Embed Size (px)
Citation preview
No
Ja
b
a
ARRAA
K1
FPAAP
1
paoiftct2
aZ
sh
0d
International Journal of Pharmaceutics 409 (2011) 62–74
Contents lists available at ScienceDirect
International Journal of Pharmaceutics
journa l homepage: www.e lsev ier .com/ locate / i jpharm
ew perspectives of 19F MAS NMR in the characterization of amorphous formsf atorvastatin in dosage formulations
iri Brusa,∗, Martina Urbanovaa, Ivana Sedenkovaa, Hana Brusovab
Institute of Macromolecular Chemistry, Academy of Sciences of the Czech Republic, Heyrovsky Sq. 2, 162 06 Prague 6, Czech RepublicZentiva K.S., U Kabelovny 130, 102 37 Prague 10, Czech Republic
r t i c l e i n f o
rticle history:eceived 13 September 2010eceived in revised form 16 February 2011ccepted 20 February 2011vailable online 26 February 2011
eywords:9F MAS NMRactor analysisolymorphismmorphous solidstorvastatinharmaceuticals
a b s t r a c t
Despite recent advances in solid-state NMR spectroscopy, the structural characterization of amorphousactive pharmaceutical ingredients (APIs) in solid dosage forms continues to be a monumental challenge.To circumvent complications following from low concentrations of APIs in tablet formulations, we pro-pose a new time-saving procedure based on chemometric approach: factor analysis of 19F MAS NMRspectra. Capability of the proposed method is demonstrated on atorvastatin – a typical representative offluorinated pharmaceutical substances exhibiting extensive polymorphism. Applying the factor analysison the recorded 19F MAS NMR spectra, unique parameters for every sample were derived. In this wayevery solid form of atorvastatin was characterized and clearly distinguishable even among various amor-phous and disordered forms. The proposed method was also found to be suitable for both qualitativeand quantitative analysis of mixtures of various forms of atorvastatin. Reliability of the proposed methodwas extensively examined by comparing the obtained results with other experimental techniques such
13
as C CP/MAS NMR, FTIR and XRPD. As highly linear correlations between the sets of parameters obtainedfrom different experimental data were found, the perspectives of the applied comparative factor analy-sis to obtain detail structural view on variability of amorphous forms of atorvastatin are also discussed.Although the reported method was tested on atorvastatin, authors expect wider application for any flu-orinated compound to give the routine, fast and reliable characterization of amorphous forms of APIsin drug products even at low concentrations (1–5%). Bear in mind that 20–25% of currently developedat lea
pharmaceuticals contain. Introduction
In pharmaceutical development, amorphous forms of activeharmaceutical ingredients (APIs) currently attract significantttention (Zakrzewski and Zakrzewski, 2006). Amorphous or low-rdered systems show in general higher solubility but less stabilityn comparison to crystalline, well-ordered systems. Big challengeor solid-state analysis is thus to detect any change of manufac-ured amorphous form of the API, because the undesired structuralhanges can affect quality (properties of drugs) and also can leado complicated patent litigations (Bauer et al., 2001; Blagden et al.,
007; Graeser et al., 2008; Greco and Bogner, 2010).In general there is a wide range of physical methods suit-ble for characterization of pharmaceutical solids (Zakrzewski andakrzewski, 2006). X-ray powder diffraction (XRPD), 13C cross-
∗ Corresponding author. Tel.: +420 296 809 380; fax: +420 296 809 410.E-mail addresses: [email protected] (J. Brus), [email protected] (M. Urbanova),
[email protected] (I. Sedenkova),[email protected] (H. Brusova).
378-5173/$ – see front matter © 2011 Elsevier B.V. All rights reserved.oi:10.1016/j.ijpharm.2011.02.030
st one fluorine atom in the molecule.© 2011 Elsevier B.V. All rights reserved.
polarization (CP) magic-angle spinning (MAS) NMR and vibrationspectroscopy are traditional tools to characterize typical well-ordered polymorphs (Brus and Jegorov, 2004; Harris, 2006, 2007;Harris et al., 2007; Husak et al., 2008, 2010), but sophisticatedanalytical tool to describe subtle differences between differentamorphous forms of a particular API is still missing. In particular, intablet formulations where concentrations of the API are very lowand the strong signals of filler compounds dominate, we are bal-ancing on physical and detection limits of conventional analyticalmethods. In these cases special techniques must be used (Griffinet al., 2007). That is why the characterization of amorphous formsof APIs and their unambiguous identification in tablet formulationsis a priority that still remains a challenge.
