Physical stability and recrystallization kinetics of amorphous ibipinabant drug product by fourier transform raman spectroscopy

Physical Stability and Recrystallization Kineticsof Amorphous Ibipinabant Drug Product by FourierTransform Raman Spectroscopy

WAYNE SINCLAIR,1 MICHAEL LEANE,2 GRAHAM CLARKE,1 ANDREW DENNIS,2 MIKE TOBYN,2 PETER TIMMINS2

1Analytical and Bioanalytical Development, Bristol-Myers Squibb, Reeds Lane, Moreton, Merseyside CH46 1QW, UK

2Drug Product Science and Technology, Bristol–Myers Squibb, Reeds Lane, Moreton, Merseyside CH46 1QW, UK

Received 25 January 2011; revised 8 April 2011; accepted 18 May 2011

Published online 16 June 2011 in Wiley Online Library (wileyonlinelibrary.com). DOI 10.1002/jps.22658

ABSTRACT: The solid-state physical stability and recrystallization kinetics during storagestability are described for an amorphous solid dispersed drug substance, ibipinabant, at a lowconcentration (1.0%, w/w) in a solid oral dosage form (tablet). The recrystallization behaviorof the amorphous ibipinabant–polyvinylpyrrolidone solid dispersion in the tablet product wascharacterized by Fourier transform (FT) Raman spectroscopy. A partial least-square analy-sis used for multivariate calibration based on Raman spectra was developed and validated todetect less than 5% (w/w) of the crystalline form (equivalent to less than 0.05% of the totalmass of the tablet). The method provided reliable and highly accurate predictive crystallinityassessments after exposure to a variety of stability storage conditions. It was determined thatexposure to moisture had a significant impact on the crystallinity of amorphous ibipinabant. Theinformation provided by the method has potential utility for predictive physical stability assess-ments. Dissolution testing demonstrated that the predicted crystallinity had a direct correlationwith this physical property of the drug product. Recrystallization kinetics was measured usingFT Raman spectroscopy for the solid dispersion from the tablet product stored at controlledtemperature and relative humidity. The measurements were evaluated by application of theJohnson–Mehl–Avrami (JMA) kinetic model to determine recrystallization rate constants andAvrami exponent (n = 2). The analysis showed that the JMA equation could describe the processvery well, and indicated that the recrystallization kinetics observed was a two-step process withan induction period (nucleation) followed by rod-like crystal growth. © 2011 Wiley-Liss, Inc. andthe American Pharmacists Association J Pharm Sci 100:4687–4699, 2011Keywords: ibipinabant; FT-Raman spectroscopy; Raman microscopy; amorphous; recrystal-lization; Avrami kinetics; physical stability; solid-state kinetics; chemometrics

INTRODUCTION

Amorphous pharmaceuticals have attracted signifi-cant research due to their potential for enhancing thedissolution rate of the increasing number of poorlywater-soluble drug candidates.1–3 However, due totheir inherently metastable nature with respect tocrystallization, the physical character, be it an amor-phous or molecularly dispersed form of the drug ina formulation, has to be closely monitored as it ex-hibits the risk of recrystallization.4–6 The quality at-tributes of solid pharmaceutical product dosage form

Correspondence to: Wayne Sinclair (Telephone: +44-151-5521624; Fax: +44-151-5521500; E-mail: [email protected])Journal of Pharmaceutical Sciences, Vol. 100, 4687–4699 (2011)© 2011 Wiley-Liss, Inc. and the American Pharmacists Association

includes physical and chemical stability, for exam-ple, impurties/degradants, drug release, disintegra-tion, and tablet hardness. All these attributes can beadversely affected by solid-state phase transforma-tion process such as crystallization. The influence ofcrystallization on dissolution and bioavailability is ofmajor concern for poorly soluble drugs as it can resultin changes in in vivo drug delivery and product activ-ity. In cases where the crystallinity affects safety andefficacy, the prediction of recrystallization behavior isimportant for establishing a quality control strategy.

It is recognized that amorphous active pharma-ceutical ingredients (APIs) may undergo recrystal-lization during processing and/or storage. Storageabove the glass transition temperature (Tg) will leadto rapid conversion due to the high mobility of the

JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 100, NO. 11, NOVEMBER 2011 4687

4688 SINCLAIR ET AL.

amorphous form above the Tg. Both temperature andmoisture are important factors that can reduce thestability of amorphous pharmaceuticals7; therefore,an understanding of the crystallization kinetics fromthe amorphous state and the prediction of physicalinstability are important in the pharmaceutical in-dustry. Characterizing the solid-state kinetics of theamorphous drug recrystallization process allows thestability to be treated mathematically and enhancesthe fundamental understanding of the transforma-tion mechanism.8,9 The characterization of the recrys-tallization kinetics typically involves modeling thefraction transformed as a function of time on the basisof a mechanistic model at each temperature. Discrim-ination between kinetic models is generally based onregression analyses and relies on having high-qualitydata to quantify the transformation process. There-fore, an accurate physical form quantification methodfor the drug product is desired.

Consequently, the detection, characterization, andquantification of the amorphous form in the finaldosage form is an essential part of the developmentprocess to confirm sufficient physical stability overthe extended shelf life of the product such that it canbe utilized to improve dissolution performance. Nu-merous analytical techniques have been reported forquantification of amorphous/crystalline phase of thepure drug substance.10 Several of these methods in-cluding powder X-ray diffraction,10 differential scan-ning calorimetry,11 and solid-state nuclear magneticresonance,12 have been used to evaluate the crys-tallinity of drug/polymer substances in solid disper-sions. However, the characterization becomes consid-erably more challenging in a formulated product dueto the presence of a multiple component excipient ma-trix. Moreover, the increasing use of highly potentAPIs dictates the development of very low dosage lev-els and poses a further significant detection challengefor physical stability analysis.

