A Multi-Technique Approach to the Study of Structural Stability and Desolvation of Two Unusual Channel Hydrate Solvates of Finasteride

A Multi-Technique Approach to the Study of Structural Stability and Desolvation of Two Unusual Channel Hydrate Solvates of Finasteride STEPHEN BYARD,1 ANUJI ABRAHAM,2 PAUL J. T. BOULTON,2 ROBIN K. HARRIS,2 PAUL HODGKINSON2 1

Department of Analytical Sciences, Sanofi-aventis, Alnwick, Northumberland NE66 2JH, UK

2

Department of Chemistry, University of Durham, Durham DH1 3LE, UK

Received 28 March 2011; revised 22 June 2011; accepted 4 August 2011 Published online 9 September 2011 in Wiley Online Library (wileyonlinelibrary.com). DOI 10.1002/jps.22740 ABSTRACT: The dehydration/desolvation of two hydrate solvates of the pharmaceutically important compound finasteride (namely, bisfinasteride monohydrate monotetrahydrofuran and bisfinasteride monohydrate mono-1,4-dioxane) has been studied by solid-state nuclear magnetic resonance, powder X-ray diffraction, thermogravimetric analysis (including coupling with mass spectrometry) and dynamic vapour sorption. The structure is unusual in that water holds the host finasteride molecules together by hydrogen bonding to form channels in which the solvent is sited. Whilst the solvent guest molecules are not strongly bound to the host, their presence is essential for structural stability. Desolvation is not found to occur at a well-defined temperature or even to consistently produce the same anhydrous form (form I vs. form II), but is instead highly dependent on the physical environment and, therefore, on the technique used. This behaviour complicates investigations, but the combination of complementary methods does allow the desolvation to be understood. Water and solvent are shown to be lost simultaneously, with no evidence of an intermediate form or increased mobility of the hydrogen-bonded water molecules. The results are consistent with a model in which structural collapse and rearrangement follows the loss of a small fraction of the solvent molecules from the channel structure, with the final form produced being very sensitive to the presence of water vapour during desolvation. © 2011 Wiley Periodicals, Inc. and the American Pharmacists Association J Pharm Sci 101:176–186, 2012 Keywords: finasteride; hydrates/solvates; desolvation; thermal analysis; TGA-MS; DVS; magic-angle spinning; solid-state NMR; X-ray powder diffraction

INTRODUCTION The pharmaceutical manufacture of a drug substance and subsequent development of a drug product will generally, at some stage in the program, involve interactions between a solvent and the active pharmaceutical ingredients or excipients. This may, for example, include recrystallisation1 and drying of the drug substance, wet granulation with excipients in the presence of an appropriate solvent,2 lyophilisation, or spray drying. Solvent interactions can be critical to the quality and performance of the final product,2,3,4 whether it be through controlling the polymorphic form, formation of solvates, or desolvaCorrespondence to: Paul Hodgkinson (Telephone: +44-191334-2019; Fax: +44-191-384-4737; E-mail: paul.hodgkinson@ durham.ac.uk) Journal of Pharmaceutical Sciences, Vol. 101, 176–186 (2012) © 2011 Wiley Periodicals, Inc. and the American Pharmacists Association

176

tion. The deliberate formation of solvates can offer significant advantages, such as formation of a highly crystalline drug substance with improved physical and chemical stability as compared with nonsolvated forms.5 The cosolvent of marketed products is normally water and the materials are often stoichiometric hydrates. However, the nature and type of solvation can be highly variable depending on the interactions involved,6 which in turn depend on the individual characteristics of the drug substance or excipients. Consequently, if solvates are involved as intermediates, the nature of such systems and (in particular) their desolvation behaviour need to be thoroughly understood. Solvates can be classified broadly into two types: stoichiometric7 or nonstoichiometric.8 The former are generally stable above a critical partial pressure of the solvent (frequently water), below which desolvation will occur. Alternatively, nonstoichiometric

JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 101, NO. 1, JANUARY 2012

DESOLVATION OF TWO UNUSUAL CHANNEL HYDRATE SOLVATES OF FINASTERIDE

solvates may form, in which the solvent is loosely bound in channels within the crystal lattice. Drying processes may yield an isomorphous desolvate, a polymorphic form of the desolvate having a different crystallographic structure, or amorphous material. In other cases, several different hydrates of the same compound have been identified and characterised.9 Because of the complexity and the different mechanisms that may occur, characterising the desolvation process using a number of techniques is necessary. These are needed to directly probe interactions at both molecular and bulk levels and must also be capable of detecting different physical forms of individual components in a complex mixture. There have been many studies of stoichiometric organic hydrates using a range of techniques. For example, Ref. 10 lists (in 2006) 36 investigations of pharmaceutical hydrates involving solid-state nuclear magnetic resonance (NMR). However, usually the examination of dehydration had been limited to thermogravimetric analysis (TGA) and related thermal methods, together with identification of the resulting desolvate structures. Detailed spectroscopic and thermal investigations of dehydration have been less common (but see Vogt et al.11 for one extensive example). Stoichiometric solvates and inclusion complexes other than hydrates have also been the object of many studies (for example Refs.12–16 ), but again the desolvation process has generally not been studied in detail. Cases when the material in question is stoichiometrically both a hydrate and a solvate are less common.17 Desolvation is then clearly more complex, with issues arising as to whether water and solvate are lost simultaneously or consecutively. Detailed examination of desolvation/dehydration for such compounds is very unusual. We have, therefore, chosen to study such a case, namely, that of a channel structure provided by the host molecule finasteride (N-t-butyl-3oxo-4-aza-5"-androst-1-ene-17$-carboxamide), which is a steroidal antiandrogen used in the treatment of benign prostatic hyperplasia. The molecular structure of finasteride is shown below. There are several polymorphs and a number of solvates known (including several solvate hydrates), but not a simple hydrate. At least two anhydrous asolvate enantiotropic polymorphs are known, denoted as I and II. The former is stable at ambient temperature, the latter being formed above about 175◦ C. According to Wenslow et al.,18 the true thermodynamic transition temperature (obtained from solubility extrapolations) is calculated to be 129◦ C. Seven of the currently known solvates form a series of isomorphous bisfinasteride monohydrate monosolvates, which have a common crystal structure with onedimensional channels19,20,21 accommodating the solvent molecules, but in which the water molecules play an essential structural role in holding the finasteride DOI 10.1002/jps