In this context we see the potential of multivariate analysis ofexperimental data as a possible way to bridge this gap (Graeseret al., 2008; Heinz et al., 2008, 2009), especially in a combination
with 19F MAS NMR spectroscopy (Aso et al., 2009; Harris, 2006).It has become evident that fluorinated compounds have a remark-able record in medicinal chemistry and will play an important rolein providing lead compounds for therapeutic applications (Purseret al., 2008). Nowadays approximately 20–25% of drugs in the phar-
al of Pharmaceutics 409 (2011) 62–74 63
mnostrcNaoNwb
Nsasm
dptpqfissdaipTwC
Noaoooipd
2
2
SC
2
atpeaFh
N
O
NH
CH3CH3
F
OH
OH
O-
O
Ca2+
_
J. Brus et al. / International Journ
aceutical pipeline contain at least one fluorine atom, and in 2007early 50% of 20 best selling pharmaceutical products were basedn fluorinated APIs. Consequently, the application of 19F MAS NMRpectroscopy in pharmaceutical research receives growing impor-ance. The increasing interest also follows from high gyromagneticatio and 100% isotopic abundance of fluorine atoms (19F). Just thisombination makes the 19F isotope to be one of the most sensitiveMR-active nuclei ever. The receptivity for 19F is 83% of that for 1H,nd is 4700 times higher than that for 13C. Moreover, almost nonef pharmaceutical excipients contain fluorine atoms. Therefore, 19FMR spectra can be acquired very quickly even for diluted systemsithout the danger that 19F NMR signals of the API are overlapped
y the signals of excipients.On the other hand, there are still some limitations for fluorine
MR spectra to be generally used in characterizing amorphousolids. Predominantly the number of sites occupied by fluorinetoms in the molecules of common active substances is usuallymall. 19F MAS NMR spectra also remain poorly resolved even wheneasured at high spinning rates and at high magnetic fields.A typical example of such behavior is currently the most pro-
uced drug – atorvastatin. This work, however, shows that even theoorly resolved 19F MAS NMR spectra of atorvastatin are sufficiento distinguish different amorphous forms of the given molecule,articularly when factor analysis (FA) is employed. As a conse-uence, subtle structural differences between different amorphousorms of atorvastatin are revealed and quantitative parametersdentifying every form are provided. The ability of the proposedtrategy to recognize various mixtures of amorphous forms is sub-equently shown and the detection limits of minor components areiscussed. Finally, the capability of the proposed strategy to char-cterize various low-ordered (amorphous) forms of atorvastatinn solid dosage formulations is demonstrated. In parallel, the pro-osed method is subjected to the extensive evaluation, validation.he results obtained by 19F MAS NMR spectroscopy are comparedith structurally highly receptive experimental techniques like 13CP/MAS NMR, FTIR and XRPD.
Overall, we tested potential of multivariate analysis of 19F MASMR spectra in terms of distinguishing different amorphous formsf both atorvastatin as a pure active pharmaceutical ingredient andtorvastatin as a low-concentration tablet formulation. As a resultf this testing we suppose that the proposed strategy represents notnly fast, routine and reliable tool to characterize amorphous formsf any fluorinated API in solid dosage forms, but also provides a newnsight into the systems exhibiting extensive pseudopolyamor-hism or polyamorphism for which high-resolution spectroscopicata cannot be obtained.
. Materials and methods
.1. Materials
Atorvastatin hemicalcium amorphous [(C33H35FN2O5−)2 Ca2+,
cheme 1] by Biocon Laboratories, Bangalore, India; Sucrose andorn Starch by Sigma-Aldrich were used as received.