In the case described herein, the selection of aquantitative technique is based on several considera-tions, most notable being the specificity and sensi-tivity of the technique to subtle changes in physi-cal form. Raman spectroscopy offers the ability forhigh API form sensitivity and specificity. Taylor andLangkilde13 demonstrated Raman spectroscopy to bea particularly effective technique for the analysis ofthe active ingredient in an intact dosage form asmost excipients are poor Raman scatterers, relativeto drug substances. Taylor and Zografi14 investigateda quantitative method for crystallinity of amorphousindomethacin using Fourier transform (FT) Ramanspectroscopy and highlighted nonhomogeneity in mix-ing to be the largest source of error in quantifyingthe degree of crystallinity down to 1% amorphous orcrystalline component of indomethacin. In addition,multivariate methods, such as partial least-square

regression (PLSR), have been extensively applied asspectra quantification methods by correlating spec-tral regions with differences in the properties of thesamples. Okumura and Otsuka15 applied FT Ramanspectroscopy in combination with PLSR to quantifythe degree of crystallinity in powders and tablets,using indomethacin as a model compound and in-domethacin/mannitol mixtures containing 10 wt %indomethacin, as a model pharmaceutical product. Inthe study, the authors used the method to detect 2%of a crystalline material in indomethacin, containedin the model product (0.2% of the total mass of thetablet).

In this study, we quantitatively determined thephysical stability of a solid dispersion of the API,ibipinabant (Fig. 1), in a solid dosage form tablet.We show that FT Raman spectroscopy is effectivelyapplied to detect and quantify the extent of recrys-tallization in a product containing a low dose amor-phous API content and detect the crystalline materialwhen present at less than 0.05% of the total mass ofthe tablet. Furthermore, this method was used to cor-relate the impact of this variable to critical processparameters of the solid dosage form. Additionally, themethod was used to investigate the kinetics of therecrystallization process in the tablet product. Ibip-inabant is a member of a novel class of diarylpyra-zolines and a potent and selective CB1-subtype selec-tive receptor antagonist explored for the treatment ofobesity and diabetes. Because of its very poor aque-ous solubility, a solid dispersion of the drug substancewith polyvinylpyrrolidone (PVP) was prepared to im-prove overall bioavailability. The solid dispersion wascompressed into tablets with a resulting API con-centration of 1.0% (w/w). The physical stability wasevaluated by means of FT Raman spectroscopy com-bined with PLSR. The method was applied to quantifythe spectra in order to predict the recrystallizationbehavior in the product after exposure to a varietyof accelerated stability storage conditions. Dissolu-tion behavior at 37◦C of the product was also studied

Figure 1. Chemical structure of ibipinabant. Molecularformula: C23H20Cl2N4O2S. Molecular weight 487 g/mol.

JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 100, NO. 11, NOVEMBER 2011 DOI 10.1002/jps

PHYSICAL STABILITY AND RECRYSTALLIZATION KINETICS OF IBIPINABANT 4689

in order to demonstrate the impact of the predictedcrystallinity on this critical quality attribute. Finally,the recrystallization kinetics of amorphous ibipina-bant in the tablet was studied in detail using FTRaman spectroscopy. Johnson–Mehl–Avrami (JMA)theory was applied to determine the reaction order,rate constant, and mechanistic information for the re-crystallization process. Similar types of studies havebeen carried out from pure amorphous drugs16,17 andtheir solid dispersions,18 but until now, little infor-mation is conducted in the final dosage form. In thisway, FT Raman spectroscopy was utilized as a sensi-tive method to assess the physical stability of a lowconcentration amorphous API during tablet manufac-ture or stability testing. The application of FT Ramanspectroscopy to the detection and characterization ofthe ibipinabant solid-state form in such a low dosedrug product and quantitative treatment of the phys-ical stability would be difficult to achieve by any othermethod.

EXPERIMENTAL

Materials

The crystalline form of the drug, ibipinabant, 3-(4-chlorophenyl)-N-[(4-chlorophenyl) sulfonyl]-4,5-dihydro-N′-methyl-4-phenyl-1H-pyrazole-1-carboxi-midamide, was manufactured by Solvay Pharmaceu-ticals B.V (Weesp, the Netherlands). The crystallineform of ibipinabant has a room temperature aqueoussolubility of less than 0.05 :g/mL at pH 7.0 andmelting point of 159◦C–162◦C. The amorphous formof ibipinabant was prepared from the crystalline formby quench cooling in liquid nitrogen after meltingthe bulk powder. The amorphous form was groundin a mortar and pestle and stored at −5◦C. The soliddispersion was manufactured via spray drying atBristol–Myers Squibb Company. (New Jersey) usinga process similar to that previously employed19,20

and consisted of 20% drug substance, 5% sodiumlauryl sulfate (SLS), and 75% PVP K-30 and storedat 2◦C–8◦C. The Tg for the solid dispersion was 132◦Ccompared with 70◦C for the pure amorphous drug.The crystallinity of the pure amorphous form andspray-dried intermediate (SDI) was assumed to be0%. PVP K-30 was purchased from ISP, Wayne, NewJersey. Lactose monohydrate (LMH) was purchasedfrom Sheffield, Norwich, New York. Avicel PH101 R©

microcrystalline cellulose (MCC) was purchasedfrom FMC International, Cork, Ireland. SLS waspurchased from Spectrum Laboratory ProductsInc., Gardena, California. AcDiSol R© Croscarmellosesodium was purchased from FMC International.Magnesium stearate was purchased from Mallinck-rodt Inc., St Louis, Missouri. Silicon dioxide (Syloid)244 R© was purchased from Grace Davison, Columbia.

Table 1. Formulation Details of Amorphous Ibipinabant Product

Component Level (%)

API/PVP/SLS (20:75:5) 5.0MCC: Lactose monohydrate 40:45.5Croscarmellose sodium 5.0SLS 2.0Silicon dioxide 2.0Magnesium stearate 0.5

Preparation of Solid Dosage Form

The amorphous tablet formulation (Table 1) was pre-pared using a dry granulation process. The amor-phous SDI of ibipinabant and SLS were passedthrough a 1.0 mm screen prior to high-shear mixingwith croscarmellose sodium, LMH, and MCC. Mag-nesium stearate was passed through a 0.5 mm screenand added to the mixture and blended. Prior to tablet-ting, the blends were roller compacted, milled usinga cone mill, and then lubricated with screened mag-nesium stearate. Tablets of 9 mm diameter were com-pressed with a fill weight of 250 mg and with a crush-ing strength value between 70–125 N. Each tabletcontained 12.5 mg of amorphous SDI [equivalent to2.5 mg (1.0%, w/w) of the API]. Tablets were eval-uated for assay (95%–105%) and content uniformity(<6% RSD), which were found to be acceptable.

