Transdermal Chronopharmaceutical Drug Delivery: Microneedles, Intradermal Infusion Experiments and a Delivery Device

Michael Vosseler

Transdermal chronopharmaceutical drug delivery: microneedles, intradermal infusion experiments and a delivery device

Dissertation zur Erlangung des Doktorgrades der Technischen Fakultät der Albert-Ludwigs-Universität Freiburg im Breisgau

April 3rd 2013

Dekan Prof. Dr. Yiannos Manoli

Gutachter Prof. Dr. Roland Zengerle Prof. Dr. Holger Reinecke

Tag der Prüfung 05.02.2014

Michael Vosseler Hahn-Schickard-Gesellschaft für Angewandte Forschung e.V. Institut für Mikro- und Informationstechnik (HSG-IMIT) Villingen-Schwenningen

Erklärung

Ich erkläre, dass ich die vorliegende Arbeit ohne unzulässige Hilfe Dritter und ohne Benutzung anderer als der angegebenen Hilfsmittel angefertigt habe. Die aus anderen Quellen direkt oder indirekt übernommenen Daten und Konzepte sind unter An- gabe der Quelle gekennzeichnet. Insbesondere habe ich hierfür nicht die entgeltliche Hilfe von Vermittlungs- oder Beratungsdiensten (Promotionsberaterinnen oder Pro- motionsberater oder anderer Personen) in Anspruch genommen. Niemand hat von mir unmittelbar oder mittelbar geldwerte Leistungen für Arbeiten erhalten, die im Zu- sammenhang mit dem Inhalt der vorgelegten Dissertation stehen. Die Arbeit wurde bisher weder im In- noch im Ausland in gleicher oder ähnlicher Form einer anderen Prüfungsbehörde vorgelegt.

Ich erkläre hiermit, dass ich mich noch nie an einer in- oder ausländischen wissen- schaftlichen Hochschule um die Promotion beworben habe oder gleichzeitig be- werbe.

Datum/date: Unterschrift/signature:

III Abstract

The beginning of modern chronotherapy was the delivery of cancer medication according to circadian (about 24 h) rhythms in the 70s and 80s of the last century. Nowadays, there is a lot of knowledge on rhythms of medical conditions and especially on the exacerbation of symptoms with time of day. Chronic conditions ranging from asthma to stroke are studied intensively. For a patient suffering from such a disease it is important to get the appropriate amount of his drug at the right time. Typically, this is an outpatient setting. In this case pills are the most accepted delivery technology. However, drug absorption from the gastro intestinal tract is site specific. This is a challenge for the development of time controlled release formulations. Minimal invasive transdermal time controlled drug delivery could be a potential solution for this issue. A new approach was developed within this thesis.

Transdermal drug delivery is a discipline that generated impressive success stories in the last two decades. Drug loaded patches enable to deliver a drug without first pass effect and other gastro intestinal adverse effects. However, there are only approximately 20 drug molecules that can be absorbed across the human skin. The reason is the barrier properties of the upper skin layer. One way to circumvent this barrier is the application of microneedles. Hollow microneedles enable the time controlled delivery of drug solutions via the skin to the patient. Fabrication of such needles is a challenging task. Within this thesis two approaches were investigated: microneedles generated by thermoforming processes such as hot embossing and injection moulding as well as steel cannulas that were inserted via a small but well defined angle into the skin. Later on the second approach will be named intradermal infusion set.

The generation of the lumen of hollow microneedles by thermoforming processes is quite challenging. It was demonstrated that a soft mould in combination with wires can be used. It was proofed that aspect ratios (lumen diameter to needle length) larger than 10 can be realized easily. The outer diameter of a specific moulded microneedle design is 0.36 mm, the inner diameter is 0.13 mm and its length is 0.60 mm. The microneedle is placed on a pedestal with a height of 0.80 mm. The aspect ratio is computed with the height of the microneedle, the height of the pedestal and the lumen diameter. Including the base plate with a thickness of 1.00 mm the aspect ratio increases to approximately 20.

The cycle time of the hot embossing process is more than one hour. Fast production is demonstrated by thermoforming with an injection moulding machine. Short fabrication cycles are inherent to this process. However, mould development is quite challenging.

As an alternative approach a cannula was inserted at a very low but well-defined angle of 10° into the skin. The tip of the cannula could reliably be placed approximately 0.70 mm ± 0.08 mm below the skin surface during injection and infusion. Parameters for intradermal infusion and injection processes were characterized, which are necessary for the development and selection of small actuators for skin attachable drug delivery devices. Extensive ex vivo and in vivo experiments with the intradermal infusion set demonstrated the feasibility of this approach. Flow rates ranging from 0.1 ml/h to 180 ml/h were studied. Data on back pressure is presented. It ranges from 10 kPa to 700 kPa. Practical leak tightness of

IV the delivery process was demonstrated, too. This is of outmost importance for the drug developer. It ensures a precise and consistent delivery mechanism.

To realize an electronic and battery free infusion device a hydrogel actuator technology was developed. The design is very simple with less than 5 fabrication steps. If the hydrogel gets in contact with water it starts to swell immediately. Thereby the hydrogel delivers volume work. The influence of the most important design parameters was studied extensively. The dynamics of the actuators is characterized by a fast initial swelling followed by slow continuous increase in volume. Depending on the design parameters a volume of approximately 0.2 ml to 2.0 ml can be displaced within the first 5 min. This is followed by an additional volume displacement of up to 1 ml within the following 4 hours.

Finally a chronotherapeutic drug delivery device, called ChronopaDD was realized. It integrates the intradermal infusion set, the hydrogel actuator and an activation mechanism with time delay without the need for electronics or batteries. Fabrication of the ChronopaDD is based on processes well-known from the packaging industry. The activation mechanism consists of three pouches that can be manufactured from suitable films in modified pouching machines. The drug solution is also stored in a pouch. The time delay of the activation mechanism is realized with a sucrose tablet. After activation of the ChronopaDD the swelling agent must dissolve this tablet before it can enter the hydrogel actuator. Delay times between 30 minutes and 280 minutes were demonstrated. The delay mechanism is a critical step because after dissolution of the sucrose tablet a plug of highly concentrated sucrose solution may be formed. This plug can result in additional undesired delay times. Apart from that the hydrogel actuator is able to displace volume at quite high sucrose concentrations, up to 50% by weight. Of course, its power is reduced in this case. Under well defined conditions the ChronopaDD can deliver e.g. 0.5 ml of liquid in combination with a time delay of 30 min. In another configuration it can deliver a volume of 0.4 ml with a time delay of 280 min. Finally, time delayed intradermal delivery of 0.35 ml liquid with the ChronopaDD was demonstrated ex vivo.

V Zusammenfassung

Die Gabe von Medikamenten gegen Krebs in Abhängigkeit von circadianen (ca. 24 Stunden) Rhythmen in den 70er und 80er Jahren des letzten Jahrhunderts markiert den Beginn der modernen Chronotherapie. Heutzutage ist reichlich Wissen über den rhythmischen Verlauf von Krankheiten und insbesondere über die rhythmisch ausge- prägte Verschlimmerung deren Symptome vorhanden. Chronische Krankheiten von Asthma bis Schlaganfall wurden diesbezüglich intensiv untersucht. Für Patienten, die unter solch einer Krankheit leiden, ist es von Bedeutung, die richtige Menge der richtigen Substanz zur richtigen Zeit zu erhalten. Dabei gilt es die Anforderungen der ambulanten Versorgung zu berücksichtigen. In diesem Fall sind Tabletten die am besten akzeptierte Lösung. Jedoch ist die Wirkstoffaufnahme aus dem Verdauungs- trakt ortsabhängig. Dies ist eine Herausforderung für die Entwicklung von zeitgesteu- erten Verabreichungssystemen. Eine mögliche Lösung dieses Problems ist die mini- malinvasive zeitlich gesteuerte Wirkstoffgabe über die Haut. In dieser Forschungsar- beit wird solch ein Ansatz verfolgt.

Die transdermale Wirkstoffgabe ist eine Disziplin, die in den letzten zwanzig Jahren beeindruckende Erfolge vorweisen konnte. Mit Wirkstoff beladenen Pflastern können Wirkstoffe unter Umgehung der Leber und ohne direkte Nebenwirkungen im Magen- Darm-Trakt verabreicht werden. Aktuell gibt es jedoch lediglich ca. zwanzig Wirk- stoffe, die der Körper über die Haut aufnehmen kann. Der Grund liegt in den Barrie- reeigenschaften der oberen Hautschichten. Ein Weg, diese Barriere zu umgehen, ist der Einsatz von Mikronadeln. Hohle Mikronadeln ermöglichen die zeitlich gesteuerte Verabreichung von Wirkstofflösung über die Haut an einen Patienten. Die Herstellung solcher Nadeln ist jedoch eine Herausforderung. Die Möglichkeit Thermoformpro- zesse wie Heißprägen und Spritzguss dafür zu verwenden wurde in dieser Arbeit ge- prüft. Darüber hinaus wurden intradermale Verabreichungssysteme auf Basis von Stahlkanülen entwickelt.

Die Herstellung des Hohlraums der Mikronadeln im Heißprägeverfahren ist eine große Herausforderung. Es wurde gezeigt, dass eine Form aus Gummi mit eingesetzten Drähten genutzt werden kann. In diesem Fall können Aspektverhältnisse (Länge des Hohlraums zu Durchmesser der Nadel) größer 10 realisiert werden. Dabei beträgt der Durchmesser der Mikronadeln 0,36 mm, der innere Durchmesser 0,13 mm und die Länge 0,60 mm. Die Mikronadel selbst ist auf einer Plattform mit einer Höhe von 0,80 mm platziert. Das Aspektverhältnis wurde mit der Höhe der Nadel, der Höhe der Plattform und dem Innendurchmesser berechnet. Rechnet man die Grundplatte mit einer Dicke von 1,00 mm hinzu steigt das Aspektverhältnis auf ungefähr 20 an.

Die Zykluszeit für diesen Heißprägeprozess ist länger als eine Stunde. Eine schnellere Produktion wurde mit einem Spritzgussprozess gezeigt. Kurze Zykluszeiten liegen in der Natur dieses Prozesses. Die Herstellung der Form ist jedoch sehr aufwendig.

Mit dem intradermalen Verabreichungssystemen wurden Kanülen unter einem kleinen Winkel von 10° in die Haut eingestochen. Dabei befand sich die Spitze der Ka- nüle, während der Injektion bzw. Infusion, ca. 0,70 mm ± 0,08 mm unter der Hautoberfläche.

VI Der nächste Schritt nach der Herstellung von Mikronadeln und intradermalen Verab- reichungssystemen war die Charakterisierung von intradermalen Injektionen und Infusionen mit diesen Systemen. Diese Versuche sind notwendig für die Entwicklung bzw. Auswahl von kleinen Aktoren für Medikamentendosiersysteme, die auf der Haut getragen werden können. Umfangreiche ex vivo und in vivo Versuche mit dem intradermalen Verabreichungssystem bestätigten eine Realisierungsmöglichkeit solcher Medikamentendosiersysteme. Dazu wurden intradermale Infusionen und Injektionen mit Flussraten von 0,1 ml/h bis 180 ml/h durchgeführt. Daten bezüglich des Gegendrucks während der Verabreichung zwischen 10 kPa und 700 kPa wurden erhoben. Darüber hinaus wurde die Dichtigkeit demonstriert. Davon hängt auch die Zuverlässigkeit und Sicherheit über die verabreichte Menge ab die von großer Bedeutung für den Entwickler ist.

Als Antriebselement für ein elektronik- und batteriefreies Infusionssystem wurde eine Hydrogelaktortechnologie entwickelt. Die sehr einfache Konstruktion kann in weniger als 5 Fertigungsschritten hergestellt werden. Der Aktor beginnt sofort zu quellen, wenn er mit Wasser in Berührung kommt. Dabei wird Volumenarbeit geleistet. Die Leistung hängt von der gewählten Konstruktion ab. Daher wurde der Einfluss der wichtigsten Konstruktionsparameter ausführlich untersucht. Charakteristisch für die Dynamik der Aktoren ist eine schnelle Quellphase zu Beginn gefolgt von einem lang- samen kontinuierlichen Quellen. Abhängig von den Designparametern wurde ein Vo- lumen von 0,2 ml bis 2,0 ml innerhalb der ersten 5 Minuten verdrängt. In den folgenden Stunden wurde mit geringerer Geschwindigkeit weiter Volumen, von bis zu 1 ml, verdrängt.

Abschließend wird im Rahmen dieser Arbeit eine als ChronopaDD bezeichnete Komplettlösung eines transdermalen Medikamentendosiersystems präsentiert. Es beinhaltet ein intradermales Verabreichungsset, einen Hydrogelaktor und einen Aktivierungsmechanismus mit Zeitverzögerung in einem elektronik- und batteriefreien System. Die Konstruktion basiert auf Prozessen aus der Verpackungsindustrie. Der Aktivierungsmechanismus besteht aus drei Beuteln, die aus geeigneten Folien auf modifizierten Beutelmaschinen hergestellt werden können. Die Wirkstofflösung wird ebenfalls in einem Beutel bevorratet. Die Zeitverzögerung ist durch eine Zuckertablette realisiert worden. Nach der Aktivierung des ChronopaDD muss das Quellmittel zunächst diese Zuckertablette auflösen, bevor es in den Hydrogelaktor gelangen kann. Dies ist ein kritischer Schritt, da nach der Auflösung der Zuckertablette am Zugang zum Hydrogelaktor eine hochviskose gesättigte Zuckerlösung vorliegt. Dies kann zu einer unerwünschten weiteren Verzögerung der Wirkstoffabgabe führen. Nichts desto trotz ist der Hydrogelaktor in der Lage, bei Zufuhr von Quellmittel mit einer relevanten Glukosekonzentration, von bis zu 50 Gew.-%, Volumenarbeit zu verrichten. Natürlich ist dann die Leistung des Aktors deutlich reduziert. Unter definierten Bedingungen im Labor konnte das ChronopaDD z.B. ein Volumen von 0,5 ml innerhalb von 30 Minuten verabreichen. In einer anderen Konfiguration wurde ein Volumen von 0,4 ml mit einer Verzögerung von 280 min verdrängt. Zuletzt wurde mit dem ChronopaDD die zeitverzögerte intradermal Infusion von 0,35 ml im ex vivo Versuch gezeigt.

VII Publications

Parts of this work have been published in the following journals, conference proceedings and patent publications.

Journal Papers

[P1] Vosseler M, Jugl M, Zengerle R. A smart interface for reliable intradermal injection and infusion of high and low viscosity solutions. Pharm Res. 2011;28:647-661

[P2] Vosseler M, Clemenz M, Zengerle R. A flat and cost effective actuator based on superabsorbent polymer driving a skin attachable drug delivery system. Smart Mater Struct. 21 (2012) 105002.

Reviewed Papers at International Conferences

[P3] Günzler R, Schumacher A, Gronmaier R, Göttsche T, Vosseler M, Kain A, Reiterer M, Hradetzky D, Kohnle J, Müller C, Messner S, Zengerle R. Personal and Chronotherapeutic Drug Delivery: Intraoral, Extracorporeal and Transdermal Devices. IEEE Personalized Health Conference. Porto Carras. Greece 2007

[P4] Vosseler M, Dobmeier M, Blaesing M, Stehr M, Aslanidis G, Zengerle R. Gesteuerte Verdampfung als fluidischer Transportmechanismus in geregelten Medikamentendosiersystemen. Mikrosystemtechnik Kongress. Dresden. Deutschland 2007:473-476

[P5] Vosseler M, Hradetzky D, Zengerle R. Transdermale Medikamentendosiersysteme – Eine Option für die Chronopharmakologie? Biomedizinische Technik. Aachen. Deutschland 2007

[P6] Häberle S, Schumacher A, Gronmaier R, Göttsche T, Vosseler M, Kain A, Reiterer M, Hradetzky D, Müller C, Messner S, Zengerle R. Mobile Drug Delivery for Ambient Assisted Living: Implantable and Extracorporeal Devices. Ambient Assisted Living. Berlin. Deutschland 2008:33-36

[P7] Vosseler M, Kain A, Béchu M, Bauer M, Alabsi A, Hradetzky D, Müller C, Zengerle R. Medical grade plastic microneedles and an evaluation-kit for the characterization of intradermal micro infusion. Perspectives in Percutaneous Penetration, La Grande Motte, France 2008

[P8] Vosseler M, Jugl M, Blaesing M, Hradetzky D, Messner S, Zengerle R. Integration of microneedle-arrays and micro pumps in disposable and cheap drug delivery devices. European Biomedical Engineering Congress. Antwerp. Belgium 2008:2364-2367

[P9] Haeberle S, Hradetzky D, Schumacher A, Vosseler M, Messner S, Zengerle R. Microfluidics for Drug Delivery. Medical Physics and Biomedical World Congress. Munich. Germany 2009;25:359-362

[P10] Messner S, Schumacher A, Vosseler M, Herrlich S, Günzler R, Zengerle R. Mikrotechnische Systeme zur Selbstverabreichung von Medikamenten –

VIII MEMS based systems for self-medication. Ambient Assisted Living. Berlin. Germany 2010

[P11] Botzelmann T, Mayer V, Vosseler M, Spieth S, Kück H. Injection moulded microneedle arrays as an interface to the human vascular system. Multi- Material Micro Manufacture. Stuttgart. Germany 2011;8:59-63

[P12] Vosseler M, Botzelmann T, Schilling P, Spieth S, Messner S, Mayer V, Schmelz M, Kück H, Zengerle R. In vivo and ex vivo intradermal infusion and injection experiments with microneedles. Microneedles. Cork. Ireland. 2012

Patents and Patent Applications1,2

[P13] Schilling P, Botzelmann T, Mayer V, Vosseler M. Nadel, Nadelanordnung, Spritzgussform und Verfahren zum Herstellen. DE 10 2008 052 749 B4. Priority date: 22nd August 2008

[P14] Botzelmann T, Mayer V, Schilling P, Vosseler M. Nadelanordnung, Spitzgussform und Verfahren zum Herstellen. DE 10 2008 052 702 A1. Priority date: 22nd August 2008

[P15] Vosseler M, Hradetzky D, Messner S, Zengerle R. Vorrichtung zur dermalen Verabreichung einer Substanz. DE 10 2008 057 822 B4. Priority date: 18th November 2008

[P16] Vosseler M, Hiltmann K, Jugl M. Dermal Access Device. WO 2011/095494 A1. Also published as: DE 10 2010 001 506 A1, Priority date: 2nd February 2010

[P17] Vosseler M, Swellable substance actuator with an electrically driven fluidic transportation device. WO 2011/161254 A2. Also published as: DE 10 2010 030 504 A1. Priority date: 24th June 2010

[P18] Vosseler M, Swellable substance actuator and delivery device, in particular for a medical active substance. WO 2011/161246 A2. Also published as: DE 10 2010 030 502 A1. Priority date: 24th June 2010

1 Kind codes of the European Patent Office. EP B1: European patent specifications (granted patent). Kind Codes of the World Intellectual Property Organization. WO A1: International application published with international search report, WO A2: International application published without international search report or international application published with declaration under Article 17(2)(a) 2 Schriftenartencodes des Deutschen Patent- und Markenamtes. DE A1: Offenlegungsschrift, DE B4: Patentschrift

IX Content

1. Introduction ...... 1 1.1. Drug delivery ...... 1 1.2. Chronotherapy ...... 4 1.3. Motivation for this thesis ...... 5 1.4. Goal and Structure of this thesis ...... 6

2. Fundamentals ...... 9 2.1. Structure of the skin...... 9 2.2. Drug transport across the skin barrier ...... 10 2.3. Design of experiments (DOE) ...... 12 2.3.1. Concept of DOE...... 12 2.3.2. Range of factor levels ...... 14 2.3.3. Sample size ...... 14 2.3.4. Experimental Design ...... 14 2.3.5. Center Points ...... 15 2.3.6. Randomization and blocking ...... 15 2.3.7. Statistical Analysis ...... 15 2.3.8. Residual Analysis and model adequacy ...... 16 2.4. Superabsorbent polymers (SAPs) ...... 16 2.4.1. Traditional Absorbent Materials ...... 16 2.4.2. Superabsorbent Materials ...... 17

3. State-of-the-art ...... 23 3.1. Microneedles ...... 23 3.1.1. Microneedle arrays made of metal...... 23 3.1.2. Microneedle arrays made of silicon ...... 25 3.1.3. Microneedle arrays made of polymer ...... 27 3.1.4. Other fabrication technologies and other applications ...... 29 3.2. Intradermal drug delivery with microneedles ...... 30 3.2.1. Insertion of microneedle arrays into the skin ...... 30 3.2.2. Operating regimes of microneedle arrays ...... 30 3.2.3. Intradermal injection and infusion ...... 31 3.2.4. Intradermally injected vaccines ...... 32 3.2.5. Intradermal delivery of insulin and other biopharmaceuticals ...... 32

X 3.2.6. Other aspects of drug delivery via microneedles ...... 34 3.3. Drug delivery devices for chronotherapeutic drug delivery ...... 34 3.3.1. Oral chronotherapeutic drug delivery devices ...... 35 3.3.2. Transdermal chronotherapeutic drug delivery devices ...... 36 3.3.3. Programmable implantable Infusionpumps ...... 40

4. Fabrication of hollow microneedles for intradermal injections and infusions ...... 41 4.1. Hot embossing of hollow microneedle arrays ...... 41 4.1.1. Preparation of the hot embossing tool ...... 42 4.1.2. Soft embossing process ...... 45 4.1.3. Results & Discussion ...... 46 4.1.4. Summary ...... 49 4.2. Injection moulding of hollow microneedle arrays ...... 50 4.3. Intradermal infusion set based on the Mantoux method ...... 52 4.3.1. Laboratory setup for ex vivo experiments ...... 52 4.3.2. Infusion set for in vivo experiments...... 54

5. Ex vivo and in vivo intradermal infusion experiments ...... 57 5.1. Ex vivo experiments with the laboratory setup ...... 57 5.1.1. Pig skin ...... 57 5.1.2. Solutions for injection and infusion ...... 58 5.1.3. Fluidic setup ...... 58 5.1.4. Injection and infusion experiments...... 59 5.1.5. Penetration Results ...... 60 5.1.6. Leakage ...... 62 5.1.7. Spot formation ...... 64 5.1.8. Infusion / injection process ...... 68 5.1.9. Relevance of the technology ...... 75 5.1.10. Summary ...... 75 5.2. In vivo experiments with the infusion set ...... 76 5.2.1. Materials & Methods ...... 76 5.2.2. Results ...... 77 5.2.3. Summary ...... 79 5.3. Ex vivo and in vivo experiments with injection moulded microneedle arrays ..... 79 5.3.1. Spring powered applicator ...... 79

XI 5.3.2. Experimental procedure ...... 80 5.3.3. Results ...... 81 5.3.4. Summary ...... 85

6. A chronopharmaceutical transdermal drug delivery device: ChronopaDD 87 6.1. Hydrogel actuator ...... 88 6.1.1. Materials and Methods ...... 89 6.1.2. Results and Discussion ...... 91 6.1.3. Factor Screening ...... 93 6.1.4. Summary ...... 99 6.2. The ChronopaDD system ...... 99 6.2.1. Components of the ChronopaDD and device assembly ...... 101 6.2.2. Characterization of SAP actuators with sucrose solution ...... 107 6.2.3. Characterization of SAP actuator powered intradermal infusions .... 108 6.2.4. Characterization of the ChronopaDD ...... 110 6.2.5. Summary ...... 113

7. Conclusions ...... 115

References ...... 117

Abbreviations ...... 125

Appendix A – Model diagnostics for chapter 6.1 ...... 126

Appendix B – Temperature of the hot embossing tool ...... 128

Appendix C – Applicator ...... 129

Danksagung ...... 132

XII 1. Introduction

Scientific facts presented in chapter 1.1 and 1.2 are usually of general nature. Specific scientific facts are referenced with a relevant publication. The contribution of the author to these chapters is the compilation of the relevant information with respect to the context of this thesis. The author is not and does not claim to be the discoverer, inventor or originating scientist of any scientific fact in chapter 1.1 and 1.2. 1.1. Drug delivery

Drug delivery is a challenging task. The correct amount of the appropriate drug must get to the proper organ at the right time [1]. Many current drug therapies are not optimised according to all these requirements.

To determine the appropriate drug for a certain disease is the task of the physician. Of course, provisions must be made in order that the patient always receives exactly this drug. Such a simple task can be challenging e.g. in homes for the elderly.

There are two major factors influencing the correct amount of the drug. The first one is the physician who needs to determine the dosage. Typically unknown inter-individual differences can make this task awkward for all involved parties. Some individuals show strong side effects while others don’t respond to the same drug. The field of research that analyses these effects and develops potential solutions is called per- sonalised medicine. One goal is to develop diagnostic tests that result in adequate predictions of the correct amount of drug. The second source for variations of the correct amount of drug can be the patient. At home patients might not take their medicine (“drug holidays”) or they might even take too much.

Many drug therapies do not discriminate diseased organs form healthy organs. This is caused by the limited number of routes for a substance into the body. The most popular route is the oral route. After swallowing a pill the drug substance is taken up by the blood vessels of the digestive system. Next, the blood transports the drug to the liver where a first stage of metabolisation takes place (“1st pass metabolism”). Finally, the amount of drug left in the blood stream is distributed in the whole body (systemic drug delivery). This unspecific distribution in the body can cause systemic side effects that may lead to the abort of the otherwise promising therapy. The science of dealing with this issue is called drug targeting. The goal is to develop technologies for the selective accumulation of the drug at the diseased organ.

For many years the philosophy of designing drug therapies was to obtain constant blood plasma levels of the drug. In the 70s and 80s of the last century doctors started to give cancer medications according to circadian (approximately 24 h) rhythms [2]. This is the beginning of a change in the generally accepted philosophy at that time and the onset of modern chronopharmacology.

1 In general, drug delivery or pharmaceutical technology is a multidisciplinary field of research. Traditional approaches range from chemistry to engineering and new technologies e.g. nanotechnology and micro technology provide additional solutions.

There are several routes for a drug to get into the body (Table 1). The transdermal route in its general sense is an umbrella term for many different more specific terms. Nowadays, it is strongly related to transdermal therapeutic systems (TTS). These devices consist of thin films loaded with a specific drug. They stick to the skin of a patient and deliver their content via the top skin layers into the body.

Another important aspect of drug delivery is the intended action at the delivery site. Topical drug delivery refers to a local therapy where the drug can be applied direct to the affected site (inhalation, eye drops, creams, ointments). In this case the site of delivery is also the site of action.

Enteral drug delivery refers to oral technologies (pill) that are usually supposed to result in a systemic effect. In this case the site of delivery is just an interstation before the drug gets in to the systemic circulation. Parenteral drug delivery refers to approaches that circumvent the enteral route usually to reach the systemic circulation (cannula).

Pills are the most common drug delivery technology. Indeed, patients prefer drug therapies based on a once a day pill to be swallowed in the morning. It is discreet, non-invasive and usually an easy task for the patient. Usually, there are no practical limitations regarding the maximum amount of drug that can be delivered. However, the gastrointestinal (GI) tract is a complicated organ with varying sites (stomach to rectum) with very different drug absorption rates [3]. The GI tract environment may even destroy the drug molecules (especially biological molecules such as antibod- ies). Additionally, GI tract transit times can vary [4, 5, 6]. Therefore, other routes of delivery must be chosen quite often during development of a pharmaceutical product. Nevertheless, oral chronopharmaceutical drug delivery technologies can be efficient.

Table 1. List of important routes of administration for drugs. The table gives the anatomic sites and typical technologies together with the terms of the routes.

Route Anatomic site Typical technology subcutaneous fatty tissue of the skin cannula transdermal intradermal top skin layer microneedles intravenous vein cannula oral gastrointestinal (GI) tract pill nasal respiratory tract spray ocular eye eye drops buccal/sublingual oral cavity chewing gum / lolly vaginal vagina suppository rectal rectum suppository intrathekal into the spinal canal implantable infusion pump

2 Intravenous drug delivery via cannulas or peripheral venous catheters is very common in a clinical environment. There is no faster access for a drug, dissolved in an isotonic solution, into the systemic blood circulation. There is no first pass metabolism and effects can occur very fast. It is possible to give a bolus via a syringe or slowly infuse a drug solution. Time varying infusion rates can be easily generated with infusion pumps. Hence, chronopharmaceutical intravenous drug delivery can be realised easily e.g. with peripheral intravenous catheters. There is usually no practical restriction on the volume or the amount of drug that can be delivered. The first drawback of this approach is the infusion line to a rack that carries the infusion pump. The patient either must stay in bed or he needs to carry the pump with him. The second drawback is a high risk for infections. Peripheral venous catheters require a lot of nursing that cannot be realised at home.

Subcutaneous drug delivery can be performed with syringes or subcutaneous infusion sets. The risk of infection is reduced compared to intravenous drug delivery via peripheral venous catheters. Therefore subcutaneous infusions can be performed at home. The drug transport mechanism from the subcutaneous space to the blood capillaries is diffusion. Hence, the resorption of the drug into the systemic circulation is delayed. Lipophilic molecules tend to form a depot in this fatty tissue and biomolecules might be too large for practical diffusion rates. The capacity of the subcutaneous tissue to absorb fluid is limited. Additionally, high drug concentrations may result in skin irritations. Therefore, the amount of drug that can be delivered via the subcutaneous route is limited. Chronopharmaceutcial drug delivery via the subcutaneous rote is an option. It is already realised for type I diabetics that use insulin infusion pumps.

Intradermal drug delivery is even less invasive than subcutaneous drug delivery. So, there is potential for a further reduced risk of infection. The diffusion distance for the drug to get into the systemic circulation is reduced. Hence, there is potentially better drug absorption. However, the amount of drug that can be absorbed is probably smaller because of the smaller volume of the intradermal compartment compared to the subcutaneous compartment. It should be an interesting option for chronopharmaceutical drug delivery. However, there is currently no intradermal infusion set available on the market.

The technology for chronopharmaceutical drug delivery to the intrathekal space via programmable implantable infusion pumps is state-of-the-art. However, infections in the intrathekal space can lead to disastrous conditions. Therefore, it is usually ad- dressed only in case of pain and spasticity where no other option is at hand. Never- theless, programmable implantable pumps can be combined with catheters that reach into other spaces of the human body, e.g. a central venous catheter. In this case all the benefits of intravenous delivery can be realised with a low risk of infection. However, surgery is necessary to place the pump inside the human body. The amount of drug is restricted to the volume of the device. Only small infusion rates can be realised if acceptable refill visits are favoured. Additionally, drug solutions might not be stable at body temperature.

The nasal, ocular, vaginal and rectal routes are usually not considered for chronotherapeutic drug delivery. Devices for these routes would interfere too much with ev- eryday life. The buccal route is an interesting option if one or two molar teeth can be replaced with a drug delivery device. Such a device can be programmed for chronotherapeutic drug delivery. However, the device must be able to cope with tempera-

3 ture changes and taste masking of the drug might be a challenge. Additionally, the amount of the drug is limited to the reservoir volume of the consequently very small device. This can be compensated by frequent device changes that are possible in this case.

The oral, the buccal, the subcutaneous and the intradermal route are promising approaches for the development of chronotherapeutic drug delivery devices. The oral route with its issues related to chronopharmacology is described in more detail in the next chapter. The buccal route is very challenging due to the limited space in the oral cavity. Limitations of the subcutaneous route (diffusion distance, potential higher risk for infections) can be improved by using the intradermal route. Therefore, the intradermal route is investigated in detail in this work. 1.2. Chronotherapy

Chronopharmacology is the study of pharmacokinetics and pharmacodynamics with respect to biological rhythms. In man the periods of rhythms range from fractions of a second up to one year. Of all these rhythms circadian (about a day) rhythms got a lot of attention in pharmacology. Predictable circadian variations in the intensity of symptoms of many diseases are well documented [7, 8]. Allergic rhinitis, arthritis, asthma, myocardial infarction, congestive heart failure, stroke and peptic ulcer are examples of some diseases with symptoms worst in the early morning hours. Other conditions show symptoms worst at other times of the day. So, the highest plasma levels of drugs should be achieved during periods with exacerbation of symptoms. During periods with mild or even without symptoms the plasma level of drugs can be lowered to reduce the negative outcome of side effects. In this way the therapeutic effect of the drug is optimised.

A once-a-day oral formulation taken in the morning is considered the most accepted treatment for patients with chronic conditions. However, absorption of a therapeutic entity is not constant during the transit through the gastrointestinal tract [9]. The stomach can be an attractive absorption site for drugs stable in its acidic environment. Other drugs need to be protected by an enteric coating that dissolves at higher pH levels. The small intestine is the main absorption site because of its large surface area [10]. The colon shows very low absorption especially for lipophilic drugs and in particular those with low permeability. Nevertheless, substantial absorption especially of hydrophilic drugs may occur.

The residence time of drugs in the stomach is approximately 0.5 h to 1 h in the fasted state [11] (Figure 1.1). The small intestine transit time is approximately 3 h. Gener- ally, residence times in the stomach and the small intestine are highly variable. Transit through the colon on average is 20 h but it can be as long as 72 h.

Figure 1.1. Drug absorption of a fictitious drug in the GI tract. The night time dose is taken at 11 pm. Drug absorption is practically stopped between 2 am an 3 am when the pill is transferred to the colon. The plasma concentration is below the therapeutic window from 5 am on. The therapy is stopped. (This diagram is based on pooled data from 1985 to 2003 of Pharma- ceutical Profiles Ltd. acquired by Quotient Bioresearch Ltd.) 1.3. Motivation for this thesis

The gastrointestinal transit times, site specific absorption and drug elimination times need to be considered in the design of chronopharmaceuticals. The ideal concept of a once-a-day formulation taken in the morning is an extremely challenging task for some diseases. Especially when symptoms are worse in the early morning hours there is need for a drug delivery device that generates a peak plasma level at this time. At least there are oral drug delivery technologies that expel their content with a time delay of 4 h to 5 h. These systems are useful in the delivery of drugs with adequate colonic absorption. Taken at bedtime they can be used in treating symptoms worse in the early morning.

Apart from the very basic challenges related to gastrointestinal transit times and preferred absorption in the small intestine many other issues need to be considered. For some drugs there are even shorter sections for absorption in the small intestine [12]. Other major issues are extensive first pass metabolism and gastrointestinal adverse effects. Therefore, other delivery routes are considered for chronopharmaceutical drug delivery.