.2. Methods: sample preparation – crystallization
Two spectroscopically well-defined highly crystalline forms andwide range of amorphous (disordered, semicrystalline) modifica-
ions of atorvastatin hemicalcium were prepared according to the
rocedures described in literature (Aronhime et al., 2006; Ayalont al., 2008; Briggs et al., 1999; Tessler et al., 2008). Followingre the examples of the preparation of atorvastatin hemicalciumorm I, Form X and Form V, respectively: (i) 90 mg of atorvastatinemicalcium as purchased was dissolved in hot water (9 ml, 70 ◦C)2
Scheme 1. Structural formula of atorvastatin.
and was allowed to stand 1 h at elevated temperature (70 ◦C) andsubsequently cooled down to room temperature. The crystals ofthe Form I were filtered out (Briggs et al., 1999). (ii) Form X wasprepared by dissolving 90 mg of atorvastatin hemicalcium as pur-chased in a mixture of hot ethanol and water (10 ml, 70 ◦C, wt.ratio 5:1) and was allowed to stand 2 h at elevated temperature(70 ◦C) and subsequently cooled down to room temperature. Thecrystals of the Form X were filtered out (Tessler et al., 2008). (iii)Form V was prepared by dissolving 90 mg of atorvastatin hemi-calcium as purchased in the mixture of hot methanol and water(10 ml, 70 ◦C, wt. ratio 4:1) and was allowed to stand 3 h at ele-vated temperature (70 ◦C) and subsequently cooled down to roomtemperature. The powders of the Form V were filtered out (Briggset al., 1999). Other modifications of atorvastatin were preparedby similar procedures by dissolving of atorvastatin hemicalciumin various solvents such as methanol, water, ethanol, isopropanol,acetone, acetonitrile and mixtures thereof in varying proportions,at elevated temperature. The prepared solutions were allowed tostand at elevated temperature (70 ◦C) in the range from 5 min toseveral hours and subsequently cooled down to room tempera-ture. The obtained powders were filtered out. In all these casesdry powdered samples were used for further spectroscopic analysiswithout any purification. In summary, approximately 100 crystal-lization procedures were tested. In this way we tried to prepare asmuch forms of atorvastatin as possible to obtain a large set of differ-ent spectra for factor analysis. From all the prepared systems, thecrystalline Forms I, V and X were unambiguously identified basedon the 13C CP/MAS NMR and XRPD experimental data (Fig. 1, Briggset al., 1999; Tessler et al., 2008). In general, however; we did not tryto assign the prepared forms of atorvastatin to the products referredin patent literature, because it was not the purpose of our study. Theunassigned amorphous modifications of atorvastatin hemicalciumare thus referred as Forms A1–An. The product directly receivedform the supplier is referred herein as Form A0.
2.3. Models of dosage forms
The models of dosage forms consisting of the prepared formsof atorvastatin (ca. 10 mg) and powdered Sucrose or Corn Starch(ca. 100–200 mg) were prepared by simple physical mixing. Theresulting concentration of atorvastatin in the prepared models oflow-dose formulations was about 1–10 wt.%.
2.4. NMR experiments
All NMR experiments were performed on a Bruker Avance 500WB/US NMR spectrometer in 4-mm and 2.5-mm double-resonanceprobeheads at carrier frequencies 500.18, 470.35 and 125.78 MHzfor 1H, 19F, and 13C nuclei, respectively. Standard cross-polarization
64 J. Brus et al. / International Journal of P
-150-125-100-75
0.0
0.2
0.4
0.6
0.8
1.0
chemical shift, ppm
50100150200
0.0
0.2
0.4
0.6
0.8
1.0
chemical shift, ppm
40353025201510
0
5000
10000
15000
20000
25000
30000
2ϑ, °
10001500300035004000
0.0
0.2
0.4
0.6
0.8
1.0
Absorb
ance
Wavenumbers, cm−1
Fig. 1. Overlay of 19F MAS NMR, 13C CP/MAS NMR, XRPD patterns and FTIR spectraof the representative modifications of atorvastatin. The red lines mark crystallineFtc
eMpedidTaOts
resents an ideal system that can be used to explore explanatory
orm I; the blue lines reflect crystalline Form X, and the black lines correspond tohe amorphous forms of the API (A0–A7). (For interpretation of the references toolor in this figure legend, the reader is referred to the web version of the article.)
