol. 35, no. 33, pp. 9114-9125, 2014.

[101] Y. Tian, H. Cao, Y. Qiao, F. Meng and X. Liu, "Antibacterial activity and cytocompatibility of titanium oxide coating modified by iron ion implantation," ACTA BIOMATERIALIA, vol. 10, no. 10, pp. 4505-4517, 2014.

[102] H. Kakinuma, K. Ishii, H. Ishihama, M. Honda, Y. Toyama, M. Matsumoto and M. Aizawa, "Antibacterial polyetheretherketone implants immobilized with silver ions based on chelate‐bonding ability of inositol phosphate: Processing, material characterization, cytotoxicity, and antibacterial properties," Journal of Biomedical Materials Research Part A, vol. 103, no. 1, pp. 57-64, 20104.

[103] W. Badiou, J. P. Lavigne, P. J. Bousquet, D. O’Callaghan, P. Marès and R. de Tayrac, "In vitro and in vivo assessment of silver-coated polypropylene mesh to prevent infection in a rat model," International Urogynecology Journal, vol. 22, no. 3, pp. 265-272, 2011.

[104] H. Cheng, Y. Li, K. Huo, B. Gao and W. Xiong, "Long‐lasting in vivo and in vitro antibacterial ability of nanostructured titania coating incorporated with silver nanoparticles," Journal of Biomedical Materials Research Part A, vol. 102, no. 10, pp. 3488-3499, 2014.

[105] S. H. Uhm, S. B. Lee, D. H. Song, J. S. Kwon, J. G. Han and K. N. Kim, "Fabrication of Bioactive, Antibacterial TiO 2 Nanotube Surfaces, Coated with Magnetron Sputtered Ag Nanostructures for Dental Applications.," Journal of Nanoscience and Nanotechnology, vol. 14, no. 10, pp. 7847-7854, 2014.

[106] G. Gosheger, C. Von Eiff, J. Hardes, H. Ahrens, A. Streitburger and H. Buerger, "Silver-coated megaendoprostheses in a rabbit model--an analysis of the infection rate and toxicological side effects," Biomaterials, vol. 25, no. 24, pp. 5547-5556, 2004.

153

[107] implantcast GmbH, "MUTARS® (Modular Universal Tumor And Revision System)," 2011. [Online]. Available: http://www.implantcast.info/index.php?option=com_content&view=category&layout =blog&id=1&Itemid=57&lang=en. [Accessed 18 6 2013].

[108] J. Hardes, C. von Eiff, A. Streitbuerger, M. Balke, T. Budny, M. P. Henrichs, G. Hauschild and H. Ahrens, "Reduction of periprosthetic infection with silver-coated megaprostheses in patients with bone sarcoma," Journal of surgical oncology, vol. 101, no. 5, p. 389, 2010.

[109] W. R. Fordham, S. Redmond, A. Westerland, E. G. Cortes, C. Walker, C. Gallagher and R. R. Krchnavek, "Silver as a Bactericidal Coating for Biomedical Implants," SURFACE & COATINGS TECHNOLOGY, vol. 253, pp. 52-57, 2014.

[110] D. Stinner, S. Waterman, B. Masini and J. Wenke, "Silver dressings augment the ability of negative pressure wound therapy to reduce bacteria in a contaminated open fracture model," J Trauma, vol. 71, no. 1 Suppl, pp. S147-150, 2011.

[111] N. Tran, P. Tran, J. Jarrell, J. Engiles, N. Thomas, M. Young, R. Hayda and C. T. Born, "In vivo caprine model for osteomyelitis and evaluation of biofilm-resistant intramedullary nails," Biomed Res Int, vol. 2013, p. 674378, 2013.

[112] K. Secinti, M. Ayten, G. Kahilogullari, G. Kaygusuz, H. Ugur and A. Attar, "Antibacterial effects of electrically activated vertebral implants," J Clin Neurosci, vol. 15, no. 4, pp. 434-439, 2008.

[113] J. Liu, Z. Wang, F. Liu., A. Kane and R. Hurt, "Chemical Transformations of Nanosilver in Biological Environments," ACS Nano, vol. 6, no. 11, pp. 9887-9899, 2012.

