Characterization of amorphous API:Polymer mixtures using X-ray powder diffraction

$: Characterization of amorphous API:Polymer mixtures using X-ray powder diffraction$
Characterization of Amorphous API:Polymer MixturesUsing X-Ray Powder Diffraction

ANN NEWMAN,1 DAVID ENGERS,1 SIMON BATES,1 IGOR IVANISEVIC,1 RON C. KELLY,1

GEORGE ZOGRAFI2

1SSCI, An Aptuit Company, West Lafayette, Indiana

2University of Wisconsin-Madison, Madison, Wisconsin

Received 8 November 2007; accepted 16 January 2008

Published online 19 March 2008 in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/jps.21352

Ron C. Kellycisco, CA.

Corresponde0112; Fax: 765-

Journal of Pharm

� 2008 Wiley-Liss

4840 JOURN

ABSTRACT: Recognizing limitations with the standard method of determining whetheran amorphous API–polymer mixture is miscible based on the number of glass transitiontemperatures (Tg) using differential scanning calorimetry (DSC) measurements, wehave developed an X-ray powder diffraction (XRPD) method coupled with computation ofpair distribution functions (PDF), to more fully assess miscibility in such systems. Themixtures chosen were: dextran–poly(vinylpyrrolidone) (PVP) and trehalose–dextran,both prepared by lyophilization; and indomethacin–PVP, prepared by evaporation fromorganic solvent. Immiscibility is detected when the PDF profiles of each individualcomponent taken in proportion to their compositions in the mixture agree with the PDFof the mixture, indicating phase separation into independent amorphous phases. A lackof agreement of the PDF profiles indicates that the mixture with a unique PDF ismiscible. In agreement with DSC measurements that detected two independent Tg

values for the dextran–PVP mixture, the PDF profiles of the mixture matched very wellindicating a phase separated system. From the PDF analysis, indomethacin–PVP wasshown to be completely miscible in agreement with the single Tg value measured for themixture. In the case of the trehalose–dextran mixture, where only one Tg value wasdetected, however, PDF analysis clearly revealed phase separation. Since DSC can notdetect two Tg values when phase separation produces amorphous domains with sizesless than approximately 30 nm, it is concluded that the trehalose–dextran system is aphase separated mixture with a structure equivalent to a solid nanosuspension havingnanosize domains. Such systems would be expected to have properties intermediate tothose observed for miscible and macroscopically phase separated amorphous disper-sions. However, since phase separation has occurred, the solid nanosuspensions wouldbe expected to exhibit a greater tendency for physical instability under a given stress,that is, crystallization, than would a miscible system. � 2008 Wiley-Liss, Inc. and the

American Pharmacists Association J Pharm Sci 97:4840–4856, 2008

Keywords: amorphous; X-ray powder
diffractometry; calorimetry (DSC); polymers;solid dispersion
’s present address is Amgen, South San Fran-

nce to: Ann Newman (Telephone: 765-463-463-2497; E-mail: [email protected])

aceutical Sciences, Vol. 97, 4840–4856 (2008)

, Inc. and the American Pharmacists Association

AL OF PHARMACEUTICAL SCIENCES, VOL. 97, NO. 11, NO

INTRODUCTION

In recent years it has become apparent that in theearly stages of the drug development process, aspotential new drug candidates are being evalu-ated, more and more of those compounds showing

VEMBER 2008

CHARACTERIZATION OF AMORPHOUS API:POLYMER MIXTURES 4841

in vitro efficacy through high affinity preferentialbinding to desired receptors and biomarkers andthe potential for desirable in vivo performance,tend to exhibit very poor aqueous solubility. Suchbehavior is believed to arise because higheraffinities of API from aqueous solution forbiomarkers and receptors occur more readilywhen molecular weights of the small moleculeAPIs are increased and molecules are renderedmore hydrophobic. In increasing the molecularweight of these molecules we can also expect thatthey, as crystals, will exhibit stronger intermole-cular interactions, higher melting points andlower solubilities as a result of increased hydrogenbonding and stronger lattice energies. If dissolu-tion rate is the rate-limiting step in oral absorp-tion, we would expect such reduced solubility tolead to reduced oral bioavailability, and indeed,this is very often the case.1

Numerous strategies have been used to enhancedissolution rates through processing and formula-tion, for example, reducing particle size or usingdisintegrants, surfactants and solubilizers in thedrug product. Increases in the intrinsic aqueoussolubility of active pharmaceutical ingredients(API) can be accomplished through salt forma-tion2 if the compound contains ionizable groups, oras has been demonstrated more recently, byforming water-soluble cocrystals3 and amorphousforms.4 This article addresses the situation whereamorphous solid forms of the API are desirable toattain greater dissolution rates, with the negativeeffect that amorphous forms are metastablerelative to the crystalline state with a thermo-dynamic tendency to recrystallize, thus offsettingthe potential beneficial effects of the greateraqueous solubility.

To counter this tendency for recrystallization,the API can be processed with certain excipients,specifically polymers, to form what are describedas ‘‘miscible’’ solid dispersions, as opposed tophase-separated mixtures of two or more compo-nents.5 Such dispersions, usually have one glasstransition temperature intermediate to those ofthe polymer and API, and appear to inhibitnucleation and crystal growth through preferen-tial API–polymer interactions6 and possible stericeffects.7 Thus, the polymer maintains the API inthe more soluble amorphous state, while, becauseof its own water solubility, it does not interferewith the solubility of the API. Useful pharma-ceutically acceptable polymers that can formmiscible amorphous dispersions with the APIinclude polyvinylpyrrolidone (PVP), polyvinylpyr-

DOI 10.1002/jps JOURNA

rolidone vinyl acetate (PVP/VA), polyethyleneglycol (PEG), various hydroxyproylmethyl cellu-lose (HPMC) derivatives, for example, HPMC,HPMC-acetate succinate, or HPMC-phthalate,and polymethacrylate derivatives. Processes usedto prepare amorphous solid dispersions includehot melt extrusion, spray drying, freeze dryingand cryogrinding. It should be mentioned that theformation of the amorphous dispersion is essentialfor inhibiting crystallization. It, however, ispossible that such polymers, as well as surfac-tants, phase-separated as solids from the API,could still enhance aqueous dissolution ratesthrough enhanced wetting of the solid system inthe dissolution medium.

