A MECHANISTIC STUDY TO IDENTIFY PHYSICALLY STABLE AMORPHOUS SOLID DISPERSION

Amorphous solid dispersions are known to improve the oral bioavailability of poorly water-soluble drugs. However, the physical instability of solid dispersions leading to phase separation and subsequent crystallization is limiting their commercial use. Polymers used in solid dispersions have shown to inhibit drug crystallization by either increasing the glass transition temperature (T g) of the mixture; and/or by interacting with the drug. To effectively inhibit drug crystallization, the polymer has to remain miscible with the drug. Drug crystallization could occur if drug-carrier miscibility is adversely affected by heat or humidity. It is therefore important to understand the drug-carrier miscibility to define strategies for ensuring the physical stability of amorphous solid dispersions. In this study, an approach is presented with which one can determine the "solid solubility", defined as the amount of drug that remains miscible with the carrier under specified heat and humidity condition. Modulated differential scanning calorimeter (MDSC) was used to determine the solid solubility of trehalose, griseofulvin, and indoprofen, in amorphous polymers like PVP and dextran . Solid dispersions that exhibit single Tg, were exposed to accelerated storage conditions and their Tg(s) were monitored periodically. An increase in Tg or the fonnation of multiple Tgs indicated the phase separation of drug from the polymer. The Tg was monitored until a plateau was reached, indicating an equilibrium state and no further phase separation. The solid solubility of drug in the polymer was detennined from the calibration plot of Tg of freshly prepared solid dispersion vs. drug-polymer ratio. The mechanism of solid solubility was elucidated and hydrogen bonding was shown to play a critical role in enhancing the solid solubility. lndoprofen that hydrogen bonds to PVP had solid solubility of l 3%w/w, whereas griseofulvin which had no hydrogen bonding to the polymer crystallized completely from the solid dispersions when stored at accelerated conditions of 40°C and 69% relative humidity. In the case of trehalose and dextran miscible solid dispersions the solid solubility of trehalose seemed to decrease with increasing storage temperatures and moisture levels . The kinetic rate of indoprofen and griseofulvin phase separation from the polymer was estimated by fitting the Tg of miscible mixtures, to the first-order rate equation. The phase separation rate of indoprofen was found to be at least 10 times lower than that of griseofulvin. The effect of surfactants on the drug solid solubility in the polymer was also investigated. To probe into the factors responsible for physical instability of amorphous solid dispersions, the molecular mobility of a model drug-griseofulvin and the polymer-PVP was studied. Isothermal and non-isothermal crystallization studies were conducted on griseofulvin to determine the activation energy for crystallization. Molecular mobility of PVP was studied using thermally stimulated current spectroscopy and the factors influencing the molecular motions were studied.

B-V-11 TSC net current plotted against the storage time for PVP aged at respective Tg-15°C for PVP Kl 7, Relationship between the enthalpy recovery measurements from MDSC and mobility recovery measurements from modified TSPC. .. 196

xvii SECTION A Overall Thesis Introduction
Pharmaceutical solid dispersion technology is generally accepted as a technique to enhance the dissolution characteristics of drugs with poor water solubility [1][2][3] . The drug substance is dispersed in a water soluble inert polymer matrix, sometimes at the molecular level, and the higher surface area due to the presence of polymer may increase the drug solubility and dissolution rate [4]. If the drug is not dispersed on a molecular scale, it is often physically modified from its crystalline to amorphous fonn during processing [5]. The presence of high-energy amorphous state leads to enhanced dissolution rate and hence bioavailability owing to the lack of crystalline lattice [6]. The prime reason for the increasing popularity of solid dispersion technology for oral solid dosage formulations is the improvement in the solubility and bioavailability of poorly water soluble drugs [7].

Solid dispersion technology: an attractive alternative
For over centuries the water solubility of pharmaceutical drugs was improved by techniques like particle size reduction through grinding [8]. Eventually sophisticated techniques like modification of the chemical structure of drug molecule via pro-drug formation, selection of suitable salt form, and complexation with solubilizing agents generated increasing attention [9]. Lately however, increasing number of highly lipophilic drug molecules are being synthesized through combinatorial chemistry synthesis program and the conventional techniques are unable to enhance the drug dissolution rate and bioavailability of such hydrophobic drugs [I OJ. Majority of the drug molecules that are recently entering into the pharmaceutical company pipeline have intrinsic water solubility of less than 1 µg/mL when compared to conventional I 00 µg/mL. Therefore there is a need for newer formulation approaches that could help meet the challenge of providing adequate drug plasma concentration for therapeutic efficacy.
Solid dispersion technology proves to be an attractive alternative for such drug molecules.
Conventionally, converting the physical state of drug substance from crystalline to its high-energy amorphous state was used as an approach to improve the drug dissolution performance. This was primarily done since the aqueous solubility of amorphous drug can be as high as 1000 fold greater than the crystalline form [11] . Such an amorphous state, however, is thermodynamically unstable and under certain conditions of temperature and relative humidity during storage it could crystallize [12]. Solid dispersion formu lation could be an attractive alternative to stabilize such amorphous drug substances while improving the drug dissolution rate and bioavailability [ 13].

Solid dispersions: challenges and opportunities
Although solid dispersion technique has generated several research contributions for over four decades, the number of marketed formulations that use this technique is rather disappointing. Only a griseofulvin-poly (ethylene glycol) solid dispersion (Gris-PEG, Novartis) and a nabilone-povidone solid dispersion (Cesamet, Lilly) were commercially available during the past four decades. There can be various subjective reasons for such a scenario. A drug candidate that is suitable to be formulated via this technique can 'die' during the pharmaceutical pipeline due to reasons like toxicity manifestation, low therapeutic efficacy, chemical instability, extremely low water solubility, and various formulation challenges. However in the case of solid dispersion, the challenges have been on a more generalized scale to prevent commercial products for several years. One of the challenges associated with this technology is the physical instability of the amorphous drug substance dispersed in the polymeric matrix during shelf life storage [14]. Other challenges include the inability to scale-up the solid dosage formulation from the typical small-scale melt-quench, or solvent-evaporation technique; inadequate knowledge of the mechanism of dissolution of drugs from the dosage forms , prevention of crystallization of some drug substances in the gastric fluids; and poor understanding of the in vitro /in vivo correlation in these dosage forms [7,15].
The purpose of this dissertation is to address the physical instability of the amorphous drug substance dispersed in the polymeric matrix. Specifically, to understand the crystallization behavior of the amorphous component from solid dispersions during their storage, and to evaluate the factors that could improve the physical stability of these dosage forms.
In this chapter an extensive literature review on the field of solid dispersion is provided, more specifically relating to physical stability issues of amorphous components. The purpose of this chapter is to increase the appetite of the reader for this field and to get familiarized with some of the latest developments.
Physical state of drug dispersion: amorphous solid solutions Solid dispersions have been traditionally formulated by heating the physical mixture of drug and carrier to the molten state and solidifying by cooling it either rapidly or slowly.
Alternate methods involved dissolving the physical mixtures in a common volatile solvent and evaporating the solvent to obtain the residual dispersion. The solvent can also be spray-dried (organic volatile solvent) or freeze dried (aqueous) to obtain the solid dispersion. Recently supercritical fluid processing and hot melt extrusion are being explored extensively for a large-scale manufacturing of solid dispersion by pharmaceutical companies [16,17] . Although there are several approaches to formulate solid dispersions, the effect of processing conditions on the physicochemical properties of the overall formulation and the exact physical state of the drug that is dispersed through the polymer is yet to be ascertained. The nature of the drug dispersion i.e. molecular vs.
phase segregation, amorphous vs. crystalline and the presence of polymer can influence the overall free energy and the physical stability of the product.
In 1961, Sekiguchi and Obi performed the initial work on pharmaceutical solid dispersion [ 18]. They classified the physical state of active components in the inert matrix into several groups as depicted in table A-1. The group under which a solid dispersion product would belong is entirely dependent upon the intrinsic property of the drug and carrier.
Historically solid dispersions have been preferred in their eutectic fom1 in which the drug 4 and carrier are dispersed as fine crystalline particles in close proximity to each other. At the eutectic composition, as indicated by the depression in melting temperature of the fused material, the active component and the carrier have the maximum possible interface with each other. At this composition the active component experiences a maximum increase in the surface area and hence the dissolution rate is enhanced. However with storage time, crystal growth of active component has been reported leading to reduction in surface area and hence dissolution rate [19] . The crystal growth could possibly be inhibited if the carrier is dispersed on a molecular level with the active component. The molecular dispersion is achieved through formation of solid solution. Since a crystalline solid solution is very rare and difficult to formulate, amorphous solid solutions are being investigated as a means for drug crystal inhibition. In amorphous solid solution, the drug and carrier is dispersed on a molecular level leading to an amorphous product that exhibits a single glass transition temperature (T g) . The glass transition temperature is the characteristic feature of the miscible mixture and is generally considered as an indicator of the degree of molecular mobility. In the following sections some fundamental information relating to amorphous forms that is needed to address physical stability aspects is provided.
Glass transition temperature (T J In order to elaborate the glass transition phenomenon, let us examine closely into the thermodynamics of molecular motion of a crystalline material that is liquefied and is maintained above its melting temperature (Tm). Fig A-I describes the changes in specific volume of the liquid as it is cooled from high temperatures. Upon cooling, the molecules of the liquid may come in close proximity and be bound, resulting in crystalline material at T rn· Melting or crystallization is a first order phase transition, where there is a change in the first derivative of Gibbs free energy with respect to temperature, and the process of crystallization results in a drastic reduction in the specific volume (water is an exception).
If the liquid manages not to crystallize at Tm, then its specific volume continues to decrease with corresponding increase in the viscosity. The system is now termed as a 'supercooled liquid' and is considered to be in 'equi librium' with properties similar to linear extrapolation of enthalpy of a liquid. As a supercooled liquid is cooled to lower temperatures, the molecules move more and more slowly and at some temperature they do not have a chance to significantly rearrange or to mobilize themselves. At this point the system is thrown out of equilibrium and this temperature is termed as the glass transition temperature (T g). At temperatures lower than the T g, the time scales of molecular motion become drastically long and the material is essentially considered in a frozen state and is termed as ' amorphous' or 'glass'. Glass transition temperature is a second order transition in which there is no change in the free energy of the system, but changes in enthalpy, entropy or viscosity of the material do occur.
Solid solutions or miscible dispersions are characterized based on their single glass transition temperature since both polymer and drug undergo the transformation from liquid state to corresponding unique amorphous phase simultaneously upon cooling. The presence of two T gS in the mixtures represents a phase separated product. Interested 6 readers on the thermodynamics of amorphous materials are referred to an excellent review and references therein by Ediger and co-authors [20] .

Molecular mobility in amorphous phase
Molecular mobility in amorphous phase is typically described in terms of relaxation time (<) . Relaxation time is the time taken for one molecular motion i.e. rotational or translational to occur. In molecular liquids near T g, it may take minutes or hours for a molecule less than 10° A in diameter to reorient [21]. Although the molecules in amorphous state at temperatures below the T g are in a fro zen state, they do possess mobility in the order of hundred of hours. The temperature dependence ofrelaxation time for supercooled liquids is often described at least approximately by the Vogel-Tammann-Fulcher (VTF) equation: where <o can be considered as a pre-exponential factor, if the equation is considered as Arrhenius equation (i .e. when T., = 0). B is then equal to Elk where E is the energy of activation for molecular motion and k is the Boltzmann constant. When T., > 0, the temperature dependence of relaxation time is non-Arrhenius and the relaxation time is predicted to become infinite at T .,. In other words, the molecular mobility is very low at temperatures well below the T g· The primary cause of such slow dynamics is due to high viscosity and low kinetic energy at such low temperatures. However the slow dynamics of amorphous state could in some cases be significant enough for the molecules to orient and transform to the low free energy crystalline state [22].

Factors influencing physical stability of solid solutions
In amorphous solid solutions, the presence of polymer having a higher glass transition temperature has been suggested as an approach to inhibit drug crystallization [23].
Typically a storage temperature of 50°C below the T g of amorphous system is considered suitable to minimize the crystallization tendency of the material [24]. When the drug substance is mixed molecularly with the polymer, the Tg of the molecular dispersion is increased relative to the Tg of drug alone and hence the molecular mobility of the drug that is needed for crystallization is reduced . It has also been shown, in specific cases, that presence of drug-polymer interactions e.g. hydrogen bonding, could play a role in delaying or preventing the drug crystallization from molecular dispersions [25]. However mechanism of prevention of drug crystallization from such dispersions is not yet fully understood in the presence of moisture and elevated temperature conditions. Moisture is a well known plasticizer and is naturally present in the atmosphere. It has a tendency to be absorbed by the amorphous materials and reduces the glass transition temperature, thereby increasing the molecular mobility and the crystallization tendency of drug components [26][27][28]. The exact extent of increase in the molecular mobility by moisture and crystallization rate is not yet clear. Also the effect of moisture on the extent of drugpolymer interaction is poorly understood .
The storage temperature is also believed to play a role in crystallization. It is generally known that increasing the storage temperature leads to enhanced crystallization kinetics.
Temperature could also influence the extent of drug-polymer interactions, by decreasing the strength of interaction with its increase. Therefore all the issues suggest that molecular dispersions are kinetically stable and their physical stability is largely influenced by storage conditions i.e. temperature and relative humidity, choice of polymer i.e. less hygroscopic polymer is better, absence of nuclei or seed crystals to prevent secondary nucleation of drug, nature of drug product packaging i.e. presence of desiccants, etc. All of these restrictions make it nearly impossible to formulate a successful physically stable molecular dispersion.
Determination of solid solubility to improve physical stability In this thesis, I address the issue of physical stability of solid dispersions by resolving the problem into two categories: kinetic stability and thermodynamic stability. A list of factors that differentiate the two have been tabulated (table A-II). When a drug is formulated as a miscible solid dispersion with a carrier, although it is miscible initially, i.e. under low moisture content levels, its extent of miscibility may not be retained in the presence of moisture that can be gained by samples during shelf-life storage. Under such conditions, the drug or polyn1er could phase separate ffom the mixtures thereby making the drug vulnerable to crystallization.
If a solid dispersion system can be designed where the drug concentration is such that the presence of moisture cannot cause phase separation and hence drug crystallization, physical stability can be assured. The concentration of the drug that remains miscible with the polymer in the presence of accelerated stress conditions i.e. high relative humidity and temperature is defined as the "solid solubility" of drug in the polymer. To characterize such solid solubility is indeed challenging, as stated by Duncan Q. M. Craig in one of his recent reviews on solid dispersions [4], " ... unequivocal demonstration of solid solubility is not as simple as one may imagine. " The concept of solid solution has been addressed for several years in the metallurgy and polymer science fields [29]. However, techniques to characterize solid solubility are very limited and restricted to specific fields. Calorimetric techniques are utilized in characterizing drug-polymer miscibility (30], and can be of valuable support to quantify drug solid solubility as performed in the present research.