The skin is an attractive place for a small and flat chronopharmaceutical drug delivery device. It is easily accessible. If necessary, the device can be removed at any time in contrast to swallowed devices. Many different skin sites can be selected to attach such a device. So, discretion is warranted.

Hydrophilic drugs are suitable for the design of pills with delay coatings. Released after several hours they can be absorbed in the colon. This approach is not feasible with lipophilic drugs.

Time controlled delivery of lipophilic drugs via the subcutaneous fatty tissue is prone to undesired and variable delay times. Interactions between the lipophilic molecule and the fatty tissue can occur. Such an uncertainty in delay time is not acceptable for chronopharmaceutical drug delivery devices that are supposed to generate a relevant plasma concentration in a short time. Therefore, the intradermal route is to be preferred to the subcutaneous route.

5 1.4. Goal and Structure of this thesis

The goal of this work, motivated in the previous chapter, is the realisation of: an extracorporeal chronotherapeutic drug delivery device to be worn on the skin with an intradermal infusion set (Figure 1.2). Due to the precise time delay of the technology and the potentially fast drug absorption from the intradermal compartment it is possible to develop promising chronotherapeutic therapies. The design of the device is especially intended to solve the night time dosing problem. Drugs that are eliminated fast usually show very low plasma levels in the early morning hours (Figure 1.1). This is an issue with medical conditions that show sever symptoms in the early morning hours.

The chronopharmaceutical drug delivery device (ChronopaDD) developed in this work is supposed to be activated and attached to the skin at bedtime. After a certain delay time the device starts to infuse a drug solution into the intradermal compartment of the skin. From this location the drug is absorbed into the systemic circulation and the patient is treated when it is most essential.

Fundamentals required for the understanding of this work are dealt with in the second chapter. It focuses on the structure of the skin and drug transport across the skin barrier. The actuator technology of the ChronopaDD is characterised with the formal procedure of designed experiments. Additionally, there is a condensed survey of this method. The actuator contains superabsorbent polymers. Therefore, there is also a short overview of these polymers.

The intradermal infusion set is a building block of the ChronopaDD. However, such infusion sets are currently not available on the market. Although, various technologies especially microneedles are developed at university and industry labs there is only little information on the intradermal infusion processes. Chapter 3 gives an overview of the state of the art in microneedles and intradermal infusion processes. Addi- tionally, there is a review of various technologies that can be used for chronotherapeutic drug delivery.

Chapter 4 focuses on the fabrication of the intradermal infusion set. Fabrication technologies like hot embossing, injection moulding and assembling of steel cannulas are investigated.

Figure 1.2. The ChronopaDD, a chronopharmaceutical drug delivery device, infuses the drug via an intradermal infusion set into the body.

6 Chapter 5 gives information on the intradermal infusion process with the devices introduced in chapter 4. Ex vivo and in vivo experiments were performed to get an adequate data set.

In chapter 6 the actuator technology based on superabsorbent polymers is investigated before the design of the ChronpaDD is presented. Characterisation of the ChronopaDD includes basic data as well as ex vivo experiments.

This thesis finishes with conclusions in chapter 7.

8 2. Fundamentals

The fundamentals presented in chapter 2 are necessary to understand the context, the specific experiments and the conclusions of this thesis. These fundamentals are collected from various sources. The sources are listed in the chapter Reference at the end of the text and are referenced in the text. Some fundamentals are taken from sources that are monographs. The monographs are usually referenced just once. The contribution of the author to this chapter is the compilation of the relevant information in a readable way. The author is not and does not claim to be the discoverer, inventor or originating scientist of any scientific fact in chapter 2. 2.1. Structure of the skin

The skin is the largest organ of man [13]. It covers a surface of approximately 2 m². The weight of the skin including the subcutaneous tissue is about 20 kg. It is a multi- function organ. One important task is to protect the innards from physical, chemical and biological stress. It prevents from excessive water loss. It is used by the body to dissipate thermal energy or to prevent heat loss. Finally, it provides sensation.

The skin consists of several layers. The outer layer is the epidermis (Figure 2.1). Be- low is the dermis. The inner layer is the hypodermis (subcutaneous fatty tissue).

The epidermis consists of 5 sublayers with a total thickness of approximately 200 µm [15, 16, 17]. It can vary from 80 µm to 1400 µm depending on the site of the body. The absolute outer layer is the stratum corneum. Its thickness varies between 10 µm and 40 µm. It consists of dead cells (corneocytes) with lipids in between. The cells and lipids are arranged in a brick and mortar fashion. Due to this structure the stratum corneum is the main barrier against chemicals and pathogens. Additionally, it prevents evaporation of body fluids. The stratum basale is the bottom layer of the epidermis. Keratinocytes proliferate from this layer. They travel through the other sublayers of the epidermis in a 14 day journey. After their cell death the keratinocytes are included as corneocytes in the stratum corneum. Finally, they are separated from the body by desquamation. The epidermis is a layer without vessels. Nutrients diffuse from the capillary loops just below the stratum basale into the epidermis. Besides a few other specific cells Langerhans cells are included into the epidermis. These cells are part of the immune system of the human body.

Figure 2.1. Structure of the human skin [14].

The dermal layer is much more complicated than the epidermis. The most important structures in the dermis are: capillary loops, sebaceous glands, sweat glands, hair follicles, lymph vessels, nerve fibres and blood vessels. The top capillary loops can be found in a papillary layer. The large surface area of this structure reduces the diffusion length of nutrients into the epidermis and increases the adhesive strength between epidermis and dermis. Below this papillary region there is the reticular region of the dermis. The structures already mentioned are embedded into a matrix of connective tissue. It consists of collagen fibres. These elastic and cross linked (reticular) fibres result in the mechanical strength, extensibility and elasticity of the dermal layer. The thickness of the dermal layer varies from 300 µm to 2400 µm.

The hypodermis (subcutaneous fatty tissue) is the layer below the dermis. Its main components are fibroblasts, macrophages and adipocytes. Fibroblasts produce collagen fibres and play a critical role in wound healing. Macrophages digest pathogens. Adipocytes are also known as fat cells. The hypodermis contains roughly 50% of the fat of man. This layer is a thermal insulator and prevents from heat loss.

The dermis and hypodermis contain a complex blood vessel system (Figure 2.2). It is structured in layers. The layer with the finest capillaries is found in the papillaries between the epidermis and the dermis. With increasing depth the diameter and the distance between the capillaries in a layer increases. Vertical vessels connect the different layers. 2.2. Drug transport across the skin barrier

There are two different motivations for drug delivery to the skin. The first one is local drug delivery in order to treat medical conditions of the skin. The second one is systemic drug delivery in order to increase the blood plasma level of a drug.

Local drug delivery to the skin is common with skin diseases. In skin diseases the skin barrier function is aggrieved. Hence, drug molecules are mixed into creams and ointments. Drug uptake into the diseased skin layer takes place by diffusion from these dosage forms.

Healthy skin is impermeable to most molecules. In fact, one purpose is to protect man from chemical and biological entities. However, in case it is necessary to deliver a certain drug to the body this feature is objectionable.

Figure 2.2. Hierarchical structure of the dermal capillaries [18].

The intracellular (between the corneocytes) route along the lipid matrix is a potential route for drug molecules into the human body [19]. However, several specific molecular properties must be fulfilled in order to establish a relevant flow of molecules into the skin. The molecular mass should be less than 600 Da. The solubility in oil and water should be adequate. Hence, the octanol / water partition coefficient should be neither too low nor too high. The partition coefficient of solute between membrane and donor solution should be high but not too high. If it is too high the solute will stay in the membrane instead of trespassing into the circulation.

The skin is easily accessible. Therefore, it is of interest as site for drug delivery. Con- sequently, several non-invasive or minimal invasive transdermal drug delivery technologies have been studied for decades. However, several modern technologies boosted the development in recent years [20, 21].

The application of skin barrier modulating substances is quite old. By chemically dis- rupting the lipid matrix between the corneocytes the permeability of the skin can be increased. However, very effective substances usually result in skin irritations. Therefore, penetration enhancers are limited in increasing drug flux across the skin barrier.

More potent approaches require the employment of technology. Ultrasound can be used to disturb the organisation of the lipid matrix in the stratum corneum by cavita- tion. With low frequency ( 20 kHz) ultrasound an increase in drug flux up to a thou- sand fold can be reached. A drawback of ultrasound technology is the size of the necessary ultrasound head. Probably, it will be limited to ambulatory usage.

Iontophoresis uses an electric current to increase drug transport into the skin. The current travels along a path composed of a drug loaded electrode, the skin and a counter electrode. Of course, both electrodes are in contact with the skin. The positively charged drug molecules travel in opposite direction to the electrons. Uncharged molecules may be transported into the skin by electroosmosis. Most of the current passes through low resistance routes like hair follicles. So, the current density in these structures can be quite high potentially resulting in damage. Another problem is the ratio of dose to transported charges that varies from patient to patient. Closed loop approaches might solve this problem.

Apart from all these non-invasive technologies minimal invasive methods promise to increase the drug flux independent of drug properties. Electroporation generates

11 aqueous pores in the lipid layer by the application of short electric pulses. Thermo- poration is another approach to generate channels with similar properties. In this case thermal energy is used to burn tiny holes into the stratum corneum. In both cases a drug loaded patch is applied to the treated skin. The molecules are transported by diffusion through the new generated channels.

Another minimal invasive approach to generate pores in the stratum corneum is the application of microneedle arrays. Typical dimensions of microneedles are: length below 1 mm and diameter below 0.3 mm. There are various fabrication technologies to manufacture needles from different materials. Microneedles made of metal, polymer or silicon are common. Microneedles offer the design of quite different dosage forms. The first option is to use a microneedle array to generate holes in the skin. After application and removal of the microneedle array a drug loaded patch can be applied to the treated skin. This process is comparable to the processes based on electroporation and thermoporation. The second option is to coat the microneedles with a drug of interest. Such microneedle arrays would stay a certain time in the skin to allow the molecules to diffuse into the skin. A little different approach is to manufacture microneedles from materials that dissolve in the human body. During manufacturing the microneedle material can be mixed with the drug of interest. After application to the skin the microneedles dissolve and the drug is released. The last option is to use hollow microneedles. Liquid drug solutions can be injected or infused with such microneedles. This technology offers the easiest method to deliver drugs in a discontinuous way. For such a purpose the hollow microneedle arrays need to be combined with a programmable infusion device. 2.3. Design of experiments (DOE) 2.3.1. Concept of DOE

Just for the benefit of the reader the chapter 2.3 is compiled from [22]. The author is not and does not claim to be the discoverer, inventor or originating scientist of any scientific fact in this chapter. The only purpose of this chapter is to brush up the fundamentals of designed experiments.

The rationale of a designed experiment is to gain information in an efficient manner. It is based on an empirical model of a process or system usually associated with variation. The process or system can be people, plants, animals, factories, objects, materials, etc. A designed experiment can be performed with already existing data or data that is going to be obtained in a series of trials. Therefore, DOE is a powerful tool in natural science, social science and engineering.

The empirical model of a process is visualized in Figure 2.3. The process transforms some input into an output that has one or more observable response variables y. Some of the process variables x1, x2, … xp are controllable, whereas other variables z1, z2, … zq are uncontrollable.

Figure 2.3. General model of a process or system. (Source: Figure 1.1 of [22])

The objectives of the experiment may include the following: 1. Determining which variables are most influential on the response y 2. Determining where to set the influential x’s so that y is almost always near the desired nominal value 3. Determining where to set the influential x’s so that variability in y is small 4. Determining where to set the influential x’s so that the effects of the uncontrollable variables z1, z2, …zq are minimized.

One strategy to get information on a process is to change one factor and study its influence on the output y. The drawback of this approach is that interactions between factors cannot be identified. The correct approach is to do factorial experiments. In these experiments several controllable factors are changed at the same time. The design of such an experiment follows a 7 step protocol: 1. Recognition and statement of the problem There are many possible objectives including: optimization, confirmation, dis- covery, stability and robustness 2. Selection of the response variable Most often, the average or standard deviation of the measured characteristics will be the response variable. Multiple responses are not unusual. 3. Choice of factors, levels and ranges Definition of a set of design factors, held constant factors and allowed to vary factors. 4. Choice of experimental design Sample size, suitable run order, blocking and randomization need to be considered. Selection of an experimental design described e.g. in [22]. 5. Performing the experiment Careful running of the designed experiment is mandatory. Errors in experimental procedure usually destroy validity. 6. Statistical analysis of the data Results and conclusions are objective with statistical methods. Residual analysis and model adequacy checking are important. 7. Conclusions and recommendations Graphical methods are often useful in presenting the results. Confirmation testing should be performed.

Steps 1 to 3 are part of the pre-experimental planning. Most of the decisions are closely linked to the nature of the process to study. Hence, generalization at this stage is not possible.

13 2.3.2. Range of factor levels

The range of the factor levels (difference between low and high level) must be determined carefully. If the range is too small noise can result in erroneous conclusions. If the range is too large nonlinear effects can affect the result and lead to wrong conclusions. Another, important issue is gauge capability. If the measurement error is large only large effects can be detected. 2.3.3. Sample size

At step 4 the DOE gets more formal. The determination of the sample size is done by calculation. Before the calculation can start some parameters need to be defined. The difference in effect to be detected must be determined. It is the resolution of the experiment. A very small value of can always be accepted. On the other side it is easy to detect large differences. So, it is advisable to determine as large as possible. The standard deviation must be known. The acceptable type 1 error must be defined. It is the risk of detecting an insignificant effect. Usually is very low e.g. 5%. The power is the probability to detect a desired effect. It is based on the type II error . It is the risk of rejecting a significant effect. Usually, is larger than , e.g. 20%. The number of factor levels is also necessary for the determination of the sample size. With this information the number of replications can be obtained from tables (e.g. Table 4.28 of [23]). Alternatively, it is possible to use the no.reps function of the dae package of R3.

2.3.4. Experimental Design

The basic design is a full factorial experiment. With two factors at two levels it consists of 4 trials (Table 2). It is called a 22 design. It gives information on the mean, the influence of factors 1 and 2 as well as the interaction of factor 1 and 2. The formal expression is the empirical model:

(2.1) y is the response, the ’s are the coefficients, x1 and x2 are the variables representing factors 1 and 2 and is the random error.

The number of trials increases with increasing number of factors according to:

(2.2) m is the number of trials and k is the number of factors. With increasing k the model includes coefficients for k-1 interaction terms. Supposing that three and more factor interaction terms are not relevant it is possible to reduce the number of trials. The design of fractional factorial experiments is described in [22].

3 www.r-project.org. The R Project for Statistical Computing.

14 Table 2. 22 design [22].

Run Factor 1 Factor 2 1 low low 2 low high 3 high low 4 high High

2.3.5. Center Points

A general 2k factorial experiment results in a first order model. However, it might be possible, that a first order model is a poor approximation of the data. This is called lack of fit. It occurs if the range between factor levels is too large. To be able to detect this defect it is necessary to add centre points to the experimental design. If the difference between the predicted model value at the centre point and the average measurement value at the centre point is high there is lack of fit. So, by including the centre point in the experimental design it is possible to partition the residual sum of squares into the residual sum of squares due to pure error and residual sum of squares due to lack of fit. In case of significant lack of fit an experimental design (central composite design) based on a second order model is advisable. 2.3.6. Randomization and blocking

Blocking and randomization are strategies to reduce variability. The sources of the variability are nuisance factors. Nuisance factors that can result in large effects are explained in more detail, now. There are unknown and known nuisance factors. Some of them can be controlled others are uncontrollable. Unknown nuisance factors are uncontrolled. Randomization helps to minimize the influence of these factors. Without randomization a trend could corrupt the data. Known but uncontrollable nuisance factors can be managed if they can be measured. In this case analysis of covariance can be used to get rid of the interfering effect. Blocking is the strategy to deal with known and controllable nuisance factors. A block should contain all factor level combinations. For example, if an experimental design requires two replications and there is not enough material from one charge it should be done in two blocks. The first block is the first replication of the experiment with the first charge. The second block is the second replication with material of the next charge. Differences between charges can be accounted for. 2.3.7. Statistical Analysis

The summary of the statistical analysis is an ANOVA (analysis of variance) table (Table 3). The first column gives the name of the source of the variation. The second and third columns give the sum of squares and the corresponding degree of freedom. The fourth column gives the mean square which is the sum of squares divided by the degrees of freedom. The fifth column gives the F statistics which is the quotient of the mean square of the source of variation and the mean square of the error. The P- value gives the significance of the source of variation. The source of variation is considered significant if the P-Value is lower than 0.05.

15 Table 3. Structure of an ANOVA table of a 22 design [22].

Sum of Degrees of Squares Freedom Mean Square F0 P-Value

A SSA 1 MSA=SSA MSA/MSE

B SSB 1 MSB=SSB MSB/MSE

AB SSAB 1 MSAB=SSAB MSAB/MSE 2 Residuals SSE 2 (n-1) MSE=SSE/DOFE - - 2 Cor Total SST n2 -1 - - - 2.3.8. Residual Analysis and model adequacy

The statistical analysis is based on the assumption that the residuals show zero mean value and constant variance. Therefore it is necessary to check if these assumptions are not violated. Nevertheless, the analysis of variance is robust to deviations from normality.

The first step is to check if the residuals are normally distributed. This check can be performed graphically or with the Shapiro-Wilk test. In the graphical method the residuals should lie on a straight line in a normal probability plot. The Shapiro-Wilk test is implemented e.g. in Origin4 or R3. The second step is to plot the residuals vs. the fitted values. This plot should not show any structure. Otherwise there is non-constant variance. A technique to compensate this defect is a transformation called Box- Cox Method. The third plot shows the residuals vs. run order. With proper randomization this plot should not show any structure. Additionally, the residuals can be plotted as a function of factor levels. Again, these plots should not show any structure.

Outliers in the data can be detected by computing the standardized residuals. By di- viding the residuals with the estimate of the standard deviation the residuals show approximately unit variance. Standardized residuals outside the interval -3 to 3 are potential outliers. 2.4. Superabsorbent polymers (SAPs)

Just for the benefit of the reader the chapter 2.4 is compiled from [24]. The author is not and does not claim to be the discoverer, inventor or originating scientist of any scientific fact in this chapter. The only purpose of this chapter is to introduce superabsorbent polymers to the reader. 2.4.1. Traditional Absorbent Materials

Traditional absorbent materials are e.g. cellulose acetate sponges, polyurethane sponges, filter paper and cotton balls. The microscopic structure of these materials is either fibrous or porous. They can absorb up to 19 g liquid per gram dry substance.

Capillary pressure is the most important force that can generate a flow of fluid into the hollow open spaces of the absorbent material. Other forces like gravity and hydrostatic pressure can also generate a flow of liquid into the material. Generally, it is

4 www.originlab.de

16 the capillary pressure that results in the desired absorption of liquid even against other forces e.g. gravitation.

(2.3)

Equation (2.3) enables to calculate the capillary pressure . is the density of the fluid. g is the gravity. is the height of the capillary. is the surface tension of the liquid and is the radius of the capillary. is the wetting angle. This equation is valid for a simple cylindrical tube. In absorbent materials many fibers or pores form fluidic paths. The collective of these paths can be regarded as a bunch of parallel capillaries with a corresponding effective diameter.

Another force that can result in a flow of liquid into the absorbent material is hydrostatic pressure. Contrary to the capillary pressure this force requires an external pressure gradient to generate a flow of liquid into the absorbent material. Darcy’s law of porous media is an empirical description of this phenomenon:

(2.4)

is the flow rate, is the pressure difference across a piece of absorbent material with cross sectional area and length , is the viscosity of the liquid and is the specific permeability of the material. The specific permeability represents the specific geometry of the material of interest. Of course, Darcy’s law is always relevant if there is a flow in a porous material.

External mechanical compression results in flow of liquid out of the absorbent material. The modulus of the porous structure determines the extent of compression depending on the external pressure. Mechanical release of water is a deficiency of traditional absorbent materials. 2.4.2. Superabsorbent Materials

Superabsorbent polymers are cross linked networks of flexible polymer chains with functional groups. The functional group can be either ionic or nonionic. Ionic groups can be either anionic or cationic. The molecular structure of an anionic polymer is presented in Figure 2.4. The absorbent properties originate from the functional groups at the polymer chains. Hence, the absorption of liquid into the superabsorbent material is a diffusive mechanism. It is driven by the difference of the chemical potential of water molecules outside and inside of a particle of superabsorbent polymer. The cross links between the polymer chains enable the tremendous volume change of the material without dissolution of the material. Of course, the diffusive uptake of water molecules is a temperature dependent process. Take up of liquid is no longer determined by capillary pressure or hydrostatic pressure.

Figure 2.4. Representation of a superabsorbent polymer. The dots represent the crosslinks between the polymer chains. The functional groups are made of carboxylic acid groups that are partially neutralized with sodium ions. (Source: Figure 1.2 of [24])

During the swelling process of a superabsorbent polymer both ions, the polymer linked ions and the free counter ions are solvated strongly. Of course, charge neu- trality must be obtained. So, the distance between the two ions is limited. Therefore, the swelling of the polymer can be considered an osmotic process without osmotic membrane.

The cross links of the polymer can be either covalent, ionic or hydrogen bonds. How- ever, only covalent cross links result in a stable material. If ions are present in the absorbed liquid the cross links change in an unpredictable way. Hydrogen bond cross links can be easily destroyed by simply heating the material. So, commercial relevant superabsorbent polymers use covalent cross links.

The typical superabsorbent polymer is an anionic polymer. The most relevant type is crosslinked, partially neutralized polyacrylic acid. The starting materials are acrylic acid and sodium hydroxide or sodium carbonate. These materials are very cheap. Other anionic polymers are e.g. polyacrylamide or polyvinyl alcohol. Acrylamide is toxic. Hence the monomer content in the polymer is a critical issue. Superabsorbent polymers made from polyvinyl alcohol are moderate expensive and its absorbent properties fall short of polyacrylic acid. Cationic polymers are expensive with low efficiency.

Superabsorbent polyacrylates are fabricated by free-radical initiated polymerization of acrylic acid and its salts. Partially neutralized acrylic acid at 20 to 40 wt% in aqueous solution, cross linker and initiator are mixed in a temperature controlled (30 °C – 80 °C) reaction vessel. Cross linker can be di- to tetra-functional groups. Ethylengly- col dimethacrylate is a di-functional cross linker. The initiator is a free-radical source. Sodium persulfate is a potential molecule to start the reaction (Figure 2.5).

After the polymerization reaction the superabsorbent material can be removed from the reaction vessel as gel. The gel is spread on a tray and dried in the oven. After- wards, the dry substance is crushed to particles typically ranging from 150 µm to 850 µm.

Figure 2.5. Polymerization reaction. Top: Acrylic acid is partially neutralized. Bottom: Partially neutralized acrylic acid, cross linker and initiator are required to start the polymerization and cross linking reaction. (Source: Figure 2.1 of [24])

Gel blocking is an unwanted effect in the absorption of liquids. It occurs when water is absorbed very fast by a tightly packed layer of very fine particles. The increase in particle volume, the particle deformation and the adhesion between particles reduce the interparticle porosity. Hence, the swelling rate of the superabsorbent polymer mass is reduced to the diffusion rate of water through the partially swollen mass.

One solution is to remove the very fine particles. Even more effective is the generation of a more rigid shell (Figure 2.6) around the superabsorbent polymers. This shell can be obtained by additional cross linking of the particle surface in a post processing step. These more rigid particles result in a higher open porosity of the particle bed even if some water is already absorbed. The higher open porosity results in a larger surface area hence in a higher absorption rate.

Suspension polymerization is another fabrication process. Droplets of monomer containing solution are dispersed in a phase immiscible with the continuous phase. Polymerization takes place in the dispersed phase. Typically, round particles form in a suspension polymerization. Their diameter can be controlled by agitation of the reaction vessel. Hence, the particle size distribution is very narrow. The surface is reduced compared to ground particles, so the absorption rate is lower. To circumvent this issue several techniques were developed to generate round and porous particles. With suspension polymerization the continuous phase must be recycled to be cost effective. Additionally, round particles roll around in diapers and can result in an unfavorable distribution in the product. Therefore, solution polymerization is commer- cially less important.

Figure 2.6. Surface cross linking is a post process (Source: Figure 3.9 of [24]).

19 Absorption, the rate of absorption and retaining of liquid in a superabsorbent polymer are the most important properties of interest. Unfortunately, there is no mathematical model of superabsorbent polymers that can explain all these properties based only on the molecular structure of the polymer. The following list gives just a brief overview of the factors influencing the molecular structure: 1. The molecular weight of the polymer chain between crosslinks 2. The molecular weight of polymer if no crosslinks were present 3. The amount of uncrosslinked polymer present (the extractible fraction of polymer) 4. The extent of entangling of the chains 5. The concentration of monomer at polymer preparation 6. The spacing of the charges along the chains 7. The thermodynamic interaction of the solvent with the polymer chains 8. The number of polymer strands that meet at a crosslink point

The shear modulus is an important parameter with great influence on the absorption of the polymer. It depends on the structure of the network of the molecules. Apart from the spacing of the charges along the chains it depends on all the listed factors. Many cross links, hence short polymer chains between cross links result in a gel with high shear modulus. Such a gel shows only a small maximum absorption of liquid. During the swelling process the shear modulus decreases with increasing amount of absorbed liquid. This is due to the fact that the number of elastic chains per volume decreases when liquid is incorporated into the gel.

The superabsorbent polymer gel is a network of ionized polymer chains. The number of fixed charges is equal to the number of free ions. The higher the concentration of ions in the gel the more ions in the external liquid are excluded from the gel phase. So, the gel can be viewed as a polymer solution confined by an osmotic membrane. The maximum absorption of liquid is decreased by decreasing the number of fixed charges in the gel or by increasing the number of ions in the liquid. Hence, the performance of superabsorbent polymers is reduced if body fluids like urine (0.9 % NaCl) need to be absorbed.

The swelling pressure of a superabsorbent polymer decreases with decreasing polymer content of the gel. In other words the swelling pressure decreases with increasing solvent concentration. The swelling pressure is the sum of three terms:

(2.5) the first contribution accounts for the interaction between the solvent and the polymer molecules. It tends to increase the solvent content of the gel. The second contribution accounts for the elasticity of the gel. It is against the uptake of solvent into the gel. The third contribution accounts for the ionic interaction. It increases the uptake of solvent into the gel. The highest swelling pressure , associated with a high , can be obtained with DI water. Superabsorbent polymer particles with core shell structure can be manufactured by cross linking of the polymer chains at the surface of the particle. A larger crosslink density results in a stiffer polymer with reduced swelling capacity. Therefore, it is obvious that the swelling capacity and the elastic modulus vary radially through a superabsorbent polymer particle. Typically, the swelling rate of a mass of superabsorbent particles with core shell structure is increased. The hard shell of the particles

20 prevents early deformation of the swelling particles under load. Consequently, a small space between the particles enables water to enter the core of the particle bed. This space is closed at much higher pressures compared to polymer particles without cores shell structure. At this state the mass of polymer particles transformed into a homogenous gel. Slow diffusion is the only transport mechanism, left. This state is known as “gel blocking”.

22 3. State-of-the-art

The state-of-the-art presented in chapter 3 is researched by the author usually in scientific journals and conference abstracts. The sources are listed in the chapter Reference at the end of the text and are referenced in the text. The contribution of the author to this chapter is the compilation of the relevant information in a readable way. The author is not and does not claim to be the discoverer, inventor or originating scientist of any scientific fact in chapter 3. 3.1. Microneedles

The concept of microneedle enhanced drug delivery goes back to a patent of Alza Corporation filed in 1971. It describes a plurality of projections made of a drug release rate controlling material and a reservoir communicating with said projections containing the drug [25]. The basic idea is to breach the stratum corneum barrier with tiny solid structures called microneedles. This chapter focuses on microneedles for this purpose. In the meantime microneedles are also used for other applications. 3.1.1. Microneedle arrays made of metal

Solid metal microneedles can be manufactured from a plain sheet of metal. The plain sheet of metal is structured e.g. by a 2 dimensional erosive process (Figure 3.1). In a post processing step the microneedles may be bent by 90° to obtain an out of plane microneedle array. Alternatively, it is possible to use a single row of these in plane microneedles. Examples of solid metal microneedle arrays are published in [26, 27]. A sheet of titanium was etched and the microneedles were bent in a post processing step. The length of a single needle is 330 µm. The needle density is 190 needles per cm² and the size of an array ranges from 1 cm² to 2 cm² (Figure 3.2, left). Application of these short microneedles requires an impact applicator.

Figure 3.1. Typical fabrication process of solid microneedle arrays. a) The process starts with a plain sheet of metal. b) The microneedle structures are generated with a 2 dimensional struc- turing process. c) The microneedles are bent by 90° to obtain an out of plain microneedle array. The schematic shows a side view of the 9 x 9 microneedle array of b). d) A single row of 3 in plane microneedles. 23

Figure 3.2. Left: Microneedle array manufactured with an etching process. (Source: Figure 1b of [26]) Right: Microneedle array manufactured with a laser cutting process. (Source: Figure 3a of [29])

The length of the solid microneedles must increase if an impact applicator is objectionable. A microneedle array design with 1000 µm long microneedles is reported in [28]. The needles with a tip angle of 20° and a base width of 200 µm were manufactured by laser cutting from stainless steel sheets with a thickness of 75 µm. The array configuration is 7 x 15 needles. A similar geometry with a needle length of 650 µm, a needle width of 160 µm and an array configuration of 5 x 10 is presented in [29] (Figure 3.2, right).

Hollow metal microneedle arrays are usually manufactured by electroplating. This process requires a negative mould of a microneedle array. Electrochemical addition of metal to this mould generates the desired microneedle array.

A process to manufacture a negative mould with very few fabrication steps is presented in Figure 3.3, left [30]. Tapered holes were drilled in a 500 µm thick sheet of biaxially oriented polyethylene terephthalate (BoPET) with a UV laser. The tapered holes were obtained by considering the energy distribution of the laser beam. The circular beam showed a high energy density in its centre and a lower energy density at its circumference. By programming a circular path with a diameter smaller than the diameter of the circular laser beam the desired tapered holes were generated. After- wards, the conductive seed layer was added by sputter deposition. It is a stack of Ti, Cu and Ti. The microneedle array is formed by electroplating of Ni. Finally, the polymer mould material was removed by wet etching in boiling NaOH. The result of the described process is shown in Figure 3.3, right. The diameter of the microneedles at the base is 250 µm, 50 µm at the exit hole with a height of 500 µm.

Different fabrication protocols are published [31, 32, 33, 34] to generate electroplated microneedle arrays. There are two distinct features common to this approach. First, the opening of the microneedles looks like a punch. Second, the material used for electroplating is Ni. It is likely, that the punch like geometry of the microneedles prevents successful intradermal infusion. Most probably the needles get blocked by a piece of punched skin. None of the referenced publications includes data on successful infusion experiments in ex vivo or in vivo skin. Ni is a material widely used in microtechnology for electroplating. However, nickel allergies are not uncommon. So, it is suggested to coat the microneedles with a biocompatible polymer. If this option is a reliable solution to the problem is unclear and questionable. Other metals that are suitable for electroplating (Au, Pd, Pt) are more expensive.

Figure 3.3. Left: Typical fabrication process for hollow metal microneedle arrays. a) Laser drilling of tapered holes in polymer. b) Deposition of a conductive seed layer. c) Electroplating. d) Selective etching of the mould material. (Source: Figure 3 of [30]) Right: Microneedle array fabricated with the process described in the left picture. A 27 gauge needle is shown for comparison. (Source: Figure 4 of [30])

Conventional cannulas with gauge index 30 and higher are also frequently considered as microneedles. The outer diameter of such cannulas is 310 µm or smaller. Therefore, the comparison to micro fabricated microneedles is appropriate. The fabrication of microneedle arrays from single cannulas is possible [34]. However, it is a quite laborious process potentially unsuitable for mass fabrication. 3.1.2. Microneedle arrays made of silicon

Many fabrication technologies for micromechanical structures are derived from fabrication processes of the microelectronics semiconductor industry. With these technologies small structures can be fabricated and combined with electronic components. The manufacturing costs for small quantities are usually very high due to the initial setup of the fabrication line. To fabricate hollow silicon microneedle arrays for less than 1 € per piece it may be necessary to manufacture several million pieces per year (especially if dry etching processes are used).

Figure 3.4. Left: Side view of a 280 µm tall microneedle fabricated in a wet KOH etch process. Right: Top view. (Source: Figures 4a and 4b of [36])

Figure 3.5. Left: Microneedle array composed of microneedles with a height of 150 µm. (Source: Figure 1a of [63]) Right: Microneedle with a height of 70 µm. (Source: Figure 5 of [37])

Solid silicon microneedles can be manufactured by various fabrication protocols [35, 36, 37]. Very simple fabrication protocols are based on a single wet etching process. The design of microneedles generated with a wet KOH etching process is strongly influenced by the anisotropic silicon crystal (Figure 3.4). With an isotropic dry etching process (e.g. reactive ion etching) it is possible to generate round microneedles (Figure 3.5, left). A combination of isotropic and anisotropic etching processes results in very small and sharp microneedles. Such microneedles can be used to pierce the stratum corneum without pain [38] (Figure 3.5, right).

Hollow silicon microneedles always require an anisotropic etching process to generate the lumen of the microneedle. This is a time consuming process because the needle lumen extends from the backside of the silicon wafer to somewhere close to the needle tip. Additionally, this is a single wafer process. Parallelization requires investments in expensive anisotropic etching equipment. So, hollow silicon microneedle arrays are quite expensive.

Very sophisticated fabrication protocols result in very nice hollow microneedle geometries [39, 40] (Figure 3.6). More simple protocols restrict the design variety. Nevertheless, it is still possible to generate hollow silicon microneedle arrays [41] (Figure 3.7, left). An uncommon approach to fabricate microneedles uses a wafer saw to define the basic shape of the microneedles [42] (Figure 3.7, right). The dicing process is followed by a wet chemical etch.