[114] H. Tsuchiya, T. Shirai, H. Nishida, H. Murakami, T. Kabata, N. Yamamoto, K. Watanabe and J. Nakase, "Innovative antimicrobial coating of titanium implants with iodine," Journal of Orthopaedic Science, pp. 595-604, 5 17 2012.

[115] Y. Kuroyanagi, E. Kim and N. Shioya, "Evaluation of a synthetic wound dressing capable of releasing silver sulfadiazine," Journal of Burn Care & Rehabilitation, vol. 12, no. 2, p. 106–115, 1991.

[116] J. White, A. Powell, K. Brady and R. Russell-Jones, "Severe generalized argyria secondary to ingestion of colloidal silver protein," Clinical and Experimental Dermatology, vol. 28, no. 3, p. 254–256, 2003.

154

[117] V. K. Poon and A. Burd, "In vitro cytotoxity of silver: implication for clinical wound care," Burns , vol. 30, p. 140–147, 2004.

[118] N. Hadrup, K. Loeschner, A. Mortensen, A. K. Sharma, K. Qvortrup, E. H. Larsen and H. R. Lam, "The similar neurotoxic effects of nanoparticulate and ionic silver in vivo and in vitro," Neurotoxicology, vol. 33, no. 3, pp. 416-423, 2012.

[119] G. Danscher and L. Locht, "In vivo liberation of silver ions from metallic silver surfaces," Histochem Cell Biol, vol. 133, p. 359–366, 2010.

[120] R. Foldbjerg, P. Olesen, M. Hougaard, D. A. Dang, H. J. Hoffmann and H. Autrup, "PVP-coated silver nanoparticles and silver ions induce reactive oxygen species, apoptosis and necrosis in THP-1 monocytes," Toxicology Letters, vol. 190, no. 28, pp. 156-162, 2009.

[121] J. S. Kim, E. Kuk, K. N. Yu, J. H. Kim, S. J. Park, H. J. Lee, S. H. Kim, Y. K. Park, Y. H. Park, C. Y. Hwang, Y. K. Kim, Y. S. Lee, D. H. Jeong and M. H. Cho, "Antimicrobial effects of silver nanoparticles," Nanomed. Nanotechnol. Biol. Med., vol. 3, pp. 95-101, 2007.

[122] E. Navarro, F. Piccapietra, B. Wagner, F. Marconi, R. Kaegi, N. Odzak, L. Sigg and R. Behra, "Toxicity of Silver Nanoparticles to Chlamydomonas reinhardtii," Environmental Science & Technology, vol. 42, no. 23, pp. 8959-8964, 2008.

[123] S. Hussain, K. Hess, J. Gearhart, K. Geiss and J. Schlager, "In vitro toxicity of nanoparticles in BRL 3A rat liver cells," Toxicology in Vitro, vol. 19, no. 7, pp. 975- 983, 2005.

[124] P. AshaRani, G. Mun, M. Hande and S. Valiyaveettil, "Cytotoxicity and Genotoxicity of Silver Nanoparticles in Human Cells," ACS NANO, vol. 3, no. 2, pp. 279-290, 2009.

[125] V. Hill and D. Pillsbury, Argyria: the pharmacology of silver, Baltimore: Williams and Wilkins Co., 1939.

[126] S. Maitre, K. Jaber, J. Perrot, C. Guy and F. Cambazard, "Increased serum and urinary levels of silver during treatment with topical silver sulfadiazine," Ann Dermatol Venereol, vol. 129, no. 2, p. 217–219, 2002.

[127] A. Brutel de la Riviere, K. Dossche, D. Birnbaum and R. Hacker, "First clinical experience with a mechanical valve with silver coating," The Journal of Heart Valve Disease, vol. 9, no. 1, p. 123–129, 2000.

155

[128] US Environmental Protection Agency, "Integrated Risk Information System: Silver," 31 1 1987. [Online]. Available: http://www.epa.gov/iris/subst/0099.htm. [Accessed 4 4 2013].