1.
To design and evaluate a screening technique to identify the "solid solubility" of drug substances in commonly employed polymeric carriers using calorimetric approaches. Also to verify if solid dispersions prepared at solid solubility levels are physically stable by storing them under accelerated stability conditions for at least six months. 10 2.
To investigate the influence of factors like storage temperature, moisture content drug-polymer interactions, drug loading etc. on the solid solubility and drug crystallization from solid dispersions.

3.
To validate the universal applicability of the screening technique by employing different hydrophobic drug models and using various analytical tools like X-ray diffraction and Fourier transform infrared spectroscopy.

4.
To understand the mechanism of drug crystallization from solid miscible dispersions and performing extensive literature review to substantiate the findings theoretically.

5.
To characterize the physical stability of model amorphous hydrophobic compounds by conducting crystallization studies to investigate the influence of moisture and storage temperature on the crystallization kinetics. Eventually compare the drug crystallization rate with the kinetics of drug phase separation and crystallization from solid dispersions.

6.
If circumstances allow, to design fonnulation approaches like addition of surfactant or complexing agents to decrease the kinetics of drug phase separation or crystallization and to enhance the drug solid solubility in the carrier.

7.
Finally, to employ a newer sensitive thermally stimulated current spectroscopy (TSC) measurement to characterize the molecular mobility of commonly employed polymer in solid dispersion. To evaluate the potential of TSC in characterizing amorphous formulations and comparing the results with experimental results obtained using modulated differential scanning calorimeter (MDSC).
28. S. L. Shambline, and G. Zografi. The effects of absorbed water on the properties of amorphous mixtures containing sucrose. Pharmaceutical Research. 16: 1119-1124 (1999 [1]. However, despite the extensive R&D efforts only a few solid dispersion products are commercially available. One of the reasons for this rarity is the thermodynamic instability of amorphous active component in dispersed systems [2,3).
Although recrystallization of amorphous drugs from solid dispersions is reported in several publications, the exact mechanism of such recrystallization is not clear. The investigators, who extensively evaluated the physical stability of amorphous materials, attributed the recrystallization phenomenon mainly to the molecular motions of amorphous components even below the glass transition temperature (T g) [4][5][6][7][8].
Lt is generall y understood that polymers improve the physical stability of aniorphous drugs in solid dispersions by either: a) increasing the T g of the miscible mixture, thereby reducing the molecular mobility at regular storage temperatures, or b) specifically interacting (e.g. hydrogen bonding) with functional groups of amorphous drug substance [9][10][11]. For a polymer to be effective in preventing crystallization by either of these mechanisms, it has to be molecularly miscible with the amorphous substance. For complete miscibility, interactions between the two components are necessary [ 12]. A miscible system can phase-separate and become unstable if specific interactions between the components are adversely affected by a third component like water. The extent of moisture uptake depends upon factors like hygroscopicity of the drug and/or carrier, the 2 1 concentration of the active substance in the drug product and the storage temperature.
Due to this complexity, the prediction of physical stability of a drug substance in a solid dispersion becomes challenging. In order that the solid dispersion systems can be effectively utilized by the pharmaceutical industry, it is critical to understand the role of time, temperature, and humidity in determination of the physical stability of solid dispersion during its shelf-life.
The objective of this study was two-fold: The first was to study the phase behavior of the selected solid dispersions via thermal analysis, in order to determine the extent of solid state miscibility under accelerated stability conditions, and the second was to understand the mechanism of physical stability of these amorphous molecular dispersions.
Molecular dispersions of trehalose-dextran and trehalose-poly (vinylpyrrolidone) (PVP) were chosen as models to understand the phase behavior of a crystallizable amorphous substance that was molecularly mixed with a polymer. Trehalose is a well characterized compound [13]. Amorphous trehalose, which is highly susceptible to moisture, crystallizes to a stable di hydrate that can be characterized thermally [ 13]. Although hydrogen-bond interactions between trehalose and PVP have been cited [14] , the influence of water on these interactions, the phase behavior of the mixtures and the solidstate miscibility has not yet been demonstrated in the literature. 22

Preparation of Amorphous Mixtures by Lyophilization:
Dextran and PVP-K29/32 were dried in a vacuum oven overnight at I 05°C until a constant weight was obtained. Trehalose dihydrate was used without any further treatment. Physical mixtures of each polymer and trehalose dihydrate were made in dry glove bags (RH < I5%) to obtain weight proportions oftrehalose dihydrate in the polymer from 0% to 100% at 10% w/w intervals. These mixtures were dissolved in purified water to obtain I 0% w/v concentration and stirred overnight at 35°C. Aqueous solutions (2 mL) were freeze-dried using a commercial freeze dryer (Dura-Stop µP, FTS System, Stone Ridge, NY). Vials containing solutions were transferred onto pre-cooled shelves (-45°C).
Primary drying was carried out at shelf temperature of -27°C for 22-24 hrs, followed by drying at -l 2°C for 2 hrs, 0°C for 16 hrs, and secondary drying at 30°C for 20 hrs. 23 Vacuum pressure of 15 mTorr was applied throughout the drying. All samples were stored over anhydrous calcium sulfate at room temperature, after capping the vials. X-ray analysis of these samples showed that all the mixtures were amorphous .

Isoth ermal Stability Studies:
Lyophilized amorphous mixtures of dextran containing I 0, 20, 30, 40 and 70% by weight of trehalose were accurately weighed (8-12 mg) in standard DSC pans in a glove bag (RH< l5%). These pans were placed in desiccators containing saturated sodium chloride solution (75% RH) to expose the miscible mixtures to excess humidity. The desiccators were kept at room temperature (23 ± I 0 C) and incubators (Precision Scientific Inc., Chicago, IL) at 40 ± I °C and 50 ± I °C. Saturated NaCl solution provides relative humidity of 75 ± 0.5% over the temperature range of 23° to 50°C [ 15] . By storing the samples in discrete DSC pans, the reproducibility of all measurements was ascertained without any interference of sample handling. Samples were taken out for analysis at predetermined time intervals for a period of up to six months. They were crimped with aluminum lids having I 0 pinholes to facilitate removal of absorbed and bound water liberated upon melting of any recrystallized trehalose.

Thermal Analysis:
Modulated Differential Scanning Calorimeter (MDSC) (TA Instruments 2920, New Castle, DE), with a liquid nitrogen cooling accessory was used for thermal analysis. The analysis was performed under a purge of dry nitrogen gas (60cc/min). High purity indium 24 and sapphire were used frequently to calibrate for the heat flow and heat capacity of the instrument. Samples were initially cooled to -30°C for 10 minutes and were heated to 245°C at 1 °C/min with modulations of 0.266° every 50 seconds. MDSC offers an advantage over conventional DSC because it can detect reversible glass transition phenomenon in the presence of several other non-reversible relaxation processes. Heat capacity changes can be resolved from moisture removal endotherm, thereby improving the signal to noise ratio and improving the elucidation of the T g·

Characterization of lrehalose:
DSC scans of trehalose before and after lyophilization are shown in Fig. B-1-la. A heating rate of 1°C/min was employed to minimize polymorphic conversion of trehalose dihydrate during heating [ 16]. Upon heating trehalose dihydrate, an endotherm at 96° ± 1°C (Th) was observed. This was followed by a Tg at 122°C and recrystallization at 171°C. The recrystallized anhydrous trehalose melted at 2 l 2°C. Thermo-gravimetric analysis (TGA) performed on trehalose dihydrate showed a 9.5% weight loss at 96°C confirming that the endothenn at 96° ± 1°C was due to the removal of water of hydration.
Hot-stage microscopy was conducted on trehalose dihydrate by heating at 1°C/min and it was observed that water removal from trehalose resulted in a structural collapse and loss of birefringence (indicating amorphous state). At approximately ! 35°C, trehalose 25 liquefied and recrystallized to the anhydrate form. These results agreed with DSC characterization and with observations of Taylor and York (13]. Lyophilized trehalose was amorphous as determined by X-ray diffraction. DSC scan of lyophilized trehalose (Fig. B-1-1 a) showed no endotherm at 96°C, indicating the absence of crystalline hydrate structure. For verification, amorphous trehalose that was exposed to 50°C and 40°C with 75%RH were heated in DSC and examined. One would expect the amorphous trehalose to crystallize upon storage and indeed as seen in Fig Trehalose-dextran and Trehalose-PVP molecular mixtures: physical stability studies MDSC reversing heat flow scans of mixtures containing 40% w/w trehalose-in-dextran, which were stored at 50°C/ 75% RH for a period of 2 -50 days are shown in Fig. B-l-3a.
A freshly prepared sample showed a single Tg at 169°C, indicating miscibility. 26 Amorphous miscible mixtures when exposed to 50°CI 75% RH absorbed moisture, and the plastici zed Tg of such a system, from the total heat flow scan (data not shown), was detected at 22° ± 3°C. The absorbed moisture was removed from the mixtures as the samples were heated in MDSC. As opposed to a single Tg of 169°C in the original mixture, two distinct Tgs were observed in the 2 and 4-day old samples; Tg values at 123° and l21°C (referred to as Tg1) , and Tg at 203°C (Tg2) . In the 16 -34 day samples, Tg1 was not observed. However a water-loss endotherm at 96° ±l °C {Th), characteristic of crystalline trehalose, had appeared. In addition, T g 2 values had increased and seemed to plateau, which is termed as equi librium Tg (Tg•q). There was also a distinct reduction in the heat capacity step change, (fl Cp) at T gl· These values decreased initially and seemed to reach a plateau as seen from data in Table B-1-1. The thermal events in dextran mi xtures containing 30% w/w trehalose were also similar to those that were observed for 40% w/w mixtures when stored at 50°C/ 75%RH (Fig. B-I-3b).
MDSC reversing heat flow scans of mixtures containing 20% and 40% w/w trehalose-in-PVP which were stored at 23°C/ 69% RH are shown in Fig. B-l-3c. With this polymer, the phase separation of amorphous trehalose was observed (T gl at I 20°C) in the 190,195 and 200 day old samples. A water-loss endotherm at 96° ± l °C (T,,) was also observed.
For 20% w/w samples, the T gl and T,, were not noticeable even after 200 days of storage.
A reduction in llCp values at Tg 2 (given within brackets in Fig. B-l-3c) fol lowed by a plateau was noticed in these systems as well . The heat flow scans that are depicted in Fig. B-l-3d are to demonstrate the presence ofTg1 & Tg2 values for some selected mixtures. 27 A plot of T g2 values of dextran mixtures containing different weight fraction of trehalose, stored at 50°, 40° and 23°C with 75% RH is shown in Fig. B-1 -4. It should be noted from the figure that T g'q is always Jess than 225°C, which is the Tg of pure amorphous dextran (dotted line). This reduction in Tg indicated that a fraction oftrehalose remained miscible with the polymer even after being exposed to excessive temperature and humidity conditions for up to six months. This fraction was identified as the "solid solubility'' of trehalose in dextran. The T g •q of all amorphous mixtures ranged from 207° -210°C, when they were stored at 40° and 50°C with 75% RH, and within 196° -203°C for mixtures stored at 23°C/75% RH conditions. These findin gs suggested that, at equilibrium, irrespective of the initial trehalose-concentrations, the solid solubility limit oftrehalose in dextran was the same for a specified storage condition. For instance, when stored at 50°C/ 75% RH , 70% w/w trehalose mixture which was initially miscible with dextran, formed a dihydrate and demonstrated Tg 2 values at 207°-210°C upon "equilibrium". This concentration was accepted as its solid solubility limit.
To exclude the possible interference of moisture on reduction of T gZ. amorphous dextran that was equi librated at 50°C I 75%RH for a month and contained 11 % w/w water was heated in MDSC. The Tg of the sample was 225°C, which was identical to the Tg of the original sample with 3% w/w water content. These findings proved that moisture was not the cause ofT g 2 reduction.
On the other hand, trehalose that was liquefied above its Tg of l 18°C could plasticize dextran and lower the T g below 225°C. Such a possibility was suggested by Six et al. [17] for HPM C-itraconazole dispersions. The MDSC scan of 1: 1 phys ical mixture of 28 amorphous trehalose and dextran, demonstrated Tgs at 119°± 0.5°C and 223° ±0.9°C (n; 3) upon heating at 1°C/min rate. The two Tg values coincided with the Tgs of pure trehalose and dextran respectively indicating that heating in MOSC did not cause any mixing. The reduction in T g2 values from 225°C to T g cq' is indicating the presence of the miscible fraction oftrehalose in the polymer.