Figure 3.6 Left and right: microneedles made of silicon manufactured with sophisticated fabrication protocols. (Source: Figure 6a of [39] and Figure 6 of [40])

Figure 3.7. Left: Microneedles made of silicon manufactured with a simple fabrication protocol. (Source: Figure 4c of [41]) Right: Microneedles manufactured with a fabrication process based on a wafer dicing saw with subsequent wet etching. (Source: Figure 2 of [42]) 3.1.3. Microneedle arrays made of polymer

The majority of fabrication protocols for microneedle arrays made of polymer require a negative mould. This negative mould is filled with the polymer to obtain the desired microneedle array. Once the negative mould is generated the fabrication of several polymer microneedle arrays by reusing the same negative mould is possible. There- fore, polymer microneedle arrays can be potentially manufactured very cost effective.

Hot embossing is one process used to manufacture polymer microneedle arrays. The hot embossing equipment usually consists of two plates that can be heated and pressed against each other. The negative mould and the polymer are placed between these two plates during the hot embossing process. Optionally, the process can be performed in vacuum.

Fabrication of a micro structured negative mould can be challenging. A workaround of this issue is to make a negative mould from a positive master [43]. This can be done by pouring a liquid moulding compound on the master (Figure 3.8, left and middle). Silicone is a typical material that can be used. It is a duroplast that can with- stand the temperatures during the hot embossing process easily. The hot embossing process with a silicone mould is usually referred to as soft embossing.

Figure 3.8. Left: Process of transforming a silicon microneedle array into a negative silicone mould. Middle: Hot embossing process. Right: Microneedle array made by hot embossing of COC. (Source: Figures 1,7 and 8 of [43])

27 A silicon microneedle array manufactured with micro fabrication technologies can be used as a positive master. After curing of the mould material the silicon master can be demoulded mechanically or by wet etching with KOH. Successful replication by hot embossing from the silicone mould is demonstrated in Figure 3.8, right.

Another mould fabrication process uses an electroplating process to transform a SU-8 photoresist structure to a metallic in plane row of microneedles. The in plane rows are assembled to a microneedle array. Next, this array is transformed into a negative mould made of PDMS for soft embossing [44]. Three more mould fabrication technologies are described in [45].

After negative mould fabrication the generation of microneedles can be done with the already described soft embossing process. Alternative processes include pouring of liquid material into the mould with subsequent solidification in the mould. Solidifica- tion can be either cooling of a melt [45, 46] or ultrasonic welding of small polymer particles [47]. Of course, these processes are usually performed in a vacuum cham- ber to ensure bubble free filling of the mould.

With the fabrication methods mentioned so far solid plastic microneedles can be manufactured. The microneedles can be used to pierce the skin followed by the application of a drug loaded patch. Alternatively, the drug can be either coated onto the microneedles or it can be incorporated into a biodegradable polymer.

Fabrication of hollow polymer microneedle arrays requires the generation of moulds with pins that generate the microneedle lumen. The slim design of microneedles usually results in high aspect ratios for the pins with very small diameters. Therefore, mechanical stability of the pins is a critical issue in mould development. An alternative approach to pins is the generation of the lumen in a post processing procedure.

A quite inventive approach to the fabrication of in plane polymer microneedle rows is based on a silicon mould [48]. The microneedle geometries are defined by an anisotropic KOH etch process. Bond wires were mounted and aligned in the cavities (Figure 3.9, left). An injection moulding machine was used to fill the mould with polymer. After demoulding the bond wire made of aluminium was dissolved with a selective etchant.

Figure 3.9. Left: Microneedle tip before and after dissolution of the bond wire. Right: Fabrica- tion of hollow microneedles with laser drilling as a post processing step. (Source: Figure 22 of [48] and Figures 1 and 3 of [49])

28 Laser drilling is a suitable post process for the generation of hollow microneedles [49]. First, the microneedles were fabricated by a hot embossing process. Counter- part of the negative mould was not just a simple plane plate but a microneedle array itself (Figure 3.9, right). The microneedles were slightly smaller than the microneedle cavities of the negative mould. After demoulding the polymer microneedles already featured a lumen from the backside of the microneedle array to just below the tip. With a laser drilling process the residual polymer layer was removed to generate the hollow microneedles. 3.1.4. Other fabrication technologies and other applications

Most of the fabrication technologies for metal, silicon and polymer microneedles can be considered mainstream. Apart from this there are quite exotic approaches. Two of them are presented in this chapter.

Hollow microneedles can be generated by two-photon polymerization [50] (Figure 3.10, left). Ormocer which is an amorphous organic-inorganic hybrid material can be used for this purpose. With femtosecond short laser pulses a two photon absorption process is started which results in breakage of chemical bonds of starter photo initiator molecules. Unexposed material can be washed away with alcohol. Finally, the structures need to be cured with ultraviolet light. With a scanner and a vertically moveable reservoir it is possible to generate complex 3 dimensional structures. The lateral resolution of this process is very high (100 nm).

Fabrication of hollow microneedles made from PMMA requires synchrotron radiation [52] (Figure 3.10). Exposed to this radiation PMMA changes its chemical properties and can be dissolved selectively. Hollow microneedles with a length of 350 µm to 800 µm, an outer diameter of 300 µm and very sharp tips can be obtained.

Intradermal drug delivery is not the only application of microneedles. Blood sampling, drug delivery to the sclera (white of the eye) and cell surgery are applications proposed in the publications [53], [54], [55] and [56].

Figure 3.10. Left: Microneedles generated with a two-photon polymerization process. The length of the microneedles is 800 µm. Right: a) Sharp microneedle made of PMMA b) Top view. (Source: [51] and Figure 3 of [52])

29 Neuroscience is a discipline where microneedles are used for various purposes. Re- cording and electrical as well as simultaneous chemical stimulation is done with microneedles. Such needles are usually fabricated with technologies used to manufacture microelectromechanical systems (MEMS). With this technology the generation of tiny electrodes on the surface of hollow microneedles is possible [57, 58, 59, 60]. 3.2. Intradermal drug delivery with microneedles

Intradermal drug delivery goes back to 1796 when the English physician Edward Jenner vaccinated a boy. He scratched his skin and introduced body fluids of a maid- servant diseased with cowpox. Afterwards, the boy was immune against pox.

In 1910 Charles Mantoux published his research on the efficacy of the intradermal performed tuberculin sensitivity test. To perform this test a volume of 0.1 ml containing 5 Tuberculin units is injected with a gauge 27 (outer diameter 0.41 mm) needle into the upper skin layers. The needle is inserted at a very small angle. The test was so successful that the procedure of intradermal injection became known as Mantoux technique.

Scratching of the skin is not very precise and the Mantoux technique requires trained personal. Microneedles promise to simplify intradermal drug delivery and make it precise in delivering a defined amount of drug to the body. 3.2.1. Insertion of microneedle arrays into the skin

Very sharp microneedles can be applied by simply pushing the microneedle array into the skin with the fingers [61]. Other microneedle arrays especially with short and densely arranged microneedles require an impact insertion method. In publication [62] the penetration of microneedle arrays with 4 x 4 to 9 x 9 microneedles was investigated. Solid microneedles made of steel with an outer diameter of 0.2 mm, a length of 0.3 mm and sharp tips were used. The spacing between the needles was 1.25 mm. With an impact speed of 1 m/s to 3 m/s it was possible to successfully pierce human skin in vitro. It was proven by permeation studies with a blue dye.

Another method to control the successful piercing of the skin in vivo is the application of histamine [43]. Successful piercing can be recognized easily by the formation of a wheal. Transepidermal water loss (TEWL) measurements can be performed to measure the disruption of the skin barrier in in vivo experiments [61]. With this method no chemicals or drugs are necessary. A final method established by engi- neers relies on measuring the change in electrical impedance after successful piercing of the skin [64, 65]. 3.2.2. Operating regimes of microneedle arrays

There are four major operating schemes of microneedle arrays: 1. Human skin can be pierced with solid microneedle arrays. Subsequently, a drug formulation can be applied to the treated skin. 2. The tips of the microneedles of an array can be coated with a certain drug. 3. Microneedles can be fabricated form polymers that dissolve in the human skin. In this case the drug of interest is mixed with the polymer. 4. Drug solutions can be intradermally injected and infused with hollow microneedles.

30 For the first operating regime there is literature data available obtained with in vitro experiments using Franz diffusion cells. Skin samples can be mounted in such cells. It is possible to apply a drug formulation on one side of the skin and at the same time the other side is in contact with the receiver solution. This receiver solution can be sampled to determine the permeation rate of the drug. Of course, the permeation of a drug through skin pretreated with a microneedle array can be studied. In [66] the permeation of calcein, a small hydrophilic molecule (623 g/mol), and insulin, a pep- tide hormone (5.8 kg/mol), was studied. Both molecules permeated through microneedle pretreated skin. The permeation rate of calcein was higher compared to insulin.

Coated microneedles require very potent drugs because the absolute surface of microneedles is very small. Vaccines are very potent. Microneedle assisted intradermal vaccination was demonstrated with a row of 5 coated microneedles in an in vivo animal study [67]. The vaccine coated stainless steel needles were inserted into the skin and left for 10 min to dissolve the vaccine in the skin. The immune response of this procedure provided full protection and it was in some aspects significantly stronger than the control group that received the vaccine via the intramuscular route.

Microneedles made of biodegradable polymers dissolve in the skin. Typical biodegradable materials are e.g. polylactic acid, polyglycolic acid and polylactic-co-glycolic acid. It is possible to fabricate quite impressive microneedle structures from these polymers. Lengths of up to 1.5 mm and aspect ratios of up to 8 are reported. The biodegradable microneedles are strong enough to pierce the skin [45]. The biodegradable microneedles can be loaded with drug particles. However, the biodegradable polymer is fluidized by heating during manufacturing. This heat may destroy heat sensitive drugs. Additionally, the drug load decreases the mechanical stability of the polymer microneedles. Drug loads of 10 % can be critical. The maximum total dose that can be administered is likely to be less than 1 mg (1000 microneedles with a mass of 10 µg and a drug load of 10%). Nevertheless, with additional modifications it is possible to design microneedle patches with sustained delivery of the incorporated drug [46]. 3.2.3. Intradermal injection and infusion

Qualitative characterization of intradermal drug delivery in vitro with hollow microneedles can be performed with coloured or fluorescent dyes, polymeric beads and cells [68, 69] usually dissolved in phosphate buffered saline (PBS). Successful delivery of all these substances is observed with methods like microscopy, fluorescence microscopy, confocal microscopy and microscopy of histological sections. Accurate quantitative results can be obtained with radio labeled serum albumin.

For the design of small intradermal drug delivery devices with hollow microneedles it is necessary to know the back pressure that builds up at the desired flow rate. Only with this information it is possible to select an adequate micro actuator. Back pressure data of intradermal infusion is published in [70]. A hollow microneedle made of glass was inserted perpendicular to the skin surface. Only with retraction of the microneedle it was possible to obtain high flow rates. A typical flow rate measured is 100 µl/h at a pressure of 150 kPa if the microneedle was inserted 1080 µm and retracted 540 µm. Addition of hyaluronidase, rapidly breaking down hyaluron within skin collagen fibers, increased the flow rate 7 fold.

31 Intradermal infusions in vivo can generate purely physical data or pharmacokinetic as well as pharmacodynamic data. While physical data can be generated with sterile isotonic solution the selection of a suitable compound is necessary for pharmaceutical experiments. Pharmacodynamic data can be generated very simple e.g. with methyl nicotinate. This drug is a vasodilator. It results in widening of blood vessels. Injected or infused intradermally [71] it dilates the surface capillaries. This results in a change in skin color. Alternatively, quantitative data can be generated with a laser perfusion monitor. Other drugs may require more elaborate methods based on blood samples. 3.2.4. Intradermally injected vaccines

The most advanced application of hollow microneedles is intradermal vaccination. One fifth of the intramuscular dose injected intradermally results in equivalent protection [72]. An intradermal injection device filled with flu vaccine is available from Sanofi Pasteur (Intanza). The device is based on a short microneedle made of a hollow steel cannula [73]. The length of the needle is 1.5 mm with an outer diameter of 0.31 mm (30G) and a short bevel. The microneedle is inserted perpendicular to the skin surface. It was tested in 645 human volunteers. In 2524 experiments 120 µl of isotonic solution was injected with the microneedle device. There is less injury of capillary loops compared to the traditional Mantoux (cannula insertion at very small angle) method because of the perpendicular insertion process. Fluid leakage was less than 5 %.

Perceived pain during injection was rated “faint pain” and quantified to 10.2 mm on a visual analog scale (VAS) for the microneedle device compared to 27.6 mm for the traditional Mantoux technique. The following adverse events were observed: small drop of blood (43%), skin redness (35%), itching (5%). These events were spontane- ously reversed within 20 to 30 min. without sequel or requirement of medical inter- vention. Overall the microneedle device is safe and easy to use. The microneedle device results in superior immunogenicity in elderly adults [74].

Hollow silicon microneedles were also used in another vaccination study [75]. A single row of 4 microneedles with a length of 0.45 mm and the design of Figure 3.6, right was used. Leakage was observed in the study with 180 human volunteers but not quantified. 3.2.5. Intradermal delivery of insulin and other biopharmaceuticals

Insulin was infused intradermally in rats with a microneedle array made of silicon [76]. It was a 4 x 4 design with 0.4 mm long microneedles and an edge length of 4.0 mm. The array was integrated in a micro dispenser that was powered by a thermally expanding silicone material. The heat for the actuator was generated with an electric current. Insulin Lispro (100 U/ml), a fast acting insulin, was infused intradermally (n=18), subcutaneously (n=9) and intravenously (n=9). The data presented shows no disadvantage using the intradermal route compared to the subcutaneous route. The minimal invasive and painless intradermal procedure is supposed to improve compli- ance and result in less frequent complications in the long term.

An intradermal insulin study in 2 humans is reported in [77]. Microneedles made of glass with a micropipette puller were used. Single microneedles were inserted perpendicular into the skin in depths of 1 mm, 3 mm and 5 mm. Subcutaneous delivery

32 of insulin was performed to obtain control data. A 50 U/ml insulin lispro was prepared form 100 U/ml insulin lispro. The subjects received insulin at a flow rate of 1 ml/min. The absolute amount of insulin was determined according to the subject’s blood glucose level and individual insulin requirements. Successful intradermal insulin infusion was acknowledged by the observation of a skin wheal. It disappeared 2 h after administration. Mild erythema disappeared 30 min. after administration. Overall, patients preferred microneedle based delivery over subcutaneous delivery. Microneedle based intradermal insulin delivery at a depth of 1 mm was at least as effective as subcutaneous delivery. There might be faster pharmacokinetics.

More advanced studies in humans are published in [78, 79]. 29 patients were enrolled in a study investigating pharmacokinetic and pharmacodynamic postprandial glycemia5 of intradermally delivered insulin versus subcutaneous delivered insulin. Single microneedles made of steel with an outer diameter of 0.18 mm (34G) and a length of 1.5 mm were used. The blood glucose level of the patients was adjusted to 120 mg/dl with a Biostator (glucose clamp)6. Regular human insulin was administered (0.125 U/kg) 17 min. or 2 min. before a standardized liquid meal was given. Insulin lispro was given 2 min. prior to the meal. Postprandial glycemia was 14 % and 11 % lower if regular human insulin was given 17 min. and 2 min. prior to the meal compared to subcutaneous delivery. Postprandial glycemia was lower with intradermal insulin lispro compared to subcutaneous insulin lispro. However, it was not statistically significant at a P-value of 0.10. Microneedle delivery of insulin was safe and generally well tolerated in this study.

Another advanced study [80] enrolled 10 healthy volunteers. The study was performed under euglycemic7 glucose clamp conditions. Stainless steel microneedles with lengths of 1.25 mm to 1.75 mm and an external diameter of approximately 0.26 mm were used. The volunteers were administered 10 U of insulin lispro at 3 microneedle lengths on days 1 to 3 followed by a repetitive intradermal injection with the 1.5 mm long needle and a subcutaneous injection on day 5. The result of the study showed accelerated pharmacokinetics of intradermally delivered insulin compared to subcutaneous delivery. The time to maximum plasma concentration tmax=36.0±2.8 min. of intradermal delivered insulin (1.25 mm long microneedle) is almost half of tmax=64.3±5.7 min. of subcutaneous delivered insulin. The pharmacodynamics of intradermally administered insulin is also accelerated. With 106.2±4.3 min. the time to maximum glucose infusion rate after intradermal injection (1.25 mm long microneedle) is shorter than the time to maximum glucose infusion rate of 130.0±5.9 min. obtained with subcutaneous injections. Intradermal insulin was generally well tolerated. Localized wheal formation and redness were observed at injection sites.

Pharmacokinetic data of other biomolecules intradermally delivered with microneedles was obtained in an animal study [81]. Etanercept a 132 kDa protein, somatropin a 22 kDa protein, human insulin a 5.8 kDa protein that typically exists as a hexamer with 35 kDa and insulin lispro a 11.6 kDa protein were studied. All formulations were administered with stainless steel microneedles with a length of 1 mm and an outer diameter of 34G. The microneedles were connected to syringes mounted in syringe

5 increase of blood glucose after meal 6 blood glucose level is measured continuously and adjusted by adequate insulin infusion 7 normal level of glucose in blood

33 pumps. Volumes ranging from 50 µl to 250 µl were infused within 2.2 min. to 12.5 min in female Yorkshire swine. A 2 to 4 fold acceleration in pharmacokinetics was observed for intradermal delivery compared to subcutaneous delivery. A 2 to 4 fold increase in maximum concentration (Cmax) was also observed. The bioavailability of intradermal Etanercept is 50 % higher. Data is presented that supports at least partial uptake of the delivered agent via dermal lymphatic vessels.

Liquids were infused in another animal study with hollow plastic microneedles [82]. An array with 18 hollow microneedles distributed on an area of 1.27 cm² was used. The length of the microneedles ranged from 500 µm to 900 µm. It was demonstrated that quite large liquid volumes of up to 1.5 ml can be infused intradermally in 6 min. A domestic swine model was used. 3.2.6. Other aspects of drug delivery via microneedles

The peril of penetration of microbes into the skin is reduced with microneedles. The difference was studied by comparing a 6 x 5 microneedle (height 280 µm, diameter 250 µm) array made of silicon to a 21G (outer diameter 0,82 mm) hypodermic needle [83]. The microneedle design corresponds to the design shown in Figure 3.4. It was shown that microbes can still penetrate the skin after application of microneedles. However, based on the experimental data the authors concluded that the infection risk is reduced with a microneedle array compared to a 21G hypodermic needle.

Extraction of interstitial fluid with microneedles is reported several times. In one study microneedles made of glass were used [84]. The tip radii ranged from 15 µm to 40 µm. The needles were inserted 700 µm to 1500 µm deep into the skin of animals and human subjects. Interstitial fluid was extracted with a suction pressure of 27 kPa to 67 kPa within 2 to 10 min. 1 µl to 10 µl of interstitial fluid can be extracted with this procedure. The glucose concentration of the samples was measured. With a linear calibration factor the glucose concentration of the interstitial fluid was correlated to blood glucose values. The data was compared to reference measurements and considered clinically acceptable. 3.3. Drug delivery devices for chronotherapeutic drug delivery

Chronotherapeutic drug delivery requires at least a delayed release of the drug. A typical situation is the “night time dosing” problem. Bed time is the last chance for a patient to take a medicine before he starts to sleep for 7 to 8 hours.

A delivery device with a simple time delay can release the drug e.g. in the early morning hours to treat critical issues during this time of the day. Pulsatile delivery devices are necessary for high frequent dosing schemes (e.g. every 30 min.) or circadian schemes if the device is used for several days.

Chronotherapeutic drug delivery devices are in operation at least for several hours. Therefore, it is important that such devices don’t impose unacceptable constraints to the patient. Of course, these constraints are different for different conditions depending on their criticality. So, it is obvious that the oral route is preferred. Flat and light weight devices can be worn on the skin for several days without disturbing the patient too much. Implanted devices are also very convenient. However, surgery is required to place them in the body. The comfort of other routes is either very awkward, e.g. respiratory route, or almost impossible, e.g. ophthalmic route. Therefore, state of the art of chronotherapeutic drug delivery devices for the oral route and the 34 transdermal route as well as implanted devices is presented in the following sub- chapters. 3.3.1. Oral chronotherapeutic drug delivery devices

Oral drug delivery devices are very comfortable for the patient. As long as the device is not too big it just needs to be swallowed. Other tasks are not necessary. There are two approaches to realise such systems, the engineering approach and the chemical approach. There is work on electronic pills that release the drug controlled by electronics and micro actuators. The chemical approach is based on materials that dissolve with time. Manufacturing of such devices is more economic than the assembling of electronic devices.

To illustrate an oral chronotherapeutic drug delivery device the geoclock technology8 is presented. With geoclock technology, the active ingredient is packed inside a press-coated tablet (Figure 3.11, left). This active core is protected from gastrointestinal fluids by an inert shell; it is this shell that allows for the delayed release. On ingestion, the shell retains its integrity as it passes through the gastrointestinal tract and is unaffected by the pH of the environment at any point. Approximately four hours after ingestion, the release of the active core is triggered by the penetration of water through the inert shell of the tablet. Because of the chemical composition of this shell, breakdown and dissolution occur very rapidly. In this way, an oral drug can be ingested but the active ingredient is not released until four hours later (Figure 3.11, right) [85].

Figure 3.11. The geoclock tablet releases its content instantly after a delay time of approximately 4 hours. Left: Photo of several dissolution stages of the tablet. Right: Pharmacokinetic profile of prednisolon (Lodotra) delivered with geoclock [86].

8 Geoclock by SkyePharma, London, Uk

35 3.3.2. Transdermal chronotherapeutic drug delivery devices

The transdermal route is an aggregation of many more specific routes like: intravenous route, subcutaneous route, intradermal route, intramuscular route and others. The transdermal route is nowadays strongly related to so-called transdermal therapeutic systems (TTS). TTS contain the drug dissolved in a solid film [87]. Non-invasive diffusion of the drug across the skin barrier is the transport mechanism. Hence, these devices are primarily not suitable for chronotherapeutic applications. However, diffusion strongly depends on thermal energy. So, it is possible to generate a time dependent flux of drug by connecting a time controlled heat generator to the TTS [88]. Although TTS are quite new they are well known in public due to the success of the analgesic pain patch.

Another approach is to use drug solutions that are transported to the skin and removed from the skin with an electronically controlled pump. This is realised by a product of Chrono Therapeutics, Hamilton, NJ, USA. The device is worn like a wrist watch [89]. It contains a peristaltic pump that transports liquid to the skin according to a pre-programmed time varying profile. The liquid contains the drug specific for the intended therapeutic effect and a penetration enhancer.

The timely controlled pumping of liquid to the skin is quite straight forward from a pharmaceutical point of view. However, the solubility of lipophilic molecules in an aqueous phase results in potentially large liquid volumes. Additionally, the pumping technology also increases the device volume. Both issues are negligible if iontophoresis is used as transportation mechanism.

The basic principle of iontophoresis is the application of a small electric current to the skin. This current provides the driving force to primarily enable penetration of charged molecules into the skin. A drug reservoir is placed on the skin under the active electrode with the same charge as the penetrant. An indifferent counter electrode is positioned elsewhere on the body. The active electrode effectively repels the active substance and forces it into the skin (Figure 3.12). This simple electrorepulsion is known as the main mechanism responsible for penetration enhancement by iontophoresis. The number of charged molecules which are moved across the barrier correlates directly to the applied current and thus can be controlled by the current density. Other factors include the possibility to increase the permeability of the skin barrier in the presence of a flow of electric current and electroosmosis. Contrary to electrorepulsion, electroosmosis can be used to transport uncharged and larger molecules. Elec- troosmosis results when an electric field is applied to a charged membrane such as the skin and causes a solvent flow across this membrane. This stream of solvent carries along with it dissolved molecules. It enhances the penetration of neutral and especially polar substances [91]. The electric flow of iontophoretic applications can lead to local erythema [92].

A first product is close to commercialisation by Incline Therapeutics, Redwood City, CA, USA (Figure 3.13, left). It is called Ionsys. The electronic patch contains the opioid drug fentanyl. Once attached to the patient he can increase the flux of the pain relief drug by pressing a button. So, the Ionsys is a smart system because it is part of a feedback loop. The patient senses the pain and demands the necessary amount of drug. In this way variable disturbances caused e.g. by interindividual different skin properties are compensated [93, 94].

Figure 3.12. Basic principle of iontophoresis. A current passed between the active electrode and the indifferent electrode repelling drug away from the active electrode and into the skin [90].

A second product is already available on the US market. It is called LidoSite and it is developed by Vyteris, Extension Fair Lawn, NJ, USA (Figure 3.13, right). LidoSite contains lidocaine a non opioid analgesic. LidoSite provides fast and effective analgesia before blood draws. It is intended to be used with patients showing distinct fear of needles.

Non-invasive iontophoresis is practically limited to small, lipophilic and non-ionic molecules. To circumvent this limitation it is necessary to generate holes in the stratum corneum. Subcutaneous infusions sets are state-of-the-art technology. With the integrated cannula they generate a hole in the stratum corneum. The tip of the cannula reaches into the subcutaneous tissue. It can be used for up to three days. A liquid drug solution can be pumped through this set into the human body. Therefore, a chronopharmaceutical transdermal drug delivery device can be a combination of a time controlled drug pump and a subcutaneous infusion set. An alternative to subcutaneous infusions sets would be intradermal infusion sets. Technologies for intradermal infusions would be less invasive because of the very short channels associated with this technology. Cicatrisation can be reduced. Apart from this advantages the different pharmacokinetics already described in chapter 3.2.5 can be beneficial. How- ever, fabrication of hollow microneedles that enable leak tight intradermal infusions is challenging.

Figure 3.13. Left: Ionsys iontophoretic patch by Incline Therapeutics, Redwood City, CA, USA. Right: LidoSite by Vyteris, Extension Fair Lawn, NJ, USA.

37 State-of-the-art technology for transdermal chronotherapeutic drug delivery devices are insulin infusion pumps. These pumps provide time varying pumping functionality plus additional bolus functions. Medtronic, Roche and Animas are the most important players in the western world. Their pumps contain an exchangeable syringe reservoir with a volume of 2 ml to 3 ml. This reservoir is emptied by a plunger actuated by an electronic motor and a gear train. These devices require large batteries and are quite heavy. Usually, they are carried in the pants pockets or with a belt pouch. A pump originally developed for the infusion of insulin is used to infuse hormones with respect to chronomedical needs. The pump is available by Ferring Pharmaceuticals.

Smaller and lighter pumps can be carried directly on the skin. Such pumps are called “patch pumps”. The first patch pump on the market is the Omnipod by Insulet. This pump uses a two-way shape memory actuator and a gear train to drive the plunger in a cylinder. It is a tube free pump that automatically inserts a subcutaneous needle upon activation. The pump is a complete disposable. Everything from drug reservoir to batteries and electronics is thrown away after usage.

Several insulin pumps are currently under development at various companies. Table 4 gives an overview of the companies, the product names and the basic technology.

A disposable device for programmed or user controlled drug delivery is the Medipad infusor (Figure 3.14, left). It is based on a pressurized gas generated by a controlled electrolysis process. The technology of the Medipad infusor was described first in [95]. Its approximate dimensions are 8.5 cm x 6.2 cm x 2.1 cm with a mass of less than 50 g. It contains approximately 3.3 ml of drug solution. The flat design of the device is persuasive. Volunteers considered it comfortable to wear. Drug solution can be infused e.g. at 69 µl/h for 48 hours [96, 97].

Another disposable programmable drug delivery device is currently developed at SteadyMed Therapeutics, Tel Aviv, Israel (Figure 3.14, right). The actuator of the device is a battery. Draining current from this battery results in a volume increase caused by movement of lithium ions. The theoretical maximum expansion of this specific chemical process (intercalation) is approximately 260%. A high pressure can be generated. The technology is described in and protected by [98].

Table 4. Short description of technologies used for insulin pumps under development.

Company / Product Technology Asante Solutions / Pearl traditional gear and drive mechanism with flexible shaft Cellnovo / n.a. plastic micro pump actuated by phase change wax Debiotech / JewelPUMP silicon micro pump actuated by a piezo ceramic Medingo (now Roche) / Solo4You traditional gear and drive mechanism Medipacs / n.a. electroactive polymer Picosulin / n.a. electromagnetic plastic piston pump with gear train Sensile Medical / n.a. electromagnetic plastic piston pump Spring / Zone Insulin Pump capillary/spring combination with pressure sensors Tandem Diabetes Care / t:slim unclear

Figure 3.14. Left: Medipad infusor, by DALI Medical Devices, Rishon LeZion, Israel. The device enables electronically programmed subcutaneous infusions. Right: PatchPump by SteadyMed Therapeutics, Tel-Aviv, Israel. It uses an expanding battery as actuator.

The transdermal drug delivery devices presented so far determine the time varying flow rate with an electronic control. Such a universal approach is not necessary for simple chronotherapeutic devices. A time delay can be enough. This functionality does not necessarily need electronics. A device with time delay based on a hydrogel actuator is published by Richter A et al. of the Laboratory for Physical Chemistry and Electrochemistry, Dresden University of Technology in [99] and [100].

The swelling behaviour of the hydrogel actuator is a saturating exponential process. To achieve a linear swelling behaviour the supply of swelling agent to a fast swelling polymer was restricted. Additionally, the authors added a time delay mechanism based on a friction piston (Figure 3.15). The high counter force of the friction piston reduced the speed of the swelling process. After a certain distance the diameter of the cylinder increases. Hence the friction decreases and the speed of the swelling process increases. Due to this piston based mechanism the device shows an unfavourable rod shaped design. It can pump a fluid at a flow rate of 250 µl/h for two hours. The time delay functionality may help in dealing with the dawn phenomenon9 of diabetic patients or it might enable other chronotherapeutic [101] applications.

Figure 3.15. A pen shaped transdermal chronotherapeutic drug delivery device. (Source: Figure 2 of [99])

9 Increase in blood sugar usually between 2 am and 8 am.

39 3.3.3. Programmable implantable Infusionpumps

Implantable drug delivery devices might be a very convenient solution for chronotherapeutic applications. Once implanted in the body the device delivers the drug according to a programmed schedule without additional user action. From time to time a visit is required to refill the reservoir of the device. After several years the battery must be changed. Again surgery is necessary for this step.

A state of the art implantable infusion pump for time varying drug delivery is the Syn- chromed II by Medtronic (Figure 3.16). It is available with reservoir volumes of 20 ml or 40 ml. The battery needs to be replaced typically after 7 years. The main field of application is drug delivery via an intrathekal catheter for pain relief.

The device combines two actuators. The first one generates a constant flow rate the second one generates a time varying flow rate. The first actuator is based on a fluid that is in a liquid-gas phase at body temperature. In this state it generates a constant pressure on the collapsible reservoir independent of the reservoir fill level [103]. The second one is a peristaltic pump that is powered by electricity [104]. This configuration is very beneficial for a long battery life time if the peristaltic pump is rarely used, e.g. once a day for a short period of time.

Figure 3.16. Synchromed II by Medtronic [102].

40 4. Fabrication of hollow microneedles for intradermal injections and infusions

4.1. Hot embossing of hollow microneedle arrays

Hot embossing is a thermoforming process. A thermoplastic material is softened by heating usually in a vacuum box. Resting in a tray it can be structured by pressing a mould into the tray. When the forming process is finished the material is cooled to solidify in the new shape. Finally the moulded piece of plastic is demoulded.

The moulding temperature is usually close (above or below) to the melting temperature of the plastic. For thermoplastics the melting temperature is the midpoint of the melting range. Polycarbonate starts to soften at a temperature of 145 °C. At its hot forming temperature (≈200 °C), which is far below its flow temperature (240 °C) and very far below the typical injection moulding temperature (≈300 °C), it is soft enough to fill the evacuated hot mould under pressure. Micro- and nanometre sized features can be replicated in this way.

The duration of a hot embossing process is very long compared to injection moulding. Additionally, complete evacuation of the moulding tool and relative movement of mould and tray must be possible. So, the gap between mould and tray must be as small as possible. Thus, the viscous polymer melt is confined as good as possible in the moulding tool. However, the polymer will always flow into gaps to some extent. Hence, the polymer will fill this gap and form a thin film.

Hot embossing of microneedle arrays was a joint development of the Laboratory for Process Technology at IMTEK and HSG-IMT. The equipment for hot embossing was provided by the Laboratory of Process Technology. The students Kain A and van Goer M were financed by HSG-IMIT. The hot embossing know how was provided by Müller C of the Laboratory of Process Technology.

The original concept of using a strained tungsten wire in a hard mould to generate hollow microneedles was introduced by Müller C. Experiments with such a mould were performed by Kain A at the Laboratory for Process Technology under the supervision of Müller C and Vosseler M. The original concept was advanced by Kain A, Müller C and Vosseler M. This advanced concept used non strained tungsten wires and a soft mould. These concepts and experiments are documented in [122] by Kain A. This document was additionally supervised by Busolt U.

Further advancement of the mould concept was done by Müller C, van Goer M and Vosseler M. Mold design and the molding experiments were performed by van Goer M under the supervision of Müller C and Vosseler M. This further advanced mould concept and the experiments by van Goer M are documented in [106, 107] and presented in chapter 4.1.1 to chapter 4.1.3 of this work. The document [107] was additionally supervised by Schön H.

41 4.1.1. Preparation of the hot embossing tool

The design of a mould for a hot embossing tool starts with an analysis of the structure to be generated by the moulding process. The desired geometry of a single hollow microneedle and the corresponding layout of the desired microneedle array are presented in Figure 4.1. The microneedle is a cylinder with a bench, a bevelled tip and a through-going capillary. Remarkable is the aspect ratio of the through-going capillary. It is approximately 10 with a small diameter of 125 µm. Provided that the replicated microneedle is placed on a substrate with a thickness of 1.0 mm the aspect ratio of the core increases to approximately 20.

Manufacturing of a mould with the negative geometry of Figure 4.1 is challenging. With traditional methods of drilling and milling it is impossible to generate cavities representing the bevelled tips of the microneedles. These structures might be generated by sinker electrical discharge machining (EDM). Another alternative is to generate a positive microneedle array and cast a negative soft mould from this structure. This is a fast and cheap approach pursued in this work. A thermoforming process based on such a mould is called soft embossing.

