Faculty of Pharmacy University of Helsinki

Finland

Granulation in Miniaturised Fluid Bed Using

Electrostatic Atomisation

Niina Kivikero

ACADEMIC DISSERTATION

To be presented, with the permission of the Faculty of Pharmacy of the University of Helsinki, for public examination in lecture room 1015, Biokeskus 2 (Viikinkaari 5),

on 29th May 2010, at 12 noon.

Helsinki 2010

Supervisors Professor Anne Juppo Industrial Pharmacy Faculty of Pharmacy University of Helsinki Finland

Docent Jukka-Pekka Mannermaa

Oy Verman Ab Kerava Finland Professor Niklas Sandler Laboratory of Pharmaceutical Sciences

Åbo Akademi University Finland

Docent Eetu-Pekka Räsänen South-Karelian Hospital Pharmacy

Lappeenranta Finland Reviewers Professor Peter Kleinebudde

Heinrich-Heine University Düsseldorf Germany Docent Pirjo Tajarobi AstraZeneca R&D Mölndal Sweden

Opponent Professor Jonathan Seville Warwick University UK

ISBN 978-952-10-6249-0 (paperback) ISBN 978-952-10-6250-6 (PDF, http://ethesis.helsinki.fi) ISSN 1795-7079 Helsinki University Printing House Helsinki 2010

Abstract

Kivikero N., 2010. Granulation in Miniaturised Fluid Bed Using Electrostatic Atomisation.

Dissertationes bioscientiarum molecularium Universitatis Helsingiensis in Viikki, 10/2010 pp. 46.

ISBN 978-952-10-6249-0 (paperback), ISBN 978-952-10-6250-6 (PDF), ISSN 1795-7079

The development of a new drug is extremely expensive and the development process is very slow, up to 15 years. Especially the early formulation development phase is a challenge for the pharmaceutical industry, as the amount of a new active pharmaceutical ingredient may only be a few grams. A small amount of drug should be used to produce as much data as rapidly possible.

In this thesis, a small scale fluid bed device (Multipart Microscale Fluid bed powder Processor, MMFP) with electrostatic atomisation is used for the first time to perform granulations. The aims of this thesis were to develop and characterise a suitable spraying method for MMFP and to characterise the spray. The process parameters of the electrostatic atomisation system and fluid bed granulation affecting the granule size were studied. Also the applicability of the set-up to early formulation studies was evaluated.

Electrostatic atomisation was found to be applicable for spray production in a small device with a specially constructed nozzle. With particle tracking velocimetry, it was possible to generate droplet size distributions of the produced spray. Also a high speed imaging system provided information about the spray. The granulation liquid flow rate affected the granule size the most, although the atomisation voltage and binder concentration of the granulation liquid had also an impact on it.

MMFP provides possibilities to study granulation properties of different materials with a small sample size. Also early formulation screening studies are possible to be conducted. The granulation process, as well as the following analysis is fast, although a small sample size may be challenging for traditional methods.

Acknowledgements

This study was conducted in the Industrial Pharmacy and Division of Pharmaceutical Technology, Faculty of Pharmacy in University of Helsinki.

I am most grateful to all my supervisors. Professor Anne Juppo is thanked for all her support and encouragement especially in challenging times. Docent Jukka-Pekka Mannermaa is thanked for introducing me to the fascinating world of pharmaceutical technology and believing in me since my Master’s thesis, Professor Niklas Sandler is thanked for the fruitful conversations and new ideas at the moments of despair and Docent Eetu Räsänen for being my true mentor in the field of miniaturised fluid beds. I also would like to thank Professor Jouko Yliruusi for giving me the opportunity to begin this work in the Division of Pharmaceutical Technology.

Professor Peter Kleinebudde and Docent Pirjo Tajarobi are deeply acknowledged for reviewing this thesis and for giving the constructive comments for it in very tight schedule.

I want to thank everyone who has ever been involved in this work (you are many!), especially my co-authors. Docent Matti Murtomaa is deeply acknowledged for the original idea of combining the electrostatic atomisation and fluid bed granulation and for building up the nozzle. Dr Osmo Antikainen is thanked for his valuable expertise in the field of modelling and tabletting. M. Sc. (Pharm.) Henrik Ehlers is thanked for encouraging enthusiasm and help with the droplet size measurements. Juha Hatara is thanked for his marvellous skills in image processing. Also technical staff is acknowledged for their help in the laboratory work.

I want to express my warm thanks to all my former and present colleagues at the Industrial Pharmacy and Division of Pharmaceutical Technology, this work would not have been possible without all of you! Especially Dr Marja Savolainen is thanked for her friendship, help and encouragement. Dr Satu Lakio is thanked for her friendship at work and in implementing the concept of “field pharmacy” into Finnish scouting. I would also like to thank all my room mates who have shared all my moments of happiness and despair. I am grateful to pharmacy owners Seija Mannermaa and Eeva-Kaisa Mäki-Moijala for giving me the opportunity to keep up my practical skills at pharmacy during these years.

Last, I would like to express my gratitude to my beloved family and friends. My dear mother Ritva is thanked for her love, understanding and support during the almost endless years of my education. All my friends are deeply thanked for reminding me of the world outside science. Päivi, Salla and Anna, you have been my anchors to the everyday life and my shoulders in the agony of balancing between work and family life.

Finally, my warmest and most sincere thanks are addressed to my husband Olli for his never ending love during all these years (especially the last one) and to my children Panu and Maisa who have brought so much joy and happiness to my life.

Helsinki, May 2010 Niina Kivikero

Table of contents

Abstract Acknowledgements Contents List of original publications Abbreviations and Greek symbols 1. Introduction 1 2. Review of the literature 2 2.1 Granulation 2 2.2 Theory of granule formation in general 3

2.2.1 Wetting and nucleation 4 2.2.2 Granule growth and consolidation 4 2.2.3 Granule attrition and breakage 6

2.3 Granule formation in fluid bed 7 2.4 Granule characterization 8 2.5 Fluidised bed granulation 9

2.5.1 Equipment parameters 10 2.5.2 Process parameters 11 2.5.3 Droplet size and droplet size distribution 12 2.5.4 Raw material parameters 12

2.6 Electrostatic atomisation 13 2.6.1 Definitions and theory of electrostatic atomisation 13 2.6.2 Pharmaceutical applications of electrostatic atomisation 14 2.6.3 Other applications of electrostatic atomisation 14

2.7 Miniaturisation 15 2.8 Modelling and multivariate analysis 16

2.8.1 Experimental design 16 2.8.2 Principal component analysis (PCA) 17 2.8.3 Partial least squares (PLS) regression analysis 17

3. Aims of the study 19

4. Experimental 20 4.1 Materials 20 4.2 Multipart Microscale Fluid bed powder Processor 20 4.3 Granulation in Multipart Microscale Fluid bed powder Processor 21 4.4 Methods for droplet and particle size analysis 23

4.4.1 Droplet size measurement using laser diffractometry (I) 23 4.4.2 Droplet size measurement with particle tracking velocimetry (IV) 23 4.4.3 High-speed imaging (IV) 24 4.4.4 Granule size measurement with sieve analysis (II) 25 4.4.5 Granule size measurement with image analysis (II) 25 4.4.6 Granule size measurement with spatial filtering technique (III) 25

4.5 Granule characterization 26 4.5.1 Friability (III) 26 4.5.2 Compactibility (III) 26

4.6 Multivariate analysis and modelling (II, III, IV) 27 5. Results and discussion 28 5.1 Electrostatic atomisation 28

5.1.1 Droplet formation (I) 28 5.1.2 Nozzle performance (IV) 29 5.1.3 Droplet size and droplet size distribution (IV) 31

5.2 Granule properties 33

5.2.1 Granule size and size distribution (II,III) 33 5.2.2 Granule size by multivariate visualisation and principal component analysis (II) 33

5.3 Early formulation and modelling (III) 35

6. Conclusions 38 References 39

List of original publications

This thesis is based on the following publications, which are referred to in the text by their roman numerals: I Murtomaa M, Kivikero N, Mannermaa J-P, Lehto V-P., 2005

Electrostatic atomization in microscale granulation. Journal of Electrostatics 63 (6), 891-897.

II Kivikero N, Murtomaa M, Ingelbeen B, Antikainen O, Räsänen E,

Mannermaa J-P, Juppo A-M., 2009 Microscale granulation in a fluid bed powder processor using electrostatic atomisation. European Journal of Pharmaceutics and Biopharmaceutics 71 (1), 130-137.

III Kivikero N, Murtomaa M, Antikainen O, Hatara J, Juppo A-M, Sandler

N., 2009. Rapid formulation screening with a Multipart Microscale Fluid bed Powder Processor. Accepted in: Pharmaceutical Development and Technology

IV Kivikero N, Ehlers H, Larjo J, Murtomaa M, Hatara J, Antikainen O,

Juppo A-M., 2009. Electrostatic atomization in miniaturized fluid bed granulation: Droplet size measurements. Submitted

Abbreviations

API Active pharmaceutical ingredient ASA Acetylsalicylic acid BF Breaking force FBG Fluid bed granulation CD Drag constant d10/50/90 Particle diameter at which 10% / 50% / 90% of the particles have diameters that are smaller than the d10/50/90 value DSC Differential scanning calorimetry DoE Design of Experiments Fc Collecting electrostatic force Fd Air drag force Fg Gravitational force Fi Electrostatic force g Gravitational constant HV High voltage LMH Lactose monohydrate MCC Microcrystalline cellulose MLR Multiple linear regression MMFP Multipart Microscale Fluid bed powder Processor MVA Multivariate analysis NIR Near infra red NME New molecular entity PC Principal component PCA Principal component analysis PLS Partial least squares PTFE polytetrafluoroethylene PVP Polyvinylpyrrolidone r droplet radius R&D Research and development SFT Spatial filtering technique v Relative velocity between droplet and air w/w Weight per weight

Greek symbols

γ Liquid surface tension ε Liquid permittivity ε0 Permittivity of free space η Viscosity σ Surface tension ρ Electrical resistivity ρg density of the air

________________________________________________________________________

1

1 Introduction

It has been stated, that the average cost to develop new drugs ranges from 802 million to 1.7 billion USD [1, 2]. Only R&D costs have been estimated to be 802 million USD, even if a new molecular entity (NME) was discovered and developed in-house. The process timeline is crucial, as it takes approximately 12 to 15 years to discover and develop a new medicine. Variation in development time is high, from basic research to application for clinical trials from 3 months to 30 years [1].

During the early development phase of a new drug product, the lack of new active pharmaceutical ingredient (API) causes serious problems. Drug substance performance as well as possible critical process parameters should be defined with the smallest amount of material as possible. Small scale manufacturing systems allow the production of multiple smaller batches and thus, possibilities to examine more process variables with the same amount of API [3].

Granulation is a routine process in pharmaceutical manufacturing in order to improve processing properties such as flowability and to minimise handling problems of powder masses as well as segregation [4]. More than 70% of the pharmaceutical industry’s granulations are performed by using wet granulation [5]. Fluid bed granulation is the most widely used technique, as it provides several advantages such as short process times as the whole process is conducted using the same device.

Electrostatic atomisation is a method, which can be utilised to produce droplets with a narrow size distribution [6]. The spraying process does not need any pressurised air and therefore it is a suitable method for small scale fluid bed granulation process.