Crystallization of trchalose:
A plot of percent crystallinity of trehalose from the miscible mixtures containing 40% and 30% w/w trehalose as a function of storage time is shown in Fig. B-1-5. The amount of trehalose crystallized as the dihydrate was calculated from the heat of fusion under the water-loss-endotherm (Th) of the total heat flow scan. Amorphous trehalose by itself when exposed to 75% RH had enthalpy of fusion equal to untreated crystalline trehalose dihydrate indicating 100% crystall ization (Fig. B-1-1 b ). On the other hand, sol id dispersion samples showed a maximum of 84% crystallization for samples stored at 50°C/75% RH for six months. It was also seen that the rate of crystallization, after an initial rapid increase, was much less in 30%w/w dispersions when compared to 40%w/w dispersions. These findings are consistent with results which suggest that a fraction of trehalose remains miscible with the polymer.
Quantifying the limit of solid solubility: To determine the molecularly-dispersed trehalose concentration in the polymer, the plot of Tg of the mixtures against the% w/w trehalose concentration was utilized (Fig. B-1 -29 2c,d). In this graph, the T g'q values were used to determine the fraction of trehalose that remained miscible with dextran (Fig. B-l-2c) and PYP . In these graphs, the solubility limits are shown by arrows. The corresponding value on the x-axis represented the solid solubility of trehalose in the polymer. Table B-I-II demonstrates the solid solubility limits of trehalose in dextran at different storage conditions. In this table the difference between the storage temperature and the plasticized Tg due to presence of moisture (Tgw) is shown as 6 Tg. It is important to note that the samples were stored at temperatures above their plasticized Tg. At this region, the system has a high degree of molecular mobility and consequently, a higher rate of destabilization. The solid solubility of trehalose-in-dextran decreased from 18% w/w at 23°C to 12% w/w at 50°C, whereas the total water content decreased from 16.5% w/w to 11% w/w. However, no appreciable difference in solid solubility was observed between 40°C and 50°C storage conditions, although total water content decreased from 13% to 11 %.

Mechanism of Phase Separation & Solid Solubility :
The hypothesized mechanism for phase separation and solid solubility is depicted schematically in Fig. B-I-6 for a dextran-trehalose model. The repeat units of dextran and trehalose are labeled as ID, 20, and IT, 2T, etc respectively.
In the absence of moisture, freshly prepared mixtures have high Tg values and hence the molecular mobility is very low at the storage temperatures . Trehalose, therefore, remains 30 in a kinetically frozen state of miscibility. Upon exposure to moisture, the mixtures are plasticized, and the molecular mobility increases. The exact role of water is not clear yet.
Water may either weaken the H-bond interaction (between -OH oftrehalose and -OH or -0-groups in dextran, or C=O groups in PVP) by bridging with polymer and trehalose structural units, or may merely increase the molecular mobility by plasticizing the mixtures. ln either case, diffusion of trehalose through the polymeric matrix could result in separation of trehalose into an amorphous phase, and subsequent crystall ization. As more and more trehalose phase separates, increasing amounts of "free" polymer units are left to interact with the remaining trehalose. Such units possibly orient around and arrest the non-diffused trehalose molecules by satisfying their H-bond requirements ( 1 T,. At equilibrium, localized pockets of trehalose molecules are almost entirely bonded to the polymer, reaching solid solubility as characterized by T g?.

Thermodynamic Interpretation of Solid Solubility:
To analyze the solid-solubility of trehalose in the polymers thermodynamically, the energetics of interactions between trehalose and polymers were calculated using the Couchman & Karasz theoretical equations [18]. Similar approach was used by Shamblin et al. to determine the excess thermodynamic properties of mixing in the binary mixtures [ 19]. Using the equations described in detail in the appendix , the enthalpy ( !1H mu ) and free energy ( t.G,,," ) of mixing as a function of trehalose weight fraction were calculated and plotted in Fig. B-1 -7. Since the influence of water on the free energy of mixing is not understood, water was not taken into considerat ion in the thermodynamic analysis. The enthalpy of mixing (!!Jim,. ) shown in Fig. B-I-7 was positive for the entire composition range and was similar in magnitude to those reported for mixtures of sucrose-PVP or sucrose-dextran [ 19). The positive !!Ji mi< at all compositions indicates that greater numbers of H-bonds are broken between trehalose-trehalose and polymer-polymer than are fonned between trehalose-polymer during mixing. Since lesser number of bonds was formed between trehalose and polymer when compared to the bonds that they had in their individual states, the positive !!Ji"'" would favor trehalose and polymer to self-associate or phase separate. The excess entropy of mixtures was positive (not shown) throughout the trehalose composition favoring mixing. The negative free energy values for low trehalose-containing mixtures are thus driven by the excess entropy. Although the dextran-trehalose mixtures studied contained moisture, and a three-phase system could be significantly different from the two-phase system, the analysis cond ucted supports the hypothesis that a component dispersed in a solid dispersion with certain solubility can be thennodynamically favored.

Mechanism of cry stal growth :
Upon phase-separation, amorphous trehalose, as characterized by T gt crystallizes in the presence of water, and is characterized by T 11 (Fig. B-I-3). The crystallization is dependent upon the initial concentration of trehalose in the mixture and the storage conditions employed (Fig. B-1 -5).
The portion of polymer located at the amorphous-crystalline interface has low molecular mobility and is termed as "rigid amorphous fraction" as suggested by Lee and Kim (20].
This fraction does not contribute significantly to the heat capacity step change values at T g 2 (6C 0 ) [20][21][22]. The fraction of amorphous bulk distant from crystalline region is the "mobile amorphous phase" and could be responsible for the 6C 0 at T g2 as indicated by Craig et al. (21 ). As the phase separated trehalose crystallizes, the fraction of polymer in close contact with the crystalline surface increases. This explains for the reduction in the 6C 0 values at T g2 (Table B- , until no more "rigid amorphous fraction" is formed. Formation of rigid amorphous fraction may also explain the decrease in the rate and extent of crystallization in 30% and 40% w/w mixtures with time (Fig. B-1 -5). As suggested by Crowley and Zografi [23] , the rigid amorphous fraction impedes the kinetics of crystallization of trehalose that has already phase separated. This kinetic phenomenon is different from the solid solubility, which is characterized by T g2, as described in the previous section.
Effect of temperature and moisture on solid solubility: Table B-1-11 demonstrates the effect of storage temperature and moisture content on the solid solubility of trehalose. Reduction of solid solubility limit wi th an increase in the temperature is probably due to the higher molecul ar mobili ty of the system at elevated 33 temperatures, leading to an enhanced phase separation and crystallization. In addition, at 50°C, thermal expansion of the mixtures may reduce the degree of interaction between the sugar and polymer thereby decreasing the miscibility limit as suggested by Tang et al. (24]. Finally, since the miscible mixtures have a higher enthalpy and free energy, it is not surprising to see that at higher temperatures, the system proceeds for de-mixing causing trehalose to crystallize. Although with an increase in temperature from 23°C to 40°C solubility reduction was seen, no difference in solubility was noticed between 40°C and 50°C. This is possibly due to an associated reduction in water content, where the reduced water level compensates the temperature effects. Therefore, water seems to have an indirect effect on T g2 reduction and hence solid solubility. Overall, the data obtained suggest that both temperature and moisture affect the phase behavior and extent of solid solubility.

CONCLUSIONS
ln this paper, we have shown application of MDSC in monitoring the phase separation and crystallization behavior of trehalose from its amorphous miscible mixtures, at ambient temperature and accelerated stability conditions. Upon exposure to moisture, trehalose separated as an intermediate amorphous phase which later crystallized to the dihydrate form. Based on the experimental data and thermodynamic as well as mechanistic assessment, it is postulated that a fraction of trehalose remained miscible with the polymers due to extensive hydrogen bonding between the two components. This 34 fraction was analyzed by the changes in the glass transition temperatures and was referred to as the "solid solubility" of trehalose in the polymer. Molecular mixtures containing I 0% w/w trehalose (i.e., trehalose concentration below solid solubility) did not phase separate or crystallize during the six months storage at accelerated stability conditions, which substantiated our findings. We also propose that the MDSC technique described can be applied to determine the phase behavior and solid solubility limits of the hydrophobic drugs that are dispersed in water-soluble carriers. This method can be useful in identifying the solid dispersions that will remain physically stable during the she! f-life.   ' Determined by thermogravimetric analysis by heating the sample at 5°C/min. § 6 Tg is the difference between storage temperature and the plasticized Tg due to presence of moisture (T g w), when the samples were exposed to stability conditions. ' Mean value of Tg2 (i.e. , Tg of polymer-trehalose mixture) after T g2 reaches a steady state. 1 Calculation of solid solubility limit is described in Fig. B-l-2c,d and the text.  ..      Results: ln the freshly prepared solid dispersions, both griseofulvin and indoprofen were molecularly miscible with PVP up to 30% w/w drug concentration. FTlR demonstrated the presence of hydrogen bonding in indoprofen-PVP dispersions, but not in griseofulvin-PVP dispersions. When exposed to 40°C/69%RH for 15 days, griseofulvin-PVP mixtures resulted in complete phase separation and crystallization which was monitored by the changes in their T g· A complete phase separation indicated absence of solid solubility.

ACKNOWLEDGMENTS
XRD detected crystallinity in 10%w/w griseofulvin dispersions, supporting the MDSC analysis. On the other hand, the solid solubi lity of indoprofen in PVP was determined as 13% w/w and no crystallinity could be detected around this concentration using XRD.
The rate of phase separation of drug from the polymer was estimated by fitting the fraction of drug phase separated from the polymer as determined by analyzing the Tg of drug-polymer miscible phase to the use of first order kinetics. The rate constants for 10%, 20% and 30% w/w griseofulvin-in-PVP were 4.66, 5.19, and 12.50 (xl0 2 ) [day" 1 ] respectively. lndoprofen-PVP dispersions had rate constants of 0.62 , and 1.25 (x 10 2 )

INTRODUCTION
Currently the development of a successful solid dispersion product is dependent upon the ability to stabilize the metastable amorphous form of the drug substance. Because amorphous form offers high solubility, it is important of retain this form of the drug in the solid dispersion [1][2][3]. Several studies have however demonstrated the unpredictable crystallization of amorphous drugs by itself and from solid dispersions, which is limiting its commercial potential [4][5][6][7]. Water soluble polymers have been proposed as a means to inhibit or delay the drug crystallization by forming a miscible phase with the drug. In the miscible phase, the drug and polymer are molecularly dispersed and crystallization inhibition is possible via drug-polymer hydrogen bonding, and/or the anti-plasticizing effect (i.e. increase in Tg of solid dispersions) by the polymer [8,9]. A detailed evaluation of the factors influencing the physical stability of amorphous solid dispersions reveals the presence of two opposing forces. On one hand, polymer helps in retaining the miscibility and prevents the drug crystallization. On the other hand, the uptake of moisture and/or exposure to elevated temperatures increases the molecular mobility, thereby increasing the potential for phase separation (de-mixing) and crystallization. The physical integrity of the miscible solid dispersion would therefore depend upon the relative magnitude of the two opposing forces.
In an earlier study, we demonstrated that depending upon the nature of drug-polymer interactions, a fraction of the drug can remain dissolved in the polymer in the solid-state even under high heat and humidity conditions [10]. This fraction is termed as the "solid solubility" and one of the factors influencing the solid solubility is the affinity between the two components of the dispersion .
In view of the earlier findings, the purpose of this paper is to provide further evidence to substantiate the importance of hydrogen bonding in the solid solubility and physical stability of solid dispersions . The solid solubility of two different hydrophobic drugs, namely griseofulvin and indoprofen in poly (vinyl pyrrolidone) (PVP), were determined.
Griseofulvin and indoprofen are highly crystalline in nature as indicated by their high melting temperatures and heats of fusion and therefore serve as good model compounds.
In addition, based on their chemical structures, griseofulvin cannot hydrogen bond with PVP whereas indoprofen can participate in hydrogen bonding, thus enabling assessing the impact of drug-polymer affinity on the solid solubility.
The second objective of this study was to detennine experimentally the kinetics of phase separation of the drug from the polymer. One may intuitively expect that for a drug to crystallize from a miscible solid dispersion, it would have to phase separate first from the polymer matrix. Therefore detem1ining the rate of phase separation of drug would serve to provide valuable information on the overall rate of physical instability since 53 crystallization of drug substance is often complex to predict. In some cases the solid so lubility of the drug in the so lid dispersion may need to be very low in order to fonnulate a physically stable miscible solid dispersion that would meet the high therapeutic dose requirement. By understanding the drug phase separation and crystallization kinetics one can possibly formulate miscible solid dispersions that are above the drug's solid solubility and retain their physical stability by either defining the shelf life or by choosing appropriate storage conditions. In this study we have employed a theoretical rate equation that is typically used in determining crystallization kinetics, to analyze the phase separation rate of griseofulvin and indoprofen from their solid dispersions.

Methods:
Preparation of solid dispersions using solvent evaporation technique: Solid dispersions of drug (I 0-50% w/w) in PVP were prepared using solvent evaporation technique. Crystalline griseofulvin and crystalline indoprofen were dissolved (0. 3 -l .5g) in dichloromethane and methanol respectively by stirring the solution in a waterbath at 37°C for about 15 minutes. PVP was added to the drug solution to make the total weight of solids as 3g. After all the PVP was dissolved, as determined visually, the solvent was evaporated using a rotary vacuum evaporator (Rotavap®) by heating the solution at 37°C in a water bath. The residual solid dispersions thus obtained were further dried in a vacuum oven at room temperature for 24 hours to remove the remaining residual solvent.
They were ground with a mortar and pestle and were sifted through sieve # 70 to result in particle size $ 21 Oµm. The solid dispersions were stored in vials over anhydrous CaS04 in a freezer when not being studied in order to minimize the molecular mobility of the system.

Determination of solid solubility
To determine the solid solubility of drugs in the polymer, solid dispersions containing I 0, 20, and 30% w/w of griseofulvin and indoprofen in PVP were accurately weighed (8l 2mg) in standard DSC pans. The pans were placed in desiccators containing saturated copper chloride solution (69% RH) and an oven at 40 ± I °C (Precision Scientific Inc. ,

55
Chicago, IL). The 69% RH was chosen in order to prevent deliquescence of PVP at higher moisture level s and yet to provide an accelerated storage condition. By storing the samples in discrete DSC pans, the reproducibility could be assessed without di sturbing the bulk properties of the aging samples. Samples were taken out from the desiccators for analysis at regular intervals for up to 90 days. They were crimped with aluminium lids having five pinholes to facilitate the removal of absorbed water during heating. The resulting thermal scans were analyzed for their changes in the drug-polymer phase compositions by measuring the glass transition temperatures (T ,). An increase in the T g of the miscible phase implied phase separation of drug from the solid dispersion since lesser fraction of drug is present in the polymer phase to plasticize the system (provided the Tg of drug is less than Tg of polymer). The increase in Tg of the solid dispersion was monitored as a function of storage time until no further increase in the Tg was observed.
To quantify the solid solubility, the equilibrium Tg value was utilized to identify the percentage of drug that would provide the same Tg value in a freshly prepared solid dispersion.

