university of groningen a novel bottom-up process to prepare drug nanocrystals … ·...

University of Groningen

A novel bottom-up process to prepare drug nanocrystalsWaard, Hans de

IMPORTANT NOTE: You are advised to consult the publisher's version (publisher's PDF) if you wish to cite fromit. Please check the document version below.

Document VersionPublisher's PDF, also known as Version of record

Publication date:2011

Link to publication in University of Groningen/UMCG research database

Citation for published version (APA):Waard, H. D. (2011). A novel bottom-up process to prepare drug nanocrystals: the art of the soluble.Groningen: s.n.

CopyrightOther than for strictly personal use, it is not permitted to download or to forward/distribute the text or part of it without the consent of theauthor(s) and/or copyright holder(s), unless the work is under an open content license (like Creative Commons).

Take-down policyIf you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediatelyand investigate your claim.

Downloaded from the University of Groningen/UMCG research database (Pure): http://www.rug.nl/research/portal. For technical reasons thenumber of authors shown on this cover page is limited to 10 maximum.

Download date: 06-07-2020

https://www.rug.nl/research/portal/en/publications/a-novel-bottomup-process-to-prepare-drug-nanocrystals(38342350-e7e4-4168-a0fa-3595d959d167).html

https://www.rug.nl/research/portal/en/publications/a-novel-bottomup-process-to-prepare-drug-nanocrystals(38342350-e7e4-4168-a0fa-3595d959d167).html

A novel bottom-up process to prepare drug nanocrystals

The art of the soluble

Hans de Waard

2

Paranimfen: Bart Anneveld Wouter Tonnis

Thesis subtitle“The art of the soluble” - Peter B. Medawar, 1967.

Printing of this thesis was supported by generous contributions from:University of GroningenFaculty of Mathematics and Natural Sciences of the University of GroningenSP Scientific

© Copyright 2011, H. de WaardAll rights are reserved. No part of this thesis may be reproduced, stored in a retrieval system or transmitted in any form or by any means, mechanically, by photocopying, recording or otherwise, without the written permission of the author.

Cover design: Christine van Gemert and Hans de WaardLay-out design: Hans de WaardPrinted by: Off Page - Amsterdam

ColofonThe research presented in this thesis was performed within the framework of project T5-105 of the Dutch Top Institute Pharma.

RIjkSUNIveRSITeIT GRONINGeN

A novel bottom-up process to prepare drug nanocrystals

The art of the soluble

Proefschrift

ter verkrijging van het doctoraat in deWiskunde en Natuurwetenschappenaan de Rijksuniversiteit Groningen

op gezag van deRector Magnificus, dr. E. Sterken,in het openbaar te verdedigen op

vrijdag 11 maart 2011om 14.45 uur

door

Hans de Waard

geboren op 14 augustus 1982te Oss

Promotor: Prof. dr. H.W. Frijlink

Copromotor: Dr. W.L.j. Hinrichs

Beoordelingscommissie: Prof. dr. j. Anwar Prof. dr. W.e. Hennink Prof. dr. S.C. De Smedt

ISBN: 978-90-367-4695-3 (printed version) 978-90-367-4696-0 (electronic version)

Fortuna est quae fit cum praeparatio in occasionem inciditLuck is what happens when preparation meets opportunity Seneca (ca. 3 BC-65 AD)

7

Table of contents

Page

Chapter 1 Introduction 9

Chapter 2 Controlled crystallization during freeze-drying. J. Control. Release 128 (2008) 179-183.

15

Chapter 3 Application of a 3-way nozzle to prevent premature crystallization for large scale production. Eur. J. Pharm. Sci. 38 (2009) 224-229.

29

Chapter 4 elucidation of the mechanism by in-line Raman spectroscopy. AAPS J. 12 (2010) 569-575.

43

Chapter 5 CLSM as quantitative method to determine the size of drug crystals in a solid dispersion. Submitted for publication (2011).

57

Chapter 6 Evaluation and optimization of a force field for the crystalline forms of mannitol and sorbitol. J. Phys. Chem. B 114 (2010) 429-436.

71

Chapter 7 Concluding remarks and perspectivesPharm. Res. (2010) in press.

89

Chapter 8 Summary 97

appendix a Samenvatting 103

appendix B References 109

appendix C Nomenclature 117

appendix d Curriculum vitae & List of publications 119

appendix e Dankwoord 123

table of contents

chapter1introduction

chapter 1

10

Introduction

As a result of modern drug discovery techniques, the emerging drug candidates increasingly tend to be highly lipophillic [1]. These drugs can be categorized as class II drugs according to the biopharmaceutics classification system. These drugs are poorly soluble in water, but once dissolved they are easily absorbed over the gastro-intestinal membrane [2]. Therefore, dissolution is the rate limiting step for the bioavailability of these drugs after oral administration [3].

There are many strategies to increase the dissolution rate of poorly soluble drugs. Classical approaches include, amongst others, particle size reduction, pH adjustment, salt formation, the use of co-solvents, complexation, addition of surfactants, formulation as emulsion or other lipid based systems, or entrapment in liposomes [4]. However, many of these techniques are limited to specific drugs or formulations. pH-adjustment is only effective for ionizable drugs, salt formation can only be applied to ionizable drugs, co-solvents are often needed in too large amounts, the use of complexing agents such as cyclodextrins is limited to molecules that fit in the cyclodextrin conus, large amounts of surfactants are not well tolerated, and the entrapment capacity of liposomes is limited.

Another frequently described approach is the application of solid dispersions. Solid dispersions consist of a hydrophilic matrix in which the lipophilic drug is dispersed. The drug can be dispersed at the molecular level (solid solution) or as small particles. According to the Ostwald-Freundlich equation [5], the saturation concentration at the surface of such small particles is larger than that of large particles. Furthermore, by decreasing the particle size, the total surface area available for dissolution increases (Noyes-Whitney). Both effects result in an increase of the dissolution rate [6]. In addition, when the drug is formulated as an amorphous solid dispersion, the drug is in a higher energy state than a crystalline drug, which further increases the dissolution rate [7].

However, since molecules in the amorphous state are thermodynamically less stable than molecules in the crystalline state, uncontrolled crystallization upon storage may occur [8]. This uncontrolled crystallization may lead to the formation of large crystals that dissolve slowly. Uncontrolled crystallization of the hydrophilic matrix can be prevented by choosing a matrix with a high Tg, but the Tg of a given drug can not be changed. Because the risk of uncontrolled crystallization is especially high for low Tg drugs, for drugs with a higher Tg the application of amorphous solid dispersion technology can be a successful formulation approach. In a previous study for example, amorphous solid dispersions were prepared by freeze-drying a solution of drug in TBA mixed with a solution of a matrix material in water [9]. To obtain an amorphous solid dispersion, the temperature during the primary drying was kept well below the Tg’.

Since low Tg drugs may be physically unstable when they are formulated as amorphous solid dispersions, it is better to formulate them in a most stable state: the crystalline state. However, to obtain still a drug product with the desired fast dissolving dissolution rate, the size of the drug crystals should be in the nanometer range (i.e. < 1 μm). Methods to prepare drug nanocrystals

11

introduction

can be divided into bottom-up and top-down methods. Typical bottom-up methods are precipitation techniques and supercritical fluid technologies. Disadvantages of these techniques are, amongst others, the difficulty to control the drug crystal size and limited solubility of the drug in the solvent. Typical top-down methods are high pressure homogenization and wet ball milling. Disadvantages of these methods are amongst others, the use of large amounts of surfactants, low yields, high energy input, and possible contamination from grinding media.

The aim of this thesis was to develop a novel bottom-up process based on freeze-drying to prepare drug nanocrystals, that overcomes some of the above mentioned disadvantages. Furthermore, other aspects related to the development of this process are discussed in this thesis. The possibility to use the process for large scale production, a PAT tool to gain insight in the crystallization process, a novel tool to analyze the end product, and molecular modeling of the used matrix material to be able to increase the understanding of the physical stability of drug nanocrystals embedded in this matrix material.

In Chapter 2, the development of the novel process to prepare drug nanocrystals is described. Usually freeze-drying is performed well below the Tg’ of the freeze-concentrated fraction, which yields amorphous solid dispersions. However, during this novel process, the temperature was increased to a temperature above the Tg’, but below the Te. At this elevated temperature, the freeze-concentrated fraction is in the rubbery state and the mobility of the molecules is relatively high, this enables drug crystallization. It was tested whether for a typical lipophilic low Tg drug (fenofibrate) indeed a nanocrystalline dispersion could be prepared by freeze-drying at a relatively high temperature (between Tg’ and Te). The possibility to use this method for highly dosed drugs was studied by increasing the drug load in the crystalline dispersion. Secondly, the influence of two process parameters, freezing rate and the ratio water to TBA, on the drug crystal size and consequently the dissolution rate were investigated. The size and morphology of the crystals were determined by SeM.

Since the mixture of drug, matrix, tertiary butyl alcohol (TBA), and water is thermodynamically unstable, it should be frozen immediately (before phase separation occurs) and fast after mixing. Therefore, during the development of the process (chapter 2), only small quantities were mixed in glass vials and frozen in liquid nitrogen. However, since this process is not suitable for large scale production, in was investigated in Chapter 3 whether a 3-way nozzle can be used to change the small scale freeze-drying process into a semi-continuous process that is suitable for scale-up. The 3-way nozzle consists of two channels through which the aqueous- and TBA-solution flow separately. The two solutions were then mixed just outside the nozzle by the atomizing airflow from the third channel. Immediately after mixing, the solutions were frozen by spraying the droplets into liquid nitrogen. To investigate whether this nozzle could be used for this process, first the mixing quality of the nozzle was tested. Subsequently, crystalline dispersions of fenofibrate in mannitol were prepared by the freeze-drying and spray freeze-drying process. To test whether spray freeze-drying could be used to scale-up the process, the crystallinity and the dissolution of the drug from the solid dispersions prepared by both methods were compared to each other.

chapter 1

12

Although the influence of several process parameters is described in chapter 2 and chapter 3, it was not completely clear at which stage during the process the different components crystallized and which mechanism determined the drug crystal size. Hereto Chapter 4 describes in-line Raman spectroscopy as PAT tool to investigate at which stage during the freeze-drying process drug, matrix material, and solvents crystallized. A Raman probe was placed immediately above the sample in the freeze-dryer. By using this in-line tool, crystallization of each of the components (fenofibrate, mannitol, TBA, and water) can be monitored. Furthermore, two process conditions, freezing rate and temperature of the freeze-dryer shelf were varied. Monitoring the crystallization process was used to study the moment and length of the crystallization of the different components, while varying the process conditions was used to elucidate the mechanism that determines the drug crystal size.

Scanning electron Microscopy (SeM) pictures in chapter 2 and chapter 3 showed that the individual particles after freeze-drying were smaller than 1 μm, implying that the size of the drug crystals in these particles was of nanoscale. However, the exact size could not be determined, because the drug and the matrix could not be distinguished. Therefore it was studied in Chapter 5 whether Confocal Laser Scanning Microscopy (CLSM) in combination with image analysis could be used as novel analytical tool to determine the drug crystal size in the end product. A prerequisite for using this analytical tool is that the drug is autofluorescent. The drug dipyridamole meets this requirement and was therefore used as model drug in this study. To validate whether CLSM could be used to determine the particle size of the drug while the drug was mixed with a second component, the pure drug crystal size as determined by laser diffraction was compared with the dipyridamole crystal size in a physical mixture as determined by CLSM. To test whether CLSM could be used to determine the size of drug crystals in a solid dispersion, the size of dipyridamole in different solid dispersions was determined by CLSM and the obtained results were related to the dissolution behavior of dipyridamole from these solid dispersions.

A molecular modeling study in which the force field parameters of the matrix material were evaluated and optimized is described in Chapter 6. The optimisation of the forcefield parameters is the first step to simulating nanocrystals embedded in mannitol and sorbitol. In this chapter two force fields, the GROMOS and the AMBER force field parameter sets were evaluated for their ability to reproduce the polymorphs of mannitol and sorbitol. each of the polymorphs was simulated at 10 k. Secondly, the parameter sets were systematically optimized using sensitivity analysis. Finally, the ability of both the original and the optimized parameter sets to reproduce the crystal structures at room temperature was tested in extended molecular dynamics simulations at 298 k.

In Chapter 7, a historical review of bottom-up preparation techniques for drug nanocrystals is given. This review was used to discuss why only a limited number of drug products prepared by bottom-up techniques reached the market.

13

introduction

chapter2controlled crystallization during freeze-drying

16

chapter 2

2.1. Abstract

To improve the dissolution behavior of lipophilic drugs, a novel bottom-up process based upon freeze drying which allows for the production of nanocrystalline particles was developed: “controlled crystallization during freeze drying”. This novel process could strongly increase the dissolution behavior of fenofibrate. For example at a drug load of 30% w/w, 80% of the drug dissolved within 10 minutes from tablets prepared from the controlled crystallized dispersions, while from tablets prepared from the physical mixture only 50% was dissolved after 120 minutes. Furthermore it was found that faster freezing or using a solution with a lower water/TBA ratio resulted in faster dissolution, indicating that the crystalline dispersions contained smaller crystals. Crystallization of the drug could occur during freezing or during drying. When crystallization occurs during freezing, faster freezing or using solutions with a lower water/TBA ratio results in the formation of more nuclei and consequently smaller crystals. When crystallization occurs during drying, faster freezing or using solutions with a higher water/TBA ratio results in the formation of smaller solvent crystals and therefore smaller interstitial spaces which contain the freeze-concentrated fraction. Since crystallization occurs in the freeze-concentrated fraction and the size of the crystals are limited to the size of the interstitial spaces, smaller crystals are formed in these situations.

H. de Waard, W.L.j. Hinrichs, and H.W. Frijlink

Department of Pharmaceutical Technology and Biopharmacy, University of Groningen, Antonius Deusinglaan 1, 9713 AV, Groningen, the Netherlands.

Published as “A novel bottom-up process to produce drug nanocrystals: controlled crystallization during freeze-drying” in: j. Control Release 2008. 128(2), 179-183.

17

controlled crystallization during freeze-drying

2.2. Introduction

Many new drugs can be considered to be class II drugs according to the Biopharmaceutics Classification System [1]. These drugs are poorly water soluble, but once they are dissolved, they are easily absorbed over the gastro-intestinal membrane [2; 10]. Therefore the bioavailability after oral administration can be improved by enhancement of the dissolution rate [3].

One of the approaches to enhance the dissolution rate is the application of amorphous solid dispersions [7; 11]. Such solid dispersions are composed of a hydrophilic matrix in which a poorly soluble drug is dispersed [12]. Application of these fully amorphous solid dispersions is theoretically an ideal method to improve the dissolution rate, because the saturation concentration as well as the surface area available for dissolution increases [7]. However, molecules in the amorphous state are thermodynamically unstable relative to the crystalline state. Therefore uncontrolled crystallization of the drug or matrix material could occur during processing or storage of the amorphous solid dispersion [8; 13-15]. This crystallization of the drug is unwanted, because it may affect the dissolution behavior. In many cases crystallization of the carrier can be prevented by choosing a carrier with a high glass transition temperature (Tg), but for a given drug the Tg cannot be changed. When a drug has a low Tg the risk of uncontrolled crystallization is high, in particular when the drug is incorporated in the solid dispersion as clusters [16]. An example of such a drug is fenofibrate, which has a Tg of -21.3 °C [17].

The undesired crystallization can be prevented by using a system which is already in its most stable state: the crystalline state. However, the drug crystals have to be sufficiently small (i.e. nanoscale) to obtain a large surface area as well as an increased saturation concentration (kelvin law) and therefore high dissolution rate [5-6]. Processes to produce nanocrystals can be categorized as top-down and bottom-up processes. Typical top-down processes are high pressure homogenization [18] and wet ball milling [19]. Disadvantages of these processes are the use of surfactants, the long processing times, the difficulty in achieving a uniform size distribution, low yields, high energy input and possible contamination from the grinding media [20-22]. Bottom-up processes are basically precipitation processes. A disadvantage of most currently applied bottom-up processes is that the final drug crystal size cannot be controlled adequately [22-23].

To overcome these disadvantages, we propose a novel bottom-up process to produce drug nanocrystals. This process is based on a technique used to produce amorphous solid dispersions developed within our laboratory [9]. These amorphous solid dispersions were prepared by freeze-drying a solution of drug and sugar in a mixture of water and TBA. In contrast to this previous process, we now present a new process that produces fully crystalline solid dispersions. By taking a drug with a low Tg (fenofibrate), a carrier (mannitol) that easily crystallizes during freeze-drying, and freeze-drying at a relatively high shelf temperature, we envisage that crystallization can occur either during freezing or during drying. The freeze-drying conditions (freezing rate and composition of the solution to be freeze-dried) were manipulated to evaluate whether crystalline dispersions could be obtained and whether the particle size and therefore the dissolution rate can be controlled.

18

chapter 2

2.3. Materials

Fenofibrate and TBA were obtained from Sigma-Aldrich Chemie B.V. Zwijndrecht, the Netherlands. Mannitol was provided by Roquette, France. Demineralized water was used in all experiments.

2.4. Methods

2.4.1. Preparation of the crystalline solid dispersionTwo separate solutions were prepared: one of mannitol in water and the other of fenofibrate in TBA (for compositions see Table 2-I). After heating the solutions to about 60 °C, the aqueous solution was mixed with the TBA solution in 10 mL glass injection vials. In all cases (except for the drip and freeze process), the total liquid volume in the vial was 4 mL. Immediately after mixing, the solution was frozen (for differences in freezing process see Table 2-II) and subsequently lyophilized.

Table 2.I. Composition of the different solutions to produce the crystalline solid dispersions.

Before lyophilization After lyophilization

Cmannitol/water

(mg/mL)

Cfenofibrate/TBA

(mg/mL)

Ratio water/TBA

(v/v)

Drug load

(% w/w)

150 25 6/4 10

67 25 6/4 20

29 50 8/2 30

33 33 7/3 30

39 25 6/4 30

78 14 3/7 30

25 25 6/4 40

Table 2.II. Description of the processes to freeze the water/TBA solution

Process Description

0.5 ºC/min Vial placed on a plate of 20 ºC and cooled to -25 ºC at 0.5 ºC/min

Pre-cooled plate vial placed on a precooled plate of -25 ºC

Liquid nitrogen vial immersed in liquid nitrogen

Drip and freeze Mixture of the solution was dripped in liquid nitrogen by using a pipette

All solutions had a water/TBA ratio of 6/4. The drug load in the obtained dispersions was 30% w/w.

Freeze-drying was performed using a Christ model epsilon 2-4 lyophilizer (Salm en kipp, Breukelen, The Netherlands). The frozen solutions were lyophilized at a shelf temperature of -25 °C, a condensor temperature of -85 °C, and a pressure of 0.220 mBar for one day. Thereafter, the shelf temperature was gradually raised to 25 °C and pressure was gradually decreased to 0.05 mBar over another day. Physical mixtures were prepared using a spatula and a mortar.

19


2.4.2. Differential Scanning CalorimetryDSC was used to determine the degree of crystallinity of the fenofibrate and mannitol in the crystallized dispersions. This was defined as the ratio between the heat of fusion of the drug or carrier in the solid dispersion and the heat of fusion of the drug or carrier as received multiplied by fraction drug or carrier in the mixture. It was assumed that the pure fenofibrate and mannitol were completely crystalline. The samples (4-10 mg) were placed in a hermetically closed aluminum pan. The heating rate of the scanning calorimeter (Q2000, TA Instruments, Ghent, Belgium) was 2 °C/min, from -50 to 200 °C.

2.4.3. X-Ray Powder DiffractionXRPD was performed using CuKα radiation with a wavelength of 1.5405 Å at 40 kV and 40 mA from an X’Pert PRO MPD diffractometer (PANalytical, Almelo, the Netherlands). Samples were scanned from 9-25 °2θ with a step size of 0.004° and a time per step of 110 s. The sample powders were placed on a zero-background silicon holder. 2.4.4. Scanning Electron MicroscopyScanning electron micrographs (SeM) were recorded with a jeOL jSM 6301-F Microscope (jeOL, japan). The powder was dispersed on top of double-sided sticky carbon tape on metal disks and coated with 80 nm of gold/palladium in a Balzers 120B sputtering device (Balzers UNION, Liechtenstein).

2.4.5. TablettingAll formulations were compressed to 9 mm round and flat tablets having a weight of approximately 100 mg. The maximum compaction load was 5 kN, and the speed of compaction was 5 kN/s. All tablets were prepared on an eSH compaction apparatus (Hydro Mooi, Appingedam, The Netherlands), and the die was lubricated with magnesium stearate. The tablets were stored in a vacuum desiccator over silica gel at room temperature for at least one day before further processing.

2.4.6. DissolutionThe dissolution behavior of the tablets was tested by using a USP dissolution apparatus II (Rowa Techniek, Leiderdorp, the Netherlands) with a paddle speed of 100 rpm. The dissolution medium was 1000 ml of a 0.5% w/v sodium dodecyl sulphate in demineralized water solution at 37 °C. The concentration of fenofibrate was measured spectrophotometrically (UV-VIS spectrophotometer Uv-1601, Shimadzu) at a wavelength of 290 nm.

2.5. Results

2.5.1. Drug loadThe degree of crystallinity of the dispersions (prepared from solutions at a water/TBA ratio of 6/4; vials frozen in liquid nitrogen, as described in table 2-II) was determined by XRPD and DSC. Crystalline mannitol can have three anhydrous polymorphic forms (α, β, and δ mannitol).

20

chapter 2

Fig. 2.1. A) Typical examples of X-ray diffraction patterns of fenofibrate as received, α, β, and δ–mannitol,

and a controlled crystallized dispersion of 30% w/w fenofibrate in mannitol (patterns of the mannitol

polymorphs were taken from ICDD-library). B) Typical examples of DSC-thermograms of fenofibrate

as received (solid line), pure mannitol (dotted line) and a controlled crystallized dispersion of 30% w/w

fenofibrate in mannitol (dashed line).

21


The melting points (166.5 °C, 166 °C , and 155 °C respectively) and melting enthalpies (52.1 kJ/mol, 53.5 kJ/mol, and 53.7 kJ/mol respectively) of these polymorphic forms differ only slightly [24]. Therefore XRPD was used to determine which polymorph was formed during freeze-drying, while DSC was used to quantify the fraction fenofibrate and mannitol which was crystalline. Despite the small differences between the polymorphic forms, the degree of crystallinity of mannitol was calculated by using the melting enthalpy of the mannitol as received since the largest difference in melting enthalpy between the different polymorphic forms is only 3%.

Fig. 2.1A shows typical examples of XRPD-diffractograms of fenofibrate as received, α, β, and δ-mannitol (the diffractograms of mannitol were taken from ICDD-library) and a dispersion containing 30% w/w fenofibrate in mannitol (prepared from a solution at a water/TBA ratio of 6/4 [Table 2-I] with the vial being frozen in liquid nitrogen [Table 2-II]). The last diffractogram shows that a dispersion consisting of crystalline fenofibrate and δ-mannitol was obtained. As can be seen in the diffractogram of the crystallized dispersion (Fig. 2.1A), δ-mannitol was obtained after freeze-drying.

The DSC-thermograms were in agreement with the XRPD-diffractograms and showed clearly the melting of crystalline fenofibrate and mannitol (Fig. 2.1B). At the tested drug loads (10-40%

Fig. 2.2. Comparison between tablets composed of physical mixtures (open symbols) and controlled

crystallized dispersions (closed symbols) of different amounts of fenofibrate (10 [◇ ,◆ ], 20 [△ ,▲ ], 30

[○ ,●] and 40 [□ ,■ ] % w/w) in mannitol. All solutions had a water/TBA ratio of 6/4 and were frozen in

liquid nitrogen. (n=6; mean).

22

chapter 2

w/w), the samples were all substantially crystalline. The degree of crystallinity of fenofibrate was 86-91%, and the degree of crystallinity of mannitol 77-87% (details not shown). However, the DSC-thermograms did not show a glass transition, which may imply that the remaining non-crystalline fraction is in the rigid amorphous state [25-27].

The dissolution behavior of tablets composed of these crystalline dispersions was compared with the dissolution behavior of the corresponding physical mixtures (Fig. 2.2; previous page). The controlled crystallized formulations showed a much higher dissolution rate than the corresponding physical mixtures. Furthermore, the dissolution rate increased when the drug load was decreased. This could be caused by the presence of smaller drug crystals in the dispersions which have a lower drug load, or the smaller amount of drug to be dissolved.