FT Raman Spectroscopy

Raman spectroscopy was employed to investigate thecrystallinity of amorphous tablets of ibipinabant dur-ing the storage stability study. An FTIR spectrometer(Nicolet Nexus 870; Thermo Fisher, Waltham, MA) in-terfaced with a FT Raman module was used to collectRaman spectra using Omnic software. Analysis wascarried out at room temperature with an air-cooleddiode pumped Nd–YAG laser (1064 nm) as the excita-tion source, CaF2 beam splitter and indium–galliumarsenide (InGaAs) detector. Calibration samples (100mg) were measured in 5 mm diameter NMR tubes.The laser power was focused on the sample with apower of about 500 mW. For each spectrum, 64 scanswere averaged at a resolution of 4 cm−1 over the spec-tral range of 3700 to 100 cm−1 in a 180◦ scatteringconfiguration in order to provide Raman spectra witha low signal-to-noise (s/n) ratio in combination withwell-resolved Raman signals. Tablet samples weremeasured by placing them directly facing the Ramanlaser. Duplicate measurements were taken in eachcase. The laser power was focused on the sample witha power of about 800 mW. For each spectrum, 256scans were averaged at a resolution of 4 cm−1 over thespectral range of 3700 to 100 cm−1 in a 180◦ scatter-ing configuration in order to provide Raman spectrawith a low s/n ratio in combination with well-resolvedRaman signals. Sulfur was used as the reference stan-dard to monitor wavenumber accuracy.

DOI 10.1002/jps JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 100, NO. 11, NOVEMBER 2011


Micro-Raman Spectroscopy

Raman microscopy studies on the surface of the intacttablets were performed by using a Jobin-Yvon/Horibamicro-Raman spectrometer (Model Labram HR800),equipped with a 633 nm He/Ne laser source (12 mW),1800 1/mm grating with an Olympus BX41 micro-scope system. Sample mapping in two dimension (2D)was performed at a step increment of 2 :m in both xand y directions over approximately 100 × 100 :marea of the tablet.

Calibration Standard Preparation

An important factor in determining crystallinity ofbinary mixtures using FT Raman spectroscopy is theestablishment of homogeneity. To reduce sample het-erogeneity caused by particle size differences, purecrystalline and amorphous forms were gently groundusing a pestle and agate mortar. The maintenance ofthe original solid form of the ground samples wereverified by comparison of the Raman spectra to thecorresponding spectra from unground material. Phys-ical mixtures corresponding to a known ratio of crys-talline and amorphous forms of ibipinabant were pre-pared by weighing each component on a microbalance(Sartorius model RC210D), followed by gentle mix-ing using an agate mortar. The resulting mixtureswere then mixed to 10% (w/w) in the placebo ma-trix. This helped to minimize the difference betweenthe s/n ratio of Raman spectra of calibration sam-ples and the real tablet samples to be used in predic-tion where the drug concentration was at 1% (w/w).Moreover, tablet samples were measured with higherlaser power and a greater number of scans to increasethe s/n ratio in the Raman spectra of these samplesand further minimize any potential impact. The to-tal weight of each standard was 200 mg. The glassvial containing the mixture was held against a run-ning vortex mixer to homogenize the powders. A totalof 11 binary mixtures were prepared. Standard sam-ples were made over the concentration range of crys-tallinity content from 0% to 100% (w/w) at 10% (w/w)intervals. Compositions at the low end of the concen-tration range, 5% and 15% (w/w) samples were alsoincluded. Each composition was prepared in tripli-cate. All mixtures were stored in closed glass vials andmeasurements were performed on the day of prepa-ration. The mixtures were used as reference samplesfor quantitative analysis of the weight percent of thecrystalline API in the tablet product. An indepen-dent validation set consisting of equivalent compo-sitions to the standard set was prepared in triplicate.These compositions were not incorporated into thecalibration models but were used for assessing modelvalidity.

Chemometric PLS Method

Thermo Nicolet TQ Analyst version 7.2 chemometricsoftware was used to construct the PLS model. Thepure samples of the 100% crystalline and 100% amor-phous form as well as the calibration mixtures weremeasured in triplicate. Several different preprocess-ing and scaling techniques were applied to eliminatesources of nonlinearity or minimize features uncorre-lated with the concentration of ibipinabant. Standardnormal variate (SNV) preprocessing was applied tothe raw data and mean centering was used for scal-ing the variables. The spectral region was chosen forthe analysis in a wavenumber range that showed thelargest spectral contrast between components, thusensuring the greatest contribution to the linear equa-tion for the property of interest. The linear correlationcoefficient (R2), root mean square error of calibration(RMSEC), and root mean square error of prediction(RMSEP) were compared to evaluate the quality ofthe model. Cross validation, using leave-one-out tech-nique, was performed to estimate the robustness ofthe model and the root mean square error of cross-validation (RMSECV) was calculated to select theoptimum of factors for the PLS models.

Stability Storage Experiments

The conditions used in this study are standard forearly-stage investigational new drug stability stud-ies. More extensive stability mapping would be usefulduring full commercial development. Physical stabil-ity of the ibipinabant formulation was assessed byRaman spectroscopy for 6 months. Tablets wereplaced into incubators under a variety of stabilitystorage environments: 2◦C–8◦C, 25◦C/60% relativehumidity (RH) open and closed dish, and 40◦C/75%RH closed dish. Tablets were stored either in glasspetri dishes or sealed bottles for open and closed stor-age conditions, respectively. The closed storage condi-tion was achieved with a heat induction sealed high-density polyethylene bottle (75 mL, round, white, Cap38/400 mm). Tablets (40) were packaged per con-tainer. One desiccant (Sorb-It 0.75 g Canister XFR/XEN) and cotton coil 16 g/yd was included per bot-tle. Previous developmental stability studies had con-cluded that the use of desiccants in the packagingimproved physical stability, with one selected as op-timum. Periodically (initial, 3 and 6 months), sam-ples were removed and characterized for presence ofcrystallinity by Raman spectroscopy analysis. In ad-dition to Raman analysis, samples from drug productstability studies were assayed for content unifor-mity, degradants, and impurities by a validatedhigh-performance liquid chromatography method,moisture content by Karl Fisher titration, tabletcrushing strength, and disintegration times. In addi-tion to the standard storage conditions, tablets were



also stored for up to 20 days at ambient 95% RH,as these conditions were likely to accelerate crystal-lization, by placing them into a constant humiditychamber of saturated K2SO4, characterized daily byRaman spectroscopy, after which (∼10 min) they werereturned to chamber until the next sampling time,when the same samples were reevaluated.