177

Figure 1. The structure of the bisfinasteride monohydrate monosolvates, viewed down the crystallographic a axis, showing how the structure is held together by hydrogen bonds involving water protons. The solvent molecules are omitted, but the channels which contain them are visible perpendicular to the figure.

molecules together by hydrogen bonding (see Fig. 1), thus requiring the finasteride–water stoichiometry of 2:1. In what follows below, the term “solvent” will specifically refer to tetrahydrofuran (THF) or dioxane, as appropriate, and water will be mentioned separately. The present study focuses on two of these isomorphous solvates, namely, bisfinasteride monohydrate monotetrahydrofuran and bisfinasteride monohydrate mono-1,4-dioxane. In contrast to water, which is rigidly held in the structure, the solvent molecules are highly disordered and very mobile inside the channels.22 Their precise positions are not revealed by single-crystal X-ray diffraction, though their presence (and their molecular nature) is clearly shown by NMR.20 However, the guest molecules occupy defined translational positions in the channels so that the disorder seems to be rotational in character and the system appears to be stoichiometric. The solvent molecules appear to be an integral part of the crystal structures, so that their removal at relatively low temperatures results in collapse of the original structure, as described in the present account. The unidimensional nature of the channels presumably facilitates the loss of the solvent molecules. The X-ray results reported in Ref.20 showed the existence of a small amount of extra electron density in the channels, probably implying some additional water content. The aim of the present work was to characterise the desolvation of these channel solvates as their structure is destabilised by heating or by the presence of high levels of humidity or solvent vapour. However, understanding desolvation processes is not a trivial JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 101, NO. 1, JANUARY 2012

178

BYARD ET AL.

task, and a combination of techniques such as diffraction, NMR and gravimetric analysis is invaluable for the study of such complex solvate systems.11 Variabletemperature powder X-ray diffraction (PXRD) and magic-angle spinning (MAS) 13 C NMR can identify what happens to the crystal structure as solvent leaves the system. In the case of the finasteride solvates, the two techniques can clearly differentiate between the solvate structure and the polymorphic forms (I and II) which result from desolvation.20 Mixtures of structures can be readily characterised both qualitatively and (in principle) quantitatively.23 In the case of NMR, the loss of solvent peaks can be directly monitored. However, the PXRD and NMR techniques are not capable of determining the nature of the gases evolved or of measuring their amounts. Conversely, TGA can quantify the amount of material lost, and the combination of TGA and mass spectrometry (MS)24,25 can also identify the gases evolved; the MS results separately and quantitatively identify the different species evolved during desolvation in real time, which is crucial for mixed solvates such as these. In particular, the TGA–MS experiment can determine whether water and solvent are lost at the same time or consecutively. The final technique used herein is dynamic vapour sorption (DVS).26,27 This is used to assess the propensity of the solvates to either lose solvent/water or gain water as a function of ambient relative humidity (RH) or partial pressure of solvent vapour. Schultheiss et al.21 have recently published data on three other isomorphous hydrate solvates (acetone, methyl ethyl ketone (MEK) and toluene) of finasteride and found it difficult to obtain reliable information on desolvation. They used PXRD and TGA, together with solution-state NMR and single-crystal XRD, but apparently did not have access to solid-state NMR, TGA–MS or DVS.

EXPERIMENTAL Synthesis of Finasteride Solvates Finasteride–THF (bisfinasteride monohydrate monotetrahydrofuran, 2C23 H36 N2 O2 ·H2 O·C4 H8 O) and finasteride–dioxane (bisfinasteride monohydrate mono-1,4-dioxane, 2C23 H36 N2 O2 ·H2 O·C4 H8 O2 ) were prepared by dissolving 0.5 g of form I finasteride (supplied by Hikma Pharmaceuticals, Amman, Jordan) in up to 8 mL of THF (Sigma –Aldrich) and 1,4-dioxane (Sigma–Aldrich), respectively. The mixtures were then heated and maintained at 55◦ C in an oil bath, under magnetic stirring until they began to crystallise. The solutions were allowed to slowly cool to room temperature. The samples were left to evaporate to dryness in a fume hood overnight. Adventitious water vapour in the atmosphere suffices for the formation of hydrate solvates. JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 101, NO. 1, JANUARY 2012

The sample of the THF solvate deuterated at exchangeable hydrogens was prepared as above, but using a mix of approximately 2.5 mL of THF and 2.0 g D2 O (99.98% 2 H, Sigma–Aldrich). All the samples were kept refrigerated, in tightly sealed containers, to minimise loss of solvent from the structure at room temperature.