Since the DSC results show that the solid dispersions were highly crystalline, the improved dissolution behavior of the controlled crystallized dispersions compared to the corresponding physical mixtures must result from a difference of the drug crystal size. Therefore it could be concluded that drug crystals were formed which were much smaller than the initial crystals (x50 = 13 μm; determined by laser diffraction) in the physical mixture.

Fig. 2.3. Dissolution profiles of tablets composed of physical mixtures (○) and controlled crystallized

dispersions of 30% w/w fenofibrate in mannitol. The controlled crystallized dispersions were prepared at

different freezing rates: 0.5 °C/min (◆), on a pre-cooled plate (▲ ), in liquid nitrogen (●), and by dripping

the solution in liquid nitrogen (■ ). All solutions had a water/TBA ratio of 6/4. (n=6; mean).

23


Fig. 2.4. Scanning electron micrographs of crystallized dispersions prepared at different freezing rates.

All dispersions contain 30% w/w fenofibrate in mannitol. (Magnification left column 1500x; magnification

right column 10000x).

24

chapter 2

2.5.2. Freezing rateWater/TBA (ratio 6/4) solutions were frozen at different rates to investigate whether differences in freezing influences the drug crystal size and thereby the dissolution behavior. The drug load used was 30% w/w in all cases. At all freezing rates samples were highly crystalline: 81-91% for fenofibrate and 79-94% for mannitol. Also in these samples a Tg was absent. In Fig. 2.3 the results of the dissolution experiments of tablets consisting of dispersions produced using different freezing rates are shown. When the solutions were frozen slowly (at 0.5 °C/min or on a precooled shelf), the dissolution rate was low. However, when the freezing rate was increased by freezing the vials into liquid nitrogen or dripping the water/TBA solution in liquid nitrogen, the dissolution rate was much higher. When the SeM pictures (Fig. 2.4) of dispersions prepared at different freezing rates are compared, it is clear that a lower freezing rate results in coarser particles. Since all samples are highly crystalline and they do not differ substantially in their degree of crystallinity, changes in crystal size should be the cause of the changes in dissolution rate. Therefore it is assumed that a change in freezing rate influences the drug crystal size. Furthermore, the pictures show that the size of the particles found upon rapid freezing was below 1 μm, implying that the size of the drug crystals in these particles should be of nanoscale. However, the exact size of the drug crystals can not be determined from Fig. 2.4 since the particles observed consist besides the drug also of mannitol.

Table 2.III. Degree of crystallinity of crystallized dispersions pepared from different ratios water/TBA.

Ratio water/TBA

(v/v)

Degree of crystallinity (%)

Fenofibrate Mannitol

8-2 67 ± 7 96 ± 2

7-3 80 ± 4 91 ± 2

6-4 91 ± 1 84 ± 2

3-7 88 ± 1 91 ± 0

The drug load in the obtained dispersions was 30% w/w. (n=3-6; mean ± SD).

Fig. 2.5 shows the dissolution behavior of tablets composed of crystallized dispersions prepared from solutions containing different water/TBA ratios. All tablets contained an equal drug load of 30% w/w. Nevertheless, large differences in dissolution rate were found. When the water/TBA ratio of the solution to be freeze dried was relatively high (8/2 v/v), the dissolution rate was very low. When the water/TBA ratio was changed to 7/3 v/v, the dissolution rate strongly increased, but still not satisfactory. However, when the water/TBA ratio of the solution was further decreased to 6/4 v/v the dissolution rate was very high. A further reduction of the water/TBA ratio of the solution to 3/7 v/v did not affect the dissolution behavior.

2.6. Discussion

In this study we present a novel process to prepare drug nanocrystals. These nanocrystals were prepared by freeze-drying a solution of a drug and a sugar in a mixture of water and TBA. A drug with a low Tg (fenofibrate) and a carrier (mannitol) which easily crystallizes were used. In

25


addition, freeze drying was performed at a relatively high temperature. With this novel process, highly crystalline dispersions could be prepared which showed a strongly improved dissolution behavior compared to the physical mixtures. For example at a drug load of 30% w/w fenofibrate in mannitol, 80% of the drug dissolved within 10 minutes from tablets prepared from the crystalline dispersion, while only 50% was dissolved after 120 minutes from tablets prepared from the corresponding physical mixture. Therefore it can be concluded that the size of the drug crystals in the dispersion must be much smaller than the size of the initial drug crystals. This crystal size could be controlled by changing the process conditions. Two process conditions were evaluated: freezing rate and water/TBA ratio of the solutions to be freeze dried. When the freezing rate was increased, the crystal size was decreased. And when the water/TBA ratio was decreased the crystal size was also decreased. SeM pictures showed that the size of the particles obtained after rapid freezing was below 1 μm. This implied that the size of the drug crystals in these particles was of nanoscale.

Crystallization of fenofibrate and mannitol in these solutions can occur during different stages of the freeze-drying process. It could occur during freezing, during drying or during a combination of both. Therefore we propose two mechanisms to explain the differences in crystal size when the freezing rate or the water/TBA ratio was changed. First, when crystallization occurs during freezing, faster freezing or a smaller water content of the solution results in a higher nucleation

Fig. 2.5. Dissolution profiles of tablets composed of physical mixtures (○) and controlled crystallized

dispersions of 30% w/w fenofibrate in mannitol. The controlled crystallized dispersions were prepared with

different ratios water/TBA (v/v): 8/2 (■ ), 7/3 (▲ ), 6/4 (●), and 3/7 (◆). All vials were frozen in liquid

nitrogen. (n=3-6; mean).

26

chapter 2

rate. Consequently more nuclei are formed resulting in more crystals, which are consequently smaller.

On the other hand, crystallization during drying could also occur when the temperature is above the Tg’ and under the Te. In this situation, the freeze-concentrated fraction is in its rubbery state and crystallization can occur. The growth of the crystals is limited to the size of the interstitial spaces between the solvent crystals. We hypothesize that the size of the interstitial spaces between the solvent crystals containing the freeze-concentrated fraction can be adjusted by changing the freezing rate or the water/TBA ratio: when the solution is frozen slowly (or the water/TBA ratio in the solution is high), large solvent crystals will be formed, resulting in large interstitial spaces. However, small solvent crystals will be formed when the solution is frozen quickly (or the amount of water in the solution is low), resulting in small interstitial spaces. Since the crystallization occurs in the freeze-concentrated fraction, the size of the interstitial spaces between the solvent crystals will limit the final size of the drug crystals.

In conclusion, the freezing rate as well as the water/TBA ratio of the solutions before freeze drying can be used to control the crystallization of the drug. When the solutions were frozen rapidly or when the fraction water was decreased, the dissolution rate was strongly increased. Whether this crystallization process occurs during freezing or during drying will be studied in further research by in-line Raman spectroscopy [Chapter 4]. Another challenge could be the application of this process to drugs which are more difficult to crystallize than fenofibrate. As mentioned earlier, this process was especially designed for drugs which have a low Tg. For drugs with a higher Tg an amorphous solid dispersion could be used to improve the dissolution rate. However, since a crystalline drug is more stable and therefore often more favorable, in future studies also this process will also be applied to drugs which have a higher Tg.

27


application of a 3-way nozzle to prevent premature crystallization for large scale production

chapter3

30

chapter 3

H. de Waarda, N. Grasmeijera, W.L.j. Hinrichsa, A.C. eissensa, P.P.F. Pfaffenbachb, and H.W. Frijlinka

a Department of Pharmaceutical Technology and Biopharmacy, University of Groningen, Antonius Deusinglaan 1, 9713 AV, Groningen, the Netherlands. b Infra Bad Hönningen strategical-technical chamical analysis, Solvay Infra Bad Hönningen GmbH, Hans-Böckler-Allee 20, 30173, Hannover, Germany.

Published as “Preparation of drug nanocrystals by controlled crystallization: application of a 3-way nozzle to prevent premature crystallization for large scale production” in: eur. j. Pharm. Sci. 2009. 38(3), 224-229.

3.1. Abstract

In a previous study we have developed a novel process to produce drug nanocrystals. This process, “controlled crystallization during freeze-drying” has shown to be a successful method to increase the dissolution rate of poorly water soluble drugs [Chapter 2]. This process consisted of two steps: a solution of a matrix material (mannitol) in water was mixed with a solution of a drug (fenofibrate) in TBA. This mixture was frozen and subsequently freeze-dried at relatively high temperature (-25 °C). Since the solution of matrix and drug in the water-TBA mixture is thermodynamically unstable, it had to be frozen immediately and fast after preparation to prevent premature crystallization of the drug resulting in the formation too large drug crystals. Therefore, small quantities were manually mixed in a vial and this vial was immersed in liquid nitrogen. To make this process ready for large scale production, the modification of this batch process to a semi-continuous process by the application of a 3-way nozzle was studied. With this nozzle, the aqueous and TBA-solutions were pumped into the nozzle via two separate channels and mixed just at the moment they left the nozzle. Thorough mixing was facilitated by the atomizing air, supplied via the third channel. Since the mixture was sprayed immediately into liquid nitrogen, premature crystallization was prevented. A further advantage was that the atomizing air generated small droplets which were directly immersed into liquid nitrogen. Consequently, the mixture was frozen even faster than in the batch process. This resulted in a reduced size of the drug crystals and hence a higher dissolution rate. Therefore, using the semi-continuous process does not only result in successfully making this process suitable for large scale production of the controlled crystallized dispersions, but it also results in a better product.

31


3.2. Introduction

Many new drugs evolving from drug discovery techniques like High Throughput Screening can be categorized according to the Biopharmaceutical Classification System as class II drugs [1]. These drugs have a poor oral bioavailability due to their low aqueous solubility. However, since they are easily absorbed throughout the gastro-intestinal membranes [2; 10] their bioavailability can be improved by increasing their dissolution rate [3; 19; 28].

The dissolution rate can be improved by decreasing their size. According to the Ostwald-Freundlich equation, the saturation concentration at the surface of small particles, especially in the nanorange, is higher than the saturation concentration at the surface of large particles [5]. In addition, by decreasing the particle size, the total surface area available for dissolution increases and consequently also the dissolution rate [6]. Therefore, the use of drug nanocrystals has been shown to be a suitable strategy to improve the dissolution rate [20; 28].

Recently, we have developed a novel bottom-up method to prepare such drug nanocrystals with improved dissolution behavior: “controlled crystallization during freeze-drying” [29]. In short, two solutions were prepared: one of a lipophilic drug in TBA, and another of a matrix material (i.e. mannitol) in water. Both solutions were mixed and subsequently immediately and quickly frozen to a temperature well below the Tg’, after which the solvents were removed by freeze-drying. This freeze-drying step was performed at a relatively high temperature (above the Tg’, but below the Te) to allow the drug and matrix material to crystallize in the freeze-concentrated fraction. It was shown that the size of the drug crystals could be varied by adjusting several process parameters (such as freezing rate and water/TBA ratio).

Although the above described process is suitable for lab-scale production, technical problems will be encountered when this process is changed to a large scale production process. The most important problem is that the solution of drug and matrix material in a mixture of TBA and water is thermodynamically unstable [9]. To prevent premature crystallization, the mixture should be frozen immediately after mixing and quickly to a temperature well below the Tg’. Therefore, until now only small quantities of the two solutions were mixed in a glass vial and the vial was immediately frozen in liquid nitrogen. However, when this batch process is scaled-up to a large scale production it is difficult to freeze it immediately after mixing and to freeze it sufficiently fast.

The aim of this study was to evaluate whether the above mentioned problems can be circumvented by using a 3-way nozzle. This nozzle consists of three channels through which the aqueous solution, TBA solution, and an atomizing air flow separately. The nozzle is designed in such a way that the atomizing air mixes the two solutions just at the moment they leave the nozzle. When this mixture is sprayed directly into liquid nitrogen it is immediately frozen after mixing. In addition, due to the atomizing air, small droplets are formed, which are therefore frozen almost instantaneously.

To validate whether the 3-way nozzle can be used for scale-up of the controlled crystallization

32

chapter 3

process from batch to semi-continuous, we first studied whether two liquids were mixed efficiently by using this 3-way nozzle. Secondly, the crystallinity and the dissolution rate of controlled crystallized dispersions prepared by the batch process and the semi-continuous process were compared. Fenofibrate was used as model drug, while mannitol was used as model matrix material.

3.3. Materials

Potassium iodate (kIO3), sodium hydroxide (NaOH) and sulfuric acid (H2SO4) were obtained from Merck (Darmstadt, Germany). Boric acid (H3BO3) was supplied by Lamers & Indemans (‘s Hertogenbosch, Netherlands). Potassium iodine (kI) was supplied by Genfarma B.v. (Maarssen, Netherlands). Sodium dodecyl sulphate (SDS) was provided by BUFA B.v. Uitgeest, The Netherlands. Fenofibrate and TBA were supplied by Sigma-Aldrich Chemie B.V. (Zwijndrecht, the Netherlands). Mannitol was obtained from Roquette (France). Demineralized water was used in all experiments.

3.4. Methods

3.4.1. Mixing quality of the 3-way nozzleTo determine whether two liquids mix rapidly after they left the 3-way nozzle-tip, the adapted Villermaux/Dushman method was used [30]. This method has been originally developed to measure the mixing quality of microfluidic devices, by using two parallel reactions taking place when an acidic and a buffered iodine/iodate solution are mixed [31]. During mixing, either the acid is involved in the formation of triiodine (in case of slow mixing) or the acid is neutralized by the buffer by which no triiodine is formed (in case of fast mixing). Therefore the amount of triiodine formed can be used as a measure for the mixing quality, since a poor mixing quality will cause some of the acid to react with iodine and iodate to form triiodine, which can be detected spectrophotometrically.

For this purpose two solutions were prepared: one consisting of 0.00635 M kIO3 and 0.0319 M KI in a 0.0909 M NaOH/H3BO3 buffer and a second solution of 0.030 M H2SO4 in water. Perfusion pumps (Ne300, New era Pump Systems Inc., Wantagh, NY, United States of America) were used to pump both solutions in a 1:1 v/v ratio separately through the 3-way nozzle (Büchi Labortechnik GmbH, the Netherlands). The nozzle consists of three separate channels: 1) the inner channel for the first liquid, 2) the middle channel for the second liquid, and 3) the outer channel for the atomizing air. After spraying the solutions, triiodine formation in the mixtures was measured spectrophotometrically at 286 nm (ThermoSpectronic, unicam Uv 500). The tested spray-settings were similar to the settings used during spray freeze-drying: an atomizing airflow of 500 Ln/hr (i.e. 500 L of air at 1 atm and 0 °C), a total liquid flow of 15 mL/min and a distance to the sprayed surface of 60 mm. As control, the two solutions were mixed without atomizing airflow.

33


3.4.2. Preparation of the controlled crystallized dispersionsTo prepare the controlled crystallized dispersions, two solutions were prepared. Fenofibrate was dissolved in TBA and mannitol was dissolved in water (for compositions see Table 3-I). Both solutions were heated to approximately 60 °C before they were processed using the batch or semi-continuous process.

Table 3.I. Composition of the different solutions used to prepare the controlled crystallized dispersions.

Water and TBA were mixed in a ratio of 6/4.


Cmannitol/water

(mg/mL)

Cfenofibrate/TBA

(mg/mL)

Drug load

(%/%)

39 25 30

25 25 40

17 25 50

For the batch process, both solutions were mixed in a 20 ml glass vial. The volumes of the aqueous- and TBA-solutions used were 1.2 and 0.8 mL, respectively. Immediately after mixing, the vials containing the mixture were immersed in liquid nitrogen and subsequently lyophilized. For the semi-continuous process the heated solutions were pumped separately through a heated 3-way nozzle, with a flow of 9 and 6 mL/min for the aqueous- and TBA-solutions respectively, using perfusion pumps. The liquid (total liquid volumes of 55, 70, and 90 mL for the dispersions having a drugload of 30, 40, and 50% w/w respectively) was then sprayed, with an atomizing airflow of approximately 500 Ln/hr, directly into a metal tray filled with liquid nitrogen. Subsequently the frozen material was freeze-dried.

Freeze-drying was done in a Christ model epsilon 2-4 lyophilizer (Salm en kip, Breukelen, The Netherlands). First the temperature of the samples was equilibrated on a pre-cooled shelf (-50 °C) for 1.5 hours. Secondly, the temperature was increased to -25 °C to allow the drug and matrix material in the freeze-concentrated fraction to crystallize. To remove the solvents, the pressure was decreased to 0.220 mBar after three hours. After at least 10 hours the temperature was gradually increased to 20 °C. The samples were stored in a dessicator over silica gel at room temperature for at least 1 day before further processing.

Physical mixtures of fenofibrate (x50 = 13 μm; determined by laser diffraction) and mannitol (x50 = 170 μm; according to Roquette specifications) were prepared using a spatula and mortar to mix both components.

3.4.3. TablettingRound and flat shaped tablets with a mass of 100 mg and a diameter of 9 mm were compressed on an eSH compaction apparatus (Hydro Mooi, Appingedam, The Netherlands) with a compaction pressure of 5 kN and a compaction rate of 5 kN/s. The die was lubricated with magnesium stearate. The tablets were stored in a vacuum desiccator over silica gel at room temperature for at least 1 day before further processing.

34

chapter 3

3.4.4. Differential Scanning CalorimetryDSC measurements were used to determine the degree of crystallinity of the components in the solid dispersions. The degree of crystallinity was calculated by dividing the heat of fusion of the component in the solid dispersion by the heat of fusion of the pure component as received, multiplied by the fraction of the component in the controlled crystallized dispersion. It was assumed that the supplied materials were fully crystalline. Samples of 2-10 mg were filled into hermetically closed aluminum pans. The samples were scanned at a heating rate of 2 °C/min, from -50 to 200 °C on a Q2000 calorimeter (TA Instruments, Ghent, Belgium). 3.4.5. X-Ray Powder DiffractionXRPD measurements were performed by using an X’pert Pro MPD diffractometer (PANalaytical, Almelo, the Netherlands). CuKα radiation with a wavelength of 1.5405 Å at 40 kV and 40 mA was used. Samples were scanned from 4-60° 2θ with a step size of 0.008° and a time per step of 35 s. The powders were placed on a zero-background silicon holder of 32 mm in diameter and 2 mm thickness (PANalaytical, Almelo, the Netherlands).

3.4.6. Scanning Electron MicroscopyA jeOL jSM 6301-F Microscope (jeOL, japan) was used to record SeM-pictures. The powder was dispersed on top of double-sided sticky carbon tape on metal disks and coated with a thin layer of gold/palladium in a Balzers 120B sputtering device (Balzers UNION, Liechtenstein).

3.4.7. DissolutionDissolution measurements were performed in 1 L 0.5% w/v sodium dodecylsulphate solutions at 37 °C, using a USP dissolution apparatus II (Rowa Techniek, Leiderdorp, The Netherlands). The paddle speed was set at 100 RPM. The fenofibrate concentration during dissolution was measured spectrophotometrically (Ultrospec III, Pharmacia LkB) at a wavelength of 290 nm.

3.5. Results and discussion

3.5.1. Mixing quality of the 3-way nozzleThe nozzle is designed to rapidly mix two liquids using the atomizing air, which results in the formation of homogeneous droplets. To test whether the two solutions were indeed effectively mixed, the degree of mixing of a solution prepared with an atomizing airflow of 500 Ln/hr (settings as used for the preparation of the controlled crystallized dispersions) was compared with a solution which was prepared without atomizing air, using the adapted Villermaux/Dushman method.

When the atomizing airflow of 500 Ln/hr was used, the average absorption was much lower (0.222 ± 0.06) compared to the solution prepared without atomizing air (1.163 ± 0.08). The lower absorption shows that less triiodine has been formed. Since the formation of triiodine is an indication for slow mixing, a lower absorption indicates faster and thus better mixing. Therefore the lower absorption shows that the two solutions were properly mixed by the

35


Fig. 3.1. A) Typical examples of DSC thermograms of a physical mixture (30% w/w) of fenofibrate and

mannitol (top), and two controlled crystallized dispersions containing 30% w/w fenofibrate in mannitol:

one prepared by the batch process (middle) and the other by the semi-continuous process (bottom).

B) Typical examples of X-ray diffraction patterns of fenofibrate as received, α-, β- and δ-mannitol and two

controlled crystallized dispersions containing 30% w/w fenofibrate in mannitol: one prepared by the batch

process (freeze-dried) and the other by the semi-continuous process (spray freeze-dried) (patterns of the

mannitol polymorphs were taken from the ICDD-library).

36

chapter 3

atomizing air. Hence it can be expected that the use of this nozzle to mix the aqueous- and TBA-solution, necessary for the preparation of the controlled crystallized dispersion, results in a homogeneous mixture.

3.5.2. From batch to semi-continuousDSC and XRPD measurements were performed to determine the crystallinity of fenofibrate and mannitol in the dispersions. The DSC thermograms (see Fig. 3.1A) show that the obtained dispersions (30% fenofibrate in mannitol prepared by the batch and the semi-continuous process) are indeed crystalline, as can be seen from the peaks in the DSC thermograms that correspond to the peaks at the same temperature as the peaks appearing in the thermogram of the physical mixture. Three crystalline anhydrous polymorphs of mannitol have been identified and their melting points differ only slightly [24]. Therefore also XRPD was used to identify which polymorphs were formed. Fig. 3.1B shows typical examples of XRPD-diffractograms of fenofibrate as received, α-, β- and δ-mannitol and two controlled crystallized dispersions containing 30% w/w fenofibrate in mannitol: one prepared by the batch process and the other by the semi-continuous process. The diffractograms of the samples clearly show that both dispersions contained crystalline fenofibrate and crystalline δ-mannitol. Since δ-mannitol was shown to be stable at ambient conditions for a few years, no shelf-life problems are to be expected [24].

Fig. 3.2. Dissolution profiles of tablets composed of a physical mixture (○) and controlled crystallized

dispersions of 30% w/w fenofibrate in mannitol. Controlled crystallized dispersions were prepared by the

batch (●) and semi-continuous process (● ). (n=3-6; mean).

37


Dissolution rates of fenofibrate from tablets prepared from a physical mixture were compared to those of crystallized dispersions of 30% w/w fenofibrate in mannitol prepared by the batch and semi-continuous process (Fig. 3.2). The dissolution rate of fenofibrate from tablets composed of a physical mixture, was low (less than 50% dissolved after 2 hours). The dissolution rate of fenofibrate was clearly improved when the tablets consisted of the controlled crystallized dispersions prepared by the batch process, as was also shown in a previous study [29]. The tablets composed of the dispersion prepared by the semi-continuous process showed an almost identical dissolution profile. Therefore it can be concluded that the modification of the batch process to a semi-continuous process was successful.

The dissolution rate of the samples prepared by the semi-continuous process appeared to be even slightly higher than the samples prepared by the batch process which may indicate the formation of smaller drug crystals. As described earlier [29], a higher freezing rate results in the formation of smaller drug nanocrystals which can be explained by the fact that when the temperature of the samples during freeze-drying is above the Tg’ but below the Te, crystallization of both drug and matrix material in the freeze-concentrated fraction can occur. The growth of the crystals is limited by the size of the interstitial spaces between the solvent crystals. Since a higher freezing rate will result in smaller solvent crystals and consequently in smaller interstitial spaces, the final drug crystals will be smaller. With the semi-continuous process, droplets of the mixture were formed by atomizing the feed mixture which were directly immersed into

Fig. 3.3. Scanning electron micrographs of controlled crystallized dispersions of 30% w/w fenofibrate in

mannitol prepared by the batch and semi-continuous process (the pictures on the left have a magnification

of 1500x and the pictures on the right a magnification of 10000x).

38

chapter 3

liquid nitrogen (see Fig. 3.3 for a SeM picture of a freeze-dried droplet), while for the batch process 2 mL of liquid in glass vials was frozen in liquid nitrogen. Although the Leidenfrost effect could occur during spraying the mixture into liquid nitrogen [32-33], it has been shown that extremely rapid freezing rates could be achieved by spray-freezing into liquid nitrogen [34], because the volume of the individual droplets is smaller than the volume of the liquid in the glass vial. Therefore we assumed that the freezing rate for the semi-continuous process was higher than that of the batch process, and consequently smaller drug crystals were formed during the semi-continuous process.