Modulated differential scanning calorimeter (MDSC):
Thermal analysis was performed using MDSC 2920 (TA Instruments, New Castle, DE), with a liquid nitrogen cooling accessory. The analysis was performed under a purge of dry nitrogen gas (60cc/min). High purity indium and sapphi re were used frequently to calibrate the heat flow and heat capacity of the instrument. Thermal history of the samples was not erased. The events were recorded and anal yzed during the first heating 56 scan. Samples (-8-12 mg) were initially cooled to -10°C for 10 minutes and were heated to 245°C at I °C/min with modulations of 0.266° every 50 seconds. The specified amplitude and period were optimized to provide the best results for analysis. Each measurement was repeated three times at every time point to ascertain the reproducibility of the experiments.

X-ray Powder Diffraction (XRPD):
The X-ray powder diffractometer (Rigaku RINT (D/Max) 2200 that was equipped with an Ultimagoniometer) consisted of a 40 KV, 40mA generator with a Cu Ko radiation anode tube. XRD pattern was used to test the presence of crystallinity in the solid dispersions. The samples were sifted through 70-mesh and placed on a 0.5 mm quartz plate holder prior to exposure to X-rays. They were scanned over a 20 range between 2° and 40° at a scan rate of 2° per minute in 0.02° step size. The divergence and scattering slits were set at 1.00°, receiving slit at O. I 5mm and monochromator was used at 0.45mm.

Fourier Transform Infra-red Spectroscopy (FTIR):
Nicolet spectrometer equipped with a KBr beam splitter was used to obtain the infrared spectra. Calibration for wavenumber accuracy was perfonned by using polystyrene sample. Dry nitrogen gas was used to purge the beam splitter and the sample compartment. IR spectra were obtained using an attenuated total reflectance (A TR) accessory (single reflection bounce diamond crystal; Golden Gate accessory). For each spectrum, 32 scans were performed and a resolution of 4 cm was chosen. The ATR accessory enabled to obtain the spectra by pressing the solid material onto the diamond 57 crystal. This accessory eliminates the use of KBr and yet produces comparable results [ 12]. All samples were dried under vacuum for 20 hrs prior to obtaining any spectra, to remove the influence of residual moisture.

Characterization of drug substance
A comparison of the physical -chemical properties of griseofulvin and indoprofen is provided in Table B-Il-1. Both compounds were seen to have high melting temperatures and heats of fusion. The glass transition temperature was obtained by heating the amorphous drug substance in DSC at I °C/min. The amorphous drug was formed by rapidly cooling the molten crystalline drug. Upon heating both the amorphous drugs underwent recrystallization at temperatures above their Tg, indicating their propensity to crystallize.

Characterization of amorphous molecular dispersions
Griseofulvin-PVP and indoprofen-PYP solid dispersions showed a compositiondependent single Tg for up to 30% w/w drug concentrations (Fig. B-Il-2). This behavior indicated the presence of molecularly dispersed drug in the polymer at these concentrations. Above 30%w/w, multiple Tgs were observed in the solid dispersions of griseofulvin-PYP, indicating a phase-separated mixture. In the indoprofen-PVP dispersions, the Tg of 40%w/w dispersion was higher than the Tg of 30%w/w dispersion 58 thus indicating immiscibility in the sample. For both the drug solid dispersions no crystalline peaks were seen at drug levels of 30%w/w or less using x-ray diffraction.
Although x-ray pattern revealed crystallinity for 40%w/w solid di spersions, no significant melting endotherm was detected at this composition for the crystalline drug using MDSC . This could most likely be due to the solubilizing effect of polymer on the crystalline drug while heating in the MDSC.
In order to probe the interactions between the drug and PVP, the FTIR spectra of freshly prepared solid dispersions were compared with the drug -PVP physical mixtures. In Fig. B-II-3a, the carbonyl stretching of a 30%w/w griseofulvin-in-PVP solid dispersion is compared with the corresponding physical mixture of amorphous griseofulvin and PVP.
No significant differences were observed between the two spectra, indicating the absence of interactions between griseofulvin and PVP in the solid dispersions. Even with an increase in griseofulvin concentration from 10 to 30%w/w in solid dispersions, the peak carbonyl position of PVP did not shift significantly from I 667cm· 1 • Only a slight broadening of the peak and a shift in the wave numbers towards values of pure amorphous griseofulvin was seen most likely due to the presence of higher concentrations of griseofulvin.
For indoprofen-PVP solid dispersions, addition of I 0% and 30%w/w indoprofen caused a significant shift in the carbonyl stretching of PVP from 1667cm· 1 to 1654cm· 1 (Fig. B-II-3b), suggesting the presence of hydrogen bonding between the two components. The spectrum of a 30%w/w solid dispersion, which had the maximum miscibility as suggested by MDSC, was clearly different from the corresponding physical mixture. The 59 physical mixtures were prepared using crystalline indoprofen and PVP , because of the rapid tendency of indoprofen to crystallize during the preparation and characterization of physical mixtures. Pure amorphous indoprofen was however scanned immediately after its preparation as was confirmed amorphous even after the scan using optical microscopy.
To ensure that the crystalline nature of indoprofen in the physical mixture did not contribute to the peak shift, the spectra of amorphous and crystalline indoprofen were compared and no significant differences were observed in the peak region of interest (i.e. 1667cm· 1 to 1654cm" 1 ).

Griseofiilvin and PVP solid dispersions
Solid dispersions of griseofulvin and PVP that were exposed to accelerated stability conditions of 40°C/69%RH were analyzed periodically by monitoring their Tg, to determine any alteration in the miscible phase composition. MDSC reversing heat flow scans of 30% w/w griseofulvin and PVP solid dispersions, which were stored at 40°C/ 69% RH for a period of I -15 days are shown in Fig. B-ll-4a. While a freshly prepared solid dispersion showed a single Tg at 131°C, two Tgs referred to as Tg1 and Tg2, were observed in the stability samples during the 5 day-storage period. T gt corresponded to the T g of PVP-in-griseofulvin dispersions and T g2 corresponded to the T g of griseofulvin-in-PVP dispersions.
The formation of two T gs indicates the phase separation of griseofulvin and PVP. Upon storage, the heat capacity change at T gt decreased and that of T gZ increased in a time-60 dependent manner. The T go and T gz values were seen to increase with storage time, indicating a progressive phase separation, except for the 1 day-old sample, where a decrease in T gl was noticed . While T gl disappeared in the 9 day-old or older samples, T g2 values increased and seemed to reach a plateau at 166° ± l.5°C. This plateau value was tem1ed as equilibrium Tg (Tg'q).
The thermal events for 20%w/w griseofulvin and PVP solid dispersions followed the same pattern when they were exposed to 40°C/69%RH (Fig. B-ll-4b). However, unlike 30% w/w dispersions, the T gl decreased during the first 4 days and then disappeared subsequently. Such a decrease in T gl is possibly due to the increased plasticization of griseofulvin in the griseofulvin-rich fraction of solid dispersion. The Tg 1 disappears after reducing to a certain value suggesting a threshold level of griseofulvin in the griseofulvin-rich phase above which crystallization would occur. Indeed as expected the T gl of 20% and 30%w/w solid dispersions reach the same value before disappearing, indicating the threshold composition needed for crystallization. This possibility may also explain the crystallinity in a freshly prepared 40%w/w solid dispersion which is most likely above the threshold drug level.

Jndoprofen and PVP solid dispersions
MDSC scans of 30% w/w indoprofen-in-PVP solid dispersions, which were stored at 40°C/69% RH for a period of 1 -90 days are shown in Fig

Quantification of solid solubility
A plot of T g values of I 0%, 20% and 30% w/w solid dispersions of griseofulvin-in-PVP and indoprofen-in-PVP that were subjected to 40°C/69% RH for 3 months is shown in

X-ray diffraction analysis
In order to detennine ifthe phase separation and/or crystallization of drugs from the solid dispersions indeed resulted in crystallization of drug substance, X-ray diffraction was utilized. MDSC was unable to detect a melting endothenn of crystalline drug substance possibly due to the solvent effect of polymer at high temperatures as described earlier.
Hence X-ray diffraction was chosen to validate the experimental solid solubility values. In Fig. B-II-6b, the X-ray diffraction patterns of 10% and 20% w/w indoprofen and PVP solid dispersions that were exposed to 40°C/69% RH for 90 days are compared with the corresponding physical mixtures. No crystallinity in the 1 Oo/ow/w solid dispersions was noticed even at the end of 90 days storage period. However 20% w/w solid dispersion had small fraction of crystalline drug, as indicated by the smaller peak hei ght in the solid dispersion when compared to the corresponding physical mixtures. These findings 63 support the solid solubility of l 3%w/w of indoprofen-in-PVP that was determined using MDSC.

Hydrogen bonding and solid solubility
The data presented in this study suggests that the presence of hydrogen bonding facilitated phase miscibility between indoprofen and PYP even under accelerated stability conditions. Griseofulvin and indoprofen are easily crystallisable drugs as seen from their high melting temperatures, high heats of fusion and recrystallization. Further both drugs demonstrate relatively low glass transition temperatures in their amorphous states and recrystallized instantaneously when exposed to relative humidity greater than 57%.
Therefore to maintain the drugs in their amorphous form in the solid dispersions under stability conditions is indeed a challenging and an important task.
If the chemical structures of PYP, griseofulvin and indoprofen, illustrated in Fig. B -11-1 , are examined, it is seen that PVP can act only as a proton acceptor through either the 0 or N atoms of the pyrrolidone ring. However due to the steric constraints, the N atom may not participate actively in hydrogen bonding [ 13]. The chemical structure of griseofulvin reveals two carbonyl and three methoxy groups which can actively participate in hydrogen bonding as proton acceptors. On the other hand, indoprofen has four hydrogen bond acceptors namely the carbonyl group of -COOH, oxygen atom of hydroxyl group of -COOH, the carbonyl group of the pyrrolidone ring and the N atom of the pyrrolidone 64 ring. lndoprofen also has one proton donor which is the proton of the hydroxyl group of-COOH. Since hydrogen bonding occurs between a hydrogen bond acceptor and a donor, for two components to interact one of them should have at least one donor and the other at least one acceptor [14].
When griseofulvin and PVP are molecularly mixed, no hydrogen bonding between the two components is demonstrated through the FTIR spectra since neither component have hydrogen bond donor. On the other hand, a hydrogen bond interaction between the hydroxyl group of -COOH of indoprofen and the carbonyl group of the pyrrolidone ring in PVP is likely to occur. This is evidenced by a decrease in the carbonyl stretching of the carbonyl group of PVP from 1667cm·' to 1654cm·' in the 10% and 30%w/w solid dispersions.
Despite the absence of hydrogen bonding in griseofulvin and PVP solid dispersions, a 30% w/w initial miscibility is seen in these dispersions, similar to the indoprofen-PVP solid dispersions that demonstrate hydrogen bonding. Such miscibility in the absence of interactions is possibly favored through gain in the entropy of the system. Griseofulvin is dispersed in PVP in its amorphous phase, which is entropically favored when compared to its crystalline form. The solid dispersions thus formed remained miscible at regular storage temperature most likely due to the lack of molecular mobility, due to their high T g values. In the presence of moisture however, plasticization of solid dispersion could lead to an increase in the molecular mobility and thereby induce de-mixing and crystallization.
One of the factors offering resistance for physical instability could be the extent of drugpolymer affinity. Although indoprofen and PVP solid dispersions have moisture levels comparable to those of griseofulvin and PVP solid dispersions, the presence of hydrogen bonding between indoprofen and PVP could possibly explain for the enhanced solid solubility.

Kinetics of phase separation
Often times the therapeutic dose requirements of the drug substance is very high and the solid solubility of drug in the solid dispersions may not be high enough to formulate a physically stable solid dispersion. ln such a case, it is important to understand the rate of drug phase separation under the regular storage conditions. Such estimation should provide a general understanding as to how various storage conditions influence the phase separation process. Phase separation of drug from the miscible phase could be the very first step towards drug crystallization and hence monitoring this process is very essential to better define physical stability. It is evident from Fig Since griseofulvin and indoprofen have tendency to crystalli ze instantaneously when stored at 40°C/69%RH , a phase separation would immediately result into crystallization of these drug substances. Therefore, phase separation (i .e. fonnation of a heterogeneous phase), is considered to be the rate-limiting step for destabilization. Destabilization is defined as the physical state where a two-phase solid dispersion results. In order to estimate the rate constant for destabilization, the Kolmogorov-Johnson-Mehl-Avrami (KJMA) first order equation was used. KJMA equation is typically used to fit experimental crystallization patterns [ 15] and to obtain their rate constants and is described as: where (I-a) is the fraction phase separated, k is the rate constant for phase separation, and I is the storage time.
In equation (1), the fraction crystallized (I-a) was obtained using the follow ing expression: The fraction of drug phase separated was obtained by determining the average of triplicate measurements of T g of the solid dispersions at storage time t and using equation 2. The rate constant for phase separation was obtained using equation I and determining the slope of the linear equation. By obtaining the rates of phase separation, the physical 67 instability (i.e. crystallization) rate can be estimated, since the phase separation step is considered as the rate limiting step (/qe.-ysto mw;on) >> k (phoseseparat;on))-As inferred from the kinetic rate constants provided in Table B -11-11, the phase separation rate is higher for solid dispersions containing higher drug levels. This behavior agrees with the regular solution theory that higher the extent of super-saturation, the greater is the difference in the chemical potential in the supersaturated and saturated state and hence higher would be the driving force for crystallization. For a given drug level, the phase separation rate was found to be higher for griseofulvin-PVP solid dispersions when compared to indoprofen-PVP system. This behavior may be explained on the basis of the differences in the extent of interactions between individual drug and the polymer.
Hydrogen bonding therefore seems to influence the solid solubility and thereby affect the kinetic rate of phase separation in amorphous solid dispersions.