Even more challenging is the realization of the lumen of the microneedles during the thermoforming process. It is clear that a rod like structure is necessary to generate the lumen. It might be possible to generate such a structure with a sink EDM process in a metallic mould. However, due to the high aspect ratio of 10 and the small diameter of 125 µm the contact area between the inner surface of the moulded plastic lumen and the outer surface of the rod is very high. During the cooling phase of the hot embossing process there will be a shrink fit of the lumen to the rod. Hence, the shrink fit and the large surface result in high demoulding forces. Nevertheless, demoulding might be possible with a metallic mould. However, it is a challenging task and tearing of one rod of an array would destroy the complete mould of this array. For the same reasons demoulding of a soft silicone mould is also extremely challenging. Therefore, the rods of the soft embossing mould are implemented as free moving and replaceable tungsten wires. The creation of the soft embossing mould is described in the following paragraphs in more detail.

Figure 4.1. Design of desired microneedles and layout of the microneedle array. Left: The microneedle with a length of 600 µm, an outer diameter of 360 µm, an inner diameter of 125 µm and a bevel angle of 40 ° is placed on a cylindrical bench with an outer diameter of 600 µm and a length of 800 µm. (b) The layout of the microneedle-array is a square with a side length of 2,3 mm.

Figure 4.2. Fabrication of the PDMS-mould. Left: The master-carrier (brass) with the master- needles (steel) is placed exact opposite to the PDMS-mould-carrier (brass) with the guide- sleeves (steel). The cavities of the master-needles and the guide-sleeves are filled with a re- movable wire (tungsten). Right: PDMS-mould consisting of PDMS-mould-carrier (brass), guide sleeves and the casted PDMS.

Fabrication of the mould started with the fabrication of a master carrier with master needles and a mould carrier with guide sleeves (Figure 4.2, left). Master needles and guide sleeves were manufactured on a wire EDM turning machine. Stainless steel capillaries with an outer diameter of 1/16” (1.59 mm) and an inner diameter of 125 µm were used to fabricate these structures. Four master needles and the guide sleeves were fixed with cyanoacrylate glue in the master carrier and the mould carrier according to the layout in Figure 4.1, right.

In the next step the mould is cast from the master needles. To accomplish this task the master carrier and the mould carrier were arranged exactly opposite in a tray. Tungsten wires with an outer diameter of 125 µm were used to fill the lumen of the four master needles and the corresponding guide sleeves (Figure 4.2). The space between the master carrier and the mould carrier was filled with polydimethylsiloxane (PDMS). After the curing process (30 min. at 80 °C) the tungsten wires and the master carrier were removed (Figure 4.2, right). An additional post bake (2 h at 200 °C, preferably in vacuum) is necessary to get rid of dissolved volatiles in the soft mould. They can cause dome shaped bumps with a diameter of several mm on the polymer surface. The first embossed structures can be destroyed by these bumps.

The PDMS used in this work (Elastosil RT 607 A/B) is a room temperature curable grade with above-average thermal conductivity (0.4 W m-1 K-1). The crosslinking is a platinum-complex activated addition curing process that can take place at temperatures from 10 °C to 200 °C. Therefore, no by products are formed during vulcanisa- tion. Nevertheless, the cured PDMS polymer typically contains up to 2% volatiles [105].

Table 5. Thermal expansion coefficients of the various materials of the soft embossing tool.

Thermal expansion Material coefficient (10-6 K-1) steel 12 brass 18 tungsten 5 polycarbonate 65 PDMS 200-400, typically 300

43 The thermal expansion coefficients of the various materials used to manufacture the mould are listed in Table 5. The thermal expansion coefficient of PDMS is not known exactly but it is one order of magnitude larger than the others. So, it is obvious that compressive stress is going to build up in the soft mould if the temperature of the soft moulding tool increases. Consequently, the cavities of the microneedles decrease in size resulting in a smaller outer diameter of the replicated microneedles. To reduce this effect the PDMS is cut at its circumference. The following equation was used to calculate the unconfined lateral increase of the PDMS:

(4.1)

is the thermal expansion coefficient, is the length of a side of the PDMS mould at room temperature and is the temperature difference. With of 50 mm and 70 mm for the short and the long side, an average thermal expansion coefficient of 300·10-6 K-1 and a temperature difference of 170 °C the lateral increases are 2.6 mm and 3.6 mm. Therefore, the short sides were cut by 1.3 mm ( ) and the long sides by 1.8 mm. No action is required in direction of the mould pressure. Due to the force controlled embossing process the PDMS can expand in vertical direction until the programmed pressure level is reached.

Preparation of the soft embossing tool for the soft embossing process was performed in the following way. A sheet of silicone was placed on the bottom of the tray (Figure 4.3). A guide plate is placed on top of it. Next, a sheet of thermoplastic material (polycarbonate, PC) with drilled (diameter 1 mm) and deburred holes is placed on the guide plate in the tray. The soft mould on its carrier plate is placed upside down on the thermoplastic material. Tungsten wires were inserted through the guide sleeves. The conical structures in the guide plate centred the tungsten wires before they were fixed in the silicone sheet at the bottom of the tray. The wires can move free in the guide sleeves with no external longitudinal forces acting on them. By placing the piston on top of the stack the soft embossing tool is ready for the soft embossing process. It can be transferred to the hot embossing machine.

Figure 4.3. Soft embossing tool prepared for the embossing process (not to scale). The stack in the tray starts with a PDMS-sheet, which holds the wires (tungsten) in place. It is followed by: the guide-plate (brass) to keep the wires centred, a plate of thermoplastic material (polycarbonate) prepared with drilled and deburred holes and the PDMS-mould mould side down. The stack is finished by the piston (brass).

44 4.1.2. Soft embossing process

The hot embossing machine used in this work is equipped with two yokes in a recipient. The yokes can be heated by oil and cooled by heat exchange water. The lower yoke can be moved against the upper yoke with a hydraulic cylinder. The oil temperature, the position of the lower yoke and the force of the cylinder are controlled by a programmable controller.

The soft embossing sequence is a time controlled program that activates the actuators (heating, heat exchange water pump, hydraulic pump, vacuum pump) of the hot embossing machine. For every step of the program temperature, force or yoke position, time and recipient pressure must be defined.

After the hot embossing tool is placed on the lower yoke and the recipient is closed the program can be started by the user. Figure 4.4 shows the chronological sequence of the parameters temperature and force. At the beginning of the first step (tB to tO) the recipient is evacuated, the heating starts to warm up the oil and the lower yoke moves up until a pre force FP of 0.1 kN is built up. The hot embossing machine is in the force controlled mode.

This pre force corresponds to a low pressure of approximately 3 N/cm² (7 cm x 5 cm piston area). This pre pressure is practically not relevant. However, it ensures the contact of the upper yoke with the embossing tool which is very important for a defined flow of heat from the lower and upper yoke into the embossing tool. The first step is finished when the oil temperature reaches the desired value TE (reasonable values 190 °C to 240 °C). With a temperature increase of approximately 10 °C/min it takes roughly 20 minutes.

The second step (tO to tE) is an idle time. Although, the oil temperature is already at operating temperature the thermoplastic polymer (polycarbonate) in the soft embossing tool is still below its forming temperature. An experimentally determined idle time (see Appendix B) of 20 minutes is necessary for this embossing tool to heat up the thermoplastic polymer (polycarbonate) to its forming temperature.

Figure 4.4. Chronological sequence of the parameters temperature and force during a hot embossing experiment. The solid line represents the programmed sequence. The dashed line represents the assumed values of the parameters in the soft embossing tool. TA: ambient temperature, TE: embossing temperature, tB: begin of sequence, tO: operating temperature of machine reached, tE: begin of soft embossing, tC: begin of cooling phase, tF: sequence finished.

45 Table 6. Typical parameters of the hot embossing process.

Time / min. Temperature / °C Force / kN

tB 0 TA 25 FP 0.1

tE 45 TE 190 / 215 / 240 FE 1.5 / 3 / 4.5

tC 50 / 60 / 70

tF 80 / 90 / 120

At this stage (tE to tC) the force is switched to the embossing force FE (1.5 kN to 5.1 kN). It is the actual soft embossing step performed with durations ranging from 1 min. to 29 min.

During the last step the embossing tool and the machine are cooled down to ambient temperature TA. At a cooling speed of approximately 10 °C/min it takes at least another 20 minutes for the cool down procedure.

Cooling of the yokes of the embossing machine results in cooling of the soft embossing tool. Of course, the cooling of the thermoplastic material lags relative to the embossing tool. As the temperature of the thermoplastic material decreases the difference in the thermal expansion coefficients between the thermoplastic polymer (polycarbonate), PDMS and brass results in lateral polymer flow between the inter- faces polycarbonate/brass and polycarbonate/PDMS. To prevent this lateral polymer flow as far as possible it is necessary to maintain the embossing force until the polycarbonate temperature dropped approximately 30 °C below its glass transition temperature (145 °C). Additionally, the thermoplastic material shrinks on the tungsten wires. Therefore, it is necessary to remove them with a gripper after disassembling of the embossing tool.

Altogether the soft embossing sequence runs roughly 1.5 hours. 18 microneedle arrays were moulded with the parameters listed in Table 6. The parameters tc, TE and FE were varied within the given ranges. 4.1.3. Results & Discussion

A picture of a demoulded microneedle array is presented in Figure 4.5, left10. The orientation of the microneedles in the array is arbitrary because no special structures for rotational alignment were realised. The roughness of the microneedles results from the wire EDM turning process used to manufacture the master microneedles. Burrs were observed at the outlet of the microneedle vias. Although, removed as good as possible they can be easily identified in Figure 4.5, left. Originally, the length of the burrs was up to 1 mm. Obviously the fit between the PDMS and the tungsten wires is not tight enough. One solution can be to skip cutting the PDMS. This would lead to larger stress, hence a tighter fit between PDMS and wires. In this case the size reduction of the microneedles can be compensated by designing larger master needles.

10 Embossing experiments and data acquisition were performed by van Goer M under the supervision of Müller C, Schön H and Vosseler M [107].

46 All 18 fabricated microneedle arrays showed one to three unambiguously off centred lumina (Figure 4.5, right). The fact, that this is a random effect supports the argument, that there is no principle alignment problem. Rather, it is caused by bending tungsten wires. This results from a tight fit between the wires and the PDMS. This fit is formed during the heating and the simultaneous expansion of the PDMS mould. At the same time the tungsten wires are straightened due to the vertical elongation of the PDMS mould. However, during the cooling phase the PDMS mould shrinks. If the tight fit between PDMS mould and tungsten wire does not open the tungsten wire is bent. Apparently, this is a stochastic process. The dilemma is, that on the one side the tungsten wires should be able to move freely in the vertical direction. On the other side there must be a tight fit to prevent the formation of burrs.

The length of the master needles and the length of the hot embossed microneedles were measured with a stereo microscope (Zeiss Axiotech vario ) combined with a length gauge (Heidenhain MT 60K). A single length measurement was performed in the following way to compensate wedge errors:

 the surface of the substrate was brought into focus at a point of the circumference of the round socket of the microneedle,  the length gauge was set to zero,  the relative height of a second, third and fourth point at the circumference with 90° distance were measured,  the relative height of the microneedle tip was measured,  the length of the microneedle was determined by calculating the mean of the four points at the circumference and subtracting it from the relative height of the microneedle (The error due to the off-centre tip is neglected).

Figure 4.5. 2x2 Microneedle arrays manufactured by hot embossing. Left Oblique view. Right: Top view of a microneedle array with one unambiguously off centred via (indicated by arrow) of a microneedle.

47 Table 7. Length of master and moulded needles. Each master needle was measured 5 times. Each moulded needle was measured 5 times (CV < 0.5%). The data in the table gives the mean of 18 experiments for moulded needles.

Difference Master needles Moulded needles moulded - master Length / µm CV Length / µm CV Length / µm 1 1384 ± 1 0.0 % 1509 ± 23 1.5 % 125 2 1379 ± 1 0.1 % 1430 ± 39 2.7 % 51 3 1425 ± 1 0.0 % 1439 ± 26 1.8 % 14 4 1382 ± 1 0.1 % 1457 ± 20 1.4 % 75 1-4 1392 ± 19 1.4 % 1459 ± 40 2.7 % 67

The length data of 18 experiments with varying parameters is given in Table 7. The CV of the hot embossed microneedles is very low (<3%) indicating good process stability. However, there is one experiment which produced needles with exceptional short lengths up to 9.4% shorter than the average length. This experiment was affiliated with a low embossing temperature of 190 °C and a short embossing time (tE to tC) of 5 min. So, the embossing temperature should be at least 215 °C or the embossing time should be at least 15 min.

The hot embossed microneedles are on average 74 µm longer than the master needles. This is caused by the elongation of the soft embossing mould due to thermal expansion.

One of the moulded microneedles (index 1 in Table 7) is exceptionally longer than the other microneedles. The reason for this effect is the varying lengths of the guide sleeves in the soft mould carrier. If forces are applied a shorter sleeve results in less compression of the soft mould. Hence, longer needles are generated. The length of the microneedles which were moulded with identical hot embossing parameters (TE=215 °C, tE to tC=15 min., FE=3 kN, 5 experiments) is plotted vs. run index in Figure 4.6. The microneedle length decreases by 16 µm ± 12 µm (1.1%) after 18 experiments. This is explained by shrinkage of the soft mould due to aging.

Figure 4.6. Needle height vs. run index.

48 The diameter of the microneedles was measured with a profile projector (Nikon V-12B ) combined with a geometric readout (Quadra-Check 200). After focusing the circumference of each needle four points on the circumference with a spacing of 90° were approached. The software of the geometric readout calculated the diameter by a best fit of the four data points to a circle. The average deviation from the calculated diameter of a single circle is 12 µm ± 10 µm. The average diameter of the master needles and the moulded needles is given in Table 8. The measured diameter of the master needles is 3.7% (22 µm) smaller compared to the specified diameter of 600 µm. The measured moulded diameter is 7.4% (43 µm) and 10.8% (65 µm) smaller compared to the measured diameter of the master needles and the specified diameter, respectively. There is also a trend in the data of the measured diameters of the moulded microneedles with identical parameters. The diameter increases by 6.3% (34 µm ± 7 µm) within these experiments. Again this effect is explained by shrinkage of the soft mould due to aging. The smaller mould results in larger diameters due to the larger thermal expansion of the soft mould.

4.1.4. Summary

With the developed moulding tool it is possible to manufacture microneedle arrays with a soft embossing process. The moulded microneedles are longer and thinner as the master needles. These effects were explained by the coefficient of thermal expansion of PDMS and the difference of coefficients of thermal expansion of PDMS and brass. The 18 performed experiments showed a trend in the data of the length and diameter of the microneedles which is explained by aging of the soft mould.

Although the hot embossing parameters were changed over a quite large parameter range the CV of the microneedle height and diameter is below 3%. Due to this result the experiments were not replicated and no further modelling was done. Overall the following parameter set can be recommended for soft embossing of microneedles arrays with the introduced mould geometry: TE=215 °C, tC=60 min. (corresponding to an embossing time of 15 min.) and FE=3 kN.

The realised mould concept is very well suitable for embossing structures with high aspect ratios. However, the microneedles showed burrs at the opening of the lumina. Additionally, one to three vias of the microneedles were positioned off-centre. So, further investigation of the effects and improvement of the mould is necessary before it can be used reliably.

Table 8. Diameter of master and moulded needles. The data represents the average of 4 master needles and 18 times 4 (n=72) moulded needles.

Needle diameter / µm CV master 578 ± 7 1.2 % moulded 535 ± 15 2.8 %

49 The problem of burrs can be solved by using a hard mould. Of course, with such a mould burr formation would be promoted. The hard material definitely won’t be able to generate a tight fit between the guide wires and the corresponding bearing. Flow of polymer into this gap is anticipated. Therefore, a post processing step is proposed. With a wafer saw it is possible to cut the microneedles to a desired length. With a suitable angle a bevelled tip can be generated. The feasibility of this approach is demonstrated in Figure 4.7. A very napless surface and quite sharp edges can be generated with this process. Additionally, it might be possible to solve the issue with off centred vias, too. By manufacturing a hard mould e.g. made in stainless steel one can focus on the position of the bearings for the guide wires. Additionally, there is no relevant change in dimensions when heated.

4.2. Injection moulding of hollow microneedle arrays

Manufacturing of injection moulded microneedles was a joint development of HSG-IMAT and HSG-IMIT. The project was financed by AiF e.V. The design of the microneedle array is presented in Figure 4.8. It was developed by Botzelmann T, Mayer V, Schilling P and Vosseler M. The corresponding mould concept was engineered by Botzelmann T, Mayer V and Schilling P. Microneedle arrays were generated with this mould by Botzelmann T and Schilling P. Figure 4.8, Figure 4.9 and Figure 4.10 were published in [108].

In principle, this design is quite similar to the one used for the soft embossing process. Major deviations are the conical shaped benches, the larger interval of 5.0 mm between the microneedles and the tip geometry. The needle consists of two cylinders in order to reduce the tip radius. The first one with a diameter of 0.32 mm is the major cylinder. It is the basic structure of the microneedle. The second cylinder with a diameter of 0.10 mm is off centred to the major cylinder. After generation of the bevel it forms a tip with a smaller radius. The penetration force in agar gel of a microneedle with this geometry (2.2 mN ± 0.4 mN, n=3) is less than a third compared to a microneedle without this feature (7.4 mN ± 0.8 mN, n=3). Therefore, it is supposed that microneedles with this feature can be easier inserted into the skin.

Figure 4.7. Moulded microneedles with sawn off tips to demonstrate the feasibility of a post processing step with a waver saw.

Figure 4.8. Dimensions of the microneedle array [108].

The small dimensions of the microneedle array and the high aspect ratio of the microneedle lumen are quite a challenge for the design of the corresponding injection mould. The realised concept based on an available milling process is presented in Figure 4.9. The mould is composed of several plates. The essential structure that shapes the negative microneedle array consists of two plates. With this approach it is possible to manufacture the mould by milling & drilling and theoretically obtain a perfect edge. Additionally, it is possible to generate the tip geometry mentioned above. First, a small hole defining the small tip radius is drilled. Second, the large hole defining the outer microneedle diameter is drilled, of course with a suitable shift of the centre of the holes. The microneedle lumen is generated by pin inserts. The pin inserts were manufactured by grinding of tungsten carbide. The length of the slim part with an outer diameter of 0.120 mm ± 0.001 mm is 2.15 mm. The pins are supported on the 3. plate of the mould if it is closed. Without support the pins would fail during the moulding process. This finding was obtained by simulation. The sprue is on the front side of the microneedle array. The ejectors work from the backside.

Figure 4.9. Left: 3 plate mould concept. The outer shape of the microneedles is formed by plates 1. and 2. The lumen of the microneedles is generated with a pin. It is supported in the 3. plate if the mould is closed. The 3. plate contains the ejector pins. Right: Mould presented in its closed state. The injected polymer is depicted in red. [108]

Figure 4.10. Left: REM pictures of a injection moulded microneedle array [108]. Right: Picture of an injection molded microneedle array mounted in a housing with tube.

More than 300 microneedle arrays were moulded on an injection moulding machine (Microsystem 50 by Battenfeld). A high melt flow rate polymer (styrene-acrylonitrile resin) was used. REM pictures of one of the first moulded arrays are presented in Figure 4.10, left. All four microneedles are present, with sharp edges and with through-going lumina.

For the pressure driven intradermal infusion experiments it was necessary to glue the microneedle arrays into an adapter (PMMA) that is connected with a tube (1/16” FEP Figure 4.10, right). Plasma activation before the gluing process (Epotek 301) was necessary to get a good bond quality. The design of the adapter and the bonding process were developed by Hiltmann K. 4.3. Intradermal infusion set based on the Mantoux method 4.3.1. Laboratory setup for ex vivo experiments

Contributions to the development of the laboratory setup: Vosseler M: concept of the penetration setup, Jugl M: concept, design and assembly of the laboratory penetration setup.

Penetration of a cannula almost parallel to the skin surface without stretching the skin generates wrinkles. In order to reduce those wrinkles together with all the complex failure mechanisms affiliated with them a wrinkle-free penetration mechanism is implemented as follows: By simply pushing a slider on a carrier plate, the cannula is inserted by a well-defined displacement x (Figure 4.11) at a well-defined angle of 10° into the skin. By releasing the slider the cannula is pushed back automatically by a second well-defined displacement (x-y) in order to relax the skin. Finally that mechanism results in a well-defined lateral penetration length w and vertical penetration depth zw of the tip of the cannula in the skin.

Figure 4.11. Left: Sketch of the laboratory setup for well-controlled insertion of a cannula into a skin specimen with a small defined angle of 10°. Right: Cannula at infusion position. Details explained in the text.

The laboratory setup is shown in Figure 4.11 and Figure 4.12. It consists of a vertically moveable carrier plate (a) made of clear PMMA. It can be locked several centi- metres above the base plate (b). A replaceable cannula (c) is fixed to a slider (d) which can move on two shafts (e) between the stop bar (f) and the carrier plate (a). In its initial state the spring (g) pushes the slider (d) towards the stop bar (f). The cannula is mounted into the slider (d) in such a way that the distance of its tip to the skin surface is zero when no forces are applied. In that position the distance of the tip to the bottom of the carrier plate (a) equals z1, which corresponds to the thickness of the stack of double side adhesive tape (i) and medical tape (j) (Hansaplast classic tape by Beiersdorf AG, Hamburg, Germany). The sticky side of the medical tape is directed downwards to fix the pig skin. The stack of tapes has a width of 25 mm with a circular hole of 4 mm radius around the cannula. This hole forms a bulge defining cavity. The skin deforms and can fill this cavity if forces are applied.

The maximum displacement x of the cannula during insertion is defined by the distance between the stop bar (f) and the carrier plate (a). The theoretical maximum vertical penetration depth of the cannula tip in the tissue during the penetration is z3, assuming that the skin bulges into the bulge defining cavity and is limited by direct contact with the bottom of the carrier plate.

Figure 4.12. Close-up photo of the laboratory setup for well-controllable insertion of a cannula into a skin specimen.

53 By releasing the slider the cannula is pushed back due to the spring (g). The final vertical penetration depth z2 of the cannula tip during the infusion experiments is defined by an exchangeable spacer (h) that is placed in between the slider (d) and the stop bar (f) before releasing the slider. z2 can be calculated to be (Figure 4.11):

(4.1)

By adjusting the stop bar (f) it is possible to vary the maximum displacement x of the cannula (d). The insertion parameters of all infusion / injection experiments are sum- marised in Table 9. Infusion experiments were performed with insertion parameter sets 3/1, 4/2, 5/3. x/y denotes the displacement sets x and y as depicted in Figure 4.11 and tabulated in Table 9.

For all experiments 31 gauge stainless steel cannulas with an outer diameter of 0.26 mm (31G), an inner diameter of 0.13 mm (min. 0.11 mm, max. 0.15 mm), a length of 30 mm and a bevelled tip (14°) were used. The length of the tip is 1.0 mm. The cannula was always mounted bevelled side up. Its fluidic resistivity was theoretically calculated according to Hagen-Poiseuille to 1.2 kPa h/ml. Hydrodynamic effects caused by the orifice of the cannula are neglected due the high aspect ratio (length per diameter) of 115. The experimental verification resulted in a fluidic resistivity of 0.9 kPa h/ml ± 0.1 kPa h/ml corresponding to an effective inner diameter of 0.14 mm. 4.3.2. Infusion set for in vivo experiments

Contributions to the development of the infusion set: Vosseler M: concept of the infusion setup, Kunze M: concept, design and assembly of the infusion setup.

The laboratory setup is very convenient and versatile. However, it is not suitable for in vivo experiments. Therefore, a second design based on a single set of insertion parameters was created (Figure 4.13). The functionality is the same. By pushing a slider to a stop the cannula is placed in the skin with a defined mechanism and wrinkles are prevented by the slight retraction of the needle. A commercial cannula (30G, BD, Franklin Lakes, NJ, USA) with luer connector can be mounted in the device.

The length and width of the supporting plate are 90 mm and 30 mm. The device can be mounted on the skin by adhesive tape. The insertion and retraction mechanism is integrated in the bearing support. The selected insertion parameter set is 4/2. z2 was measured with a profile projector and determined to 0.84 mm ± 0.10 mm (n=15). It corresponds well with the expected value of 0.80 mm.

Table 9. Sets of insertion parameters of the injection and infusion experiments. x is the maximum displacement of the cannula (c) in Figure 4.11. y is the final displacement of the cannula for infusions / injections. z1 is the depth of the bulge defining cavity and defined by the thickness of the stack of double side adhesive (i) and tape (j). z2 and z3 are the vertical distance of the cannula tip to the bottom of the carrier plate. They are defined by the displacements y and x, respectively.

x / mm y / mm z1 / mm z2 / mm z3 / mm 3 1 0.45 0.62 0.97 4 2 0.45 0.80 1.14 5 3 0.45 0.97 1.32

Figure 4.13. Tool to perform in vivo intradermal infusion experiments.

56 5. Ex vivo and in vivo intradermal infusion experiments

Almost all paragraphs of chapter 5.1 were published as original research paper (publication P1, see list of publications) in Pharmaceutical Research in 2011 (Volume 28, pp 647-661). Contributions to this publication: Vosseler M: literature search and analysis, overall concept and concept of the penetration setup, planning of all experiments, almost all experimental work, all data analysis, manuscript preparation. Jugl M: concept and design of the laboratory penetration setup depicted in Figure 4.11, Figure 4.12 and Figure 5.2. Zengerle R: scientific advice during experiments and manuscript preparation.

The purpose of the intradermal infusion experiments was to get more information on the infusion process. The most relevant issue was to identify the back pressure as a function of flow rate. This data is important for the selection of a suitable actuator for the chronopharmaceutical drug delivery device. Additionally, it is very important to know if there is leakage and the likeliness of its occurrence. As a side effect of the performed experiments the distribution of solvent and a solute was observed. This data may be helpful to get a rough estimate of the distribution of a specific drug molecule in the skin. Of course, one must consider the molecular difference between the drug molecule and the solute solvent combination used in this work.

The data in this chapter was acquired with the laboratory setup, the infusion set and the microneedle arrays introduced in the preceding chapter. Excised pig skin was used as a model for human skin in the ex vivo experiments. Pigs were used as animal model for man in the in vivo experiments. The ex vivo experiments generated an extensive set of data that was checked with few in vivo experiments. So, the intradermal infusion and injection processes were characterized in an efficient manner. In this case injection experiments refer to a process duration of 10 s and infusion experiments refer to a duration of up to 4 h. 5.1. Ex vivo experiments with the laboratory setup 5.1.1. Pig skin

Excised pig skin was used in the experiments because of its easy availability. It shows similarity to human skin in terms of morphology and permeability characteristics [109]. However, there is a higher density of collagen and other fibre bundles in the reticular (lower part of the) dermis [73] that results in reduced skin elasticity and increased resistance to skin deformability.

Non blanched pig ears were obtained from a local slaughterhouse. The ears were cleaned with fresh water. Skin specimens were prepared by removing the skin from the backside of the ear with a scalpel and punching of discs with 30 mm diameter. The average weight of the skin specimens, is 1.36 g ± 0.34 g. Ears and skin specimens were stored in a freezer at -18 °C until needed. The average thickness, measured in the frozen state, is 2.51 mm ± 0.45 mm resulting in a specimen volume of

57 1.77 ml ± 0.32 ml. No ear and no skin specimen were stored longer than three months. The skin was not shaved. 5.1.2. Solutions for injection and infusion

Infusion / injection experiments were performed with water (vehicle) based solution containing 0.90 wt. % NaCl and 0.02 wt. % gentian violet (solute) for staining. A second solution based on PEG 200 (vehicle) with 0.02 wt. % gentian violet (solute) was prepared, too. All chemicals were obtained from Merck KGaA, Darmstadt, Germany.

PEG 200 is a Newtonian fluid11. It is completely miscible with water and octanol [110], it is chemically inert and has low toxicity. Its viscosity was determined to be 55 mPas. PEG 200 is a low molecular weight (200 g/mol) species with diol characteristics. It has a high vascular permeability compared to high molecular weight PEG such as PEG 30.000 [111].

Gentian violet is positively charged in neutral aqueous solutions, has a molecular weight of 408 g/mol and an octanol water partition coefficient of log P = 0.51. Gentian violet can bind to proteins e.g. Human Serum Albumin [112] and human cells e. g. epidermal cells [113]. Its maximum solubility is 1.1 g in 100 ml of water and 13.9 g in 100 ml ethanol. It’s solubility in PEG 200 is good due to the diol characteristics of the low molecular weight species. Therefore, gentian violet in PEG is less staining than gentian violet in water if it is used e. g. in the eradication of bacteria afflicted skin [114]. 5.1.3. Fluidic setup

The fluidic setup consists of the cannula, the tubing, a flow sensor, a pressure sensor and a syringe made of glass. To keep the fluidic capacity of the whole setup as small as possible stiff FEP (fluorinated ethylene-propylene) tubings were used. FEP is a transparent material, so bubble free priming can be controlled easily. The fluidic resistance of the tubing (outer diameter: 1.6 mm, inner diameter: 0.8 mm) per centime- tre is four orders of magnitude smaller compared to the fluidic resistance of the cannula. Therefore, it can be neglected. Glass syringes (10 ml volume for infusion experiments, 0.1 ml and 0.5 ml volume for injection experiments) were used because of their negligible fluidic capacity. A laboratory syringe pump (MDSP3f by MMT Micro Mechatronic Technologies GmbH, Siegen, Germany) was used to enable flow controlled liquid delivery in the infusion experiments. The glass syringe was driven manually in the injection experiments. A metronome was used for the timing of this manual injection process. The flow rate in the tubing was monitored with a flow sensor (Liqui-Flow by Bronkhorst, Ruurlo, Netherlands) with a maximum measurement range of 5 ml/h. Depending on the expected back pressure one out four pressure sensors (two different sensors of type EPX by Entran Sensors, Potterspury, United Kingdom; PDCR 200 by GE Sensing, Billerica, USA and CTE8000 by SensorTech- nics, Puchheim, Germany) was integrated in the setup to measure the back pressure generated by the needle and the tissue. The measurement ranges of the sensors are 100 kPa, 350 kPa, 600 kPa, and 1000 kPa, respectively. The selected sensor was mounted downstream (after the flow sensor) and the pressure offset was levelled to

11 Overview Brochure Carbowax PEG, The Dow Chemical Company, Midland, Michigan, USA

58 zero before the infusion took place. This way, the pressure difference along the needle plus the tissue was recorded.

Based on the above statements the following fluidic model for the infusion / injection experiments is assumed (Figure 5.1). It is a parallel circuit of a fluidic capacity and a fluidic resistor. The experimental setup is modelled as a capacity (Csetup), neglecting its fluidic resistivity. The cannula is modelled as a constant resistor (Rcannula). Its capacity is neglected due to the high stiffness of metal tubing. The tissue is also modelled as a resistor (Rtissue). This assumption is checked in the results section. The flow q, generated by the syringe pump or by pressing the syringe generates the pressure drop p. 5.1.4. Injection and infusion experiments

To perform an infusion or injection experiment, the carrier plate was locked well above the base plate. A piece of artificial skin (skin suture trainer by 3B Scientific, Hamburg, Germany) was wrapped in cling wrap and placed on the base plate. A pig skin specimen was placed on top of it and the carrier plate was lowered down until it got in contact with the skin. Immediately, the skin specimen stuck to the sticky tape at the bottom of the carrier plate. The cannula was inserted in the skin according to the method described before.

After the insertion of the cannula and before starting the experiments, the carrier plate with the skin specimen attached to it was lifted again and locked in its upper position. This way, there was no mechanical pressure applied on the skin during the infusion / injection experiments.

Infusion experiments were performed with all three sets of insertion parameters listed in Table 9. A volume of 0.20 ml and 1.00 ml was infused in 2 hours at flow rates of 0.10 ml/h and 0.50 ml/h, respectively. Experiments with the high viscosity PEG solution were done with both flow rates but only with the insertion parameter set x = 4 mm and y = 2 mm.

Injection experiments were always performed with insertion parameters x = 4 mm and y = 2 mm. Only the water based solution was used for the injection experiments. Two different sets of injection experiments were performed. In the first set 0.10 ml of volume was injected in 10 s (average flow rate: 36 ml/h). In the second set a volume of 0.50 ml was injected in 10 s (average flow rate: 180 ml/h).

Figure 5.1. Assumed model for the intradermal infusion / injection experiments.

59 For an infusion / injection experiment with a fixed set of insertion parameters at least nine skin specimens were prepared from skin of three different pigs. At least three skin specimens came from the same individual pig, respectively. All together 92 experiments were performed: 56 infusion experiments and 18 injection experiments with the water based solutions as well as 18 infusion experiments with the PEG based solutions.

To estimate the diffusion of the dye in the skin 0.50 ml of aqueous solution and 0.10 ml of PEG solution were injected in skin specimens. The spot area was measured after injection and 2 hours later.

Leakage was detected and quantified by the following method. If any leakage was observed a droplet of infusion / injection solution with a volume of 5 µl was placed on top of the skin with a microliter syringe and a cannula. The stained area of the leakage was compared to the stained area of the 5 µl droplet. In this way it was possible to estimate leakage to be below or above 5 µl.

The clear PMMA material of the carrier plate enabled taking photos of the skin specimen and the cannula (c) by placing it under a microscope with a camera attached to it. Photographs of the cannula (c) were taken in its infusion / injection position with and without skin. Comparing the two pictures, it was possible to experimentally determine the lateral penetration length w of the cannula. After removing the cannula photographs of the infusion spot and its cross section were taken with a cali- brated microscope / camera combination. The flow sensor and pressure sensor data were recorded during infusion with a sample rate of 10 Hz. 5.1.5. Penetration Results

For the penetration experiments the carrier plate (a) was lowered onto the skin specimen (k) (Figure 4.11, Figure 5.2) resting on the artificial skin (l). The skin bulged into the bulge defining cavity due to the weight of the carrier plate ( 500 g). This cavity is defined by the stack of double side adhesive tape (i) and medical tape (j). A notch (Figure 5.2) formed around the cannula (c) which made it easy to penetrate the skin.

Figure 5.2. Penetration setup resting on pig skin (k) and artificial skin (l). Bulging of the skin into the bulge defining cavity is shown.

60 The cannula insertion process is presented in Figure 5.3. From the pictures it gets clear, that the skin deforms prior to tip penetration. So, the maximum lateral tip penetration is not as deep as theoretical calculations suggest. Although the main effect of the retraction function is to regress the skin deformations some effective retraction of the cannula cannot be excluded.

Skin deformations might be prevented if there is no gap in the medical tape. In this case the skin infusion / injection site is no longer observable which is probably of no concern in the final end user device.