In discussing amorphous dispersions it isimportant to define such systems in as funda-mental a manner as possible. When we speak of‘‘miscibility’’ we are not generally speaking of anAPI–polymer mixture in which the API exhibitsequilibrium solubility in the polymer. Rather, weshould think of miscible systems generally asrepresenting a single supersaturated metastablephase of API and polymer, where the componentsare able to mutually influence the solid structureof the other phase, molecular mobility andintrinsic properties, for example, tendency torecrystallize or to undergo chemical instability,over practical storage times. Important thermo-dynamic criteria for the formation of metastableamorphous molecular dispersions, include asufficiently positive combinatorial entropy plusintermolecular interactions, for example, hydro-gen bonding, between each component thatcan offset interactions between the individualcomponents.8

Currently, the major means of experimentallydetermining whether or not a miscible amorphousAPI–polymer dispersion has been produced is touse differential scanning calorimetery, DSC, tomeasure any glass transition temperature, Tg,associated with the amorphous phase. If thesystem is completely miscible it will exhibit onlyone Tg, intermediate to the Tg values of theindividual components. If the system is fullyphase separated into the individual amorphouscomponents, two Tg values, those of the individualcomponents in a binary mixture, will be detected.If the components are partially miscible, two Tg

values will still be observed, and they will beintermediate between the Tg values of the pureindividual components. To observe Tg by DSC,however, the individual values generally must beabout 108C apart and the size of the phase

L OF PHARMACEUTICAL SCIENCES, VOL. 97, NO. 11, NOVEMBER 2008

4842 NEWMAN ET AL.

separated domains must be large enough to bedetected by the thermal changes measured byDSC, that is, greater than �30 nm.9 Therefore, itis possible that an API–polymer dispersion mayexhibit only one Tg, while the system is actuallyphase separated with individual amorphousphases having domain sizes that are less thanthat detectable by DSC measurements. Becausethe size of the domains is less than �30 nm, such asystem can be considered to be a solid nanosus-pension. Characterization of a solid nanosuspen-sion has been reported with agglomerates ofindividual particles exhibiting sizes smaller than5–10 nm.10 A further indication of the limitationsof DSC in correctly establishing miscibility orimmiscibility in certain situations is a recentreport of a blend of poly(oxyethylene) andpoly(methyl methacrylate) that exhibits two Tg

values, those of the individual components, yet byother criteria, such as small angle neutronscattering, are found to be miscible.11 It is criticalthat samples and data are analyzed usingtechniques which will readily determine whethera phase-separated mixture or miscible dispersionhas been produced, since physical and chemicalstability can be markedly different for the twotypes of systems.

In view of the inability of DSC alone to alwaysdetermine the miscibility or immiscibility ofamorphous dispersions, we show in this articlehow powder X-ray diffraction (XRPD) measure-ments combined with computational approachescan provide additional insight to assess thephysical state of amorphous solid dispersions.We do this by studying three systems represent-ing a miscible, phase separated, and solidnanosuspension of API and polymer, respectively,and subjecting each system to a computationalanalysis of measured XRPD patterns using XPRDdata of the individual components.

EXPERIMENTAL

Materials

Indomethacin, polyvinylpyrrolidone K90, anddextran (Mn: 64–76 kDa) were obtained fromSigma (St. Louis, MO). D(þ)-trehalose dihydratewas obtained from Fluka (Steinheim, GER).Solvents used to prepare dispersions, namelywater and dichloromethane were obtained fromMallinckrodt (Phillipsburg, NJ) and EMD (Gibbs-town, NJ), respectively, and used as received.

JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 97, NO. 11, NOVEMBER 20

Preparation of Amorphous Samplesby Freeze-Drying

Amorphous trehalose, dextran, and PVP wereprepared from aqueous solutions containing �10wt% solids. The vials containing solution werecapped and stirred at �18 rpm overnight in aheated orbital shaker block maintained at �358C.Solutions were dispensed into a 20 mL glass vial(�1–3 mL/vial) and into a custom vial (6-mm ID)with an aluminum foil backing film (�100 mL/vial). Small samples prepared in this manner wererecovered from the container after drying forXRPD with the samples held intact in transmis-sion (see below) to provide the best representationof a sample ‘as prepared.’ Solutions were subse-quently frozen by immersing the filled vials into abath of liquid nitrogen. Drying was completedunder vacuum for 48 h, with the samplesmaintained as cold as possible during the first12 h using an isopropanol-dry ice bath; dryingafter 12 h was complete at ambient temperature.Dried samples were stored at �208C over desic-cant. Mixtures of trehalose in dextran and PVP indextran were prepared at known compositionsusing the procedure described above.

Preparation of Indomethacin–PVP Samples

Samples of indomethacin in PVP were prepared atknown compositions by flash evaporation fromdichloromethane. Solids were dissolved in dichlor-omethane with stirring. The resulting solutionwas filtered through a 0.2 mm nylon filter into a50 mL round bottom flask. The solution was flashevaporated under vacuum with the flask rotatedin a water bath maintained at 508C. Secondarydrying of the resulting solids was performedovernight in a vacuum oven maintained at408C. The solids were yellow in color with a flakemorphology, as described previously.6 Driedsamples were stored at �208C over desiccant.

Preparation of Amorphous Indomethacinby Quench-Cooling

Amorphous indomethacin was prepared by pla-cing crystalline gamma indomethacin in a glassvial immersed in an oil bath heated to maintainthe molten liquid at �1658C. The molten liquidwas poured into a bath of liquid nitrogen toquench-cool and then ground to a fine powderusing a mortar and pestle. The bulk amorphousindomethacin was stored over desiccant at �208Cuntil used.

08 DOI 10.1002/jps


Methods

X-Ray Powder Diffraction

XRPD analyses were performed for trehalose–dextran and PVP–dextran dispersions, and theirindividual components, using an Inel XRG-3000diffractometer equipped with a CPS (CurvedPosition Sensitive) detector with a 2u range of1208. Real time data were collected after a heliumpurge (e.g., to minimize instrumental backgrounddue to air scatter) using Cu K( radiation at aresolution of 0.038 2u. The tube voltage andamperage were set to 40 kV and 30 mA,respectively. The monochromator slit was set at3 mm� 160 mm. Samples were removed intactfrom the custom sample vials; the net weight andapproximate thickness of each sample was accu-rately recorded. Samples were mounted on the endof a copper pin positioned on the goniometer fordata acquisition in transmission; no other samplesupport was employed. This approach was used toexamine samples ‘as prepared’ and to avoid theneed to mechanically alter the specimen for XRPDanalysis. Powdering of freeze-dried samples iscommonly performed when preparing samples foranalysis in sample holders for a Bragg–Bretano(reflection) geometry. In addition to the disruptionof the microstructure generated during processing,the resulting powder is highly electrostatic and isdifficult to properly prepare ‘full and level’ in theXRPD holder. The data acquisition time for eachsample and all blanks (e.g., helium purged)examined using the Inel XRG-3000 was limited to15 min. Instrument calibration was performedusing a silicon reference standard.