The aim of this study was not to make a true down-scaling study, since the liquid spraying method and the dimensions of the used device are unique. In this thesis, a small scale fluid bed, Multipart Microscale Fluid bed powder Processor is used for the first time to perform granulation studies.

________________________________________________________________________

2

2 Review of the literature

2.1 Granulation

Granulation is defined as a size enlargement process, where particles are gathered to form larger permanent aggregates, where original smaller particles are still detectable. Besides in the pharmaceutical industry, granules are produced also in e.g. mineral processing, the manufacture of agricultural chemicals and food [7]. Although the same theories apply to granulations in every field, the challenges encountered are different, for instance the production rate may vary from kg/day to tons/hour [8]. Typically the size range for pharmaceutical granules is from 100 μm to 2 mm [9]. Granules are seldom used as a dosage form, but there are some products in the market.

Granules are common intermediates in capsule or tablet processing, because most pharmaceutical powders do not possess optimal tabletting behaviour naturally. Powders are granulated before tabletting to improve the flowability and compression properties. Many pharmaceutical powders flow poorly, because particles are small in size and they may possess irregular shape and unoptimal surface characteristics.

Granulation also prevents the segregation of different components, which is due to their differences in particle size or density. In an ideal granule, all components are mixed in right proportions, as the segregation in the granular level may affect the weight or the content uniformity of the final tablet. With granulation process it is possible to densify the material, which may prevent caking of hygroscopic materials during storage. Denser material also requires less volume in storage. Also, it is possible to promote the dissolution of active pharmaceutical ingredient (API) by controlling the porosity of the granule structure. Granulation improves safety in work as well as environmental safety, as dust formation is decreased with granulation [10]. Also the risk of contamination in production is decreased.

Granulation can be divided into dry and wet methods. Continous and thus, cost-effective dry granulation methods, such as roller compaction are used especially, when the API is sensitive to heat or moisture. Wet granulation methods, such as fluid bed granulation and high shear granulation utilise liquid to bind primary particles together to produce porous granules for various applications. The granulation process can be either very simple or very complex depending on material characteristics, the target quality of the final granules and the equipment available.

Granulation process research is important for several reasons. Process improvement in production scale ensures the quality of granules. Also, with good process control it is possible to minimise time consuming size classification processes, such as sieving [11]. The high potency of present and future drug molecules requires process research too. Therapeutic dose in the drug products i.e. dose in tablet or capsule is maybe relatively low, e.g. 0,25 μg [2]. These low dose formulations are challenges for formulation and manufacturing, as the granules needed for producing

________________________________________________________________________

3

low-dose tablets are mainly excipients (drug concentration may be as low as 0.001% w/w) and achieving the content uniformity is difficult.

2.2 Theory of granule formation in general

In the first systematic granulation studies in the 1950’s [12, 13] it was assumed, that granules grow by coalescence when limiting moisture content has been reached and after that mechanical agitation modifies granule shape. Modern theories of granulation [7, 14, 15] state, that there are three or four (depending on the source and the grouping of mechanisms) key mechanisms in the granulation process: powder wetting and nucleation (1), granule coalescence or growth (2), granule consolidation (3) and granule attrition or breakage (4). Granulation is a complex combination of these subsequent processes and also referred to be more art than science [11]. Several authors have also contributed in modelling granule growth [16], in which population balance equations have been the most popular approach [17].

Figure 1 Key mechanisms of granulation, modified from ref. [18].

________________________________________________________________________

4

2.2.1 Wetting and nucleation

Wetting of the powder is influenced by granulation liquid distribution [19], properties of the powder [20] as well as liquid [21] and powder bed flow [22]. The feeding rate and the extent of granulation liquid are important to control, as bed overwetting leads to broad nuclei (i.e. seed granules) distribution and poor wetting of the bed to droplet coalescence and hence, fewer and larger nuclei with ungranulated powder [18]. Recently, after recognition of the effect that nucleation affects the final granule properties, population balance models have been applied to predict the nuclei size distribution based on the properties of binder liquid spray, liquid feeding method and powder bed flow properties [23].

There are several methods to measure powder wetting such as contact angle measurement which explains how the drop spreads on the powder surface. Wetting of a powder surface depends on its physical nature e.g. particle size, porosity, pore size and roughness as well as the properties of the liquid, such as viscosity and density [18].

The nuclei formation is dependent on powder wetting. A direct relationship between the droplet size and the granule size is found in the beginning of granulation [24]. As the wetting process proceeds, the granulation liquid first penetrates into the pores of the particle and forms a nucleus. Then the liquid migrates outwards as the nucleus grows [7]. The nucleation mechanism depends on the relative sizes of the droplets and particles [25]. Large droplets lead to immersion of smaller particles into droplets, leading to nuclei with saturated pores. Nuclei with air trapped inside may be produced, when small droplets do not cover the bigger particle.

2.2.2 Granule growth and consolidation

There are two major mechanisms for granule growth, coalescence and layering. In coalescence two particles collide, stick together and form a new, larger particle. In layering, smaller particles are attached to the surface of previously formed nuclei. Granules also consolidate during the granulation process due to collisions with granulator chamber walls and other granules. During consolidation, the granules become less porous and smaller [18].

The amount of liquid is crucial for granule growth. Until permanent bonds between particles are formed, the liquid binds the particles together by capillary pressure, surface tension and viscous forces [7]. There also has to be enough liquid in the granules to ensure the coalescence of the particles. The five different bonding mechanisms that hold the primary particles together are presented in Table 1.

________________________________________________________________________

5

Table 1. The bonding mechanisms in the granules

Mechanism Forces Additional information Immobile liquid films Adhesional and cohesional

forces Between primary particles

Mobile liquid films Interfacial forces, capillary pressure

Strong bonds, prerequisite for solid bridges

Solid bridges Hardening of binders, partial melting or recrystallisation of materials

Main mechanism in wet granulation

Attractive forces Van der Waals forces, electrostatic forces

Does not necessarily need any liquid

Mechanical interlocking Often fibrous or flat particles

The states of granulation liquid saturation representing the stages of water distribution in a bed of solid particles are pendular, funicular and capillary state [13] (Fig. 2). The fourth state is the droplet state, where the liquid completely surrounds the particles. The end-point for the wet granulation depends on the granulation method. In conventional wet granulation, where the water is poured on the material the end point lies near the capillary state [26]. In modern high-shear granulation, the aim is at the funicular state, whereas in extrusion-spheronisation more liquid is needed.

Figure 2 The different states of liquid saturation [13]

Granules consolidate, as they collide in the granulator. Consolidation reduces the granule size and porosity (densification), as the air trapped inside the granule comes out. [7]. If the collisions are intense enough, the granulation liquid may be squeezed out of the granules and trigger granule coalescence. With brittle granules, collisions may also lead to attrition [8].

________________________________________________________________________

6

2.2.3 Granule attrition and breakage

Granule breakage during granulation includes the breakage of wet granules and attrition or fracture of dried granules. Wet granule breakage affects binder distribution and thus, the final granule size. Attrition generates dust or fines to the process, which is usually unwanted [7, 8]. At the process scale, breakage is important in enhancing the material distribution and eventual strength of the product granules.

Granule breakage may occur by erosion or abrasion (surface breakage) or by fracture or shattering (particle breakage). The mechanism is influenced by material properties, such as elastic modulus, hardness and particle shape [8]. Granule breakage can be modified in many ways as presented in Figure 3. Also other granule properties can be modified to affect breakage, e.g. granule hardness and mass can be altered to reduce wear [18].

Figure 3 Examples of granule breakage control [18].

________________________________________________________________________

7

2.3 Granule formation in fluid bed

In the beginning of the spraying phase of the fluid bed granulation (FBG), the primary particles form nuclei and are held together by liquid bridges (pendular state, Fig. 2.). Nuclei are loose and very porous. The size of the nuclei is dependent on droplet size, as stated by various authors [27-29]. As the liquid addition continues, more and more nuclei agglomerate and continue the transition from the pendular state to the capillary state [30]. In FBG, liquid saturation in the granules is increased only by adding granulation liquid to the process and thus, more granulation liquid is needed than in high shear granulation where the liquid saturation is increased by densification of granules. Granule growth in the fluid bed granulator is presented in Figure 4.

Liquid redistribution in the agglomerate is rate-determining and crucial for granule growth in fluidised bed [28, 31] as the granule mean diameter and operational moisture content during granulation have linear and positive correlation. The capillary pressure difference between the granule surface and partially saturated pores inside the granule induces growth. When wetting saturation is reached, there is not sufficient amount of binder left at the surface of the granule to bind any new particles. In general, the fluid bed process requires less wetting than high shear granulation. The binding between primary particles is mainly dependent on binder, physicochemical properties of starting materials and process conditions [30].

Fluid beds are more sensitive to the wetting stage of the granulation, as there is no mechanical redispersion of the liquid after initial spraying. Low shear forces are not able to squeeze the granulation liquid out of the agglomerates, therefore the initial spraying determines the final granule size [32]. Collapse of the bed as well as overwetting may occur if surface mixing is not adequate [24]. Droplet size, humidity of the bed and contact between the droplet and the powder are important factors [32]. Generally, granules produced in the fluid bed are porous and have loose structure. Granules also disperse well in water and have lower bulk densities compared to granules prepared by high shear granulation [10].

________________________________________________________________________

8

Figure 4 The granule growth in the fluid bed granulator. Modified from ref. [30].

2.4 Granule characterisation

In large scale tablet production tablet mass properties have to be excellent in order to maintain a high quality of tablets. Size distribution, flowability and homogeneity of mass are crucial properties of powder mass for the tabletting process, but also bulk density, porosity, residual moisture content are considered to be important for the optimal tabletting process [5]. Drug content uniformity in granules is not often determined, but in the case of low drug content and hydrophobic active pharmaceutical ingredient (API) it would be an important factor to control [33].

________________________________________________________________________

9

Also voidage i.e. the empty space between granules is an important factor. Bulk density of granules and granule attrition decrease, when granule voidage increases [18]. Porosity controls volume reduction in tablet compression, increase in porosity increases fragmentation and hence, results in tablets with high crushing strength [34]. On the other hand, porous and weak granules are challenging to handle and store.

There are several shape factors to characterise granules, such as aspect ratio and circularity, although shape and surface properties of the granules are not easy to measure. Shape factors are often derived from a microscopic image of a particle. Sometimes, target of the granulation process may be to produce spherical granules with a smooth surface. That is obtained mainly by selection of excipients, e.g. with microcrystalline cellulose (MCC) it is possible to produce such granules easily [35].

2.5 Fluidised bed granulation

Fluidised bed granulation (FBG) is one of the most important and frequently used unit operations in pharmaceutical manufacturing [30]. Wurster introduced the process and performed the first pharmaceutical fluid bed experiments [36] in the late 1950’s. Since then, numerous researchers have contributed to the field, and especially the effect of different process parameters to the final granule properties have been studied extensively [31, 37-39] Typical fluid bed granulator (top spray) set-up is presented in Figure 5. The advantages and disadvantages of FBG are listed in Table 2.