3.5.3. High drugloadsIn the previous paragraph it was shown that modification of the batch process to a semi-continuous process was successful. But it seemed to have an additional advantage. The dissolution rate of the samples prepared by the semi-continuous process was slightly higher than the dissolution rate of the samples prepared by the batch method. As described above, we speculated that the increased dissolution rate was the result of a smaller drug crystal size of fenofibrate in the controlled crystallized dispersions prepared by the semi-continuous process. However, the difference in dissolution rate between the samples prepared by the semi-continuous and the batch process was small. Obviously this is explained by the fact that the dissolution rate of the samples prepared by the small batch process was already quite high.

Fig. 3.4. Dissolution profiles of tablets composed of physical mixtures (open symbols) and controlled

crystallized dispersions (closed symbols) containing 40% w/w (□ ,■ ,■ ) and 50% w/w (▽ ,▼ ,▼ )

fenofibrate in mannitol. Controlled crystallized dispersions were prepared either by the small batch (black

symbols) or the semi-continuous (grey symbols) process. (n=3-6; mean).

39


Since it can be envisaged that the dissolution rate of fenofibrate from a tablet with a higher drug load would be lower, it was expected that differences in dissolution rate of samples prepared by the two methods would be more pronounced. Therefore, the dissolution rate of fenofibrate from tablets with higher drug loads of 40 and 50%, prepared both by the batch and the semi-continuous process, was determined.

Fig. 3.4 shows that by increasing the drug load from 30 to 40 or even 50% w/w, the dissolution rate decreases. This can be explained by the fact that the mass of all tablets is similar and that the tablets become therefore more lipophilic. Due to the higher lipophilicity, the wetting of the tablets will be slower, and hence the dissolution rate of the drug from the tablets will decrease. This effect was observed for the dispersions prepared by both the batch process and the semi-continuous process. Fig. 3.4 also shows that the difference in dissolution rate between the tablets composed of controlled crystallized dispersions prepared by the batch and the semi-continuous process is much larger than when a lower (30% w/w) drug load is used. Since this difference could be explained by differences in crystallinity or in differences in particle size, DSC and SeM measurements were performed.

DSC measurements indicated that controlled crystallized dispersions of both drug loads were highly crystalline and had a similar degree of crystallinity (the degree of crystallinity of fenofibrate was 86-93% and 82-86% for the samples prepared by the batch process and the

Fig. 3.5. Scanning electron micrographs of controlled crystallized dispersions of 40% (left) and 50%

(right) w/w fenofibrate in mannitol prepared by the batch (top) and semi-continuous (bottom) process

(magnification 10,000x).

40

chapter 3

semi-continuous process, respectively). Therefore, the large difference in dissolution rate must be attributed to a smaller drug crystal size. SeM pictures show that the particles of both spray freeze-dried samples are much smaller than the particles of the freeze-dried samples (Fig. 3.5). This indicates that the crystal size in the spray freeze-dried samples is indeed smaller than in de freeze-dried samples. Furthermore, these SeM-pictures illustrate that the particle architecture consists of agglomerates of crystals and was similar for all three drug loads (see also Fig. 3.3). In addition, it can clearly be seen that the particles are smaller than 1 μm, i.e. that they are of nanoscale. Therefore it is concluded that using the 3-way nozzle not only allows scaling-up of the controlled crystallization process, but also results in smaller crystals with a higher dissolution rate.

3.6. Conclusions

In this study, it was demonstrated that the controlled crystallization process can be modified from a batch process to a semi-continuous process by using a 3-way nozzle. The adapted Villermaux/Dushman method indicated efficient and fast mixing of solutions immediately after they left the nozzle. Furthermore it was shown that the dissolution rate of controlled crystallized dispersions of 30, 40, and 50% w/w fenofibrate in mannitol prepared by the semi-continuous process were all high. In fact, the dissolution rate was higher than that of controlled crystallized dispersions prepared by the batch method. Since the degree of crystallinity is similar for the dispersions obtained from both processes, the differences in dissolution rate can be explained by the differences in crystal size. The freezing rate during the semi-continuous process was higher than during the batch process. The higher freezing rate resulted in smaller drug crystals and consequently a higher dissolution rate. Therefore we concluded that the controlled crystallization process can successfully be made ready for large scale production by applying a 3-way nozzle. In addition, controlled crystallized dispersions prepared by the semi-continuous process also showed a higher dissolution rate and therefore tablets with a higher drug load can be prepared. Therefore this process is not only ready for large scale production, but it also results in a better product.

41


chapterelucidation of the mechanism by in-line raman spectroscopy 4

44

chapter 4

4.1. Abstract

We developed a novel process, “controlled crystallization during freeze-drying” to produce drug nanocrystals of poorly water soluble drugs. This process involves freeze-drying at a relatively high temperature of a drug and a matrix material from a mixture of tertiary butyl alcohol and water, resulting in drug nanocrystals incorporated in a matrix. The aim of this study was to elucidate the mechanisms that determine the size of the drug crystals. Fenofibrate was used as a model lipophilic drug. To monitor the crystallization during freeze-drying, a Raman probe was placed just above the sample in the freeze-dryer. These in-line Raman spectroscopy measurements clearly revealed when the different components crystallized during freeze-drying. The solvents crystallized only during the freezing step, while the solutes only crystallized after the temperature was increased, but before drying started. Although the solutes crystallized only after the freezing step, both the freezing rate and the shelf temperature were critical parameters that determined the final crystal size. At a higher freezing rate, smaller interstitial spaces containing the freeze-concentrated fraction were formed, resulting in smaller drug crystals (based on dissolution data). On the other hand, when the solutes crystallized at a lower shelf temperature, the degree of supersaturation is higher, resulting in a higher nucleation rate and consequently more and therefore smaller crystals. In conclusion, for the model drug fenofibrate, a high freezing rate and a relatively low crystallization temperature resulted in the smallest crystals and therefore the highest dissolution rate.

H. de Waarda, T. De Beerb, W.L.j. Hinrichsa, C. vervaetc, j.P. Remonc, and H.W. Frijlinka

a Department of Pharmaceutical Technology and Biopharmacy, University of Groningen, Antonius Deusinglaan 1, 9713 AV, Groningen, the Netherlands. b Department of Pharmaceutical Analysis, Ghent University, Harelbekestraat 72, 9000, Ghent, Belgium. c Department of Pharmaceutical Technology, Ghent University, Harelbekestraat 72, 9000, Ghent, Belgium.

Published as “Controlled crystallization of the lipophilic drug fenofibrate during freeze-drying: elucidation of the mechanism by in-line Raman spectroscopy” in: AAPS j. 2010. 12(4), 569-575.

45

elucidation of the mechanism by in-line raman spectroscopy

4.2. Introduction

Many newly developed drugs are categorized as class II drugs according to the Biopharmaceutics Classification System [1]. These drugs have a low solubility in water, but once dissolved, they are easily absorbed over the gastro-intestinal membrane [2; 10]. For many of these drugs, the dissolution rate is the rate-limiting step for absorption. Therefore the bioavailability can be improved by increasing the dissolution rate [3]. One strategy for increasing the dissolution rate is the application of drug nanocrystals [28; 35-37]. The dissolution rate of nanocrystals is increased by their increased surface area (Noyes-Whitney equation) [6], the decreased thickness of the diffusion boundary layer [38], and the increased saturation concentration around the small particles (kelvin’s law) [5].

Current methods to prepare drug nanocrystals can be divided into top-down and bottom-up methods. Most currently used top-down methods are high-pressure homogenization [18] and wet-ball milling [19]. Disadvantages of these methods include, the use of surfactants, the difficulty of achieving a uniform size distribution, low yields, and possible contamination from grinding media [20-22]. For currently used bottom-up methods, disadvantages include the use of toxic solvents and the difficulty of adequately controlling the particle size [22-23].

To overcome these disadvantages, we developed a novel bottom-up process to produce drug nanocrystals: controlled crystallization during freeze-drying [29]. Briefly, this method consisted of the following steps. Firstly, two solutions are prepared: one is a solution of the poorly water-soluble drug in TBA and the other is a solution of a matrix-material in water. These solutions are mixed, frozen, and then freeze-dried. Freeze-drying is performed at a relatively high temperature to allow both the drug and the matrix to crystallize. By using this process, nanocrystalline dispersions with improved dissolution behavior were obtained.

When we developed this process, we found that the size of the nanocrystals was affected by the process conditions, such as the freezing rate. However, it was not clear during which stage of the process the solutes crystallized and how the freezing rate caused its effect on the crystal size. Since the mixture is thermodynamically unstable, the drug and the matrix could crystallize 1) during freezing, 2) after freezing at a temperature above the Tg’, or 3) during drying. knowledge on when and how crystallization occurs can help to elucidate the mechanisms through which parameters such as freezing rate affect the final size of the crystals. Crystallization processes can be monitored with Raman spectroscopy. The effects of the process conditions on the drug particle size could be elucidated if the crystallization of both the solvents and the solutes could be monitored during the freeze-drying process.

The aim of the study was to elucidate the mechanisms that determine the size of the drug crystals formed during freeze-drying. Thereto, in-line Raman spectroscopy was used as process analytical tool. Since Raman spectroscopy is a fast, non-invasive technique, it offers the opportunity to monitor physical changes of the formulation during freeze-drying. One of the advantages of Raman spectroscopy is that Raman spectra show sharp peaks by which individual components generally do not overlap each other. This makes it relatively easy to

46

chapter 4

determine which component crystallizes at which stage during the process [39].

In this study, a Raman probe was placed immediately above the sample in the freeze-dryer (as described by De Beer et al. [40]). By using this in-line method, the crystallization of the four components (a model drug, matrix, TBA, and water) during the process could be monitored. The different stages during the CCDF-process (freezing, increase in temperature, and drying) were separated from each other and prolonged to ensure that the complete phase changes during each stage were measured. In addition, two process conditions, the freezing rate and the temperature of the freeze-dryer shelf, were varied. Monitoring the crystallization process yields insight into the moment and length of the crystallization of the different components, while varying the process conditions reveals which factors influence the model drug crystal size and therefore the dissolution rate. Fenofibrate and mannitol were used as the model drug and the matrix material, respectively.

4.3. Materials

Fenofibrate and TBA were obtained from Sigma-Aldrich Chemie B.V. Zwijndrecht, The Netherlands. Mannitol was purchased from vWR international (Fontenay sous Bois, France).

4.4. Methods

4.4.1. Preparation of the crystalline dispersionsCrystallized dispersions were prepared similarly to the method described previously [29]. Briefly, the drug, fenofibrate, was dissolved in TBA (25 mg/mL) and the matrix material, mannitol, in water (31 mg/mL). The solutions were heated to approximately 60 °C and 2.4 ml of the aqueous solution was mixed with 1.6 mL of the TBA solution. After mixing, the vials were frozen. Two freezing rates were used: immediately freezing the vials on a precooled (-50 °C) freeze-dryer shelf (cooling rate of the sample approximately 2.5 ºC/min); and freezing at 1 °C/min by placing the vial on the shelf at 20 °C and cooling the shelf to -50 °C at a rate of 1 °C/min (cooling rate of the sample approximately 0.8 ºC/min). In both cases, the temperature of -50 ºC was maintained for two hours. The temperature of the shelf was then increased to either -25 °C or -15 °C in 30 minutes and this temperature was maintained for another seven hours. Next, sublimation of the solvents was initiated by decreasing the pressure to 0.8-1.0 mBar in 20 minutes and the samples were dried for 30 h (see Table 4-I). The temperature of the shelf was then gradually (two hours) raised to room temperature. Finally, the freeze-dryer (Amsco-Finn Aqua GT4 freeze-dryer) was emptied and the samples, having a drug load of 35% w/w, were stored in a vacuum desiccator over silica gel for at least one day before further processing.

47


Table 4-I. Description of the process and its variables to crystallize and freeze-dry the water/TBA

mixtures.

Step Temperature (ºC) Length of step (h) Pressure (mBar)

1. Freezing variable 1 1000

-50 2 1000

2. Ramping to Tc variable 2 0.5 1000

3. Crystallization variable 2 7 1000

Drying variable 2 30 0.8-1.0

Ramping to Tend 20 2 0.8-1.0

20 1 0.8-1.0

emptying 20 - 0.8-1.0

Process variable 1: freezing rate variable 2: Tc

1. (Slow; -25 ºC) From 20 to -50 ºC at 1 ºC/min -25 ºC

2. (Slow; -15 ºC) From 20 to -50 ºC at 1 ºC/min -15 ºC

3. (Fast; -25 ºC) vial on pre-cooled shelf (-50 ºC) -25 ºC

4. (Fast; -15 ºC) vial on pre-cooled shelf (-50 ºC) -15 ºC

4.4.2. In-line Raman spectroscopyA non-contact probe coupled to a RamanRxn1 spectrometer (kaiser Optical Systems, Ann Arbor, Michigan, USA) via a glass fiber optic cable, was placed immediately above a vial inside the freeze-dryer. The spectrometer was equipped with an air-cooled CCD detector (back-illuminated deep depletion design). The laser wavelength was the 785 nm line from a 785 Invictus NIR diode laser. Spectra were collected every minute during freeze-drying and the exposure time was 30 seconds. All spectra were recorded at a resolution of 4 cm-1 using a laser power of 400 mW. The HoloREACT reaction analysis and profiling software package, the Matlab Software package (version 6.5), and the Grams/AI-PLSplusIQ software package (version 7.02) were used for data-collection, –transfer, and –analysis. Spectra were preprocessed by baseline correction (Pearson’s method).

4.4.3. Differential scanning calorimetryThe degree of crystallinity (defined as the ratio between the heat of fusion of the drug or carrier in the solid dispersion and the heat of fusion of the drug or carrier as received multiplied by the fraction drug or carrier in the mixture) of fenofibrate and mannitol in the crystallized dispersions was determined by DSC. Hermetically closed aluminum pans were filled with 2-8 mg of the sample and heated at 2 °C/min, from -50 to 200 °C in the scanning calorimeter (Q2000, TA Instruments, Ghent, Belgium).

4.4.4. X-ray powder diffractionXRPD was used to determine the crystallinity and identify the polymorphic forms of both solutes. Therefore a CuKα radiation with a wavelength of 1.5405 Å at 40 kV and 40 mA from an X’Pert PRO MPD diffractometer (PANalytical, Almelo, The Netherlands) was used. The sample

48

chapter 4

Fig. 4.1. A) X-ray diffraction patterns of fenofibrate as received, α-, β- and δ-mannitol and B) the

controlled crystallized dispersions containing 35% w/w fenofibrate in mannitol (patterns of the mannitol

polymorphs were taken from the ICDD-library).

49


powders were placed on a zero-background silicon holder (diameter, 32 mm; thickness, 2 mm) and scanned from 4–60° 2θ with a step size of 0.008° and a time per step of 35 s.

4.4.5. TablettingThe obtained powders were compressed to 9-mm tablets of 100 mg on a eSH compaction apparatus (Hydro Mooi, Appingedam, The Netherlands). The compaction rate was 5 kN/s to a maximum compaction load of 5 kN. The tablets were stored in a vacuum desiccator over silica gel at room temperature for at least one day before further processing.

4.4.6. DissolutionThe dissolution rate of fenofibrate from these tablets was tested in a USP dissolution apparatus II, Rowa Techniek. Dissolution was performed in 1 L of a 0.5% w/v sodium dodecyl sulphate solution at 37 °C; the paddle speed was 100 rpm. The concentration of fenofibrate was measured spectrophotometrically (Uv-vIS spectrophotometer Uv-1601, Shimadzu) at a wavelength of 290 nm.

4.5. Results

4.5.1. Physicochemical characterization of the dispersionsThe physical properties of the obtained freeze-dried product were determined by both DSC and XRPD. The DSC-thermograms of all four dispersions showed melting peaks corresponding to crystalline fenofibrate and crystalline mannitol. Mannitol exists as three anhydrous polymorphs whose melting points differ only slightly (166.5, 166, and 155 ºC for the α-, β-, and δ-polymorphs, respectively [24]). Therefore DSC is not the most suitable method to identify which polymorph has been formed. The XRPD patterns showed that all samples contained only the δ-polymorph (Fig. 4.1). Although the pure mannitol as received did not consist of δ-mannitol, DSC can be used to determine the degree of crystallinity of mannitol since the melting enthalpies of α-, β-, and δ-mannitol are also similar [24]. Fenofibrate and mannitol in all four dispersions were highly crystalline and had a similar degree of crystallinity (96-102 % for fenofibrate and 92-95% for mannitol; Table 4-II).

Table 4-II. Degree of crystallinity of both drug and matrix (mean ± standard deviation; n=3).

Degree of crystallinity

Process Fenofibrate Mannitol

1. (Slow; -25 ºC) 102.3 ± 0.8 91.9 ± 0.2

2. (Slow; -15 ºC) 99.5 ± 0.6 92.5 ± 0.6

3. (Fast; -25 ºC) 96.4 ± 2.2 95.1 ± 1.2

4. (Fast; -15 ºC) 100.2 ± 1.2 91.5 ± 2.1

4.5.2. In-line Raman spectroscopyeach crystalline component had a characteristic peak that did not overlap with characteristic peaks of any of the other components (Fig. 4.2). The characteristic peak used for fenofibrate

50

chapter 4

was at 1580-1610 cm-1, for mannitol at 865-895 cm-1, for TBA at 725-763 cm-1, and for water at 208-226 cm-1 [41-43]. Since these peaks can be clearly distinguished from each other, it is possible to use in-line Raman spectroscopy to monitor the crystallization of the individual components. To determine when crystallization of fenofibrate, δ-mannitol, and water started and ended, the peak intensity of each individual characteristic peak was determined. An increase in peak intensity indicates the start of crystallization. The peak intensity of the TBA peak can not be used for the determination of the start and end of the TBA crystallization, since liquid TBA already shows a peak with a high intensity. Therefore the width of the TBA peak was used to determine the crystallization of TBA. A narrowing of the peak indicates the start of the crystallization of TBA [44]. Furthermore, it should be mentioned that Raman spectroscopy only allows to measure the surface of the sample.

The intensity of the water peak increased and the width of the TBA peak decreased during the freezing stage of process 1 (Fig. 4.3), corresponding to the crystallization of the solvents. The intensity of the fenofibrate peak decreased slightly due to the decrease in temperature and the the peak of crystalline mannitol was not detected during this stage of the CCDF-process. Thus only the solvents and not the solutes crystallized during the freezing stage. After approximately 80 minutes, the intensity of the solvent peaks did not change anymore, indicating that almost all solvent had crystallized. At this point during the CCDF-process, the system consisted of crystalline water (ice), crystalline TBA, and a vitrified freeze-concentrated solution of fenofibrate and mannitol in a mixture of water and TBA.

Fig. 4.2. Raman spectrum of the sample during freeze-drying it according to process 1 after 500 minutes.

The peaks used to determine the intensities of the individual components are encircled.

51


After the temperature was increased to -25 ºC, the intensity of the peaks corresponding to both fenofibrate and mannitol increased. Indicating that the solutes crystallized at this stage. Since the intensities did not change any further after approximately 460 and 480 minutes for fenofibrate and mannitol, respectively (Table 4-III), which was well before the vacuum was applied (after 640 minutes), these data show that fenofibrate and mannitol crystallized completely before the drying step started. After 650 minutes the width of the TBA peak increased and the intensity of the water peak decreased, indicating that sublimation started immediately after the pressure was decreased. Due to the sublimation of the solvents, the concentration of the solutes, and therefore the intensities of the solute peaks, increased. After 1144 and 980 minutes the peaks of TBA and water, respectively, disappeared. Since only the surface of the sample can be measured by Raman spectroscopy, this indicates that the top of the sample was dried in approximately 8.5 hours. To end with a completely dry crystalline dispersion, the total drying time used was 30 hours.

Fig. 4.3. The intensities of fenofibrate (1580-1610 cm-1), mannitol (865-895 cm-1), and water (208-226

cm-1) peaks and the peak width of TBA (725-763 cm-1) during the first 1000 minutes of process 1. During

this process the freeze-drier shelf was cooled during the first 70 minutes from 20 ºC to -50 ºC, then the

temperature was kept constant at -50 ºC for 120 minutes, then the temperature of the freeze-drier shelf was

increased to -25 ºC in 30 minutes and kept constant for 420 minutes. Finally, the pressure was decreased

to 0.8-1.0 mBar to dry the samples.

52

chapter 4

Table 4-III. Summary of the average time at which the crystallization of each individual component

starts and ends.

Crystallization times (min)

Process 1 Process 2 Process 3 Process 4

Start end Start end Start end Start end

Fenofibrate 291 461 273 353 243 439 207 307

Mannitol 237 480 231 379 222 475 199 314

TBA 59 75 44 67 15 30 14 25

Water 71 80 63 68 19 33 21 26

The solutes and solvents of the other three processes crystallized at the same stages of the freeze-drying process as found for process 1 (Table 4-III; note that step 2 of process 3 and 4 start 70 minutes earlier than step 2 of process 1 and 2, due to the different freezing rate). The solvent crystallized upon cooling, while crystallization of the solutes was not observed at this stage. The solutes crystallized after the temperature had been increased to -25 ºC (processes 1 and 3) or -15 ºC (processes 2 and 4). Even if the mixture was cooled at a rate of at least 1 ºC/min, the solutes did not crystallize during freezing, but only after the temperature was increased. Since the intensity of the solute peaks reached a constant value before the vacuum was applied, the solutes indeed crystallized only during the time between freezing and drying.

4.5.3. DissolutionAs expected from earlier studies [29; 45], the dissolution rate of samples that were made by rapid freezing (vials placed on a pre-cooled freeze-dryer shelf) was higher than the dissolution rate of samples that were frozen more slowly (at 1 ºC/min) (see Fig. 4.4). Furthermore, a relatively low shelf temperature during the crystallization stage (-25 ºC) resulted in faster dissolution than a relatively high shelf temperature (-15 ºC). Between the samples that were crystallized at a relatively low temperature, the effect of the freezing rate on the dissolution behavior is negligible, indicating that crystallization at a lower temperature could diminish the effect of the freezing rate. Since all samples had a similar degree of crystallinity and consisted of the same polymorphic forms, differences in dissolution rate can be attributed to differences in crystal size [29; 45]. Thus, both the rate of freezing and the temperature at which crystallization occurs influence the final particle size. Based on the dissolution data, we conclude that a higher freezing rate and a lower temperature at which the solutes crystallized, resulted in smaller drug crystals.

4.6. Discussion

In this study, we clearly elucidated the mechanism of crystallization during the CCDF-process. The in-line Raman spectroscopy measurements showed that the solvents and the solutes crystallized at different steps during the CCDF-process. During the freezing step, the solvents crystallize. Due to the crystallization of the solvents, the remaining solution becomes more concentrated. Since the remaining solution (containing fenofibrate, mannitol, TBA, and water) is

53


thermodynamically unstable, the solutes can start to crystallize as well. However, if the freezing occurs rapidly, and if the system is cooled to a temperature well below the Tg’, the mobility is strongly reduced and a rigid glass is formed. Due to the strongly reduced mobility neither the solvents nor the solutes can crystallize in the freeze-concentrated fraction. The in-line¬ Raman data show that the solutes did not crystallize at the freezing stage of the CCDF-process. Thus, the freezing rate (even at 1 ºC/min) is apparently fast enough to prevent crystallization of the solutes during freezing and the reached temperature of -50 ºC is well below the Tg’.

After step 2, when the temperature was increased to either -25 ºC or -15 ºC, the solutes crystallized and the crystallization was finished before the next step, drying, was started. The Tg’ of the solutes could not be properly measured in this complex system by DSC (data not shown), but since the solutes crystallized at these temperatures, the temperature of the system was apparently above the Tg’. At temperatures below the Tg’, the freeze-concentrated fraction is glassy and the solutes can not crystallize.

Although the solutes crystallized solely during the step after the freezing step, not only the crystallization temperature, but also the freezing rate determined the particle size of the

Fig. 4.4. Dissolution profiles of tablets composed of a physical mixture (■ ) and controlled crystallized

dispersions of 35% w/w fenofibrate in mannitol. Controlled crystallized dispersions were prepared by slow

freezing (closed symbols) or fast freezing (open symbols) and a low (-25 ºC) crystallization temperature

(circles) or a high (-15 ºC) crystallization temperature (triangles). Process 1 corresponds to (●), process

2 corresponds to (▲ ), process 3 corresponds to (○), and process 4 corresponds to (△ ) (n=3-6; mean ±

standard deviation).

54

chapter 4

powder. Therefore we conclude that two phenomena play a role in the formation and growth of the crystals during CCDF.