The vertical penetration depth zw (Figure 4.11) of the cannula tip at the infusion / injection position in the skin was determined indirectly to be:

(5.2)

with the measured lateral penetration depth w (Figure 4.11, right), the outer diameter d of the cannula (c) and a penetration angle of 10°. The vertical penetration depths vary from 0.47 mm to 0.70 mm depending on the penetration parameters (Figure 5.4) (ANOVA, p  0,001). These absolute values are based on indirect measurements. However, it can be stated with very high probability that the cannula tip penetrates into the dermis. This statement is based on the assumption that the thickness of the epidermal layer is 0.20 mm. Although, there is some uncertainty in the absolute values the relative values are supposed to be accurate.

Penetration of a cannula into the dermis is associated with stimulation of nerve end- ings. So, the insertion process of the cannula is not pain free. If it is less painful compared to the insertion of a subcutaneous infusion set needs to be studied. Pain sensation during infusion / injection is a combination of pain caused by the introduction of fluid into the skin and chemical stimulation e.g. by preservatives. Injections of 0.12 ml of fluid causes faint pain [73].

Figure 5.3. Cannula insertion process. Left: The cannula tip is in its initial position on top of the skin. Middle: The cannula was pushed with the slider a distance of 4 mm at an angle of 10°. The tip punctured the skin. The deformation of the skin is indicated by the black line on the skin. Right: The slider is retracted by 2 mm. The cannula tip is in its infusion / injection position. The deformation of the skin is regressed.

Figure 5.4. Vertical penetration depths zw of the cannula tip into the skin specimens determined by indirect measurements (see text). Each bar shows the mean and the standard deviation of at least 18 penetration experiments based on different skin storage conditions (fresh, cooled, frozen). At every set of insertion parameters (defined in chapter 4.3.2 and listed in Table 10) the samples came from at least three different pigs.

The original expectations regarding the vertical penetration depths were based on the assumption that the skin bulges completely into the bulge defining cavity and is in direct contact with the bottom of the carrier plate when the tip of the cannula starts penetration of the skin as illustrated in Figure 5.2. This would lead to an expected vertical penetration depth of z2 as illustrated in Figure 4.11. Comparing the indirect experimental results with those expectations it is found that the experimental values of vertical penetration depths were approximately 0.23 mm less than expected (Table 9) which equals roughly half the depth of the bulge defining cavity.

No relevant influence between different individual skin samples or different storage methods (fresh, cooled up to two days, and frozen) on the vertical penetration depths could be identified. However, in one out of 92 experiments it was impossible to penetrate the skin with the cannula and in three out of 92 experiments two or three at- tempts were necessary to place the cannula successfully in the tissue. Yet, all four out of the 92 experiments that created difficulties were affiliated with skin specimens that were frozen and defrosted several times. 5.1.6. Leakage

A very important prerequisite for clinically acceptance of hollow microneedles is leak tightness to allow delivery of consistent amounts of drug in a well-defined manner. In these tests 83 out of 92 experiments were completely leakage-free. For the remaining 9 cases three different leakage failure modes were identified (Table 11).  Leakage mode 1 (occurred 2 times within 92 experiments): That failure mode results from hair follicles that were damaged during the insertion of the cannula. In that case the stained fluid was leaking at the location of the hair follicle and could be detected on the skin surface. There was no leakage at the puncture site. That leakage mode is expected to depend on the hair follicles density which is quite high for skin from pig ears. It is expected that this failure mode would happen less often in case of penetration of human skin. It is also expect that due to proper selection of the application site the frequency of this failure mode could be reduced to an insignificant level for humans.

62 Table 11. Overview of all experiments done to investigate leakage behaviour. Failure modes and failure frequencies are discussed in the text.

Infusion Injection Water based PEG based Water based 0.10 0.50 0.10 0.50 180 # of ml/h ml/h ml/h ml/h 36 ml/h ml/h incidences No leakage 24 26 9 8 9 7 83 Failure mode 1 1 0 0 0 0 1 2 Failure mode 2 0 1 0 0 0 1 2 Failure mode 3 2 2 0 1 0 0 5 Experiments 27 29 9 9 9 9 92

 Leakage mode 2 (occurred 2 times in 92 experiments): In that failure mode a liquid volume of more than 5 µl, could be detected on the skin surface. In all those cases only clear liquid without staining (0.02 wt. % gentian violet, molecular weight of 408 g/mol) came out of the skin. In clinical infusion experiments it would have to be determined if in this leakage mode there is really a leakage of therapeutic relevant molecules or if only a small fraction of solvent or interstitial fluid without drug molecules is leaking. This investigation has to be done for every specific drug.

It is important to mention that the leakage occurred at sites close to the puncture site but not at the puncture site itself. Injections based on the traditional manual Mantoux method (with a 31 gauge cannula) were performed to investigate this in more detail. Also in those injections the same leakage mode occurred.

It was checked if this failure mode was affiliated with variations in the back pressure that occurred during the infusion. No significant differences between the experiments with and without leakage could be identified. There is no relevant influence of the individual skin samples or storage methods.

 Leakage mode 3 (occurred 5 times in 92 experiments): In that failure mode less than 5 µl of clear liquid without staining could be detected on the skin surface. As the leaking volumes are less than 2% of the delivered liquid volumes this is regarded to be not relevant.

In conclusion, the presented intradermal tool enables leakage-free drug delivery when an application site with low hair density is properly selected. In the remaining leakage modes 2 and 3 only clear liquid without staining was detectable. It is very likely that this way no therapeutically relevant drug molecules are lost. Of course this has to be finally confirmed in further experiments with humans and by using the drug of interest.

The penetration of the skin with a hollow cannula at an inclined angle enables intradermal infusions / injections with very good leakproof properties. Another device which also demonstrated leakproof intradermal delivery is described in [73]. It consists of a single 30 gauge (outer diameter 0.31 mm) cannula which protrudes 1.5 mm of a specifically engineered structure. This structure was optimised to maximise the contact with the skin and minimise fluid leakage. The cannula is inserted perpendicular to the skin surface. During the injection it is necessary to keep light contact of the

63 syringe tip in the skin’s surface probably in order to realise the leak proof properties of the specifically engineered structure. This device is definitely optimised for injection applications.

With the presented inclined penetration method and the traditional Mantoux method no special structure is necessary which maximises skin contact and minimises leakage. It is not necessary to press the structure against the skin which is prone to result with time in pain sensation (pressure to high) or leakage (pressure too low). There- fore, the proposed intradermal interface is best utilised in long term (several minutes to days) applications. 5.1.7. Spot formation

The spreading of the vehicle and solute of injected and infused solution in the skin is a combination of pressure driven flow and diffusion with potential solute-vehicle, vehicle-skin and solute-skin interactions. The extent of diffusion was determined by measuring the spot area of injected liquid (0.50 ml of stained water and 0.10 ml of PEG solution) after injection and after 2 h. Within that time the spot area caused by injected stained water increased by 15 % ± 9 % (n=3) and the spot area of injected PEG solution increased by 30 % ± 14 % (n=3). The cross sections of the spots were also measured and the data was consistent with the surface data (see below). Spreading of the stained spot in the skin by diffusion is only a minor effect if water is used as vehicle. However, if PEG is used as vehicle spreading of the stained spot by diffusion is substantial.

The spot formation at the surface of the skin during an experiment with insertion parameters 5/2 and a flow rate of 0.50 ml/h of stained water is presented in Figure 5.5. The three pictures were taken at the beginning, after one hour and after two hours respectively. The spot geometry was approximated by an ellipse. The area of the ellipse increased 62 % during the second hour of the experiment. The increase is mainly caused by the pressure driven infusion. This statement is supported by the following observations. The penetration point of the cannula is very close to the circumference of the spot and the spot is directed away from the cannula opening. This observation is in agreement with previous observations in [41]. Finally, the minor extent of diffusion was already determined in separate experiments.

Figure 5.5. Formation of a blue stained spot (top view) during a two hour infusion of 0.02 wt. % gentian violet in 0.9 % NaCl solution at 0.50 ml/h into the intradermal compartment of ex vivo pig skin. Left: before the infusion started, middle: after one hour, right: after two hours.

Figure 5.6. Cross-sections of pig skin specimens after 2 h of infusion at a flow rate of 0.10 ml/h and insertion parameters 4/2. Left: water based solution. An ellipse with vertical (minor) and horizontal (major) axes was drawn manually to describe the cross-section of the spot as good as possible. Right: PEG based solution. In vertical direction the skin specimen is stained completely below the tissue surface resulting in an almost perfect rectangular spot in the cross- section.

The cross section of skin specimens after infusion of water based solution and PEG based solution at a flow rate of 0.10 ml/h and insertion parameters 4/2 is presented in Figure 5.6. The maximum vertical depth where gentian violet can be found seems to be limited to the dermal layer at a depth of approximately 1.5 mm if water is used as vehicle. The dashed line, following the skin surface, indicates a dome shaped wheal which formed due to the injection of stained water. The dimensions of the wheal are larger than the dimensions of the stained spot. Obviously, the dye was separated from the water molecules. Unfortunately, the size of the wheal could not be determined because determining its circumference is very difficult.

With PEG as vehicle the maximum vertical depth where the dye can be found extends to the fatty tissue underneath the dermal layer. The dashed line, following the skin surface, is flat. The stained skin site felt harder compared to non stained skin sites. This increase in hardness is due to the higher viscosity of PEG. Due to the fact, that only stained skin sites felt harder one can conclude that there is none or only minor separation between the dye and PEG.

For statistical analysis the spots and the cross sections of water based infusions / injections were approximated by ellipses. The variations of stained tissue areas monitored from the top with delivered liquid volumes are given in Figure 5.7, left for water and PEG based solutions. The plot shows data obtained with the insertion parameter set 4/2. The size of the stained spots caused by the water based solutions increases linearly (R² = 0.98) from 3.5 mm² to 15.4 mm² with the delivered volume. This is a lateral increase of the stained area of approximately 440 %, affiliated with a factor 10 between infused / injected volumes. The stained tissue areas monitored from the top caused by the PEG based solutions are almost one order of magnitude larger compared to the spots caused by the water based solutions.

Investigations to check if the lateral and/or vertical penetration depths of the stained areas are influenced by the insertion parameter sets were performed. Water based solutions with the insertion parameter combinations 3/1, 4/2, 5/3 was used. A trend could be identified that the area of the spot viewed from the top was maximum for the 4/2 insertion parameter set, but the differences were small and finally not significant

65 (data not shown). The maximum vertical penetration depth of the stained area was determined by the minor axis of the stained cross sectional ellipses. Data for insertion parameters 4/2 and variable infusion volumes are shown in Figure 5.7, right. Even when the infused volume varies by a factor of 5 the maximum penetration depths of the stained tissue increases only from 0.75 mm to 1.18 mm which is an increase of about 60%.

Further analysis of the lateral extension of the tissue staining from the cross sections of the skin specimens was performed. The major axis of the ellipse of the stained cross section area was used as a measure for the maximum lateral extension of the tissue staining. The analysis based on the cross section view of the sample is completely consistent with the top view analysis.

Unfortunately, it is not possible to compare the size of wheals resulting from infusions with the two vehicles. The reason is lack of data for wheals obtained by infusion of aqueous solution as already mentioned. So, it is impossible to determine if the area of skin wheals depends on the vehicle.

Due to the infusion / injection of stained water into the skin specimen liquid escaped from the skin. Typically it appeared at the circumference of the skin specimen. It was transported from the infusion / injection site by the capillaries as indicated by Figure 5.8. It seems that water prefers to spread in the dermis partly forming a dome shaped skin wheal and partly entering the capillaries instead of substantially penetrating into the hypodermis (vehicle-skin interaction). At the same time the mainly pressure driven distribution of gentian violet with water as vehicle is also confined to the dermis. So, it is concluded, that the introduced stained water spreads in the dermis before it enters the capillaries and finally escapes the skin specimen at its circumference. The escaping liquid is almost clear. There are several reasons. As already mentioned the water molecules separate from the dye and spread more in the dermis. So, there are not only capillaries that take up water from the stained site but also capillaries that take up “filtered” water. This results in a dilution effect. Additionally, the dye can bind to the tissue (solute-skin interaction) and finally there is some mix- ing with interstitial fluid which probably can be neglected due to the total introduced volume ranging from 0.1 ml to 1.0 ml.

Figure 5.7. Left: Spot size of intradermal infusions and injections into ex vivo pig skin, depending on delivered volume. Data of experiments with water and PEG based solution is given. Each data point represents 7 to 11 experiments. Right: The maximum vertical penetration depth of the staining in the tissue. Data of infusion/injection experiments with water based solution and insertion parameters 4/2 is shown.

Figure 5.8. Backside of a skin specimen after infusion of stained water with a flow rate of 0.10 ml/h and insertion parameter set 3/2. The capillaries turned blue and conducted liquid from the infusion site to the circumference of the skin specimen, indicated by the arrow. With PEG as a vehicle the staining is not limited to the dermis and can also be found in the fatty hypodermis. This is due to the good solubility of PEG in polar and non- polar solvents. Capillary transport was also observed. However, the liquid escaping the skin sample was more stained compared to the case with water as vehicle. This is due to the fact that there are almost only capillaries that take up PEG from stained sites. Additionally, with PEG as vehicle the affinity of gentian violet to bind to the tissue is reduced due to the good solubility in this vehicle. This is an indication for solute-vehicle interactions.

None of the experiments provided any hint that the spot size would depend on skin samples of different pigs or different storage methods. Only the data of the skin of two out of 23 pigs showed statistically different results (ANOVA, p<0.05).

To finally achieve systemic drug delivery by intradermal administration it is necessary that the drug is taken up by the dermal capillaries. The efficiency of that route depends inter alia on the affected area within the highly capillarised dermal layer. The results clearly show that the lateral distribution of gentian violet with PEG as vehicle covers approximately ten times larger areas compared to water based solutions. Ad- ditionally, the gentian violet spreads into the hypodermis with PEG as vehicle.

There are pros and cons of using PEG based solutions: An advantage is the fact that by just using one individual cannula a quite large area within the dermis can be loaded with drug. In contrast a potential risk in using PEG based solutions is the fact that potentially some drug delivery to the blood stream is delayed when the drug solution penetrates into the hypodermis and deeper tissues as it was illustrated in Figure 5.6, right. For water based solutions a single microneedle can stain a tissue area monitored from the top from 3.5 mm² ± 2.7 mm to 15 mm² ± 4.4 mm², depending on the infused volume.

It is shown that a solute can be separated from its vehicle during pressure driven infusion / injection e. g. by binding of the solute to the tissue. Drug molecules probably show less tissue binding compared to a dye like gentian violet. So, the affected skin area of such drugs by pressure driven intradermal infusion / injection are supposed to be larger compared to the values given above.

67 Three major factors need to be considered in designing microneedle arrays. One major factor is the rate of absorption which is supposed to be higher the larger the affected skin area is. So, the higher the desired absorption rate the more microneedles are necessary. Another major factor is the affected skin area by a single microneedle because it determines the density of microneedles in an array. This factor can be influenced by the injected volume and interactions of the vehicle with the skin, the solute and the skin and the solute and the vehicle. The third major factor is the depth in the skin where the liquid containing the drug is infused / injected. With very small, hollow, leakproof and short microneedles intraepidermal infusions / injections would be possible.

A general limiting factor for intradermal infusions / injections is the solubility of the drug in the solution hence the total volume of delivered liquid. Injection volumes are probably limited to a volume of 0.20 ml which is usually the maximum volume injected with the Mantoux method. It is supposed that larger volumes need to be infused at low infusion rates over longer periods of time so that the vehicle can be removed by the body. Of course, this is only possible if the duration of administration is not critical.

In the design of a microneedle tool one might end up with a solution based on a single microneedle. Continuous infusion of insulin might be such an application because insulin is currently infused via a single subcutaneous cannula. 5.1.8. Infusion / injection process

A transient back pressure was observed in the infusion / injection experiments. The back pressure signal of two infusion experiments with aqueous solution, a flow rate of 0.10 ml/h and insertion parameter sets 3/2 and 5/2 is presented in Figure 5.9. The transient back pressure of the experiment with insertion parameters 3/2 shows an initial back pressure overshoot. The experiment with insertion parameters 5/2 shows almost no back pressure overshoot. Both signals finally reached a steady state back pressure level for the rest of the experiment.

To get the steady state back pressure level and information on the dynamics of the process the data of the infusion experiments without back pressure overshoot was fit to the following equation:

(5.3)

The time dependent back pressure is denoted as p, the asymptotic steady state back pressure as pA, the time as t and the characteristic time constant as . The back pressure p reaches 63 % of its asymptotic value within a time span of  and increases to 99 % of the asymptotic value within five times of . The fit worked well for data without back pressure overshoot. The signals presented in Figure 5.9 resulted in coefficients of determination of 0.90 and 0.04 for experiments without and with back pressure overshoot, respectively. Table 12 gives an overview of the average coefficients of determination for infusion experiments without back pressure overshoot and identical parameters.

Figure 5.9. Back pressure as a function of time for 2 intradermal infusion experiments with aqueous solution, a flow rate of 0.10 ml/h and insertion parameters 3/2 and 5/2. The dotted lines represent the fit of equation (5.3) to the data.

The series of experiments with a flow rate of 0.50 ml/h and insertion parameter set 4/2 showed a low coefficient of determination of 0.61. The main reason is a slight continuous decrease of -15 %/h ± 4 %/h (-4.9 kPa/h ±1.3 kPa/h) of back pressure with time instead of a constant steady state back pressure. The most probable reason is an undetected leak in the experimental setup. Two of the experiments with insertion parameter set 3/2 also showed this slight continuous decrease which reduced the average coefficient of determination to 0.74. Experiments with coefficients of determination higher than 0.80 showed unstructured residuals.

If equation (5.3) is fit to data with back pressure overshoot the parameter pA is over- estimated (Figure 5.9). Therefore, in case of back pressure overshoot the steady state back pressure was determined by calculating the mean back pressure between 0.5 h and 2 h after start of the experiment.

The obtained steady state back pressures give the total back pressures generated by the cannula and the tissue. Table 13 gives an overview of total (measured) and cannula generated (see materials section) back pressures at different flow rates and media.

Table 12. Coefficients of determination for data of infusion experiments with aqueous solution fit to equation (5.3). Experiments with back pressure overshoot are excluded.

Flow rate Insertion parameter set 0.10 ml/h 0.50 ml/h 3/2 0.85 ± 0.01 (n=2) 0.74 ± 0.13 (n=6) 4/2 0.84 ± 0.13 (n=7) 0.61 ± 0.13 (n=812) 5/2 0.81 ± 0.13 (n=7) 0.90 ± 0.06 (n=7)

12 One Experiment was omitted due to an exceptional constant increase (11 kPa/h) in back pressure. It could not be related to a full occlusion because stained liquid was infused into the skin and formed a spot with an area not distinguishable from other experiments with the same parameters.

69 Table 13. Total (measured) and cannula generated (see materials section) back pressures at different flow rates with media with different viscosities.

Back pressure / kPa Aqueous solution PEG solution 0.10 ml/h 0.50 ml/h 36 ml/h 180 ml/h 0.10 ml/h 0.50 ml/h Total 11 ± 7 31 ± 18 261 ± 101 881 ± 76 47 ± 25 198 ± 149 Cannula 0.09 0.45 32 162 5 25

The back pressure generated by the cannula is below 2% of the measured total value for infusion experiments with aqueous solution and flow rates ≤ 0.50 ml/h. However, in all other cases, with flow rates ≥ 36 ml/h, the back pressure generated by the cannula accounts for 11% to19 % of the measured values. Therefore, the back pressure generated by the cannula was subtracted from the total back pressure in order to plot the steady state back pressure of the tissue as a function of flow rate in Figure 5.10. The graph also shows data of the injection experiments. The back pressure generated by the experimental setup is neglected because the pressure sensor was mounted close to the cannula.

At this stage it is interesting to further characterise the tissue in terms of the back pressure that is needed to penetrate it with a defined liquid flow rate. The steady state back pressure values of experiments with water based solution (Figure 5.10) can be fitted as p ~ q0.53. It is consistent with previous work published in ([70], Fig. 5). This data can be fitted as p ~ q0.36±0.10. There is only a minor difference between the two exponents which is most probably due to the fact that human cadaver skin instead of pig skin was used. The data clearly demonstrates that there is a nonlinear relationship between back pressure and flow rate. In other words, the fluidic resistance of the excised tissue is a function of flow rate. Additionally, the back pressures needed for delivering PEG and water based solutions into the tissue were compared. Although the viscosity of the two solutions differs by a factor of 55, the maximum pressure values only change roughly by a factor of 5 (factor 4.3 for 0.10 ml/h; factor 6.4 for 0.50 ml/h). This is a clear indication for vehicle-skin interactions.

The slightly higher exponent of the fitted back pressure data obtained with pig skin compared to the exponent of the fit to data from [70] obtained with human cadaver skin can be explained. Due to its higher density of collagen and other fibre bundles in the reticular dermis pig skin shows an increased resistance to skin deformations [73] which are e. g. caused by the intradermal infusion / injection of liquid volume. So, the back pressure of intradermal infusions / injections is higher in pig skin compared to human skin. Therefore, pig skin is a reasonable model to study the back pressure of intradermal infusions / injections if the focus is on obtaining an upper design limit.

Figure 5.10. Steady state back pressures across the tissue for intradermal injections/infusions at different liquid delivery rates with water based and PEG based solutions. Delivery rates below 1 ml/h belong to infusion experiments. Delivery rates above 1 ml/h belong to injection experiments. The error bars show the standard deviation (0.10 ml/h: n=27, aqueous sol.; 0.50 ml/h: n=29, aqueous sol.; all other data points n=9).

The dynamics of infusion experiments without back pressure overshoot is described by the parameter  of equation (5.3). The fit didn’t work with data showing back pressure overshoots because of the different shapes of the measured curve and the model curve. Therefore,  could not be determined. Instead, an equivalent number was generated. In this case it was obtained by reading the time when the signal reached 63% of the steady state value. The slope of the initial back pressure increase is almost the same for experiments with and without back pressure overshoot. Based on the assumption that the increase in back pressure is caused by the same process  and the equivalent number can be averaged together. The result is a time constant of at least 2.3 min ± 1.2 min. (0.50 ml/h aqueous solution, n=27). Further investigations revealed that the dynamics of the infusion process was mainly governed by the syringe pump13. This result confirms the previously stated assumption that it is always the same process that determines the slope of the back pressure increase. (The data of 2 experiments was excluded from the mean, 11 min. and 14 min., because of its strong influence on the mean suggesting that there is a different unknown but rare mechanism at work.).

During the infusion experiments the delivered liquid volumes were cross checked with the integrated data of the flow sensor in the setup. Both values were within 10% deviation.

13 According to the subsequent text the time constant of the experimental setup including the tissue can be calculated to be 0.7 min, with Rtotal,0.50 ml/h=62 kPa h/ml (calculated with data of Table 13) and

Csetup=0.2 µl/kPa (see subsequent text).

71 Table 14. Summary of back pressure overshoots in infusion experiments with aqueous solution. Every flow rate and insertion parameter set combination was realised 9 times.

Flow rate Insertion parameter set 0.10 ml/h 0.50 ml/h 3/1 7 4 4/2 2 2 5/3 3 2

Back pressure overshoot was observed in 20 out of 56 infusion experiments with aqueous solution. Table 14 gives an overview of back pressure overshoots sorted by insertion parameter set and flow rate. The trend in the data suggests that the risk of an overshoot reduces with increasing tip depth and flow rate.

The duration of a back pressure overshoot was obtained by determining the time the signal exceeded the steady state back pressure by 10%. This data is presented in Table 15 together with the recorded average maximum back pressure. The recorded maximum values in a single experiment are 38 kPa (352% of steady state back pressure) and 112 kPa (361% of steady state back pressure) for flow rates of 0.10 ml/h and 0.50 ml/h, respectively.

Experiments performed with PEG solution showed no back pressure overshoot. This might be related with the aspect that in general higher pressure levels were needed in order to deliver PEG solutions at the same flow rate. Finally it is concluded that an initial back pressure overshoot can be present. This phenomenon is related to initial clogging of the cannula after its insertion. An initial back pressure overshoot of less than 400 % of the steady state back pressure was sufficient to remove the clogging.

In manual injection experiments, only performed with water as vehicle, the back pressure followed a steep increase before it reaches a constant steady state level with superimposed irregular oscillations (Figure 5.11). The nature of these oscillations was most likely due to the manual control of the injection procedure. The duration of the metronome synchronised manual injection process is 9.3 s ± 0.4 s (n=18). It was measured from the beginning of the steep increase to the beginning of the decay.

Table 15. Duration and average maximum back pressure of back pressure overshoots with aqueous solution at flow rates 0.10 ml/h and 0.50 ml/h.

Flow rate / ml/h Cases Duration / min. Average max. / kPa 0.10 12 of 27 20 ± 9 18 ± 9 0.50 8 of 34 15 ± 9 40 ± 3314

14 Excluding one experiment, with an exceptional back pressure overshoot of 112 kPa, the average absolute max. is 30 kPa ± 16 kPa.

72 To extract the values of the constant steady state back pressure level the back pressure values within the band of oscillations were averaged. This band is defined by the maximum back pressure and the minimum back pressure between the steep increase and final decrease of the signal (Figure 5.11). To derive quantitative results the characteristic time t1 for which 90% of the constant back pressure level was reached was extracted. t1 is determined to 5.7 s ± 2.3 s. Further analysis is disputa- ble because it is mixed with the manual injection process.

The back pressure decay fits very well to an exponential behaviour described by:

(5.4)

The time dependent back pressure is denoted as p, the residual back pressure at the end of the decay with p0, the back pressure at the beginning of the decay is the sum of pD and p0, and the time constant is denoted with . The coefficient of determination is 0.97 ± 0.02 (n=18). The average residual pressure p0 is less than 1 % of the average constant steady state back pressure level. The time constants of the back pressure decays are 5.1 ± 1.8 s (n=9) and 3.1 ± 0.9 s (n=9), for injection flow rates of 36 ml/h and 180 ml/h respectively. The ratio of the time constants is 1.6.

The high coefficient of determination is a strong indication that changes in  during back pressure decay of a single experiment can be neglected. Taking this as a basis the time constant can be calculated according to the model assumption (see materials section) to:

(5.5)

The time constant affiliated with the flow rate q is denoted as q, the fluidic resistivity of the cannula as Rcannula, the fluidic resistivity at flow rate q of the tissue as Rtissue,q and the fluidic capacity of the setup as Csetup. With Rtissue as a function of flow rate and assuming Csetup is constant one can calculate the ratio of time constants according to:

(5.6)

Figure 5.11. Back pressure as a function of time for a manual intradermal injection of 0.10 ml of aqueous solution with insertion parameter set 4/2. The asymptotic value was determined by averaging the data within the band of oscillations. t1 is the time when the signal reached 90% of the asymptotic value.

73 The subscripts low and high denote the corresponding time constant  and fluidic resistivity R in the case of injection experiments with low (36 ml/h) flow rate and high (180 ml/h) flow rate. With Rtissue,low of 6.3 kPa h/ml, Rtissue,high of 4.0 kPa h/ml (both values calculated with data of Table 13) and Rcannula of 0.9 kPa h/ml (see materials section) the ratio of the time constants is 1.5. This is just a minor deviation from the ratio of the measured time constants. So, the structure of the assumed model is acknowledged with a flow rate dependent tissue resistivity.

The fluidic capacity of the experimental setup can be calculated by solving equation (5.5). The measured fluidic resistivities and measured time constants of intradermal injection experiments with water resulted in a fluidic capacity of 0.19 µl/kPa. The data of experiments without skin ( = 0.50 s ± 0.04 s, q=180 ml/h, n=3) results in a fluidic capacity of 0.15 µl/kPa which is close to the value obtained from the measured data. The fluidic capacity of the experimental setup is quite high. A simple syringe with cannula is supposed to exhibit a much lower capacity.

Due to the high back pressures observed in the injection experiments it demands as a good practice to leave the cannula inserted into the skin for a multiple of time constants when the injection is finished. Otherwise, a small droplet of liquid appears on the skin surface which might result in inconsistent drug delivery. The time constants observed in this work are quite long due to the high capacity of the experimental setup. A simple syringe cannula combination with a substantial lower capacity will result in shorter time constants.

Neither for infusion experiments nor for injection experiments there is any hint that the asymptotic back pressure levels or the time constants would depend on skin samples from different pigs or different storage conditions.

The phenomena of varying flow resistance of the tissue and the overshoot in back pressure might influence the proper selection of drug delivery actuation mechanisms for intradermal drug delivery devices.

The data presented so far was obtained in ex vivo experiments with excised tissue. It is assumed, that the infusion at low flow rates (0.1 ml/h) and injection of small volumes (0.1 ml) in vivo results in steady state back pressures comparable to the presented values because an infusion / injection volume of 0.10 ml is less than 10 % of the average volume (1.77 ml ± 0.32 ml) of the skin specimens. This assumption needs to be checked in in vivo experiments.

The limiting issue for intradermal infusions is the initial back pressure overshoot which is almost 400 % higher compared to the steady state back pressure. At this stage, there is no reason to believe that the data of the initial back pressure overshoot obtained in the ex vivo experiments is not valid in the in vivo case. So, for reliable infusions a suitable device must be able to generate pressures from 38 kPa to 112 kPa for infusion flow rates of 0.10 ml/h to 0.50 ml/h, respectively. This is not self- evident for a micro actuator [115, 116]. Currently, small skin attachable programmable or intelligent infusion devices based on micro actuators seem to be feasible if a flow rate of approximately 0.10 ml/h is sufficient for the application. Substantial higher flow rates resulting in higher back pressures in the order of more than 100 kPa definitely require more powerful drive systems for intelligent infusion systems, e.g. elec- tromechanical drives.

74 5.1.9. Relevance of the technology

The laboratory setup was designed in order to study the influence of different penetration parameters. For the sake of convenience in performing these experiments its size was set for easy handling. Therefore, it is quite large. If the penetration parameters are fixed its size can be shrunk to a minimum. However, even with estimated dimensions of 14 mm x 12 mm x 10 mm (corresponding to a volume of 1.7 ml) the shrunk intradermal tool will be much bulkier compared to out of plane microneedle designs. This is not necessarily a drawback. The intradermal tool is best utilised in intelligent infusion devices which are worn for longer periods of time. Hence, the device also needs to store a supply of drug solution, a drive mechanism, electronics, battery and a user interface. This functionality is e. g. realised by insulin pumps. The currently smallest one on the market (Omnipod by Insulet, Bedford, MA, USA) has a size of 45 ml. Therefore, the presented mechanism shrunk to the estimated size is considered an alternative tool for intelligent infusion devices.

The presented intradermal tool in combination with a suitable delivery device can infuse liquids in continuous or intermittent fashion. It is conceivable to control the intermittent infusion e. g. in response to a sensor input. Of course, it can also be used to inject liquids into the skin. The infused / injected liquids can be solutions of hydrophilic, ionic and macromolecular drugs. However, the rate of uptake of these molecules from the tissue into the systemic circulation needs to be studied. 5.1.10. Summary

An intradermal tool which enables to place a cannula into the dermis by simply pushing and releasing a slider is presented. It is minimum invasive and allows to infuse or inject liquids containing hydrophilic, ionic and macromolecular drugs into the dermis. This mechanism could be an important building block of complete drug delivery systems for time, user or sensor controlled drug delivery. The cannula insertion mechanism is so simple that it could be applied for self-administration of drugs.

Three different failure modes for leakage were investigated and it is proven that the intradermal tool showed excellent leak tightness. Although leakage occurred in 9 out of 92 experiments, this leakage never happened at the puncture site. A first leakage failure mode can be excluded by properly selecting an application site with low hair follicle density. In a second and third leakage failure mode only a small leakage of clear carrier fluid without staining could be detected. So it is very likely that no drug molecules are lost by those leakage failure modes. This needs to be checked in vivo experiments preferably in humans.

The volume of the stained tissue can depend on vehicle-skin, vehicle-solute and solute-skin interactions. With water as vehicle the distribution of the model drug (gentian violet) and the vehicle were limited to the dermal layer involving solute-skin and vehicle-skin interactions. The maximum vertical staining depth was 1.16 mm ± 0.14 mm. The lateral penetration depth resulted in stained areas of 3.5 mm² ± 2.7 mm² to 15 mm² ± 4.4 mm² monitored from the top. With PEG as vehicle the stained tissue area is almost one order of magnitude larger compared to the water based solution involving vehicle-solute interactions. In this case penetration of stained PEG into the hypodermis was observed. If a given tissue area should be loaded with drug these interactions need to be considered. For water based solutions

75 the presented data can be regarded as lower limit because drug molecules show probably less tissue binding compared to the dye gentian violet.

The infusion / injection process initiated with a flow source showed occasionally initial back pressure overshoot. Values ranging from 38 kPa to 112 kPa for flow rates from 0.10 ml/h to 0.50 ml/h were measured. This back pressure overshoot was identified as the limiting parameter for the design of small programmable or intelligent devices based on micro actuators.

The ex vivo experiments revealed a flow rate dependent tissue back pressure. The back pressure needed for infusing PEG based solutions was only about a factor 5 higher compared to the back pressure needed to infuse water, although the viscosities vary roughly by a factor of 55. This is due to solute-skin interactions. In general, it is found that steady state back pressures in the order of 11 kPa ± 7 kPa to 198 kPa ± 149 kPa are needed for the infusion of water and PEG based solutions at flow rates of 0.10 ml/h and 0.50 ml/h, respectively. In vivo back pressures might be higher and need to be studied.

With the presented intradermal tool it is possible to deliver liquids reliably to the intradermal layer of the skin. A large variety of drugs dissolved in this liquid can be infused with high temporal and quantitative precision compared to oral drug administration if the intradermal tool is combined with a suitable skin attachable drug delivery actuation mechanism. Consequently, this tool has the potential to be a clinically relevant building block for easy-to-use self-administration of drugs. 5.2. In vivo experiments with the infusion set 5.2.1. Materials & Methods

In vivo experiments were performed with the intradermal infusion set. Infusion flow rates from 0.1 ml/h to 2.5 ml/h and injections (0.1 ml or 0.5 ml in 10 sec.) with flow rates of 36 ml/h and 180 ml/h were accomplished. Aqueous (infusions and injections) and PEG (infusions) solutions were used (chapter 5.1.2). Evans blue was used as dye instead of gentian violet. Evans blue is a dye with a molecular mass of 980 g/mol that binds to serum albumin. Domestic pigs (3 months old) were applied as laboratory animals in the animal research facilities of the Universitätsmedizin Mannheim on the campus of the University Heidelberg.