Thermogravimetric Analysis Thermogravimetric analysis experiments were performed using a Mettler Toledo TGA850 module (Mettler-Toledo). Data acquisition and processing were carried out using Mettler-Toledo Stare software (version 9.01). Typically, 15–20 mg of the test substance were gently packed into a 70-:L alumina crucible and the loss in mass recorded over the temperature range 25◦ C–200◦ C, using heating rates as specified in the Results and Discussion section, with a purge gas of dry nitrogen, nominally at atmospheric pressure and at a flow rate of 70 mL/min. A blank curve acquired using identical experimental conditions, but with an empty crucible, was subtracted from each thermogram. The temperature and the temperature lag (of the furnace and related components, compensated for in the heating rates) were calibrated using the melting points of traceable standards of indium and zinc. The balance calibration was verified using standard check weights in conjunction with the loss observed for a traceable standard of calcium oxalate monohydrate.

TGA with MS A PerkinElmer Pyris 1 thermogravimetric analyser (PerkinElmer) coupled to a HPR20 Hiden Analytical mass spectrometer (Hiden Analytical) was used for the TGA–MS experiments. Finely powdered sample (5–10 mg) was placed in a small open-top crucible and heated at defined rates ranging from 0.5 to 10◦ C/min over a temperature range of 25◦ C–200◦ C (monitored by a calibrated thermocouple very close to the sample). All TGA analyses were carried out at a nominal atmospheric pressure of the carrier gas (dry helium) passing over the heated sample at approximately 20 mL/min, although the exact pressure and flow conditions at the sample are not well established. The entire system was purged by dry helium gas for 3 h prior to each run. Helium is used because it is more efficient than nitrogen at removing adventitious solvent, though by the same token it may affect the temperature profile of desolvation. Calcium oxalate monohydrate was used to verify that the amount of water desorbed from both the solvate samples was consistent with the known stoichiometry. DOI 10.1002/jps

DESOLVATION OF TWO UNUSUAL CHANNEL HYDRATE SOLVATES OF FINASTERIDE

Dynamic Vapour Sorption Dynamic vapour sorption experiments were performed using a DVS-1 instrument supplied by Surface Measurement Systems (London, UK). Prior to analysing the test substances, performance checks were carried out using saturated solutions of sodium chloride and lithium chloride traceable standards. The RH cycle was taken from 30%–95% to 0%–30% at 25◦ C. For water-vapour-based experiments, RH changes of 10% were used, except for the step between 90% and 95% RH. This cycle was completed twice for each sample. A target equilibrium condition of mass change less than 0.0005% over 5 min was set, with minimum and maximum equilibrium periods of 10 and 360 min, respectively. The carrier gas was nitrogen, with a flow rate of 100 mL/min in sample and reference ports. Typically, 40–45 mg of drug substance were placed in a gauze basket for the analysis. Vapour sorption–desorption experiments involving vapour pressures of dioxane in the absence of water vapour were also carried out, with various cycles designed to probe the physical transformations of the drug substance; details of each experiment are described in the Results and Discussion section. NMR Spectroscopy For NMR, the samples consisted of polycrystalline material tightly packed in capped rotors of 5-mm outside diameter. Note that the rotor caps contain small diameter apertures to allow for pressure equalisation. These permit a limited degree of exchange between the interior of the rotor and the exterior atmosphere (dry air). Spectra were obtained using an InfinityPlus 500 spectrometer (Varian Inc., now Agilent Inc.), for which the Larmor frequencies of 13 C and 2 H are 125.7 and 76.71 MHz, respectively. For the 1 H → 13 C NMR work, a MAS rate of 8.0 kHz was employed, and we note that the material in rotors would be subject to high centrifugal forces at the periphery at these spinning rates. The crosspolarisation (CP) contact time was 2 ms and the recycle time was 4.0 s; 150 free induction decays were coadded for each spectrum, and proton decoupling was carried out using the TPPM (two-pulse phasemodulated) pulse sequence.28 Spectra were recorded at ambient probe temperature (ca. 25◦ C), at 10◦ C intervals between 50◦ C and 110◦ C, at 115◦ C and finally at 125◦ C, using an equilibration time of at least 15 min at each temperature change. Note that the temperatures quoted here are nominal and the reported temperatures include a 10 K correction for the heating effects due to the sample spinning. The total experimental time was approximately 8–10 h for each solvate. For the 2 H study, the spinning rate was 7 kHz and the pulse duration was 7 :s (a nominal 90◦ pulse DOI 10.1002/jps

179

angle). 1 H decoupling was not used, as this was not found to improve the spectral resolution noticeably. Spectra were recorded at 10 intervals between –50◦ C and +50◦ C, plus 65◦ C and 75◦ C (nominal temperatures–true temperatures are ca. 7 K higher at this MAS rate). Powder X-ray Diffraction A D8 Advance diffractometer (Bruker) was used to carry out PXRD experiments over the temperature range 30◦ C–150◦ C. The samples were lightly ground and then passed through sieve size 60 (25 :m). Thin uniform layers were deposited onto a silicon disc 15 mm in diameter. The samples were exposed to heated air in an Anton Paar HTK1200 furnace (Anton Paar) at ambient atmospheric pressure during the experiments. All the samples were scanned in a 2θ range between 5◦ and 40◦ , with a 6-mm variable-divergence slit, using Cu K" radiation and a step time of 2.0 s. The total experimental time for each solvate was approximately 5 h.