[129] "Secondary Drinking Water Regulations: Guidance for Nuisance Chemicals," U.S Environmental Protection Agency, 31 May 2013. [Online]. Available: http://water.epa.gov/drink/contaminants/secondarystandards.cfm. [Accessed 21 11 2013].

[130] US Environmental Protection Agency, "Chapter 3: Exposure Scenario Selection," in RCRA Delisting Technical Support Document, Dallas, 2000, p. 8.

[131] E. Rentz, "Viral Pathogens and Severe Acute Respiratory Syndrome: Oligodynamic Ag+ for Direct Immune Intervention," Journal of Nutritional & Environmental Medicine, vol. 13, no. 2, p. 109, 2003.

[132] Q. L. Feng, J. Wu, G. Q. Chen, F. Z. Cui, T. N. Kim and J. O. Kim, "A mechanistic study of the antibacterial effect of silver ions on Escherichia coli and Staphylococcus aureus," Journal of Biomedical Materials Research Part A, vol. 52, no. 4, pp. 662- 668, 2000.

[133] J. Spadaro, T. Berger, S. Barranco, S. Chapin and R. Becker, "Antibacterial effects of silver electrodes with weak direct current," Antimicrobial agents and chemotherapy, vol. 6, no. 5, pp. 637-642, 1974.

[134] C. P. Davis, N. Wagle, M. D. Anderson and M. M. Warren, "Bacterial and fungal killing by iontophoresis with long-lived electrodes," Antimicrobial Agents and Chemotherapy, vol. 35, no. 10, p. 2131–2134, 1991.

[135] T. J. Berger, J. A. Spadaro, S. E. Chapin and R. O. Becker, "Electrically Generated Silver Ions: Quantitative Effects on Bacterial and Mammalian Cells," Antimicrobial Agents and Chemotherapy, vol. 9, no. 2, pp. 357-358, 1976.

[136] R. O. Becker and J. A. Spadaro, "Treatment of orthopaedic infections with electrically generated silver ions. A preliminary report," The Journal of bone and joint surgery. American volume, vol. 60, no. 7, p. 871, 1978.

[137] D. Webster, J. Spadaro, R. Becker and S. Kramer, "Silver anode treatment of chronic osteomyelitis," Clinical Orthopaedics and Related Research, vol. 161, p. 105–114, 1981.

[138] J. J. Biemer, "Antimicrobial susceptibility testing by the Kirby-Bauer disc diffusion method," Annals of clinical laboratory science, vol. 3, no. 2, pp. 135-140, 1972.

156

[139] R. A. Serway, C. J. Moses and C. A. Moyer, "Chapter 4: The Particle Nature of Matter," in Modern Physics, David Harris, 2004, p. 143.

[140] SAS, "Procedures," in SAS/STAT(R) 9.2 User's Guide, Second Edition, Cary, SAS, 2009.

[141] Y. Shai, "Mode of action of membrane active antimicrobial peptides," Peptide Science, vol. 66, no. 4, pp. 236-248, 2002.

[142] H. Akaike, "A new look at the statistical model identification," IEEE Transactions on Automatic Control , vol. 19, no. 6, p. 716–723, 1974.

[143] K. P. Burnham and D. R. Anderson, "Multimodel inference: understanding AIC and BIC in Model Selection," Sociological Methods and Research, vol. 33, p. 261–304, 2004.

[144] M. Bosetti, A. Massè, E. Tobin and M. Cannas, "Silver coated materials for external fixation devices: in vitro biocompatibility and genotoxicity," Biomaterials, vol. 23, no. 3, pp. 887 - 892, 2002.

[145] D. Czock and F. Keller, "Mechanism-based pharmacokinetic–pharmacodynamic modeling of antimicrobial drug effects," J Pharmacokinet Pharmacodyn , vol. 34, p. 727–751, 2007.

[146] K. Skarstad, H. Steen and E. Boye, "Cell cycle parameters of slowly growing Escherichia coli B/r studied by flow cytometry," Journal of Bacteriology, vol. 154, no. 2, p. 656–662, 1983.

[147] K. Vidal, I. Grosjean, J. Revillard, C. Gespach and D. Kaiserlian, "Immortalization of mouse intestinal epithelial cells by the SV40-large T gene. Phenotypic and immune characterization of the MODE-K cell line," J. Immunol. Methods, vol. 166, p. 63, 1993.