The experimental setup for the infusion experiments is presented in Figure 5.12. Sy- ringe drivers were used for intradermal infusions experiments. Injection experiments were performed with syringes. The plunger of the syringe was manually pressed within 10 sec. and timed with a metronome. The infusion and injection pressures were recorded with pressure transducers with measurement ranges up to 20 kPa and up to 100 kPa (RPUB002D and RPUB010D, SensorTechnics, Puchheim, Germany). The intradermal infusion sets were mounted on the abdominal skin of the laboratory animal (Figure 5.12, right).

16 in vivo infusion experiments at flow rates ranging from 0.1 ml/h to 2.5 ml/h were performed with aqueous and PEG solution. Additionally, 25 in vivo injection experiments with aqueous solution were performed.

Figure 5.12. Experimental setup. Right: Position of the intradermal infusion experiments. 5.2.2. Results

Leakage of liquid from the skin was not observed in any in vivo experiment. After removal of the cannula from the skin it can happen that a droplet appears on the surface (Figure 5.13). Such a droplet can contain infused/injected liquid and blood.

Intradermal infusion and injection of liquids results in wheal formation (Figure 5.14). Injections and infusions of aqueous and PEG solutions with the same volume result in wheals comparable in size. The profile of wheals caused by introduced PEG solutions is more dome-shaped. This is due to the higher viscosity of the PEG solution. It cannot be distributed in the skin as easily as water. The diameter of a wheal can increase to 20 mm with an introduced volume of 0.8 ml.

Mild redness of the skin around the treatment site was observed with injection experiments (0.1 ml). No redness could be detected after infusion of aqueous and PEG solution. Redness after intradermal injections is frequently observed. It is an indication for skin irritation. Nevertheless, intradermal injections are well tolerated by patients [73, 74, 75]. If injections of volumes larger than 0.1 ml would also be tolerated by patients needs to be studied.

Figure 5.13. A small red colored droplet appearing on the skin surface after removal of the cannula. It is the result of an injection experiment with aqueous solution.

Figure 5.14. Left: Wheal after injection of 1 ml of aqueous solution. Irritation of the skin indicated by mild redness can be identified. Right: Wheal formation after infusion of 1 ml of PEG solution. No irritation of the skin detectable.

The transient back pressure of the in vivo injection and infusion experiments was analysed in the same way as the ex vivo experiments (chapter 5.1.8). The steady state back pressure is presented in Figure 5.15 together with the ex vivo data of Figure 5.10. Due to the limited amount of in vivo infusion data the existing data is presented as single data points. The data of in vivo injection experiments is presented with mean and standard deviation. The time constants for the infusion experiments ranged from 0.5 min to 17.9 min. The time constant decreased with increasing flow rate. There is no difference in time constant between infusion of aqueous and PEG solution. The time constant is probably determined by the syringe pump (chapter 5.1.8).

Figure 5.15. Back pressure of ex vivo and in vivo experiments.

78 Comparing the in vivo data to the ex vivo data of the infusion experiments with aqueous solution it can be concluded that there is a good correlation. The ex vivo data can be used as an upper limit for the design of actuators for intradermal infusion applications. There is no correlation for the in vivo and ex vivo infusion data with PEG solution. No change in back pressure at flow rates of 0.1 ml/h and 0.5 ml/h could be identified. The high osmotic pressure of PEG might explain the higher in vivo pressure compared to the ex vivo experiments at a flow rate of 0.1 ml/h. A flow of water to the PEG injection site can generate a force opposing the infused fluid flow. A higher back pressure is the inevitable consequence. The lower back pressure at a flow rate of 0.5 ml/h might be explained by a less dense collagen network of the abdominal skin compared to the skin removed from the pig ear. 5.2.3. Summary

The intradermal infusion set is a robust and easy to use technology. With a simple operating step a cannula can be placed in the dermis for leak proof injections and infusions. The in vivo experiments confirmed the ex vivo data for aqueous solutions. Hence, with the available data at hand, a device developer can seriously integrate the intradermal infusion set into a drug delivery device. 5.3. Ex vivo and in vivo experiments with injection moulded microneedle arrays

Defined application of the microneedle arrays was performed with a spring powered impact applicator. The spring powered applicator enabled the application of microneedles into cadaver skin at high speed. The dynamic parameters of the spring powered applicator are calculated theoretically (see appendix c). 5.3.1. Spring powered applicator

The spring powered applicator is presented in Figure 5.1615. It consists of a piston moving in a bushing, a spring and a lifting table. There are 10 notches in the bushing with a spacing of 2 mm starting at 2 mm from its lower end. For all experiments in this work a spring with a spring rate of 1.9 N/mm was used. It is compressed by pressing the piston into the bushing. Subsequent the piston can be locked with the pins in notches at the desired level by rotating it. With the selected spring the notches starting 4 mm from the lower end can be used. This notch resulted in an effective skin deflexion of s0=-1.5 mm. The notch at 20 mm from the lower end resulted in an effective spring deflexion of s0=-17.5 mm. The bushing itself is mounted on a platform fixed to the lifting table.

15 Concept by Vosseler M, design and assembly by Alabsi B [117].

Figure 5.16. Spring powered applicator. Left: Drawing of the applicator revealing its components. By moving the piston into the bushing the spring gets compressed and energy is stored. The piston can be locked at a certain position by rotating the pin into one of the notches. The arrow indicates the positive direction of the parameter s describing the position of the piston. Right: Picture of the setup with the spring powered applicator mounted in a vertically adjustable platform. 5.3.2. Experimental procedure

A microneedle array with lid and tubing (Figure 4.10, right) was inspected with a stereo microscope. Potentially present burrs were removed cautiously with a scalpel. The tubing was connected to the tubing of the experimental setup with pressure sensor and glass syringe. The tubing and the microneedle array were primed with liquid (0.02% methylene blue in 0.9 % by weight NaCl solution) for infusion / injection experiments. Subsequent, the microneedles were tested for occlusions and comparable fluidic resistivity by squirting some liquid through the microneedles. A piece of full thickness pig skin was placed on a sheet of artificial skin.

The microneedle array with lid and tubing was fixed in the microneedle adapter that is joined to the piston (Figure 5.17). It was locked in the notch at 4 mm from the lower end of the bushing affiliated with a spring deflexion of s0=-1.5 mm. Next, the lifting table was adjusted in a way that the tips of the microneedles are in slight contact with the pig skin. The vertical position of the lifting table was obtained with a calliper. The table was lifted in order to transfer the piston to the upper notch affiliated with a spring deflexion of s0=-17.5 mm. The lifting table was lowered again by an additional subsidence of 2 mm, 3 mm, 4 mm or 5 mm below the indicated position of the calliper. This way the microneedle array is constantly pressed against the skin after the transient application procedure. With a spring balance connected to the picot the effective force was measured. The application was done by manually rotating the piston at one of the pins. The spring accelerated the piston until the microneedle array plunged into the skin.

Infusions were performed with the syringe pump at flow rates of 0.10 ml/h and 0.20 ml/h for 20 min. Injection experiments were performed manually by injecting 0.10 ml of fluid within 10 s. The pressure was recorded during the infusions with an analog to digital converter at 10 Hz. Pictures of the infusion / injection site were taken after the experiments.

Figure 5.17. Microneedle array applied to pig skin. The microneedle array is mounted in the microneedle array adapter which is joined to the piston. A picot is also joined to the piston. The pig skin rests on a piece of artificial skin wrapped in cling wrap. 5.3.3. Results

The effective force by which the microneedle array is pressed against the pig skin is presented in Figure 5.18. It increases linearly from 3.9 N to 7.5 N for a subsidence of 2 mm to 5 mm. The upper limit is calculated theoretically. It is based on the assumption of a stiff and fixed plate instead of the soft tissue. The measured effective force is lower than the upper limit because the microneedle array sinks into the soft tissue.

Pictures with infusion spots are given in Figure 5.19 and Figure 5.20. In Figure 5.19, left one can easily recognise that most of the infusion solution was infused through one of the three microneedles. The extent of the diffusion of methylene blue in the tissue can be comprised by comparing Figure 5.19, left and the almost three hours later taken picture in Figure 5.19, middle. In some of the experiments leakage occurred. In Figure 5.19, right the superficial distribution of infusion liquid on the skin surface is shown. The transport of the liquid along the wrinkles of the skin can be identified easily. So, experiments without and with leakage can be readily distin- guished. Figure 5.20 shows the top and bottom side of a piece of pig skin after 4 infusion experiments. The top view shows that the dye spreads in the upper skin layers as already stated. The bottom view shows larger dermal capillaries that are stained by the dye. The staining of all 4 infusion sites is directed towards one direction. So, it is concluded that the staining in the capillaries is distributed with a substantial pressure driven component. This is a strong indication of penetration of the dermis by the microneedles.

The microneedles were inspected with a stereo microscope after the infusion experiments. The microneedles didn’t show any noticeable alteration. In fact one of the microneedle arrays was used in 8 experiments (data included in this work was always acquired with an unused microneedle) without any issue. So, the microneedle design is very robust.

Figure 5.18. Effective force by which the microneedle array is pressed against the pig skin resting on the artificial skin.

Figure 5.19. Results of two infusion experiments with flow rate of 200 µl/h for 20 min. Left: Pic- ture taken immediately after the infusion. Middle: Picture taken after 160 min. Right: Superficial distribution of infusion solution due to leakage.

Figure 5.20. Piece of pig skin with four infusion sites. Left: top view. Right: bottom view. The red circles of the bottom view were placed to the best of one’s knowledge.

Table 16. Summary of desired and undesired events of 19 infusion experiments.

Subsidence / mm Leakproof Leaky Clogged 2 1 2 - 3 7 1 1 4 3 0 1 5 3 - -

82 Table 16 gives a summary of all desired and undesired events of 19 infusion experiments. Leakage was observed in 2 of 3 experiments at a subsidence of 2 mm and in 1 of 9 experiments at a subsidence of 3 mm. Leakage was not observed in experiments with a subsidence of 4 mm or higher. At this level the microneedle array is pressed against the skin with a force of 6.2 N or higher. It is assumed that in this case the frustums of the microneedles enable the leakproof infusion. However, continuously pressing a microneedle array against the skin with a force of 6 N is un- pleasant. Therefore, it is concluded that the microneedle arrays are best used for injections with short application time.

Clogging was observed in 2 of 19 experiments. It is a serious problem. The fact that it can occur requires the designer of a corresponding drug delivery device to account for this issue. For instance one could integrate a pressure sensor to detect occlusions.

The infusion experiments were performed with a syringe pump. The transient back pressure of the leakproof experiments fits very well (R²=0.92±0.05, n=14) to the equation:

(5.7) p denotes the transient back pressure, ps is the steady state back pressure, t is the time and  is the time constant. Data of the first 20 min. of the 14 leakproof infusions with a flow rate of 200 µl/h was analysed. The time constant is 5.4 min. ± 2.2 min. The average infused volume is 51 µl ± 8 µl. The measured steady state back pressure is plotted in Figure 5.21 vs. the force by which the microneedle arrays were pressed against the skin. The back pressure increases linearly with the subsidence force (R²=0.96). Based on the finding that a force of approximately 5 N is necessary to perform a leakproof infusion it gets clear that a pressure of at least 346 kPa ± 117 kPa is necessary to perform reliable infusions.

Figure 5.21. Back pressure as a function of the effective force due to the defined subsidence of the applicator. The error bars are based on n≥3 experiments. There is no vertical error bar at the lower end of the graph because there is only one leakproof experiment. There is no horizontal error bar at one of the data points because it was calculated from data of.

83 5 more infusion experiments were performed at a flow rate of 100 µl/h. The subsidence was set to 3 mm. In one experiment the microneedle array was clogged and another experiment was leaky. The data of the 3 successful experiments showed no statistically significant difference compared to the experiments with a flow rate of 200 µl/h.

Injection experiments were performed manually. The plunger of a 100 µl syringe was pressed within 10 sec. timed by a metronome. This is equivalent to a flow rate of 36 ml/h. The subsidence was set to 3 mm. All three experiments resulted in leakproof injections. A picture of an injection spot is presented in Figure 5.22. The colour of the spot is paler compared to the infusion experiments. This is due to the short injection time affiliated with a very short diffusion of the dye in the skin. Most of the liquid was infused through one of the three microneedles. The back pressure during injection of liquid solution into the tissue was higher than 1 MPa.

In vivo infusion experiments were performed with an anaesthetised pig. The animal was lying on the back. Its ear was placed below the microneedle applicator (Figure 5.23). The applicator was adjusted to a subsidence of 3 mm. In the first experiment the pig ear was resting on a metal surface. All microneedles broke during the application procedure. In the following experiments 8 sheets of wiping tissue were folded and placed between the pig ear and the metal surface.

Figure 5.22. Picture taken immediately after an injection experiment.

Figure 5.23. Experimental setup to study in vivo infusion parameters.

84 Table 17. Summary of in vivo experiments.

No. ps / kPa  / min V / µl Duration / min Remarks 1 microneedles broken 2 252 6.8 157 40 leakproof leakproof, back pressure 3 186 5.2 139 34 overshoot (max. 270 kPa) 4 214 6.9 53 17 leaky after 17 min. 5 clogged

In the first two infusion experiments 0.02 % by weight methylene blue in 0.9 % by weight NaCl solution was used. The infusion spot was very pale. It was impossible to derive from the infusion spot whether liquid was infused or not. Therefore, the following experiments were performed with a solution with a methylene blue concentration of 1 % by weight.

The data of 5 infusion experiments with a flow rate of 300 µl/h is summarized in Table 17. One experiment was leaky after 17 min. another experiment showed an initial back pressure overshoot. The steady state back pressure level and the time constant is comparable to the ex vivo experiments. The skin was cut at the infusion site and bend over to take a look at the cross section (Figure 5.24). The cross section was clearly stained. 5.3.4. Summary

Intradermal injections and infusions with the injection moulded microneedles are possible. However, an applicator is necessary to insert the microneedles into the skin. The mechanical stability of the injection moulded microneedles is high. Successful in vivo experiments demonstrated the rugged design. Additionally, for reliable leak proof operation it is necessary to press the microneedle array into the skin. Consequently, a much higher infusion/injection pressure is necessary compared to the intradermal infusion set. Furthermore, continuous pressure on the microneedle array is uncom- fortable. Therefore, these microneedle arrays should be considered just for injection applications.

Figure 5.24. Pig ear with cut through an infusion spot.

86 6. A chronopharmaceutical transdermal drug delivery device: ChronopaDD

Almost all paragraphs of chapter 6.1 and some paragraphs of chapter 6.2 were published as original research paper (publication P2, see list of publications) in Smart Materials and Structures in 2012 (Volume 21, 105002). Contributions to this publication: Vosseler M: literature search and analysis, concept, design, planning of all experiments, most of the experimental work, all data analysis, manuscript preparation, supervision of Schmidt C and Ilieva V. Clemenz M: review of the manuscript. Zengerle R: scientific advice during experiments and manuscript preparation. Schmidt C: experiments on displaced volume as a function of back pressure and data collection (Figure 9 and Figure 10 of publication P2 corresponding to Figure 6.9 and Figure 6.10 of this text). Ilieva V: experiments on the volume displacement of the hydrogel actuator with different superabsorbent polymers and data collection (Figure 6 of publication P2 corresponding to Figure 6.6 of this text).

Chronopharmaceutical drug delivery can be easily realized with a time controlled device that infuses a liquid drug solution. The development of such a device requires a reliable fluidic interface to the human body and a flat design for wear comfort.

A suitable fluidic interface is the intradermal infusion set introduced in chapter 4.3. The need for this technology is explained in chapter 1.2. The feasibility of using this component was demonstrated in ex vivo and in vivo experiments in chapters 5.1 and 5.2.

A device for time controlled delivery of liquids with a hydrogel actuator is introduced in chapter 3.3.2. It uses a friction piston in a tube with varying inner diameter to realize the time delay. This approach results in a tubular design that doesn’t comply with required wear comfort. Apart from that mass fabrication of the friction piston with tight tolerances is probably extremely challenging. Therefore, a flat hydrogel actuator design is introduced in this work. The time delay is realized chemically with a layer that dissolves in a defined time.

The typical characteristic of a hydrogel actuator is a slow expansion process that can build up a relevant pressure. So, a hydrogel actuator is a promising technology to realize intradermal infusions. The actuator should be able to generate a reasonable flow rate of 0.1 ml/h or higher (see chapter 5.2). Its back pressure sensitivity up to 50 kPa (see chapter 5.2) should be low. Only in this case it is possible to operate a device with a hydrogel actuator without feedback loop.

The feasibility of designing a flat hydrogel actuator for the purpose of drug delivery is studied in this chapter. Subsequently, the flat hydrogel actuator is integrated in a chronopharmaceutical drug delivery device.

87 6.1. Hydrogel actuator

A flat and simple superabsorbent polymer (SAP) actuator structure is proposed. Its cross section is illustrated in Figure 6.1. It consists of a polypropylene (PP) ring. Its cavity is filled with SAP granulate. It is closed on one side with a thermoplastic elastomeric membrane and on the other side with a woven polyester mesh. This mesh is a rigid filter which is supposed to keep the SAP granulate and the hydrogel in the cavity. The mesh and the elastic membrane are bonded to the PP ring by plastics welding.

Fabrication is already quite simple. Nevertheless, mass fabrication can be even sim- pler. One could think of the following three steps. First: two component injection moulding of the ring and the thermoplastic membrane, second: filling with SAP granulate and third: sealing with the filter membrane.

The functional principle of the SAP actuator is simple. The swelling agent e.g. water can get in contact with the SAP granulate via the filter membrane. Immediately, the SAP granulate takes up the swelling agent. The swelling process is started. It continues as long as the chemical potential of the free water is higher than the chemical potential of the absorbed water. Continuously, the volume of the cavity increases by bulging the elastomeric membrane outwards. This way the chemical energy stored in the SAP is transformed into mechanical work. Anything (solid, liquid or gaseous) moveable on the other side of the elastic membrane can be displaced.

In general, the swelling rate of a SAP polymer is fast in the beginning of the absorption process. With time the swelling rate decreases. There are polymers that absorb a large fraction of the theoretically absorbable mass very fast. Other polymers show a substantial reduction in swelling rate after absorption of a small fraction of the theoretically absorbable mass.

Figure 6.1. Illustration of the cross section of a superabsorbent polymer actuator. The cavity of the ring made of polypropylene is filled with granulate of superabsorbent polymer. It is closed by welding a filter (polyester mesh) at the top and an elastomeric membrane at the bottom of the ring surface.

88 Table 18. Internal parameters with influence on the performance of the SAP actuator

Geometrical factors Material properties filter pore size type of SAP cavity volume (diameter, height) amount of SAP thickness of elastomeric membrane stiffness of the filter membrane hardness of elastomeric membrane

The capacity of the SAP actuator to take up swelling agent and its transient behaviour is influenced by external and internal parameters. External parameters are the chemical potential of the swelling agent, the external pressure on the elastomeric membrane, the input and output flow rate. The temperature is another external parameter that is not investigated in this work. Numerous internal parameters of the SAP actuator are listed in Table 18. They can be influenced by design. The main objective is to characterize the influence of important factors on the performance of the SAP volume displacement actuator. This is done with a factor screening experiment. 6.1.1. Materials and Methods

PP rings with an outer diameter of 40 mm and inner diameters of 17 mm (area 2.3 cm²), 19.5 mm (area 3.0 cm²) and 24 mm (area 4.5 cm²) were manufactured by laser cutting of PP sheets (thickness 1 mm, 1.5 mm or 2 mm, obtained from Schmidt & Bartl, Villingen-Schwenningen, Germany). The volume of the ring cavity is always 0.45 ml. The edges were deburred manually. Actuator fabrication started by welding the polyester mesh (mesh opening 40 µm, 50µm or 105 µm, Spectrum Laboratories, Rancho Dominguez, CA, USA) to the PP ring. A ring shaped hot welding tool (240 °C) with appropriate inner diameter and a width of 2 mm was used. It was pressed against the mesh and the PP ring with a pneumatic cylinder (pressure 400 kPa, time 5 s, custom made at HSG-IMIT). Then, the cavity volume was filled with SAP granulate (SX-Fines, SXM 9410 or VP-Fines, Favor, Evonik Stockhausen, Krefeld, Germany. The powder density of the granulate is in the range of 0.5 g/cm³ to 0.7 g/cm³). The (Favor) granulate is a product based on partially neutralised and crosslinked (sodium) polyacrylic acid. The granulates of type SX-Fines (average particle diameter < 150 µm) and SXM 9410 (average particle diameter 150 µm to 850 µm) consist of surface crosslinked particles. Particles of type VP-Fines (average particle diameter < 150 µm) are not surface crosslinked. The SAP actuator was closed by welding the elastomeric membrane (thickness of 0.25 mm, hardness Shore A30, A50 or A70, custom made by Tekni-Plex Europe, Erembodegem, Belgium) to it (Figure 6.2, left). Figure 6.2, right demonstrates the volume change of a SAP actuator after swelling for several days.

Figure 6.2. Left: Completely assembled SAP actuator (inner-Ø: 17 mm, outer-Ø: 30 mm) with elastomeric membrane (thickness: 0.25 mm, hardness: 50 Shore A) upside and filled with 0.1 g of SAP granulate (VP-Fines Favor). Right: SAP actuator swollen to a volume of approximately 3 ml after three days.

The experimental setup to study the dynamics of the volume displacement of the SAP actuator is shown in Figure 6.3. The actuator is mounted on a water filled flange connector made of glass (custom made by a local glassblower). Tubing (fluorinated ethylene propylene - FEP with inner-Ø: 0.8 mm, length approximately: 50 cm) connects the flange connector with a water filled beaker on an analytical balance (AC 210 S, readability 0.1 mg, Sartorius, Göttingen, Germany). The balance is connected to a PC for continuous data acquisition. The water surface in the beaker is covered with paraffin oil (Sigma-Aldrich, Munich, Germany) to prevent evaporation.

Experiments were started by dosing 2 ml or 10 ml of deionized (DI) water into the cavity of the top flange connector. It was covered (not air tight) with a lid to prevent evaporation but permit venting.

After dosing DI water to the actuator it passed immediately the filter membrane and reached the SAP granulate. Instantaneously, the transient swelling process of the particular actuator under tests starts. As a consequence the elastomeric membrane bulged and pushed the water in the flange connector through the tubing into the beaker on the analytical balance. The transient increase in mass was recorded.

Figure 6.3. Experimental setup to study the dynamics of the volume displacement of the SAP actuator at ambient pressure. The SAP actuator is mounted between two flange connectors made of glass via two o-rings and a clamp. The bottom flange connector is filled with water. Tubing connects the flange connector with a beaker which also contains water and paraffin oil (to prevent evaporation). The beaker is placed on an analytical balance.

Figure 6.4. Experimental setup to study the dynamics of the volume displacement of SAP actuators at defined back pressures. Tubing connects the flange connector with a flow sensor, a pipette and a pressure controller.

The setup with analytical balance is very accurate with regard to absolute values. However, it cannot be used to study the dynamics of the volume displacement for various back pressures. Therefore, the setup was modified according to Figure 6.4. The desired back pressure is generated with a pressure controller (DPI530, range 200 kPa, GE Sensing, Fairfield, CT, USA). The displacement of water is acquired with a mass flow sensor (LiquiFlow, range 6 g/h, Bronkhorst, Ruurlo, The Nether- lands). The suggested setup is prone to offset errors of the flow sensor. Hence, a pipette (2 ml) was used to get the absolute displacement in order to check the integrated data of the flow sensor. The working fluid of the pressure controller is air. So, the setup was configured in a way that the phase boundary water/air was located at the bottom of the pipette before an experiment was started. 6.1.2. Results and Discussion

Initial Experiments

A first series of initial experiments was done to get an impression of the transient volume displacement of SAP actuators filled with different types of SAP (SX-Fines, SXM 9410 and VP-Fines). Except of the type of SAP all other design parameters and ambient conditions were kept constant (inner-Ø: 17 mm, thickness of PP-ring: 1 mm, mesh opening: 105µm, thickness of elastomeric membrane: 0.25 mm, hardness of elastomeric membrane: Shore A50, amount of SAP: 0.1 g, room temperature, ambient pressure).

Three actuators per SAP type were investigated after delivering 2 ml of DI water. The volume displacement of each of them was recorded for 4 hours. After each experiment the actuator was removed from the setup and inspected. The elastic membrane bulged outward as expected. The filter membrane retained the hydrogel in the cavity. However, its surface felt greasy. This indicates that hydrogel reaches partly across the filter pores without losing connection to the gel inside the cavity. Apart from that, the filter membrane also bulged out slightly. Hence, the efficiency in volume displacement of the actuator is not ideal. A layer of dry SAP pellets and an entrapped air bubble could be seen through the opaque elastomeric membrane of actuators filled with VP-Fines (Figure 6.5). The SAP-granulate of the other polymers formed a homo- geneous gel and a distinct air bubble could not be detected. In this case either the air or some of it could escape the actuator or it is evenly distributed in the gel.

Figure 6.5. State of a SAP actuator with VP-Fines after the experiment. A layer of dry SAP pellets and an entrapped air bubble could be seen through the opaque elastomeric membrane.

The result of the initial series of experiments obtained with the experimental setup (Figure 6.3) is presented in Figure 6.6. The displaced volume within 4 h ranges from 0.77 ml ± 0.06 ml for VP-Fines to 0.92 ml ± 0.08 ml for SX-Fines. The main difference between the three polymers is its behaviour in the early phase. The three configura- tions show large differences in terms of how fast they react on the addition of swelling agent. Consequently they show large differences in the displaced volume during that initial phase (5 min.). The smallest initial displacement was observed with VP-Fines (0.12 ml ± 0.03 ml) and the highest initial displacement was observed with SX-Fines (0.52 ml ± 0.08 ml).

The small initial displacement with VP-Fines can be explained by the fast formation of a dense gel layer. Water needs to diffuse through this gel layer to reach the dry SAP pellets. Apparently this is gel-blocking.

To extract quantitative information regarding the duration of the initial volume displacement phase the data was fit to the following equation:

(6.1)

This empiric equation describes the displaced volume versus time as superposition of two dynamic effects modelled with asymptotic exponentials. f(t) is the displaced volume at time t,  and  are the asymptotic values and 1 as well as 2 equal the time until 63 % of the asymptotic value is reached. The experimental data fits very well to equation (1) with coefficients of determination of 0.99 and higher. The longest duration 1 of the graphs of Figure 6.6 is 1.2 min (VP-Fines). So, within 3.6 min (3 times 1) the volume displacement in the initial phase reached at least 95 % of its final value. 2 ranges from 126 min to 162 min.

Based on this data, two types of SAP were allocated to two levels for the factor screening experiment presented in the next section. VP-Fines was allocated to level -1 because the volume displacement during the initial phase is very low. SX-Fines was allocated to level 1 because the volume displacement during the initial phase is large.

Figure 6.6. Transient behavior of the volume displacement of SAP actuators filled with different superabsorbent polymer pellets. The mean of 3 SAP actuators and the standard deviation (at selected points) is plotted16. 6.1.3. Factor Screening

Design of the experiment

A factor screening experiment was set up to identify and study the influence of relevant design parameters on the transient volume displacement of the actuator. Two response variables of the screening experiment were extracted from the displaced volume over time. The first variable is the displaced volume after 5 min (V5) the second one is the gain at 4h (G235) which represents the displaced volume between 5 min and 4 h. The selected factors, their levels and ranges are given in Table 19.

Factor A is the filter area of the SAP volume displacement actuator. The filter area, its diameter and the volume of the actuator cavity are dependent dimensions. Therefore, diameters (17 mm, 19.5 mm and 24 mm) were selected that result in the same cavity volume (0.45 ml) when PP sheets with a thickness of 1.0 mm, 1.5 mm and 2.0 mm were used. The medium level of factor A (filter area) is -0.32. It is not zero because the area of the medium actuator (3.0 cm²) is smaller than the mean area (3.4 cm²) of the small and the large actuator.

Table 19: Levels of the design factors.

Levels Factors -1 -0.32 / 0 1 filter area 2.3 cm² 3.0 cm² 4.5 cm² A (inner diameter) (17 mm) (19.5 mm) (24 mm) B type of SAP VP-Fines SX-Fines hardness of mem- C Shore A30 Shore A50 Shore A70 brane

16 Experiments performed and data acquired by Ilieva V [118]

93 The actuator cavity was filled with a different mass (0.23 g VP-Fines and 0.31 g SX- Fines) of SAP due to the different powder densities. This way the cavities were completely filled with an equal volume of SAP granulate. The difference in granulate mass is not considered relevant because the SAP pellets are supposed to absorb just a tiny fraction of the theoretically absorbable mass.

The level of factor B describes the SAP material. It is set according to the results of the preliminary experiments. These results were obtained not independently of the other parameters. Therefore, just the two extremes (VP-Fines and SX-Fines) were assigned to factor levels -1 and 1. No SAP granulate was assigned to factor level 0.

The hardness of the membrane is coded with factor C. The levels were set according to the available materials. Membranes with different hardness but the same thickness (0.25 mm) were obtained.

The experiments were performed at ambient conditions in the lab. So, changes in temperature and pressure are present but small. Variations in the assembly process (e. g. variations in the position of the weld seam) might also affect the response variables. Nevertheless, all these factors were neglected.

With 3 factors at 2 levels (-1 and 1) a full factorial design results is 8 experiments. An additional true centre point cannot be realised because factor B is a categorical factor with just two levels. Instead, experiments with factor level combinations -0.32/-1/0 and -0.32/1/0 were added to the design (Table 20).

To determine the minimum number of experiments it is necessary to define a difference in the effect to be detected. This difference is set to 0.2 mL of volume displacement. With this level it is possible to distinguish actuators with substantial different V5 value and substantial different G235 value. The standard deviation is set to 0.07 ml according to the average of the preliminary experiments. The significance level (probability to detect a difference when there is no difference) is set to 5%. The power (probability of detecting a difference when there is a difference) is set to 80%. The maximum number of factor levels is 3. With these specifications the minimum number of replications was determined to 4 (The no.reps function of the dae package of R was used). With double replication of the experimental design every factor level is replicated at least 4 times while most factor levels are replicated 8 times. Overall, 20 experiments were performed.

The actuators were manufactured in blocks (depending on the filter area). Within each block fabrication was done in randomized order. The complete experimental design is presented in Table 20. The experiments were performed as described already. The data of the response variables was obtained from the transient signals and is also presented in Table 20. Additionally, to get the true amount of absorbed water the actuator cavities were opened after each experiment. The gel was removed and weighed. By subtracting the dry polymer mass one gets the desired mass of the absorbed amount of water.

94 Table 20: Experimental design and results of the 2³ factorial design with double replication and additional “centre” points. (Data presented without randomization.)

Response variables of displaced mass / g Factor levels After 5 min Gain by 4 h Set A B C 1 2 1 2 1 -1 -1 -1 0.155 0.189 1.005 0.978 2 -1 -1 1 0.117 0.072 0.850 0.843 3 -1 1 -1 0.955 0.856 0.818 0.870 4 -1 1 1 0.470 0.460 0.803 0.800 5 -0.32 -1 0 0.277 0.322 1.141 1.134 6 1 -1 -1 0.363 0.317 1.837 1.577 7 1 -1 1 0.208 0.200 1.447 1.479 8 1 1 -1 2.392 2.267 1.452 1.327 9 1 1 1 1.103 1.104 1.136 1.160 10 -0.32 1 0 0.813 0.878 1.067 0.965

Experimental Result and Data Analysis

First, the absorbed amount of water was compared to the displaced volume obtained by conversion of the acquired mass data of the balance. On average the displaced volume is 74 % ± 4 % of the total absorbed volume. Hence, not all the absorbed volume contributes to the desired displacement. This is partly caused by superseding the air between the particles of the granulate in the actuator cavity with water, compression of the remaining air and the limited stiffness of the rigid filter membrane. So, one way to increase the efficiency it to use a stiffer filter membrane. Another way to increase efficiency is to replace the air in the cavity by a suitable material or to decrease the actuator cavity volume to a minimum that is needed for a designated volume displacement.

Next, the data was analysed based on a full model considering all three single factors (A, B and C), all three two factor interactions (AB, AC and BC) and the three factor interaction (ABC). The ANOVA table for the response V5 is given in Table 21. The model is significant with very low pure error. However, there is significant lack of fit. So, there is nonlinear curvature in the data that cannot be represented by the model. Therefore, the model is accurate in the corners of the design space and in a region close to the corners. The residual plots of the model can be found in Figure A.1 & Figure A.2. There is no indication that the assumptions on normality and constant variance are violated. Outliers are not present.

The response G235 was analysed based on a full model. The model was reduced by removing insignificant terms. A transformation of the response with the ln function was indicated by the box cox17 method due to non-constant variance. The ANOVA table of the transformed response is presented in Table 22. The model is significant with very low pure error and lack of fit is insignificant. The residuals of the transformed response G235 are plotted in Figure A.3 & Figure A.4. The transformation solved the problem of non-constant variance (with Figure A.4, right as a potential ex- ception). The assumption of normality is acknowledged and outliers could not be identified.

17 The boxcox function of the MASS package of R was used.

95 Table 21. ANOVA for the response displaced volume after 5min.

Sum of Degrees of Mean Squares Freedom Square F0 P-Value Model 8.157 7 1.165 161.34 <0.0001 A 1.421 1 1.421 196.70 <0.0001 B 4.403 1 4.403 609.72 <0.0001 C 0.884 1 0.884 122.35 <0.0001 AB 0.894 1 0.894 123.77 <0.0001 AC 0.178 1 0.178 24.66 <0.0001 BC 0.528 1 0.528 73.08 <0.0001 ABC 0.132 1 0.132 18.30 0.0011 Residuals 0.087 12 0.007 Lack of Fit 0.068 2 0.034 18.34 0.0005 Pure Error 0.019 10 0.002 Cor Total 8.243 19

The full model of the data V5 is shown in Figure 6.7. The left contour plot shows the model with factor B (type of SAP) at level -1 the right contour plot shows the model with factor B (type of SAP) at level 1. The model standard deviation is ±0.09 g. With factor B (type of SAP) at level -1 (VP-Fines) there is much less influence of the factors A (diameter) and factor C (hardness of membrane) on V5 compared to level 1 (SX-Fines).