xperiments were used to record 13C CP/MAS NMR spectra atAS frequency of 11 kHz (4-mm probehead, ca. 50 mg of the sam-
le). The intensities of excitation and spin-locking fields B1(13C)xpressed in frequency units ω1/2� = �B1 were 64 kHz and theuration of cross-polarization contact time pulse was 2 ms. The
ntensity of B1(1H) field of TPPM (two-pulse phase-modulated)ecoupling was ω1/2� = 89.3 kHz. The repetition delay was 4 s.he acquisition time to record 13C CP/MAS NMR spectra of pure
torvastatin with acceptable signal-to-noise ratio was about 4.5 h.vernight experiments were run to acquire 13C CP/MAS NMR spec-ra of the corresponding models of low-dose formulations. A rotorynchronized Hahn-echo pulse sequence was used to measure 19F
harmaceutics 409 (2011) 62–74
MAS NMR spectra at MAS frequency of 31 kHz (2.5-mm probehead,ca. 8 mg of the sample). The length of 90◦ (19F) pulse was 2.5 �s,repetition delay was 4 s and the number of scans was 124–256and 512–1024 for pure atorvastatin and atorvastatin in modelsof low-dose formulations, respectively. Active cooling was usedto compensate frictional heating of rotating samples (Brus, 2000).The total experimental time to acquire one 19F MAS NMR spec-trum with acceptable signal-to-noise ratio was ca. 20 and 120 min,respectively.
2.5. IR experiments
The ATR FTIR spectra were recorded on Thermo Nicolet NEXUS870 FTIR Spectrometer with a MCT detector. Single reflection ATRaccessory Golden Gate (Specac, UK) with diamond crystal wasused. Spectral resolution was 4 cm−1 with 256 scans taken for eachspectrum. Spectral region between 1700 and 2600 cm−1 stronglyaffected by absorption of diamond ATR crystal were cut off. Thespectra were normalized to compensate the variations of pressurebetween the sample and ATR crystal.
2.6. XRPD measurements
A Philips X’PERT PRO MPD PANalytical diffractometer (Almelo,The Netherlands) with CuK� radiation (wavelength: 1.542 A) wasused at 45 kV and 40 mA. Samples were scanned 4–40◦ 2� with astep size 0.01◦ 2� and a step time 150 s. Results were analyzed usingX’PERT HighScore Plus software.
2.7. Statistical analysis: factor analysis and spectral pretreatment
Factor analysis (FA) using the singular value decomposition(SVD) algorithm was performed to extract information from theexperimental data obtained by 19F MAS NMR, 13C CP/MAS NMR, IRand XRPD data, and to visualize differences between amorphousforms of atorvastatin. Processing of spectral data was performedusing Matlab program package. The following spectral range wassubjected to the factor analysis: 19F NMR: from 50 to −200 ppm,13C NMR: from 200 to −50 ppm; IR: from 650 to 4000 cm−1; XRPD:from 6 to 40◦ 2�. All spectra were base-line corrected and nor-malized. Preparation and processing of a moderately sized data setcontaining ca. 40 spectra took about 10 min.
3. Results
3.1. Primary spectroscopic data
The potential of 19F MAS NMR spectroscopy in structural char-acterization of amorphous solids is demonstrated on atorvastatin,a moderately-sized compound exhibiting extensive polymorphism(Scheme 1). From 1985 to the end of 2006 there were registered 17patents claiming 63 forms of atorvastatin (Aronhime et al., 2003,2006; Ayalon et al., 2008; Briggs et al., 1999; Gogulapati et al.,2004; Tessler et al., 2008). These modifications include severalhighly crystalline forms like Form I (Briggs et al., 1999), Form VII(Aronhime et al., 2003) or Form X (Tessler et al., 2008), and morethan 50 disordered (semicrystalline, almost-amorphous and amor-phous) forms [e.g. Form V (Briggs et al., 1999)]. As this substancecontains one fluorine atom in its molecule – the fluorine is locatedin para position of the aromatic ring (Scheme 1) – atorvastatin rep-
power of 19F MAS NMR spectroscopy.In the first step of our study, we explored structural receptiv-
ity of 19F MAS NMR spectra (Fig. 1). In order to maintain adequateclarity of the obtained results, this part of our investigation was
J. Brus et al. / International Journal of Pharmaceutics 409 (2011) 62–74 65
-140-120-100-80
0.00
-0.04
-0.08
-0.12
S2
(19
F MAS NMR)
chemical shift, ppm
Form I
Form X
Amorphous forms
-140-120-100-80
0.00
-0.02
-0.04
-0.06
-0.08S
1
(19
F MAS NMR)
chemical shift, ppm
-140-120-100-80
-0.04
0.00
0.04
0.08
0.12
0.16
0.20 S3
(19
F MAS NMR)
chemical shift, ppm
Form X Amorphous
forms
-140-120-100-80
-0.08
-0.04
0.00
0.04
S4
(19
F MAS NMR)
chemical shift, ppm
Fig. 2. Subspectra S1(19F) to S4(19F) obtained by the factor analysis of 19F MAS NMR spectra of the representative modifications of atorvastatin (see Appendix A, A1a and A1b,where singular values wj(19F) and other high-order subspectra Sj(19F) are presented).