Figure 5 Typical fluidised bed granulator set-up

________________________________________________________________________

10

In FBG, particles are fluidised in a conical granulation chamber by air created by a turbine fan suction. The incoming air, which is filtered and often heated by an air handling unit, passes through a perforated plate at the bottom of the vessel. Process air is then removed through filter bags, which separates the product and fine particles from the air. The binder is typically added through a pneumatic nozzle, where liquid is atomised by pressurized air [40]. When granules are formed, they are dried to the target moisture content defined by the subsequent step in the manufacturing process.

The fluid bed granulation process is often divided into two different process stages; the binder addition phase and the drying phase. Before the binder addition phase, also a pre-mixing phase can be performed to ensure adequate mixing of materials. In FBG a large area of solid surface per unit particle mass is exposed to gas. The smaller the particles are, the larger the specific surface area of the particles. This large specific surface area facilitates solid-to-gas heat and mass transfer, and is a unique advantage of FBG [41].

Critical variables of the process are divided into apparatus, process and product parameters, which all affect the product. Often, granulation related problems in manufacturing require changes to more than one parameter subgroup to overcome, as the parameters are interrelated and influence each other [18].

Table 2. The advantages and disadvantages of fluid bed granulation (ref. [42-44])

Advantages Disadvantages

Short processing time Expensive

Minimum material handling and loss Laborious process optimization

Efficient heat and mass transfer Demanding down scaling and scale up

Saves in personnel costs Long and product-specific development

Complies with GMP (closed system)

Easy automation

Good reproducibility

2.5.1 Equipment parameters

Equipment parameters affecting FBG are the nozzle height, type and location [42, 45, 46] the design of the air distribution plate [47] and mechanical agitators [16] that sometimes are placed inside the granulation chamber.

Nozzles used in the FBG are usually twin-fluid devices with external mixing. Also nozzles with internal mixing may be used, although the risk for nozzle blockage is then increased [48] . Usually, nozzles are located above the powder bed, but it can also be placed in the bed, or sideways. If the nozzle is located too high according to the powder bed, the wetting of the powder may be inadequate [49] and the risk of

________________________________________________________________________

11

spray drying of the droplets and wall wetting increases [50]. Decreasing the nozzle position has been reported to lead to an increase in granule mean size and decrease in friability due to enhanced powder wetting [49], although there are also studies that consider nozzle position insignificant factor [51].

With air distribution plate, the inlet process air is distributed to the powder bed. Proper airflow is crucial for the process, as inadequate air flow causes uneven fluidisation. To ensure air flow, the pressure difference over the air distribution plate should be at least one third of pressure drop across the bed. Air distribution plates are made of steel and the openings in the plate vary between 2 % and 30 % of plate area [30].

Granulator chamber geometry has also an impact on the granulation process. Conical shape is usually preferred to control the fluidisation velocity i.e to keep the lighter particles from flying into the filters and keep the heavier particles from falling to the bottom of the chamber [30].

Mechanical agitators can be used in a fluidised bed to break up the granules in the process and thus, produce granules with narrow size distribution and high density [16].

2.5.2 Process parameters

A fluid bed granulation process can be controlled with an inlet air flow rate [48]. If the inlet air flow rate is too high, especially the fines of material may be blown out of the granulating chamber and in the worst case, may cause clogging in the filters. Too low inlet air flow rate causes defluidisation, especially when the granulation liquid increases the bed weight. Increase in the inlet air flow rate has been reported to cause a decrease in the granule size due to enhanced evaporation and attrition [27, 46]. Granule size may be decreased even more with increasing the process temperature, which also densifies the granules due to the faster rate of solvent evaporation.

The overall moisture content of the powder bed, is an important factor in the FBG, as the liquid concentration of the bed determines the adhesion between the particles [27, 31]. Large granules are produced in conditions with high relative humidity.

Also the granulation liquid flow rate is an important factor in FBG. If not controlled properly i.e. to be in balance with the evaporation rate, too high flow rate may result in overwetting of the bed and too low rate may cause spray drying of the granulation liquid droplets in the inlet air stream. There is a linear and positive correlation between the granulation liquid flow rate and granule size, found by numerous researchers [49, 52-54].

________________________________________________________________________

12

2.5.3 Droplet size and droplet size distribution

With certain conditions – when the relationship droplet size and granule size is preserved during the granulation process - there is also a relationship between the droplet size and the primary granule size, as first reported by Schæfer et al. [27]. In their work, granule breakage was not found to be significant, because the spraying of the granulation liquid ensured bonding between the particles, as later discussed by Waldie et al. [29].

In the case of pneumatic nozzles, the mean droplet size is mainly affected by nozzle construction, air-to-liquid ratio of the spray and atomising air pressure [55]. Properties of granulation liquid affect the droplet size too by surface tension, density and viscosity.

The easiest way to control the granule size in the fluid bed granulation process is to adjust the atomising air pressure of the granulation liquid as an increase in the atomising air pressure leads to a smaller droplet size [45, 56, 57] and hence, also smaller granule size [27, 49].

2.5.4 Raw material parameters

Product variables are defined as physical properties of solid material and granulation liquid [58]. Wettability is a more important factor in FBG than in high shear granulation, as there is no agitation present in the process [59].

Surface area and the water absorption properties of the starting material strongly affect granule growth [20]. For example, a well-known water uptake property by starch has been reported to show in the granulation process as a lag time in granule growth [60]. Poor wettability may result in smaller granules and can be improved with surfactants [61]. To achieve proper fluidisation is fundamental for the fluid bed granulation process. Proper fluidisation is not typical for most pharmaceutical powders, due to poor flowability and cohesiveness [62]. Differences in the densities of starting materials may cause uneven fluidisation, even more so than the difference in particle sizes [63]. Also static charging may cause problems in the fluidisation [46].

Polyvinylpyrrolidone (PVP), starch, gelatine and hydrophilic cellulose derivatives are common binders in FBG. Final granule strength depends both on binder type and concentration. Increasing binder concentration in granulation liquid has been reported to result as an increase in granule size [53, 54, 64]. Binder liquid viscosity and molecular weight of the binder polymer have an impact on the granule size, as it affects the droplet size [56].

Good solubility in granulation liquid enhances granulation [65], although the choice of binder is also affected by the following steps of the processing, such as tableting, or pharmaceutical compatibility of the components [66]. The granulation liquid solvents may be either aqueous, organic or mixes (e.g. water and ethanol), but environmental issues and safety reasons (possibility for explosion hazard) have promoted the use of the former ones.

________________________________________________________________________

13

2.6 Electrostatic atomisation

2.6.1 Definitions and theory of electrostatic atomisation

Electrostatic atomisation i.e. electrospray refers to a process where an electric field is used to generate charged droplets. When an electric field is applied to the liquid in a capillary, the liquid meniscus forms a cone and a jet of liquid is emitted from its tip and after, the jet breaks into droplets [67]. The stable cone jet (Taylor cone) of liquid produces fairly small droplets, which size is controllable by the liquid flow rate and the atomisation voltage. The charge of the same sign hinders the droplet coalescence [68].

The spray produced by electrostatic atomisation is a highly complicated function of liquid electrical resistivity ρ, viscosity η, surface tension σ and liquid permittivity ε. Generally, a liquid surface breaks in separate droplets when the Coulombic forces caused by induced charges exceed the surface tension. For stationary droplet, this limit of maximum charge q can be defined as Rayleigh limit (Eq. 1)

(1) q=8π(γε0r3)½

where γ is the liquid surface tension, ε0 is the permittivity of free space and r is the drop radius. This limit does not apply to all situations, e.g. when the droplet is very small and of high viscosity liquid [69].

One of the most important features of electrostatic atomisation are the different spraying modes, resulting in jet formation and dynamics (Fig. 6). The different spraying modes were first reported by Taylor in the mid 1960’s [67], although interactions between water and the electric field was reported by Lord Rayleigh in the 19th century [70]. With the cone jet mode it is possible to produce monosized droplets, although it is sensitive to any changes in atomisation liquid, such as temperature. Most of the studies are conducted using either the cone jet mode or the micro dripping mode [71]. Also other spraying modes have been studied, focusing on the classification of different modes [6].

Electrostatic atomisation has several advantages compared to conventional atomising systems. It is possible to produce small droplets possessing narrow droplet size distribution. Coalescence of the droplets is not an issue, as droplets undergo self dispersion. It is possible to guide the droplet motion by an electric field and e.g. guide the charged droplets easily on certain (charged) object surfaces [6].

________________________________________________________________________

14

Figure 6 The different liquid meniscus forms in electrostatic atomisation. A= cone jet, B= drop, C= microdripping, D= spindle, E= multi-jet, F= multi spindle, G= ramified jet. Modified from ref [6].

2.6.2 Pharmaceutical applications of electrostatic atomisation

There are several pharmaceutical dosage forms – for example dry powder inhalers – which do not meet their quality requirements without well-defined and optimised drug particles [72]. The desired properties depend on formulation, but particle size distribution, as well as surface properties are critical [73]. Crystallisation is a time consuming way to produce particles, micronisation or milling may cause problems in the crystal structure and unpredicted morphology.

Electrostatic atomisation of solutions or suspensions enables particle production – in certain conditions even in nanosize [71]. Also drug particles have been produced by electrostatic atomisation: acetosalicylic acid [74], indomethacin [75], paracetamol [73] and insulin [76]. In the case of solutions, after atomisation droplets are evaporated and solid drug particles are formed by crystallisation. When atomising suspensions, fine powder particles form a cluster after solvent evaporation. It is possible to control particle size of produced particles by adjusting the liquid flow rate, the strength of the electric field, concentration or ionic strength of the liquid [77].

Also droplets generated by electrostatic atomisation may be used in drug delivery. For instance in lung delivery, the aerosol droplets have to be less than 5 μm in diameter to reach the alveoli region [88].

2.6.3 Other applications of electrostatic atomisation

Nanoparticles, produced by electrostatic atomisation have been applied in numerous studies in various fields of natural sciences [78]. Electrostatic spray has numerous other applications [78]. It has been utilised e.g. in crop spraying [79], painting [69], ink-jet printing [80] and emulsion production [81]. Metal powders are also produced

________________________________________________________________________

15

by electrostatic atomisation [69]. The ink-jet printer is an application of a hybrid system, where the jet itself is not necessarily produced by electrostatic forces.

2.7 Miniaturisation

Miniaturisation - the creation of ever-smaller scales for mechanical, optical, and electronic products and devices - has been an ongoing trend also in pharmaceuticals for years [82] and maybe the most well-known application are high throughput methods developed for drug discovery [83]. Miniaturised approaches are also needed to increase understanding of physical and chemical phenomena that are important in pharmaceutical processes [84].

Lack of an active pharmaceutical ingredient (API) in the early stage of development has an impact on various stages of the development of a new drug product and thus, the formulation development phase has been carried out later in the process [85]. Besides different process equipment (granulators, mixers, pelletising machines), also whole formulation strategies as well as analysis have been miniaturised. As extremity, a miniaturized version for intrinsic dissolution testing has been developed and reported to reduce the need of API up to 10 000 fold [86]. Miniaturisation causes cost reductions due to low consumption of API, but also reagents. Small scale also diminishes the laboratory space requirements and enables easy shipping and transport [83]. There are also fascinating applications of microchips that provide unique possibilities for pharmaceutical analysis.

Also, advances in microelectromechanical systems have created biomedical implantable microdevices, which can aid in the diagnosis, monitoring or treatment of various (especially chronic) illnesses. Controlled drug delivery is possible with these devices either passively or by external activation [87].