CONCLUSIONS
The solid solubility of two crystallizable hydrophobic drugs namely griseofulvin and indoprofen in PVP has been experimentally determined using modulated DSC.
Griseofulvin did not provide any solid solubility in PVP whereas a 13%w/w indoprofen    ..: ..: Surfactants may be added in such binary mixtures of drug and polymer to improve the drug-polymer interactions and to create a miscible phase. In the miscible phase, the drug is molecularly dispersed in the carrier and could have a lower tendency to crystallize.
Although in pharmaceutical studies surfactants are added in drug-polymer dispersions mainly to improve the solid solubility and dissolution of the poorly soluble drug (2-6) , the function and behaviour of the surfactant molecules in the physical stability of the system has never been investigated. The solid solubility of griseofulvin was increased from 3%w/w to 40%w/w in polyethylene glycol 6000 (PEG 6000) by adding 5%w/w of sod ium dodecyl sulphate (SOS) (7,8). SDS also improved the solid solubility of griseofulvin in PEG 3000 and PEG 20000 in concentrations from 3%w/w to 25%w/w.
Pluronic® F-68 increased the solid solubility and dissolution rate of a poorly water soluble drug, nifedipine, in its solid dispersions (9) . Tween 40 (0.5% w/w) and SDS in the presence and absence of sucrose stearate. Ternary solid dispersions containing the three ingredients were prepared and characterized by measuring the solid solubility of the drug in the polymer [12]. The effect of surfactants on the solid solubility of the 83 hydrophobic drug that were dispersed in the amorphous polymers has not yet been reported in the literature.

Materials:
Plasdone (Poly (vinylpyrrolidone) hydrophilic with an HLB value of 14 and was amorphous in nature. We measured its glass transition temperature (T g) as 34°C.

Preparation of Solid Dispersions -Solvent Evaporation Technique:
Solid dispersions of PVP-surfactant, surfactant-drug and PVP-drug-surfactant were prepared using solvent evaporation technique. Ternary solid dispersions were prepared by 84 dissolving crystalline griseofulvin in dichloromethane. A pre-calculated amount of PVP and surfactant was added to the drug solution. The surfactant concentrations of the batches were increased from 10%w/w to 40%w/w at 10% weight increments and the weight proportion of griseofulvin was altered from 10%w/w of griseofulvin-in-PVP to 60%w/w of griseofulvin-in-PVP. After all three ingredients thoroughly dissolved in dichoromethane, the solution was sonicated for 15 minutes in a water-bath that was maintained at 37°C, to complete clarity. The solvent was removed using a rotary evaporator (Rotavap®) at 37°C, leaving the residual drug product. Solid dispersion obtained in this manner was further dried in a vacuum oven at room temperature for at least 24 hours to remove the remaining residual solvent. The batches thus obtained were ground with a mortar and pestle and were sifted through a sieve# 70 to result in particle size $ 21 Oµm. The dispersions were stored in vials over anhydrous CaS04 in a freezer until use.

Modulated Diff erential Scanning Calorimeter (MDSC):
The glass transition temperature (T g) of the solid dispersions were analyzed using Modulated DSC 2920 (TA Instruments, New Castle, DE), with a liquid nitrogen cooling accessory. The analysis was performed under a purge of dry nitrogen gas (60cc/min).
High purity indium and sapphire were used frequently to calibrate for the heat flow and heat capacity of the instrument. Thennal history of the samples was not erased unless mentioned specifically. The events were recorded and analyzed during the first heating scan. Samples (-8-12 mg) were initially cooled to -10°C for 10 minutes and were heated 85 to 245°C at I °C/min with modulations of 0.266° every 50 seconds. The specified amplitude and period were optimized to provide the best results fo r analysis.

X-ray Powder Diffraction (XRPD):
The crystallinity of solid dispersions was tested using X-ray powder diffractometer (Rigaku RINT (D/Max) 2200 that was equipped with an Ultimagoniometer). The instrument consisted of a 40 KV, 40mA generator with a Cu Ko. radiation anode tube. The samples were sifted through sieve # 70 and placed on a 0.5 mm quartz plate holder prior to exposure to X-rays. They were scanned over a 29 range between 2° and 40° at a scan rate of 2° per minute in 0.02° step size. The divergence and scattering slits were set at 1.00°, receiving slit at O. l 5mm and monochromator was used at 0.45mm.

Fourier Transform infra-red Spectroscopy (FTJR):
Nicolet spectrometer equipped with a KBr beam splitter was used to obtain inrrared spectra. Calibration for wavenumber accuracy was performed by using polystyrene sample. Dry nitrogen gas was used to purge the beam splitter and sample compartment.
IR spectra were obtained using an attenuated total reflectance (A TR) accessory (single reflection bounce diamond crystal; Golden Gate accessory) . For each spectrum, 32 scans were performed and a resolution of 4 cm was chosen. The ATR accessory enabled to obtain the spectra by pressing the solid material onto the diamond crystal. This accessory eliminates the use of KBr and yet produces comparable results. All samples were dried under vacuum for 20 hrs prior to obtaining any spectra, to remove the influence of residual moisture.

Th ermal and Spectroscopic Analysis of Griseojiilvin-Sugar ester Solid Dispersions:
The ?VP-drug-surfactant ternary solid dispersions: It was demonstrated earlier that griseofulvin had a miscibility of up to 30% w/w in PVP.
At drug concentration of 40% w/w and higher, the samples were found to be crystalline . ln order to investigate the effects of the surfactant on the extent of miscibility between griseofulvin and PVP, sucrose stearate was added to the binary mixture to provide 10%, 20%, 30% and 40% (w/w) of total sample weight. The Tg values of various mixtures along with the XRD results verifying amorphous or crystalline are shown in Table B-Ill-1. The raw XRD patterns of the mixtures are also provided in As seen rrom Table B-lll-1, an increase in the surfactant concentration, rrom 10% to 20% w/w, increased the amorphous fraction of drug in the solid dispersion . At 20% w/w surfactant concentration, 50% w/w of drug in the polymer can be incorporated in the system in the amorphous fonn whereas the solid dispersion that contains 10%w/w surfactant facilitates only 30%w/w griseofulvin in the system. However, when the surfactant loading increased from 20% w/w to 30% w/w, the fraction of drug that was present in the amorphous form was dropped from 50% w/w to 40%. Such a 89 crystallization behavior can be explained by the increased molecular mobility in the system in the presence of higher amounts oflow-Tg sucrose ester.
The Effect of sugar ester on the solid solubility limit of griseofulvin: To investigate the effect of the sucrose stearate on the solid solubility, the ternary solid dispersions were exposed to accelerated stability conditions of 40°C/69%RH for over 30 The arithmetically calculated value of surfactant in PVP in the absence of griseofulvin is 33%w/w. Therefore it can be concluded that sucrose stearate did not improve the solid solubility limit of griseofulvin in PVP.

CONCLUS IONS
ln this study a ternary solid dispersion of drug, polymer and a sugar ester surfactant was prepared and characterized using MDSC and XRD . The surfactant interacted with the polymer through hydrogen bonding and was miscible in it up to 60%w/w concentration.
Although the surfactant-drug mixtures did not exhibit composition dependent miscible Tg, FTIR demonstrated the presence of weak interactions between the two components.
Therefore, the effect of surfactant on the extent of miscibility between the drug and the polymer was investigated. When present at 20%w/w concentrations, the surfactant increased the extent of miscibility from 30%w/w in PVP to 50%w/w. Such an increase in the miscibility was attributed to the physical interactions between the three components.     .....

ABSTRACT:
Purpose: To study the isothermal crystallization of griseofulvin at low moisture contents.
Also, to determine the energy of activation (E,) for griseofulvin crystalli zation using modulated differential scanning calorimeter (MDSC) and non-isothermal crystallization studies.
Methods: Griseofulvin was made amorphous by rapidly cooling from its molten state. Xray diffraction and MDSC were used to generate the calibration plot for isothermal crystalli zation studies. Amorphous griseofulvin was exposed to 0%, 32% and 43 % has been shown to result in its higher solubility [5] . The meta-stable amorphous phase could however interact with moisture during its shelf life storage and undergo crystallization, thus influencing attributes like dissolution, bioavailability and chemical stability of the drugs [6]. Working with amorphous materials, Zografi et. al. [7] have shown that moisture could be selectively absorbed by the disordered regions of amorphous phase, thereby increasing the molecular mobility and inducing crystallization.
Crystallization of amorphous drug substances may be prevented by either minimizing exposure to high humidity conditions, or storing them in low temperatures. It is generally understood that storing the amorphous drug substance at 50°C below its glass transition temperature (T g) could significantly reduce the molecular mobility that initiates crystallization process although this may not be a feasible approach in many instances [8].
If the drug crystallization cannot be prevented, retention of the drug product attributes in a predetennined period (during the shelf life) may be acceptable. In order to do so one must understand the impact of moisture and temperature on the kinetics of drug crystallization. Crystallization kinetics is typically determined by quantifying the percent crystallinity in the drug substance or a drug product as a function of storage time. There are also reports of conducting non-isothermal crystallization studies [9]. To quantify the fraction of crystalline component in a drug substance, various techniques like X-ray diffraction [10][11][12][13][14] , DSC [15,16], solution calorimetry [17,18], water vapor sorption [17,19], thermally stimulated currents [20], density measurements [13,14,21], isothermal microcalorimetry [22,23] etc. have been used.  [ 17].
Although thermally stimulated currents are very sensitive to molecular mobility, they I 15 may not be able to accurately characterize the crystalline phases. One of the techniques that have been widely used for over three decades is the X-ray diffraction technique (24].
This technique has the advantage of being non-thermal in nature but has a disadvantage of detecting only levels as low as 5 -10%w/w.
In this study, modulated DSC was used to study the isothermal crystallization kinetics of amorphous griseofulvin at 0%, 32% and 43% relative humidity conditions. Modulated DSC was chosen since it provided greater reproducibility and had ability to differentiate between the effects of slight variations in the amorphous and crystalline griseofulvin ratios when compared to the X-ray diffraction technique. The calibration plot generated using known levels of crystalline and amorphous griseofulvin had higher slope with MDSC when compared to XRD.
Griseofulvin was chosen as the model compound since it is poorly a water-soluble hydrophobic substance and has tendency to crystallize from its amorphous state. It also has a relatively high Tg of 90°C which makes it possible to study the crystallization pattern of the drug well below its Tg, i.e. room temperature (65°C below the Tg), without refrigeration. In addition, since the amorphous samples were exposed to low relative humidity conditions (a maximum of 43%RH), their Tgs were not plasticized to lower values from 90°C. In this study the non-isothemrnl Kissinger's analysis was used to study the energy of activation for drug crystallization (E.). A review of the pharmaceutical literature to probe into the mechani sm of drug nucleation and crystallization at such low relative humidity and storage temperatures has been provided.

Methods:
Preparation of amorphous griseofulvin: Amorphous griseofulvin was prepared by melt-quenching the crystalline drug substance.
The drug was weighed (3g) in a stainless steel beaker and was placed in an oven maintained at 2 I 7°C (i .e., 5°C above the melting temperature of drug), for 5 minutes. At the end of 5 minutes, it was ensured that drug was in a complete molten state. The stainless steel beaker was immediately immersed in a pool of liquefied nitrogen to rapidly cool (quench) the molten drug. The solidified drug was then grounded in a mortar and pestle and was sifted through sieve # 70 to result in particle size less $ 21 Oµm. Drug substance obtained this way was stored in a freezer in vials over anhydrous CaS0 4 until use. 117

Assay for drug degradation:
The melt-quenched drug substance was analyzed using High Performance Liquid A HPLC calibration plot was first generated before assaying for the drug content. For the calibration plot, untreated crystalline drug substance was dissolved in a mixture of 4: I parts of methanol: water to result a concentration of3.33 mg/mL. Several dilutions of this stock solution were made using the solvent. Each solution was analyzed according to the HPLC method described above. A plot of the area under the peak against the 'concentration' of drug substance was generated. Later, a known amount of melt quench drug substance was dissolved in 4: I methanol: water solution and the percent drug was quantified by obtaining the area under the drug analysis peak and using the calibration plot.

X-ray Powder Diffraction (XRPD):
The X-ray powder diffractometer (Rigaku R!NT (D/Max) 2200, equipped with an Ultimagoniometer) consisted of a 40 KV, 40mA generator with a Cu Ka radiation anode tube. XRD pattern was needed to test the presence of any crystalline drug substance in the melt quenched drug substance. The sample was placed on a 0.5 mm quartz plate holder. They were scanned over a 20 range between 2° and 40° at a scan rate of 2° 20 per minute and step size 0.02°. The divergence and scattering slits were set at 1.00°, receiving slit at 0.15mm and monochromator was used at 0.45mm. X-ray pattern of melt quenched drug substance showed no crystalline peak indicating an amorphous material (Fig. B-lV-2). In addition no birefringence was seen under a polarized light microscope, thus supporting the X-ray analysis.

Preparation of calibration plot:
X-ray diffraction: A calibration plot was constructed in order to quantify the percent crystallinity in amorphous sample during stability studies. Conventionally, X-ray diffraction technique is utilized to study isothermal crystallization kinetics (25]. A physical mixture of amorphous griseofulvin, crystalline griseofulvin and an internal reference as lithium fluoride (20% w/w) was prepared in 300mg quantities to fit in entirely on the X-ray slide.
Mixing was achieved using Turbula MixerT" used at 46 rpm for 3 minutes. Each mixture was then X-ray scanned using the method described in triplicate. The peak height ratio of crystalline griseofulvin and lithium fluoride were then plotted against the weight fraction of amorphous griseofulvin to generate the calibration plot.
Thermal Analysis: The crystallization kinetics of amorphous griseofulvin was monitored using MDSC.
MDSC is a sensitive tool to measure the heat flow associated with thermal events like melting, crystallization, glass transition phenomenon etc. Upon heating, amorphous griseofulvin undergoes a glass transition and crystallizes in sit11 in the DSC pan. The heat of recrystallization (i.e. area under the exothermic area of non-reversing heat flow scan) was found to be experimentally reproducible and was used to obtain a measure of the percent amorphous sample in a given mixture [26]. In order to generate a calibration plot, physical mixtures of crystalline and amorphous griseofulvin, were weighed accurately in known weight proportions in a DSC pan. Calibration plot was constructed by plotting the in sit11 heat of recrystallization of the physical mixture against the percent amorphous griseofulvin.