The first process variable that determines the final crystal size was the freezing rate. A higher freezing rate resulted in smaller crystals. This effect was already shown in an earlier study, in which we also speculated about different mechanisms that could determine the drug crystal size [29]. Since we have shown in this study that both the drug and the matrix crystallized after freezing, we conclude that the size of the solvent crystals determines the final solute crystal size. A higher freezing rate results in smaller solvent crystals [46] and consequently smaller interstitial spaces, containing the freeze-concentrated fraction, between the solvent crystals. Since the solutes crystallized in the freeze-concentrated fraction, the size of the interstitial spaces limits the final size of the solute crystals. Thus, a high freezing rate results in smaller solvent crystals and therefore smaller interstitial spaces and consequently smaller solute crystals.

The second process variable that determines the crystal size is the shelf temperature. We found that a lower shelf temperature during crystallization (-25 ºC vs. -15 ºC) resulted in smaller crystals. At a lower shelf temperature, the degree of supersaturation is higher, resulting in a higher nucleation rate [47]. Since the formation of new nuclei and crystal growth are two competing processes [48], a higher nucleation rate results in more nuclei and consequently in smaller crystals. Thus, although the size of the interstitial spaces (controlled by the freezing rate) determines the maximum particle size, the crystal size is also determined by the balance between the formation of new nuclei and crystal growth (controlled by the shelf temperature).

4.7. Conclusion

In this study we revealed during which stages of the CCDF-process the solvents as well as the solutes crystallized. Based on this, we identified the critical steps for crystal formation for the model drug fenofibrate during freeze-drying. CCDF consists of three successive steps: freezing, increasing the temperature, and drying. The in-line Raman measurements showed that the first two steps, the freezing step and the crystallization step are critical steps that determine the final size of the fenofibrate crystals. When the mixture was frozen more rapidly, the interstitial spaces between the solvent crystals are smaller and because the size of the solute crystals is limited by the size of the interstitial spaces, smaller fenofibrate crystals are formed. When the solutes crystallize at a lower shelf temperature, the degree of supersaturation in the freeze-concentrated fraction is higher. A higher degree of supersaturation (as caused by lower crystallization temperatures) results in a higher nucleation rate and consequently more and therefore smaller fenofibrate crystals are formed. The combination of a high freezing rate and a relative low crystallization temperature results in the smallest fenofibrate crystals and consequently a higher dissolution rate. Furthermore these results show that the fenofibrate crystal size can be controlled by choosing the appropriate freezing rate and shelf temperature. Further studies are necessary to prove whether the critical steps as defined in this study and the mechanism of crystal formation are generally applicable to other lipophilic drugs.

55


chapterCLSM as quantitative method to determine the size of drug crystals in a solid dispersion 5

58

chapter 5

H. de Waarda, M.j.T. Hesselsa, M. Boona, k.A. Sjollemab, W.L.j. Hinrichsa, A.C. eissensa, H.W. Frijlinka

a Department of Pharmaceutical Technology and Biopharmacy, University of Groningen, Antonius Deusinglaan 1, 9713 AV, Groningen, the Netherlands. b Microscopy and Imaging Center, University Medical Center Groningen, Antonius Deusinglaan 1, 9713 AV Groningen, The Netherlands.

Published in: Submitted for publication, 2010.

5.1. Abstract

Purpose. The aim of this study was to test whether confocal laser scanning microscopy can be used as analytical tool to determine the drug crystal size in a powder mixture or a crystalline solid dispersion. Methods. To test this, crystals of the autofluorescent drug dipyridamole were incorporated in a matrix of crystalline mannitol by physical mixing or freeze-drying. Laser diffraction analysis and dissolution testing were used to validate the particle size that was found by CLSM. Results. The particle size of the pure drug as determined by laser diffraction and CLSM were similar (D50 of approximately 23 μm). CLSM showed that the dipyridamole crystals in the crystalline dispersion obtained by freeze-drying of less concentrated solutions were of sub-micron size (0.9 μm), whereas the crystals obtained by freeze-drying of more concentrated solutions were larger (1.5 μm). This trend in drug crystal size was in agreement with the dissolution behavior of the tablets prepared from these products. Conclusions. CLSM is a useful technique to determine the particle size in a powder mixture. Furthermore, CLSM can be used to determine the drug crystal size over a broad size distribution. A limitation of the method is that the drug should be autofluorescent.

59

clsm as quantitative method to determine the size of drug crystals in a solid dispersion

5.2. Introduction

Many new drug candidates can be categorized as class II drugs according to the Biopharmaceutics Classification System [1]. These drugs have a low aqueous solubility, but once they are dissolved, they are easily absorbed over the gastro-intestinal membrane [2; 10]. Since dissolution is the rate limiting step for absorption for these drugs, the bioavailability can be improved by increasing the dissolution rate [3]. A strategy to increase the dissolution rate is the application of drug nanocrystals [35; 37; 49-50]. According to the Ostwald-Freundlich equation [5] and the Noyes-Whitney equation [6], the dissolution rate of drug nanocrystals is increased due to an increased saturation concentration around the small particles and due to an increased surface area.

In a previous study, we developed a novel bottom-up process to produce such drug nanocrystals: controlled crystallization during freeze-drying [29]. Briefly, a drug and a matrix material, dissolved in a mixture of TBA and water was rapidly frozen and then freeze-dried. Freeze-drying was performed at a relatively high temperature to allow both drug and matrix to crystallize [51]. By using this process, nanocrystalline dispersions with improved dissolution behavior were obtained.

Usually, methods such as DLS and laser diffraction are used to measure the size of nanoparticles. Disadvantages of these methods are that when DLS is used only the submicron particles can be measured [52]. And when laser diffraction is used, it is difficult to apply the correct optimal parameters [53]. Nevertheless, these methods could provide accurate particle size distributions of nanosuspensions. However, to be able to measure nanoparticles with these methods, the particles should be dispersed in a liquid. When top-down methods such as high pressure homogenization or wet ball milling are used, the obtained product is a nanosuspension. However, when CCDF is used to prepare drug nanocrystals, the particles should be dispersed in a liquid first. In order to determine the size of the drug particles in the matrix, the matrix material should dissolve in the liquid, but the drug should not. Unfortunately, when the crystalline dispersions are dispersed in for aqueous or non-aqueous solvents, the drug particles can easily agglomerate or dissolve rapidly [54]. Agglomeration can be prevented by the addition of surfactants. However, due to the addition of the surfactants, the drug might partially dissolve. Since especially the smallest particles will dissolve, this may result in a wrong (too large) crystal size to be found. Also Ostwald ripening can occur by which one also finds a too large particle size [54-55].

If the particle size of the drug in the matrix could be determined in solid state, these drawbacks would be overcome. A frequently used method to distinguish between different components in solid state, is eDX. For this method, it is required that the different molecules in the mixture contain different types of atoms and preferably, that the deviating atom has a high atomic mass. However, frequently the atomic mass or the concentration of the deviating atoms in the drug is too low, or there are no deviating atoms at all. For example, in a previous study [29], we used fenofibrate and mannitol as model drug and matrix, respectively. The only atom type in fenofibrate that is absent in mannitol is chloride. Since there is only one chloride atom

60

chapter 5

per molecule fenofibrate and the atomic mass of chloride is low, EDX could not be used to determine the drug crystal size.

Another method that could be used to distinguish the drug from the matrix, is Raman spectroscopy. Raman spectroscopy is frequently used to determine the distribution of a drug in a mixture. However, since the spatial resolution is usually approximately 1 μm [56] or at optimal conditions maximally 0.25 μm [57], this technique is suitable for particles of at least a few micrometers, but not for particles in the sub-micron range.

CLSM on the other hand, could be a useful addition to current analytical tools to determine the drug crystal size in solid state. The reason is that CLSM has a high resolution and is capable of providing quantitative measurements covering several orders of magnitude [58]. However, to be able to distinguish between the drug from the matrix, the drug of interest or the matrix material should be autofluorescent. Since there are many examples of autofluorescent drugs (e.g. dipyridamole, nifedipine, tetracyclines, and acridines), CLSM could be a useful method to determine the drug particle size.

The aim of this study was to test whether CLSM can be used to determine the drug crystal size in the solid state. In order to test this, drug crystals of a poorly soluble and autofluorescent drug were prepared by CCDF. Dipyridamole was selected as model drug, because it can be categorized as BCS class II drug (logP: 3.95 [59]; aqueous solubility at neutral pH: 38 mg/L [60]) and it is autofluorescent due to its aromatic ring system [61].

5.3. Materials

Dipyridamole and TBA were obtained from Sigma-Aldrich Chemie B.V. Zwijndrecht, The Netherlands. Mannitol was obtained from Roquette (France). Demineralized water was used in all experiments.

5.4. Methods

5.4.1. Preparation of the crystalline dispersionsThe controlled crystallized dispersions were prepared according to the method described previously [29]. Briefly, two separate solutions of dipyridamole in TBA and mannitol in water were prepared (for compositions see Table 5-I). Both solutions were heated to approximately 60 ºC and mixed in a glass vial. Immediately after mixing, the solutions were frozen by placing them on a pre-cooled (-50 ºC) shelf of a freeze-dryer (Christ model epsilon 2-4 lyophilizer, Breukelen, the Netherlands). This temperature was maintained for 1.5 hour, after which the temperature was increased at a rate of 1 ºC/min to -25 ºC to allow the drug and matrix to crystallize. After three hours, the samples were dried in 10 hours at a 0.220 mBar. Finally, the temperature was gradually increased to room temperature. The obtained powders were stored in a dessicator over silica gel at room temperature for at least 1 day before further processing.

61


A physical mixture of dipyridamole and mannitol was prepared using a spatula and mortar.

Table 5-I. Compositions of the different slutions used to prepare the controlled crystallized dispersions.

In all experiments 1.2 mL aquous solution was mixed with 0.8 mL TBA solution (i.e. ratio of 6/4).


Cdipyridamole/TBA (mg/mL) Cmannitol/water (mg/mL) Drug/matrix ratio

25.0 150 10/90

50.0 300 10/90

5.4.2. X-ray powder diffractionXRPD was performed using CuKα radiation with a wavelength of 1.5405 Å at 40 kV and 40 mA from an X’Pert PRO MPD diffractometer (PANalytical, Almelo, The Netherlands). The samples were scanned from 4–60° 2θ with a step size of 0.008° and a time per step of 35 s.

5.4.3. Scanning electron microscopySeM pictures were taken using a jeOL jSM 6301-F microscope (jeOL, japan). The samples were was dispersed on top of double-sided sticky carbon tape on metal disks and coated with a thin layer of gold/palladium in a Leica EM SCD 050 sputtering device. The acceleration voltage used was 4.0 kv.

5.4.4. Surface Brunauer, Emmet, and Teller-isotherm determinationBefore analysis, the powder samples were filled into ¼ inch sample tubes and dried with nitrogen gas using a vacPrep 061 (Micromeritics, Norcross, U.S.A.). Thereafter, the average surface are was determined using a Tristar 3000, Micromeritics (Norcross, U.S.A.) BeT.

5.4.5. Laser diffractionThe particle size of the pure dipyridamole was determined by laser diffraction. Small quantities dipyridamole were dispersed by using a RODOS dispersing system at 1.0 bar. Then the geometric particle size distribution was determined with a a Sympatic HeLOS compact model KA laser diffraction apparatus (Sympatec GmbH, Clausthal-Zellerfeld, Germany). A 100-mm lens was used and calculations were based on the Fraunhofer theory.

5.4.6. Confocal Laser Scanning microscopyCLSM images were obtained using a Leica TCS SP2 microscope with an Argon laser (excitation wavelength 458 nm). HCX PL APO CS 63x 1.4 (for the freeze-dried samples) and 40x 1.25 (for the physical mixture) oil objectives were used. Both dipyridamole and mannitol were excited at 458 nm. To obtain the fluorescent image of the dipyridamole crystals, emission was measured at a wavelength of 482-551 nm. To obtain a brightfield reflection image of complete powder, emission was detected at a wavelength of 449-471 nm. To reduce the background noise, the average of six scans was used to produce one image. The digital images were recorded in a 2048 x 2048 resolution, which resulted in a pixel size of approximately 183 x 183 nm for the images taken with the 40x objective lens and 50 x 50 nm for the images taken with the 63x objective lens. Samples were prepared by using a spatula to spread the powders onto a cover glass. Only the cover glass was used during the CLSM measurement.

62

chapter 5

5.4.7. Image analysisThe obtained fluorescent CLSM pictures (Fig. 5.6A; p. 68) were first deconvoluted using the Huygens Professional 3.6 software from Scientific Volume Imaging (Fig. 5.6B). ImageJ 1.41n (National Institutes of Health, USA) was used to change the deconvoluted images into binary images (Fig. 5.6C) and to measure the surface area of the particles (Fig. 5.6D). To exclude the counting of single pixels, originating from noise, as particles, only particles with a surface area larger than 0.13 µm2 or 0.01 µm2 for the images taken with the 40x objective lens or the 63x objective lens, respectively, were measured. Since the total surface area of particles at the edge can not be determined, these particles were excluded from analysis. The average surface area of each individual particle was measured. Finally, the volume median diameters were calculated by assuming that the particles were spherical.

5.4.8. TablettingThe obtained powders were compressed into 9 mm round and flat tablets having a weight of 100 mg on a eSH compaction apparatus (Hydro Mooi, Appingedam, The Netherlands). The used compaction rate was 5 kN/s and the maximum compaction load was 5 kN. The tablets were stored in a vacuum desiccator over silica gel at room temperature for at least one day before further processing.

Fig. 5.1. X-ray diffraction patterns of δ-mannitol, dipyridamole as received, and the dispersions that

were prepared freeze-drying from solutions containing 25.0 and 50.0 mg/mL dipyridamole in TBA. In the

patterns of the solid dispersions, the major dipyridamole peak at 8.1º 2θ is clearly visible, but also minor

dipyridamole peaks at 8.9, 10.3, 18.8, 20.9, 23.5, and 26.1º 2θ were found. Both dispersions contained 10%

w/w dipyridamole in mannitol.

63


5.4.9. DissolutionThe dissolution behavior of dipyridamole from the tablets was determined by using a USP dissolution apparatus II (Sotax AT 7, Basel, Switzerland). As dissolution medium 1 L of a pH 6.8 phosphate buffer with 0.05% w/v sodium dodecyl sulphate at a temperature of 37 °C was used. The paddle speed was set at 100 rpm. The concentration of dipyridamole was measured spectrophotometrically (Evolution 300 UV–VIS spectrophotometer, Thermo Fisher Scientific, Madison, U.S.A.) at a wavelength of 285 nm.

5.5. Results

5.5.1. Physicochemical characterization of the dispersionsTo determine the crystallinity of the obtained dispersions, XRPD was used. Differential scanning calorimetry, which a common technique to quantify crystallinity, could not be used because the melting point of dipyridamole and mannitol are almost equal to each other (163 ºC and 166 ºC respectively [24; 60]). Consequently, the peaks in the thermogram overlap and therefore the degree of crystallinity could not be determined. However, the XRPD patterns clearly revealed the major dipyridamole peak at 8.1º 2θ, other smaller dipyridamole peaks, and peaks corresponding to δ-mannitol (Fig. 5.1). This indicates that the obtained dispersions

Fig. 5.2. SEM pictures of two controlled crystallized dispersions. Both contain 10% w/w dipyridamole

in mannitol, but one was prepared from a solution containing 25.0 mg/mL dipyridamole in TBA (top),

while the other was prepared from a solution containing 50.0 mg/mL dipyridamole in TBA (bottom). Both

pictures have a magnification of 5000x.

64

chapter 5

consisted of crystalline dipyridamole and δ-mannitol.

The SeM pictures show that the particle size of the particles formed during the CCDF-process increased, when the concentrations dipyridamole in TBA and mannitol water were increased (Fig. 5.2). The BeT results showed that the particles obtained from the less concentrated solutions had a larger surface area (4.6 m2/g) than the particles obtained from the more concentrated solutions (3.7 m2/g). These results agree with the SEM findings and are an indication that the particle size is smaller when the samples are freeze-dried from less concentrated solutions.

5.5.2. Comparison between laser diffraction and CLSM

To validate whether CLSM can be used to determine the particle size of the autofluorescent drug while the drug is mixed with a second component, the pure drug crystal size as determined by laser diffraction was compared with the dipyridamole crystal size in a physical mixture as determined by CLSM. The fluorescent dipyridamole signal and the non-fluorescent mannitol signal could easily be separated from each other by CLSM (data not shown). The particle size of the pure drug as determined by laser diffraction was similar to the size of the unprocessed drug in the physical mixture as determined by CLSM (Fig. 5.3). The most frequently used value to indicate the particle size, the D50 value (50% of the particles smaller than the given size with respect to the volume distribution), was found to be 22 and 24 μm by laser diffraction and

Fig. 5.3. A comparison between the particle size obtained by CLSM (black bars) and by laser diffraction

(grey bars). The particle size obtained by CLSM is the mean from 50 pictures, while the particle size

obtained by laser diffraction is the mean of 3 measurements.

65


CLSM, respectively, for the unprocessed drug. These results indicate that CLSM in combination with the image analysis, can be used to determine the particle size of an autofluorescent drug, even though it is mixed with a second component.

5.5.3. Particle size analysis and dissolution

The two previously described dispersions were used to test whether CLSM can also be used as analytical tool to determine the size of fluorescent drug crystals solid dispersions. Also in these samples, the dipyridamole and mannitol could be clearly distinguished from each other by CLSM and the fluorescent dipyridamole signal could be separated from the mannitol signal (Fig. 5.4). After deconvolution, the grey-scale images were converted into binary images and from these images the volume based particle size distribution was calculated.

The calculated particle size distributions do not only show a clear difference in particle size between the physical mixture and the crystallized dispersions, but also between the two crystallized dispersions (Fig. 5.5A). The D50 of dipyridamole in the physical mixture was found to be 24 μm, while the size was 0.9 and 1.5 μm for the crystals in the dispersions prepared from the less concentrated and more concentrated solutions, respectively.

Fig. 5.4. Examples of overlay of the reflection and fluorescent images (left), fluorescent images (middle),

and binary images (right) of the controlled crystallized dispersions obtained from the solution containing

25 mg/mL (top) and 50 mg/mL (bottom) dipyridamole in TBA. Both controlled crystallized dispersions

contain 10% w/w dipyridamole in mannitol.

66

chapter 5

Fig. 5.5. A) Particle size distribution of the crystalline dispersions obtained from the 25 mg/mL (light

grey) mg/mL and 50 (dark grey) dipyridamole in TBA solution and the physical mixture (black).

B) Dissolution profiles of dipyridamole from tablets composed of the physical mixture (●) and controlled

crystallized dispersions. The dispersions were prepared from a containing 25 mg/mL (○) and 50 mg/mL

(●) dipyridamole in TBA. The D50-values in the figure are the values as determined in (A). All tablets

contained 10% w/w dipyridamole (n = 3-6; mean).

67


Although, dissolution experiments can not be used to quantify the drug crystal size, differences between drug crystal sizes are reflected in differences in dissolution rate. The dissolution experiments showed that the dissolution rate of dipyridamole was higher from the controlled crystallized dispersions than from the physical mixture (Fig. 5.5B). Furthermore, it can be seen that the dissolution rate was faster when the crystalline dispersion was prepared from the less concentrated solutions. Since all samples were crystalline, a higher dissolution rate indicates smaller crystals. Therefore, it can be concluded that the CLSM findings, which show that the dipyridamole crystals from the less concentrated solutions are the smallest, are in agreement with the results of the dissolution experiments.

5.6. Discussion

In this study we showed that CLSM, in combination with a two step image analysis process, can be used as a tool to determine the particle size of autofluorescent drug crystals in the solid state. To validate this, first the pure dipyridamole crystal size was determined by laser diffraction analysis. The found crystal size was compared to the size of the same drug crystals physically mixed with mannitol as determined by CLSM. Both methods showed a similar particle size distribution, indicating that the combination of CLSM and image analysis can be used to determine the crystal size of dipyridamole that is mixed with a second component.

In a previous study [29], we developed CCDF as novel process to prepare drug nanocrystals. In another study [51], we showed that the size of the drug crystals was limited to the size of the interstitial spaces between the frozen solvent crystals. We hypothesized in this study that when less concentrated solutions are freeze-dried, more solvent crystals and therefore smaller interstitial spaces are formed and consequently smaller drug crystals will be formed.

The results of the BeT measurements suggested that the particles obtained from the less concentrated solutions were indeed smaller than the particles obtained from the higher concentrated solutions. Furthermore, the SeM pictures suggested that the particles obtained from the less concentrated solutions were of sub-micron size, while the particles from the more concentrated solutions were larger than a micrometer. However, both methods can not distinguish the drug from the matrix.

The CLSM analysis on the other hand can distinguish between drug and matrix and it can be used to quantify the drug crystal size in the crystalline dispersion. It showed that the crystals in the dispersion prepared from the less concentrated solution were of sub-micron size, whereas the crystals in the dispersion prepared from the more concentrated solution were approximately 1.5 μm.

This trend in drug crystal size was in agreement with the dissolution behavior of the tablets prepared from the physical mixture and the freeze-dried products. Since the pure drug, but also both freeze-dried dispersions, are all crystalline, differences in dissolution rate can be ascribed to differences in drug crystal size. Therefore, the dissolution tests showed that the

68

chapter 5

dipyridamole crystals obtained from the less concentrated solutions were indeed smaller than the dipyridamole crystals from the more concentrated solutions.

These results indicate that CLSM is a useful technique to determine the drug crystal size in a powder mixture. The fact that in the crystallized dispersions sub-micron sized particles were analyzed and in the physical mixture particles of several tens of micrometers, indicates that the CLSM method can be used to measure the particle size over a size range covering several orders of magnitude. Furthermore, the particle size of only one of the components in a mixture or dispersion, instead of that of the combination, can be determined. A limitation of the method is that it is limited to autofluorescent drugs. However, when the drug of interest is autofluorescent, CLSM could be a valuable tool.

5.7. Conclusion

In conclusion, CLSM can be used to determine the drug crystal size of an autofluorescent drug in a powder mixture independent from the mode of incorporation. Advantages are that it can be used to determine the drug crystal size in solid state and that the size of one component in a mixture or dispersion can be determined. Furthermore, CLSM can be used to determine the

Fig. 5.6. A typical example of the image analysis as applied to an image of dipyridamole as received. The

original fluorescent image (A) was deconvoluted (B), then made binary (C), and finally surface area of each

individual particle was measured (D). Note that particles at the edge of the image (for example two particles

in the bottom left corner) are omitted from the analysis and are therefore not shown on (D).

69


size of crystals of several hundreds of nanometers to tens of micrometers. A limitation is that this method can only be applied to autofluorescent compounds. In addition to this, we showed that the drug crystal size in the crystalline dispersion, prepared by CCDF, could be controlled by the concentration of drug and matrix in the initial solution.

chapterevaluation and optimization of a force field for crystalline forms of mannitol and sorbitol 6

72

chapter 6

6.1 Abstract

Two force fields, the GROMOS53A5/53A6 (united atom) and the AMBER95 (all atom) parameter sets, coupled with partial atomic charges derived from quantum mechanical calculations were evaluated for their ability to reproduce the known crystalline forms of the polyols mannitol and sorbitol. The force fields were evaluated using molecular dynamics simulations at 10 K (which is akin to potential energy minimization) with the simulation cell lengths and angles free to evolve. Both force fields performed relatively poorly, not being able to simultaneously reproduce all of the crystal structures within a 5% deviation level. The parameter sets were then systematically optimized using sensitivity analysis and a revised AMBeR95 set was found to reproduce the crystal structures with less than 5% deviation from experiment. The stability of the various crystalline forms for each of the parameter sets (original and revised) was then assessed in extended MD simulations at 298 k and 1 bar covering 1 ns simulation time. The AMBeR95 parameter sets (original and revised) were found to be effective in reproducing the crystal structures in these more stringent tests. Remarkably, the performance of the original AMBeR95 parameter set was found to be slightly better than that of the revised set in these simulations at 298k. The results of this study suggest that, whenever feasible, one should include molecular simulations at elevated temperatures when optimizing parameters.