Recent publication by Van Eerdenbrugh et al. [85] points out the importance of feasibility of batch size of miniaturised processes. The researchers investigated the production of nanosuspensions and concluded that it was possible to carry analyses, such as X-ray powder diffraction and thermal analysis (differential scanning calorimetry, DSC) with a total amount of 10 mg of API in the produced suspension.

Down-scaling of production processes is important, as it is possible to produce more data with smaller sample sizes and hence, possibly avoid unoptimal process conditions. To conduct these small scale investigations, there are several devices available for different processes, such as pelletisation and spray drying. Some commercial small scale fluid bed granulators are presented in Table 3.

________________________________________________________________________

16

Table 3. Commercial fluid bed granulators

Granulator Batch size (min-max) Size

MP Micro [88] 20 – 200 g Table top

STREA-1TM [88] 100 ml - 2 l Table top

4M8 Fluidbed 50 – 200 g Laboratory scale

Mini-Glatt 50 – 300 g Laboratory scale

The presented granulators differ from each other by batch size, but also by automatisation and monitoring possibilities. Since the granulation chamber is small, it is not possible to add probes inside the device, as in larger granulators. Although transparency makes it possible to watch the process and detect possible problems, such as material sticking to the chamber walls, not many devices offer this possibility, as the production scale devices are predominantly made of steel. Materials may act differently, as e.g. electrostatic charging is more likely in chamber made of plastic or glass.

2.8 Modelling and multivariate methods

Modelling provides advantages to the pharmaceutical industry e.g. by improving the fundamental understanding of unit operations and minimising risks of scale-up due to predictions based on process data. Modelling can be divided into several subgroups. Molecular modelling is referred to studies on e.g. molecular structure, reactivity, selectivity and thermochemical properties Process modelling focuses on unit operations, such as crystallisation, granulation, blending or coating. Systems modelling may consist of process optimisation, cost analysis or quality risk assessment. The main idea in modelling is to save time by performing a minimal amount of experiments and then use the acquired dataset to develop a model to conduct virtual experiments in silico [89].

Pharmaceutical process analysis produce large amount of data. Multivariate analysis (MVA) is extremely powerful e.g. in interpretation of spectroscopic data, where differences between observations are sometimes small and signal-to-noise ratio is low. Besides pharmaceutical process analysis, multivariate techniques are also used in early-stage preclinical research [83].

2.8.1 Experimental design

It is challenging to establish relationships between numerous different factors in pharmaceutical processes. For example, in tabletting a well thought-out experimental plan is required in the early development phase due to the the complexity of interaction between materials and processes [90].

________________________________________________________________________

17

The main idea of an experimental design (Design of Experiments, DoE) is to plan and conduct experiments in order to extract the maximum information from the collected data with the smallest number of runs. DoE starts with the definition of factors, i.e. parameters that can be used to influence responses. The goal is then to find the combination of factors that produce the most optimal set of responses. The results are usually presented as mathematical models [91].

2.8.2 Principal component analysis (PCA)

Principal component analysis (PCA) is a multivariate projection method, which is often used in the early stage of data analysis. It is a technique for a qualitative analysis to get an overview of the data set or classifying and detecting trends or outliers in the data [92].

PCA extracts and displays systematic variations in the original data matrix by dividing the information into two low-dimensional matrices, scores and loadings. Score values explain similarities and differences between observations. Loadings are used to explain the origin of the similarities and the differences. Also loadings describe how the variables influence the principal components (PC) [93]. Each PC consists of a loadings vector and a score vector, which are orthogonal to each other. The data is organised so, that the first component (PC1) explains most of the variation and the PC2 the second most.

PCA has been e.g. combined with spectroscopic techniques to quantify between different polymorphs of drug (Raman spectroscopy [94]). It has also been applied to studies of API synthesis (Raman spectroscopy [95], crystallisation (Raman spectroscopy [96]) and used in detecting the purity of a polymorph (near infra red spectroscopy, NIR [97]). Multivariate visualisation has been applied by PCA e.g. to screen for differences in the amorphous drug [98] and to compare particle size and shape distributions [99]. PCA has been used in numerous other applications, among them; testing of drug dissolution profiles [100], monitoring wet granulation [101], as a tool in the food quality improvement process [102] and quality control of herbal products [103].

2.8.3 Partial least squares (PLS) regression analysis

Partial least squares (PLS) regression analysis is a technique to correlate two data sets together. PLS is very useful especially, when the amount of variables is high, because it is possible to fit the model for all the responses simultaneously [92].

In PLS regression analysis, the variables (matrix x) are correlated with the responses (matrix y). First, PCA is performed to both of the matrices and then the best possible correlation between the matrices x and y is determined using the least squares method. As a result, a model that describes the dependence of matrices x and y is obtained [92, 93].

________________________________________________________________________

18

PLS regression analysis has numerous applications. A common use for PLS regression applied to spectral data is to determine the polymorphic state and quantitate the amount of API [104] e.g. in the case of famotidine [105], ranitidine [106] and atorvastatin [107]. Also flow properties of pharmaceutical powders [108], real time assessment of granule and tableting properties [109] and fluidised bed coating [110] have been modelled with PLS regression analysis. Besides in pharmaceutics, PLS regression analysis has been used in e.g. food science to assess meat quality [111] and detecting puree adulteration [112].

________________________________________________________________________

19

3 Aims of the study

The aim of this thesis was to:

1. Develop and characterise a suitable spraying method for Multipart Microscale Fluid bed powder Processor

2. To use particle tracking velocimetry and high-speed imaging for

characterisation of the spray in order to evaluate spray parameters

3. To investigate process parameters of the spray and fluid bed granulation affecting granule size

4. To study applicability of Multipart Microscale Fluid bed Powder processor

with electrostatic atomisation in granulation for early formulation studies

________________________________________________________________________

20

4 Experimental

4.1 Materials

The materials used in the work are listed in Table 1.

Table 4. The materials used in the work.

Excipient/API Brand Supplier Publication

Acetylsalicylic acid (ASA) Pharma-Chemicalien,

Rütgers Organics GmbH,

Germany

III

α-lactose monohydrate

(LMH)

Pharmatose 80 M

Pharmatose 200M

DMV International,

Netherlands

I, II, III, IV

Microcrystalline cellulose

(MCC)

Avicel PH 101 FMC Biopolymers, Ireland III

Polyvinylpyrrolidone

(PVP) Kollidon K-25 BASF, Germany I, II, III, IV

4.2 Multipart Microscale Fluid bed powder Processor (MMFP)

Multipart Microscale Fluid bed powder Processor (MMFP, Ariacon Oy, Turku, Finland) is a miniaturized device, which has been developed for fast characterisation of pharmaceutical materials in order to investigate processability. The set-up consists of five conical glass chambers, which are individually controllable and can be adjusted into the same line if needed. During experiments, MMFP was linked to an air control unit (Ilmasäätö Oy, Turku, Finland), which enables the use of heated (up to 70°C) or humidified process air. There were also resistor type heating elements per fluidisation chamber. The granulation set-up of the device is presented in Figure 7.

________________________________________________________________________

21

Figure 7 The schematic diagram of the granulation chamber in MMFP. HV= high voltage supply. Modified from ref.[113].

Unlike in the production scale devices, there are no filter bags in MMFP, due to precisely controlled air flow. The air flow in MMFP was created with a fan and measured by a high accuracy meter (Ilmasäätö Oy, Turku, Finland). The size and the structure of the metering unit were optimised off-line and calibrated against a valid reference (Golibrator-2, Sensidyne, Clearwater, USA). Air flow rate and the dimension of the chamber inner diameter were used to calculate air flow velocity.

Previously, MMFP has been used for studying solid state transitions during fluidisation [84], dehydration process parameters [113], flowability [114], drying of pharmaceutical materials [115, 116], fluidisation of different materials [62] and pellet tackiness determination [117].

4.3 Granulation in Multipart Microscale Fluid bed powder Processor (MMFP)

The use of pneumatic nozzles was not possible in the Multipart Microscale Fluid bed powder Processor (MMFP) due to the fact that fluidisation especially at low air flow

________________________________________________________________________

22

rates would have been disturbed. In preliminary experiments, electrostatic atomisation was found to be the most promising technique to droplet production in granulation process and thus, a specific nozzle was built up. The nozzle consists of earthed shield, polytetrafluoroethylene (PTFE) insulation and a metal capillary connected to a high voltage power supply (Spellman, High Voltage Corporation, New York, USA). The outer diameter of the capillary was 0.5 mm and the axial distance from tip to shield was adjusted to be 4 mm. The nozzle was placed to the granulation chamber, 10 cm above the powder bed. Granulation liquid was pumped through the capillary using a syringe pump (Model NE-1000, New Era Pump Systems Inc., New York, USA). The nozzle is presented in Figure 8.

Figure 8 The schematic diagram of the nozzle and a photograph of the nozzle inside the granulation chamber. In diagram: A= earthed shield, B= insulator, C= metal capillary.

In the beginning of the granulation experiments, powder materials of the batch were poured into the chamber. The binder (PVP) was either dissolved into the granulation liquid (I, II) or added as dry powder (III). The fluid bed was set in motion and the air flow was adjusted. As the air flow rate reached 100 ml/h, regardless of the target flow rate which depends on the batch size, a high voltage supply was switched on to apply an atomisation potential to the capillary. After that, pumping of the granulation liquid started.

As the bed weight increased due to the spraying of the granulation liquid, the air flow rate was manually raised to the target level at an adequate rate. The air flow rate had to be high enough to keep the wet powder in motion, but low enough to prevent a blockage in the nozzle by the powder. A typical batch size in MMFP is from 25 to 45 ml.

________________________________________________________________________

23

4.4 Methods for droplet and particle size analysis

4.4.1 Droplet size measurement using laser diffractometry (I)

Laser diffraction (Malvern System 2600, Malvern Instruments, Worcestershire, UK) was used to determine the droplet size distribution. Measurements were performed at least three times for each process condition.

4.4.2 Droplet size measurement with particle tracking velocimetry (IV)

A special imaging measurement was constructed. The system consisted of a HiWatch diode laser stroboscope (Oseir Ltd, Tampere, Finland) and industrial grade 1.5 Mpix greyscale CCD camera (Lumenera, Lumenera Corporation, Ottawa, Canada). A close- up zoom video lens (Navitar 7000, Navitar Inc, New York, USA) offered a range of magnifications to accommodate the large variation of spray density at different distance points. Oseir Osirec software (Oseir Ltd, Tambere, Finland) was used to extract the droplet speed and size information from the captured images. The measurement zone was adjusted to catch as many droplets as possible around the central line of the spray zone. The distance from the nozzle orifice was adjusted to be 4, 103 and 202 mm from the camera lens, respectively and the distance from the central line of the spray was always 107 mm.

Figure 9 The particle tracking velocimetry measurement set-up.

________________________________________________________________________

24

The diode laser and the camera were used to produce images in backlight set-up, which is useful in accurate sizing of droplets [118]. The zoom lens magnification and hence, the image area were adjusted to find the setting that produced a satisfactory number of droplet detections at each distance point from the nozzle.

The use of particle tracking velocimetry requires that a single droplet is captured multiple times to determine the droplets velocity. To achieve this, the camera shutter time used was relatively long (40 ms) and the interval of the laser pulses was approximately 1 ms. Laser pulse duration was 100 ns.