Dissolution Experiments

Discriminating dissolution conditions were identifiedthat demonstrated fast, complete dissolution fromtablets containing amorphous drug but showed slow,incomplete dissolution from the tablets containingcrystalline material. Dissolution profiles were gener-ated for 250 mg tablets containing 2.5 mg of ibipina-bant, following storage under the different conditions.Tablets were tested using a six station USP II (paddle)dissolution apparatus (Vankel) operating at 65 rpm.Phosphate buffer solution (1000 mL) of pH 6.8 con-taining 0.075% SLS was used as dissolution medium(37 ± 0.5◦C). The amount of dissolved drug in the dis-solution medium was determined by ultraviolet (UV)spectrometry (Agilent 8543 using Chemstation soft-ware) with a flow-through cuvette at a wavelengthof 313 nm. The absorbance values of aliquots of thedissolution medium were collected at 5, 10, 15, 20,30, 45, 60, 90, 120, 150, and 180 min. No UV inter-ference from excipients was confirmed by scanning asolution of the placebo tablet. Six tablets were usedin the dissolution test.

RESULTS AND DISCUSSION

Characterization by FT Raman Spectroscopy

Figure 2 (top) shows the FT Raman spectra of crys-talline, amorphous (from melt-quench method), andamorphous SDI forms of ibipinabant and PVP overthe spectral range 1500–1650 cm−1. Crystalline andamorphous forms have a prominent band region inthe range 1500–1650 cm−1. Molecular bands at 1595and 1560 cm−1 can be attributed to the ring deforma-tion mode of the benzene ring present in the molecule(Fig. 1). Spectral differences between the two formscan be found in this wavenumber region. Relative tothe crystalline form, the spectrum of the amorphousform had broader bands that reflect the local disorderin the amorphous form. Specifically, the bands at 1595and 1560 cm−1 showed an increase in full width athalf maximum and small frequency shifts comparedwith the crystalline form. A clear difference in bandposition of the 1560 cm−1 band is apparent on mov-ing from amorphous (1559 cm−1) to crystalline (1563cm−1) form. With higher temperature/humidity condi-tions, the amorphous form of ibipinabant is convertedinto the crystalline form as observed by the narrow-ing and growth of the band near 1600 cm−1. PVP hasa band at 1660 cm−1, attributed to the carbonyl group

stretching mode. In the SDI, this band has no spectralinterference in the spectral region 1500–1650 cm−1

and is significantly less intense than the band fromthe drug, demonstrating that crystalline and amor-phous forms of ibipinabant can still be readily dis-tinguished by Raman spectroscopy in the SDI withno interference of the PVP polymer. Figure 2 (bot-tom) shows the FT Raman spectra of (a) amorphousSDI form and the excipient used in the drug product(b) LMH, (c) MCC, and (d) the studied tablet prepa-ration. The API bands are well resolved from excipi-ent features, indicating no excipient interference andare observed in the tablet despite being only presentwith a low (1.0%, w/w) drug load. This is due to thesignificantly larger Raman scattering cross-sectionof the aromatic groups found in compound ibipina-bant, in contrast with excipients, which are typicallyaliphatic. Therefore, Raman spectroscopy is sensitiveto determine the API in the presence of excipients.

PLS Model Calibration by FT Raman Spectroscopy

The spectral contrast in Raman spectra between crys-talline and amorphous forms has been used to quan-titate the crystalline content in pure SDI and in arange of common excipient materials. The relativecrystallinities of crystalline ibipinabant and the soliddispersion were defined as 100% and 0% (w/w), re-spectively. The calibration for quantitative evalua-tion of the crystalline form of ibipinabant in tabletswas calculated using a multivariate PLSR model.The spectral region of the amorphous and crystallineforms used for multivariate analysis was in the re-gion between 1540 and 1640 cm−1 that has no Ra-man signal interference from excipients. The PLS1model using the raw spectral data and SNV trans-formation as pretreatment method and mean cen-tering as a scaling method was found to be optimal.Figure 3 (top) shows the Raman spectra acquired afterSNV transformation and mean centering for the stan-dard mixtures, which demonstrated significant corre-lation in the 1540–1640 cm−1 range between the spec-tra and the crystallinity of ibipinabant. The predictedresidual error sum of squares plot shown in Figure 3(middle) suggests that it is sufficient to build the PLSmodel with the use of two factors. The PLSR model,obtained by correlation between estimated and speci-fied percent (w/w) crystallinity content value showedgood linearity [RMSECV was 3.11% (w/w) and R2

was 0.9979]. From the use of the validation set stan-dards, the RMSEP was evaluated to be less than3.5% (w/w) and represents the overall difference be-tween the model predicted values and the true valuesof crystalline API in the standards. The value of R2

was 0.9962 in the concentration range from 0.0% to100.0% (w/w) of the crystalline API content. There-fore, the calibration model was verified to have goodaccuracy and precision. Figure 3 (bottom) shows the



Figure 2. (Top) Raman spectra of crystalline and amorphous forms of ibipinabant and PVPover the spectral range 1500–1650 cm−1: (a) crystalline, (b) amorphous (from melt-quenchmethod), (c) SDI, and (d) PVP. (Bottom) Comparison of Raman spectra of (a) the SDI of ibipin-abant, (b) Lactose monohydrate, (c) Avicel, and (d) tablet preparation with 1% (w/w) drug loadof ibipinabant as indicated in Table 1.

multivariate calibration curve based on Raman mea-surements using a PLS algorithm with two factors.