H. de Waarda, A. Amanib,c, j. kendrickb, W.L.j. Hinrichsa, H.W. Frijlinka, and j. Anwarb

a Department of Pharmaceutical Technology and Biopharmacy, University of Groningen, Antonius Deusinglaan 1, 9713 AV, Groningen, the Netherlands. b Computational Biophysics Laboratory, Institute of Pharmaceutical Innovation, University of Bradford, West Yorkshire, BD7 1DP, United Kingdom. c Department of Medical Nanotechnology, School of Advanced Medical Technologies, Tehran University of Medical Sciences, Tehran, Iran.

Reproduced with permission from [j. Phys. Chem. B 2010. 114(1), 429-436] Copyright 2010 American Chemical Society.

73

evaluation and optimization of a force field for crystalline forms of mannitol and sorbitol

6.2. Introduction

Polyols, or sugar alcohols, such as erythritol, isomalt, lactitol, sorbitol, mannitol, maltitol, and xylitol, are important components in many food, confectionery, and pharmaceutical products. Within the food industry, polyols are commonly used as sugar replacers, since they have a lower caloric content than sugars and they do not contribute to tooth decay. Additionally, due to the high negative heat of solution, inclusion of polyols like mannitol and sorbitol give a cooling sensation in products like, for example, chewing gum [62]. The use of polyols in pharmaceuticals is widespread [63]. Mannitol, for example, is employed as a diluent (to increase the overall mass of a tablet containing a potent drug) and as a direct-compression aid in tablet formulations [64-65], in wet granulation, as a plasticizer in soft gelatin capsules, a carrier in dry powder inhalers, and as a bulking agent or for its possible lyoprotectant effect in freeze-dried products [66-69]. In addition to these applications, it can be administered parenterally for its therapeutic role in reducing the intracranial pressure or for its function as a osmotic diuretic. Recently, we described the use of mannitol for stabilizing nanocrystals of pharmaceutically active substances [29; 45]. In this application, mannitol forms the matrix into which the drug nanocrystals are embedded. Nanocrystals of poorly water soluble pharmaceuticals, because of their immense surface area, promise to significantly enhance the bioavailability of poorly soluble compounds [18-19]. Sorbitol (also called glucitol) can be employed as an alternative to mannitol, but because of its rather different physical properties it also has other niche applications that include its use

Fig. 6.1. Molecular structures of mannitol and sorbitol showing atom indexing for mapping force field

atom types: (a) mannitol (AMBER95 force field), (b) mannitol (GROMOS53A5/53A6 force field), (c)

sorbitol (AMBER95 force field), and (d) sorbitol (GROMOS53A5/53A6 force field).

74

chapter 6

as a humectant in oral solutions, in oral suspensions, in toothpastes, and in dermal products [63]. A notable feature of many of the above applications is that the polyols are exploited for their solid state properties.

Mannitol and sorbitol are stereoisomers of hexan-1,2,3,4,5,6-hexol (HO-CH2-CHOH-CHOH-CHOH-CHOH-CH2OH) (see Fig. 6.1) and are known (or claimed) to exist in numerous crystalline polymorphs. Although the two chemical structures are similar, their physical properties differ markedly. For example, the melting points of the mannitol polymorphs vary over the range 155-166 °C, while the corresponding range for sorbitol is 93-112 °C. Furthermore, mannitol has a lower affinity for moisture, compared with the highly hygroscopic sorbitol. Mannitol is claimed to exist in at least eight polymorphic forms. Burger et al.[24] have extensively reviewed these claims and have reduced the number of anhydrous polymorphs to three: α-, β-, and δ-mannitol. The β-polymorph, which is the main commercially available product, is the thermodynamically stable form at ambient conditions. The δ-polymorph shows considerable kinetic stability, since it has been observed to show no transformation when subjected to mechanical stress or on storage (dry) for more than 5 years at 25 ºC [24]. For sorbitol even more polymorphs have been identified at room temperature: five anhydrous (Α-, Β-, Δ-, Ε-, and Γ-sorbitol) and 2 hydrated (sorbitol hydrate I and II) crystal forms. However, the structures of only three of the anhydrous forms, namely Α-sorbitol [70], E-sorbitol [71], and Γ-sorbitol [72], have been determined. A key structural variation between the various polymorphic forms of both sorbitol and mannitol tends to be in the hydrogen bonding, more specifically in terms of the choice of whether a particular OH-group engages intra- or inter-molecularly [70-74].

In addressing fundamental and technological issues in materials behavior and function and solid state chemistry, molecular simulations are increasingly adding value to experimental studies. Molecular simulations can provide molecular-level insights and enable estimation/prediction of thermodynamic quantities [75-77]. The reliability of these simulations depends on the quality of the input parameters, namely the force field that characterizes the interaction forces between the atoms. Despite the immense utility of both mannitol and sorbitol and the need to understand the molecular interactions of these materials with drug substances (including therapeutic peptides and proteins) and other formulation excipients, there are hardly any molecular simulations studies involving mannitol and sorbitol, particularly in the solid state. A possible obstacle is likely to be the reliability of the currently available force field parameters which have neither been tested nor optimized to reproduce the various crystalline forms of both sorbitol and mannitol. Indeed the accuracy of forcefield parameters is at the heart our ability (or inability, depending on one’s perspective) to predict crystalline structures from first principles using intermolecular potentials [78].

In this study we have tested two force fields, GROMOS53A5/53A6 (united atom) [79] and AMBeR95 (all-atom) [80], for their ability to reproduce the crystal structures of the known polymorphs of both mannitol and sorbitol. Both force fields in their original form were found to perform poorly in reproducing all of the experimental crystal structures using low temperature (10 k) molecular dynamics simulations, which is akin to potential energy minimization. We attempted to optimize the force fields, with the critical parameters being determined by

75


sensitivity analysis. We were able to improve on the AMBeR parameters so as to reproduce the various structures to less than 5% deviation in the cell parameters, but were unable to make any progress with the GROMOS parameters. Remarkably, on validation of the parameters in molecular dynamics simulations at room temperature, the original AMBeR parameters were found to perform slightly better than the optimized AMBeR parameters (optimized at 10 k). The study illustrates the challenges of optimizing potential parameters for systems where the potential energy surface for the molecular interactions is characterized by many potential energy minima of similar depths, as typified by the various crystalline forms of mannitol and sorbitol in which a particular OH group can engage in intra- and intermolecular hydrogen bonding simply by rotation about a bond. The study suggests that optimization of potential parameters for crystalline phases should also include molecular simulations at the relevant temperatures and pressures and not only 0 k potential energy minimization, which should now be feasible given the increase in available computational power and the highly parallel nature of the problem.

6.3. Methods

6.3.1. Determination of partial chargesA single set of derived charges was employed for both forcefields that were evaluated. To calculate the atomic partial charges, three conformationally distinct molecules each of both mannitol and sorbitol were taken from the crystal structures (DMANTL08, DMANTL09, and DMANTL10 for α-, β-, and δ-mannitol respectively [81] and GLUCIT01 [70], GLUCIT02 [71], and GLUCIT03 [72] for Α-, E-, and Γ- sorbitol respectively), and optimized using the General Atomic and Molecular electronics Structure System-Uk (GAMeSS-Uk) ab initio package [82] using the hybrid B3LYP density functional method with a 6-31G* basis set. The partial charges were calculated from the final electron densities using the electrostatic potential fitting method [83]. The fitting produced a single set of charges that provided the best fit to the electrostatic potential for all six molecular structures simultaneously, thus encapsulating any variation in the partial charges due to molecular conformation.

6.3.2. Evaluation of force fieldsThe two force fields, GROMOS53A5/53A6 (united atom) [79] and AMBER95 [80] parameter sets, (see Tables 6-I and 6-II) were evaluated by means of molecular dynamics simulations at 10 k for their ability to reproduce the known crystal structures of the anhydrous polymorphs of both mannitol and sorbitol. At this low simulation temperature the procedure is essentially equivalent to potential energy minimization, and hence very much in the same spirit as the conventional approach for testing force fields. The dynamics simulations were carried out using the DL_POLY 2.18 package [84] in a NσT ensemble with the external pressure set to 1 bar and temperature and pressure coupling constants of 0.1 and 1.0 ps respectively [85]. The NσT ensemble allows cell angles and cell lengths to change independently as a result of stress variations within the cell. The only restriction on the molecule internal degrees of freedom was that the bonds were constrained using the SHAke algorithm. The electrostatic interactions were calculated using ewald summation with a precision of 10-5. The cutoffs for the van der

76

chapter 6

Table 6-I. Mapping of the GROMOS53A5/53A6 [65] parameters to the mannitol and sorbitol molecules.

Non-bonded terms

Atoma Atom types A (kj mol-1 Å12)b B (kj mol-1 Å6)b

1-6 OA 1210000c/1505529d 2093.98

7, 12 CH2 33965584 7468.42

8-11 CH1 97022500 6068.41

13-18 HO 0 0

Bonded terms

Bonds Length (Å)

OA—CH2 1.43

OA—CH1 1.43

OA—H 1.00

CH2—CH1 1.53

CH1—CH1 1.53

Angles kθ (kj mol-1) θ0 (deg)

OA—CH2—CH1 520 109.5

OA—CH1—CH2 520 109.5

OA—CH1—CH1 520 109.5

CH2— OA—H 450 109.5

CH2—CH1—OA 520 109.5

CH2—CH1—CH1 520 109.5

CH1—OA—H 450 109.5

CH1—CH1—OA 520 109.5

CH1—CH1—CH1 520 109.5

Dihedralse kφ (kj mol-1) δ m

X—CHn— CHn —X 5.92 0 3

X—CHn— OA —X 1.26 0 3

a Atom indices correspond to those given in figure 6.1.

b c Used for OA-CH2, OA-CH1, and OA-H nonbonded interactions. d Used for OA-OA nonbonded

interactions. e For any bond between two atoms (j and k), only one set of atoms (i,j,k,l) is chosen that

define the dihedral angle.

77


Table 6-II. Mapping of the AMBeR95 [66] parameters to the mannitol and sorbitol molecule.

Non-bonded terms

Atoma Atom types εb (kcal mol-1) σb (Å)

1-6 OH 0.2104 1.5332

7-12 CT 0.1094 1.6998

13-18 HO 0.0000 0.0000

19-26 H1 0.0157 1.2357

Bonded terms

Bonds Length (Å)

OA-CT 1.410

OA-HO 0.960

CT-CT 1.526

CT-H1 1.090

Angles kθ/2(kcal mol-1) θ0 (deg)

OA-CT-CT 50.0 109.5

OA-CT-H1 50.0 109.5

CT-OA-HO 55.0 108.5

CT-CT-OA 50.0 109.5

Ct-CT-CT 40.0 109.5

CT-CT-H1 50.0 109.5

H1-Ct-H1 35.0 109.5

dihedrals vn/2 (kcal mol-1) δ m

X-CT-CT-X 0.156 0.0 3

X-CT-OH-X 0.167 0.0 3

a Atom indices correspond to those given in figure 6.1.

b

78

chapter 6

Waals interactions and the real space ewald were both 1.2 nm. The time step was 0.002 ps and the total simulation time was 50 ps. All simulations converged relatively rapidly, and the cell angles and cell lengths were averaged over the final 30 ps.

6.3.3. Sensitivity analysis and optimization of potential parametersBoth the GROMOS and the AMBeR parameters failed to reproduce the various crystal structures of sorbitol and mannitol simultaneously. In view of this, we proceeded to optimize each of the parameter sets using an approach based on sensitivity analysis, which in an earlier study proved to be very effective in optimizing potential parameters for crystalline phases [86]. Sensitivity analysis involves investigating the significance of the effect of a small change in each of the parameters, and hence enables the identification of the critical parameters which can then be optimized. Here we examined the effect of a 5% change in each of the potential parameters on the lattice parameters of the various phases in molecular dynamics simulations at 10 K as detailed above. This enabled the identification of those parameters that had the most effect in reproducing the lattice parameters of the crystalline phases. The optimum values of the identified critical parameters were then identified by varying these parameters in isolation in small increments.

The influence of the various van der Waals parameters for both force fields was determined by investigating the effect of a 5% increment in each of the homo-parameters (i.e. the A and B values for GROMOS, and ε and σ values of the fully atomistic AMBER force field) in isolation. For this study the hetero-parameters were calculated from the homo-parameters using standard mixing rules. A secondary study explored the effects of a 5% increment of the individual hetero-parameters, which of course makes the model more flexible but increases the parameter space considerably.

We also considered the force constants for the dihedral energy barriers as parameters that could benefit from optimization. This is unusual but necessary for the two molecules mannitol and sorbitol since these force constants are expected to play an important role in determining the hydrogen bonding interaction of the OH groups. One could attempt to calculate the specific force constants for the two molecules from first principles by characterizing the potential energy surface for bond rotation about the dihedrals. However, such an approach still requires the removal of the electrostatic and van der Waals contributions of the 1-4 and above interaction from the potential energy (Pe) surface, which may not be straight forward given that the van der Waals interactions themselves are not fixed and are the subject of optimization. In view of this we proceeded with the original torsional force constants (which, in general, would have been determined from first principles and optimized, albeit for other simpler molecules) and considered them as parameters for optimization. Thus the effect of 10, 20, and 40% variations in the force constants was examined for both the AMBER and GROMOS force fields. The van der Waals parameters were kept at their original values while the torsional force constants were changed.

6.3.4. Validation: molecular dynamics simulations of mannitol and sorbitol crystalsThe quality of the original and optimized parameter sets was tested in molecular dynamics

79


simulations at ambient conditions. Thus, the stability of crystals of α-mannitol (8x4x2 unit cells; 256 molecules), β-mannitol (7x4x2 unit cells; 224 molecules), δ-mannitol (8x2x7 unit cells; 224 molecules), Α-sorbitol (4x4x4 unit cells; 256 molecules), E-sorbitol (7x4x2 unit cells; 224 molecules), and Γ-sorbitol (2x2x8 unit cells; 384 molecules) was investigated in extended dynamics simulations performed in a NσT ensemble [85] at 298K and 1 bar with a temperature and pressure coupling of 0.1 and 1.0 ps respectively, using the same simulation parameters as specified in section 2.2 above. Each of the simulations accessed 1 ns simulation time.

6.4. Results and analysis

6.4.1. Partial chargesThe partial atomic charges determined from simultaneously fitting the electrostatic potentials of the various molecular structures found in the polymorphs of anhydrous polymorphs of mannitol and sorbitol are shown in Table 6-III. The standard deviation of the electrostatic potential generated by this point charge model from the quantum-mechanical electrostatic potential, was similar for all conformations and smaller than 0.009 (atomic units). The best fits to the electrostatic potential of fitting each molecule individually are all above 0.0077, while using a single charge assignment for all molecules gives values of less than 0.0084. For the united-atom force field, the charges on the aliphatic hydrogen were added to the charges of the carbons to which they are bonded, giving a total charge for the united atom.

6.4.2. Evaluation and sensitivity analysis of the GROMOS force fieldFor the GROMOS force field, the simulated crystal structures showed deviations in the cell lengths of up to 6% from the experimental values (Table S1, Supporting Information), with the b-axis of the β-polymorph of mannitol being the most problematic. In contrast, the cell angles of all crystals were found to deviate less than 0.5% from the experimental values. Sensitivity analysis of the parameters revealed that the GROMOS force field was relatively robust with respect to the dihedral energy force constants, variations in which did not cause significant changes in the lattice parameters and neither induced the desired simultaneous overall improvement in the crystal lattices. Changes in the values for the dihedral energy barrier would therefore result in the improvement of some of the lattice parameters, but compromise others. The A and B van der Waals homo-parameters for the atom (united) types CH1 (aliphatic CH1-group) and CH2 (aliphatic CH2-group) did not appear to affect the lattice parameters much. In contrast, the lattice parameters were highly sensitive to variations in both the A and B parameters for the atom type OA (the hydroxyl oxygen atoms), but the effect was never beneficial all round for reproducing all of the crystalline phases. Although the HO (hydroxyl hydrogen) van der Waals parameters in the original force field are set to zero, their influence was also examined. Variations in the HO B-parameter had a significant and positive effect toward reproducing the various crystalline lattices. The B parameter, however, represents the attractive component of the Lennard jones potential and giving this a value whilst leaving the A parameter set at zero does not yield a proper Lj potential, and hence this course was rejected, being considered to be rather ad hoc. The analysis, however, clearly indicates that the hydroxyl descriptions are critical.

80

chapter 6

Table 6-III. Atomic partial charges calculated from electrostatic potential fitting. The atom indices

refer to the numbers given in Fig. 6.1a and 6.1c for the AMBER force field and Fig. 6.1b and 6.1d for the

GROMOS force field.

AMBER force field GROMOS force field

Atom Atom type Atomic partial charge (e) Atom Atom type Atomic partial charge (e)

1 OH -0.582 1 OH -0.582

2 OH -0.590 2 OH -0.590

3 OH -0.590 3 OH -0.590

4 OH -0.590 4 OH -0.590

5 OH -0.590 5 OH -0.590

6 OH -0.582 6 OH -0.582

7 CT 0.153 7 CH2 0.211

8 CT 0.199 8 CH1 0.217

9 CT 0.199 9 CH1 0.217

10 CT 0.199 10 CH1 0.217

11 CT 0.199 11 CH1 0..217

12 CT 0.153 12 CH2 0.211

13 HO 0.373 13 HO 0.373

14 HO 0.371 14 HO 0.371

15 HO 0.371 15 HO 0.71

16 HO 0.371 16 HO 0.371

17 HO 0.371 17 HO 0.371

18 HO 0.373 18 HO 0.373

19 H1 0.029

20 H1 0.029

21 H1 0.018

22 H1 0.018

23 H1 0.018

24 H1 0.018

25 H1 0.029

26 H1 0.029

a The atom indices refer to the numbers given in Figure 6.1a and 6.1c for the AMBER force field and

Figure 6.1c and 6.1d for the GROMOS force field.

As our attempts to optimize the homo-parameters proved ineffective, we proceeded to investigate the sensitivity of the lattice parameters to variations in the individual hetero-parameters, which serve as a more flexible model. Modifying any of the interactions, with exception of the OH – HO (oxygen – hydrogen) interaction, did not result in any significant and/or overall beneficial change in the lattice parameters. Increasing the A-parameter (the repulsive component) of this interaction had a strong detrimental effect on all the cell lengths (Table S2, Supporting Information). There was no option to reduce this parameter since it is

81


already set at zero (the lowest possible value with a physical basis) in the original force field. Therefore became apparent that there was no scope for optimizing the GROMOS force field for the various crystalline forms of mannitol and sorbitol.

6.4.3. Evaluation and sensitivity analysis of the AMBER95 force fieldAttempts to reproduce the crystalline forms of mannitol and sorbitol using the AMBeR95 force field revealed δ-mannitol as the problem phase, for which both the β-angle (by almost 12 %) and the a-axis (by over 8%) were underestimated (Table S3, Supporting Information). Sensitivity analysis with respect to the dihedral energy force constants (Table S4, Supporting Information) showed that whilst the lattice parameters were sensitive to variations in the force constants, any given change in the force constants did not consistently improve the crystal structure predictions.

Table 6-IV. Optimization of the LJ σ-parameter of the hydroxyl oxygen atom (atom type OH) of the

AMBER95 force field.

Inte

ract

ion

Incr

ease

(%)

α-mannitol β-mannitol δ-mannitol

a (%) b (%) c (%) a (%) b (%) c (%) a (%) b (%) c (%)

OH – OH σ 0 -0.77 -1.78 -0.89 0.24 -0.15 -2.16 -8.37 2.17 -4.81

2 0.55 -1.11 -0.90 1.87 0.43 -2.82 -6.75 2.74 -0.37

3 1.26 -0.77 -0.89 2.72 0.74 -3.11 -5.91 2.96 -4.31

4 1.97 -0.39 -0.88 3.55 1.14 -3.39 -5.34 -1.27 0.13

5 2.68 -0.08 -0.87 4.38 1.43 -3.58 -4.35 -1.29 0.43

A-sorbitol e-sorbitol Γ-sorbitola (%) b (%) c (%) a (%) b (%) c (%) a (%) b (%) c (%)

OH – OH σ 0 -1.66 2.63 -2.57 -1.69 -0.55 0.27 0.25 0.03 -1.97

2 -0.92 3.70 -1.82 -0.05 -0.37 0.78 0.65 -0.60 0.68

3 -0.58 4.20 -1.33 0.71 -0.26 1.02 1.02 -0.87 1.05

4 -0.19 4.75 -0.92 1.16 -0.22 1.25 1.13 -1.06 2.19

5 0.12 5.24 -0.35 2.44 -0.16 1.44 1.33 -1.12 3.00

The table gives the % deviation in the cell lengths of the simulated crystal structure from the experimental

values as the LJ σ-parameter of the hydroxyl oxygen atoms is incremented.

With respect to the van der Waals homo-parameters, these too did not offer an option for optimization. It was found that changing the σ-parameter of the OH-atom had the strongest influence on the cell dimensions with an increase of 5% of this value resulting in a strong improvement in the cell dimensions of δ-mannitol. However, at the same time this variation was detrimental for the cell dimensions of the other structures e.g. β-angle of A-sorbitol. Sensitivity analysis of the van der Waals hetero-parameters revealed that a variation in the σ-parameter of the OH-OH interaction could considerably improve the predicted lattice

82

chapter 6

parameters, in particular eliminating the very large deviations (Table S5, Supporting Information). However, the effects of this parameter were not wholly beneficial as some of the lattice deviations, which previously were low, were now larger. In view of this, the σ-parameter of the OH – OH interaction was optimized by investigating the effect of stepwise increments of 1% up to a maximum 6%. The 4% increased value of the σ-parameter appeared to be optimal in characterizing all of the crystal forms of mannitol and sorbitol at 10k (Table 6-Iv). This increase in the σ-parameter represents a direct increase in the effective diameter of the hydroxyl oxygen atoms. For this optimal parameter set the lattice energies of the various forms of mannitol and sorbitol were -156.7, -161.6, -153.6 for α-, β- and δ-mannitol respectively, and -165.6, -158.6, -145.2 for A-, E- and Γ-sorbitol respectively. Unfortunately there are no experimental data against which these values can be compared.

6.4.4. Molecular dynamics simulation at ambient conditionsThe quality of the optimized as well as the original AMBeR parameters and original GROMOS parameters (which could not be optimized any further) was tested in extended molecular dynamics simulations at 298k and 1 bar pressure i.e. ambient conditions for each of the polymorphic forms of mannitol and sorbitol. In principle only one of the polymorphic forms of each of the polyols is thermodynamicallystable at ambient conditions, and an expectation may be that the other polymorphic forms would transform to the stable polymorphic form. Whilst first-order phase transformations are indeed accessible in molecular dynamics simulations, these are largely restricted to systems with low activation energy barriers, which in the laboratory occur relatively rapidly [87-88]. Systems with strong hydrogen-bonding tend to be relatively sluggish and require in excess of tens of nanoseconds of simulation time unless coerced using a biased potential method like metadynamics [89]. In view of these considerations it was anticipated that the strongly hydrogen-bonded mannitol and sorbitol structures, which in the laboratory are kinetically stable for extended periods of time, were unlikely to undergo any transformation during the limited 1 ns simulations.

The results from the MD simulations, that is, the averaged lattice parameters, deviations in the lattice parameters from experiment, and the atom-based root mean squared deviations from the experimental structures are presented in Tables 6-v and 6-vI. The data (Table 6-v) clearly reveal that the GROMOS parameters whilst being just outside the limit of respectability in the 10 k simulations are extremely poor in characterizing the various crystalline forms at ambient conditions. An extreme example is the marked deviation in the c-axis of δ-mannitol, which is overestimated by 42%. In contrast, the AMBER parameters, both original and optimized, fare much better (Table 6-v) and are able to reproduce the various crystalline structures reasonably well, within the limits of 6%. Remarkably, the original AMBER parameters (see Fig. 6.2) result in a slightly better representation of the crystals than the optimized AMBeR parameters (see Fig. 6.3).