The back transformed reduced model of the data G235 is shown in Figure 6.8. The left contour plot shows the model with factor B (type of SAP) at level -1 (VP-Fines) the right contour plot shows the model with factor B (type of SAP) at level 1 (SX- Fines). The model standard deviation was calculated explicitly at data points of the transformed model and back transformed to the original scale. It ranges from 0.05 ml to 0.10 ml with factor B at level -1 and from 0.04 ml to 0.08 ml with factor B at level 1.

With factor B (type of SAP) at level -1 (VP-Fines) the V5 value (Figure 6.7) is significantly smaller than with factor B (type of SAP) at level 1 (SX-Fines) regardless of the level of the factors A (filter area) and C (hardness of membrane). This is explained by the fast formation of a dense gel layer with VP-Fines (level -1) compared to SX-Fines (level 1) as already explained (see chapter 2.4.2). So, to get a volume displacement actuator with a small value of V5, it is beneficial to set factor B (type of SAP) to level - 1 (VP-Fines).

Table 22. ANOVA for the transformed response gain at 4 h.

Sum of Degrees of Mean Squares Freedom Square F0 P-Value Model 1.149 4 0.289 94.70 <0.0001 A 0.938 1 0.938 309.22 <0.0001 B 0.127 1 0.127 41.74 <0.0001 C 0.076 1 0.076 24.98 0.0002 AB 0.015 1 0.015 4.83 0.0442 Residuals 0.046 15 0.003 Lack of Fit 0.022 5 0.004 1.87 0.1874 Pure Error 0.024 10 0.002 Cor Total 1.195 19

Figure 6.7. Response variable V5 (displaced volume after 5 min.) as a function of diameter (factor A) and hardness of the elastomeric membrane (factor C) as contour plot. Left: VP-Fines (factor B at level -1). Right: SX-Fines (factor B at level 1).

Figure 6.8. Response variable G235 (displaced volume between 5 min. and 4h) as a function of diameter (factor A) and hardness (factor C) as contour plot. Left: VP-Fines (factor B at level -1). Right: SX-Fines (factor B at level 1).

The highest values of G235 can be achieved with factor B (type of SAP) set to level -1 (VP-Fines). This gets clear by comparing the contour plots of Figure 6.8. So, both goals, minimized V5 and maximized G235 can be accomplished with factor B (type of SAP) at level -1 (VP-Fines).

The levels of factors A (filter area) and C (hardness of membrane) are contradictory concerning the minimisation of V5 and maximisation of G235. This gets clear from the nature of these factors. If the filter area is larger the flow of water into the actuator is higher. Consequently, V5 as well as the G235 are increasing. If the hardness of the elastomeric membrane increases the pressure inside the actuator cavity is higher at identical cavity volumes. So, if the swelling process is influenced by the pressure in the cavity V5 as well as G235 are reduced if a membrane with higher hardness is used.

Test of the Model

To demonstrate the predictive capability of the model 4 experiments with two parameter sets not included in the designed experiment were performed. The first parameter set (A: -0.32, B: -1 and C: 1) should result in a quite low V5 with a predicted range of 0.16 ml ± 0.09 ml and a reasonable G235 with a predicted range of 1.03 ml ±

97 0.06 ml. The V5 of the two performed experiments is 0.15 ml and 0.14 ml. Both are within the predicted range. The G235 was experimentally determined to be 1.19 ml and 1.07 ml. So, the first result is larger (+0.1 ml) than the predicted range. However, the deviation from the model is smaller than the difference in effect to be detected (0.2 ml) defined in the design of the experiment (chapter 0). The second parameter set (A: 1, B: -1 and C: 0) should result in a moderate V5 with a predicted range of 0.29 ml ± 0.09 ml and a high G235 with a predicted range of 1.59 ml ± 0.09 ml. The V5 of the two experiments resulted in 0.33 ml and 0.29 ml. Both are inside the predicted interval. The G235 is 1.56 ml and 1.66 ml. Both values are in the predicted interval. The performed tests demonstrated the capabilities of the models that live up to the defined expectations.

Volume displacement with back pressure

The influence of back pressure on the volume displacement process of SAP actuators was studied at 5 pressure levels from 50 kPa to 200 kPa. The following actuator design was selected: The inner diameter was set to 17 mm (level A at -1). The polymer VP-Fines was used (factor B at level -1). An elastomeric membrane with shore hardness of A50 was selected (factor C at level 0). Apart from the previous actuator designs the mass of SAP polymer was set to 0.10 g. This is a reduction of 0.13 g. The conclusions of the following experiments will show that this actuator design is robust to changes in SAP mass. The mesh width of the filter membrane was set to 50 µm. This smaller mesh width is necessary for experiments with back pressure. With larger mesh widths (e.g. 105 µm) the hydrogel can leave the actuator cavity too easily. According to the model the selected actuator design is supposed to show a small V5 of 0.17 ml ± 0.09 ml and a moderate G235 value of 0.93 ml ± 0.05 ml at 0 kPa back pressure.

Figure 6.9. Transient behavior of the volume displacement of SAP actuators at different back pressures. The mean of 5 SAP actuator and the standard deviation (at selected points) is plot- ted18.

18 Experiments performed and data acquired by Schmidt C [119].

98 20 actuators were studied at 4 pressure levels ranging from 50 kPa to 200 kPa (Figure 6.9). A linear fit to the data results in coefficients of determination from 0.94 to 1.00 (Fits with y-axis-intercept=0 result in coefficients of determination from 0.75 to 1.00). Hence, the displacement of volume shows approximately zero order kinetics within the first 4 hours. The approximation is supposed to be adequate for drug delivery devices. The same data is presented as a function of back pressure after 4 h in Figure 6.10. The G235 value drops from a level of 0.81 ml ± 0.03 ml at 50 kPa to 0.28 ml ± 0.06 ml at 200 kPa. The data fits very well to a straight line with a coefficient of determination of 0.99. The extrapolated volume displacement for a back pressure of 0 Pa is 1.00 ml. It is just slightly higher than the predicted value of 0.93 ml ± 0.05 ml. This is remarkably because the mass of SAP is lower compared to the actuators used in the factor screening experiment. It supports the argument stated at the beginning of chapter 6.1.3: changes in mass of the SAP are less relevant. 6.1.4. Summary

A flat SAP volume displacement actuators is characterised in this chapter. In initial experiments some of the actuators showed a substantial fast displacement of volume in the first 5 min. followed by a slow but continuous increase of displaced mass for more than 4 hours. With a factor screening experiment the influence of the type of polymer, the hardness of the elastomeric membrane mediating the volume displacement and the filter area on the actuator performance were studied. The model that resulted from the experimental data enables to adapt the parameters to specific needs of a desired application. E.g. it was possible to identify a parameter set for an SAP actuator having a low displaced volume V5 after the initial phase of 5 minutes and a large displaced volume G235 from 5 minutes to 4 hours. Manufacturing can be as simple as suggested with anticipated low fabrication costs. An actuator design that shows almost zero order kinetics was studied further and its back pressure dependent volume displacement was characterized. It might be neglected in applications with small variations in back pressure (e.g. < 25 kPa corresponding to <10% variation in delivered volume).

With their ability to displace relevant volumes of liquid even against an elevated pressure level (e. g. 100 kPa) SAP volume displacement actuators can enable the realisation of thin skin attached drug delivery devices. A volume of approximately 1 ml can be delivered within 4 h or faster if necessary. However, in order to create a drug delivery device it is necessary to add an activation mechanism, a drug reservoir and a case in a suitable way. 6.2. The ChronopaDD system

An implementation of a drug delivery device with integrated SAP actuator is presented in Figure 6.11. The SAP actuator is placed between the activation mechanism and the drug pouch within a case. The activation mechanism consists of the expanding agent pouch, the check valve, the delay pouch and the delay pellet. The drug pouch is connected to an intradermal infusion set with a cannula. If the SAP actuator expands it compresses the drug pouch and its content is delivered through the cannula to the skin of the patient. The assembled device is presented in Figure 6.13.

Figure 6.10. Displaced mass after 4 h as a function of back pressure18.

The cannula of the infusion set is closed with a rubber cap. So, the drug solution can be stored in the drug pouch. The solution is just in contact with the pouch material, the stainless steel cannula, plastics of the intradermal infusion set, glue (mounting of the cannula) and the rubber. For storage purposes the pouch material must be transparent. So, the content of the pouch can be checked visually before application.

The SAP actuator and the drug pouch are arranged in the ChronopaDD in such a way that the actuator can squeeze the drug pouch. The drug pouch, the delay pouch and the SAP actuator must be located in a stiff packaging. Otherwise, external pressure can influence the infusion process.

Figure 6.11. Detailed view of the ChronopaDD and its components. The intradermal interface consist of the fixed block, the slider, the shafts and the cannula. The drug pouch and the SAP actuator are explained in detail in the text. The time delayed activation mechanism consist of the expanding agent pouch, the check valve, the delay pellet and the delay pouch.

100 Figure 6.11 shows the device during storage (step I.). The activation procedure starts by removing the rubber cap, fixation of the device on the skin surface and placing of the cannula in the skin (Figure 6.12, step II.). Next, the expanding agent needs to be transferred to the delay pouch. This is done by squeezing the expanding agent pouch manually. Accordingly, the peelable seal opens and the liquid is transferred through the check valve (step III.). Optionally, the expanding agent pouch can be removed, now. The check valve keeps the expanding agent in the delay pouch. Next, the delay pellet dissolves within a certain time. After this process the passage between the delay pouch and the SAP actuator is free (step IV.). Now, the expanding agent can enter the SAP actuator where it is absorbed by the SAP granulate. As a consequence the volume of the SAP actuator increases and deflects the elastomeric membrane. Thereby, it squeezes the drug pouch. Hence, the drug solution is transported via the cannula of the intradermal infusion set into the patient’s skin (step V.).

In the subsequent chapters the components of the ChronopaDD are explained in more detail. Then the performance of the ChronopaDD is demonstrated in various experiments including ex vivo infusion experiments. The possibility to tune the delay time is also described. 6.2.1. Components of the ChronopaDD and device assembly

Intradermal infusion set

A miniaturized intradermal infusion set was designed for the ChronopaDD. The miniaturized set is based on the laboratory setup (chapter 4.3.1) and realizes a fixed insertion parameter set (4/2). A first version of the miniaturized set is integrated in the ChronopaDD displayed in Figure 6.13. A second version is separated from the ChronopaDD and enabled the integration of pressure and flow sensors for experiments in the lab (Figure 6.21).

Figure 6.12. Subsequent steps of ChronopaDD usage. (The components are labeled in Figure 6.11). II. The cannula of the intradermal infusion set is inserted into the skin. III. The content of the expanding agent pouch is transferred to the delay pouch. IV. The delay pellet is completely dissolved. V. The expanding agent is absorbed by the SAP actuator and the content of the drug pouch is infused into the skin.

101

Figure 6.13. Assembled functional model of the ChronopaDD. The device dimensions are: 50 mm x 30 mm x 10 mm without expanding agent pouch.

Drug pouch

The drug pouch design is depicted in Figure 6.14. It consists of a drug reservoir with a flexible channel made of the same flexible and transparent film. The design is derived from a state-of-the-art 3 side seal pouch [120].

Pouch fabrication follows a 4 step protocol. First, the film material (PE-film with a thickness of 0.10 mm) is folded and the longitudinal seal is welded (welding tong hpl ISZ by hawo GmbH, Obrigheim, Germany) and fold back. Second, the transverse seal is welded. Third, an access hole is punched in one layer of the film at the end of the flexible channel. Finally, the flexible channel is generated by welding the form seal. A small tube (FEP, outer diameter=1.6 mm, inner diameter=0.8 mm) was placed in the flexible channel in order to decrease its flow resistance. Optionally, the superfluous film material can be cut off. The width of the pouch is 25 mm, the length of the drug reservoir and the flexible channel is 30 mm, respectively.

Without additional thermoforming of the film the pouch can only be filled with a positive pressure. It fills with 1.7 ml of water at a pressure of 1.2 kPa which corresponds to a water level of 12 cm. Of course, in combination with the fluidic resistivity of the cannula (inner diameter: 0.13 mm, length: 30 mm) this pressure generates an outflow (16 µl/min) of its content as soon as the rubber cap is removed. Although, this outflow is very low it can be further reduced by additional thermoforming of the pouch.

Figure 6.14. Layout of the pouch used as drug container. The dashed circle (diameter 17 mm to scale) marks the area where the pouch is squeezed by the SAP actuator.

102 In this work the drug pouch is bonded to the intradermal infusion set or a corresponding adapter with double side adhesive. In a large scale manufacturing environment a welding process would be utilized. Large scale pouch fabrication may follow the listed 7 steps: 1. Punching of a hole into the plain transparent film 2. Welding of the intradermal infusion set to the film 3. Pouch forming with a forming shoulder 4. Generation of the longitudinal seal 5. Generation of the form seal 6. Filling of the drug pouch 7. Sealing by generation of the transverse seal

These fabrication steps are state of the art in the packaging industry [120]. The same sequence is used e.g. for the packaging of coffee beans. Valves are welded to the film of the packaging than the pouch is formed, filled with beans and sealed. Gas generated by the beans can leave the pouch through the valve while the flavor is contained.

Superabsorbent polymer (SAP) actuator

The SAP actuator structure and its assembly are described in detail in chapter 6.1.1. The dimensions of the circular actuator cavity of the ChronopaDD are: inner diameter=17 mm and thickness=1 mm. The actuators were filled with 0.10 g or 0.16 g of SAP granulate (SXM 9410). A polyester mesh with a mesh opening of 40 µm was used as filter. Elastomeric membranes with Shore hardness of A30 and A50 were used. In some experiments the actuator cavity was vented. For this purpose 3 cannulas (G27, with outer diameter 0.41 mm and inner diameter 0.21 mm) were inserted through the weld seam of the elastomeric membrane. Swollen SAP (hydrogel) did not leave the cavity through the venting cannulas in none of the performed experiments. This is due to the high aspect ratio of the cannulas.

Delay pellet

The core component of the delay pellets are candy discs which were manufactured according to the following protocol. Sucrose was heated to get a melt without crys- tals. This melt was poured into a rubber mould with a diameter of 12 mm and a thickness of 0.7 mm. (With a density of 1.6 g/cm³ of sucrose the weight of a candy disc is approximately 0.13 g.) A sheet of rubber was placed on the filled mould and charged with a flat and heavy piece of metal until cooled. If too much melt was poured into the rubber mould the excess candy formed a thin burr. It was removed carefully before the disc was placed in a cabinet desiccator until needed.

103

Figure 6.15. Delay pellet. Top: Cross section. Bottom: Top view.

The candy disc was bonded to a stack of films. Double side adhesive (universal by Tesa, Hamburg, Germany), PP-film (Goodfellow GmbH, Bad Nauheim, Germany) with a thickness of 0.18 mm and double side adhesive were used. This stack increased the mechanical stability of the disc (Figure 6.15). A hole was punched in the centre of the stacked films for fluid communication. On its other side the candy disc was covered with clear double side adhesive (Montageband transparent, Tesa, Hamburg, Germany). It enabled to observe the dissolution of the candy disc. The circumferential area of the candy disc was covered with a stained varnish (UHU plus schnell fest by UHU, Germany, mixed with blue Acrydye by Struers, Germany). The 4 notches in the varnish were created with a handheld drill with a 0.1 mm thick cutting blade.

Just at the 4 notches the expanding agent can get access to the candy disc. In this way the dissolution speed of the candy disc is reduced. In large scale manufacturing it might be challenging to realize the circumferential coating with the 4 notches. Apart from sucrose there might be materials that don’t need the circumferential coating. In this case large scale manufacturing of the delay pellet is considered uncomplicated.

Figure 6.16. Check valve. Top: Cross section. Bottom: Top View.

104 Check valve

Manufacturing of the check valve (Figure 6.16) started by punching a hole (diameter 1.5 mm) in PP-film (thickness 0.3mm, Goodfellow GmbH, Bad Nauheim, Germany). A thermoplastic elastomeric (TPE) film (hardness Shore A50, thickness=0.2 mm) was welded to the PP-film above the hole with a ring shaped welding tool (inner diameter 4 mm and width 1.5 mm). A scalpel was used to create a cut ( 2 mm to 3 mm) in the elastomeric film. Double side adhesive was punched and bonded to both sides of the PP-Film. The ring shaped PP-spacer (thickness 1 mm) was attached to the face of the PP-film with the elastomeric membrane.

A positive pressure gradient in flow direction deflects the TPE-film and generates a flow path at the site of the cut in the TPE-film. A negative pressure gradient in flow direction seals the hole in the PP-film and prevents liquid flow. Large scale manufacturing of the check valve is considered uncomplicated.

Expanding agent pouch

The design of the expanding agent pouch is depicted in Figure 6.17. It consists of the reservoir and a flexible channel made of translucent peel film (1275.PPT.000, Orbita- Film, Weißandt-Gölzau, Germany). The permanent weld seals were done with the welding tong at elevated temperature. The peelable seal was done with a corner shaped (angle 90°, width 1.5 mm) welding tool. Heated to a temperature of 160° C the tool was pressed against the film layers with a pneumatic cylinder at a pressure of 200 kPa for 5 sec. Filling and sealing of the expanding agent pouch is done during the final assembling of the ChronopaDD.

Delay pouch

The delay pouch is manufactured from the same PE-film as the drug pouch. It is a simple 3 side pouch with two holes on opposing sides. The width of the pouch is 25 mm and its length is 30 mm, weld seals included. The holes with a diameter of 3 mm are centered with respect to the small side and are located 8 mm from the end of the pouch, respectively. One transverse seal of the pouch is not done until final assembly. So, the delay pellet can be mounted inside the pouch centered above one of the punched holes.

Figure 6.17. Layout of the expanding agent pouch.

105 Activation mechanism with time delay

The activation mechanism with time delay consists of the delay pouch, the delay pellet, the check valve and the expanding agent pouch filled with expanding agent. It is assembled in the following way: The expanding agent pouch and the delay pouch were bonded to the check valve. Next, the final transversal seal of the delay pouch was done with a vacuum packaging machine (VP440, Vama Maschinenbau, Wild- poldried, Germany). This way, the formation of air bubbles can be neglected when the expanding agent enters the delay pouch. Finally, the expanding agent pouch is filled with expanding agent (2 ml of water or sucrose solution) and sealed by welding its final transversal seal. The seal is done carefully in order not to include any air bubble and without spilling large amounts of expanding agent. The complete activation mechanism can be bonded to the SAP actuator.

Large scale manufacturing of the activation mechanism must be done in a different order. Although, all the fabrication steps are state-of-the-art the complete sequence is new. There are several possibilities to design the fabrication sequence. One might be: 1. Punching two holes in a film for the delay pouch 2. Mounting of the delay pellet above one of the holes 3. Mounting of the expanding agent pouch (fabricated on another pouching line) 4. Forming of the delay pouch with a forming shoulder 5. Generation of the longitudinal seal 6. Generation of the transverse seal 7. Transfer to the vacuum sealing machine 8. Vacuum sealing and generation of the final transverse seal

ChronopaDD assembly

The final fabrication procedure of the ChronopaDD consists of several assembling steps. First, the drug pouch is placed on the carrier plate (Figure 6.18). Second, the SAP actuator is mounted above the drug pouch. It rests on a supporting frame with a thickness of 2 mm. Third, the actuator is covered with a frame of double side adhesive. The cavity of this frame is filled with a spacer mesh (mesh opening 0.3 mm, Figure 6.11). The double side adhesive and the spacer mesh are covered with a PP- film with a thickness of 0.18 mm and a punched access hole (diameter 1.5 mm) centred above the SAP actuator. The spacer mesh allows the expanding agent solution to enter the SAP actuator via an area much larger than the access hole with its diameter of 1.5 mm. The activation mechanism is bonded to the PP-film with double side adhesive. Certainly, the access holes for fluid communication are oriented one upon the other.

The combination of the SAP actuator (with elastic membrane diameter of 17 mm) and a frame with a thickness of 2 mm results in a maximum volume of 0.45 ml that can be displaced, theoretically.

For the characterisation experiments an adapter was connected to the drug pouch instead of the intradermal infusion set. So, it was possible to integrate pressure and flow sensors in the setup.

106 Carrier plate, frame and SAP actuator were fixed with screws for convenience (Figure 6.18). These connections need to be realised by plastic welding in a large scale manufacturing process. Such a process is considered uncomplicated. 6.2.2. Characterization of SAP actuators with sucrose solution

The time delay of the ChronopaDD is realised with a delay pellet made of sucrose. After dissolution of this pellet the expanding agent is no longer DI water. Instead a sucrose solution is the expanding agent. Sucrose solution is an osmotic agent. Therefore, it is expected, that the efficiency of the SAP actuators is reduced with sucrose solution as expanding agent. The extent of this reduction is investigated in this chapter.

Experiments with expanding agent with a sucrose concentration of 40 % by weight were performed to get a first impression on the reduction of the performance. The experimental setup depicted in Figure 6.4 was used. SAP actuators (0.1 g of SXM9140 granulate, elastomeric membrane with hardness Shore A50, inner diameter 17 mm, outer diameter 50 mm, thickness 1 mm, mesh opening 0.04 mm) were primed with 0.20 ml of expanding agent (40 % by weight sucrose solution). The primed actuators were mounted on the glass flange, the system was filled with water and the pressure controller was set to the desired back pressure level. After the decay of the pressure transient, caused by the capacity of the actuator, 2 ml of expanding agent (40 % by weight sucrose solution) were dispensed to the cavity above the SAP actuator to start the expansion process. The dynamics of the expansion process was acquired with the flow sensor at 10 Hz. The integrated data was cross- checked by comparing it to the absolute displaced mass obtained with the pipette.

Figure 6.18. Assembled functional model of the delivery device with integrated SAP actuator (cover removed). The dimensions of the carrier plate are 50 mm x 30 mm.

107

Figure 6.19. Displaced volume vs. time of SAP actuators (n=3) filled with SXM9140. Sucrose solution at 40% by weight was used as expanding agent. Displaced volume vs. time of SAP actuators expanded with di water is presented in Figure 6.6. However, the SAP actuators were not primed in the experiments of Figure 6.6 and a different mesh size of the filter was used. A comparison with no difference in experimental parameters is presented in Figure 6.20.

The dynamics of the expansion process with sucrose solution at 40% by weight at different back pressure levels is given in Figure 6.19. With 0 kPa back pressure the SAP actuators expand quite rapidly at first before it continues to swell at a rate of 0.04 ml/h. At a back pressure of 50 kPa and 100 kPa this rapid phase is no longer present. Within the studied interval of 2 h the average rates of expansion were 0.08 ml/h and 0.06 ml/h respectively.

The total displaced volume within 2 h by SAP actuators as a function of back pressure is presented in Figure 6.20. It is compared to data of actuators expanded with DI water. The displaced volume of the actuator is almost divided in halves if sucrose solution at 40 % by weight is used at 0 kPa back pressure. With increasing back pressure this ratio is even worse. Although it is possible to use sucrose solution as expanding agent this result implies practical issues in SAP actuator usage in the ChronopaDD. The efficacy of the actuator is definitively reduced. Considering a delay pellet with a mass of 0.13 g is dissolved in 2 ml DI water a sucrose solution with 6 % by weight is generated. So, with a single delay pellet and DI water as expanding agent there is probably just a minimal impairment of the hydrogel actuator performance. However, after dissolution of the delay pellet the first droplet of expanding agent reaching the SAP actuator has a much higher sucrose concentration than the theoretically calculated even concentration. Both issue need to be considered.

6.2.3. Characterization of SAP actuator powered intradermal infusions

The experimental setup to proof the feasibility of SAP actuator powered intradermal infusions by squeezing a pouch is presented in Figure 6.21. A frame was mounted on the SAP actuator (0.15 g ± 0.01 g SXM9410 granulate, elastomeric membrane with hardness Shore A30, diameter of 17 mm, thickness of 1 mm and mesh width of 40 µm) to store expanding agent. This cavity can contain approximately 1 ml of expanding agent. In order to integrate flow (1 g/h µFlow, Bronkhorst, Ruurlo, the Neth- erlands) and pressure (100 kPa, EPX, Entran Sensors, Potterspury, UK) sensors into the experiment the intradermal infusion set was connected via an adapter and tubing (FEP, outer diameter 1.6 mm, inner diameter 0.8 mm) with the drug pouch.

108

Figure 6.20. Displaced volume of SAP actuators within 2 h as a function of back pressure (n=3). DI water and sucrose solution at 40% by weight were used as expanding agent.

The drug pouch was bubble free primed with infusion solution (0.02 % by weight methylene blue in 0.9 % by weight NaCl solution) from a 12 cm elevated beaker via the fill port. This way, the drug pouch was filled with 0.68 ml ± 0.05 ml (n=9) of infusion solution.

The carrier plate with the intradermal infusion set was attached to an ex vivo pig ear with double side adhesive (Figure 6.22). The cannula was inserted into the skin by pushing the slider of the intradermal infusion set. The experiments were started by dispensing 1 ml (limited reservoir size) of expanding agent (DI water) to the cavity above the SAP actuator. The data of the sensors was recorded with an AD converter at 10 Hz.

Leak proof infusion of stained liquid into the ex vivo tissue was observed. The data of 4 infusions experiments is presented in Figure 6.23. The data was recorded for 1 h due to the limitation of the amount of expanding agent. The delivered volume ranges from 0.23 ml to 0.37 ml with back pressures ranging from 9 kPa to 18 kPa. Noticeable, is the large gradient of the delivered volume vs. back pressure. It can only be explained by an additional fluidic capacitance introduced by the flexibility of the drug pouch. This capacity is determined to be 0.02 ml/kPa.

Figure 6.21. Experimental setup for ex vivo infusion by squeezing a pouch with a SAP actuator.

109

Figure 6.22. Photo of the experimental setup according to the schematic illustration of Figure 6.21.

The data clearly demonstrates that it is possible to infuse relevant amounts of liquid into the upper skin layers by squeezing a pouch. Even with sucrose solution it should be possible to deliver a relevant amount (0.1 ml to 0.2 ml) of drug solution to the skin. The fluidic capacity introduced by the flexible pouch introduces a high level of uncertainty over the infusion rate. This is not relevant if the infusion rate is not important for the desired application. In all other cases the device needs to be redesigned in order to reduce the pouch capacity. This could be realized e.g. by structures supporting and confining the pouch. 6.2.4. Characterization of the ChronopaDD

The ChronopaDD infusion device was characterised by two different approaches. In the first one the time delay mechanism was characterised with defined back pressure. With the second approach experiments with time delayed ex vivo intradermal infusions were performed. ChronopaDD infusion devices according to Figure 6.11 and Figure 6.18 were used in all these experiments. The drug pouch was always connected via an adapter to obtain measurement data.

Figure 6.23. Delivered volume of infusion solution within 1 h as a function of infusion back pressure.

110 The time delay mechanism of the ChronopaDD was characterised with a light back pressure. The drug pouch was in fluid communication with an elevated beaker (12 cm above the ChronopaDD corresponding to a pressure of 1.2 kPa) placed on scales with a PC interface. The beaker was filled with water that was covered with paraffin oil to prevent evaporation. The time delay mechanism of the ChronopaDD was studied by activating the ChronopaDD device and recording the increase in mass on the scales. Three different expanding agents (DI water, sucrose solution 33 % by weight and 50 % by weight) were used with three ChronopaDDs, respectively. Overall 9 experiments were performed. The experiments were started by pressing the expanding agent pouch and thereby transferring its content to the delay pouch.

The time delay of the ChronopaDD with three different expanding agents is presented in Figure 6.24. The delay time ranges from 32 min. to 284 min. for sucrose concentrations from 0 % to 50 % by weight, respectively. Clearly, the time delay is a function of the concentration of sucrose in the expanding agent.

Beneficial is the possibility to adjust the time delay by varying the sucrose concentration. Unfavourable, from a fabrication point of view, is the large gradient between sucrose concentrations of 33 % by weight and 50 % by weight. It requires precise ad- justment of the sucrose concentration. Additionally, a high temperature dependence is expected.

The total displaced volume 2 h and 4 h after the activation of the ChronopaDD is given in Figure 6.25. It ranges from 0.40 ml to 0.55 ml. The increase in volume between 2 h and 4 h is marginal, indicating that the expansion process is practically finished within 2 h. The displaced volume within 2 h decreases approximately by 20 % with sucrose concentrations between 0 % and 50 % by weight. Based on a pouch fill volume of 0.68 ml this corresponds to a delivery efficiency of 60% to 80% of the stored drug. This might be an issue e.g. if the drug solution is expensive.

With these experiments the delivery of a relevant amount of liquid (> 0.4 ml) with time delay (up to 4.7 h) at a constant back pressure of 1.2 kPa was demonstrated. So, the functionality of the ChronopaDD under well-defined lab conditions is proven.

Figure 6.24. Delay time of the ChronopaDD as a function of sucrose concentration in the expanding agent (n=3).

111

Figure 6.25. Displaced volume as a function of sucrose concentration in the expanding agent (n=3).

Time delayed ex vivo intradermal infusions with the ChronopaDD were characterised with flow and pressure sensors. Of course, the intradermal infusion set was separated from the ChronopaDD devices (see Figure 6.21) to integrate the sensors. The expanding agent pouches were filled with a 40 % by weight sucrose solution. This corresponds to a delay time of 164 min. (with a maximum deviation of ± 24 min. Val- ues are obtained by interpolation from Figure 6.24.)

After preparation of the experimental setup and insertion of the cannula into the ex vivo pig skin the experiments were started by transferring the content of the expanding agent pouch to the delay pouch.

The time to dissolve the delay pellets of the 3 ChronopaDDs studied is 116 min ± 10 min. The time delay is 48 min. shorter than predicted. So, the error in delay time is most probably caused by preparing sucrose solutions with inaccurate concentration. Variations in temperature in the lab are small and can be neglected most probably.

After dissolution of the delay pellet the pressure increased instantaneous to 8.2 kPa ± 1.6 kPa (n=3). However, only in one case the intradermal infusion started as desired and delivered a volume of 0.09 ml at a back pressure of 6.4 kPa within 2 h (0.15 ml within 4h). In the other two cases additional delay times of 284 min. and 388 min. elapsed until the infusion began. During that period the pressure increased to 9.1 kPa ± 1.1 kPa (n=2). Afterwards, the pressure dropped to 1.5 kPa ± 0.0 kPa (n=2) and a volume of 0.35 ml ± 0.05 ml (n=2) was infused within 2 h (0.40 ml ± 0.05 ml within 4h). This is comparable to the result presented in Figure 6.25.

The back pressure of the experiment without additional delay time agrees very well with the data presented in Figure 5.15. The back pressure of the infusions with additional time delay is much lower than expected. Unfortunately, there is no clearly identified reason for this observation. It might be due to a highly viscous and concentrated plug of sucrose solution. However, this plug can not explain the drop in back pressure after the additional delay time. Clogging of the cannula can not be excluded. A resolved clogging condition can be the explanation for the observed pressure drop. Altogether, the additional time delay is inadmissible.

As already stated, the hydrogel actuator in general can cope with back pressures observed in the ex vivo experiments so far. Therefore it is necessary to identify other

112 soluble materials to realise the time delay functionality. The osmolarity of this material must be as small as possible. However, a highly concentrated and pontentially viscous solution of whatever material might always be a problem of time delayed activation of the ChronopaDD. 6.2.5. Summary

The experiments showed that infusing a liquid drug intradermally by squeezing a pouch is feasible. Such an approach is advantageous because the fabrication of pouches is state-of-the-art. However, the drug pouch introduced an undesirable fluidic capacity into the system. With optimisation of the pouch geometry, additional thermoforming of the film prior to pouch formation and supporting and confining structures the capacity might be substantially reduced and the efficacy of discharging the drug pouch might gain up to 100 %.

With the presented data the operability of the ChronopaDD concept is proven. Time delay and intradermal infusions were demonstrated in ex vivo pig ears. Nevertheless, the technology needs to be improved. One approach would be the identification or the development of a suitable material for the delay pellet that does not decrease the chemical potential of the expanding agent.

The costs of the ChronopaDD are anticipated low and it can be disposed easily without separating electronics or batteries from the device. It is small with a thickness of 12 mm. So, it can be worn in a comfortable way because its dimensions are comparable to a wrist watch. Fabrication is generally derived from state-of-the-art packaging processes. Consequently, this concept is suitable for large volume manufacturing because the output of packaging machines is very high.

Without further optimisation different applications need to be identified. Another approach would be the design of an application with low and constant back pressure. One could think of replacing the intradermal interface with a moist cellulose swab. In this case the lab performance of the ChronopaDD would be applicable. Such a modified ChronopaDD could be used in combination with another skin poration technology. After creation of the pores they can be occluded by the moist cellulose swab. This way they stay open until the drug solution is transferred to the swab. This could be accomplished by time delayed liquid transfer with the ChronopaDD. Then, the drug molecules can diffuse into the body via the previously generated pores.

113

114 7. Conclusions

The focus of this work is set on the proof of principle of intradermal chronopharmaceutical drug delivery with skin attachable, flat and small delivery devices not based on electronic components and batteries. A prior condition was the development of a suitable technology to perform intradermal infusions and injections reliably. The fluidic characterisation of this technology provided data that allows the assessment of the chances of successful combination with a certain actuator technology. A hydrogel actuator technology was characterised in this work and demonstrated promising performance. It enabled the design of a delivery device without electronics and batteries. The time delay functionality is based on a physico-chemical dissolution process. Large scale fabrication of the device is supposed to be economical because most fabrication steps are derived from the packaging industry.

Hollow microneedle arrays were considered to perform intradermal infusions and injections. Thermoforming of polymers is a technology that was investigated to manufacture such microneedles. Thereby, several hollow microneedles can be combined on a single substrate. The possibility to fabricate such arrays in thermoplastics sim- plifies the integration into devices that are also manufactured from thermoplastics. Welding is the joining method of choice. Although, the hot embossing process developed so far is quite laborious it offers the generation of small amounts of arrays without expensive mould fabrication. Large scale manufacturing of thermoplastic microneedle arrays was demonstrated by the development of an injection moulding tool and process. These needles are quite robust and facilitate intradermal injections and infusions. The need for high speed application needs to be considered in the product design. The required contact pressure for leak tight delivery is higher than desired. Therefore, this technology is notably qualified for its usage in short time injection applications.