-90 -100 -110 -120 -130
FV
A6
A4
A1
A0
A3
A2
A5
-90 -100 -110 -120 -130
-90 -100 -110 -120 -130
-0.2 0.0 0.2 0.4 0.6 0.8-0.6
-0.4
-0.2
0.0
0.2
0.4
0.6
A1
FI'
A2
A5
A3
A4
FV
FX'FX''
FI''
A6
A0
19F V
i3
19FV i
4
FX’
FI’
F tainedo ous m
cvoXFpfc
miac
ig. 3. Correlation plot of Vi3(19F) and Vi4(19F) coefficients. The coefficients were obf atorvastatin. The corresponding 19F MAS NMR spectra of crystalline and amorph
onducted on a relatively small set of representative forms of ator-astatin which did not contain filler compounds. The selected setf the samples consisted of well-ordered crystalline Forms I and; several amorphous forms prepared by us (thereafter referred asorm V and Forms A1–A7); and the sample received from the sup-lier (Form A0). For the purpose of this paper, the term “amorphousorm” is used to refer both amorphous and disordered (partiallyrystalline) forms of atorvastatin.
In the recorded 19F MAS NMR spectra (Fig. 1) the crystallineodifications of atorvastatin (Form I and Form X) can be easily
dentified and distinguished from the amorphous ones (Forms Vnd A1–A7). This follows from the fact that the highly orderedrystalline forms produce narrow and well-resolved NMR signals,
by the factor analysis of 19F MAS NMR spectra of the representative modificationsodifications are placed on the left and right side, respectively.
while the amorphous ones give a single broad NMR signal. Theobtained experimental data also reflect slight structural variationsbetween the amorphous products of atorvastatin (Fig. 1). How-ever, the spectral differences between them are not easy to findand quantitatively describe. Although this is the common prob-lem of all spectroscopic data including XRPD, FTIR and 13C CP/MASNMR (Fig. 1), the 19F MAS NMR spectra exhibit remarkably weak-est spectral differences. As the spectral differences between the
amorphous modifications of atorvastatin are very subtle, visualinspection of 19F MAS NMR spectra is not enough. Correct and rig-orous interpretation of the observed variations of the 19F MAS NMRspectra thus requires special statistical processing based on factoranalysis.
66 J. Brus et al. / International Journal of Pharmaceutics 409 (2011) 62–74
A0
-90 -100 -110 -120 -130
A0 (ns=512)
A0 (ns=384)
A0 (ns=256)
A0 (ns=128)
A0 ( ns= 48)
A0 (ns=100)
0.80.60.40.20.0-0.2
-0.4
-0.2
0.0
0.2
0.4
A1
FI'
A2A7
A5
A3
A10
A12
A4
FV
FX'
A6
19F V
i3
19FV i
4
-0.0402
-0.0556
-0.0048 -0.0172
Fig. 4. Correlation plot of Vi3(19F) and Vi4(19F) coefficients. The coefficients were obtained by factor analysis of the extended set of 19F MAS NMR spectra of the representativemodifications of atorvastatin. The 19F MAS NMR spectra of amorphous Form A0 recorded repeatedly over one week are placed on the right side. The expanded region of thecorrelation plot reflecting scatter of Vij coefficients of A0 samples (read dots) is placed in the upper right corner.