RESULTS AND DISCUSSION In many cases, desolvation is determined by factors such as particle size, packing, local humidity, sample history and so on, so that very different desolvation profiles may be observed by different techniques. Hence, relying on individual techniques is often misleading, particularly if the environmental conditions in a particular analysis are far removed from those encountered in production. The information provided by a range of complementary techniques allows a detailed understanding of the transformation to be inferred. Figure 2 shows 13 C CPMAS spectra of the two solvates as the temperature is raised. The spectra remain essentially unchanged up to about 110◦ C and show the resonances from the solvate guest molecules. Above that temperature, desolvation occurs; the solvent resonances (indicated by grey arrows) are gradually lost and, simultaneously, peaks arising from polymorph I (e.g, those indicated by black arrows in Fig. 2) appear.20 The assignment to the orthorhombic form I is unequivocal because, in distinction to form II, it has only one molecule in the asymmetric unit. This contrasts with the observation of Schultheiss et al.21 that form II is obtained from desolvation of acetone and MEK solvates under 72 h of vacuum drying at ambient temperature. Because the transformation takes up to 2 h at 20◦ C to complete, it is likely that the constraints of rotor containment prevent complete desolvation on the timescale of the experiment, implying that the conversion temperature is likely to be significantly lower in an open atmosphere, and the continued presence of water/solvent vapour in the rotor JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 101, NO. 1, JANUARY 2012

180

BYARD ET AL.

(a) 135°C

120°C

110°C

100°C

90°C

35°C

70

60

50

70

60

50

40 δC (ppm)

30

20

10

40

30

20

10

(b) 135°C

120°C

110°C

100°C

90°C

35°C

δC (ppm)

Figure 2. Variable-temperature 13 C CPMAS spectra of (a) the finasteride hydrate dioxane solvate and (b) the finasteride hydrate THF solvate, displaying the behaviour under desolvation. Some distinctive peaks for form I are indicated by black arrows. Grey arrows indicate solvent resonances (67.5 ppm for dioxane and 26.5/68.1 ppm for THF). The THF resonances in (b) are rather weak, presumably because the CP efficiency is low in these cases. The temperatures given at the left-hand side are calibrated. Spectra obtained between 35◦ C and 90◦ C are uninformative and not displayed.

(in spite of the small hole in the drive cap) may affect the nature of the polymorph formed. Uniquely, the 13 C NMR is able to correlate the structural changes with the loss of solvent (PXRD is not able to monitor the disordered solvent). Crucially, the signals from the finasteride carbons in the solJOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 101, NO. 1, JANUARY 2012

vate are lost (and those of form I arise) at the same time as the dioxane or THF peaks are lost. Note that there are no splittings, shifts or broadenings in the host resonances that would be expected if the solvate structure were retained but with lower fractions of included solvent. This contrasts, for example, with DOI 10.1002/jps

181

DESOLVATION OF TWO UNUSUAL CHANNEL HYDRATE SOLVATES OF FINASTERIDE

(a)

82°C

77°C

T = 28°C (Cool) T = 100°C 72°C

T = 90°C T = 80°C T = 74°C 57°C

T = 40°C

5

10

15

20 2θ (°)

25

30

35

(b) 150

100

50 2

0

-50

-100

-150

H frequency (kHz)

Figure 3. Variable-temperature 2 H MAS spectra for the finasteride THF D2 O solvate, obtained at a MAS rate of 7 kHz. True probe temperatures (calibrated) are given at the left-hand side. The same vertical scale was used for all the spectra.

T = 28°C (Cool) T = 100°C T = 90°C

the channel hydrate sildenafil citrate, where shifts in the 13 C resonances can be directly correlated with the degree of occupancy of the channels.29 Moreover, there is no indication of peaks that cannot be assigned to either the solvate or form I, that is, no intermediate state can be seen. In particular, no amorphous form can be detected. Thus, it appears that loss of solvent molecules is directly connected with the collapse of the structure. Figure 3 illustrates variable-temperature 2 H spectra recorded for the sample of the THF solvate hydrate deuterated at exchangeable hydrogens. The wide spread of the spinning sideband manifolds shows that, as expected from the crystal structure (which indicates that water molecules link the finasteride molecules by hydrogen bonding), the water is relatively rigidly held at all the temperatures – spinning sideband analysis shows that the quadrupolar parameters (χ = 165 MHz, η = 0.2) do not vary significantly over the temperature range recorded. Although the anisotropy is somewhat lower than expected for fully rigid water, the low asymmetry parameter implies, somewhat surprisingly, that D2 O is not undergoing rapid C2 flips at these temperatures.30 However, substantial loss of water (which must lead to structural collapse) is clearly occurring above 75◦ C. Note that the lower signal heights are genuinely due to reDOI 10.1002/jps

T = 80°C T = 74°C T = 40°C

5

10

15

20 2θ (°)