XRPD analyses were performed for samples ofindomethacin–PVP and the individual compo-nents using a Shimadzu XRD-6000 X-ray powderdiffractometer (Kyoto, Japan) with Cu Ka radia-tion. The Shimadzu instrument is equipped with along fine focus X-ray tube. The tube voltage andamperage were set to 40 kV and 40 mA,respectively. The divergence and scattering slitswere set at 18 and the receiving slit was set at0.15 mm. Diffracted radiation was detected by aNaI scintillation detector. A u–2u continuous scanat 1.28/min from 2.58 to 608 2u was used with aneffective 0.04 step size (2 s/step). The sample wasspun at 30 rpm. The analyses were performed atambient temperature. A silicon standard wasanalyzed to check the instrument alignment. Datawere collected and analyzed using XRD-6100/7000 v. 5.0. Samples were prepared for analysis byplacing them in an aluminum reflection sample


holder with low background silicon inserts. Thenet weight of the sample was recorded in eachcase. The dimensions of the sample well areapproximately 10 mm in diameter and 2 mmin depth.

Powder samples were prepared for analysis bypacking them into an aluminum sample holderwith low background silicon inserts, which wererequired for data acquisition of powdered testspecimens in reflection geometry rather than thepreferred use of helium as a purge gas. Furthermechanical manipulation (e.g., compression of thepowders into thin wafers) to make samplessuitable for data acquisition in transmissionwas not performed.

Differential Scanning Calorimetry

Modulated differential scanning calorimetry(mDSC) was performed using a TA InstrumentsModel 2920 calorimeter (New Castle, DE). Thesample was placed into an aluminum pan, and theweight accurately recorded. The pan was coveredwith a lid and crimped. The analysis wasperformed under a dry nitrogen purge. Sampleswere ramped to approximately 508C below theexpected glass transition temperature (Tg) andheated at 18C/min to approximately 508C abovethe highest expected Tg using modulations of 0.88every 60 s. The measured Tg is reported as theonset of the step change in the reversing heat flowsignal. Indium metal and sapphire were used asthe calibration standard.

Computational Studies

Pair Distribution Function (PDF)

One of the most generally useful tools for totaldiffraction analysis is the pair distribution func-tion or PDF. The PDF provides a fingerprint of theinteratomic distances that define a particularsolid form12 and is very useful for determiningrelationships between ordered and disorderedsystems. The PDF is generally presented asprobability against distance and represents theweighted probability of finding two atoms sepa-rated by a distance r.13 The maxima in the PDFcorrespond to commonly occurring interatomicdistances, where the product of area underthe maximum and distance gives the numberof atom units with that specific pair separation.A PDF trace is a one-dimensional representa-tion of a radially averaged three-dimensional


4844 NEWMAN ET AL.

structure.14–18 Additional information on the PDFcan be found elsewhere.19–22 PDF analysesperformed in this study used software developedinternally based on published equations.19

The PDF, g(r), is calculated using the relation-ships defined in Eqs. (1) and (2)

gðrÞ ¼ 4pr2½rðrÞ � r0 (1)

where r0 is the average atom number density ofthe structure and r(r) the atom pair density forX-ray scattering, given by Peterson21

rðrÞ ¼ 1

4pr2

� �X fpfq

fh ij j2

!dðr � rpqÞ (2)

Here fp,q are the individual atomic form factorsand hfi represents a mean atomic form factor forthe structure. The distance rpq represents theatom pair separation and r(r) represents theprobability of finding an atom pair separated bythe distance, r, weighted by the atomic formfactors and averaged over all pairs in thestructure. The pair sum must be performed overa radial distance large enough to give the full atompair relationships out to the distance of interest.

Methodology

The computational studies described belowrequire a measured XRPD pattern for theamorphous API, the polymer used in the disper-sion (e.g., PVP), as well as the solid dispersion. Forbest results, the dispersion and the amorphousreference materials should be prepared using thesame process, and the XRPD data collected underthe same conditions and ideally on the samediffractometer. Blanks should be collected whereinstrumental background cannot be easily mod-eled computationally; however, in all cases,attempts should be made to minimize backgroundwhenever possible (e.g., through the aforemen-tioned use of a helium purge). Backgroundreduction and accurate accounting of the instru-mental background are critical when attemptingto derive PDF from laboratory powder patterns ofX-ray amorphous samples. As a minimumrequirement, data should be collected from back-ground at low angles (i.e., angles below where theprimary diffraction intensity from the samplestarts) to background at high angles (i.e., angleshigher than the highest angle significant diffrac-tion event observed). For the types of materialspresented in this document, this minimumrequirement corresponds to an angular measure-ment range of approximately 2–608 2u (for Cu Ka


X-rays). If the instrument geometry allows,collecting data out to angles greater than 608 willimprove the quality of the PDF.

Once the PDF has been derived from themeasured XRPD patterns for the individualcomponents (API and polymer) and the soliddispersion, analysis proceeds by modeling thePDF from the dispersion samples as a linearcombination of the PDFs for the individualcomponents. In this approach, the PDFs derivedfrom the individual components are treated asreference patterns and the PDF from the soliddispersion as the unknown. A fitting algorithmsuch as the Brent method23 is used to obtain thebest fit ratio of the reference PDFs, with respect tothe dispersion PDF by independently scaling thereference PDFs. As the PDF traces contain bothpositive and negative points, the chosen errorfunction for the fitting process is the sum of squareddifferences in PDF probabilities at each distance r.

The PDF derived from the API and polymer X-rayamorphous patterns represent the characteristicinter- and intramolecular distances found in eachmaterial. Therefore, if the PDF of the dispersion canbe described as a linear combination of the PDFs forthe individual components, then the packingpatterns found in the dispersion can be describedby the characteristic distances found in the API andpolymer, respectively. The ability to describethe PDF of the dispersion is clear evidence thatthe dispersion has a packing pattern similar to theindividual components, which infers that thedispersion is at least partially phase separated atthe molecular level. In addition to the ‘goodness offit’ argument, the individual scale factor required fora reference PDF can be considered to be equivalentto a relative weight percentage, as the PDF isessentially a quantitative measure of the electrondensity distribution within a material. So inaddition to being able to describe the PDF of adispersion by linear combinations using the PDFs ofthe individual components, the required scalefactors are expected to correlate to the relativeweight percentages of the API and polymer presentin the dispersion for a phase separated system.Otherwise, if the best fit is poor or the scale factorsfor the PDFs of the individual components are verydifferent from known weight percentages, thedispersion may be considered miscible—the packingpattern of the dispersion is different from thatpresented in the API and the polymer only. Formiscible dispersions, the residual differencebetween the calculated PDF and the PDF derivedfrom measured XRPD data can be used to represent

08 DOI 10.1002/jps


the deviation in the packing pattern from a phaseseparate system, and therefore the greater inter-molecular interaction between the API and polymer.