Mod11lated Differential Scanning Calorimeter (MDSC):
Thermal analysis was performed using MDSC 2920 (TA Instruments, New Castle, DE), with a liquid nitrogen cooling accessory. The analysis was performed under a purge of dry nitrogen gas (60cc/min). High purity indium and sapphire were used frequently to calibrate for the heat flow and heat capacity of the instrument. Samples (-8-12 mg) heated from 0°C to 245°C at 1 °C/min with modulations of 0.266° every 50 seconds. The 120 specified amplitude and period were optimized to provide best results for analysis. The heat of crystallization was obtained by measuring the area under the non reversing heat flow signal of crystallization exothermic peak using the TA Instrument Software for Universal Analysis. Measurement were conducted times n=3 for each time-point to ascertain the reproducibility of experiments.
The amorphous drug substance was accurately weighed (8-J 2mg) in standard DSC pans.
The pans were exposed to isothermal stability condition by placing them in desiccators containing anhydrous calcium sulfate (0% RH), saturated magnesium chloride solution (32% RH) and saturated potassium carbonate solution (43% RH) [27]. By storing the samples in discrete DSC pans, the reproducibility of the results could be assessed without any interference due to sample handling. Samples were taken out for analysis at predetermined time intervals for a period of up to seventy days or until near complete crystallization. The sample pans were crimped with aluminum lids having pinholes to facilitate the removal of absorbed water during heating. The heat of recrystallization of the sample was determined and the percent crystallinity at a specific time point was quantified using the calibration plot.

121
Non isothermal crystallization studies: The energy of activation E, for the crystallization of griseofulvin was determined nonisothermally using Kissinger analysis [28,29].
In Kissinger's analysis, the energy of activation E, is determined by measuring the peak crystallization temperature (Tc) at several heating rates (/J). The slope of a plot of ln(p I T/ ) against I/Tc gives -E,/R, where R is the gas constant. Amorphous griseofulvin samples were weighed in DSC pans and were heated at scanning rates of 1, 5, 7.5, 10, 15, 20 and 25°C/min. The peak crystallization temperatures Tc was obtained from the software.

Determination of moisture content:
The moisture content of the sample was analyzed using the Karl Fischer (KF) titration.
KF titration detects the bound as well as unbound water present in a sample by dissolving it in the KF reagent. The sample is titrated coulometrically. A small quantity of sample was accurately weighed (15 mg) and was titrated with the Karl Fisher reagent to determine the moisture content. The endpoint, provides the total amount of water present was provided by the instrument.

Chromatographic analysis to assess the drug degradation:
The data pertaining to the calibration plot for HPLC is provided in Appendix I. The test solution concentration used for analysis of melt quenched drug substance was 0.3 mglmL. Table B-IV-11 provides data on the assay for the melt quench drug substance of four different lots. As seen from the assay values, no drug degradation was noticed due to melt quenching of crystalline griseofulvin.

X-ray analysis for the construction of calibration plot:
The X-ray scans for physical mixture of 40%: 40%: 20% w/w of crystalline griseofulvin, amorphous griseofulvin and lithium fluoride respectively four reproducibility scans are shown in Fig. B-lV-3. The peak height ratio of crystalline griseofulvin to lithium fluoride for a specific ratio of 40%:40%:20%w/w of amorphous griseofulvin, crystalline griseofulvin and lithium fluoride were not reproducible. Also, the peak height ratio of crystalline griseofulvin and lithium fluoride, when plotted as a function of percentage w/w of griseofulvin yielded a straight line with very poor slope of 0.0097, as seen in Fig.   B-lV-4. The slope obtained was too small to be able to accurately resolve the differences in the percent amorphous griseofulvin in the mixture. Hence X-ray analysis was not chosen as a suitable technique to study isothennal crystallization of griseofulvin. 123

Thermal analysis ofgriseofit!vin:
The DSC scans of amorphous and crystalline griseofulvin are shown in Fig B-IV-5 . The amorphous griseofulvin undergoes a non-reversible exothermic recrystallization during the heating scan with peak maxima at around 81 ° and l l 9°C. It appears from the two distinct exothermic peaks that recrystallization occurs in two phases involving a structural transformation from a metastable crystalline form at lower temperature into a stable crystalline one at elevated temperatures. A similar observation in the recrystallization behavior was found in the case of nifedipine and a solid-state transformation was accounted as the reason for two exothermic crystallization peaks [30].
The glass transition event of amorphous griseofulvin was characterized at 92°C from the reversing heat flow scan and is shown as an inset in Fig

Construction of calibration plot: MDSC
Since X-ray diffraction did not prove to be a suitable technique for this compound, increases. To generate the calibration plot, the heat of recrystallization (J/g) that was obtained from the area under the recrystallization peaks was plotted against the percentage of amorphous griseofulvin present (Fig. B-IV-7). As seen from Fig. B-IV-7, the calibration plot has a much higher slope when compared to the plot generated using X-ray analysis and hence was found more suitable to interpret the crystallization studies.

Isothermal stability studies:
Amorphous griseofulvin samples that were exposed to 0%, 32% and 43% RH conditions were analyzed periodically to monitor the changes that occurred in their crystallinity. The As can be seen from the plot, the rate and extent of crystallization of the samples stored at 43% RH was much greater than those stored at 32% RH. Even the samples stored at 0% RH showed significant crystallization although at a much smaller rate.
The total moisture absorbed by the samples was obtained using the Karl Fischer titrimetric analysis and are tabulated in Table B-IV-IV.

Non-Isothermal stability studies:
Non isothem1al crystallization studies were perfonned using Kissinger's analysis. The peak crystallization temperature of amorphous griseofulvin was recorded by heating it at different rates. The non reversing heat flow scans of amorphous griseofulvin are shown in FigB-fV-10.

Isothermal crystallization kinetics:
The crystallization kinetics was monitored with time as a function of relative humidity by storing the samples in isothennal stability conditions. The rate and extent of crystallization of griseofulvin have been plotted in Figure 9 and the crystallization kinetics was calculated based on Hancock-Sharp equation [3 I) .

Hancock-Sharp equation:
Unlike the chemical degradation rates in solution, the rate of alteration in the physical state of a powder is difficult to detennine from accelerated stability studies . However, predictions of some physical alteration pathways such as crystallization, dehydration, polymorphism etc., have been attempted. The Hancock-Sharp equation is often used to describe the kinetics of reversion of amorphous phase to its crystalline fonn and is written as follows: ln[-ln(J -a)) = lnB +mint 126 where a the fraction of drug that has crystallized, t is the storage time, B is a constant and m is a constant relating to the mechanism of griseofulvin crystallization. The kinetic equations for the most common mechanisms that are believed to operate in solid-state decomposition are presented in Table B chose the appropriate mechanism that would fit the experimentally obtained crystallization data, the results were fit to all the equations provided in Table B-IV-Ill.
The results of such fit are provided in Table E Kolmorgorov-Johnson-Mehl-Avrami (KJMA) equation: A first order degradation mechanism has often been chosen to plot crystallization kinetic plots [32 , 33]. The relationship pertaining to this mechanism is given by the KJMA equation and is described as: where k is the rate constant for crystallization, t is the storage time and to is the induction or the nucleation time. The rate constants for crystallization obtained from the slope of the linear equation are given as a function ofrelative humidity in Table B-IV-IV. The rate constants were found to be higher at high relative humidity conditions. This was not 127 surprising because at higher relative hwnidity, water in the samples could cause an increased molecular mobility and hence increased crystallization rates.

Non-isothermal crystallization:
Non isothennal crystallization was performed using Kissinger's analysis to determine E •.
The E, value was determined from the slope of the linear plot of ln(jJ I T/ ) and l/T, as shown in Fig. B-lV-13. In this plot, fJ is the heating rate and T, is the peak crystallization temperature. With an increase in the heating rate, an increase in the peak crystallization temperature was noticed (Fig. B-IV-10). Although the area under the exothermic crystallization peak appears to increase with an increase in the heating rate, the absolute values were found to be constant (Fig. B-lV-10). This indicated that irrespective of the heating rate employed in this study, amorphous griseofulvin crystallized completely.
Similarly at higher heating rates, most of the crystallization probably occurred over a narrow temperature domain and hence, a greater peak appeared compared to the ones obtained with slower heating rates, where crystall ization occurred gradually over a larger temperature domai n.
The slope of the line in Fig. B-IV-13 provided -E,/R values, where R is the gas constant, and E, was calculated as 239 KJ/mol. The E, value for the crystallization of griseofulvin has not been reported in the literature for comparison with the values obtained in this study. However, the E, value of griseofulvin was found to be much lower than the E, of 317 KJ/mol reported for crystallization of lactose by Schmitt et. al [32]. This agrees with 128 the isothermal crystallization rates which seem to be very high at experimental conditions employed.

Mechanism of crystallization:
The mechanism of crystallization of amorphous materials has been studied extensively in the literature [34][35][36] . The process of crystallization involves nucleation and crystal growth. It is generally considered that crystallization from amorphous state can occur only at temperatures above the T g of the material where the viscosity of the amorphous state is low and conducive for nucleation and subsequent crystal growth. However for amorphous indomethacin, crystallization has been shown to occur at temperatures 30°C below the Tg [37]. In our case griseofulvin crystallized at temperatures as low as 65°C below the T g· The crystallization at this low temperature was expressed by the Turnbull and Fischer equation [20]: where V* is the nucleation rate per unit volume, N is the number of molecules per unit volume, k is the Boltzman constant, h is Planck's constant, and Tis the temperature. The term llG K is the Gibbs free energy change for the formation of nucleus with a critical size and llG 0 is the free energy associated with transportation of molecules from the amorphous bulk onto the growing nucleus.
It can be seen from the equation shown above that nucleation rate is the result of net effect between two opposing factors: temperature and super-cooling. When temperature 129 decreases, the degree of super cooling increases. This may correspond to the increase in the temperature difference between the melting temperature and storage temperature.
Under such conditions, the nucleation rate would be expected to increase exponentially.
However, temperature reduction also causes reduction of the molecular mobility and hence decreases nucleation probability. Although in varying amounts, molecular mobility changes will impact both f'.G K and !'.G 0 . In equation 4, the term f'.G K is dependent upon the interfacial energy between the nucleus and amorphous bulk as well as degree of super cooling, and is less influenced by the magnitude of changes in the molecular mobility of a system at a constant temperature. f'.G 0 on the hand, encompasses the energy of activation for diffusion through the bulk phase and this factor is largely affected by molecular mobility.
For griseofulvin crystallization, it can be assumed that the molecular mobility is still strong even at temperatures as low as 65°C below the T g and hence nucleation, and crystal growth, results. A similar explanation was provided for the crystallization of amorphous indomethacin at temperatures below T g and nucleation was explained by the rotational motion of molecules. This assumption can be true for griseofulvin considering its small chemical structure and ease of rotation. The increased moisture content of the drug probably increases the ease of rotation and translational diffusion (Table B-IV-IV).
Increased molecular mobility is likely to increase the crystallization rate constant. increasing relative humidity indicated higher molecular mobility at high relative humidities, which enhanced the crystallization process. High molecular mobility was the cause for crystallization at 0% RH and temperatures as low as 65°C below T g· No crystallization from amorphous phase was noticed in samples stored at -20°C (i .e. 110°C below T g) . The theory of crystallization and the factors affecting the nucleation and crystallization growth rate have also been reviewed. According to the data obtained, it can be concluded that amorphous griseofulvin should be stored at very low temperatures in order to retain its amorphous form. 131

INTRODUCTrON
Polymers are widely used in the pharmaceutical industry as excipients in several dosage forms like tablets, capsules, trans-dermal dosage forms , inhalation products, suppositories and many other formulation types [I]. They have also been used to modify the drug dissolution profile and to create a sustained or delayed release effect to meet therapeutic requirements [2]. The usage of polymers to improve the aqueous solubility of poorly soluble drugs and to inhibit the drug crystallization has also been reported in the pharmaceutical literature extensively (3)(4)(5)(6). The mechanism of crystallization inhibition by polymers in amorphous solid dispersions has been postulated to be the ability of the polymer to form a miscible phase with the drug [7]. In this molecularly miscible phase, the coupling molecular motions of long chain polymers reduce the net molecular motion of small drug molecules and hence delay the crystallization of the amorphous drug. It is important however to monitor the molecular mobility of the system since the thermodynamic non-equilibrium nature of amorphous drug would always favor crystallization to the lower energy state, which is poorly water soluble.
An amorphous material constantly undergoes structural or enthalpy relaxation to achieve its low-energy super-cooled equilibrium state or a crystalline state as depicted in Fig. B-Y-I. Retention of the amorphous state of the solid would therefore depend upon the extent to which the material can structurally relax to lower its enthalpy or heat content.
The higher the molecular mobility of the material, the higher would be the rate at which it 157 will structurally relax. Therefore molecular mobility measurements are necessary to understand the dynamics of crystallization.
It is however, difficult to quantify directly the structural relaxation or the mean relaxation time, which is an indicator of molecular mobility for amorphous solids below their T g· Such difficulty is due to (a) long time scales of molecular motions and (b) complex relationship between the mean relaxation time and storage temperature [8,9]. Despite these challenges, various techniques including enthalpy relaxation measurements [I 0-12] , dielectric spectroscopy [ 13], viscosity measurements [14] , and nuclear magnetic resonance [ 15,16] have been used to estimate the mean relaxation time.
The purpose of this study was firstly to conduct enthalpy relaxation measurements and to

Modulated Differential Scanning Calorimeter (MDSC):
A Modulated Differential Scanning Calorimeter (MDSC) (TA Instruments 2920, New Castle, DE), with liquid nitrogen cooling accessory was used for the enthalpy recovery measurements. Analysis was performed under a purge of dry nitrogen gas (60cc/min).
High purity indium and sapphire were used bimonthly to calibrate for the heat flow and heat capacity of the instrument.