An intradermal infusion set based on the traditional Mantoux technique was developed as laboratory tool and in vivo device. It is particularly suitable for long term intradermal infusions. The device is quite simple in design and fabrication. In ex vivo and in vivo experiments the device demonstrated very good data on leak tightness. This is important to guarantee a consistent delivery of a specific amount of drug solution. The affected skin area of intradermal infusions is quite large. It ranges from 3.5 mm² to 15 mm² depending on the drug and its vehicle. This is quite impressive and demonstrates that a single microneedle might be sufficient. The back pressure during intradermal infusions and injections with the intradermal infusion set was studied by varying the fluid viscosity and the flow rates. This data is necessary for the design of a suitable delivery device. A back pressure of 112 kPa was associated as the maximum back pressure due to an initial backpressure overshoot for a flow rate of 0.5 ml/h. This back pressure level limits the number of potential interesting micro actuator technologies for small wearable devices. Average injection pressures can range up to 800 kPa. This is no obstacle for injection applications. Nevertheless, it

115 must be considered in device design. The intradermal infusion set was selected for the integration in the ChronopaDD device.

The hydrogel actuator is a simple actuator. It can be fabricated in very view steps. The necessary raw materials are cheap and currently produced in large volumes. The generation of volume work by the hydrogel actuator requires swelling agent. This is of minor importance if just several ml need to be displaced. Therefore it is an interesting actuator for small skin attachable drug delivery devices. The general performance of actuators with different design parameters was demonstrated in this work. The actuators show a high flow rate (up to 0.4 ml/min, 2 ml in 5 min.) at the beginning followed by a low flow rate ( 0.3 ml/h, 1.2 ml in 4 h) for several hours. This behavior is interesting for drug delivery devices in general. With the first fast phase it is possible to deliver an amount of drug large enough to increase the plasma concentration above the lower therapeutic limit. During the subsequent slow phase a small continuous flow of drug can help to keep the plasma concentration above this limit. Alternatively, it is possible to select an actuator with almost linear characteristics. The data on the back pressure sensitivity is encouraging. At back pressures of 50 kPa and 100 kPa the displaced volume after 4 h drops by approximately 20 % and 40 %, respectively. Based on these values the development of drug delivery devices with this actuator is warranted. Devices with catheters extending into cavities, e.g. the oral cavity, are definitely possible because of anticipated low back pressure. Devices with subcutaneous or intradermal catheters are more challenging because of the back pressure generated by infusion into the tissue. The hydrogel actuator was selected for the integration into the ChronopaDD device.

The ChronopaDD integrates the intradermal infusion set, the hydrogel actuator and an activation mechanism with time delay. The activation mechanism with time delay in combination with the hydrogel actuator is purely based on physico-chemical processes. So, the device is free of electronics and batteries. It can be disposed easily and it is supposed to be very cheap in production. Processes of the packaging industry were used as guide. Therefore, the setup of large scale production should be manageable. Sterile fabrication is mandatory for the drug pouch.

The time delay mechanism is currently based on a sucrose pill. For time delays of several hours it is required to work with expanding agents with high sucrose concentration (40 % by weight). Therefore, after dissolution of the pill the first droplet close to the hydrogel actuator is highly viscous with a very high sucrose concentration. This issue needs to be considered in the further development of the ChronopaDD. One approach is to find a different material for the delay pellet. This material should not reduce the chemical potential of the expanding agent and it should not form highly viscous plugs. Another approach is to concentrate on applications with very low back pressure (1.2 kPa).

The design of an electronic free extracorporeal chronopharmaceutical drug delivery device that can be manufactured in large scale is feasible. However, to get such a system on the market requires the identification of a disease that would benefit from time delayed extracorporeal drug delivery. Additionally, the market size must be really large. Otherwise, solutions based on the combination of already available equipment (e.g. insulin pumps) will be preferred. Actually, I’m convinced if the ChronopaDD ever hits the market it is the last action in a long process started by the combination of already existing technologies.

116 References

[1] Lemmer B. Chronopharmakologie: Tagesrhythmen und Arzneimittelwirkung. Wissenschaftliche Verlagsgesellschaft. Stuttgart. Germany. 2004 [2] Lévi F. Circadian chronotherapy for human cancers. Lancet Oncol. 2001;2:307-315 [3] Wilding I. Site-Specific Drug Delivery in the Gastrointestinal Tract. Crit Rev Ther Drug. 2000;17(6):557-620 [4] Degen LP, Phillips SF. Variability of gastrointestinal transit in healthy women and men. Gut. 1996;39:299-305 [5] Yu LX, Amidon GL. Characterisation of small intestine transit time distribution in humans. Int J Pharm. 1998;171:157-163 [6] Southwell BR, Clarke MCC, Sutcliffe J, Hutson JM. Colonic transit studies: normal values for adults and children with comparison of radiological and scintigraphic methods. Pediatr Surg Int. 2009;25:559-572 [7] Ohdo S. Chronopharmaceutics: Pharmaceutics Focused on Biological Rhythm. Biol Pharm Bull. 2010;33:159-167 [8] Baralod M. The influence of circadian rhythms on the kinetics of drugs in humans. Expert Opin Drug Metab Toxicol. 2008;4:175-195 [9] Rouge N, Buri P, Doelker E. Drug absorption sites in the gastrointestinal tract and dosage forms for site-specific delivery. Int J Pharm. 1996;136:117-139 [10] Davis SS, Hardy JG, Fara JW. Transit of pharmaceutical dosage forms through the small intestine. Gut. 1986a;27:886-892 [11] McConnell EL, Fadda HM, Basit AW. Gut instincts: Explorations in intestinal physiology and drug delivery. Int J Pharm. 2008;364:213-226 [12] Davis SS. Formulation strategies for absorption windows. Drug Discov Today. 2005;4:249-257 [13] Reuter P. Springer Lexikon Medizin. Springer. Berlin. 2004 [14] http://www.scf-online.com/english/27_e/frontpage27_e.htm [15] Maibach HI, Bronaugh RL. Percutaneous Absorption: Drugs, Cosmetics, Mechanisms, Methodology (Drugs and the Pharmaceutical Sciences). Marcel Dekker. New York. 1999 [16] www.wikipedia.com / www.wikipedia.de / 2011 [17] www.scf-online.de / 2011 [18] http://www.scf-online.com/english/40_e/bloodvessel40_e.htm [19] Barry BW. Novel mechanisms and devices to enable successful transdermal drug delivery. Eur J Pharm Sci. 2001;14:101-114 [20] Langer R. Transdermal drug delivery: past progress, current status, and future prospects. Adv Drug Deliver Rev. 2004;56:557-558 [21] Prausnitz MR, Mitragotri S, Langer R. Current status and future potential of transdermal drug delivery. Nat Rev Drug Discov. 2004;3:115-124 [22] Montgomery DC. Design and Analysis of Experiments. John Wiley & Sons. New York. USA. 2009 [23] Siebertz K, van Bebber D, Hochkirchen T. Statistische Versuchsplanung. Springer. Berlin. Deutschland. 2010

117 [24] Buchholz FL, Graham AT. Modern superabsorbent polymer technology. John Wiley & Sons. New York. USA. 1998 [25] Gerstel MS, Place VA. Drug Delivery Device. US 3,964,482 1971 [26] Matriano JA, Cormier M, Johnson J et al. Macroflux Microprojection Array Patch Technology: A New and Efficient Approach for Intracutaneous Immunization. Pharm. Res. 2002;19:63-70 [27] Cormier MJN, Nat AS, Neukermans AP et al. Device for enhancing transdermal agent delivery or sampling. US 6,230,051 2001 [28] Martanto W, Davis SP, Holiday NR et al. Transdermal Delivery of Insulin Using Microneedles in Vivo. Pharm Res. 2004;21:947-952 [29] Wermeling DP, Banks SL, Hudson DA et al. Microneedles permit transdermal delivery of a skin-impermeant medication to humans. P Nat. Acad. Sci. USA 2008;105:2058-2063 [30] Davis SP, Prausnitz MR, Allen MG. Fabrication and Characterization of Laser Micromachined Hollow Microneedles. Transducers. 2003:1435-1438. Boston USA [31] Kim K, Lee JB. High aspect ratio tapered hollow metallic microneedle arrays with microfluidic interconnector. Microsyst. Technol. 2007;13:231-235 [32] Zhu MW, Li HW, Chen XL et al. Silica needle template fabrication of metal hollow microneedle arrays. J. Micromech Microeng. 2009;19:1-6 [33] Yousef H, Lehto M, Jäderblom T et al. A Device Integrating Paraffin Microactuator, Fluidic Compartment and Microneedle Array for Fluid Injection or Sampling. 9th Int. Conf. on Min. Syst. for Chem. And Life Sci. 2005:157- 159. Boston USA [34] Hasada T, Shikida M, Sato K. Densely Arrayed Microneedles Having Flow Channels Fabricated by Mechanical Dicing and Anisotropic Wet Etching of Silicon. Transducers. 2005:1493-1496. Seoul, Korea [35] Verbaan FJ, Bal SM, van den Berg DJ et al. Assembled microneedle arrays enhance the transport of compounds varying over a large range of molecular weight across human dermatomed skin. J. Control. Release. 2007;117:238- 245 [36] Wilke N, Reed ML, Morrissey A. The evolution from convex corner undercut towards microneedle formation: theory and experimental verification. J. Micromech. Microeng. 2006;16:808-814 [37] Trautmann A, Ruther P, Paul O. Microneedle Arrays Fabricated Using Suspended Etch Mask Technology Combined with Fluidic Through Wafer Vias. MEMS. 2003:682-685. Kyoto. Japan [38] Trautmann A, Heuck F, Denfeld R et al. Detachable Silicon Microneedle Stamps for Allergy Skin Prick Testing. MEMS. 2006:434-437. Istanbul. Turkey [39] Griss P, Stemme G. Side-Opened Out-of-Plane Microneedles for Microfluidic Transdermal Liquid Transfer. J Microelectromech S. 2003;12:296-301 [40] Gardeniers HJGE, Luttge R, Berenschot EJW. Silicon Micromachined Hollow Microneedles for Transdermal Liquid Transport. J Microelectromech S. 2003;12:855-862 [41] Stoeber B, Liepmann D. Arrays of Hollow Out-of-Plane Microneedles for Drug Delivery. J Microelectromech S. 2005;14:472-479 [42] Mukerjee EV, Isseroff RR, Collins SD et al. Microneedle Array with Integrated Microchannels for Transdermal Sample Extraction and in situ Analysis. Transducer. 2003:1439-1441. Boston. USA

118 [43] Trautmann A, Heuck F, Mueller C et al. Replication of Microneedle Arrays Using Vacuum Casting and Hot Embossing. Transducers. 2005:1420-1423. Seoul. Korea [44] Han M, Hyun DH, ParkHH et al. A novel fabrication process for out-of-plane microneedle sheets of biocompatible polymer. J Micromech Microeng. 2007;17:1184-1191 [45] Park JH, Allen MG, Prausnitz MR. Biodegradable polymer microneedles: Fabrication, mechanics and transdermal drug delivery. J Control Release. 2005;104:51-66 [46] Park HJ, Allen MG, Prausnitz MR. Polymer Microneedles for Controlled- Release Drug Delivery. Pharm Res. 2006;23:1008-1019 [47] Park JH, Choi SO, Kamath R et al. Polymer particle-based micromolding to fabricate novel microstructures. Biomed Microdev. 2007;9:223-234 [48] Lippmann JM, Geiger EJ, Pisano AP. Polymer Investment Molding: Method for fabricating hollow, microscale parts. Sensor Actuat A Phys. 2006;134:2- 10 [49] Guber AE, Dittrich H, Heckele M et al. Polymer Micro Needles with Through- Going Capillaries. µTAS. 2001:155-156. Monterey. USA [50] Ovsianikov A, Chichkov B, Mente P et al. Two Photon Polymerization of Polymer-Ceramic Hybrid Materials for Transdermal Drug Delivery. Int J Appl Ceram Technol. 2007;4:22-29 [51] http://www.rsc.org/chemistryworld/News/2008/January/16010801.asp [52] Khumpuang S, Horade M, Fujioka K et al. Geometrical strengthening and tip- sharping of a microneedle array fabricated by X-ray lithography. Microsyst Technol. 2007;13:2009-214 [53] Paik SJ, Lim JM, Jung I. A Novel Microneedle Array Integrated with a PDMS Biochip for Microfluidic Systems. Transducers. 2003:1446-1449. Boston. USA [54] Liu R, Wang XH, Tang F et al. An In-Plane Microneedles Used for Sampling and Glucose Analysis. Transducers. 2005:1517-1520. Seoul. Korea [55] Jiang J, More JS, Edelhauser HF et al. Intrascleral Drug Delivery to the Eye Using Hollow Microneedles. Pharm. Res. 2009;26:395-403 [56] Shibata T, Nakanishi A, Sakai T et al. Fabrication and Mechanical Characterization of Microneedle Array for Cell Surgery. Transducers 2007;719-722. Lyon France [57] Fernández LJ, Altuna A, Tijero M, et al. Study of functionality viability of SU- 8-based microneedles for neural applications. J. Micromech. Microeng. 2009;19:1-8 [58] Wang T, Yang W, Huang H et al. A Novel Fabrication Method of Flexible Micro Electrode Array for Neural Recording. MEMS 2007;297-300. Kobe. Japan. [59] Mukerjee EV, Isseroff RR, Nuccitelli R. Microneedle Array for Measuring Wound Generated Electric Fields. EMBS. 2006:4326-4328. New York. USA [60] Bhandari R, Negi S, Rieth L et al. A novel masking method for high aspect ratio penetrating microelectrode arrays. J Micromech Microeng. 2009;19:1-8 [61] Haq MI, Smith E, John DN et al. Clinical administration of microneedles: skin puncture, pain and sensation. Biomed Microdevices, 2009;11:35-47 [62] Verbaan FJ, Bal SM, van den Berg DJ et al. Improved piercing of microneedle arrays in a dermatomed human skin by an impact insertion method. J Control Release. 2008;128:80-88

119 [63] Henry S, McAllister DV, Allen MG. Microfabricated Microneedles: A Novel Approach to Transdermal Drug Delivery. J. Pharm Sci. 1998;87:922-925 [64] Loeters PWH, Duwel RF, Verbaan FJ et al. Measuring the Insertion of Microfabricated Microneedles into Skin with a Penetration Sensor. 8th Int Conf Miniat Syst Chem Life Sci. 2004:497-499. Malmö, Sweden [65] Roxhed N, Griss P, Stemme G. Relaible In-Vivo Penetration and Transdermal Injection Using Ultra-Sharp Hollow Microneedles. Transducers. 2005:213-216. Seoul. Korea [66] McAllister DV, Wang PM, Davis SP et al. Microfabricated needles for transdermal delivery of macromolecules and nanoparticles: Fabrication methods and transport studies. P Nat Acad Sci USA. 2003;100:13755-13760 [67] Wang PM, Cornwell M, Hill J et al. Precise Microinjection into Skin Using Hollow Microneedles. J Invest Dermatol. 2006;126:1080-1087 [68] Kim YC, Quan FS, Yoo DG et al. Improved influenza vaccination in the skin using vaccine coated microneedles. Vaccine. 2009;27:6932-6938 [69] Häfeli UO, Mokhtari A, Liepmann D et al. In vivo evaluation of a microneedle- based miniature syringe for intradermal drug delivery. Biomed Microdev. 2009;11:943-950 [70] Martanto W, Moore JS, Kashlan O. Microinfusion Using Hollow Microneedles. Pharm Res. 2006;23:104-113 [71] Sivamani RK, Stoeber B, Wu GC et al. Clinical microneedle injection of methyl nicotinate: stratum corneum penetration. [72] Kenney RT, Frech SA, Muenz LR et al. Dose sparing with Intradermal Injection of Influenza Vaccine. New Engl J Med. 2004;351:2295-2301. [73] Laurent PE, Bonnet S, Alchas P et al. Evaluation of the clinical performance of a new intradermal vaccine administration technique and associated delivery system. Vaccine. 2007;25:8833-8842 [74] Holland D, Booy R, De Looze F et al. Intradermal Influenza Vaccination Administered Using a New Microinjection System Produces Superior Immunogenicity in Elderly Adults: A Randomized Controlled Trial. J Infect Dis. 2008;198:650-658 [75] Van Damme P, Oosterhuis-Kafeja F, Van der Wielen M et al. Safety and efficacy of a novel microneedle device for dose sparing intradermal influenza vaccination in healthy adults. Vaccine. 2009;27:454-459 [76] Nordquist L, Roxhed N, Griss P et al. Novel Microneedle Patches for Active Insulin Delivery are Efficient in Maintaining Glycaemic Control: An Initial Comparison with Subcutaneous Administration. Pharm Res. 2007;24:1381- 1388 [77] Gupta J, Felner EI, Prausnitz MR, Minimally Invasive Insulin Delivery in Subjects with Type 1 Diabetes Using Hollow Microneedles. Diabetes Technol Ther. 2009;11:329-337 [78] Pettis RJ, Hirsch LJ, Nosek L et al. Microneedle-based intradermal injection of lispro of human regular insulin accelerates insulin uptake and reduces post-prandial glycaemia. Diabetologia. 2009;52:S378 [79] Pettis RJ, Ginsberg B, Hirsch L, Sutter D, Keith S, McVey E, Harvey NG, Hompesch M, Nosek L, Kapitza C, Heinemann L. Intradermal Microneedle Delivery of Insulin Lispro Achieves Faster Insulin Absorption and Insulin Action than Subcutaneous Injection. Diabetes Technol Ther. 2011;13:435- 442 [80] Pettis RJ, Hirsch L, Kapitza C, Nosek L, Hövelmann U, Kurth HJ, Sutter DE, Harvey NG, Heinemann L. Microneedle-Based Intradermal Versus

120 Subcutaneous Administration of Regular Human Insulin or Insulin Lispro: Pharmacokinetics and Postprandial Glycemic Excursions in Patients with Type 1 Diabetes. Diabetes Technol Ther. 2011;13:443-450 [81] Harvey AJ, Kaestner SA, Sutter DE, Harvey NG, Mikszta JA, Pettis RJ. Microneedle-Based Intradermal Delivery Enables Rapid Lymphatic Uptake and Distribution of Protein Drugs. Pharm Res. 2011;28:107-116 [82] Burton SA, Ng CY, Simmers R, Moeckly C, Brandwein D, Gilbert T, Johnson N, Brown K, Alston T, Prochnow G, Siebenaler K, Hansen K. Rapid Intradermal Delivery of Liquid Formulations Using a Hollow Microstructured Array. Pharm Res. 2011;28:31-40 [83] Donnelly RF, Sing TRRS, Tunney MM et al. Microneedle Arrays Allow Lower Microbial Penetration Than Hypodermic Needles In Vitro. Pharm Res. 2009;26:2513-2522 [84] Wang PM, Cornwell M, Prausnitz MR et al. Minimally Invasive Extraction of Dermal Interstitial Fluid for Glucose Monitoring Using Microneedles. Diabetes Technol Ther. 2005;7:131-141 [85] Schäffler A. Delivery to Match Patient Needs. ipt. 2007;22:59-62 [86] http://www.lodotra.com/healthcare-press/lodotrar/what-is-unique-about- lodotra.html [87] Prausnitz MR, Langer R. Transderma Drug Delivery, Nat. Biotechn. 2008;26:1261-1268 [88] Morrow DIJ, McCarron PA, Woolfson AD, Donenelly RF. Innovative Strategies for Enhancing Topical and Transdermal Drug Delivery. Open Drug Deliv J. 2007;1:36-59 [89] Huang JZ, DiPierro G. Portable Drug Delivery Device Including a Detachable and Replaceable Administration or Dosing Element. US 2006 0 271 020 A1. 24.05.2005 [90] http://www.scf-online.com/english/37_e/skinpenetration37_e.htm [91] Daniels R. Strategies for Skin Penetration Enhancement. www.scf- online.com. 2004;37 [92] Cross SE, Roberts MS. Physical Enhancement of Transdermal Drug Applications: Is Delivery Technology Keeping up with Pharmaceutical Development? Curr Drug Del. 2004;1:81-92 [93] Escoffier C, de Rigal J, Rochefort A, Vasselet R, Lévêque JL, Agache PG. Age-Related Mechanical Properties of Human Skin: An In Vivo Study. J Invest Dermatol. 1989;93:353-357 [94] Norlén L, Nicander I, Rozell BL, Ollmar S and Forslind B. Inter- and Intra- Individual Differences in Human Stratum Corneum Lipid Content Related to Physical Parameters of Skin Barrier Function In Vivo. J Invest Dermatol. 1999;112:72-77 [95] Gross J. Liquid Material Dispenser. EP 0 481 601 B1. 19.10.1990 [96] Lynch PM, Butler J, Huerta D, Tsals I, Davidson D, Hamm S. A Pharmacokinetic and Tolerability Evaluation of Two Continuous Subcutaneous Infusion Systems Compared to an Oral Controlled-Release Morphine. J Pain Symptom Manage. 2000;19:348-356 [97] Meehan E, Gross Y, Davidson D, Martin M, Tsals I. A microinfusor device for the delivery of therapeutic levels of peptides and macromolecules. J Control Release. 1996;46:107-116 [98] Genosar A, Goldstein J, Fleisher N, Bachar Y, Aldar Y. Controllable Drug Delivery Device. US 7 918 843 B2. 23.04.2007

121 [99] Richter A, Klenke C, Arndt KF. Adjustable Low Dynamic Pumps Based on Hydrogels. Macromol. Symp. 2004;210:377-384 [100] Richter A, Klenke C, Arndt KF, Schiltges G. Automatic Conveyor Driven by Hydrogels, Provided with an Adjustable Output Characteristic for Conveying a Medium. US 7 479 135 B2. 13.01.2003 [101] Lemmer B, Chronobiology, drug delivery, and chronotherapeutics, Adv Drug Del Rev. 2007;59:825-827 [102] http://www.cancernetwork.com/palliative-and-supportive- care/content/article/10165/88598?pageNumber=2 [103] Srisathapat C, Indravudh V, Medication Infusion Pump with Improved Pressure Reservoir, US 5 514 103, 14.06.1994 [104] Rogers CR, Seifert GP, Implantable Infusion Device with Optimized Peristaltic Pump Motor Drive, US 7 122 026, 22.04.2002 [105] Arkels B, Larson G. Silicon Compounds Silanes & Silicones. 2008. Gelest. Morrisville. PA. USA. Company catalogue [106] van Goer M. Weiterentwicklung von Heißprägeverfahren zur Erzeugung von Mikronadeln mit Kavität aus thermoplastischem Polymer. 31.07.2008. Bericht Praktisches Studiensemester. Hochschule Furtwangen University. Betreuer: Müller C, Vosseler M [107] van Goer M. Herstellung eines Werkzeugs zum Prägen von Mikronadel- Arrays mit Kavität aus thermoplastischem Polymer. 30.03.2009. Diplomarbeit. Hochschule Furtwangen University. Betreuer: Müller C, Schön H, Vosseler M [108] Botzelmann T, Mayer V, Vosseler M. Etablierung von spritzgegossenen Mikronadel-Arrays als mikroinvasive transdermale Schnittstelle zum vaskulären System des Menschen. 31.07.2010. AiF Abschlussbericht. FV- Nr.: 15516 N [109] Henning A, Neumann D, Kostka KH, Lehr CM, Schaefer UF. Influence of human skin specimens consisting of different skin layers on the result of in vitro permeation experiments. Skin Pharmacol Physiol. 2008;21:81-88 [110] Saji M, Taguchi S, Uchiyama K, Osono E, Hayama N, Ohkuni H. Efficacy of gentian violet in the eradication of methicillin-resistant staphylococcus aureus from skin lesions. J Hosp Infect. 1995;31:225-228 [111] Bonnekoh B, Wevers A, Jugert F, Merk H, Mahrle G. Colorimetric growth assay for epidermal cell cultures by their crystal violet binding capacity. Arch Dermatol Res. 1989;281:487-490 [112] Sepasi K, Yalkowsky SH. Solubility prediction in octanol: a technical note. AAPS Pharm Sci Tech. 2006;7:E26 [113] Yamaoka T, Tabata Y, Ikada Y. Distribution and tissue uptake of poly(ethylene glycol) with different molecular weights after intravenous administration to mice. J Pharm Sci US. 1994;83:601-608 [114] Santhanalaksmi J, Balaji S. Binding studies of crystal violet on proteins. Colloid Surf A. 2001;186:173-177 [115] Nisar A, Afzulpurkar N, Mahaisavariya B, Tuantranont A. MEMS based micropumps in drug delivery and biomedical applications. Sens Act B Chem. 2008;130:917-942 [116] Iverson BD, Garimella SV. Recent advances in microscale pumping technologies: a review and evaluation. Microfluid Nanofluid. 2008;5:145-174 [117] Alabsi B. Versuchsaufbau zur Applikation von Mikronadeln. 30.08.2007. Bericht Praktisches Studiensemester. Hochschule Furtwangen University. Betreuer: Vosseler M

122 [118] Ilieva V. Vorgangsverzögerung durch Auflösungsprozesse. 26.02.2010. Bericht Praktisches Studiensemester. Hochschule Furtwangen University. Betreuer: Reule W, Vosseler M [119] Schmidt C. Charakterisierung der Gegendruckabhängigkeit eines Hydrogelaktors für ein Chronotherapeutisches Medikamentendosiersystem. 26.02.2010. Masterarbeit. Hochschule Furtwangen University. Betreuer: Haimerl G, Vosseler M. [120] Dean DA, Evans ER, Hall IH. Pharmaceutical Packaging Technology. Taylor & Francis. London. UK. 2000 [121] Dehnert J. Piezoresisitve high-g Accelerometers. 2010. HSG-IMIT. Villingen- Schwenningen. Germany. Application Report [122] Kain A. Innovative Prozesse für das Heißprägen von Mikronadeln. 26.02.2010. Diplomarbeit. Hochschule Furtwangen University. Betreuer: Busolt U, Müller C, Vosseler M

123

124 Abbreviations

AD analog to digital

ANOVA analysis of variance

DI deionised

DOE design of experiments

EDM electrical discharge machining

FEP fluorinated ethylene propylene

G235 gain at 4 h, see chapter 6.1.3

GI gastro intestinal

PC polycarbonate

PE polethylne

PEG polyethylene glycol

PDMS polydimethylsiloxane

PMMA polymethyl methacrylate

PP polypropylene

SAP superabsorbent polymer

TPE thermoplastic elastomer

V5 displaced volume after 5 min, see chapter 6.1.3

125 Appendix A – Model diagnostics for chapter 6.1

Figure A.1. Residuals of the full model of the variable displaced mass after 5 min. Top left: Residuals versus fitted values. Top right: Residuals versus run number. Bottom: Normal plot of residuals.

Figure A.2. Residuals of the full model of the variable displaced mass after 5 min. versus factor levels.

126

Figure A.3. Residuals of the full model of the transformed data gain by 4 h. Top left: Residuals versus fitted values. Top right: Residuals versus run number. Bottom: Normal plot of residuals.

Figure A.4. Residuals of the full model of the transformed data gain by 4 h versus factor levels.

127 Appendix B – Temperature of the hot embossing tool

Figure B.1. Temperature of the hot plates of the hot embossing machine and temperature of the embossing tool vs. time. The embossing tool temperature lags behind during the transient heating of the embossing setup. Within less than 20 min. the difference in temperatures is balanced19.

Figure B.2. Temperature of the hot plates of the hot embossing machine and temperature of the embossing tool vs. time. The embossing tool temperature lags behind during the transient cooling of the embossing setup19.

19 Experimental data acquired by van Goer M.

128 Appendix C – Applicator

The dynamics of the spring powered applicator can be calculated based on Hooke’s law, Newton’s second law and the fundamental equations of motion. The acceleration as a function of the position of the piston is calculated according to Newton’s second law and Hooke’s law by:

(C.1) a(s) is the acceleration of the piston as a function of its position s (see Figure 5.16). s0 is the initial position of the piston. m is the mass of the piston (20.6 g), the mass of the spring is neglected. g is the acceleration of gravity. k is the spring rate (1.9 N/mm). The equation is valid for s0 ≤ s ≤ 0. If s is larger than 0 the piston is only accelerated by g. Friction is also neglected by this equation.

Starting from the fundamental equations of motion ( one can derive the velocity as a function of position:

(C.2)

v(s) is the velocity as a function of position s. v0 is the velocity at time t=0. a(s) is the acceleration as a function of position s.

Applying equation (C.1) to equation (C.2) the velocity as a function of position s is:

(C.3)

The meaning of the symbols is not changed. It is assumed v0=0 m/s and s0 ≤ s ≤ 0.

With the duration of the motion can be calculated according to:

(C.4)

t(s) is the time elapsed if the piston moves from the position s0 to s. This integral is solved numerically.

The spring powered applicator was characterised with an acceleration sensor (developed and manufactured at HSG IMIT [121]) mounted on the piston. The measurement range of this acceleration sensor is - 2000 g to 2000 g. The mass of the piston

129 and the sensor is 20.6 g. However, due to the cable to the sensor which is also partly moved during the experiments the effective mass is unknown.

The acceleration sensor was connected to a an amplifier (7A22, Tektronix Inc., Bea- verton, Oregon, USA) and the data was acquired with an analog to digital converter (NI 92, National Instruments, Austin, TX, USA) at 100 kHz. The sensitivity (0.37 mV/g) and offset (26.83 mV) of the amplified sensor signal was determined on a shaker (TV 51140-C, Tira GmbH, Schalkau, Germany) with a reference sensor (7253C-10, San Juan Capistrano, CA, USA).

The spring powered applicator was characterised with spring deflexions of 7.5 mm, 13.5 mm and 17.5 mm. The height of the bushing above a sheet of rubber was adjusted in such a way that the piston can move a distance as far as the spring deflexion, respectively. 3 experiments were performed at each deflexion. The data of a single experiment with a spring deflexion of 13.5 mm is presented in Figure C.1. In real- ity the acceleration does not follow an initial step as the theory suggests. The signal shows a first slow increase (-1 ms to 0 ms) of the acceleration followed by a steep but still limited increase (0 ms to 1 ms). The slope of the steep increase of the signal (0 ms to 1 ms) was determined. Its intercept point with the x-axis was defined as the beginning of the motion. The maximum acceleration is lower than the theoretical maximum. The duration of the motion is longer than predicted. The velocity of the integrated data (Figure C.1, right) is also below the predicted value. Additionally, noise and other disturbances are added to the signal.

The very slow increase of the signal at its beginning (-1 ms to 0 ms) is most probably due to the manual release process of the piston. Due to friction forces the steep increase of the acceleration is limited. Additionally, the maximum acceleration is smaller and the duration of the motion is longer than expected. Consequently, the lower velocity is explained by friction.

Figure C.1. Left: Acceleration as a function of time. Data of an acquired experiment with a spring deflection of 13.5 mm is presented together with the theoretical calculated graph. Right: The data of the same experiment was integrated and plotted together with the theoretical graph.

130 The data obtained in the experiments is compiled in Figure C.2. Acceleration, velocity, duration and distance are given for three spring deflexions. The maximum acceleration values are 17 % ± 10 % lower than expected. The maximum velocity values are 21 % ± 2 % smaller than expected. The duration of the motion values are 29 % ± 3 % higher than expected. The maximum distance values are 15 % ± 3 % smaller than expected. Based on this data it is concluded, that theoretically predicted characteristics and experimental data are in good agreement within an error range of ± 20 %.

Figure C.2. Results of the experiments to characterise the spring powered applicator. The error bars represent the standard error with n=3. Top left: maximum acceleration as a function of spring deflexion. Top right: velocity as a function of spring deflexion. Bottom left: duration as a function of spring deflexion. Bottom right: distance as a function of spring deflexion.

131 Danksagung

Herrn Prof. Dr. Roland Zengerle danke ich für die Möglichkeit, die Ergebnisse meiner Arbeiten am HSG-IMIT in einer Dokotorarbeit veröffentlichen zu können. Für die Be- treuung der Arbeit bin ich ebenfalls dankbar. Ein weiterer Dank gilt Herrn Prof. Dr. Holger Reinecke, dem Zweitgutachter dieser Arbeit. Der Baden-Württemberg Stiftung gGmbH und der AiF e.V. danke ich für die Finanzie- rung der Projekte ChronopaDD, dermalPort und Microneedles. Herrn Prof. Dr. Claas Müller danke ich für die Zusammenarbeit hinsichtlich der Ferti- gung von hohlen Mikronadeln im Heißprägeverfahren. Herrn Prof. Dr. Heinz Kück und Herrn Dr. Volker Mayer danke ich für die gemeinsame Leitung und Koordination des AiF-Projektes Microneedles. Herrn Prof. Dr. Martin Schmelz danke ich für die gemeinsame Leitung und Koordina- tion des Projektes dermalPort. Den Herren Prof. Dr. Stephan Messner, Prof. Dr. David Hradetzky und Sven Spieth danke ich für die Organisation des Bereichs Mikrofluidik und der Produktgruppe MikroMedizin. Dankbar bin ich allen Mitarbeitern des HSG-IMIT und des HSG-IMAT die im Rahmen der Projekte ChronopaDD, Microneedles und dermalPort mitgearbeitet haben. Ein weiterer Dank gilt allen Studenten, die im Rahmen ihrer Abschlussarbeiten, Prak- tika und Hiwi-Tätigkeiten einen Beitrag zu dieser Arbeit geleistet haben. Bei den Firmen Across Barriers GmbH, B. Braun Melsungen AG, ECMTEC GmbH, Hochschule Furtwangen (Prof. Dr. Hartmut Schiefer), Horst Scholz GmbH, Novineon Healthcare Technology Partners GmbH, Optimags GmbH, RKT Rodinger Kunststoff- Technik GmbH, Schreuers-Tools GmbH, UPT Optik Wodak GmbH bedanke ich mich für die Mitarbeit im projektbegleitenden Ausschuss des Projektes Microneedles. Dankbar bin ich auch allen Firmen, die kostenfrei Material zur Verfügung gestellt haben. Dazu gehören unter anderem die Firmen Evonik Stockhausen GmbH und Orbita Film GmbH. Den Tierärzten Dr. Andrej Barke und Dr. Friedhelm Erschig danke ich für die Bereit- stellung von Schweineohren. Herrn Prof. Dr. Heinz Gründemann danke ich für die Durchsicht des Manuskriptes.

132