It should be noted that, while the entire powderpattern is used to calculate the PDF out to a fewhundred angstroms, only a subset of the PDF isused in the fitting procedure. This is done to avoidfitting to artifacts in the PDF (e.g., terminationripples) which can be significant for PDFs derivedfrom X-ray amorphous powder patterns. Thefitting cutoff should be selected for each systembased on the last clearly observed feature (maxi-mum) in the PDF, typically found around 20 A formost X-ray amorphous pharmaceuticals.

Before presenting the results, it should berecognized that a similar analysis can be usedfor directly modeling measured XRPD patterns ofthe dispersions using linear combinations of themeasured XRPD patterns for the individualcomponents. This approach will typically yieldlower random errors than the PDF method whendirectly modeling the XRPD patterns is applic-able. Although directly modeling the powderpatterns is straightforward to implement andcan give lower errors, the approach is less robustthan the PDF modeling approach. Because it isrelatively common that the X-ray amorphouspatterns of the individual components (API andpolymer) are quite similar, the least squaresmodeling procedure is almost always able toachieve a good fit. The PDFs on the other hand,tend to be sufficiently different to allow a morerobust fitting procedure. In addition, the indivi-dual scale factors used to achieve the best fit whenusing XRPD data cannot be related to weightpercent of the components present in the disper-sion unless an appropriate normalization/calibra-tion scheme is adopted. This removes one of thekey success criteria from the method as thecalculated weight percent will be dependent onthe normalization scheme as well as the indivi-dual scale factors. In the course of calculatingthe PDF, the data reduction procedure provides theXRPD pattern in electron units or on an absolutescale. This gives a rational and unbiased approachfor determining the weight percent values.

RESULTS

Dextran–PVP

The first system investigated was dextran–PVPdispersions prepared by freeze-drying. The indivi-


dual components were analyzed initially to form abasis for comparison with the dispersion samples.Dextran used in this study (Mn: 64–76 kDa)exhibited a Tg onset value of 2218C as measuredby mDSC. A Tg value of 2118C has been reported fordextran T10 (average molecular weight 9300) and2258C has been reported for dextran T500 (averagemolecular weight 520,000),8 therefore the valueobtained in this study is in agreement with previousreports. PVP was used as received and exhibited aTg onset of approximately 1778C, which is alsosimilar to literature values.24

A number of compositions, ranging from 20 to80 wt% dextran in PVP, were prepared by freeze-drying aqueous solutions. The DSC thermogramsfor two of these samples, 30 and 70 wt% dextran inPVP, are shown in Figure 1. The lyophilizedsamples exhibit two distinct glass transitiontemperatures, indicating that these are phaseseparated even after lyophilization. Similar datawere obtained for the other ratios. Previousstudies on lyophilized dextran–PVP also showedtwo distinct glass transition temperatures, asexpected for most polymer blends because ofreduced combinatorial entropy that leads toimmiscibility.8 In this study and in previouswork, the Tg values reported for the mixtureswere within 2–38C of those observed for theindividual components, suggesting essentiallycomplete phase separation.

The XRPD data confirm that the samples areX-ray amorphous, as shown in Figure 2. However,the shape of the patterns for each dispersiondisplays significant differences between the sam-ples. Lyophilized dextran exhibits a broad halowith the position of the maximum at approxi-mately 178 2u, while lyophilized PVP exhibits twonarrower maxima at approximately 118 and198 2u. The mixtures containing less PVP showa significant decrease in the size of the maximumat 118 2u when compared to the PVP pattern. Thisindicates that the amorphous halos are directlyrelated to the composition of the individualcomponents and supports the thermal data thatthe samples are phase separated.

Computational studies were performed on thissystem using linear combinations of XRPD or PDFdata. The correlation plots given here are derivedfrom the PDF data; similar results were obtainedusing the XRPD data. When a system is phaseseparated, the PDF for the dispersion samplesderived from measured XRPD data can bedescribed by a linear combination of the PDFsfor the individual components.


Figure 1. Overlay of modulated DSC, reversing heat flow, from top to bottom: PVP, 70and 30 wt% dextran in PVP dispersions, and dextran (Mn: 64–76 kDa) prepared byfreeze-drying.

4846 NEWMAN ET AL.

Furthermore, the scale factors required for thestarting materials to achieve the best fit to thedispersion sample are used to provide a calculat-ed estimate of the weight percents of PVP anddextran present in the sample. These calculatedweight percents can be compared to theoreticalweight percents as final confirmation that thesystem diffracts X-rays as a simple phase sepa-rated mixture. A plot of calculated versustheoretical composition for dextran and PVP isgiven in Figure 3. An ideal correlation, where thecalculated and theoretical results are equal, is

Figure 2. Overlay of XRPD patterns (as measured)from top to bottom: PVP, 20, 30, 40, 50, 60, 70, 80 wt%dextran in PVP dispersions, and dextran (Mn: 64–76kDa) prepared by freeze-drying.


represented by the solid line. The lower panel foreach component represents the calculated totalweight percent of PVP plus dextran in eachdispersion as determined from the analysis. Asthe PVP and dextran reference PDF data wereindependently scaled, there is no constraintenforced that their combined weight percentshould be equal to 100. Minimal deviations fromthe target value of 100% for the combined weightpercentages are a good indication that the PDFs ofthe individual components provide a good descrip-tion of the packing patterns found in the disper-sion, and provide support for the presence ofimmiscibility.