-90 -100 -110 -120 -130
A0:FX (99:1)
A0
A0:FX (95:5)
A0:FX (80:20)
A0:FX (50:50)
A0:FX (10:90)
FX
0.40.20.0-0.2-0.4-0.6-0.5
-0.4
-0.3
-0.2
-0.1
0.0
0.1
0.2
0.3
0.4
0.5
A1
FI'
A2
A7
A5
A3
A10
A12
A6
FV
A14
A4
19FV i
4
19F V
i3
A0
FX''
F ined bf s cono the re
3
e(1isa
Y
i
ig. 5. Correlation plot of Vi3(19F) and Vi4(19F) coefficients. The coefficients were obtaorms of atorvastatin. Red dots correspond with the crystalline–amorphous mixturen the right side. (For interpretation of the references to color in this figure legend,
.2. Factor analysis
In general, factor analysis (FA) provides a versatile tool toxplore complex changes in large sets of experimental dataClaydenaq et al., 1997; Eads et al., 2004; Henry and Hofrichter,992; Koonst and Ellis, 1995; Levine, 1998). Specifically, the exper-
mental spectra Yi are converted into the set of orthonormalubspectra Sj (Eq. (1)) using singular value decomposition (SVD)lgorithm.
n∑
i =j=1
wjVijSj (1)
The calculated subspectra are linear combinations of the exper-mental data (n – number of experimental spectra) and vice versa
y the factor analysis of the extended set of the 19F MAS NMR spectra of representativesisting of Form X and Form A0. The corresponding 19F MAS NMR spectra are placedader is referred to the web version of the article.)
the experimental data can be given as the linear combination of thesubspectra (Eq. (1)). Each subspectrum Sj represents a specific spec-tral feature that is typical for a given type of analyzed samples. Thestatistical importance and hence the order of each subspectrum Sj isexpressed by the singular value, wj. Significantly, the ability of a par-ticular subspectrum Sj to describe the experimental spectrum Yi isexpressed by the normalized coefficient Vij. Consequently the coef-ficients Vij (i.e. scores) represent quantitative parameters reflectingspectral differences between the analyzed samples. In this way anymodification of a particular API can be unambiguously identified
via the set of Vij coefficients. In this way, the large set of experi-mental data can be reduced into just several (m, m < n) essentialfactors (subspectra) that are sufficient to appropriate description ofthe experimental data series when only the subspectra associatedwith the highest singular values (wj) are taken to approximation.
J. Brus et al. / International Journal of Pharmaceutics 409 (2011) 62–74 67
-90 -100 -110 -120 -130
0.80.60.40.20.0-0.2
-0.4
-0.2
0.0
0.2
0.4
19F V
i3
19FV i
4
A3
A6
20:80 (A6:A3)
70:30 (A6:A3)
50:50 (A6:A3)
10:90 (A6:A3)
90:10 (A6:A3)
A3
A6
20:80 (A6:A3)
50:50 (A6:A3)
70:30 (A6:A3)
1.21.11.00.90.80.70.60.50.40.30.20.10.0
-0.2
-0.1
0.0
0.1
0.2
0.3
0.4
20:80 (A6:A3)
70:30 (A6:A3)
50:50 (A6:A3)
10:90 (A6:A3)
90:10 (A6:A3)
A3
A6
wt. % of A6
19FV i
4
Fig. 6. Correlation plot of Vi3(19F) and Vi4(19F) coefficients. The coefficients were obtained by the factor analysis of the extended set of the 19F MAS NMR spectra of ther orphN comp( rred to
3
1
sutlsSnSlt3Fib
epresentative modifications of atorvastatin. Red dots correspond to the scores of amMR spectra are placed on the right side. The dependence of Vi4(19F) factors on the
For interpretation of the references to color in this figure legend, the reader is refe
.3. Factor analysis of 19F MAS NMR spectra
Applying this statistical data treatment procedure, the recorded9F MAS NMR spectra of atorvastatin were analyzed and the set ofubspectra Sj(19F) together with their corresponding singular val-es wj(19F) were derived. As only the singular parameters w1(19F)o w4(19F) reached reasonable values (Appendix A and A1a) the ana-yzed dataset is almost completely described by the correspondingubspectra S1(19F) to S4(19F) (Fig. 2). Other high-order subspectra5(19F) to Sn(19F) then reflect random variations in the electronicoise. Interpreting the obtained results, the 1st-order subspectrum1(19F) shows the mean signal representing the whole set of ana-yzed samples. In our particular case this mean signal resembles
ypical signal of amorphous forms of atorvastatin. The 2nd- andrd-order subspectra then depict specific markers of crystallineorms I and X. The 3rd-order subspectrum S3(19F) can also benterpreted as a difference spectrum reflecting spectral differencesetween the crystalline forms of atorvastatin. The spectral regionsous–amorphous mixtures consisting of Form A6 and Form A3. The relevant 19F MASosition of analyzed mixtures (Form A6/Form A3) is depicted on the bottom chart.the web version of the article.)