MDSC Experimental Protocol:
The experimental protocol is provided in Fig. B-V-2. The glass transition temperatures of samples (8-l 2mg) were determined by first heating to 25°C above the T g (Tr in Fig. B-V-2), to erase the thermal history, and then quench cooling to l 50°C below their Tg (T 0 ).
The Tg was detected upon a second heating scan as the midpoint of the inflection in the heat capacity or the reversing heat flow signal. A heating rate of 3°C/min, with modulations of ±I 0 every 60 seconds was used for all analysis. Following the Tg determination, the samples were aged at T., which were 5°, 15° and 30°C below the respective T g for a period (T,)

Thermally Stimulated Polarization Current (TSPC):
The principle of TSPC is depicted schematically in Fig. B-V-3 . The sample is heated across its transition temperature(s) in the presence of a high voltage electric field and the movement of the polar groups (dipoles) of molecules is measured as they orient to the applied field. The sample is placed between two electrodes of a parallel plane capacitor.
As the temperature is increased, the molecular mobility rises and the dipole orients to the applied field when sufficient mobility exists. The motion of the dipole generates a polarization current (I) which is amplified and detected. The intensity of the polarization 160 current (output signal) provides a direct probe into the degree of molecular mobility of the materials as a function of temperature.

Th ermally Stimulated Depolarization Current (TSDC):
The TSDC experiments are similar to TSPC measurements, with an exception that the samples are first heated to temperature above the suspected transition and are then polarized by applying an electric field. The polarization is frozen by rapidly cooling the samples to low temperatures. The depolarization current (opposite in direction to polarization current) is measured in the absence of electric field, by heating the sample across the suspected transition(s). The driving force for molecular motion upon reheating the sample is the restoration of neutrality or the original configuration. The molecular motion causes a depolarization current which is amplified and detected. The glass transition temperature is characterized by the temperature of peak maximum of the broad peak as shown schematically in Fig. B-V-3b.

TSPC Experimental protocol:
In this study, TSPC experiments were perfom1ed unless specified. TSPC or TSDC experiments were carried out using TSC/RMA 9000 Instrument (TherMold Partners, Stamford, CT). All experiments were conducted in an atmosphere of high purity helium gas. Samples (13.5 ± O.Smg) were placed in aluminium pans, covered with teflon lid and placed between the screw electrodes. This assembly was enclosed in a Faraday cage that was evacuated to 10 4 mbar and flushed several times with helium gas prior to experiments. The experimental procedure used to measure enthalpy recovery using TSPC 16 1 is depicted in Fig. B

MDSC:
The changes in heat capacity for different molecular weight grades of PVP obtained using MDSC are shown in Fig. B-V-4a. Glass transition temperatures T g, is denoted as the midpoint of inflection in the heat capacity step-change. As seen from the Fig B-V-4a, each molecular weight grade of PVP exhibits a characteristic Tg that increased with an 162 increase in the molecular weight. ln addition, an increase in the heat capacity step-change !!. Cp was also observed with increase in the molecular weight.

TSPC:
In Fig. B Similarly the T2max of a specified polymer corresponded to respective Tg obtained from MDSC as seen in Table B-V-1. In addition, the current intensity was proportional to the applied electric voltage, indicating that the signal generated was indeed from the sample and not due to any mechanical or charge dissipation.
It was interesting to see that there was a difference of at least 30°C between the two depolarization peaks Pl and P2 for all molecular weights reported so far. The peak differences may indicate entirely different modes of molecular motion.
It is generally understood that the activation energy (Eae1) for the relaxation of dipoles is not singular in nature and is a function of temperature [ 17]. Usually there is no significant difference between the activation energies for dipole relaxations that constitute a specific molecular motion e.g. side chain motion of a polymer, enthalpy relaxation, main chain 163 ( relaxation etc. In such cases, the Eae1 of different dipoles that constitute the overall relaxation overlaps to result in a continuous distribution as a function of temperature. The result is a single broad peak in the depolarization or the polarization current output. In the present case, however, a difference of at least 30°C was seen between two peaks indicating that molecular motions ti-om entirely different domains exist. This can be explained further by noting that the dipolar side group -pyrrolidone can rotate locally through its C-N amide linkage, besides moving in union with the main chain . With increase in temperature, a ' localized cooperative rotational relaxation motion' of the pyrrolidone side group is possible and this may constitute the P-current peak (P 1 ). A Prelaxation peak has been reported fo r PVP samples containing 23% w/w water at -60°C using dielectric analysis [18]. ln a different study, PVP was shown to exhibit P-relaxation at approximately -73°C (MW 40 KDa) [ 19]. A similar explanation for P-relaxation peak was given in the case of polyester liquid crystalline polymer films, polycarbonate phenyl rotation, polystyrene phenyl rotation etc. [20].
The molecular motions of PVP that are believed to be due to the cooperative rotational motion of the side chains did not produce appreciable changes in the heat capacity and therefore remained undetected by MDSC. At higher temperature regions, the main chain motion results in the CL-peak (P2) that is characterized by the glass transition temperature by MDSC as well. Moreover, since the two dipole peaks are broad, they can be considered to occur ti-om distribution of Eact of several side group and main-chain dipoles. We also observed that with an increase in the applied voltage, the intensity of polarization current increases at P 1 and P 2 . This phenomenon further indicated that the peaks are indeed due to the molecular motions in the sample rather than resulting from di ssipation of unknown trapped charges or any other mechanical stress.

Enthalpy relaxation measurements with MDSC and TSPC:
The protocol described in Fig. B-V-2 was followed to understand the structural relaxation behavior in PVP.

Enthalpy relaxation measurements using MDSC:
The non-reversing heat flow signals of different molecular weight grades of PVP that were aged at an aging temperature (T,) of l 5°C below the respective Tg is shown in Fig. B-V-5. An increase in the area under the non-reversing heat flow was observed with aging time for all the three polymers. As the sample was held isothennally at temperature below its Tg, it underwent structural relaxation process, which led to its densification. The increase in the area under the curve is due to the increasing magnitude of heat input that is necessary to overcome the densified state of the polymer and to reach the supercooled equilibrium state. In addition, the temperature of heat-flow maximum (Tn,,x) shifts to higher values, indicating the need for higher energy of activation to mobilize the structurally relaxed segments of the polymer. These results are consistent with the results published by other groups [21 ]. The enthalpy recovery values were obtained by integrating the area under the curve and were plotted as a function of aging time (Fig. B-V -6). No significant differences were seen in the enthalpy recovery values for a corresponding aging time and temperature for the samples with different molecular weights. 165

Estimation of relaxation time constants:
From the enthalpy relaxation measurements, it appears that significant molecular mobility exists below the glass transition temperature. ln order to compare the degree of molecular mobility between different polymers, it is necessary to calculate their relaxation time constants ("r). Relaxation time constants (-r) is the average time taken for one relaxation event. These constants can be calculated by fitting the enthalpy recovery data to stretched-exponential equation known as Kolrausch-William-Watts equation [22,23) and obtaining adjustable relaxation time constants T and Pas follows : where<P is a distribution function of different relaxation times occurring for a specific molecular relaxation process, t is the storage period, -r is the mean relaxation time constant and j3 is the relaxation time distribution parameter with a values between 0 and 1.
The relaxation distribution function can be obtained from the enthalpy recovery measurements by using the relationship: (2) where<P, is the extent to which the material relaxes; !!Ji, is the experimentally measured enthalpy recovery and !!Ji w is the enthalpy recovery of a fully relaxed material. !!Ji w is calculated as :

Ml~ =(T, -T)t:,Cp
where Tg is the glass transition temperature, T is the experimental storage temperature and t:,C Pis the heat capacity step-change at the glass transition.
By rearranging and equating equation 1-3, an expression to estimate the relaxation time constant can be obtained as follows: The experimental enthalpy recovery values were then fit to equation 4, using MicroMath® Scientist software (version 2.0) to obtain i: and p values (Table B- The empirical KWW function has been shown to describe the various relaxation processes of amorphous materials during enthalpy relaxation process [25). However data interpretation should be done cautiously, because distribution of the relaxation time constants is as much important as the average relaxation time constants to access information pertaining to molecular mobility and associated stability. In Table B temperature [26]. Similarly a decrease in the mean relaxation time indicates an increase in the molecular mobility. Such a pattern is expected since the molecular mobility is known to increase with an increase in temperature. Further, the relaxation time constant for all the polymers tend to approach each other near the T g·

Enthalpy relaxation measurements using TSPC:
Since the electric field was turned off in TSPC measurements during sample aging, a comparison of enthalpy relaxation measurements between the two techniques is valid.
When the samples were heated following the protocol described in Fig. B-V-2, a series of relaxation process generate the electrical signal. Jn step 1 and 2, heating the samples to temperatures above T g removes any electric and/ or mechanical stress, which otherwise would have produced noise peaks in TSC analysis. Also the thermal history of the samples was erased in steps 1 and 2 to generate 'fresh' amorphous samples. ln step 3, the samples were stored at very low temperatures, and hence the dipoles were kinetically frozen. The mean relaxation time (i:) (i.e. average time taken for one molecular motion to occur) is very large at low temperatures and thereby no molecular mobility and hence polari zation current was generated. As the samples were heated in step 4, the relaxation time of the dipoles decreases, and at a given temperature domain, the rate of molecular relaxation matches the experimental time scales thus producing polarization current. The temperature where the polymer main chain exhibits maximum mobility (i.e. peak 168 polarization current) is the glass transition temperature. In step 5, beyond the glass transition temperature, the average relaxation time of the molecules exceeds the heating rate and the motion of dipoles cancel out. The net current thereby lowers and eventually decays to the baseline.
The TSPC scans of PVP K 17 and PVP K30 that were aged at I 5°C below their respective T g for a period of up to 8 hours are shown in Fig. B-V-8. The TSPC scan demonstrates two distinct global relaxation peaks analogous to the TSDC scans except for the negative sign. As a trend, for either polymers, the peak area at PI decreased in magnitude whereas the peak area at P2 increase with aging time. However, the total area under the peaks at P 1 and P2 remained constant.

Polarization current at P1 and Pi:
The reduction in the area of current at P 1 was observed in case of all different molecular weight grades of PVP. Since the aging temperature (T,) was always greater than the temperature domain corresponding to P 1 , the dipoles responsible for producing current at P 1 could always be considered to be in a state of sufficient mobility so that aging does not influence them. Hence one would expect no change in their polarization current intensity with aging time, when compared to that of a freshly prepared amorphous sample.
However a decrease in the polarization current peak suggesting a progressive decrease in the molecular mobility of dipoles was observed. An explanation to such behavior could be the possible entrapment of side chains of the polymer with the main chain upon 169 densification. Upon subsequent heating of sample at the end of aging, cooperative molecular motions between the molecular side chains and the main chain could possibly explain the reduced mobility and hence polarization current. When the samples are heated above their T g, the main chain and side chains are free of conformational restrictions and could be considered 'fresh'.
The area under the polarization current peak at P 2 was seen to increase with aging period unlike the peak at P 1 . Also an increase in Tmax (i .e. temperature corresponding to peak maximum intensity) at P 2 was seen as function of aging time ( Figure B-V -8). The increase in the area under signal at P 2 is analogous to the increase in the enthalpy relaxation endotherrn that is obtained at Tg from MDSC measurement. Since the aging temperature is below the temperature domain corresponding to P2, as the sample ages, it undergoes structural relaxation to attain the equilibrium state. The density of sample increases with an associated decrease in the free-volume. When the sample is reheated, the densified molecules regain their super-cooled state at their Tg. During this process, an increase in the molecular mobility, most likely due to a "burst effect", causes an increase the area of current at P 2 • This increase in the area at P2 was found proportional to the aging time.

Polarization current p eak cleaning (P 1 ) :
Based on the data generated so far it was not possible to obtain the enthalpy relaxation due to aging at the Tg, primarily due to interference from peak P 1 • Therefore the 170 polarization peak at Pl was 'removed' by modifying the experimental protocol. The modified protocol is shown in Fig. B-V-9. As seen from the protocol, a polarization step for (T p) 15 min . was added at the end of each aging period (step 6) to polarize the dipoles responsible for peak P 1 • Since the samples were rapidly cooled (step 7) in the presence of electric field, the now polarized dipoles responsible for P1 remained in a frozen polarized state. Upon subsequent heating, the polarized dipoles would not produce any significant polarization current in the presence of electric field and the motion of main chain alone could be characterized.
Following this protocol, the modified TSPC plots of PVP samples after aging for periods up to 8 hours are shown in Fig. B-V-10. No polarization current peak at P1 was seen in these scans when compared with TSPC plots in Fig. B-V-8. Favorably, the area under the polarization current peak at P 2 could now be integrated. This area when subtracted from the area of polarization current due to freshly prepared amorphous sample provided the cumulative molecular mobility of all the dipoles that underwent enthalpy recovery.

Correlation ofTSPC and MDSC results to estimate molecular mobility:
Plots of cumulative normalized area of PVP samples after aging for a period up to 8 hrs following modified TSPC method are shown in Fig. B-V-11. Similar to the MDSC experiments, an initial increase in the polarization current followed by a plateau in the cumulative area under the peak at P 2 was observed with the TSPC (modified) experiments. This indicated that the heat flow involving and non-heat flow involving techniques characterize the same phenomenon however in a different fashion. ln MDSC experiments, the heat input that is necessary to bring the structurally relaxed state of material to its super cooled state is measured. Whereas in TSPC, the molecular mobility generated when the structurally relaxed material regains its equilibrium state is obtained from the polarization current. Therefore, the two techniques correlate in principle. ln order to investigate if the experimental results correlated a plot of the net current generated at peak P 1 of structurally relaxed material and the enthalpy of recovery of the material that were aged for the same duration at a given aging temperature was constructed (Fig. B-V-8). An empirical linear relationship was observed between the data generated from the enthalpy recovery (MDSC) and the molecular mobility recovery (TSPC) techniques.
The slope of the linear line should then represent the rate of enthalpy relaxation, which indicates the extent of molecular mobility. Therefore, the relative extent of molecular mobility of a polymer at a specified storage temperature below its respective glass transition temperature could be obtained. From and least for PVP K30 (Mw -58 KDa ±-5KDa). Although it is intuitive to expect a decrease in the molecular mobility between PVP K17 and PVP K90 (Mw-130KDa± -5KDa), due to a reduction in the molecular weight, it is counter-intuitive to expect PVP K90 to have higher degree of mobility when compared to PVP K30. The mean relaxation time data (Fig. B-V-5) also suggest the rank order for decrease in the molecular mobility as PVP Kl 7 > PVP K90 > PVP K30, which agrees with TSPC results. It is not completely clear as to why there is such a trend in the results. One possible explanation for such a behavior could be the heat capacity step-change (ti. Cp) at the Tg for the polymer, which is the indicator of the non-equilibrium state of the amorphous phase with reference to its extrapolated supercooled liquid state (Fig. B-V-1 ). A larger heat capacity step-change (L\.Cp) at Tg would imply a greater driving force for the sample to reach equilibrium and hence a greater degree of molecular mobility. The higher (ti. Cp) values for PVP K90 when compared to PVP K30, could possibly explain for the higher degree of mobility in the former. However, there seems to be a balance between the two factors of molecular weight and heat capacity change at Tg, since although the (L\.Cp) values for PVP K90 is greater than PVP Kl 7, the large differences in their molecular weights could potentially guide the extent of molecular mobility.