It is evident from Figure 3 that the calculatedweight percent of PVP follows closely with theexpected values for a phase separated system,although there is some random scatter in the datapoints. Figure 3 on the other hand shows that thecalculated dextran weight percent consistentlyfalls below the expected composition. This isreflected in the calculated combined weightpercent which falls to approximately 90% forthe samples containing more dextran. Althoughthe calculated dextran weight percent fallsbelow the predicted composition, the slope of thecalculated values line follows very closely withthe expected trend line. This indicates thatthe phase separated mixture model is validfor the phase separated dispersion system but

08 DOI 10.1002/jps

Figure 3. Correlation of calculated versus theoretical weight percent of PVP (PVP; toppanel) and dextran (DEX; top panel) in dextran–PVP dispersion samples. The solid linerepresents a direct correlation of the calculated and theoretical values. The total %(lower panel) is the sum of the calculated best fit result for each component. The twoopen circle markers denote control measurements performed using a second powderdiffractometer (reflection geometry).


that the dextran is scattering more efficiently inthe dispersion matrix than in the pure form.Differences in scattering efficiency observed in amixture matrix rather than in the pure form canbe described in terms of differences in packingdensity and a relatively large granular/domainstructure in the phase separated dispersion.However, it should be noted that such micro-absorption like phenomena are generally verysmall unless significant differences in electrondensity exist between the different domains/grains in a mixture, and this point is still beinginvestigated.

Figure 4 shows the PDF trace derived for the50 wt% dextran in PVP dispersion sample(measured) compared with the (calculated) bestfit achieved using linear combinations of the PDFsfor the individual components, dextran and PVP.The difference between the calculated and mea-


sured PDF, Dg(r), is displayed in the lower paneland shows low amplitude, random fluctuationsover the region of interest, 4–20 A

´. This suggests

that the measured PDF is well described by thecalculated PDF determined from the referencematerials. It can be assumed the value of Dg(r)represents differences in the local molecularpacking distances which infers the existence ofadditional molecular interactions in the dextran–PVP system not described by the API–API pluspolymer–polymer model. The residual differencesshould then show larger fluctuations and amarked intensity decrease on going from nearestneighbor to next nearest neighbor distances. Thisobservation provides strong support that dex-tran–PVP samples are phase separated mixturesand not miscible dispersions.

To investigate differences in instrument geo-metry, two control measurements were performed


Figure 4. PDF analysis for 50 wt% dextran in PVPdispersion with the circles representing measured data,and the solid line representing the calculated data. Theresiduals (lower panel) greater than 4 A are small andrandom indicating that the measured PDF is welldescribed by the PDF calculated from the two referencematerials as expected for a phase separated mixture.

4848 NEWMAN ET AL.

for the dextran–PVP system using 30 and 70 wt%dextran in PVP compositions. The results from thePDF analysis of these control samples are plottedin Figure 3 along with the results from thetransmission geometry. The control measure-ments lie within the trend observed for thetransmission measurements.

It is important to note that even though bothcomponents initially were dissolved in solution,separation of the components occurred duringprocessing, in this case lyophilization, thusshowing the lack of solid-state affinity for thetwo polymers. A similar phenomenon would beexpected to occur with other methods of proces-sing, such as spray-drying.

Indomethacin–PVP

The second system investigated was indometha-cin–PVP dispersions prepared by rotary evapora-tion from dichloromethane solutions. Thereference materials for amorphous indomethacinwas produced in this study by a melt/quench,which exhibited a Tg onset of approximately 438C,similar to values previously reported.24–26 PVP


after processing (e.g., evaporation from dichlor-omethane) resulted in a Tg onset of approximately1778C, which is also similar to literaturevalues.24,25

Various dispersions ranging from 20 to 80 wt%indomethacin in PVP were prepared. The samplecontaining 30 wt% PVP exhibited a Tg onset ofapproximately 708C, and the sample containing70 wt% PVP exhibited a significantly higher Tg

onset of 1168C, as shown in Figure 5. The Tg

values for this system have been reported to varyfrom approximately 558C with 5 wt% polymer to1608C with 90 wt% polymer,25 and our Tg valuesare very similar. As stated earlier, the presence ofone Tg is generally attributed to the componentsbeing miscible over the entire range of composi-tions, with the Tg varying with concentration, andthis has been reported for indomethacin–PVPdispersions.25 It also has been shown using IRspectroscopy that indomethacin–PVP dispersionsmade from solution result in disruption of theindomethacin dimer found in the gamma crystal-line form. This disruption is another indication ofa miscible dispersion, whereas IR measurementsof phase separated mixtures of amorphous indo-methacin and PVP show that the dimer is intact.6

Examining this system more closely withXRPD, as shown in Figure 6, indicates that thesamples are X-ray amorphous; however, the shapeof the interference patterns is found to be differentfrom that displayed for the individual compo-nents, indomethacin and PVP. One broad halo isobserved at approximately 218 2u for the disper-sions, which is shifted compared to PVP processedin the same manner. The lower maximum atapproximately 118 2u is significantly reduced andbroader for the indomethacin–PVP samples anddoes not appear to directly track with the amountof polymer in the sample. These observationsindicate that the local packing arrangement, andtherefore the interactions, in the dispersion isdifferent than that found in the referencematerials, which is in agreement with the thermaldata that suggest that a miscible dispersion hasbeen formed.

Linear combinations of XRPD and PDF datawere also used to analyze the miscibility of thissystem. A plot of calculated versus theoreticalpercentages for indomethacin and PVP is given inFigure 7 using the PDF data; similar results wereobtained using the XRPD data. As in Figure 3 forthe dextran–PVP system, the expected correlationis represented by the solid line and variation of thetotal calculated weight percent from the target of

08 DOI 10.1002/jps

Figure 5. Overlay of modulated DSC, reversing heat flow, from top to bottom:indomethacin prepared by melt quench, 70 and 30 wt% indomethacin in PVP dispersionsand PVP prepared by flash evaporation.


100 is presented in the lower panel for eachcomponent. In Figure 7, the calculated indo-methacin weight percent is seen to be significantlyhigher than the expected composition, althoughthe general slope of the calculated weight percentdoes follow the expected trend line. The calculatedPVP weight percent shown in Figure 7 issignificantly scattered and the general trend isfar from the slope of the expected trend line. Thesignificant deviation of the calculated weightpercent is reflected in the values for the total

Figure 6. Overlay of XRPD patterns (as measured)from top to bottom: indomethacin prepared by meltquench, 80, 70, 60, 50, 40, 30, 20 wt% indomethacinin PVP and PVP prepared by flash evaporation.


weight percent that range from approximately80% up to 170%. Such large deviations away fromthe expected behavior for a phase separatedsystem supports the previous observations thatindomethacin–PVP form a miscible dispersion;that is, the packing patterns of the individualcomponents do not represent those displayed inthe PDF analysis of the dispersion. It is interest-ing to note that the largest deviations away fromthe phase separated mixture model are seen forthose samples with higher PVP content.