Quantitative Analysis of Crystallization During TabletStability Studies

The current International Conference on Harmoniza-tion guidelines on long-term stability testing require-ments for a registration application provides the

conditions for assessing pharmaceutical products.21

The utility of the Raman method was demonstratedby evaluating the effect of these conditions on thephysical stability of the amorphous SDI of ibipinabantafter manufacture in the form of a low dose tabletfor 6 months. Tablet stability storage conditions at2◦C–8◦C, 25◦C / 60% RH and 40◦C / 75% RH were se-lected for stress testing the long-term stability of the



Figure 3. (Top) Overlaid comparison of SNV transformedspectra of ibipinabant standard mixtures of varying crys-tallinity over the spectral range 1520–1640 cm−1. Percent-ages show the degree of ibipinabant crystallinity in the sam-ple. (Middle) RMSECV values calculated for crystallinitymodeling in tablets based on Raman spectra. (Bottom) Cor-relation between the actual and predicted percentages ofcrystalline form for the calibration set and test set of crys-tallinity content of physical mixtures.

amorphous form in the tablet. At initial, the resultsshow that the drug retains its amorphous form in thetablet, indicating that crystallization growth is notinduced during the course of the manufacturing pro-cess. The intact tablet samples were tested after 3 and6 months in both closed (induction sealed containers)and open dish conditions. Figure 4 (top) shows the Ra-man spectra of tablets samples at the range of storage

Figure 4. (Top) FT Raman spectra of amorphous ibip-inabant tablets showing the benzene deformation region(1550–1650 cm−1) following exposure for 3 months to2◦C–8◦C, 25◦C/60% RH (closed), 25◦C/60% RH (open), and40◦C/75% RH (closed). (Bottom) Calculated degree of crys-tallinity of intact tablet with a 1.0% drug load followingexposure for 3 and 6 months to 2◦C–8◦C, 25◦C/60% RH(closed), 25◦C/60% RH (open), and 40◦C/75% RH (closed).

conditions after 3 months. At 25◦C/60% RH (open) and40◦C/75% RH (closed), the band at 1600 cm−1 beginsto sharpen, and with concomitant increase in band in-tensity relative to the bands of the excipient, confirm-ing the conversion to the crystalline form. Figure 4(bottom) shows the predicted percent (w/w) crys-tallinity in the tablet at 3 and 6 months by using thePLS Raman method. The drug retains its amorphousform when stored at 2◦C–8◦C after 6 months, whereasminimal levels of crystallinity are predicted at 25◦C/60% RH (closed). In general, increases in crystallinitywere observed correlating to increases in temperatureand humidity. A significant increase in crystallinitywas observed at 25◦C/60% RH (open) with about 35%conversion after 6 months. At 40◦C/75% RH (closed),the tablet showed faster conversion, with about 80%crystallinity after 6 months. Higher RH and tempera-ture accelerated the transformation even though thecompound is very hydrophobic. In contrast, the SDImaterial itself showed no detectable recrystallizationafter 3 months at 25◦C/60% RH in the open dish



condition. This difference in the crystallizationpropensity of ibipinabant in the SDI and groundtablet when exposed to the high humidity environ-ment indicates compression effects could be responsi-ble for the relative decrease in physical stability.

Raman Mapping of Tablets during Stability Studies

In order to investigate the spatial distribution of therecrystallization behavior of the active compound inthe tablet, XY (2D) micro-Raman spectroscopy map-ping was conducted at drug-rich locations on the sur-face of the tablet for each stability storage condition.A mapped image was acquired over a 100 × 100 :marea. The data were analyzed based on the band po-sition of the band near 1560 cm−1 and shows cleardifference in band position on moving from the amor-phous (1559 cm−1) to crystalline (1564 cm−1) form.Data analysis consisted of extraction of the data inthe 1545–1575 cm−1 region, thresholding to removeall data corresponding to “no drug content” and bandfitting of the 1560 cm−1 band to the positions ob-served for the pure amorphous and crystalline spec-tra. Figure 5 shows the results of XY Raman map-ping of the tablet at the range of stability conditionsafter 3 months. The color bar in Figure 5 repre-sents the position of the band near 1560 cm−1 andthe band positions at 1559 cm−1 (green areas) and1564 cm−1 (yellow areas) described the largely amor-phous and largely crystalline regions, respectively. Atthe 2◦C–8◦C condition, there is no evidence of the crys-talline form evidenced by only seeing the green areas.

Typical domain sizes range from about 5–20 :m. At25◦C /60% RH (closed), there were no clear areas ofcrystalline material visible. At 25◦C/60% RH (open),the majority of the API appears to be amorphous, al-though there are some clear regions of crystalline ma-terial (red areas). The mapped region clearly showsthat at the 40◦C/75% RH (closed) condition, thereis a large degree of crystalline aggregates (yellow/red areas) but nonetheless significant contribution ofamorphous material (blue/green areas). In general,the micro-Raman results are consistent with the ob-served trends observed from FT Raman analysis andperhaps with further method development may havethe potential to provide even further sensitivity fordetecting the crystalline form in the product.

Influence of Crystallinity Content on Physical Properties

As previously described, the critical quality attributesof a solid product include chemical stability, dissolu-tion, crushing strength, density, disintegration, andso on. All these attributes can be affected by solid-state phase transformation processes such as crystal-lization of an amorphous phase.

The influence of crystallization on the dissolutionand hence bioavailability is of major concern forpoorly soluble drugs, as it may result in changes indissolution rate of the product. In addition to Ra-man spectroscopy, several tests were conducted toassess the quality of the tablets after storage, sum-marized in Table 2. There were no significant dif-ferences in physical properties after storage for up

Figure 5. XY Micro Raman mapping of intact tablet and accompanying spectra showing dis-tribution of amorphous and crystalline ibipinabant following exposure for 3 months to 2◦C–8◦C,25◦C/60%RH (closed), 25◦C/60%RH (open), and 40◦C/75%RH (closed).The scale bars indicatethe position of the band at 1560 cm−1 and the band positions at 1559 cm−1 (green areas)and 1564 cm−1 (yellow areas) described the largely amorphous and largely crystalline regions,respectively.