83


Table 6-V. Deviation (%) in the cell lengths of the simulated crystal structure from the experimental

values for various polymorphic forms of mannitol and sorbitol averaged for the latter part of a molecular

dynamics simulation trajectory at 298k and 1 bar pressure.


a (%) b (%) c (%) a (%) b (%) c (%) a (%) b (%) c (%)

GROMOS 10.71 6.01 -6.19 -0.76 8.79 1.18 10.50 -11.92 42.41

AMBeR95 0.75 -0.04 -0.39 0.89 1.84 -1.21 -5.97 -1.09 0.70

AMBeR optimized 3.80 1.40 -0.27 4.40 3.08 -1.87 -1.97 0.25 5.39

A-sorbitol e-sorbitol Γ-sorbitola (%) b (%) c (%) a (%) b (%) c (%) a (%) b (%) c (%)

GROMOS 1.45 3.34 1.80 6.94 4.14 -2.18 6.88 -1.67 9.88

AMBeR95 0.01 3.54 -1.34 1.26 0.56 -0.04 0.25 1.50 -0.31

AMBeR optimized 1.70 5.74 0.35 5.70 1.41 0.22 1.67 1.92 2.97

Table 6-VI. Root mean square deviation (Å2/atom) of the atomic coordinates of the simulated crystal

structure from the experimental values for various polymorphic forms of mannitol and sorbitol averaged

for the latter part of a molecular dynamics simulation trajectory at 298k and 1 bar pressure.


GROMOS 0.299 0.201 0.239

AMBeR95 0.265 0.187 0.218

AMBeR optimized 0.505 0.231 0.330

A-sorbitol e-sorbitol Γ-sorbitol

GROMOS 0.244 0.199 0.262

AMBeR 0.373 0.300 0.541

AMBeR optimized 0.721 0.268 0.417

In general, the various polymorphs essentially retained their structures in the MD trajectories other than the δ-polymorph of mannitol which showed some changes in its hydrogen-bonding interactions. The hydroxyl group (O3, H15) attached to carbon atom C9 appears to reveal a tendency to form an intramolecular h-bond whilst compromising an intermolecular H-bond interaction (see Fig. 6.4).

6.5. Discussion

The objectives of the study were to identify a set of force field parameters that could simultaneously reproduce all known crystal structures of the various polymorphic forms of the polyols, mannitol and sorbitol. We have tested two force fields, namely the united atom GROMOS53A5/53A6 and the all atom AMBER95 parameter set, and then attempted to optimize them using sensitivity analysis. The worst case GROMOS performance was a 6% deviation in the cell parameters, just putting it outside the respectable 5% limit. The critical parameters

84

chapter 6

Fig. 6.2. Deviation (%) in the cell dimensions between the simulated and experimental structure as a

function of simulation time in a molecular dynamics simulation at 298 k and 1 bar using the original

AMBER95 force field. (a) α-mannitol, (b) β-mannitol, (c) δ-mannitol, (d) Α-sorbitol, (e) Ε-sorbitol, and (f)

Γ-sorbitol; (—) a-axis, (- - - ) b-axis, and (.....) c-axis.

were found to be the van der Waals parameters of the hydroxyl hydrogen atoms, increase in which caused a larger deviation between the calculated and experimental cell parameters. For the GROMOS force field these parameters are in fact set at zero; hence there was little scope for any further optimization.

The AMBeR95 parameter set appeared to perform even worse (maximum deviation for a lattice dimension >8%; for a cell angle > 11%), with the critical parameters as identified by sensitivity analysis being the Lennard-Jones σ-parameter of the OH – OH interaction and the X—CT—OH—X dihedral energy barrier. Optimizing the former enhanced the performance to a respectable level (worst cell deviation just over 5%), whilst the effect of varying the dihedral energy barrier was found to be inconsistent and irregular. We then proceeded to test the force fields, the original GROMOS and AMBeR parameter sets and the optimized AMBeR parameters, in extended molecular dynamics simulations on mannitol and sorbitol crystals at 298 k and 1 bar. Whilst the GROMOS parameter set resulted in almost complete instability in some of the structures, the AMBeR95 parameters performed reasonably well. Remarkably, at 298 k the original AMBeR95 parameter set was a little better than the optimized parameters. The underlying issue of force fields that have been optimized using low temperature simulations performing badly at elevated temperatures is well understood. The optimization process is typically based on only a few crystal structures, limiting the number of molecular orientations that are sampled.

85


Fig. 6.3. Deviation (%) in the cell dimensions between the simulated and experimental structure as a

function of simulation time in a molecular dynamics simulation at 298 k and 1 bar using the optimized

AMBER95 force field. (a) α-mannitol, (b) β-mannitol, (c) δ-mannitol, (d) Α-sorbitol, (e) Ε-sorbitol, and (f)

Γ-sorbitol; (—) a-axis, (- - - -) b-axis, and (…..) c-axis.

Fig. 6.4. Comparison between the experimental structure (A) and a snap shot from molecular dynamics

trajectory at ambient conditions (B) for δ-mannitol showing the slight variation in the hydrogen-bonding

interactions as a result of thermal energy.

At higher temperatures the molecules sample an entire spectrum of orientational interactions of which the optimized parameters are un-informed. The puzzling aspect here is the converse observation that of the AMBeR parameters failing grossly in low temperatures simulations whilst performing well at elevated temperatures. Whilst the cause here is difficult to identify, it is probably pertinent to note that the AMBeR non-bonded parameters were adopted from OPLS [90] and also derived using the OPLS philosophy of fitting parameters to simulations

86

chapter 6

of liquid phases i.e. simulations at elevated temperatures [80]. Despite the original AMBeR95 parameters performing slightly better at 298 k, we propose the revised set (in which only the Lj σ-parameter of the OH – OH interaction is modified) as the definitive parameter set for future application since it has better temperature dependence i.e. from 0 k to ambient temperature.

It is legitimate to ask the question ‘why is it not possible to improve the accuracy of the force field parameters beyond that achieved?’. The general response would be to highlight the difficulty of fitting a complex response surface (which is a function of the atomic coordinates in the crystal and their dependence as a function of time and temperature) with a limited set of parameters. There are however at least a couple more specific issues. Firstly, a point charge description of the electrostatics is unable to describe either the conformational dependence of the charge distribution or the anisotropy of the charge-charge interactions; for greater accuracy one needs to employ approaches like the atomic multipole model [91]. Most molecular simulation codes, however, only contain a point charge description, which restricts the use of atomic multipole-based force fields. A secondary issue is the anisotropy of the repulsion component of the van der Waals interaction [92], though this is not expected to be major problem for the atom types present in the polyols studied. Furthermore, in the present study the atomic partial charges, which were determined from the electrostatic potential, were kept fixed and not subjected to systematic optimization. Whilst this is the accepted philosophy, partial charges obtained from first principles can on occasion fare worse than fitted charges as exemplified by the crystal structure prediction of glycine [93]. However, empirical optimization of the charges could in principle yield a set of charges (which may or may not be intuitively sensible) that reproduce the structures extremely well in the testing protocol but completely fail in a real application. Furthermore, the partial charges are likely to be highly correlated with the van der Waals parameters, particularly the Lennard-Jones well-depth, ε, or equivalent. Because of these considerations we thought it prudent not to optimize the charges.

While the study identifies a parameter set for future simulations of mannitol and sorbitol in the solid state, there are clearly some lessons to be learnt with respect to the approaches commonly employed (and also employed here) for optimizing parameters for the solid state. The usual protocol in optimizing the parameters for the crystalline state is to attempt to reproduce the crystalline structure in a 0 k structural optimization, typically involving potential energy minimization with the lattice parameters, internal symmetry and the atomic coordinates being free to evolve. In the present study we have employed low temperature (10 k) MD simulation with a fully flexible simulation cell, which essentially equates to potential energy minimization since the thermal energy is very low, keeping the system close to the nearest local energy minima. We are, of course, interested in the molecular behavior at ambient temperature and pressure, and the optimized potential parameters are likely to be employed in molecular simulations at these conditions. Furthermore, the crystal structures to which the parameters are being fitted are certainly never determined at 0 k. Our results, which illustrate that a good performance of a force field at 10 K may not per se be reflected in an elevated temperature study, suggest that we should also include the stability assessment of the structures in molecular simulations (i.e. molecular dynamics or Monte Carlo simulation involving reasonable sampling or simulation time) at temperatures and pressures of interest in the optimization iteration cycle. Note that

87


we cannot replace the potential energy minimization entirely by, for example, MD at ambient temperature, since it is very likely that some crystalline structures are likely to diverge from the experimental structure because they happen to be thermodynamically unstable under those conditions. In fact, contrary to what some might expect, obtaining such a result would confirm the accuracy of the employed force field.

Summarizing, we have tested and attempted to optimize the two force field parameters sets, GROMOS and AMBeR95, to model the many crystalline forms of the polyols, mannitol and sorbitol. The AMBER95 parameter set with a modified LJ σ-parameter for the OH – OH interaction coupled with the derived atomic partial charges has been found capable of reproducing all the known crystalline structures in MD simulations at 298k for up to 1 ns. We intend now to employ this parameter set in exploring drug-polyol interfacial properties that underpin the stability of nanocrystals of pharmaceuticals embedded in polyol (crystalline and amorphous) matrices.

Supporting Information Available: Tables showing evaluation and sensitivity analysis of the nonbonded homoparameters and hetero-parameters of the GROMOS53A5/53A6 force field, evaluation and sensitivity analysis of the nonbonded homo-parameters of the AMBER95 force field, and sensitivity analysis of the dihedral energy force constants and the nonbonded hetero-parameters of the AMBER95 force field. This information is available free of charge via the Internet at http://pubs.acs.org.

chapterconcluding remarks and perspectives7

90

chapter 7

H. de Waard, H.W. Frijlink, and W.L.j. Hinrichs

Department of Pharmaceutical Technology and Biopharmacy, University of Groningen, Antonius Deusinglaan 1, 9713 AV, Groningen, the Netherlands.

Accepted for publication as “Bottom-up preparation techniques for nanocrystals of lipophilic drugs” in Pharm. Res. 2010.

91

concluding remarks and perspectives

7.1. Introduction

Modern drug discovery techniques result in the development of an increasing number of highly lipophilic drug candidates. Their lipophilicity results in a slow dissolution of these drugs in the aqueous gastro-intestinal fluids, resulting in a poor bioavailability. Over the past decades, many strategies have been developed to increase the dissolution rate of these types of drugs. The application of drug nanocrystals is one of those. The dissolution rate of drugs from nanocrystals is increased due to an increase of the saturation concentration around these particles (kelvin law) and an increase of the surface area available for dissolution.

Preparation techniques for drug nanocrystals can be divided into top-down and bottom-up techniques. Top-down techniques are based on size reduction of relatively large particles into smaller particles, whereas bottom-up techniques consist of the growth of small particles from individual molecules. The driving force for the growth of a crystal from individual molecules is supersaturation. Supersaturation of a drug in a solution can amongst others be obtained by decreasing the temperature or addition of an anti-solvent.

The size of crystals formed from supersaturated pure drug solutions depends on the balance between the nucleation rate and crystal growth. This balance between nucleation and growth is determined by the extent of supersaturation. At a higher extent of supersaturation, the nucleation rate increases and hence the crystal size decreases, whereas at a lower extent of supersaturation, the growth rate increases and hence the crystal size increases. Therefore, process conditions, such as temperature or the ratio between solvent and anti-solvent and their mixing rate can be used to control the drug crystal size. When drug crystals are formed in drug composites, the size of the crystals is also determined by the interstitial spaces between matrix molecules.

While there are many excellent reviews on top-down techniques (e.g. [4; 94]), bottom-up techniques are usually just referred to as ‘precipitation techniques’. Bottom-up techniques can be considered as the oldest techniques to prepare drug nanocrystals. However, recently interesting new developments in the field of bottom-up preparation of drug nanocrystals have evolved. This commentary focuses on these techniques. First an overview of bottom-up preparation techniques of crystalline drug nanoparticles will be given (Table 7-I). Secondly, the process related barriers for products prepared by these techniques to reach the market will be discussed. For an introduction on top-down methods to prepare drug nanocrystals and the application thereof, the reader is referred to an excellent recently published review [94].

7.2. Bottom-up techniques

Since bottom-up techniques to prepare drug nanocrystals are commonly referred to as ‘precipitation methods’, solid dispersions are usually not considered to be the first product containing drug nanocrystals. However, some types of solid dispersions consist of nanocrystalline drug particles embedded in a matrix. Already in the early 1960s Sekiguchi et al.

92

chapter 7

Tab

le 7

-I. O

verv

iew

of b

otto

m-u

p te

chni

ques

to p

repa

re d

rug

nano

crys

tals

Ref

eren

ce

[95]

[96]

[97]

[98]

[99]

[29]

Maj

or d

isad

vant

age

Poss

ible

deco

mpo

siti

on

of th

e dr

ug

Con

tam

inat

ion

from

toxi

c or

gani

c

solv

ents

Diffi

cult

y to

cont

rol t

he s

ize

Diffi

cult

y to

rem

ove

orga

nic

solv

ent

Solu

bilit

y of

the

drug

in th

e so

lven

t

Long

pro

cess

ing

tim

es

Com

posi

tion

Dru

g

com

posi

te

Dru

g

com

posi

te

Pure

dru

g

or d

rug

com

posi

te

Pure

dru

g

or d

rug

com

posi

te

Pure

dru

g

or d

rug

com

posi

te

Dru

g

com

posi

te

Des

crip

tion

Rap

idly

coo

ling

of a

eut

ecti

c

mel

t of d

rug

and

mat

rix

eva

pora

tion

of t

he s

olve

nt

in w

hich

the

drug

and

mat

rix

are

diss

olve

d

Prec

ipit

atio

n fr

om a

n or

gani

c

solv

ent b

y ad

diti

on o

f wat

er

Satu

rati

on o

f a d

rug

solu

tion

wit

h a

supe

rcri

tica

l flui

d

Rap

id d

ecre

ase

of th

e pr

essu

re

of a

sup

ercr

itic

al fl

uid

in w

hich

the

lipop

hilic

dru

g is

dis

solv

ed.

Free

ze-d

ryin

g at

a r

elat

ivel

y hi

gh

tem

pera

ture

of a

mix

ture

of d

rug,

mat

rix,

org

anic

sol

vent

, and

wat

er.

Year

1961

1965

1988

1991

1999

2008

Inve

ntor

s

k. S

ekig

uchi

and

N. O

bi

T. T

achi

bana

and

A.

Nak

amur

a

M. L

ist a

nd

H. S

ucke

r

v.j

.

kru

koni

s

et a

l.

G.W

. Pac

e

et a

l.

H. d

e W

aard

et e

l.

Abb

revi

atio

n

- - - GA

S

Re

SS

CC

DF

Prep

arat

ion

tech

niqu

e

Hot

mel

t met

hod

Solv

ent e

vapo

rati

on

met

hod

Hyd

roso

l

Gas

ant

i-so

lven

t

recr

ysta

lliza

tion

Rap

id e

xpan

sion

of

supe

rcri

tica

l sol

utio

ns

Con

trol

led

crys

talli

zati

on d

urin

g

free

ze-d

ryin

g

93


prepared a solid dispersion of the poorly soluble sulfathiazole embedded in urea [95]. The small particles were prepared by melting the drug and matrix and subsequently cooling this mixture in an ice bath. This resulted in a carrier matrix with drug particles entrapped in it. The size of the nanoparticles could be controlled by process parameters such as the cooling rate.

A little later, Tachibani et al. developed the solvent evaporation method, to prepare solid dispersions [96]. To obtain a solid dispersion by this method, the drug and matrix are dissolved in a common solvent (e.g. chloroform), after which the solvent is evaporated under vacuum. When Tachibani developed this method, the drug was incorporated in the matrix monomolecularly or as amorphous nanoparticles. Although this can be a successful strategy to increase the dissolution rate, there are concerns regarding the physical stability of these solid dispersions. Since amorphous solid dispersions have the risk of uncontrolled crystallization during processing or storage, a crystalline product is usually preferred. After the solvent evaporation method was adjusted, it could be used to prepare nanocrystalline solid dispersions by this method as well [7].

The first traditional precipitation technique that has been developed, is the so called ‘hydrosol’ technique, invented by List and Sucker [97]. To obtain drug nanocrystals by this technique, the lipophilic drug is first dissolved in an organic solvent (e.g. ethanol) and then mixed with a large amount of anti-solvent that is miscible with the organic solvent, usually water. Mixing of the organic solution and the anti-solvent should be performed rapidly, to assure fast nucleation and thereby small particles. To stabilize the nanocrystals, a carrier (e.g. gelatin) can be dissolved in the aqueous phase. Hereafter the solvents can then be removed by evaporation or lyophilization to obtain a dry nanocrystalline drug product.

Another category of bottom-up techniques that was developed, is commonly known as supercritical fluid technologies. Although there are currently many different processes to prepare drug nanocrystals based on supercritical fluid technologies, they are all more or less based on ‘gas anti-solvent recrystallization (GAS)’ and ‘rapid expansion of supercritical solutions (RESS)’. The application of GAS to obtain drug nanocrystals was for the first time described in 1991 by Krukonis et al. [98]. In this process, the supercritical fluid acts as an anti-solvent. A solution of the lipophilic drug in an organic solvent is saturated with supercritical fluid (such as supercritical CO2), thereby decreasing the solubility of the drug in the solvent and consequently causing the drug to precipitate. The other supercritical fluid technique, RESS, was already widely described in other applications before it was for the first time described in 1999 by Pace et al. to prepare drug nanocrystals [99]. In this process, the supercritical fluid acts as a solvent. The lipophilic drug is dissolved in the supercritical fluid after which the pressure is rapidly decreased. Due to the decrease of solvent power, the drug rapidly precipitates from the supercritical fluid.

More recently, a freeze-drying technique to prepare drug nanocrystals, controlled crystallization during freeze-drying, was developed [29]. This technology is based on freeze-drying a mixture of a non-toxic organic solvent (tertiary butyl alcohol) in which the drug is dissolved and an aqueous solution containing a matrix material (e.g. mannitol). Freeze-drying is performed at

chapter 1

94

a relatively high temperature (above the glass transition temperature of the maximally freeze-concentrated fraction) to allow the drug and matrix to crystallize. In the future, other bottom-up techniques to prepare drug nanocrystals based on existing processes may be developed. Bottom-up techniques, such as spray-freezing into liquid [34] or nanoprecipitation in microfluidic reactors [100], are promising techniques to prepare drug nanocrystals. Currently they are used to prepare amorphous nanoparticles, but they may be adapted in such a way that they can be used to produce a nanocrystalline drug product. SFL for example, consists of similar process steps as CCDF. A mixture of drug, polymer, organic solvent, and water is rapidly frozen and then freeze-dried. Although there are more differences between these two processes, the most important difference is that the solutions used during SFL are, at least at room temperature, thermodynamically stable, while the mixture used during CCDF is not. So far SFL is only reported to yield amorphous nanoparticles, but when process conditions such as the concentration organic solvent or the freeze-drying temperature are modified, crystalline nanoparticles may be obtained.

7.3. Bottom-up prepared nanocrystals on the market

Whereas there are already several nanocrystalline products on the market prepared by top-down methods [94], the number of products prepared by bottom-up methods is very limited. Rare examples are products such as Gris-PeG® (griseofulvin in PeG8000) and Cesamet®

(nabilone in PvP). The few bottom-up prepared nanocrystalline products are still based on the solid dispersion technology as described by Sekiguchi and Obi, which dates back almost fifty years ago. Although there are many promising in-vitro results published, only a few of these efforts resulted in a marketed product.

It is therefore worthwhile to analyze why there are so few nanocrystalline drug products based on bottom-up preparation techniques. Commonly mentioned disadvantages of these techniques are possible decomposition of the drug (hot melt method), contamination from toxic organic solvents (solvent evaporation method), the difficulty to control the size of the drug crystals (hydrosol), limited solubility of the drug in the solvent (supercritical fluid technologies), and long process times (CCDF). However, also top-down methods suffer from disadvantages such as contamination from grinding media, the difficulty to achieve a uniform size distribution, and the long processing times. These disadvantages did not limit the development of commercialized nanocrystalline drug products based on top-down preparation techniques.

Therefore it seems to be unlikely that the above mentioned disadvantages are the most important reason that only a limited number of bottom-up prepared nanocrystalline drug products reached the market. A more plausible explanation may be the difficulty to scale-up the bottom-up processes. For obvious reasons, most bottom-preparation methods are developed on lab-scale. However, to utilize these potentially useful bottom-up techniques, one should think about industrial production. This means that for each technique the possibility to change the process into a continuous and/or large scale process should be kept in mind. Also, the use of toxic organic solvents should be minimized to reduce the risks during manufacturing and the

95


risk of high concentrations of residual impurities in the end product.

Large-scale production of drug nanocrystals by bottom-up techniques is not necessarily problematic. Techniques such as the hot melt method, some supercritical fluid methods, and CCDF seem to be suitable for scale-up. For example, the lab-scale hot melt method can be changed into a large-scale holt melt extrusion process and freeze-drying during CCDF can be changed into continuous spray freeze-drying. However, more research should be performed on the actual feasibility of these techniques to be used for large-scale production, before they can be seen as an alternative to the currently used top-down methods.

In conclusion, although there are only a limited number of products based on bottom-up preparation of drug nanocrystals on the market, promising technologies are available. examples of these techniques are hot melt extrusion, supercritical fluid technologies, and CCDF. However, products prepared by these or other bottom-up techniques will only reach the market, if the industrial production is already kept in mind during lab-scale development.

chaptersummary8

98

chapter 8

Summary

The application of drug nanocrystals is an approach to increase the bioavailability of BCS class II drugs after oral administration. The dissolution rate of drugs, formulated as nanocrystals, is increased due to the increased saturation concentration around these small particles and due to their increased surface area. However, currently used methods to prepare drug nanocrystrals have disadvantages such as the use of large amounts of surfactants, the difficulty to control the particle size, low yields, possible contamination from grinding media, and difficulties related to scale-up of the process. Therefore, a novel bottom-up method to prepare drug nanocrystals, controlled crystallization during freeze-drying was developed. This thesis furthermore describes aspects related process scale-up, in-line Raman measurements as PAT tool, CLSM as tool to analyze the obtained product, and molecular modeling of the matrix material.

In chapter 2, the development of “controlled crystallization during freeze-drying” as process for the preparation of drug nanocrystals is described. This process consists of three steps: first, a solution of a matrix material in water was mixed with a solution of drug in TBA. This mixture was then frozen and subsequently freeze-dried at a relatively high temperature. It was envisioned that at this temperature, the freeze-concentrated fraction is in the rubbery state and that therefore the drug and matrix material may crystallize. To test this, solid dispersions containing concentrations of 10-40% w/w fenofibrate (a highly lipophilic model drug) in mannitol (the matrix material) were prepared. DSC and XRPD analysis showed that all dispersions were highly crystalline. It was also shown that the dissolution rate of fenofibrate from tablets prepared from the crystalline dispersions was higher than the dissolution rate of fenofibrate from tablets prepared from a physical mixture. SEM pictures showed that the obtained particles were of nanoscale when the mixture was frozen rapidly. To test the influence of two process conditions, the freezing rate and the water/TBA ratio were varied. It was found that a higher freezing rate or a lower water/TBA resulted in smaller drug crystals. These two process conditions can be used to control the drug crystal size.

In chapter 3, the novel process, as described in chapter 2, was modified into a process that is more suitable for large scale production. The initial process consisted of freeze-drying a mixture of drug, matrix material, TBA, and water. This mixture had to be frozen immediately after preparation, since the mixture is thermodynamically instable. Thereto small fluid quantities were mixed in a glass vial and the vial was immersed in liquid nitrogen. However, to change this process into a process that is more suitable for large scale production, the application of a 3-way nozzle was tested in this chapter. The nozzle is designed to have two solutions flow separately through the nozzle, after which they rapidly mix just outside the nozzle by the atomizing airflow from a third channel. First it was shown by the adapted Villermaux/Dushman method that two solutions indeed mix rapidly after they left the nozzle. Secondly, it was shown that both the freeze-drying process and the spray freeze-drying process, resulted in the formation of fast dissolving crystalline dispersions of fenofibrate in mannitol. Thirdly, it was shown that the spray freeze-drying process resulted in smaller drug crystals than the small-scale freeze-drying process. Again a higher freezing rate results in the formation of smaller drug crystals. When the semi-continuous spray freeze-drying process is used, the freezing rate was higher than when

99

summary

the freeze-drying process was used. Therefore, the 3-way nozzle can not only be used to scale-up the batch process into a semi-continuous process, but it also yields smaller drug crystals with a higher dissolution rate.