The powder data used for the miscibilityanalysis of indomethacin and PVP was measuredusing a reflection geometry diffractometer over alimited measurement range. The transmissiongeometry was not used for this sample as it was aloose powder. Mounting loose powders for trans-mission measurements requires the use of acapillary or polymer film sandwich that willcontribute additional background to the measure-ment. The reflection measurements over areduced angular range are less ideal for PDFanalysis than the transmission measurementsacquired from the other systems. To verify thatthe results obtained for indomethacin–PVP werenot a result of the differences in instrumentgeometry, the two control measurements per-formed for the dextran–PVP system using 30 and70 wt% dextran in PVP compositions were


Figure 7. Correlation of calculated versus theoretical weight percent of indomethacin(IMC; top panel) and PVP (PVP; top panel) in indomethacin–PVP dispersion samples.The solid line represents a direct correlation of the calculated and theoretical values. Thetotal % (lower panel) is the sum of the calculated best fit result for each component.

4850 NEWMAN ET AL.

collected, as previously discussed in Dextran–PVPSection. The control measurements lie within thetrend observed for the transmission measure-ments indicating that the change in instrumentgeometry is not responsible for the indomethacin–PVP results.

Figure 8 shows the PDF trace derived for the50 wt% indomethacin in PVP dispersion sample(measured) compared with the (calculated) best fitachieved using linear combinations of the PDFsfor the individual components, indomethacin andPVP. Again, the difference between the calculatedand measured PDF, Dg(r), is displayed in thelower panel where the expected decay in theamplitude of the PDF oscillations is observed,with a much larger residual difference from 4 to20 A

´. The dotted lines following the amplitude

decay are predicted curves for a highly disorderedsystem.27 The larger residual difference suggeststhat the measured PDF is poorly described by thecalculated PDF determined from the reference


materials and therefore suggests that the mole-cular packing patterns present in one or both purephases are different from that found in theprocessed dispersion. In this view, the profile ofDg(r) may be taken as evidence of this difference,which is corroborated by the IR measurementsshowing disruption of the indomethacin dimer inindomethacin:PVP miscible dispersions.6

Trehalose–Dextran

The third system studied was trehalose–dextranproduced by lyophilization of aqueous solutionscontaining both components. The Tg values of119.68C, obtained for trehalose, and 225.68C fordextran, are in good agreement with valuesreported previously for these materials.28

Dispersion samples ranging from 20 and 80 wt%trehalose in dextran were prepared. DSC datacollected on selected freshly prepared materials

08 DOI 10.1002/jps

Figure 8. PDF analysis for 50 wt% indomethacin inPVP dispersion with the circles representing measureddata, and the solid line representing the calculateddata. The residuals (lower panel) greater than 4 Aare not random and relatively large with a systematicdecay indicating that the measured PDF is poorlydescribed by the PDF calculated from the two referencematerials.

Figure 9. Overlay of modulated DSC, revamorphous trehalose, 70 and 30 wt% trehal(Mn: 64–76 kDa) prepared by freeze drying.



indicated a single Tg value, as shown in Figure 9for samples containing 30 and 70 wt% trehalose.These samples displayed onset Tg values ofapproximately 180 and 1368C, respectively. Pre-vious studies with this system also have shownsingle initial Tg values in good agreement with thevalues reported above.28 It is interesting to notethat in the present study the values of Tg changedwith concentration, as expected of a miscibledispersion, and that calculation of the predictedTg at these two concentrations from the individualTg values of trehalose and dextran using the Foxequation is in close agreement with the measuredvalues as expected for a miscible dispersion.29

However, in the earlier study with this system,28

two Tg values were observed upon stressing withrelative humidity, with samples eventually exhi-biting crystallization of trehalose. Thus, it wasinteresting to subject this system to the computa-tional analysis described above.

The XRPD data shown in Figure 10 confirm thatthe samples of trehalose–dextran are X-rayamorphous. As observed for the previous systems,the profile for each pattern is different. For theamorphous trehalose, the main maximum isobserved at approximately 198 2u with an obviousshoulder at approximately 258 2u. Dextran exhi-bits a less broad halo at approximately 178 2u withno obvious shoulders. The dispersion compositions

ersing heat flow, from top to bottom:ose in dextran dispersions and dextran


Figure 10. Overlay of XRPD patterns (as measured)from top to bottom: trehalose, 80, 70, 60, 50, 40, 30,20 wt% trehalose in dextran dispersions, and dextran(Mn: 64–76 kDa) prepared by freeze drying.

4852 NEWMAN ET AL.

containing more trehalose more closely resemblesthe pattern of amorphous trehalose and thesamples containing more dextran exhibits a halomore similar to that of pure dextran. Based onvisual inspection of the shapes of the amorphoushaloes and positions of the maxima, it wouldappear that the data are directly correlated to theindividual components. XRPD figures were notincluded in the previous report, therefore nocomparisons can be made in this regard betweenthe two studies.28

A plot of calculated versus theoretical composi-tion for trehalose and dextran is given inFigure 11 using the PDF data; similar resultswere obtained using the XRPD data. The expectedcorrelation is again represented by the solid lineand the variation from the target 100 wt%is presented in the lower panel. The calculatedweigh percent for both trehalose and dextranclosely follows the expected behavior of a phaseseparated system. The correlation between thecalculated and theoretical weight percent is betterfor the trehalose–dextran system than for thedextran–PVP system. This observation stands inapparent contradiction to the thermal measure-ments which show a single glass transitiontemperature for the trehalose–dextran system.The close correspondence between the totalcalculated weight percent for trehalose anddextran with the target weight percent of 100reflects the almost ideal behavior of this system asdetermined by the phase separated mixturemodel. The close to ideal behavior suggests a


much finer grain/domain structure in trehalose–dextran than that observed for dextran–PVP.

Figure 12 shows the PDF trace derived for the50 wt% trehalose in dextran dispersion comparedwith the (calculated) best fit achieved using linearcombinations of the PDFs for the individualcomponents, trehalose and dextran. The differ-ence (residual) between the calculated andmeasured PDF is displayed in the lower plot withan expanded y-axis scale. The difference betweenthe calculated and measured PDF, Dg(r), isdisplayed in the lower panel and, as was observedfor dextran–PVP, shows low amplitude, randomfluctuations from 4 to 20 A

´. Again, this suggests

that the measured PDF is well described by thecalculated PDF determined from the referencematerials. This analysis provides support that thetrehalose–dextran samples are phase separatedmixtures and not miscible dispersions. Based onwork conducted in our laboratory on a variety ofamorphous dispersion systems using diffraction,thermal, and computational analyses, we havecalled these systems solid nanosuspensions.