Table 2. Physical Stability Data for Amorphous Ibipinabant Product After Storage

Test Time(months) Storage Condition Moisture (%)

HardnessValues (N)

Disintegration Time(min) (n = 6)

Initial Initial 5.20 100 ± 6 N/R3 2◦C–8◦C 5.22 79.1 ± 6 6.03 25◦C/60%RH (closed) 5.14 81.2 ± 5 6.53 25◦C/60%RH (open) 5.38 65.8 ± 7 6.53 40◦C/75%RH (closed) 5.32 95.9 ± 7 7.5

to 3 months at 2◦C–8◦C and 25◦C/60% RH (closed);the tablet crushing strength and mean water contentwere comparable to the initially prepared samples.When the tablets were stored in the more aggres-sive conditions (25◦C/60% RH open and 40◦C/75%RH closed), no major changes in the physical prop-erties were noted, and confirming that crystallizationof the amorphous form on storage had produced noeffect on the crushing strength and disintegrationof the tablet. In addition, no significant increasesin chemical degradation or presence of impuritieswere observed. A discriminating dissolution methodwas developed to discriminate for the formation ofcrystalline material in the amorphous formulations.Figure 6 shows the dissolution profiles at each stabil-ity storage condition after 3 months and is comparedwith the corresponding predicted percent (w/w) levelsof crystallinity by the Raman spectroscopy method.Also shown is the dissolution profile obtained for atablet formulated with the same amount (2.5 mg) ofthe crystalline form of ibipinabant. Clear differencesin release profile can be seen depending on the envi-ronmental conditions at which the tablet was stored,with decreasing release profiles correlating with in-creasing levels of crystalline conversion. The fastestand most complete dissolution profiles were seen fortablets stored at 2◦C–8◦C (no crystallinity detected).These were followed closely by tablets stored at 25◦C/60% RH (closed) with less than 10% crystalline con-version detected. When tablets were stored at the ag-gressive conditions (25◦C/60% RH open and 40◦C/75%RH closed), significant changes in the dissolution pro-

files were observed. Tablets stored at 40◦C/75% RH(closed) showed the slowest profiles; the dissolutionrate decreased to around 65% released after 180 minfor approximately 50% predicted crystalline contentand was almost comparable to the dissolution profileof a tablet containing 2.5 mg of crystalline drug ina similar lactose and MCC filler formulation matrix.The dissolution property and apparent solubility ofthe solid dispersion of ibipinabant in the product canbe explained on the basis of the amount of the recrys-tallization, which occurred after 3 months storage asdetermined by the Raman spectroscopy method.

Kinetics Studies of the Recrystallization Process

The recrystallization kinetics of ibipinabant in theformulation was conducted as follows. Samples werestored at 95% RH and room temperature to expe-dite the solid-state phase transformation and the de-gree of crystallinity with time was evaluated basedon prediction by the Raman method. A ground tabletand the SDI were stored unprotected at 95% RH androom temperature and analyzed at different time in-tervals (initial, daily). The FT Raman spectrum of theSDI showed no significant change during the durationof the experiment, suggesting that it remained as astable amorphous form. In contrast, Figure 7 (top)shows significant changes in the Raman profiles ofthe ground tablet with storage time at 95% RH. Again,like the FT Raman spectra in Figure 2, the changes inthe Raman spectra of the tablet are consistent with in-creased crystallinity. Crystallization usually involvesa two-step process of nucleation and growth. Recrys-

Figure 6. Calculated degree of crystallinity and dissolution profile of intact tablet with a1.0% drug load following exposure for 3 months to 2◦C–8◦C, 25◦C/60%RH (closed), 25◦C/60%RH(open), and 40◦C/75% RH (closed).



Figure 7. (Top) FT Raman spectra of amorphous ibip-inabant tablets showing the benzene deformation region(1550–1650 cm−1) following exposure for various periods oftime (up to 2 weeks) to room temperature and 95% RH.(Bottom) Calculated crystallization profile, conversion ver-sus time, for the crystallization of amorphous ibipinabanttablets at room temperature and 95% RH.

tallization kinetics of ibipinabant in the tablet can bequantified by application of the JMA model, widelyknown as the “random nucleation and growth” modelin which the amorphous phase coexists with the crys-talline phase during the phase transformation.22–25

The JMA model can be used to describe the develop-ment of the relative degree of recrystallization as afunction of time according to the following idealizedequation:

" = 1 − exp {− [k(t − tind)]} (1)

where α is the fraction of crystalline drug at the stor-age time t, tind is the induction period to the nu-clei growth, k is the apparent temperature-dependantrate constant (time−1), which includes nucleationand growth, and n is the temperature-independentAvrami exponent (dimensionless) that depends on thenucleation and growth mechanism and the number ofdimensions, m, in which crystal growth occurs.26 Theequation describes a random nucleation and growth

model, in which transformations start at a particularsite and continue by progressive propagation of the re-action interface. Various nucleation mechanisms areknown such as sporadic (random) nucleation and in-stantaneous nucleation. In sporadic nucleation (ther-mal), the nuclei are formed continuously (linearly)with first-order time dependence. In instantaneousnucleation (athermal), all the nuclei that are to ap-pear will appear at one time with zero-order timedependence. Crystal growth proceeds on the nucleias soon as they are formed. Crystal growth dimen-sion, m, is either, one, two, or three corresponding torod-like (unidimensional), disk-like (bidimensional),and spherulitic (tridimensional) crystal habitat, re-spectively. The reaction order, n, is equal to m + 1,if nucleation is sporadic or equal to m if nucleationis instantaneous. According to the model, the rate ofcrystallization (growth rate) and the dimensionalityof the growth in the model are assumed to be lin-ear and constant. The parameters n and k can beused to interpret qualitatively the nucleation mech-anism, morphology, and overall crystallization rateof the drug in the formulation. Table 3 describes thecrystal growth dimensions for different values of n forsporadic and instantaneous nucleation mechanisms.The values of k and n can graphically be determinedby the classical double-logarithm expression of Eq. 1.

ln[−ln (1 − ")

] = n ln (t − tind) + n ln (k) (2)