25

30

35

Figure 4. Variable-temperature powder X-ray diffractograms displaying the desolvation behaviour of (a) finasteride hydrate dioxane solvate and (b) finasteride hydrate THF solvate.

duced NMR signal, rather than to broadening caused by motional processes. Similar experiments were not performed on the dioxane solvate because the 2 H spectra in this case are complicated by the effects of conformational exchange in the dioxane ring.22 Figure 4 depicts the variable-temperature PXRD patterns for the two solvates. Desolvation is occurring between 80◦ C and 90◦ C, which is somewhat lower than in the 13 C NMR variable-temperature experiments. Moreover, comparison with the literature and the PXRD patterns predicted from the Cambridge Crystallographic Structure Database shows that, in contrast to the NMR case, form II is produced. This is JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 101, NO. 1, JANUARY 2012

182

BYARD ET AL.

quite clear because this form has three strong diffraction lines below 12◦ in 2θ, whereas form I has none in that region. Indeed, these lines suggest the presence of a small amount of form II below 80◦ C in both cases. No other forms can be detected at any stage. The relatively open nature of the sample in the PXRD experiment (in contrast to the NMR situation) means that following expulsion from the finasteride channels, solvent or water molecules are not retained near the sample. The resulting vapour pressure gradient promotes desolvation. Thermogravimetric analysis may be used to determine both the temperature profile of desolvation and the mass of the solvent lost. The theoretical values for total mass loss of water from the bisfinasteride monohydrate monosolvates are 2.2% and 2.1% for the THF and dioxane solvates, respectively. The corresponding values for loss of solvent are 8.6% and 10.4%, respectively, making totals for loss of both water and solvent 10.8% and 12.5%, respectively. Our previous TGA study of finasteride solvates20 (which used a heating rate of 2◦ C/min) showed a total loss in mass of 13.0% and 13.2% for the THF and dioxane solvates, respectively, but these numbers probably include some adventitious solvent or water on crystallite surfaces or between particles. Figure 5 shows TGA experiments for the dioxane solvate conducted at a range of heating rates (similar plots were obtained for the THF solvate). The temperature of maximum rate of change in mass for the dioxane solvate increased substantially from 59◦ C, when the rate was 0.5◦ C/min, to 114◦ C at a rate of 20◦ C/min. Such behaviour is not particularly unusual for kinetically controlled desolvation of a stoichiometric solvate. Small mass losses are observed in the build up to the main desolvation process. These are less obvious in the TGA–MS results (discussed below), where heating was preceded by a long purge

10%

0.5°C / min 2 °C / min 5 °C / min 10 °C / min 20 °C / min

40

60

80

100 120 T (°C)

140

160

180

Figure 5. Thermogravimetric analysis traces of mass loss versus temperature for desolvation of the finasteride dioxane solvate hydrate at various heating rates. JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 101, NO. 1, JANUARY 2012

period with dry helium. This suggests that these initial losses are largely associated with adventitious water/solvent, which is consistent with the indicated total mass losses being in excess of the theoretical value. Schultheiss et al.,21 in their study of three other finasteride hydrate solvates, had difficulty in fully ascribing the mass losses during TGA, and concluded that “additional solvent and/or water within the lattice” was present. Clearly, it is difficult to relate experimental mass losses to theoretical values based on the expected stoichiometry in this case. Because the diffraction-derived structure shows these systems to be monohydrates, it appears to be likely (as suggested by the DVS experiments; see below) that additional water may be incorporated in some way into the samples (either on particle surfaces or in the channels), thus creating variability in content. Indeed, when a sample of the THF solvate exposed to a D2 O atmosphere for 2 weeks was subjected to analysis by 2 H MAS NMR, it was discovered that, whilst there was no broad signal (i.e., no chemical exchange of H between deuterium oxide and NH/structural water), a relatively narrow peak was seen (at 4.6 ppm), presumably arising from mobile adventitious deuterated water. Further analysis was carried out using DVS at 25◦ C to assess the lability of the solvates under a range of different RH and solvent vapour pressures. The DVS isotherm plots (Fig. 6) show that the solvate structure is stable at RH values between 30% and 90% (as implied by the long-term stability of the solvate under ambient conditions). There is a small additional uptake of water as the RH increases, consistent with the presence of adventitious (i.e., nonstructural) water at ambient conditions. The experiment also shows that exposing the solvate to very high humidity levels (RH > 90%) leads to loss of solvent and water, with anhydrous form I being formed (as determined by PXRD of the final sample). Although the formation of the structure with less water than the starting material at high humidities seems paradoxical, this is consistent with the lack of evidence for a simple hydrate (which would otherwise be expected to be formed in these conditions). In the case of the finasteride solvates, the solvent clearly plays a significant role in maintaining lattice structure, as shown below by the TGA–MS experiments; the channel structure is not adopted in the absence of organic solvent. It seems plausible that at elevated RH, additional water can displace solvent from the channels, “destabilising” the channel structure. It is important to note that the exact mechanism of desolvation cannot be observed experimentally, so that alternative mechanisms such as plasticisation of surface amorphous or disordered material cannot be ruled out. It seems unlikely that the transformation is determined by the presence of amorphous/disordered DOI 10.1002/jps