In chapter 4, in-line Raman spectroscopy was used as PAT tool to reveal at which stage of the process the drug, matrix material, and solvents crystallized. The mechanism that determines the drug crystal size during CCDF was elucidated. Although it was shown in chapter 2 and chapter 3 that process conditions, such as the freezing rate, influenced the drug nanocrystal size, it was not exactly clear which mechanism determined the drug crystal size. To elucidate this mechanism, the moment and length of crystallization was determined, whereas the factors that influenced the drug crystal size were determined by varying process conditions. It was shown that both solvents, TBA and water, crystallized upon cooling the sample to -50 ºC, while the solutes, fenofibrate and mannitol crystallized after the freezing step when the temperature of the freeze-dryer shelf was increased to -25 ºC or -15 ºC. Although the solutes crystallized after the freezing step, both the freezing rate and the shelf temperature were critical parameters that determined the drug crystal size. At a higher freezing rate, smaller solvent crystals and consequently smaller interstitial spaces containing the freeze-concentrated fraction are being formed. Since the drug and matrix crystallize in these interstitial spaces, the size of these interstitial spaces limits the size of the final crystals. On the other hand, when the temperature of the freeze-dryer shelf is at a lower temperature during the crystallization step, the degree of supersaturation is higher. A higher supersaturation results in a higher nucleation rate and consequently more and therefore smaller crystals. Therefore, both the freezing rate and the temperature of the freeze-dryer shelf can be used to control the drug crystal size.

In chapter 5, it was shown that CLSM combined with image analysis can be used to determine the size of drug crystals dispersed in a matrix. To validate this method, the size of pure dipyridamole crystals as determined by laser diffraction was compared to the size of the same dipyridamole crystals in a mixture with mannitol as determined by CLSM. Both methods showed a similar particle size distribution, demonstrating that CLSM can be used to determine the size of drug crystals, even though they were mixed with a second component. Secondly, two solid dispersions of dipyridamole crystals in mannitol were prepared by CCDF. The concentration drug and matrix in the water/TBA mixture was varied to obtain crystals of a different size, while the ratio drug/matrix was kept constant to obtain solid dispersions with the same drug load. CLSM showed that the size of the drug crystals prepared from the less concentrated solutions was of sub-micron size, whereas the size of the drug crystals prepared from the more concentrated solutions was larger. This trend in crystal size was in agreement with the trend in dissolution rate. Therefore, it was shown that CLSM combined with image analysis can be used to determine the drug crystal size, even though the drug is mixed with or dispersed in a second component. The fact that the smallest crystals in this study were of sub-micron size and the largest crystals several tens of micrometers, shows that CLSM can be used to measure the drug crystal size over a broad size range

In chapter 6, the evaluation and optimization of a molecular modeling force field for the matrix materials mannitol and sorbitol is described. Two force field parameters sets,

100

chapter 8

GROMOS53A5/53A6 and AMBER95 parameter sets were evaluated for their ability to reproduce the crystalline structure of the known polymorphs of mannitol and sorbitol. Molecular dynamics simulations at 10 k with the simulation cell lengths and angles free to evolve were performed to evaluate the force fields. Both force fields failed to reproduce the crystal structure of all polymorphic forms. Therefore both force fields were systematically optimized by sensitivity analysis. This approach did not result in an optimized GROMOS force field that was able to reproduce all polymorphic forms of both mannitol and sorbitol. However, the sensitivity analysis led to a revised AMBeR95 parameter set that was able to reproduce the crystal structures with less than 5% deviation from the experimental structures. Hereafter, the revised and the original AMBeR95 parameter set were tested in extended MD simulations at 298 k. Remarkably both the original and the revised parameter sets effectively reproduced the crystal structures under these conditions. Even more remarkable was the finding that the original force field parameter set reproduced the crystal structures slightly better. Therefore it was suggested that, if possible, force field parameters should not only be optimized by potential energy minimization (MD simulation at low temperature), but also at elevated temperatures. Since the optimized AMBeR95 parameter set has been found capable of reproducing all the crystal structures of mannitol, it can now be used to explore drug-mannitol interfacial properties. These properties can be used to gain knowledge about the physical stability of drug nanocrystals embedded in a mannitol matrix.

In chapter 7, an overview of the currently existing bottom-up preparation techniques of drug nanocrystals is given. Although some of these techniques are promising techniques to be used for the preparation of marketed drug products, there is only a limited number of products on the market produced with these techniques. It is discussed in this chapter that products based on bottom-up preparation techniques will only reach the market, if the industrial production is already kept in mind during lab-scale development.

101

summary

samenvattingAappendix

104

appendix a

Samenvatting

De toepassing van nanokristallen is één van de mogelijke methodes om de biologische beschikbaarheid van BCS klasse II (slecht wateroplosbaar, maar goede opname over de darmwand) geneesmiddelen na orale toediening te verhogen. De dissolutiesnelheid van geneesmiddelen, geformuleerd als nanokristallen, wordt verhoogd door de verhoogde verzadigingsconcentratie rond deze kleine deeltjes en het vergrote oppervlak. De huidige methodes om nanokristallen te maken hebben echter nadelen zoals het gebruik van grote hoeveelheden oppervlakte actieve stoffen, problemen bij het beheersen van de deeltjesgrootte, lage opbrengst, mogelijke contaminatie van maalkogels en problemen gerelateerd aan het opschalen van het proces. In dit proefschrift wordt de ontwikkeling van een nieuwe methode om nanokristallen te maken, “gecontroleerde kristallisatie tijdens vriesdrogen”, beschreven. verder worden in dit proefschrift aspecten gerelateerd aan het opschalen van het proces, de toepassing van Raman spectroscopie als PAT instrument, de toepassing van CLSM als instrument om de kristalgrootte van het geneesmiddel in het eindproduct te analyseren en het moleculair modeleren van het gebruikte matrix materiaal beschreven.

In hoofdstuk 2, wordt de ontwikkeling van “gecontroleerde kristallisatie tijdens vriesdrogen” als proces voor het maken van nanokristallen beschreven. Dit proces bestaat uit drie stappen: eerst wordt een oplossing van het geneesmiddel in TBA gemengd met een oplossing van het matrix materiaal (mannitol) in water. Dit mengsel wordt daarna bevroren en vervolgens gevriesdroogd bij een relatief hoge temperatuur. er werd verwacht dat bij de relatief hoge temperatuur de ‘freeze-concentrated fraction’ zich in de rubber-toestand bevond en dat daarom het geneesmiddel en matrix materiaal zouden kunnen kristalliseren. Om dit te testen werden vaste dispersies met 10-40% fenofibraat (een zeer slecht wateroplosbaar) in mannitol bereid. DSC en XRPD analyse toonden aan dat alle dispersies zeer kristallijn waren. SeM foto’s lieten zien dat de deeltjes die verkregen werden na snel invriezen kleiner dan een micrometer (nanoschaal) waren. Om de invloed van procesparameters te testen, werden de invriessnelheid en de verhouding water/TBA gevarieerd. Een hogere invriessnelheid of een lagere water/TBA verhouding resulteerde in kleinere geneesmiddelkristallen, die sneller oplosten. Deze proces parameters kunnen dus gebruikt worden om de kristalgrootte te beïnvloeden.

In hoofdstuk 3, werd het nieuwe proces, zoals beschreven in hoofdstuk 2, aangepast zodat het beter geschikt was voor productie op grote schaal. Het oorspronkelijke proces bestond uit het vriesdrogen van een mengsel van een oplossing van geneesmiddel in TBA en oplossing van een matrix materiaal inwater. Dit mengsel moest onmiddellijk na mengen bevroren worden, omdat het mengsel thermodynamisch instabiel is. Daarom werden kleine hoeveelheden vloeistof in glazen flesjes gemengd en werden deze flesjes ingevroren in vloeibare stikstof. Om dit proces beter geschikt te maken voor productie op grote schaal, werd het gebruik van een 3-weg sproeikop om het vriesdroog-proces te veranderen in een sproeivriesdroog-proces getest. Deze sproeikop is zodanig ontworpen dat twee verschillende opolossingen gescheiden door de sproeikop kunnen worden gepompt en gemengd kunnen worden vlak nadat ze de sproeikop verlaten hebben gemengd kunnen worden door middel van een persluchtstroom uit een derde kanaal. Met de aangepaste Villermaux/Dushman methode werd aangetoond

105

samenvatting

dat de twee oplossingen daadwerkelijk snel mengden nadat ze de sproeikop verlieten. verder werd aangetoond dat zowel het vriesdroog als het sproeivriesdroog-proces gebruikt kunnen worden om snel oplossende kristallijne dispersies van fenofibraat in mannitol te maken. Ook werd aangetoond dat het sproeivriesdroog-proces kleinere geneesmiddelkristallen opleverde dan het vriesdroog-proces. Wanneer het sproeivriesdroog-proces werd toegepast, was de invriessnelheid hoger dan wanneer het vriesdroog-proces werd gebruikt. Wederom leverde een hogere invriessnelheid kleinere geneesmiddelkristallen op. De 3-weg sproeikop kan daarom niet alleen gebruikt worden om het batch proces (vriesdrogen) op te schalen naar een semicontinu proces (sproeivriesdrogen), maar het levert ook kleinere geneesmiddelkristallen met een hogere dissolutiesnelheid op.

In hoofdstuk 4 werd in-line Raman spectroscopie als PAT instrument gebruikt om aan te tonen in welke fase van het vriesdroog-proces het geneesmiddel, het matrixmateriaal en de oplosmiddelen kristalliseerden. Op basis hiervan werd het mechanisme beschreven dat de grootte van de geneesmiddelkristallen tijdens het vriesdrogen bepaalt. Ondanks dat in hoofdstuk 2 en hoofdstuk 3 duidelijk was geworden dat proces parameters zoals de invriessnelheid van invloed waren op de grootte van de kristallen, was het nog niet precies duidelijk waarom deze parameters de grootte beïnvloedden. Om dit mechanisme op te helderen werd het moment waarop de verschillende componenten kristalliseerden en hoe lang het proces van kristallisatie duurde bepaald. Om de invloed van verschillende procesparameters te bepalen werden de invriessnelheid en de temperatuur tijdens vriesdrogen gevarieerd. er werd aangetoond dat beide oplosmiddelen, TBA en water, kristalliseerden tijdens het koelen naar -50 ºC, terwijl beide opgeloste stoffen, fenofibraat en mannitol, kristalliseerden in de fase waarin de temperatuur in de vriesdroger was verhoogd naar -25 ºC of -15 ºC. De kristalgrootte werd beïnvloed door zowel de invriessnelheid als de temperatuur waarbij werd gevriesdroogd, ondanks dat de opgeloste stoffen pas kristalliseerden na het invriezen. Hieruit kon worden afgeleid dat bij een hogere invriessnelheid kleinere kristallen van de oplosmiddelen gevormd werden, met als gevolg kleinere tussenruimtes met daarin de ‘freeze-concentrated fraction’. Aangezien het geneesmiddel en het matrix materiaal alleen in deze tussenruimtes kunnen kristalliseren, is de maximale grootte van de uiteindelijke kristallen beperkt tot de grootte van deze tussenruimtes. Aan de andere kant is de mate van oververzadiging hoger bij een lagere temperatuur in de vriesdroger. een hogere oververzadiging resulteert in een hogere nucleatie snelheid en daarom meer en dus kleinere kristallen. kortom, zowel de invriessnelheid als de temperatuur tijdens vriesdrogen kunnen gebruikt worden om de grootte van de kristallen te reguleren.

In hoofdstuk 5 is aangetoond dat CLSM gecombineerd met beeldverwerking gebruikt kan worden om de grootte van geneesmiddelkristallen, gedispergeerd in een matrix, te bepalen. Om deze methode te valideren werd de grootte van pure dipyridamol kristallen bepaald met behulp van laser diffractie en vergeleken met de grootte van dipyridamol kristallen in een mengsel met mannitol, bepaald met behulp van CLSM. Beide methodes gaven een vergelijkbare deeltjesgrootte verdeling, wat aantoont dat CLSM gebruikt kan worden om de deeltjesgrootte te bepalen, ook al is het geneesmiddel gemengd met een tweede stof. vervolgens werden twee verschillende dispersies van dipyridamol in mannitol gemaakt door middel van “gecontroleerde kristallisatie tijdens vriesdrogen”. De concentratie geneesmiddel en matrix in het water/TBA

106

appendix a

mengsel werd gevarieerd om kristallen van verschillende grootte te verkrijgen. Met CLSM werd aangetoond dat de kristallen bereid uit de laag geconcentreerde oplossingen van sub-micron grootte waren terwijl de kristallen bereid van de hoog geconcentreerde oplossingen groter waren. Deze trend in kristalgrootte kwam overeen met de gevonden trend in dissolutiesnelheid. Dit toont aan dat CLSM ook gebruikt kan worden om de grootte van kristallen, gedispergeerd in een tweede stof, te bepalen. Het feit dat de kleinste kristallen in deze studie van sub-micron grootte waren en de grootste kristallen enkele tientallen micrometers groot waren, toont aan dat CLSM gebruikt kan worden om de kristalgrootte over een breed bereik te bepalen.

In hoofdstuk 6 wordt de evaluatie en optimalisatie van een ‘force field’ voor het moleculair modeleren van de matrix materialen mannitol en sorbitol beschreven. Van twee force field parameter sets, GROMOS53A5/53A6 en AMBER95, werd bepaald in hoeverre ze geschikt zijn om de kristalstructuur van de bekende polymorfen van mannitol en sorbitol weer te geven. Om de force fields te testen werden moleculaire simulaties bij een temperatuur van 10 K waarbij de lengtes en hoeken van de cel zich vrij konden ontwikkelen uitgevoerd. voor beide force fields gold dat ze niet geschikt waren om alle polymorfen van beide stoffen goed weer te geven. Daarom werden beide force fields systematisch geoptimaliseerd door de invloed van de individuele parameters op de afmetingen van de kristallen te bestuderen. Deze benadering leidde niet tot een verbeterde versie van het GROMOS force field dat gebruikt kon worden om de polymorfen van beide stoffen weer te geven. Dit leidde echter wel tot een verbeterde versie van het AMBER force field dat alle polymorfen kon weergeven met een afwijking in de kristalstructuur van minder dan 5%. Hierna werden het originele en geoptimaliseerde AMBER force field getest in langere simulaties bij 298 K. Verassend genoeg waren onder deze omstandigheden beide force fields geschikt om alle kristalstructuren weer te geven. Nog verrassender was de uitkomst dat het originele force field de kristalstructuren iets beter weer gaf. Op basis hiervan werd dan ook gesuggereerd dat, indien mogelijk, force field parameters niet alleen moeten worden geoptimaliseerd door simulaties bij lage temperatuur, maar juist ook bij hogere temperaturen. Aangezien de geoptimaliseerde AMBeR parameter set geschikt was om alle kristalstructuren van mannitol weer te geven, kan deze gebruikt worden om de geneesmiddel-mannitol interacties te bestuderen. Deze kennis kan bijdragen om de stabiliteit van geneesmiddelnanokristallen in een mannitol matrix beter te begrijpen.

In hoofdstuk 7 is een overzicht van de huidige methodes om geneesmiddelnanokristallen te maken weergegeven. Tot nu toe zijn er slechts een paar geneesmiddelproducten op de markt gebracht die bereid zijn met één van deze technieken, ondanks het feit dat sommige van deze technieken veelbelovend zijn om te gebruiken voor commerciële productie. In dit hoofdstuk wordt geponeerd dat er alleen producten op de markt zullen komen, bereid met behulp van deze technieken, wanneer gedurende de ontwikkeling van de technieken al rekening wordt gehouden met productie op grote schaal.

107

samenvatting

referencesBappendix

110

appendix b

References

1. Lipinski, C.A., et al. (2001). experimental and computational approaches to estimate solubility and permeability in drug discovery and development settings. Adv Drug Deliv Rev 46, 3-26.

2. Lobenberg, R., and Amidon, G.L. (2000). Modern bioavailability, bioequivalence and biopharmaceutics classification system. New scientific approaches to international regulatory standards. eur j Pharm Biopharm 50, 3-12.

3. Curatolo, W. (1998). Physical chemical properties of oral drug candidates in the discovery and exploratory development settings. Pharm Sci Tech Today 1, 387-393.

4. Rabinow, B.e. (2004). Nanosuspensions in drug delivery. Nature Rev 3, 785-796.

5. Hou, D., et al. (2003). The production and characteristics of solid lipid nanoparticles (SLNs). Biomaterials 24, 1781-1785.

6. Noyes, A.A., and Whitney, W.R. (1897). The rate of solution of solid substances in their own solutions. j Am Chem Soc 19, 930-934.

7. Leuner, C., and Dressman, j. (2000). Improving drug solubility for oral delivery using solid dispersions. eur j Pharm Biopharm 50, 47-60.

8. Serajuddin, A.T. (1999). Solid dispersion of poorly water-soluble drugs: early promises, subsequent problems, and recent breakthroughs. j Pharm Sci 88, 1058-1066.

9. van Drooge, D.j., Hinrichs, W.L.j., and Frijlink, H.W. (2004). Incorporation of lipophilic drugs in sugar glasses by lyophilization using a mixture of water and tertiary butyl alcohol as solvent. j Pharm Sci 93, 713-725.

10. Amidon, G.L., et al. (1995). A theoretical basis for a biopharmaceutic drug classification: the correlation of in vitro drug product dissolution and in vivo bioavailability. Pharm Res 12, 413-420.

11. Law, D., et al. (2003). Properties of rapidly dissolving eutectic mixtures of poly(ethylene glycol) and fenofibrate: the eutectic microstructure. J Pharm Sci 92, 505-515.

12. Chiou, W.L., and Riegelman, S. (1971). Pharmaceutical applications of solid dispersion systems. j Pharm Sci 60, 1281-1302.

13. Chauhan, B., Shimpi, S., and Paradkar, A. (2005). Preparation and evaluation of glibenclamide-polyglycolized glycerides solid dispersions with silicon dioxide by spray drying technique. eur j Pharm Sci 26, 219-230.

14. Hancock, B.C., and Zografi, G. (1997). Characteristics and significance of the amorphous state in pharmaceutical systems. j Pharm Sci 86, 1-12.

15. Pokharkar, v.B., et al. (2006). Development, characterization and stabilization of amorphous form of a low Tg drug. Powder Technol 167, 20-25.

16. van Drooge, D.j., et al. (2006). Characterization of the molecular distribution of

111

references

drugs in glassy solid dispersions at the nano-meter scale, using differential scanning calorimetry and gravimetric water vapour sorption techniques. Int j Pharm 310, 220-229.

17. de Waard, H., et al. (2008). Unexpected differences in dissolution behavior of tablets prepared from solid dispersions with a surfactant physically mixed or incorporated. Int j Pharm 349, 66-73.

18. Müller, R.H., and Peters, K. (1998). Nanosuspensions for the formulation of poorly soluble drugs: I. Preparation by a size-reduction technique. Int j Pharm 160, 229-237.

19. Merisko-Liversidge, e., Liversidge, G.G., and Cooper, e.R. (2003). Nanosizing: a formulation approach for poorly-water-soluble compounds. eur j Pharm Sci 18, 113-120.

20. keck, C.M., and Muller, R.H. (2006). Drug nanocrystals of poorly soluble drugs produced by high pressure homogenisation. eur j Pharm Biopharm 62, 3-16.

21. Sarkari, M., et al. (2002). enhanced drug dissolution using evaporative precipitation into aqueous solution. Int j Pharm 243, 17-31.

22. Shekunov, B.Y., et al. (2006). Nanoparticles of poorly water-soluble drugs prepared by supercritical fluid extraction of emulsions. Pharm Res 23, 196-204.

23. kipp, j.e. (2004). The role of solid nanoparticle technology in the parenteral delivery of poorly water-soluble drugs. Int j Pharm 284, 109-122.

24. Burger, A., et al. (2000). Energy/temperature diagram and compression behavior of the polymorphs of D-mannitol. j Pharm Sci 89, 457-468.

25. Menczel, J., and Jaffe, M. (2007). How did we find the rigid amorphous phase? J Therm Anal Cal 89, 357-362.

26. Song, M. (2001). Rigid amorphous phase and low temperature melting endotherm of poly(ethylene terephthalate) studied by modulated differential scanning calorimetry. j Appl Polym Sci 81, 2779-2785.

27. Craig, D.Q.M., et al. (2001). The measurement of small quantities of amorphous material—should we be considering the rigid amorphous fraction? Pharm Res 18, 1081-1082.

28. jinno, j., et al. (2006). effect of particle size reduction on dissolution and oral absorption of a poorly water-soluble drug, cilostazol, in beagle dogs. j Control Release 111, 56-64.

29. de Waard, H., Hinrichs, W.L.j., and Frijlink, H.W. (2008). A novel bottom-up process to produce drug nanocrystals: controlled crystallization during freeze drying. j Control Release 128, 179-183.

30. Panic, S., et al. (2004). experimental approaches to a better understanding of mixing performance of microfluidic devices. Chem Eng J 101, 409-419.

112

appendix b

31. Fournier, M.C., Falk, L., and villermaux, j. (1996). A new parallel competing reaction system for assessing micromixing efficiency--Experimental approach. Chem Eng Sci 51, 5053-5064.

32. Overhoff, k.A., et al. (2007). Novel ultra-rapid freezing particle engineering process for enhancement of dissolution rates of poorly water-soluble drugs. eur j Pharm Biopharm 65, 57-67.

33. engstrom, j.D., et al. (2007). Morphology of protein particles produced by spray freezing of concentrated solutions. eur j Pharm Biopharm 65, 149-162.

34. Hu, j., et al. (2002). Improvement of dissolution rates of poorly water soluble APIs using novel spray freezing into liquid technology. Pharm Res 19, 1278-1284.

35. Hu, j., johnston, k.P., and Williams, R.O., 3rd (2004). Nanoparticle engineering processes for enhancing the dissolution rates of poorly water soluble drugs. Drug Dev Ind Pharm 30, 233-245.

36. Möschwitzer, J., and Müller, R.H. (2006). New method for the effective production of ultrafine drug nanocrystals. Journal of nanoscience and nanotechnology 6, 3145-3153.

37. Blagden, N., et al. (2007). Crystal engineering of active pharmaceutical ingredients to improve solubility and dissolution rates. Adv Drug Deliv Rev 57, 617-630.

38. Bisrat, M., and Nyström, C. (1988). Physicochemical aspects of drug release. vIII. The relation between particle size and surface specific dissolution rate in agitated suspensions. Int j Pharm 47, 223-231.

39. De Beer, T.R.M., et al. (2007). Influence of particle size on the quantitative determination of salicylic acid in a pharmaceutical ointment using FT-Raman spectroscopy. eur j Pharm Sci 30, 229-235.

40. De Beer, T.R.M., et al. (2007). Implementation of a process analytical technology system in a freeze-drying process using raman spectroscopy for in-line process monitoring. Anal Chem 79, 7992-8003.

41. D’Arrigo, G., et al. (1981). Raman scattering and structure of normal and supercooled water. The journal of chemical physics 75, 4264-4270.

42. Taylor, L.S., et al. (1997). Characterization of frozen glucose solutions. Pharm Dev Tech 2, 395 - 402.

43. Romero-Torres, S., et al. (2007). Monitoring of mannitol phase behavior during freeze-drying using non-invasive raman spectroscopy. PDA journal of pharmaceutical science and technology / PDA 61, 131-145.

44. McCreery, R.L. (2000). Raman spectroscopy for chemical analysis, 1st edn (New York, Wiley-Interscience).

45. de Waard, H., et al. (2009). Preparation of drug nanocrystals by controlled crystallization: application of a 3-way nozzle to prevent premature crystallization for

113

references

large scale production. eur j Pharm Sci 38, 224-229.

46. Pikal, M.j. (2007). Freeze drying. In encyclopedia of pharmaceutical technology, j. Swarbrick, ed. (New York, Informa healthcare USA, Inc.).

47. Rodríguez-Hornedo, N., et al. (2007). Crystallization: general principles and significance on product development. In Encyclopedia of pharmaceutical technology, j. Swarbrick, ed. (New York, Informa healthcare USA, Inc.).

48. Qian, F., et al. (2007). Mechanistic investigation of pluronic(R) based nano-crystalline drug-polymer solid dispersions. Pharm Res 24, 1551-1560.

49. jinno, j., et al. (2008). In vitro-in vivo correlation for wet-milled tablet of poorly water-soluble cilostazol. j Control Release 130, 29-37.

50. Möschwitzer, j., et al. (2004). Development of an intravenously injectable chemically stable aqueous omeprazole formulation using nanosuspension technology. eur j Pharm Biopharm 58, 615-619.

51. de Waard, H., et al. (2010). Controlled crystallization of the lipophilic drug fenofibrate during freeze-drying: elucidation of the mechanism by in-line Raman spectroscopy. AAPS j 12, 569-575.