The slope corresponds to the Avrami exponent, n.After assigning the appropriate order, the linearizedform of Eq. 1

[−ln (1 − ")]1/n = k(t − tind) (3)

can be used for determination of rate constants.Figure 7 (bottom) shows the predicted changes incrystallinity of ibipinabant product over time at roomtemperature and 95% RH by using the PLS Ramanmethod (1550–1650 cm−1) and shows the sigmoidalcurve expected for reactions occurring through nu-cleation and growth, indicating an induction time fol-lowed by a rapid increase and eventually a leveling offof crystallinity. The crystallinity conversion is shown

Table 3. Values of the Avrami Exponent n for DifferentNucleation and Growth Mechanisms

Crystal Growth DescriptionAvramiexponent, n Sporadic nucleation Instantaneous nucleation

1 – Rod-like2 Rod-like Disk-like3 Disk-like Spherulitic4 Spherulitic



to be complete over the timescale of the experiment. Incontrast, the SDI showed no detectable increase in re-crystallization over the duration of the experiment. Alag time between the beginning of the experiment andthe first detection of recrystallization was observedand is estimated to be between 50 and 60 h. The totaltime for the water vapor absorption process was deter-mined to be complete within 10 h for the sample anddoes not fully account for the observation. Therefore,we define this period as the induction time tind as-sociated with the actual nucleation process (randommechanism) or the slow growth of germ nuclei (in-stantaneous mechanism). This is the time requiredfor the generation of nuclei that will be transformedinto fully active growth nuclei and identified here asthe time interval prior to conforming to the Avramikinetics. The crystal growth behavior of amorphousibipinabant in the tablet formulation and in the SDIcan be followed by spectral changes in the FT Ra-man spectra. To clarify the kinetic mechanism of re-crystallization of amorphous ibipinabant, the fractionof crystallization was evaluated based on the Ramanmethod and kinetic parameters were evaluated basedon the Avrami solid-state kinetic equation. Crystalgrowth was only reliably detected at concentrationsgreater than 5% (w/w). In Figure 8 (top), the crys-tallized fraction was plotted against time using thelinear expression of the JMA equation (Eq. 2). Thevalue of the Avrami exponent, n, was found from theslope. Figure 8 (middle) shows the experimental dataand fit of Eq 3 using the derived value of n, for crystal-lization at room temperature in the conversion range0.1 < " < 0.9. The value of the rate constant, k, wasfound from the slope. The results are shown in Table 4.The correlation coefficients (R2) were greater than0.99 and indicates that the JMA equation correctlydescribes the crystallization process. The Avrami ex-ponent was determined to be 1.92, suggesting a singletransformation mechanism and gave a JMA rate con-stant of 14.6 × 10−3 h−n. Figure 8 (bottom) showschanges in crystallinity over time corrected by theinduction time. The solid line represents the fittedAvrami curve (according to Eq. 1; n = 1.92). In gen-eral, these results indicated that crystallization con-sisted of either a sporadic nucleation process with aninduction period followed by a rod-like crystal growthor an instantaneous nucleation process with disk-likecrystal growth. Optical microscopy observation of the

Table 4. Avrami Parameters Experimentally Derived for theRecrystallization of Amorphous Ibipinabant Product at RoomTemperature and 95% RH Monitored by FT Raman Spectroscopy

Kinetic Parameter Calculated Value

n 1.92 (R2 = 0.995)k (h-n) × 103 14.6 (R2 = 0.994)tind (h) 50–60

Figure 8. (Top) Crystallization data plotted using theclassical double logarithm Avrami plot for determinationof exponent n. (Middle) Plot of [−ln (1 − ")]1/n against t −tind for data calculated with, n = 1.92, for determination ofrate constant k. (Bottom) Comparison between experimen-tal data and model prediction during crystallization.

recrystallization of amorphous ibipinabant (solubi-lized in a lipid based semisolid microemulsion system)after direct exposure to 40◦C/75% RH for 2 months ap-peared to follow a needle-like crystal growth (Fig. 9).The observation of a noticeable induction period aswell as the needle-like crystal habit growth wouldsuggest a sporadic nucleation mechanism for ibipina-bant in the low dose drug product.

CONCLUSION

Evaluation of the physical stability of amorphouspharmaceutical tablets containing low doses of po-tent drug molecules presents a challenge to currentanalytical testing methods. In combination with mul-tivariate analysis, a quantitative FT Raman spec-troscopic method was developed for the analysis of



Figure 9. Visual microscopy of needle-like crystals afterrecrystallization of amorphous ibipinabant that was solubi-lized in a lipid-based semisolid microemulsion system afterdirect exposure to 40◦C/75% RH for 2 months.

mixtures of crystalline and amorphous ibipinabant.The method was verified to be useful to predict crys-talline content in a very low dose (1%, w/w) drug prod-uct under a range of standard stability storage condi-tions. The potential for physical instability has beenexamined and showed that extremes of exposure to el-evated temperature and humidity induced significantlevels of crystallinity to occur. As the disintegrationand crushing strength of the tablets remained invari-ant, the level of crystallinity could be directly corre-lated to an observed slow down in the dissolution pro-files. This difference in the crystallization propensityof ibipinabant in the SDI and ground tablet when ex-posed to the high humidity environment indicates aninfluence from processing during formulation and willbe the subject of further study. It was demonstratedthat the crystallization process in the product can bemonitored by FT Raman spectroscopy to study recrys-tallization kinetics. The recrystallization of the SDIwas found to be a two-step process. The first step wasassociated with nucleation from the observation of aninduction period. The second step was crystal growth,which was monitored by the spectral changes in theRaman spectra. The experimental data showed excel-lent fitting to the second-order JMA model. Elevatedhumidity exposure of the formulated product for ex-tended time periods can serve as a nucleation cata-lyst that seeds crystallinity. Given the sensitivity forrecrystallization to environmental conditions, corre-lation of the rate constant depending on temperatureand RH values could further aide in understandingthe kinetics of the process. Chemometric FT Ramanspectroscopy has been shown to be a powerful tool forthe detection and evaluation of the phase transitionprocess of a low dose amorphous solid dosage form.It can be applied effectively to address uncertain-

ties regarding the stability of the amorphous phase.Thorough characterization of the kinetics can also en-hance the fundamental understanding of the underly-ing crystallization process. These results demonstratethe utility of Raman spectroscopy and multivariateanalysis for detection and monitoring the physicalstability of ibipinabant in a very low dose productduring stability storage.