DESOLVATION OF TWO UNUSUAL CHANNEL HYDRATE SOLVATES OF FINASTERIDE

(a) 18

Mass change/%

16 14 12

Sorption Desorption

10 8 6 4 2 0 0

10

20

30

40 50 60 70 Relative humidity/%

80

90

100

80

90

100

(b) 18 16

Mass change/%

14

Sorption 1 Desorption 1

12

Sorption 2 Desorption 2

10 8 6 4 2 0 10

20

30

40 50 60 70 Relative humidity/%

Figure 6. Dynamic vapour sorption isotherm plots of (a) the finasteride hydrate THF solvate and (b) the finasteride hydrate dioxane solvate, showing the loss of mass at high humidity. The mass changes are shown as a percentage of the mass at the end of the experiment (i.e., of the asolvate). The slow initial gain in mass may be ascribed to adsorption of water on crystallite surfaces or in interstices between crystallites. Note the incomplete transformation after one sorption–desorption cycle in the case of the dioxane solvate.

sites because the structural collapse can be controlled by cycling RH. This is shown in Figure 6b for a DVS trace of the dioxane solvate in which the first desorption cycle was initiated before equilibrium had been fully established with respect to mass change. We conclude that structure collapse/recrystallisation triggered by solvent displacement and/or local dissolution is the more likely origin for the change in the physical form. The DVS experiment thus adds to the evidence that the solvent molecules are vital for the structural integrity of the lattice, in spite of the fact that they are mobile and have no obvious bonding effects with finasteride. Their influence would seem to arise from a templating effect, relying solely on their shape or volume. The situation parallels that for inclusion compounds involving urea as a host with guests such as alkanes31,32 ; again, there appears to be DOI 10.1002/jps

183

little or no bonding between the host and guest,33 yet the host structure collapses as soon as the guests are removed. For the dioxane solvate hydrate, DVS experiments were also carried out as a function of the partial pressure of dioxane vapour (in nitrogen carrier gas) in the absence of water vapour at 24.6◦ C. A mass loss of approximately 6.1% occurred over 6 h under zero dioxane partial pressure. Increasing the pressure of dioxane to 10% and then to 20% of its saturation value resulted in further mass losses, but with additional dioxane vapour pressure, the mass remained relatively constant until 70% saturation (giving presumably an asolvate). There was, then, a small increase in mass as the partial pressure approached 90%, probably because of some surface adsorption. A desorption cycle resulted in a small additional mass loss, giving an overall loss of 15.6%. When the DVS experiments started in the presence of dioxane partial pressure at 95% of its saturation value, there was a small mass gain initially, ascribed to surface adsorption, followed by considerable mass loss over a period of 6 h. Similarly, during experiments carried out at a constant dioxane pressure of 60%, a small initial mass gain became a net 13% mass loss over a period of 20 h. These experiments show that the structure is unstable at all values of dioxane vapour pressure in the absence of water, as might be expected from the vital structural role of water molecules. This fact helps to explain the apparently variable stability of the structure as a function of temperature determined by different techniques, since the initial state (surface adsorption) and, particularly, the ambient conditions specific to the technique will have a major effect on the results. Interestingly, PXRD patterns recorded immediately after the dioxane-exposure DVS experiments show that the anhydrous form II is obtained, in contrast to the results following water-exposure DVS experiments. Thus, the presence of water vapour appears to determine the polymorph formed on desolvation. The TGA–MS experiments illustrated in Figure 7 provide vital information about the nature of the solvent and water loss. In the case of THF, a mass loss of just over 10% is observed (at 0.5◦ C/min heating rate) as a single event, with a peak loss rate at 76◦ C–a value which increased with heating rate as expected. For the dioxane solvate, the TGA trace (Fig. 5) shows a total loss of approximately 13%, in line with the expectations. In this case, the maximum rate of loss (at 0.5◦ C/min heating rate) occurs between 40◦ C and 50◦ C. The peak loss rates observed using the TGA–MS technique proved to be somewhat variable, and tended to differ from values obtained by simple TGA measurements at the same heating rate. The principal result deduced from these experiments is that the main losses of water and solvent occur simultaneously. This strongly suggests that the JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 101, NO. 1, JANUARY 2012

184

BYARD ET AL.

CONCLUSIONS

(a) THF Water

0

50

20

150 Time (min) T (°C) 80 100

60

200

120

250

140

mass (%)

100 98 96 94 92 90 88

40

100

0

50

100

150 Time (min)

200

250

200

250

(b) Dioxane Water

0

50

40

100

60

150 Time (min)

80

T (°C) 100

120

140

mass (%)

100 95 90 85

0

50

100

150 Time (min)

200

250

Figure 7. Thermogravimetric analysis–mass spectrometry experiments for the finasteride hydrate solvates: (a) THF solvate and (b) dioxane solvate. In each case, the lower trace is the TGA plot at a heating rate of 0.5◦ C/min, whereas the upper trace is the corresponding development of MS peaks (the solid line corresponds to THF/dioxane and the broken line corresponds to water). The vertical scales of the MS plots are arbitrary.

loss of the solvate triggers structural collapse, without the formation of an intermediate desolvated but hydrated form. High-temperature tails are observed in the TGA–MS plots for both solvates (especially the dioxane solvate). However, the very long tail observed in the MS results for water is a recognised phenomenon of TGA–MS coupling and is not relevant to the desolvation itself.

JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 101, NO. 1, JANUARY 2012

These isomorphous channel solvates are unusual in that the host structure (in which water plays an essential role in linking finasteride molecules via hydrogen bonding) appears to be templated on and stabilised by the presence of highly mobile solvent molecules; the host structure becomes unstable on removal of the guest, despite the weak nature of the host–guest (and guest–guest) interactions. This considerably complicates the characterisation of the desolvation process. Using a combination of “bulk” techniques such as DVS with molecular-level methods (PXRD and CPMAS NMR), together with the hybrid TGA–MS method, we have been able to construct a fairly comprehensive picture of the desolvation process in these systems. Whilst the details of the temperature of desolvation (accompanied by structural collapse) and of the measured mass losses depend on both the initial state of the sample (size and shape of the particles) and the sample handling and physical environment appropriate for the technique used, the structural collapse on heating can be associated with the loss of a relatively small fraction of the solvent from the channels, followed by simultaneous loss of the solvent and water as the structure collapses and reforms as an asolvate. TGA experiments show that the temperature of the structural collapse increases markedly with increased heating rate, confirming that the behaviour of the collapse is kinetically (rather than thermodynamically) controlled. The somewhat lower desolvation temperatures found for the finasteride– dioxane solvate hydrate than for the finasteride–THF solvate hydrate suggest that loss of solvent is the initiating factor when heating, rather than the loss of structural water. (The reverse would be expected to be true when structural collapse is triggered by lowered humidity levels.) Note how the desolvation temperatures are in the opposite order to that expected on the basis of the vapour pressures for pure dioxane versus THF (boiling points 101◦ C and 66◦ C, respectively), which is consistent with the stability of the solvent within the solvate structure itself being determinant. Exposure to very high RH also triggers structural collapse and formation of an anhydrate. This somewhat surprising observation is presumed to be due to the displacement of solvent molecules in the channels by water at very high humidities, although the exact mechanism of the local reordering cannot be established experimentally. Deuterium NMR shows that the structural water is relatively rigidly held throughout (consistent with its role in stabilising the host finasteride tunnel structure by hydrogen bonding). Although it is not feasible to estimate the degree of solvent/water loss required

DOI 10.1002/jps

DESOLVATION OF TWO UNUSUAL CHANNEL HYDRATE SOLVATES OF FINASTERIDE

to trigger the collapse, the 13 C NMR spectra show no evidence of a partially desolvated structure. The polymorph produced by structural collapse is found to depend, inter alia, on the experimental technique used, presumably because of the different sample environments. Variable-temperature NMR and DVS under water vapour conditions lead to form I, whereas variable-temperature PXRD and DVS under dioxane vapour conditions give form II. The difference in the polymorph formed appears to depend on whether water vapour is present (as it would be in the enclosed nature of the NMR experiments). There is no evidence for intermediate crystalline or amorphous forms, though a transient formation of such species cannot be entirely ruled out.

ACKNOWLEDGMENTS This work was supported by EPSRC grant EP/ D057159. Jeay Upton is thanked for initial exploratory work using TGA–MS. We are also grateful to Doug Carswell for most TGA–MS experiments, to Jackie Moseley for help with interpreting them, to David Apperley for some of the NMR experiments, to Abdullah Othman and Andrew Ilott for some of the 13 C NMR and PXRD work and to Rebecca Yue for some PXRD experiments. The referees are thanked for their helpful comments on the original script.

REFERENCES 1. Variankaval N, Lee C, Xu J, Calabria R, Tsou N, Ball R. 2007. Water activity-mediated control of crystalline phases of an active pharmaceutical ingredient. Org Process Res Dev 11:229–236. 2. Tantry JS, Tank J, Suryanarayanan R. 2007. Processinginduced phase transitions of theophylline–Implications on the dissolution of theophylline tablets. J Pharm Sci 96:1434–1444. 3. Griesser UJ. 2006. The importance of solvates (chapter 8). In Polymorphism in the pharmaceutical industry; Hilfiker R, Ed. Wiley-VCH Verlag GmbH & Company (Weinheim, Germany), pp 211–233. 4. Govindarajan R, Suryanarayanan R. 2006. Processinginduced phase transformations and their implications on pharmaceutical product quality (chapter 13). In Polymorphism in the pharmaceutical industry; Hilfiker R, Ed. Wiley-VCH Verlag GmbH & Company (Weinheim, Germany ), pp 333–363. 5. Hickey MB, Peterson ML, Manas ES, Alvarez J, Haeffner F, Almarsson O. 2007. Hydrates and solid-state reactivity: A survey of beta-lactam antibiotics. J Pharm Sci 96:1090–1099. 6. Newman AW, Reutzel-Edens SM, Zografi G. 2008. Characterisation of the “hygroscopic” properties of active pharmaceutical ingredients. J Pharm Sci 96:1047–1059. 7. Khankari RK, Grant DJW. 1995. Pharmaceutical hydrates. Thermochim Acta 248:61–79. 8. Authelin JR. 2005. Thermodynamics of non-stoichiometric pharmaceutical hydrates. Int J Pharm 303:37–53. 9. Reutzel-Edens SM, Bush JK, Magee PA, Stephenson GA, Byrn, SR. 2003. Anhydrates and hydrates of olanzapine: Crystallisation, solid-state characterization, and structural relationships. Cryst Growth Des 3:897–907. DOI 10.1002/jps