DISCUSSION

The results of this study, as summarized inTable 1, demonstrate the need to support theuse of DSC analysis to ascertain whether or not amixture of amorphous components is miscible orimmiscible by using a technique that moredirectly characterizes the structural features ofsuch a system. Since thermal analysis and themeasurement of Tg for polymers has beentypically described as a measure of the diffusionalrelaxation kinetics of an amorphous molecule, itcan not distinguish phase separation if theindividual components have glass transitiontemperature ranges that might overlap, or if thedomain size of the individual separated compo-nents are below a critical dimension, taken to beon the order of 30 nm.9 Further, the effect ofthermal stress during relatively slow thermalanalysis measurements over a potentially broadtemperature range, as is the case during modu-lated DSC, can not be ruled out as a potentialdriving force for changes in the microstructureleading to phase separation. Through the use ofXRPD data obtained for individual componentsand mixtures, combined with computationalanalysis, we have presented an approach thatsupports the conclusions of DSC data when theyare applicable, while more importantly allowing

08 DOI 10.1002/jps

Figure 11. Correlation calculated versus theoretical percent weight of trehalose(TRE; left-top panel) and dextran (DEX; right-top panel) in trehalose–dextran dispersionsamples. The solid line represents a direct correlation of the calculated and theoreticalvalues. The total % (lower panel) is the sum of the calculated best fit result for eachcomponent.


analysis of phase-separated mixtures not detect-able by the DSC method. Recognizing thatprocessing of some materials to form dispersionsof API and polymer can provide conditions thatallow the formation of a phase separated colloidaldispersion, analogous to a lyophobic colloidaldispersion of, for example, gold particles inwater,30 it seems reasonable to expect that thebehavior observed by the trehalose–dextransystem in forming a solid nanosuspension withone detectable glass transition temperature mayoccur more often with other systems than onemight anticipate. Thus the need for the kinds ofanalyses presented herein to supplement thethermal analysis measurements that are moreroutinely performed.

To understand possible causes of such behavior,it is important to recall that the thermodynamicfactors driving the formation of an amorphousdispersion are the same as those driving themixing of molecules to form a solution. Ordinarilywith small molecules the major driving force is the


increase in combinatorial entropy or the entropyof mixing. This large entropy increase uponmixing can overwhelm any negative effects thatmight come from a positive enthalpy change dueto a lack of strong affinity between the individualcomponents, and produce a thermodynamicallystable miscible mixture. Since it is known thatpolymers upon mixing exhibit much smallerentropies of mixing due to their large molecularsizes, it is not surprising that miscible thermo-dynamically stable polymer blends occur rarely,and only when there is a very large negativeenthalpy of interaction between the polymers toprovide an overall negative free energy of mixing.Such phenomena would be expected to occur alsowhen attempting to form a solid dispersion of twopolymers, as indeed, is observed with the phaseseparation of dextran and PVP when colyophilizedfrom aqueous solution, as shown above. Formixtures of a small molecule, such as an API,with a polymer, we might expect a larger positiveentropy of mixing than observed with two


Figure 12. PDF analysis of 50 wt% trehalose in dex-tran dispersion with the circles representing measureddata and the solid line representing the calculated data.The residuals (lower panel) greater than 4 A are smalland random indicating that the measured PDF is welldescribed by the PDF calculated from the two referencematerials—a phase separated mixture.

4854 NEWMAN ET AL.

polymers, but the presence of the one polymerwould also increasingly reduce the tendency formixing with the API as the molecular weight ofthe polymer increases. Therefore, unless there is astrong tendency of the API and polymer to interactrelative to the API–API and polymer–polymerinteractions miscibility will be decreased. Sucheffects may even be more important in determin-ing whether one can maintain supersaturatedsolutions that most often are desired whenmaking a solid amorphous dispersion for phar-maceutical use. In the case of indomethacin–PVP,which by all standards appears to form a very

Table 1. Summary of Thermal and Computational Analys

Components PreparationTg ValuesObserveda

Dextran–PVP Freeze-driedfrom solution

2

Indomethacin–PVP Flash evaporatedfrom solution

1

Trehalose–dextran Freeze-driedfrom solution

1b

aNumber of Tg values observed upon analysis of the sample.bOne Tg value may split into two upon standing, storage, or stre


stable supersaturated miscible dispersion, we canconclude that the interactions between indo-methacin and PVP, as reported previously,6 offsetthe indomethacin–indomethacin and PVP–PVPinteractions, as well as any loss of entropy ofmixing due to the presence of PVP.

That trehalose forms a phase separated mixturewith dextran at first glance would not appear to belikely since chemically both molecules essentiallyconsist of glucose units that should hydrogen bondquite easily. The concentration dependence of Tg

for this system would also favor the formation of asingle phase system. The fact that subsequentstressing of the system with relative humidityproduced two Tg values, and eventual crystal-lization,28 suggests, however, a very marginalthermodynamic tendency for mixing initially,while the results of this study show that indeedthe system can be described as a phase separatedmixture from the very beginning despite havingonly one Tg. Again, we would conclude that in alllikelihood this tendency for phase separationoccurs because the trehalose–trehalose and par-ticularly the dextran–dextran interactions,coupled with a reduced entropy of mixing, offsetany tendencies for trehalose–dextran interactionas a driving force for thermodynamic mixing.Having suggested that phase separation ofnanosized amorphous domains of trehalose anddextran are formed upon lyophilization, itwould be of interest to consider why such a solidnanosuspension occurs with trehalose–dextranbut not with PVP–dextran. We first might suggestthat this tendency to form a colloidal dispersionmay arise because of a kinetic factor geared tothe nature of the lyophilization process andthe relative sizes of the two components, bothdifferently affecting the rate at which smalldomains would become frozen in as the glassforms upon freezing and drying. If this is so, theformation of such solid nanosuspensions with

es

Results Type of System

Can be modeled fromindividual components

Phase separated mixture

Cannot be modeled fromindividual components

Miscible amorphousdispersion

Can be modeled fromindividual components

Solid nanosuspension

ss.

08 DOI 10.1002/jps


small API and large polymer molecules may bevery widespread, particularly when API–API andpolymer–polymer interactions are quite strongrelative to API–polymer interactions. Futurework to investigate this may include using aPVP–drug dispersion where the drug is not aH-bond donor.