[148] C. Pautke, M. Schieker, T. Tischer, A. Kolk, P. Neth, W. Mutschler and S. Milz, "Characterization of Osteosarcoma Cell Lines MG-63, Saos-2 and U-2 OS in Comparison to Human Osteoblasts," Anticancer Research, vol. 24, pp. 3743-3748, 2004.

[149] B. Necula, J. van Leeuwen, L. Fratila-Apachitei, S. Zaat, I. Apachitei and J. Duszczyk, "In vitro cytotoxicity evaluation of porous TiO2–Ag antibacterial coatings for human fetal osteoblasts," Acta Biomaterialia, vol. 8, no. 11, pp. 1742-7061, 2012.

157

APPENDICES

158

Appendix A: VB Codes in MS Excel®

Private Sub modelfit_Click()

Dim k As Single, delta As Single, k_op As Single, delta_op As Single, x0 As Single

Dim data1 As Single, data2 As Single, data3 As Single, data4 As Single, data5 As Single, data6 As Single, data7 As Single, data8 As Single, datamean As Single

Dim BD0 As Single, BD3 As Single, BD6 As Single, BD9 As Single, BD12 As Single, BD24 As Single, BD36

As Single, BD48 As Single

Dim Rtt As Single, Rer As Single, minRer As Single

data1 = ThisWorkbook.Worksheets("Sheet1").Range("d4").Value data2 = ThisWorkbook.Worksheets("Sheet1").Range("d5").Value data3 = ThisWorkbook.Worksheets("Sheet1").Range("d6").Value data4 = ThisWorkbook.Worksheets("Sheet1").Range("d7").Value data5 = ThisWorkbook.Worksheets("Sheet1").Range("d8").Value data6 = ThisWorkbook.Worksheets("Sheet1").Range("d9").Value data7 = ThisWorkbook.Worksheets("Sheet1").Range("d10").Value data8 = ThisWorkbook.Worksheets("Sheet1").Range("d11").Value x0 = data1

BD0 = x0 datamean = WorksheetFunction.Average(data1, data2, data3, data4, data5, data6, data7, data8)

Rtt = (data1 - datamean) ^ 2 + (data2 - datamean) ^ 2 + (data3 - datamean) ^ 2 + (data4 - datamean) ^ 2 + (data5

- datamean) ^ 2 + (data6 - datamean) ^ 2 + (data6 - datamean) ^ 2 + (data8 - datamean) ^ 2

ThisWorkbook.Worksheets("Sheet1").Range("d14").Value = Rtt minRer = 9999

ThisWorkbook.Worksheets("Sheet1").Range("e4").Value = BD0

ThisWorkbook.Worksheets("Sheet1").Range("e5").Value = BD3 159

ThisWorkbook.Worksheets("Sheet1").Range("e6").Value = BD6

ThisWorkbook.Worksheets("Sheet1").Range("e7").Value = BD9

ThisWorkbook.Worksheets("Sheet1").Range("e8").Value = BD12

k = 0.3

Do While k < 1.3 delta = 0.01

Do While delta < 0.3

If BD0 = 0 Then

BD3 = 0

Else

BD3 = WorksheetFunction.Max(0, 8.89 - WorksheetFunction.Log10(1 + (10 ^ 8.89 / 10 ^ x0 - 1) * _

Exp(-1.35 * 3 + k / Log(1 - delta) * (3 * (1 - delta) ^ 3 - ((1 - delta) ^ 3 - 1) / Log(1 - delta)))))

End If

If BD3 = 0 Then

BD6 = 0

Else

BD6 = WorksheetFunction.Max(0, 8.89 - WorksheetFunction.Log10(1 + (10 ^ 8.89 / 10 ^ x0 - 1) * _

Exp(-1.35 * 6 + k / Log(1 - delta) * (6 * (1 - delta) ^ 6 - ((1 - delta) ^ 6 - 1) / Log(1 - delta)))))

End If

If BD6 = 0 Then

BD9 = 0

Else

BD9 = WorksheetFunction.Max(0, 8.89 - WorksheetFunction.Log10(1 + (10 ^ 8.89 / 10 ^ x0 - 1) * _