Figure 10 An example photo of the droplets taken with a CCD camera.

4.4.3 High-speed imaging (IV)

Videos of the spray were shot using a high speed camera (Casio EX-F1, Tokyo, Japan) at 600 frames per second.

________________________________________________________________________

25

4.4.4 Granule size measurement with sieve analysis (II)

Granule size distribution was analysed with miniaturised sieves (diameter 10 cm, Fritsch GmbH, Germany) with apertures of 45, 65, 125, 250, 355, 500, 710 and 1000 μm. The amplitude of the sieve shaker was 4 mm and the sieve stack was kept in the shaker for 4 min. The granule diameters d10, d50 and d90 were obtained from the sieve analysis data using the method of Meshali [119]:

(2)

where xi is the relative weight of the given fraction with a geometrical mean diameter di equal to

(3)

It was not possible to repeate any of the sieve analysis to evaluate the standard deviation of the procedure as a result of small batch size. The sample sizes were at a range 9.5 – 15.2 g, as the manufacturer of the sieves recommends sample sizes not less than 10 g.

4.4.5 Granule size measurement with image analysis (II)

Projected area diameters i.e. the diameters of a circle having the same area as the projected particle were obtained by optical microscopy and image analysis (IA). For the analysis, four samples (0.5 g) of each batch were randomly collected from storage and poured into a petri dish with grid lines. Five of the checks on the petri dish were blackened and only particles on those were analysed. Altogether 20 images per batch were taken by an optical microscope (Leica MZ6, Leica Mikroskopie und Systeme GmbH, Bensheim, Germany). N-values varied between 292-799 per measured batches. Granule size was determined using image analysis software (Leca Qwin, Leica Imaging Systems Ltd., Cambridge, UK).

4.4.6 Granule size measurement with spatial filtering technique (III)

Granule size distribution was determined using a spatial filtering technique (SFT) with 2.5 g samples. To ensure representativeness of the sample, the whole batch was first poured through a small scale sample divider (SBX-450-R, Endecotts, London, UK). The SFT apparatus (Parsum® IPP 70-Se, Gesellschaft für Partikel-, Strömungs- und Umweltmesstechnik GmbH, Chemnitz, Germany) was placed on the table and the sample was poured through a 4 mm orifice. Pressurized air was used to disperse the

________________________________________________________________________

26

particles. Raw data from the measurements was collected via an A7D converter to a PC. The transformation from the number based particle size distribution was transformed into volume distribution by SFT apparatus software.

4.5 Granule characterisation

4.5.1 Friability (III)

The friability of the granules was determined from change in particle size detected by spatial filtering technique (SFT). First 5 g of each granule batch was measured with SFT. Then, the friability of the granules was tested. The same amount of granules (5 g) and glass beads (diameter 5 mm) were weighed in a 100 ml container and the container was spinned in a Turbula® mixer (type T2F, WA Bachofen, Basel, Switzerland) for 5 min at 96 rpm. Then, the 5 g batch of granules was measured again with SFT. The size fraction d50 values of the batches before and after the friability tests were compared and the difference between the d50 values was calculated.

4.5.2 Compactibility (III)

The compression properties of the granules were studied with an instrumented eccentric tablet machine (Korsch EK0, Erweka Apparatebau, Germany). Three samples (250 mg each) of every batch were manually poured into a die and tabletted at various compression forces. Tablet punches used were flat faced and 9 mm in diameter. The starting compression force was determined individually for each batch to obtain tablets strong enough for handling. The tablet die was lubricated before every tabletting with a solution of magnesium stearate and acetone (20 % w/w). For batches, at least five tablets of 250 mg (+/- 10 mg) were compressed. For every compression force, one tablet (n=1) was compressed because of the small batch size. Thereafter, the breaking force (BF) of the tablets was measured using a tablet hardness tester (Erweka, Germany).

Compactibility of the granules was evaluated by plotting the breaking force (BF) of the tablets as a function of the maximum force of compression. The curve obtained is material dependent and describes the tablettability of the mass i.e. how strong bonds the particles develop during compression as the compression force is increased. In the paper III slope (K) of the linear part of this curve has been used as a response in modelling.

________________________________________________________________________

27

4.6 Multivariate analysis and modelling (I, II, III, IV)

Design of Experiments (DoE) in papers II, III and IV was carried out with MODDE® software (MODDE®- for Windows, Version 7.0, Umetrics, Umeå, Sweden).

Principal component analysis (PCA) was applied to the granule data in paper II to better differentiate between the produced batches. PCA was carried out using SIMCA-P® software (Version 10.5, Umetrics, Umeå, Sweden).

Partial least squares regression (PLS) was applied to the granule size and the granulation process data (III). The variables for the model were the volume proportions of the excipients and API in the composition, the particle size distribution obtained by spatial filtering technique (SFT) and particle size fractions d10, d50 and d90. Breaking forces and K-values (calclulated from the slope obtained by plotting the breaking force as a function of maximum compression force) were used as responses. A correlations scaled loadings plot was used to visualise the findings. Also multiple linear regression (MLR) was used in paper III to confirm the relationship between the K-value and the proportions of the excipients. PLS and MLR were carried out using SIMCA-P® software (Version 10.5, Umetrics, Umeå, Sweden) and MODDE® software (Version 7.0, Umetrics, Umeå, Sweden).

In paper IV the MLR was applied to the data to find the dependence of the droplet size fractions from the atomisation voltage, granulation liquid flow rate and nozzle position. MODDE® software (Version 7.0, Umetrics, Umeå, Sweden) was used to do the second order polynomial fitting. The acquired models were simplified by excluding the least significant terms from the model as long as the predictive power (Q2) of the model was increased and the statistical significance was adequate.

Modelling of the collecting electric field in paper I was performed with Opera-3D software (Vector Fields Ltd., Oxford, UK).

K-values for the modelling in paper III were calculated with Sigma Plot program (version 6.10 2000 for Windows, SPSS Inc. USA).

________________________________________________________________________

28

5 Results and discussion

5.1 Electrostatic atomisation

5.1.1 Droplet formation (I)

Electrostatic atomisation is applicable to produce droplets needed in the fluid bed granulation process in MMFP (I). There are several advantages in using electrostatic atomisation in the miniaturised fluid bed system. There is no need for pressurised air – which could easily destroy the fluidisation and decrease the yield by blowing material to the chamber walls - in the production of droplets. Also the use of a collecting electrical field facilitates the droplet movement towards the powder bed, which is necessary due to the dimensions of the conical granulation chamber. To assess the potential required to guide the droplets to the powder bed, forces affecting the individual droplets were studied.

When highly charged droplets were emitted from the capillary tip, they encountered various forces. Gravitational force (Fg), which was directed downwards from the nozzle tip, air drag force (Fd) which is caused by upwards moving air and droplet velocity, electrostatic force (Fi) induced at the earthed shield and the repulsive charge force which is caused by the unipolarly charged droplets. Another high voltage source, which operated at opposite polarity than the droplets, was attached to the metallic air distributor. This caused a collecting electrostatic force Fc, which accelerated charged droplets downwards into the fluid bed.

The terminal velocity for the droplets in upwards moving gas with radius 100 μm (acquired from the laser diffraction studies) was estimated to be 0.603 ms-1 (I). The value was calculated using Equation 4, which includes the assumption that gravitational force and air drag force are equal. If there was no collecting electric field and the droplet would be relatively far from the nozzle (image force = 0), the droplet velocity would be a function of air flow velocity. Thus, if the air flow was the same as the terminal velocity, where the air drag force and gravitation are equal, the droplet would stay stationary. Calculation of terminal velocity seems logical parameter to be calculated, when the necessary collecting electric field is assessed. In other words, when air drag force and gravitation are equal, the only accelerating force acting on the particle, is Fc (the collecting electric field).

In the equation, m is the mass of the droplet, g is the gravitational constant, CD is the drag coefficient, ρg is the density of air, d droplet diameter, v the relative velocity between the droplet and the surrounding air and η is the viscosity of air.

________________________________________________________________________

29

(4) Numerically solved equation for a droplet in upwards moving gas.

Figure 11 presents different forces as a function of distance from the nozzle. Also an example of a collecting electric field (500V) is presented.

Figure 11 Induction force (Fi), gravitation force (Fg), air drag force (Fd), collecting force (Fc, 500V) and the total force Ftot (500 V) as functions of distance from the electrostatic nozzle. Modified from paper I

5.1.2 Nozzle performance (IV)

Nozzle performance was further evaluated in paper IV. The experiments were performed with ion-exchanged water. When the atomisation voltage is increased at constant liquid flow rate, the frequency of dripping also increases, as the

________________________________________________________________________

30

electrostatic forces grow at water surface [71]. Increasing the liquid flow rate at constant atomisation voltage increases droplet size.

The spray is not very dense, as the number of the droplets per mm2 varies from 0.05 to 3. Trajectories of the droplets are mainly directed downwards, also small satellite droplets with unoptimal trajectories can be observed. Especially these satellite droplets gather to the nozzle shield and as process time elapses, produce a large droplet, which can be observed from Figure 12. This droplet may cause at least some of the big aggregates (lumps) found in granulating experiments. The nozzle performance is presented in Figure 12.

Figure 12 Nozzle performance as a function of the granulation liquid flow rate. The large droplet caused by gathered smaller droplets is present in figures 60 ml/h, 90 ml/h and 110 ml/h.

Electrostatic atomisation was found to be applicable for fluid bed granulation. However, in order to get even more information about the nozzle and atomisation method, other granulation liquids, than mixtures of water and polyvinylpyrrolidone could have been studied. Certain constraints were set by the device, e.g. level of viscosity (too viscous fluids will not disperse into droplets and stickyness of the solution. Also safety was an issue, as the use of solvent-water mixtures would have increased the risk for sparks and thus, possible explosion. One option to modify the droplet size could also have been through the resistivity modification by adding salt to the solution.

________________________________________________________________________

31

5.1.3 Droplet size and droplet size distribution (IV)

The dependence of droplet size fractiles d10 d50 and d90 from the atomisation voltage, granulation liquid flow rate and nozzle position were modelled separately using multilinear regression. The models are presented in Table 5. What is notable, is that the atomisation voltage was not significant term compared to the others – as also stated previously [120].

Table 5. The coefficients, R2 and Q2 values of models for droplet size fractiles d10, d50 and d90. V= atomisation voltage, F= granulation liquid flow rate, N= measure- ment distance from the nozzle, a0=constant.

d10 d50 d90 V - - - F 0.163*** 0.184*** 0.194*** N 0.116*** 0.070** 0.049* V·V - - - F·F - - 0.109*** -0.140*** N·N -0.93** -0.077* - V·F - - - F·N 0.105*** 0.074* - V·N - - - a0 2.564*** 2.705*** 2.757*** R2 0.848 0.878 0.856 Q2 0.779 0.796 0.799

*p<0.05 **p<0.01 ***p<0.001

Increase in the granulation liquid flow rate increased all fractile values measured at all three distances from the nozzle. This is in line with literature [120], although there are situations (e.g. very low flow rates), when there is no impact to the droplet size [121]. The largest droplets are observed at the furthest distance from the nozzle at high granulation liquid flow rate possibly due to the lack of a collecting electrical field.