CONCLUSIONS
In this study, we have estimated the relative extent of molecular mobility as a function of molecular weights for three different grades of PVP . Two analytical instruments which measure on different principles, one that involves heat-flow measurements and the other that involves a direct measurement of molecular motions, were compared.
MDSC was used to perform enthalpy relaxation measurements to gain insight into the   Figure  Thennally stimulated polarization currents following the modified experimental protocol for PVP Kl 7 that was aged at Tg-l 5°C. Using the modified protocol, the sub-Tg peak current peak is not produced (compare with Fig. B-V-8). Thermally stimulated polarization currents following the modified experimental protocol for PVP K30 that was aged at Tg-l 5°C. Using the modified protocol , the sub-Tg peak current peak is not produced (compare with Fig. B-V-8).

CUITent
l'urrent Amps ..c: ..... One of the limitations of DSC is the difficulty in interpreting the heat flow if multiple processes are involved over the same temperature range. For instance, a moisture removal and glass transition processes can occur between 60° -I 00°C, and hence it is not possible to resolve the two signals. The sensitivity of the experiment can however be improved by decreasing the sample mass or heating rate; whereas the resolution to separate transitions 197 can be increased by increasing the sample mass or heating rate. Therefore to obtain a balance between the sensitivity and resolution is challenging.

Principles of Operation:
In DSC, the differential heat flow or heat flux is expressed using the thermal equivalent of Ohm's Law as: 198 where dQ = heat flow dt t:i.T = temperature difference between the reference and sample R 0 = thermal resistance of constantan disc (1) In MDSC, the same heat flux is modified using sinusoidal temperature oscillation. The heating rate is dependent upon three experimental variables: underlying heating rate, the amplitude of modulation and the period (frequency) of modulation.
A different way of expressing equation 1 is as follows: The heat capacity term multiplied by underlying heating rate gives the reversing heat flow . The non-reversing heat flow is obtained from the difference between the total heat flow and the reversing heat flow signals. (} · ·O· · · · · · ·0-· · · · · ·O · · · · · ·O· ··ff ·°il: :     During the DSC experiments, water is driven away from the sample (8-12 mg), when it is slowly heated at t °C/min rate, at temperatures above 100°C.
To test whether the reduction in T g2 values was the result of the moisture retained in the dispersions, dextran (containing no trehalose), which was previously exposed to 50°C I 75% RH for a month and contained l 1% w/w water, was heated. Its Tg 2 was 225°C ; which was identical to the T g of original sample that had -3% w/w moisture content. Th.is finding indicates that water did not influence the T g2 values during the heating scan in DSC.
Secondly, the Tg2 values of heat flow scans of l 0% w/w trehalose-dextran mixtures that were exposed to various stability conditions and contained varying levels of moisture are demonstrated in Fig B-16. They all are consistent with that of the freshly prepared samples (Fig. 2a in the manuscript) indicating that moisture did not influence Tg2. 2 16 Reduction in Tg2 (from 225°C) can be explained by the solubilization of trehalose in the polymer. For instance, trehalose mixtures above their solubility limit (i.e. 20%, 30%, 40% and 70%) were crystallized when they were exposed to stability conditions and their Tg2 values increased to 207-2 I 0°C (at 50°C/75% RH). On the other hand, the samples containing I 0 % trehalose (the concentration below the solubility limit), showed no crystallization and Tg2 changes confirming that, reduction in Tg2 values reflects the complete miscibility oftrehalose with the polymer.
Water however, seems to impact Tg2 by facilitating the phase separation and crystallization of trehalose, as described in detail on PI 2,under  We have ongoing studies specifically addressing the impact of temperature and moi sture on the solid solubility. We plan to submit that work to this journal as a follow up study. 2 17  DSC scans of dextran that was exposed to 50°C/75%RH for one month and containing 11 % w/w moisture. The sample was heated in MDSC at the rate of I °C/min, with ±0.266° every 50seconds to 240°C. Response: The FTIR spectroscopy was perfonned on the freshly prepared samples to detennine the nature of interactions between trehalose and dextran. However, no meaningful interpretation of the -OH stretch (free vs. bound), could be made as a result of the presence of multiple hydroxyls of trehalose, dextran and water (Fig. B-17). XRD and NMR were not used for these mixtures due to accessibility problems. Currently we have easier access to XRD and are using it as the supporting tool in the remaining studies.
Regarding the concerns that you raised about the certainty of deviations in Fig. 2   "Why there arc no recrystallization/melting peaks in Fig. 3? If the phase segregation happened as illustrated in Fig. 7, anhydrate form could be formed.
Why not?" Response: Yes it could. However, under hot-stage microscope, we observed that trehalose undergoes the glass transition and subsequently liquefies before nucleating to anhydrate form . The presence of polymer is likely to inhibit the nucleation of anhydrate form and as the result; no recrystallization/melting peaks are seen.
"Is there any fundamental difference between dextran-trchalosc and PVPtrehalose? Tg (eq) of the former case is so close to the melting point of trehalose, but not for the PVP dispersed system.".
Response: The difference between trehalose-dextran and trehalose-PVP mixtures is, the most likely, the extent of miscibility of the sugar and the respective polymers.
In general, the degree of miscibility of the components may depend upon factors including (1) molecular sizes of two components i.e. two sugars are easy to mix when compared to two polymers; (2) the differences in the Tg of individual components i.e. the smaller the differences in Tgs, the lower is the difference in their relative molecular mobility at mixing temperature; (3) structural similarities between two components i.e. like dissolves like; and ( 4) potential for interactions between the components i.e.
interactions promotes mixing.
The chemical structure of trehalose and the dextran repeat unit are very similar and also there is a potential for hydrogen bonding between the two components. For PVP mixtures the difference between the Tg of PVP (167°C) and trehalose (119°C) is less when compared to dextran (225°C) and trehalose. Also, the interactions between trehalose -OH and the basic C=O of PVP is considered stronger than trehalose-OH and dextran -OH .
The Tg'q found in the trehalose-dextran mixture was lowered from 225°C to the values that happen to coincide with the melting temperature of anhydrous trehalose for samples 226 that were stored at 40°C and 50°C/75%RH storage conditions. This was a coincidence! Similar reduction in Tg values for PVP mixtures was seen from I 68°C to I 58°C due to trehalose dissolution in PVP.

Calculations of the thermodvnamic parameters
The excess enthalpy, entropy and free energy of mixing of trehalose with dextran and PVP were calculated using the theoretical equations provided below. The calculation involves binary mixtures and their phase stability behavior. ln the presence of water, the mixtures are however considered as ternary. Although there is a difference between a two-phase mixing and three-phase mixing, the free energy minimum of mixing ( Fig. 7; manuscript I), shows that the thermodynamic stability of the system is expected at its maximum around 20%w/w trehalose-concentration. Our observations showed that T g2 values reached a plateau suggesting "equilibrium" of the three phase system that corresponded to 12-18% trehalose concentration, i.e. values closer for the predicted system with no water. Although the calculations are made based on the trehalose-polymer interactions, it may be more than a coincidence.
The molar enthalpy of polymer I, having glass transition temperature T gl , in its glassy state, at a temperature T, is given as:

{I)
Similarly, the molar enthalpy in liquid state H: (T) can be described as: According to Ehrenfest's theory of second-order transition; H : (T, ,) = H , • (T, ,), and so equations (I) and (2) can be equated at T g l .
For a miscible blend containing polymer P1 and sugar S2 with weight fractions w 1 and w2 respecti vely, and undergoing a single glass transition temperature Tg 12 , the enthalpy of mixture, H 11 (Ji g) can be described by the relation (30): where H 1 and H 2 are the enthalpy in either liquid or glassy states of polymer and sugar respectively, and l'!.H.,ix is the enthalpy of mixing.

229
Upon calculating the enthalpy and entropy, the free energy of mixing can be calculated using the following equation: Heat capacity-Temperature scans oflyophilized dextran .....    Since, the enthalpy of mixing, Af-l mL,= f!J/ ;deatmit +Af-/!u , andM-l it1ealmix =0, the excess enthalpy of mixing is equal to total enthalpy of mixing as given in the above indoprofen is a relatively stable crystalline molecule due to its high heat of fusion.
Also it is expected to be relatively unstable in its amorphous phase due to its lower T 8 (50°C) and lower recrystallization temperature (72°C).  In order to understand the nature of interactions between drug molecules in their amorphous and crystalline forms a spectroscopic investigation was conducted.
Such a study wo uld provide information on the various routes through which the drug molecules crystallize. Vibrational spectroscopy is a technique that is well established and is typically used to study specific interactions. IT-IR spectroscopy was used in this study to probe into the state of functio nal groups of drugs in their amorphous and crystalline states and the spectra are shown in Fig. C-3 . Differences are evident 248 between the two forms of both the drugs in the carbonyl stretch region of 1750 cm·' and 1600 cm·' .
Griseoful vi n has three methoxy units and two carbonyl groups as shown in Fig. la (chapter 2). On its FT-IR spectra, in addition to broadening of peaks in the amorphous sample, a weak peak at l 739cm·' was seen for crystalline fo rm indicating the presence of unbound C=O groups (Fig. C-3A). Also a strong peak at l 598cm·' presumably due lo the carbon-carbon structural ordering of molecules is seen in crystalline form and is absent in amorphous phase. Since there are no signi ficant differences in the peak pattern of carbon yl groups between the amorphous and crystalline griseoful vin, it is possible that the functional groups do not involve through interactions during drug crystallization.
lndoprofen has two carbonyl groups: a) C=O of -COOH group and b) C=O of the amide in pyrrolidone ring (Fig. C-38). ln the case of crystalline indoprofen, a weak peak at 1728cm·' and a strong peak at 1697 cm·' was seen. The presence of a high intensity peak at lower wave-numbers relati ve to peak at 1728cm·' (t. v=J lcm·') indicates that the C=O of -COOH is largely present in a bound stale. The band at I68 Icm· 1 is assigned lo the C=O of pyrrolidone ring. In the case of amorphous indoprofen, the peak al l 728cm·' has much higher intensity when compared to the peak at 1697cm·', indicating that the C=O of -COOH are relati vely "free". Actually, the peak at l 728cm·' is characteristic of free C=O of -COOH groups where the -OH group is hydrogen bonded lo a di ffe rent functional group. The carbonyl band of -COOH group where both C=O and -OH groups are free appear at l 750cm·' . It the amorphous state, in addition to band at 168 lcm·' , a broad band at 1648cm·' was 24 9 observed. This band was assigned to the hydrogen bonded C=O the pyrrolidone ring (ti. v=33cm-1 ). The C=O of pyrrolidone ring could most likely be bonded to -OH group of -COOH. This would agree with previous observation of free carbonyl group of -COOH stretch at l 728cm·' instead of l 750cm·'. Therefore it can be concluded that indoprofen molecules crystallize by fom1ing cyclic dimers between two -COOH units of individual molecules. Fi gure C-20 MDSC reversing heat flow scans of PVP solid dispersions containing (a) I 0%w/w indoprofen exposed to 40°C/0%RH for 34 days. To estimate the kinetic rate of phase separation of the drug from the polymers, the changes in the Tg of the mixture were utilized to calculate the fraction of the drug that had phase-separated. It is assumed that wi th increasing phase-separation of drug from the polymer, the Tg of the mixture would increase and eventually reach to Tg of pure polymer when no drug remains miscible.
Based on this assumption, a mixture having Tg equalt lo that of a freshly prepared sample was accounted as I 00% miscible mixture. Likewise a Tg of I 67°C indicated a 0% miscible mixture. Accordingly, the percentage of drug miscible was calculated at a specific storage time (t) as follows: (I-a), =l

Tg(poly~r) -Tg(initiaf)
where a is the amount of drug phase separated, Tg(polymc•) is the 167°C, Tg(;n;1;,1) is the Tg of freshly prepared sample and is dependent on the composition of drug present, T g2(t) is the Tg of the mixture at the specified storage time/.
Upon obtaining the fraction of drug phase separated with storage time, the phase separation rate constant was estimated by using a first -order rate equation: (-ln{l -a)] =kt where k is the rate constant for phase separation.
A first order rate equation was applied since it seemed to best fit the data when compared to other rate equations.
Phase separation rates were determined from average ofn=3 Tg values at every time point.      those were stored at 40°C/69%RH for 2 days. Phase separation of griseofulvin is seen from the shift in T 8 from I 13°C to I 62°C. In addition, the surfactant -Pluronic F87 crystallizes completely as indicated by the melting at 70°C and the shi ft in Tg of PVP to 162°C. The approach described in this study was fo und useful to identify the physical stabi lity of this product at 40°C/69%RH conditions.   where a the fraction of drug that has crystallized, t is the storage time, B is a constant and m is a constant relating to the mechanism of griseofulvin crystallization.
The kinetic equations for the most common mechanisms that are believed to operate in solid-state alteration are presented in Table B-IV-lll, chapter 4.
In order to find the equation that best fits the crystallization rate, the equation that has the m value closest to the slope obtained in the Hancock-Sharp fit was chosen, which is KJMA rate equation (m = I). To counter-verify, the crystallization data obtained is fit to all the equations listed in Table B-IV-III, chapter 4.      .. c.