ACKNOWLEDGMENTS

We would like to acknowledge Solvay Pharmaceuti-cals B.V for supply of ibipinabant drug substance, thecontribution of Feng Qian and Raja Haddadin for themanufacture of the SDI and that of Sandra Costall,Terry McMahon and Sarah Brown for the manufac-ture and dissolution testing of the amorphous tabletformulation. We also wish to thank Dr. Simon Fitzger-ald, Jobin Yvon Ltd. for the technical assistancein Raman microscopy experiments. We also thankGary McGeorge, Bristol–Myers Squibb, for helpfuldiscussions.

REFERENCES

1. Serajuddin ATM. 1999. Solid dispersion of poor water-solubledrugs: Early promises, subsequent problems, and recentbreakthroughs. J Pharm Sci 88:1058–1066.

2. Hancock BC, Parks M. 2000. What is the true solubility advan-tage for amorphous pharmaceuticals? Pharm Res 17:397–404.

3. Yu L. 2001. Amorphous pharmaceutical solids: Preparation,characterization and stabilization. Adv Drug Delivery Rev48:27–42.

4. Hancock BC, Zografi G. 1997. Characteristics and significanceof the amorphous state in pharmaceutical systems. J PharmSci 86:1–12.

5. Yoshioka S, Stella VJ. 2000. Stability of drugs and dosageforms. New York: Kluwer Academic/Plenum Publishers,pp 139–150.

6. Bhugra C, Pikal MJ. 2008. Role of thermodynamic, molecular,and kinetic factors in crystallization from the amorphous state.J Pharm Sci 97:1329–1349.

7. Ahlneck C, Zografi G. 1990. The molecular basis for moistureeffects on the physical and chemical stability of drugs in thesolid state. Int J Pharm 62:87–95.

8. Carstensen JT. 1995. Drug stability: Principles and practices.2nd ed. New York: Marcel Dekker.

9. Feth PF, Volz J, Hess U, Sturm E, Hummel RF. 2007.Physicochemical, crystallographic, thermal, and spectroscopicbehaviour of crystalline and x-ray amorphous ciclesonide.J Pharm Sci 97(9):3765–3780.

10. Shah B, Kakumanu VK, Bansal AK. 2006. Analytical tech-niques for quantification of amorphous/crystalline phases inpharmaceutical solids. J Pharm Sci 95(8):1641–1665.

11. Miyazaki T, Yoshioka S, Aso Y, Kawanishi T, 2007. Crystal-lization rate of amorphous nifedipine analogues unrelated tothe glass transition temperature. Int J Pharm 336:191–195.

12. Ito A, Watanabe T, Yada S, Hamaura T, Nakagami H, HigashiK, Moribe K, Yamamoto K. 2010. Prediction of recrystalliza-tion behaviour of troglitazone/polyvinylpyrrolidone solid dis-persion by solid-state NMR. Int J Pharm 383: 18–23.

13. Taylor LS, Langkilde FW. 2000. Evaluation of solid-stateforms present in tablets by Raman spectroscopy. J Pharm Sci89(10):1342–1353.



14. Taylor LS, Zografi G. 1998. The quantitative analysisof crystallinity using FT-Raman spectroscopy. Pharm Res15(5):755–761.

15. Okumara T, Otsuka M. 2005. Evaluation of microcrystallinityof a drug substance, indomethacin, in a pharmaceutical modeltablet by chemometric FT-Raman spectroscopy. Pharm Res22(8):1350–1357.

16. Otsuka M, Kato F, Matsuda Y. 2002. Physiochemical stabil-ity of cimetidine amorphous forms estimated by isothermalmicrocalorimetry. AAP PharmSciTech 3(4):1–13.

17. Feth MP, Volz J, Hess U, Sturm E, Hummel RP. 2008.Physicochemical, crystallographic, thermal, and spectroscopicbehaviour of crystalline and X-ray amorphous ciclesonide.J Pharm Sci 97(9):3765–3780.

18. Yang J, Grey K, Doney J. 2010. An improved kinetics approachto describe the physical stability of amorphous solid disper-sions. Int J Pharm 384:24–31.

19. Fakes MG, Blisse JV, Qian F, Desikan S, Gandhi RB, Lai C,Hsieh A, Franchini MK, Toale H, Brown J. 2009. Enhance-ment of oral bioavailability of an HIV-attachment inhibitorby nanosizing and amorphous formulation approaches. Int JPharm 370:167–174.

20. Tobyn M, Brown J, Dennis A, Fakes M, Gao Q, GambleJ, Khimyak Y, McGeorge G, Patel C, Sinclair W, TimminsP, Yin S. 2009. Amorphous drug–PVP dispersions: Applica-tion of theoretical, thermal and spectroscopic analytical tech-niques to study of a molecule with intermolecular bonds inboth crystalline and pure amorphous state. J Pharm Sci98(9):3456–3468.

21. ICH Guideline Q1A (R2), Stability testing of new drug sub-stances and products. CPMP/ICH/2736/99 and the effectivedate August 2003.

22. Avrami M. 1939. Kinetics of phase change. I General theory.J Chem Phys 7:1103–1112.

23. Avrami, M. 1940. Kinetics of phase change. II Transformation-time relations for random distribution of nuclei. J Chem Phys8:212–224.

24. Avrami M. 1941. Kinetics of phase change. III Granulation,phase change, and microstructure. J Chem Phys 9:177–184.

25. Johnson WA, Mehl RF. 1939. Reaction kinetics in processes ofnucleation and growth. Trans Am Inst Min Eng 123:416–441.

26. Maffezzoli A, Kenny JM, Torre L. 1995. On the physicaldimensions of the Avrami constant. Thermochim Acta 269/270:185–190.


Documents

Physical stability and recrystallization kinetics of amorphous ibipinabant drug product by fourier transform raman spectroscopy