Typical droplet size distributions were observed when atomisation voltage and granulation liquid flow rate were studied. Type A distribution was relatively narrow (range 150 μm – 350 μm) and wider Type B distribution (range 150 μm – 650 μm). Low granulation liquid flow rate (10 ml/h) was typical for type A distribution, whereas type B distributions were produced at higher flow rates. The low number of observations (droplets) diminishes the relevance of the data, especially at the furthest distance from the nozzle when the spray density was not adequate enough for particle tracking velocimetry. Producing a narrow droplet size distribution is not

________________________________________________________________________

32

simple, as the operating zone is small. Droplet size distributions at 4 mm distance from the tip are presented in Figure 13.

Droplet size distribution studies would have been more reliable, if the spray had been denser. This could have been achieved by nozzle modification, e.g. using larger capillar and higher liquid flow rates.

Figure 13 Droplet number size distributions at 4 mm distance from the tip. Type A distribution is found in batches N1 and N2.

________________________________________________________________________

33

5.2 Granule properties (II, III)

5.2.1 Granule size and size distribution

In paper II, granule sizes were determined by sieve analysis. Granule size was found to be fairly similar despite a different quantity of granulation liquid used in the study. Low granulation liquid flow rate and low atomisation voltage seemed to produce small granules. Increase in binder concentration increased the granule size, as expected (paper II, Fig 8). Sieve analysis was not considered as a reliable analysis method for produced batches, as the quantity of samples was small and hence each batch was only possible to be analysed once. The reliability of this method could have been better, if the sample divider would have been in use.

It would have been also informative to add the size data of original lactose monohydrate as a figure to the paper II, in order to be able to better compare the change after granulation. According to the manufacturer (DMV International, Netherlands), less than 20 % of the particles have been reported to be smaller than 100 μm, 70-90 % of the particles have been reported to be smaller than 250 μm and 95 % of the particles have been reported to be smaller than 315 μm.

Based on image analysis (IA, paper II), the amount of larger granules clearly increased, when the atomisation voltage was increased. This was not expected, as higher voltages are expected to produce smaller droplets and hence, smaller granules. Further studies of the nozzle have revealed (paper IV, figure 4) that droplet size decreases with increased atomisation voltage. However, the increase in the amount of lumps and larger granules (II) detected by IA was concluded to be strongly related to the experimental set-up (choice of material, amount of granulation liquid) and process conditions (adherence of small droplets to nozzle shield and formation of a large droplet).

5.2.2 Granule size by multivariate visualisation and principal component analysis (II)

Granule size data (d10, d50 and d90 values) were plotted in three dimensional space as a function of the granulation liquid flow rate (paper II, Fig.9). Critical value of the flow rate was found to be 36 g/h and it was possible to cluster the data by higher granulation liquid flow rates. It was not possible to cluster the data based on the atomisation voltage or binder concentration.

Principal component analysis (PCA) was applied to the sieve analysis data of the batches to provide visualization and an overview of it. Batch 12 was removed from the model, as it was an outlier (both R2 and Q2 values of the model were improved after the removal). Possible explanation for this is the combination of low atomisation voltage and high granulation liquid flow rate. The atomisation voltage has not been high enough to produce droplets of PVP solution and thus, the

________________________________________________________________________

34

granulation process has not been optimal. At least batch 12 has been different from other batches to fit in the model. This difference could probably have been found from the process data, if it had been added to the model. Two principal components were needed to explain ~ 70% of the variation in the data (PC1: 51%, PC2:19%).

Q2 value (predictivity) of the model was 23.6%. There are several explanations for the low predictivity value; to acquire more predictive model, the number of experiments should have been higher. More experiments would most likely have provided also normal distribution, as required. There are several factors that could have added to the model, such as temperatures and humidities from the process, as well as the final moisture content of the granules. However, certain trends in this very complex system were possible to find using this fairly rough model.

Scores and loadings plots are presented in Figure 14. Figure 14 is plotted in order to reveal the effects of binder concentration and atomisation voltage, which are much weaker than the effect of granulation liquid flow rate.

The first principal component was related to the particle size. Further on the left the scores are in the plot (Figure 14, left side), the more dominating is the granule mass fraction on sieves with the widest apertures (S5-S8), i.e. the amount of large granules is high in the batch. The smaller granules are dominating in the batches that are located on the right hand side of the plot. The scores plot can be divided into segments by changes in the atomisation voltage and granulation liquid binder concentration. Also, batches that were far apart in PCA plot differed in the sieve analysis data. It was not possible to explain the variation in PC2.

The loadings plot (Figure 14, right side) explains, how the measured sieve fractions contribute to the separation of the batches. For instance, in batch 16 (circled black square) the mass on sieve S2 (710 μm) is dominating.

Figure 14 Scores and loadings plots of 21 batches. Key for process conditions:. ■ = X, high oltage, high binder concentration, ▲= X, medium voltage, medium binder concentration, ●= X, low voltage, low binder concentration, ○= any other process condition. X= any rate of granulation liquid flow rate. S1-S8= sieve numbers, where S1 is the sieve with the smallest and S8 with the widest aperture. Modified from II.

________________________________________________________________________

35

To analyse further, PCA was applied separately to the sieve analysis data from subset A and B. The PCA of subset A needed two principal components to explain the variation and Q2 value of the model was 51%. The acquired results were parallel with the PCA of the whole data set. PCA of subset B resulted in high negative Q2-values for both principal components and was discarded. This was most likely due to the low number of batches.

5.3 Early formulation screening and modelling (III)

An attempt to mimic early formulation screening studies with well-known materials on a small scale was performed in paper III (Figure 15). The focus of the study was on the mechanical properties of the granules: particle size, compactibility and friability.

Figure 15 The outline of preformulation study performed in paper III.

Breaking force and K from the compactibility studies and delta d50 from the friability studies were used as the responses in partial least squares (PLS) regression analysis. The variables for the model were the volume proportions of excipients and active pharmaceutical ingredient (API) in composition, particle size distributions obtained by spatial filtering technique (SFT) and particle size fractions calculated from previous. Three factors were needed for a PLS model. The R2X, R2Y and Q2Y values for the model were 68.4%, 81.6% and 49.7%, respectively. The first factor explained more than half of the variation of the batches (56.5%), whereas the other two factors explained 13.6% and 11.6% of the variation.

________________________________________________________________________

36

Correlation scaled loadings plot (or loadings Bi-plot) was used to interpretate the results (Figure 16). The correlation scaled loadings plot is a useful tool to get an overview of the whole data set simultaneously.

In this plot, the scores and loadings are plotted in the same figure to facilitate the interpretation of the data. This plot is used in order to detect correlations between factors and responses. Positively correlated factors and/or responses are located near each other and negatively correlated are in the opposite direction (180 degrees via origin). For example, MCC (microcrystalline cellulose) and BF (breaking force) are located near one another in the plot and thus, when the MCC amount is high in the composition, the breaking force of the produced tablets is high. Also, the batches that produced the weakest tablets are located further in the opposite side of the plot than BF.

It is possible to distinguish between the excipient proportions; batch (A9) containing only lactose monohydrate (LMH) is near the LMH in the plot. Also, the friability testing results can be seen from the plot, as the strongest granules according to the data (paper III, table II) are located furthest from the delta d50 in the plot.

Figure 16 Correlations scaled loadings plot (III)

________________________________________________________________________

37

Preformulation data was also modelled with Multiple linear regression (MLR). The variables were the proportions of the excipients in the composition and the response was slope K from the tabletting studies (III, Fig.5). Value K increased when the proportion of MCC increased in the composition and value K was the lowest in the compositions where ASA was dominating. R2 for the model was 60.9 but Q2 was negative, which indicates that the model was not predictive at all. This is explained by the missing experiments that could not be produced with MMFP due to technical reasons and the noise in the data. Although the predictivity of the model was low, the findings were in line with the PLS model obtained and the literature.

With this set-up, it was possible to conduct early formulation granulation studies with relatively small amount of materials. The choice of materials was not optimal for this study, as one third of the experiments were not conducted due to malfunction in the process. As the performance of the model drug and the excipients was well-known, adding another model drug possessing different properties in granulation and compression than acetylsalicylic acid, would have been interesting. Addition of the process data to the model would have provided more information about the different compositions.

________________________________________________________________________

38

6 Conclusions

Based on the results obtained during the study, Multipart Microscale Fluid bed powder Processor (MMFP) combined with electrostatic atomisation is a suitable device for fluid bed granulation on a small scale.

1. Electrostatic atomisation was found to be applicable for fluid bed

granulation in MMFP and the constructed nozzle was found to produce narrow droplet size distributions.

2. Particle tracking velocimetry was found to provide useful information about

the droplet size distribution. High speed imaging provided valuable additional information about the spray.

3. In the granulation process in MMFP, granulation liquid flow rate has the

largest positive effect to granule size. Due to the great effect of the flow rate, the effects of binder concentration and atomisation voltage were difficult to detect. However, they both have proportional effects to granule size. With principal component analysis it was possible to detect the trends in the granule size change

4. MMFP combined with electrostatic atomisation can be utilised for studying

the granulation properties of materials with a relatively small sample size. Also early formulation screening studies can be conducted, although the analysis may be challenging due to a small batch size. The granulation process and following analysis is fast and the results are obtained with a small amount of powder compared to traditional methods. Multivariate analysis can be applied to the process data and acquire adequate results to guide the developmet process further.

________________________________________________________________________

39

References

1. D.W. Light, J.K. Andrus, R.N. Warburton, 2009. Estimated research and development costs of rotavirus vaccines. Vaccine 27, 6627-6633.

2. J.Y. Zheng, An Overview, in: J.Y. Zheng (Ed.), Formulation and Analytical Development for Low-Dose Oral Drug Products, John Wiley & Sons Inc., New Jersey, USA, 2009, pp. 2-20.

3. J.F. Gamble, A.B. Dennis, M. Tobyn, 2009. Monitoring and end-point prediction of a small scale wet granulation process using acoustic emission. Pharm. Dev. Technol., 14, 299-304.

4. M.A. Mullier, J.P.K. Seville, M.J. Adams, 1991. The effect of agglomerate strength on attrition during processing. Powder Technol., 65, 321-333.

5. M.D. Tousey, 2002. The Granulation Process 101: Basic Technologies for Tablet making. Pharm. Technol., 26, 8-13.

6. A. Jaworek, A. Krupa, 1999. Jet and drops formation in electrohydrodynamic spraying of liquids. A systematic approach. Exp. Fluids, 27, 43-52.

7. S.M. Iveson, J.D. Litster, K. Hapgood, B.J. Ennis, 2001. Nucleation, growth and breakage phenomena in agitated wet granulation processes: a review. Powder Technol., 117, 3-39.

8. P.R. Mort, 2005. Scale-up of binder agglomeration processes. Powder Technol., 150, 86-103.

9. H.G. Kristensen, T. Schæfer, 1987. Granulation: A Review on Pharmaceutical Wet-Granulation. Drug Dev. Ind. Pharm., 13, 803-872.

10. O.P. Rubino, 1999. Fluid-bed technology. Overview and criteria for process selection, Pharm. Tech., 23, 104-117.

11. S.M. Iveson, J.D. Litster, 1998. Growth regime map for liquid-bound granules. AIChE J., 44, 1510-1518.

12. H. Rumpf, 1958. Grundlagen und Methoden des Granulierens. 1. Teil: Begriffe, Anwendung und Eigenschaften der Granulate. Chem. Ing. Tech., 30,144-158.