Exp(-1.35 * 9 + k / Log(1 - delta) * (9 * (1 - delta) ^ 9 - ((1 - delta) ^ 9 - 1) / Log(1 - delta)))))

End If

If BD9 = 0 Then

160

BD12 = 0

Else

BD12 = WorksheetFunction.Max(0, 8.89 - WorksheetFunction.Log10(1 + (10 ^ 8.89 / 10 ^ x0 - 1) * _

Exp(-1.35 * 12 + k / Log(1 - delta) * (12 * (1 - delta) ^ 12 - ((1 - delta) ^ 12 - 1) / Log(1 - delta)))))

End If

If BD12 = 0 Then

BD24 = 0

Else

BD24 = WorksheetFunction.Max(0, 8.89 - WorksheetFunction.Log10(1 + (10 ^ 8.89 / 10 ^ x0 - 1) * _

Exp(-1.35 * 24 + k / Log(1 - delta) * (24 * (1 - delta) ^ 24 - ((1 - delta) ^ 24 - 1) / Log(1 - delta)))))

End If

If BD24 = 0 Then

BD36 = 0

Else

BD36 = WorksheetFunction.Max(0, 8.89 - WorksheetFunction.Log10(1 + (10 ^ 8.89 / 10 ^ x0 - 1) * _

Exp(-1.35 * 36 + k / Log(1 - delta) * (36 * (1 - delta) ^ 36 - ((1 - delta) ^ 36 - 1) / Log(1 - delta)))))

End If

If BD36 = 0 Then

BD48 = 0

Else

BD48 = WorksheetFunction.Max(0, 8.89 - WorksheetFunction.Log10(1 + (10 ^ 8.89 / 10 ^ x0 - 1) * _

Exp(-1.35 * 48 + k / Log(1 - delta) * (48 * (1 - delta) ^ 48 - ((1 - delta) ^ 48 - 1) / Log(1 - delta)))))

End If

Rer = (data1 - BD0) ^ 2 + (data2 - BD3) ^ 2 + (data3 - BD6) ^ 2 + (data4 - BD9) ^ 2 + (data5 - BD12) ^ 2 +

(data6 - BD24) ^ 2 + (data7 - BD36) ^ 2 + (data8 - BD48) ^ 2

If Rer < minRer Then

161

minRer = Rer k_op = k delta_op = delta

End If

delta = delta + 0.001

Loop k = k + 0.01

Loop

ThisWorkbook.Worksheets("Sheet1").Range("e14").Value = minRer

ThisWorkbook.Worksheets("Sheet1").Range("f14").Value = 1 - minRer / Rtt

End Sub

162

Appendix B: Table of Abbreviations

AEP antimicrobial efficacy parameters Ag+ silver ion AIC Akaike information criterion ALBC antibiotic loaded bone cements ASA anode surface area C. albicans Candida albicans CCC circumscribed central composite CCD central composite design CI current intensity CM cathode material CoNS Coagulase-negative staphylococci CRP C-reactive-protein DIS dual-metal implant system DZ diffusion zone E. coli Escherichia coli E. faecalis Enterococcus faecalis EDF empirical distribution function EPA Environmental Protection Agency ESD electrode separation distance FDA Food and Drug Administration IZ inhibition zone LIDC low intensity direct current MAE maximal antimicrobial efficacy MH Mueller Hinton MMA methyl methacrylate MRSA methicillin-resistant S. aureus MRSE Methicillin-resistant Staphylococcus epidermidis MSSA methicillin-sensitive Staphylococci aureus NHOst Normal human osteoblast cell

163

P. aeruginosa Pseudomonas aeruginosa P. mirabilis Proteus mirabilis PBS phosphate buffered saline PMMA polymethylmethacrylate PTFE Polytetrafluoroethylene

QE quantity of electric charge RfD reference dose S. aureus Staphylococcus aureus SCD surface current density SMCL secondary maximum contaminant levels SSI surgical site infection THA total hip arthroplasty TJA total joint arthroplasty TKA total knee arthroplasties

164