52. Driscoll, D.F., et al. (2001). Physicochemical assessments of parenteral lipid emulsions: light obscuration versus laser diffraction. Int j Pharm 219, 21-37.

53. Keck, C.M., and Müller, R.H. (2008). Size analysis of submicron particles by laser diffractometry-90% of the published measurements are false. Int J Pharm 355, 150-163.

54. verma, S., Gokhale, R., and Burgess, D.j. (2009). A comparative study of top-down and bottom-up approaches for the preparation of micro/nanosuspensions. Int J Pharm 380, 216-222.

55. Wu, W., and Nancollas, G.H. (1999). Determination of interfacial tension from crystallization and dissolution data: a comparison with other methods. Adv Colloid Interface Sci 79, 229-279.

56. Bugay, D.e. (2001). Characterization of the solid-state: spectroscopic techniques. Adv Drug Deliv Rev 48, 43-65.

57. Doub, W.H., et al. (2007). Raman chemical imaging for ingredient-specific particle size characterization of aqueous suspension nasal spray formulations: a progress report. Pharm Res 24, 934-945.

58. Pygall, S.R., et al. (2007). Pharmaceutical applications of confocal laser scanning microscopy: The physical characterisation of pharmaceutical systems. Adv Drug Del Rev 59, 1434-1452.

59. kostewicz, e.S., et al. (2002). Forecasting the oral absorption behavior of poorly soluble weak bases using solubility and dissolution studies in biorelevant media.

114

appendix b

Pharm Res 19, 345-349.

60. Patterson, J.E., et al. (2005). The influence of thermal and mechanical preparative techniques on the amorphous state of four poorly soluble compounds. j Pharm Sci 94, 1998-2012.

61. Iuliano, L., et al. (2000). Fluorescence quenching of dipyridamole associated to peroxyl radical scavenging: a versatile probe to measure the chain breaking antioxidant activity of biomolecules. Biochem Biophys Acta 1474, 177-182.

62. Bornet, F.R.j. (1994). Undigestible sugars in food products. Am j Clin Nutr 59, 763S-769S.

63. Rowe, R.C., Sheskey, P.j., and Owen, S.C. (2005). Handbook of pharmaceutical excipients, 5th edn (Pharmaceutical Press).

64. Yoshinari, T., et al. (2003). The improved compaction properties of mannitol after a moisture-induced polymorphic transition. Int j Pharm 258, 121-131.

65. Bolhuis, G.K., Rexwinkel, E.G., and Zuurman, K. (2009). Polyols as filler-binders for disintegrating tablets prepared by direct compaction. Drug Dev Ind Pharm 35, 671-677.

66. johnson, R.e., kirchhoff, C.F., and Gaud, H.T. (2002). Mannitol-sucrose mixtures--versatile formulations for protein lyophilization. j Pharm Sci 91, 914-922.

67. Hawe, A., and Frieß, W. (2006). Physico-chemical lyophilization behavior of mannitol, human serum albumin formulations. eur j Pharm Sci 28, 224-232.

68. Liao, X., Krishnamurthy, R., and Suryanarayanan, R. (2007). Influence of processing conditions on the physical state of mannitol-implications in freeze-drying. Pharm Res 24, 370-376.

69. Pyne, A., Chatterjee, k., and Suryanarayanan, R. (2003). Solute crystallization in mannitol–glycine systems - Implications on protein stabilization in freeze-dried formulations. j Pharm Sci 92, 2272-2283.

70. Park, Y.j., and jeffrey, G.A. (1971). Determination of the crystal structure of the A form of D-Glucitol by neutron and X-ray diffraction. Acta Cryst B27, 2393-2401.

71. Schouten, A., et al. (1998). Conformational polymorphism of d-sorbitol (d- glucitol): the crystal and molecular structures of d-glucitol 2/3-hydrate and epsilon d-glucitol. Carbohydr Res 312, 131-137.

72. Rukiah, M., et al. (2004). Ab initio structure determination of the C form of D-sorbitol (D-glucitol) by powder synchrotron X-ray diffraction. j Appl Crystallogr 37, 766-772.

73. kim, A.I., Akers, M.j., and Nail, S.L. (1998). The physical state of mannitol after freeze-drying: effects of mannitol concentration, freezing rate, and a noncrystallizing cosolute. j Pharm Sci 87, 931-935.

74. Berman, H.M., jeffrey, G.A., and Rosenstein, R.D. (1968). The crystal structures of the

115

references

α’ and β forms of D-mannitol. Acta Cryst B24, 442-449.

75. Catlow, C.R.A., et al. (1997). Introductory lecture computer modelling as a technique in solid state chemistry Faraday Discuss 106, 1-40.

76. Frenkel, D., and Smit, B. (2002). Understanding molecular simulations: From algorithms to applications, vol 1, 2nd edn (San Diego, Academic Press).

77. Chipot, C., and Pohorille, A. (2007). Free energy calculations: Theory and applications in chemistry and biology, vol 86 (Heidelberg, Springer).

78. Day, G.M., et al. (2009). Significant progress in predicting the crystal structures of small organic molecules - a report on the fourth blind test. Acta Cryst B65, 107-125.

79. Oostenbrink, C., et al. (2004). A biomolecular force field based on the free enthalpy of hydration and solvation: The GROMOS force-field parameter sets 53A5 and 53A6. J Comput Chem 25, 1656-1676.

80. Cornell, W.D., et al. (1995). A second generation force field for the simulation of proteins, nucleic acids, and organic molecules. j Am Chem Soc 117, 5179-5197.

81. Fronczek, F.R., kamelb, H.N., and Slattery, M. (2003). Three polymorphs (alpha, beta and delta) of D-mannitol at 100 k. Acta Crystallogr C 59, O567-O570.

82. Guest, M.F., et al. (2005). The GAMeSS-Uk electronic structure package: algorithms, developments and applications. Mol Phys 103, 719 - 747.

83. kendrick, j., and Fox, M. (1991). Calculation and display of electrostatic potentials j Mol Graph 9, 182-187.

84. Smith, W., and Forester, T.R. (1996). DL_POLY_2.0: A general-purpose parallel molecular dynamics simulation package. j Mol Graph 14, 136-141.

85. Berendsen, H.j.C., et al. (1984). Molecular-dynamics with coupling to an external bath. The journal of chemical physics 81, 3684-3690.

86. Chatchawalsaisin, J., et al. (2008). An optimized force field for crystalline phases of resorcinol. CrystengComm 10, 437-445.

87. Anwar, j., Tuble, S.C., and kendrick, j. (2007). Concerted molecular displacements in a thermally-induced solid-state transformation in crystals of DL-norleucine. j Am Chem Soc 129, 2542-2547.

88. Beckham, G.T., et al. (2007). Surface-mediated nucleation in the solid-state polymorph transformation of terephthalic acid. j Am Chem Soc 129, 4714-4723.

89. Martoňák, R., et al. (2005). Simulation of structural phase transitions by metadynamics. Z Kristallogr 220, 489-498.

90. jorgensen, W.L., and Tirado-Rives, j. (2002). The OPLS [optimized potentials for liquid simulations] potential functions for proteins, energy minimizations for crystals of cyclic peptides and crambin. j Am Chem Soc 110, 1657-1666.

116

appendix b

91. Stone, A.j. (2005). Distributed multipole analysis: stability for large basis sets. j Chem Theory Comput 1, 1128-1132.

92. Day, G.M., and Price, S.L. (2003). A nonempirical anisotropic atom-atom model potential for chlorobenzene crystals. j Am Chem Soc 125, 16434-16443.

93. Price, S.L., et al. (2006). Applications Of Dl_poly And Dl_multi To Organic Molecular Crystals. Mol Simulat 32, 985-997.

94. Shegokar, R., and Müller, R.H. (2010). Nanocrystals: Industrially feasible multifunctional formulation technology for poorly soluble actives. Int j Pharm 399, 129-139.

95. Sekiguchi, k., and Obi, N. (1961). Studies on absorption of eutectic mixtures. I. A comparison of the behavior of eutectic mixtures of sulphathiazole and that of ordinary sulphathiazole in man. Chemical & pharmaceutical bulletin 9, 866-872.

96. Tachibana, T., and Nakamura, A. (1965). A methode for preparing an aqueous colloidal dispersion of organic materials by using water-soluble polymers: Dispersion of β-carotene by polyvinylpyrrolidone. Kolloid-Z Polym 203, 130-133.

97. List, M., and Sucker, H. (1988). Pharmaceutical colloidal hydrosols for injection.

98. krukonis, v.j., Gallagher, P.M., and Coffey, M.P. (1991). Gas anti-solvent crystallization process.

99. Pace, G.W., et al. (1999). Processes to generate submicron particles of water-insoluble compounds.

100. Ali, H.S.M., York, P., and Blagden, N. (2009). Preparation of hydrocortisone nanosuspension through a bottom-up nanoprecipitation technique using microfluidic reactors. Int j Pharm 375, 107-113.

nomenclatureCappendix

118

appendix c

Nomenclature

BCS Biopharmaceutics Classification SystemBeT Brunauer, emmet, and Teller-isothermCCDF Controlled Crystallization During Freeze-dryingCLSM Confocal Laser Scanning MicroscopyDLS Dynamic Light ScatteringDSC Differential Scanning CalorimetryeDX energy-Dispersive X-ray spectroscopyGAS Gas Anti-Solvent recrystallizationMD Molecular DynamicsPAT Process Analytical TechnologyPe Potential energyReSS Rapid expansion of Supercritical SolutionsSeM Scanning electron MicroscopySFL Spray-Freezing into LiquidTBA Tertiary Butyl AlcoholTc Crystallization TemperatureTe eutectic temperatureTg Glass transition TemperatureTg’ Glass transition Temperature of the maximally freeze-concentrated fractionXRPD X-Ray Powder Diffraction

curriculum vitae & list of publications

appendixD

120

appendix d

Curriculum Vitae

Hans de Waard was born (14 August 1982) in Oss, the Netherlands. After finishing highschool (Utrechts Stedelijk Gymnasium) in Utrecht in 2000, he started with his study pharmaceutical sciences at the University of Groningen. From 2003 he combined this with the master chemical engineering. In 2006 he obtained his master of science degree in pharmaceutical sciences (pharmaceutical technology) as well as in chemical engineering (pharmaceutical product engineering). After finishing his masters, Hans started his PhD-study at the department of Pharmaceutical Technology and Biopharmacy (University of Groningen) under supervision of Prof. dr. H.W. Frijlink. During his PhD-study, he worked on the development of a novel bottom-up process to prepare drug nanocrystals of lipophilic drugs. As part of his PhD-study, he stayed at the Institute of Pharmaceutical Innovation (University of Bradford) for 6 months as a research visitor. During his stay at the IPI, he worked under the supervision of Prof. dr. j. Anwar on molecular modeling of nanocrystals. Currently he is working as post-doctoral reseracher at the department of Pharmaceutical Technology and Biopharmacy of the University of Groningen.

121

curriculum vitae & list of publications

List of publications and awards

Books and book chaptersj.A. Wesselingh and H. de Waard (2011) Communicate your calculations using Mathcad. (Delft, vSSD), in press.

G.k. Bolhuis and H. de Waard (2011). Compaction properties of directly compressible materials. In Pharmaceutical Powder Compaction Technology 2nd edition, M. Celik, eds. (London, Informa Healthcare), in press.

H. de Waard (2010). Nanotechnologie en farmacologie. In Nanotechnologie, P. Oomen, T. Wobbes, and T. Bemelmans, eds. (Nijmegen, valkhof Pers), pp. 106-126.

Journal articlesH. de Waard, M.j.T. Hessels, M. Boon, k.A. Sjollema, W.L.j. Hinrichs, A.C. eissens, and H.W. Frijlink (2011). CLSM as quantitative method to determine the size of drug crystals in a solid dispersion. Submitted for publication.

A. Amani, P. York, H. de Waard, and j. Anwar (2011). Molecular dynamics simulation of a polysorbate 80 micelle in water. Submitted for publication.

H. de Waard, H.W. Frijlink, and W.L.j. Hinrichs (2010). Bottom-up preparation techniques for nanocrystals of lipophilic drugs. Pharm. Res., accepted for publication; invited.

H. de Waard, T. De Beer, W.L.j. Hinrichs, C. vervaet, j.P. Remon, and H.W. Frijlink (2010). Controlled crystallization of the lipophilic drug fenofibrate during freeze-drying: Elucidation of the mechanism by in-line Raman spectroscopy. AAPS j 12, 569-575.

H. de Waard, A. Amani, j. kendrick, W.L.j. Hinrichs, H.W. Frijlink, and j. Anwar (2010). Evaluation and optimization of a force field for crystalline forms of mannitol and sorbitol. J Phys Chem B 114, 429-436.

H. de Waard, N. Grasmeijer, W.L.j. Hinrichs, A.C. eissens, P.P.F. Pfaffenbach, and H.W. Frijlink (2009). Preparation of drug nanocrystals by controlled crystallization: application of a 3-way nozzle to prevent premature crystallization for large scale production. eur j Pharm Sci 38, 224-229.

H. de Waard, W.L.j. Hinrichs, and H.W. Frijlink (2008). A novel bottom-up process to produce drug nanocrystals: controlled crystallization during freeze drying. j Control Release 128, 179-183.

H. de Waard, W.L.j. Hinrichs, M.R. visser, C. Bologna, and H.W. Frijlink (2008). Unexpected differences in dissolution behavior of tablets prepared from solid dispersions with a surfactant physically mixed or incorporated. Int j Pharm 349, 66-73.

122

appendix d

Presentations (Oral)H. de Waard (2010). Freeze-drying of co-solvent systems: rationale, opportunities, and applications. SP Scientific Webinar.

H. de Waard, N. Grasmeijer, W.L.j. Hinrichs, and H.W. Frijlink (2010). Preparation of drug nanocrystals by controlled crystallization: from a batch to a semi-continuous process, 7th World Meeting on Pharmaceutics, Biopharmaceutics and Pharmaceutial Technology, valetta, Malta.

H. de Waard, M.j.T. Hessels, W.L.j. Hinrichs, and H.W. Frijlink (2009). Preparation of nanocrystals from drugs with different Tg by “controlled crystallization during freeze-drying”, TIPharma Spring Meeting, Utrecht, the Netherlands.

H. de Waard, N. Grasmeijer, W.L.j. Hinrichs, and H.W. Frijlink (2008). Preparation of drug nanocrystals by controlled crystallization: from batch to semi-continuous, FIGON Dutch Medicine Days, Lunteren, the Netherlands.

H. de Waard, W.L.j. Hinrichs, and H.W. Frijlink (2008). Preparation of stable surfactant-free nanocrystalline drug formulations, 6th World Meeting on Pharmaceutics, Biopharmaceutics and Pharmaceutial Technology, Barcelona, Spain.

H. de Waard, W.L.J. Hinrichs, M.R. Visser, C. Bologna, and H.W. Frijlink (2006). Influence of a surfactant on the release of class II drugs from sugar glass-based solid dispersions, Biopharmacy day of the Belgian-Dutch Biopharmacy Society, Beerse, Belgium.

Presentations (Poster)H. de Waard, T. De Beer, W.L.j. Hinrichs, and H.W. Frijlink (2010). In-line Raman spectroscopy to identify the critical parameters that determine the size of drug nanocrystals formed during freeze-drying. Pharmaceutical Sciences World Congress/AAPS Annual meeting, New Orleans, United States of America.

H. de Waard, M.j.T. Hessels, W.L.j. Hinrichs, and H.W. Frijlink (2009). Preparation and size determination of dipyridamole nanocrystals prepared by “controlled crystallization during freeze-drying”, AAPS Annual meeting, Los Angeles, United States of America.

H. de Waard, W.L.j. Hinrich,s and H.W. Frijlink (2008). Preparation of stable surfactant-free nanocrystalline drug formulations, Biopharmacy Day, Leuven, Belgium.

H. de Waard, j.P. Möschwitzer, W.L.j. Hinrichs, and H.W. Frijlink (2007). Optimization of a capsule formulation containing nanocrystals of a poorly soluble drug, Pre-Satellite Meeting of the 3rd Pharmaceutical Sciences World Congress, Amsterdam, the Netherlands.

AwardsTravel award from the european Federation for Pharmaceutical Sciences (eUFePS), to attend the 2010 FIP PSWC/AAPS Annual Meeting and Exposition in New Orleans, Louisiana.

dankwoordEappendix

124

appendix e

Dankwoord

Op de voorkant van een proefschrift staat slechts één naam geschreven. Het is echter een illusie om te denken dat er maar één persoon heeft bijgedragen aan het tot stand komen van een proefschrift. Ongetwijfeld is dat één van de redenen waarom het dankwoord het meest gelezen deel van een proefschrift is. Ook voor dit proefschrift geldt dat ik het nooit had kunnen afronden zonder de hulp van velen. Ik wil allen dan ook ontzettend bedanken voor hun bijdrage aan het tot stand komen van dit proefschrift.

Allereerst wil ik mijn promotor prof. dr. H.W. Frijlink bedanken. erik, jij bent degene die me ooit de kans geboden heeft te gaan promoveren. Ik ben je altijd dankbaar geweest voor het vertrouwen dat je daarmee in me toonde. Maar ik ben je misschien nog wel meer dankbaar voor het vertrouwen dat je de jaren daarna gaf. Mede dankzij jou heb ik het promotietraject als een bijzonder leerzame, maar zeker ook leuke, tijd heb ervaren. je hebt me niet alleen de kans geboden om me te ontwikkelen als onderzoeker, maar ook om veel meer van de wetenschappelijke wereld te ontdekken dan alleen Groningen. Dank je!

Ten tweede wil ik mijn co-promotor dr. W.L.j. Hinrichs bedanken. Wouter, omdat je niet alleen mijn directe begeleider, maar ook mijn kamergenoot was heb ik het geluk gehad niet alleen op vele wetenschappelijke vlakken veel van je te leren, maar ook de nodige small-talk te kunnen delen. Ik heb beide altijd zeer gewaardeerd. Door deze open en directe samenwerking, heb ik het altijd erg naar mijn zin gehad. je bijdrage in het onderzoek zelf, je praktische manier van het oplossen van problemen en je bereidheid iedere keer weer kritisch naar mijn artikelen en presentaties te kijken is van onschatbare waarde geweest! Dank je!

I would also like to thank prof. dr. j. Anwar. jamshed, you were more than hospitable to me during my stay at the Institute of Pharmaceutical Innovation where you introduced molecular modelling to me. I will never forget how you were able to simplify difficult things such as quantum mechanics and molecular interactions. ‘People in a room’ is for me the perfect representation of the interaction between atoms. The scientific baggage you gave me has influenced my scientific thinking ever since. I would also like to thank you, as a member of the reading committee for reading my dissertation and your valuable comments and suggestions. Thank you!

Uiteraard wil ik ook de andere leden van de leescommissie, prof. dr. W.e. Hennink en prof. dr. S.C. De Smedt bedanken voor de tijd die ze hebben genomen om dit proefschrift te lezen en voor hun waardevolle commentaar en suggesties.

Mijn directe (oud)collega’s van de vakgroep farmaceutische technologie en biofarmacie wil ik bedanken voor de prettige werkomgeving. Paul, klaas, Yu San, Marinella, Tien Thanh, Gerad, kees, Anne, Doetie, Maarten, Anne, joke, Sonja, Niels, Parinda, Gieta, Marcel, Floris, Peter, Christina, Milica, Bao Tung, Hans, Senthil, jan, Herman, Reinout, Fesia en Tofan, ieder van jullie ben ik ontzettend dankbaar voor de gezellige tijd. Ik heb het altijd het gevoel gehad dat ik, wanneer dan ook bij jullie kon aankloppen. Dank je, cám on, khawp khun, terima kasih, hvala, nandri!

125

dankwoord

In het bijzonder wil ik ‘de jongens van het eerste uur’ Gerrit, jean-Pierre, en vinay bedanken. Gerrit, de tips en trucs die je me als ‘senior’ AIO bijbracht hebben er voor gezorgd dat ik snel thuis was in het reilen en zeilen van het leven als AIO. Dank je! jean-Pierre, ik heb de vele momenten tijdens en na het werk waarbij we beide even los kwamen van de dagelijkse werkzaamheden altijd als erg gezellig en ontspannend ervaren. Dank je! vinay, Main hamesha bhartiya atithya ki srahanaa karta hoon, jo tumne mere prati hamesha dikhaya aur vishesh roop se tumhari saadi ka jasn bahut saandar tha ! Dhanyawad.er ontbreekt echter nog één directe collega. Anko, je verdient het om apart genoemd te worden. je hebt niet alleen een belangrijke waardevolle bijdrage geleverd aan hoofdstuk 3 en hoofdstuk 5, maar eigenlijk heb je me gedurende mijn gehele tijd als AIO bij vele praktische zaken geholpen. Dank je!

verder wil ik Thomas De Beer bedanken voor de zeer prettige samenwerking en zijn waardevolle bijdrage aan hoofdstuk 4. Thomas, vanaf het begin vond ik dat we, ondanks de afstand en onze buitenlandse uitstapjes, op een zeer prettige manier hebben samengewerkt. Ook heb ik je gastvrijheid in Gent altijd erg gewaardeerd. Dank je!Daarnaast wil ik klaas Sjollema bedanken voor de zeer prettige samenwerking en zijn waardevolle bijdrage aan hoofdstuk 5. klaas, je stond altijd klaar om te helpen bij het opzetten, optimaliseren en analyseren van de CLSM analyse van die ‘gele poeders’. Dank je!

I have had the opportunity to perform a half year of my research at the Institute of Pharmaceutical Innovations, Bradford. My stay there was very pleasant due to all my colleagues there, but I would like to mention a few of them separately. Amir, thank you for all your help with setting up my force field files. John, thank you for always being available to help me with setting up the linux systems and your valuable support on chapter 6. victoria, thank you for all your IT-support and taking care of Garfield. Ook ben ik regelmatig op bezoek geweest bij de Universiteit van Gent. Ik wil iedereen van de afdeling farmaceutische analyse en de afdeling farmaceutische technologie dan ook van harte danken voor hun gastvrijheid. Ik wil daarbij met name jean-Paul Remon en Chris vervaet bedanken voor hun gastvrijheid en waardevolle bijdrage aan hoofdstuk 4.

Ich möchte Peter Pfaffenbach und janette Wessel (Solvay Barium Strontium, Hannover) herzlich danken für ihre Hilfe mit der XRPD-analyse. Peter, wie Du dich auch gefühlt hast, ich fand es Super das Du immer begeistert reagiert hast wenn ich wieder Samples schicken wollte. vielen herzlichen Dank, ich hoffe das es Dir gesundheitlich besser gehen wird. jannette, mit eben so viel enthusiasmus hast Du die Samples analysiert wenn Peter nicht da war. vielen Dank!

een van de leukste dingen die ik als AIO heb mogen doen is het begeleiden van studenten. jenny, Carla, Niels, kees, Martin, Nasim, juliana, Frank, eva, Lisanne, Nienke en Maarten, ieder van jullie heeft op directe of indirecte wijze een bijdrage geleverd aan dit proefschrift. Dank je, grazie, paldies!

I performed my research within the framework of project T5-105 of the Dutch Top Institute

126

appendix e

Pharma. I would like to thank all project partners for the collaboration.

Uiteraard wil ik ook graag mijn paranimfen Bart en Wouter van harte bedanken. Bart, sinds het begin van de studie zijn we onafscheidelijk opgetrokken. Ik kan me geen betere vriend indenken waarop ik in welke situatie dan ook kan bouwen. een te lang nachtje in de kroeg, een goed gesprek, een kritische noot, of wat dan ook, in alle gevallen was je er. Wouter, ondanks het feit dat we slechts een jaar als collega’s werkzaam zijn, hebben we wat mij betreft al wel een bijzondere band. De vele avonden eten, borrelen en praten over zinnige of minder zinnige zaken heb ik altijd zeer gewaardeerd!

Tot slot wil ik graag mijn Laura bedanken. Lieve Laura, je hebt me altijd gesteund in de dingen die ik vond die ik moest doen voor dit proefschrift. Zonder jou waren de afgelopen jaren lang zo leuk niet als ze nu waren. Ik had me geen betere steun, maar vooral geen lievere vriendin kunnen wensen!

127

dankwoord

university of groningen a novel bottom-up process to prepare drug nanocrystals … ·...

Documents