13. D.M. Newitt, J.M. Conway-Jones, 1958. A contribution to the theory and practice of granulation. Trans. Inst. Chem. Engrs., 36, 422-442.

14. F. Hoornaert, P.A.L. Wauters, G.M.H. Meesters, S.E. Pratsinis, B. Scarlett, 1998. Agglomeration behaviour of powders in a Lödige mixer granulator. Powder Technol., 96, 116-128.

15. A. Goldszal, J. Bousquet, 2001. Wet agglomeration of powders: from physics toward process optimization. Powder Technol., 117, 221-231.

16. S. Watano, T. Morikawa, K. Miyanami, 1995. Kinetics of Granule Growth in Fluidized Bed Granulation with Moisture Control. Chem. Pharm. Bull., 43, 1764-1771.

17. I.T. Cameron, F.Y. Wang, C.D. Immanuel, F. Stepanek, 2005. Process systems modelling and applications in granulation: A review. Chem. Eng. Sci., 60, 3723-3750.

________________________________________________________________________

40

18. B.J. Ennis, Theory of Granulation. In: D.M. Parikh (Ed.) Handbook of Pharmaceutical Granulation Technology (2nd Edition), Informa Healthcare USA Inc, New York, USA, 2007, pp.6-58.

19. A.S. Rankell, M.W. Scott, H.A. Lieberman, F.S. Chow, J.V. Battista. Continous production of tablet granulations in a fluidized bed II. Operation and performance of equipment. J.Pharm.Sci., 1964. 53, 320-324.

20. T. Schæfer, O. Wørts, 1977. Control of fluidized bed granulation I Effects of spray angle, nozzle height and starting materials on granule size and size distribution. Arch. Pharm. Chemi. Sci. Ed., 5, 51-60.

21. L.S.C. Wan, P.W.S. Heng, G. Muhuri, 1992. Analysis of growth of granules in a fluidized bed granulator. S.T.P. Pharma sci., 2, 381-386.

22. G.I. Tardos, M.I. Khan, P.R. Mort, 1997. Critical parameters and limiting conditions in binder granulation of fine powders. Powder Technol., 94, 245-258.

23. W.J. Wildeboer, J.D. Litster, I.T. Cameron, 2005. Modelling nucleation in wet granulation. Chem. Eng. Sci., 60, 3751-3761.

24. S.H. Schaafsma, N.W.F. Kossen, M.T. Mos, L. Blauw, A.C. Hoffmann, 1999. Effects and control of humidity and particle mixing in fluid-bed granulation. AIChE J., 45, 1202-1210.

25. T. Schæfer, C. Mathiesen, 1996. Melt pelletization in a high shear mixer. VIII. Effects of binder viscosity. Int. J.Pharm., 139, 125-138.

26. J.B. Schwartz, 1988.Granulation. Drug Dev. Ind. Pharm., 14, 2071-2090. 27. T. Schæfer, O. Wørts, 1978. Control of fluidized bed granulation IV. Effects of

binder solution and atomization on granule size and size distribution. Arch. Pharm. Chemi. Sci. Ed., 6,14-25.

28. S.H. Schaafsma, P. Vonk, P. Segers, N.W.F. Kossen, 1998. Description of agglomerate growth. Powder Technol., 97, 183-190.

29. B. Waldie, 1991. Growth mechanism and the dependence of granule size on drop size in fluidized-bed granulation. Chem. Eng. Sci., 46, 2781-2785.

30. D.M. Parikh, D.M. Jones, Batch Fluid Bed Granulation. In: D.M. Parikh (Ed.) Handbook of Pharmaceutical Granulation Technology (2nd Edition), Informa Healthcare USA Inc, New York, USA, 2007, pp.204-260.

31. S. Watano, T. Fukushima, K. Miyanami, 1996. Heat Transfer and Granule Growth Rate in Fluidized Bed Granulation. Chem. Pharm. Bull., 44, 572-576.

32. A. Faure, P. York, R.C. Rowe, 2001. Process control and scale-up of pharmaceutical wet granulation processes: a review. Eur. J. Pharm. Biopharm, 52, 269-277.

33. R. Zelko, Z. Budavari, S. Marton, I. Racz, 1999. The effect of solid/liquid interactions on the migration of riboflavin in tray-dried granule beds. S.T.P. Pharma sci., 9, 215-217.

34. K. Zuurman, G.K. Bolhuis, H. Vromans, 1995. Effect of binder on the relationship between bulk density and compactibility of lactose granulations. Int. J. Pharm., 119, 65-69.

________________________________________________________________________

41

35. A.M. Bouwman, J.C. Bosma, P. Vonk, J.H.A. Wesselingh, H.W. Frijlink, 2004. Which shape factor(s) best describe granules? Powder Technol., 146, 66-72.

36. D.E. Wurster, 1959. Air-suspension technique of coating drug particles. J. Am. Pharm. Ass., 48, 451-454.

37. L.S.C Wan, K.S. Lim, 1991. Action of binders in the fluidised bed granulation of lactose. S.T.P. Pharma sci., 1, 248-255.

38. L. Juslin, J. Yliruusi, 1996. Granule growth kinetics and attrition of granules made of different materials in a fluidized bed granulator. S.T.P. Pharma sci., 6, 321-327.

39. M.E. Aulton, M. Banks, 1978. The factors affecting fluidized bed granulation. Manuf. Chem. & Aerosol News, 49, 50-56.

40. T. Schæfer, 1988. Equipment for wet granulation. Acta Pharm. Suec., 25, 205-228.

41. P. Roy, M. Vashishtha, R. Khanna, D. Subbarao, 2009. Heat and mass transfer study in fluidized bed granulation - Prediction of entry length. Particuology, 7, 215-219.

42. W.J. Thiel, 1981. The theory of fluidization and application to the industrial processing of pharmaceutical products. Int. J. Pharm. Tech. & Prod. Mfr., 2, 5-8.

43. P.C. Record, 1980. A review of pharmaceutical granulation technology. Int. J. Pharm. Tech. & Prod. Mfr., 2, 32-39.

44. O.P. Rubino, 1999. Fluid-bed technology. Overview and criteria for process selection, Pharm. Tech., 23, 104-117.

45. L. Juslin, O. Antikainen, P. Merkku, J. Yliruusi, 1995. Droplet size measurement: I. Effect of three independent variables on droplet size distribution and spray angle from a pneumatic nozzle. Int. J. Pharm., 123, 247-256.

46. M.E. Aulton, M. Banks, 1981. Fluidised bed granulation - factors influencing the quality of the product. Int. J. Pharm. Tech. & Prod. Mfr., 2, 24-29.

47. Z. Ormós, K. Pataki, B. Csukás, 1973. Studies on granulation in a fluidized bed IV. Effects of the characteristics of the fluidized bed. The atomization and the air distributor upon the physical properties of the granulates. Hung. J. Ind. Chem., 1, 475-492.

48. P.G. Smith, A.W. Nienow, 1983. Particle growth mechanisms in fluidised bed granulation – I: The effect of process variables. Chem. Eng. Sci, 38, 1223-1231.

49. W.L. Davies, W.T. Gloor, 1971. Batch production of pharmaceutical granulations in a fluidized bed I: Effects of process variables on physical properties of final granulation. J.Pharm.Sci., 60, 1869-1874.

50. M. Hemati, R. Cherif, K. Saleh, V. Pont, 2003. Fluidized bed coating and granulation: influence of process-related variables and physicochemical properties on the growth kinetics. Powder Technol., 130, 18-34.

51. B. Rambali, L. Baert, D. Thoné, D.L. Massart, 2001. Using Experimental Design to Optimize the Process Parameters in Fluidized Bed Granulation. Drug Dev. Ind. Pharm., 27, 47-55.

52. T. Schæfer, O. Wørts, 1978. Control of fluidized bed granulation III. Effects of inlet air temperature and liquid flow rate on granule size and size distribution.

________________________________________________________________________

42

Control of moisture content of granules in the drying phase. Arch. Pharm. Chemi. Sci. Ed., 6, 1-13.

53. L.S.C. Wan, K.S. Lim, 1988. The effect of incorporating polyvinylpyrrolidone as a binder on fluidized bed granulation of lactose. S.T.P. Pharma sci., 4, 560-571.

54. L.S.C. Wan, P.W.S. Heng, B.L. Ling, 1995. Fluidized bed granulation with PVP K90 and PVP K 120. Drug Dev. Ind. Pharm., 21, 857-862.

55. K.Y. Kim, W.R. Marshall, 1971. Drop-size distribution from pneumatic atomizers. AIChE J., 17, 575-584.

56. T. Schæfer, O. Wørts, 1977. Control of fluidized bed II. Estimation of droplet size of atomized binder solutions. Arch. Pharm. Chemi. Sci. Ed., 5, 178-193.

57. M.E. Aulton, M. Banks, 1979. The measurement of spray droplet size distributions. J. Pharm. Pharmacol., 31, Suppl. 102P.

58. P.J. Sherrington, R. Oliver, Granulation. In: A.S. Goldberg (Ed.) Monographs in Powder Science and Technology, Heyden & Son Ltd., London. 1981. pp. 7-59.

59. R.C. Rowe, 1989. Binder-substrate interactions in granulation: a theoretical approach based on surface free energy and polarity. Int. J. Pharm., 52, 149-154.

60. T. Abberger, H. Egermann, 1994. Effects of spraying rate and of powder type on granule size growth in fluid-bed granulation. Eur. J. Pharm. Biopharm., 40, Suppl. 23S.

61. M.E. Aulton, M. Banks, D.K. Smith, 1977. The wettability of powders during fluidized bed granulation. J. Pharm. Pharmacol., 29, Suppl. 59P.

62. E. Räsänen, J. Rantanen, J.-P. Mannermaa, J. Yliruusi, 2004. The characterization of fluidization behavior using a novel multichamber microscale fluid bed. J. Pharm. Sci., 93, 780-791.

63. A.W. Nienow, P.N Rowe, L.Y.-L. Cheung, 1978. A quantitative analysis of the mixing of two segregating powders of different density in a gas-fluidized bed. Powder Technol., 20, 89-97.

64. W.L.Davies, W.T. Gloor, 1972. Batch production of pharmaceutical granulations in a fluidized bed II: Effects of various binders and their concentration on granulation and compressed tablets. J.Pharm.Sci., 61, 618-622.

65. Z. Ormós, K. Pataki, 1979. Studies on granulation in a fluidized bed VII. The effect of raw material upon the granule formation. Hung. J. Ind. Chem., 7, 89-104.

66. J.A.H. de Jong, 1991. Tablet properties as a function of the properties of granules made in a fluidized bed process. Powder Technol., 65, 293-303.

67. G.I. Taylor, 1964. Disintegration of Water Drops in an Electric Field. Proc. Roy. Soc. Lond. Math. Phys. Sci, 280, 383-397

68. J. Zeleny, 1914. The Electrical Discharge from Liquid Points, and a Hydrostatic Method of Measuring the Electric Intensity at Their Surfaces. Phys. Rev., 3, 69.