185

10. Harris RK. 2006. NMR studies of organic polymorphs and solvates. Analyst 131:351–373. 11. Vogt FG, Brum J, Katrincic LM, Flach A, Socha JM, Goodman RM, Haltiwanger RC. 2006. Physical, crystallographic, and spectroscopic characterization of a crystalline pharmaceutical hydrate. Understanding the role of water. Cryst Growth Des 6:2333–2354. 12. Reutzel-Edens SM, Kleemann RL, Lewellyn PL, Borghese AL, Antoine LJ. 2003. Crystal forms of LY334370 HCl: Isolation, solid-state characterization, and physicochemical properties. J Pharm Sci 92:1196–1205. 13. Vogt FG, Cohen DE, Bowman JD, Spoors GP, Zuber GE, Trescher GA, Dell’Orco PC, Katrincic LM, Debrosse CW, Haltiwanger RC. 2005. Structural analysis of polymorphism and solvation in tranilast. J Pharm Sci 94:651–665. 14. Harris RK, Ghi PY, Puschmann H, Apperley DC, Griesser UJ, Hammond RB, Ma C, Roberts KJ, Pearce GJ, Yates JR, Pickard CJ. 2005. Structural studies of the polymorphs of carbamazepine, its dihydrate, and two solvates. Org Process Res Dev 9:902–910. 15. Gajda J, Jeziorna A, Clesielski W, Potrzebowski WM, Prezdo WW, Potrzebowski MJ. 2007. High-resolution solid-state NMR spectroscopy as a tool for investigation of enantioselective inclusion complexation. Solid State Nucl Magn Reson 31:153– 161. 16. Harris RK, Hodgkinson P, Larsson T, Muruganantham A, Ym´en I, Yufit DS, Zorin V. 2008. Characterization of polymorphs and solvates of terbutaline sulphate. Cryst Growth Des 8:80–90. 17. G¨orbitz CH, Hersleth HP. 2000. On the inclusion of solvent molecules in the crystal structures of organic compounds. Acta Crystallogr B B56:526–534. 18. Wenslow RM, Baum MW, Ball RG, McCauley JA, Varsolona RJ. 2000. A spectroscopic and crystallographic study of polymorphism in an aza-steroid. J Pharm Sci 89:1271–1277. 19. Wawrzycka I, Stepniak K, Matyjaszczyk S, Koziol AE, Lis T, Abboud KA. 1999. Structural characterization of polymorphs and molecular complexes of finasteride. J Mol Struct 474:157– 166. 20. Othman A, Evans JSO, Evans IR, Harris RK, Hodgkinson P. 2007. Structural study of polymorphs and solvates of finasteride. J Pharm Sci 96:1380–1397. 21. Schultheiss N, Smit JP, Hanko JA. 2009. Three isostructural solvates of finasteride and their solid-state characterization. Eur J Pharm Sci 38:498–503. 22. Hogg HM, Boulton PJT, Zorin VE, Harris RK, Hodgkinson P. 2009. Use of rotary echoes in 2 H magic-angle spinning NMR for the quantitative study of molecular dynamics. Chem Phys Lett 475:58–63. 23. Apperley DC, Harris RK, Larsson T, Malmstrom T. 2003. Quantitative nuclear magnetic resonance analysis of solid formoterol fumarate and its dihydrate. J Pharm Sci 92:2496–2502. 24. Kaisersberger E, Post E. 1997. Practical aspects for the coupling of gas analytical methods with thermal-analysis instruments. Thermochim Acta 295:73–83. 25. Kamruddin M, Ajikumar PK, Dash S, Tyagi AK, Raj B. 2003. Thermogravimetry-evolved gas analysis-mass spectrometry system for materials research. Bull Mater Sci 26:449–460. 26. 2007. Water-solid interactions–Sorption isotherms, water activity. Pharmeuropa 19:151–154. 27. Burnett DJ, Thielmann F, Sokoloski TD. 2007. Investigating carbamazepine-acetone solvate formation via dynamic gravimetric vapour sorption. J Therm Anal Calorim 89:693– 698. 28. Bennett AE, Rienstra CM, Auger M, Lakshmi KV, Griffin RG. 1995. Heteronuclear decoupling in rotating solids. J Chem Phys 103:6951–6958. JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 101, NO. 1, JANUARY 2012

186

BYARD ET AL.

29. Apperley DC, Basford PA, Dallman CI, Harris RK, Kinns M, Marshall PV, Swanson AG. 2005. Nuclear magnetic resonance investigation of the interaction of water vapour with sildenafil citrate in the solid state. J Pharm Sci 94:516– 523. 30. Weiss A, Widen N. 1980. Deuteron magnetic resonance in crystal hydrates. Adv Nucl Quadrupole Reson 4:149– 248.

JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 101, NO. 1, JANUARY 2012

31. Takemoto K, Sonoda N. 1984. Inclusion compounds. Vol. 2. New York: Academic Press, p 47. 32. Harris KDM. 1997. Understanding the properties of urea and thiourea inclusion compounds. Chem Soc Rev 26:279–289. 33. Ilott AJ, Palucha S, Batsanov AS, Harris KDM, Hodgkinson P, Wilson MR. 2011. Structural properties of carboxylic dimers confined within the urea tunnel structure: An MD study. J Phys Chem B 115:2791–2800.

DOI 10.1002/jps