CONCLUSIONS

This study has demonstrated that it is not alwayspossible to determine whether an API–polymeramorphous mixture, processed with the intent ofproducing an amorphous molecular dispersion, ismiscible or phase separated based only on thedetermination of Tg using DSC; one Tg apparentlyindicates a single amorphous phase and two Tg

values indicate phase separation of amorphousphases. Evidence is presented, using XRPDmeasurements of individual components andthose of the solid dispersion containing API andpolymer coupled with analysis of these data withPDF, to show that a system exhibiting only one Tg

still can be completely phase separated as anamorphous mixture. Since DSC measurementscan not discriminate between separated bulkamorphous phases below a certain domain size,less than approximately 30 nm, we conclude thatwhen XRPD data analysis indicates phase separa-tion, and only one Tg is observed by DSC, theequivalent of a colloidal dispersion or solidnanosuspension exists in the amorphous state.Such systems would be expected to have proper-ties intermediate to those observed for miscibleand macroscopically phase separated amorphousdispersions.

ACKNOWLEDGMENTS

The authors gratefully acknowledge the analyti-cal support provided by SSCI AnalyticalResources, Paul Schields, Claire Gendron, andKevin Leach, for their assistance in acquiringthe data used in this analysis.

REFERENCES

1. Yalkowsky SH. 1999. Solubility and solubilizationin aqueous media. Oxford: Oxford University Press.

2. Stahl PH, Wermuth CG. 2002. Handbook of phar-maceutical salts: Properties, selection, and use.New York: Wiley-VCH.


3. McNamara DP, Childs SL, Giordono J, Iarriccio A,Cassidy J, Shet MS, Mannion R, O’Donnell E, ParkA. 2006. Use of a glutaric acid cocrystal to improveoral bioavailability of a low solubility API. PharmRes 23:1888–1897.

4. Hancock BC, Parks M. 2000. What is the truesolubility advantage for amorphous solids? PharmRes 17:397–404.

5. Serajuddin ATM. 1999. Solid dispersion of poorlywater soluble drugs: Early promises, subsequentproblems, and recent breakthroughs. J Pharm Sci88:1058–1066.

6. Taylor LS, Zografi G. 1997. Spectroscopic charac-terization of interactions between PVP and indo-methacin as amorphous molecular dispersions.Pharm Res 14:1691–1698.

7. Saleki-Gerhardt A, Zografi G. 1994. Non-isothermaland isothermal crystallization of sucrose fromthe amorphous state. Pharm Res 11:1166–1173.

8. Shamblin SL, Taylor LS, Zografi G. 1998. Mixingbehavior of co-lyophilized binary amorphous sys-tems. J Pharm Sci 87:694–701.

9. Krause S, Iskander M. 1977. Phase separation instyrene-a-methyl styrene block copolymers. In:Klempner D, Frisch KC, editors. Polymer scienceand technology. Vol. 10. New York: Plenum Press.pp 231–243.

10. Karavas E, Georgarakis M, Docoslis A, Bikiaris D.2007. Combining SEM, TEM, and micro-Ramantechniques to differentiate between the amorphousmolecular level dispersions and nanodispersionsof a poorly water-soluble drug within a polymermatrix. Int J Pharm 340:76–83.

11. Lodge TP, Wood ER, Haley JC. 2006. Two glasstransitions do not necessarily indicate immiscibiity:The case of PEO/PMMA. J Polym Sci Part B PolymPhys 44:756–763.

12. Billinge SJL, Thorpe MF. 1998. Local structurefrom diffraction. New York: Plenum Press.

13. Debye P. 1915. Dispersion of rontgen rays. AnnPhysik 46:809–823.

14. Proffen T. 2000. Analysis of occupational and displa-cive disorder using atomic pair distribution function:A systematic investigation. Z Kristallogr 215:1–8.

15. Proffen T, Billinge SJL, Egami T, Louca D. 2003.Structural analysis of complex materials using theatomic pair distribution function—A practicalguide. Z Kristallogr 218:132–143.

16. Bishop M, Bruin C. 1984. The pair correlationfunction: A probe of molecular order. Am J Phys52:1106–1108.

17. Guinier A. 1994. X-ray diffraction in crystals,imperfect crystals, and amorphous bodies. NewYork: Dover Publications.

18. Klug HP, Alexander LE. 1974. X-ray diffractionprocedures for polycrystalline and amorphousmaterial. New York: John Wiley & Sons.

19. Warren BE. 1990. X-ray diffraction. New York:Dover Publications.


4856 NEWMAN ET AL.

20. Toby BH, Egami T. 1992. Accuracy of pair distributionfunction analysis applied to crystalline and non-crystalline materials. Acta Cryst A48: 336–346.

21. Peterson PF, Bozin ES, Proffen T, Billinge SJL.2003. Improved measures of quality for the atomicpair distribution function. J Appl Cryst 36:53–64.

22. Petkov V, Billinge SJL, Shastri SD, Himmel B.2001. High-resolution atomic distribution functionsof disordered materials by high-energy X-ray dif-fraction. J Non-Cryst Solids 293–295:726–730.

23. Brent RP. 1973. Algorithms for minimization with-out derivatives. Englewood Cliffs, NJ: Prentice-Hall.

24. Crowley KJ, Zografi G. 2002. Water vapor absorp-tion into amorphous hydrophobic drug/poly(vinyl-pyrrolidone) dispersions. J Pharm Sci 91:2150–2165.

25. Yoshioka M, Hancock BC, Zografi G. 1995. Inhibitionof indomethacin crystallization in poly(vinylpyrroli-done) coprecipitates. J Pharm Sci 84:983–986.


26. Matsumoto T, Zografi G. 1999. Physical propertiesof solid molecular dispersions of indomethacin withpoly(vinylpyrrolidone) and poly(vinylpyrrolidone-co-vinylacetate) in relation to indomethacin crystal-lization. Pharm Res 16:1722–1728.

27. Bates S, Zografi G, Engers D, Morris K, Crowley K,Newman A. 2006. An analysis of X-ray amorphouspharmaceutical solids. Pharm Res 23:2333–2349.

28. Vasanthavada M, Tong W-Q, Joshi Y, Serpil Kisla-lioglu MS. 2004. Phase behavior of amorphousmolecular dispersions I: Determination of thedegree and mechanism of solid solubility. PharmRes 21:1598–1606.

29. Fox TG. 1956. Influence of diluent and of copolymercomposition on the glass temperature of a polymersystem. Bull Am Phys Soc 1:123.

30. Hiemenz PC, Rajagopal R. 1997. Principles of col-loid and surface chemistry. 3rd edition. New York:Marcel Dekker, Inc. pp 232–235.

08 DOI 10.1002/jps

Characterization of amorphous API:Polymer mixtures using X-ray powder diffraction

Documents

Transcript of Characterization of amorphous API:Polymer mixtures using X-ray powder diffraction