69. A.G. Bailey, 1986. The Theory and Practice of Electrostatic Spraying. Atomisation and Spray Technology, 2, 95-134.

70. L. Rayleigh, 1882. On the Equilibrium of Liquid Conducting Masses charged with Electricity. Phil. Mag., 14,184-186.

________________________________________________________________________

43

71. A. Jaworek, 2007. Micro- and nanoparticle production by electrospraying. Powder Technol., 176, 18-35.

72. X.M. Zeng, S. Jones, D. O'Leary, M. Phelan, J. Colledge, 2002. Delivery of formoterol from a novel multi-dose inhaler AirmaxTM. Resp. Med., 96, 397-403.

73. J.S. Kaerger, R. Price, 2004. Processing of Spherical Crystalline Particles via a Novel Solution Atomization and Crystallization by Sonication (SAXS) Technique. Pharm. Res., 21, 372-381.

74. S.W. Li, S.N. Jayasinghe, M.J. Edirisinghe, 2006. Aspirin particle formation by electric-field-assisted release of droplets. Chem. Eng. Sci., 61, 3091-3097.

75. M. Nyström, M. Murtomaa, J. Salonen, 2010. Fabrication and characterization of drug particles produced by electrospraying into reduced pressure. J. Electrostat., 68, 42-48.

76. A. Gomez, D. Bingham, L. de Juan, K. Tang, 1998. Production of protein nanoparticles by electrospray drying. J. Aerosol Sci., 29, 561-574.

77. A.G. Bailey, W. Balachandran, 1981. The disruption of electrically charged jets of viscous liquid. J. Electrostat., 10, 99-105.

78. A. Jaworek, A.T. Sobczyk, 2008. Electrospraying route to nanotechnology: An overview. J. Electrostat., 66,197-219.

79. R.A Coffee, P. Rivett, 1983. Electrodynamic Spraying and the Control of Pests of Agricultural Crops. EPPO Bulletin, 13, 443-448.

80. J.F. Kamphoefner, 1972. Ink-jet printing. I.E.E.E Trans.Electron. Dev., 19, 584-593.

81. J.F. Hughes, I.D. Pavey, 1981. Electrostatic emulsification. J. Electrostat., 10, 45-55.

82. R.C. Rowe, 2000. Small is beautiful. PSTT, 3, 187-188. 83. H.-L. Koh, W.-P. Yau, P.-S. Ong, A. Hegde, 2003. Current trends in modern

pharmaceutical analysis for drug discovery. Drug Discov. Today, 8, 889-897. 84. J. Aaltonen, K. Kogermann, C.J. Strachan, J. Rantanen, 2007. In-line monitoring

of solid-state transitions during fluidisation. Chem Eng Sci, 62, 408-415. 85. B. Van Eerdenbrugh, G. Van den Mooter, P. Augustijns, 2008. Top-down

production of drug nanocrystals: Nanosuspension stabilization, miniaturization and transformation into solid products. Int. J. Pharm., 364, 64-75.

86. A. Avdeef, K. Tsinman, O. Tsinman, N. Sun, D. Voloboy, 2009. Miniaturization of Powder Dissolution Measurement and Estimation of Particle Size. Chem. Biodiv., 6, 1796-1811.

87. N.M. Elman, Y. Patta, A.W. Scott, B. Masi, H.L. Ho Duc, M.J. Cima, 2009. The Next Generation of Drug-Delivery Microdevices. Clin. Pharmacol. Ther., 85, 544-547.

88. GEA Pharma Systems, brochure, Solid dosage R&D. 2009. 89. P. McKenzie, S. Kiang, J. Tom, A.E. Rubin, M. Futran, 2006. Can pharmaceutical

process development become high tech? AIChE J., 52, 3990-3994.

________________________________________________________________________

44

90. M. Andersson, A. Ringberg, C. Gustafsson, 2007. Multivariate methods in tablet formulation suitable for early drug development: Predictive models from a screening design of several linked responses. Chem. Int. Lab. Sys., 87, 125-130.

91. G.A. Lewis, D. Mathieu, R. Phan-Tan-Luu, Overview, In: Pharmaceutical experimental design. 1st ed., Marcel Dekker Inc., New York, USA. 1999, pp. 1-22.

92. J. Gabrielsson, N.-O. Lindberg, T. Lundstedt, 2002. Multivariate methods in pharmaceutical applications. J. Chemom., 16, 141-160.

93. L. Eriksson, E. Johansson, N. Kettaneh-Wold, J. Trygg, C. Wikström, S. Wold, Principal Component Analysis (PCA), In: Multi- and Megavariate Data Analysis Part I Basic Principles and Applications, Umetrics: Umeå, Sweden, 2006, pp. 39-62.

94. D. Pratiwi, J.P. Fawcett, K.C. Gordon, T. Rades, 2002. Quantitative analysis of polymorphic mixtures of ranitidine hydrochloride by Raman spectroscopy and principal components analysis. Eur. J. Pharm. Biopharm., 54, 337-341.

95. O. Svensson, M. Josefson, F.W. Langkilde, 2000. The synthesis of metoprolol monitored using Raman spectroscopy and chemometrics. Eur. J. Pharm. Sci., 11, 141-155.

96. F. Tian, H. Qu, M. Louhi-Kultanen, J. Rantanen, 2009. Mechanistic insight into the evaporative crystallization of two polymorphs of nitrofurantoin monohydrate. J. Crystal Growth, 311, 2580-2589.

97. M. Blanco, J. Coello, H. Iturriaga, S. Maspoch, C. Pérez-Maseda, 2000. Determination of polymorphic purity by near infrared spectrometry. Anal. Chim. Acta, 407, 247-254.

98. M. Savolainen, A. Heinz, C.J. Strachan, K.C. Gordon, J. Yliruusi, T. Rades, N. Sandler, 2007. Screening for differences in the amorphous state of indomethacin using multivariate visualization. Eur. J. Pharm. Sci., 30, 113-123.

99. N. Laitinen, J. Rantanen, S. Laine, O. Antikainen, E. Räsänen, S. Airaksinen, J. Yliruusi, 2002. Visualization of particle size and shape distributions using self-organizing maps. Chem. Int. Lab. Sys., 62, 47-60.

100. R.M. Maggio, P.M. Castellano, T.S. Kaufman, 2008. A new principal component analysis-based approach for testing "similarity" of drug dissolution profiles. Eur. J. Pharm. Sci., 34, 66-77.

101. A. Jørgensen, J. Rantanen, M. Karjalainen, L. Khriachtchev, E. Räsänen, J. Yliruusi, 2002. Hydrate Formation During Wet Granulation Studied by Spectroscopic Methods and Multivariate Analysis. Pharm. Res., 19, 1285-1291.

102. R. Bro, F. van den Berg, A. Thybo, C.M. Andersen, B.M. Jørgensen, H. Andersen, 2002. Multivariate data analysis as a tool in advanced quality monitoring in the food production chain. Trends Food Sci Tech , 13, 235-244.

103. S.K. Singh, S.K. Jha, A. Chaudhary, R.D.S. Yadava,S.B. Rai, 2010. Quality control of herbal medicines by using spectroscopic techniques and multivariate statistical analysis. Pharm. Biol., 48, 134-141.

104. D. Giron, 2003. Monitoring of Polymorphism - From Detection to Quantification. Eng. Life Sci., 3, 103-112.

________________________________________________________________________

45

105. Z. Német, G.C. Kis, G. Pokol, A. Demeter, 2009. Quantitative determination of famotidine polymorphs: X-ray powder diffractometric and Raman spectrometric study. J. Pharm. Biomed. Anal., 49, 338-346.

106. N. Chieng, S. Rehder, D. Saville, T. Rades, J. Aaltonen, 2009. Quantitative solid-state analysis of three solid forms of ranitidine hydrochloride in ternary mixtures using Raman spectroscopy and X-ray powder diffraction. J. Pharm. Biomed. Anal., 49, 18-25.

107. S. Mazurek, R. Szostak, 2009. Quantification of atorvastatin calcium in tablets by FT-Raman spectroscopy. J. Pharm. Biomed. Anal., 49, 168-172.

108. M.C. Sarraguça, A.V. Cruz, S.O. Soares, H.R. Amaral, P.C. Costa, J.A. Lopes, 2010. Determination of flow properties of pharmaceutical powders by near infrared spectroscopy. J. Pharm. Biomed. Anal., 52, 484-492.

109. P. Luukkonen, M. Fransson, I. Niklasson-Björn, J. Hautala, B. Lagerholm, S. Folestad, 2008. Real-Time assessment of granule and tablet properties using in-line data from a high-shear granulation process. J. Pharm. Sci., 97, 950-959.

110. K. Naelapää, P. Veski, J.G. Pedersen, D. Anov, P. Jørgensen, H.G. Kristensen, P. Bertelsen, 2007. Acoustic monitoring of a fluidized bed coating process. Int. J. Pharm., 332, 90-97.

111. P.J. Toscas, F.D. Shaw, S.L. Beilken, 1999. Partial least squares (PLS) regression for the analysis of instrument measurements and sensory meat quality data. Meat Sci., 52, 173-178.

112. J.K. Holland, E.K. Kemsley, R.H. Wilson, 1998. Use of Fourier transform infrared spectroscopy and partial least squares regression for the detection of adulteration of strawberry purées. J. Sci. Food Agric., 76, 263-269.

113. E. Räsänen, J. Rantanen, J.-P. Mannermaa, J. Yliruusi, H. Vuorela, 2003. Dehydration studies using a novel multichamber microscale fluid bed dryer with in-line near-infrared measurement. J. Pharm. Sci., 2003. 92, 2074-2081.

114. E. Räsänen, O. Antikainen, J. Yliruusi, 2003. A New Method to Predict Flowability Using a Microscale Fluid Bed. AAPS PharmSciTech, 4, 418-414.

115. S. Airaksinen, M. Karjalainen, E. Räsänen, J. Rantanen, J. Yliruusi, 2004. Comparison of the effects of two drying methods on polymorphism of theophylline. Int. J. Pharm., 276, 129-141.

116. M. Römer, J. Heinämäki, I. Miroshnyk, N. Kivikero, N. Sandler, J. Rantanen, J. Yliruusi, 2008. Phase transformation of erythromycin A dihydrate during fluid bed drying. J. Pharm. Sci., 97, 4020-4029.

117. M. Fernández-Cervera, J. Heinämäki, E. Räsänen, O. Antikainen, O. M. Nieto, A. Iraizoz-Colarte, J. Yliruusi, 2004. Determination of tackiness of chitosan film-coated pellets exploiting minimum fluidization velocity. Int. J. Pharm., 281, 119-127.

118. J. Larjo, D.L. Paisley, S. Kleinfelder, D.R. Snyder, B.J Thompson, 2005. High-power diode lasers in spray process diagnostics. SPIE, 5580, 455-465.

119. M. Meshali, H.M. El-Banna, H. El-Sabbagh, 1983. Use of a fractional factorial design to evaluate granulations prepared in a fluidized bed. Pharmazie, 38 323-325.

________________________________________________________________________

46

120. K. Tang, A. Gomez, 1994. Generation by electrospray of monodisperse water droplets for targeted drug delivery by inhalation. J. Aerosol Sci., 25, 1237-1249.

121. O. Lastow, W. Balachandran, 2007. Novel low voltage EHD spray nozzle for atomization of water in the cone jet mode. J. Electrostat., 65, 490-499.