in which the amorphous forms of the API differ from each other arehighlighted in the 4th-order subspectrum S4(19F).
As follows from the above-mentioned interpretations, struc-tural differences between the crystalline forms of atorvastatin canbe explicitly expressed by the factors Vi3(19F), while the subtle dif-ferences between the amorphous forms are preferably describedby Vi4(19F) coefficients. In a graphical representation (one of themost suitable representation seems to be the correlation plot of Vi3and Vi4 coefficients, Fig. 3) the crystalline Forms I and X are clearlyseparated in the horizontal dimension, while the amorphous forms(A0–A7) are vertically sorted. The observed systematic array ofVi4(19F) coefficients clearly reflects differences between the amor-phous forms of atorvastatin. Moreover, the considerably wide range
of Vi4(19F) values covering the interval form −0.5 to 0.5 confirmsthe high structural receptivity of this parameter.In order to verify reproducibility of the obtained results, themeasurements of 19F MAS NMR spectra including their process-ing and SVD analysis were repeated. Over a short period the
68 J. Brus et al. / International Journal of Pharmaceutics 409 (2011) 62–74
0.30.20.10.0-0.1-0.2-0.3-0.4-0.5
-0.6
-0.4
-0.2
0.0
0.2
0.4
A0
A1
FI'
A5
A7
A3
A8
A10
A11A13
FV
A14
FX'FX''
FI''
A16
A6
A15
A4
A2y=-0,0003+0,976x
R2=0,958
19F V
i4 ato + excipient
19FV i
4 a
to
-90 -100 -110 -120 -130
0.80.60.40.20.0-0.2-0.4
-0.4
-0.2
0.0
0.2
0.4
A3
A13A0
A10A14A6
A16
A2
A15
FV
A8
A7
A11
A1
FX'
FI''FI'
FX''
A5
19F V
i3
19FV i
4
A5
A2
A3
A0
A1
A10
A6
FV
F tainea f crysc spond
1
tmrNmanodopsap
ig. 7. Correlation plot of Vi3(19F) and Vi4(19F) coefficients. The coefficients were obtorvastatin in mixtures with excipients. The corresponding 19F MAS NMR spectra oorrelation of Vi4(19F) parameters obtained for pure forms of the API and their corre
9F MAS NMR spectra of selected samples were recorded severalimes at the same experimental conditions. In addition, to esti-
ate the influence of electronic noise on the quality of the obtainedesults the number of scans accumulated to acquire one 19F MASMR spectrum was systematically varied. The obtained experi-ental data were subsequently processed by the factor analysis
nd Vi3(19F)–Vi4(19F) correlation plots were created. In both testso changes in relative displacement of Vij(19F) parameters werebserved in the resulting correlation plots. The calculated standardeviation of Vi3(19F) scores was X = 4.23 × 10−3 and Y = 9.36 × 10−3
f Vi4(19F) parameters resulting from 6 repetitions for a given sam-le (Fig. 4). Therefore it is clear that the proposed procedure isufficiently robust and stable, and different forms of atorvastatinre unambiguously identified if the differences in their Vij(19F)arameters exceed ca. two-times the above-mentioned standard
d by the factor analysis of 19F MAS NMR spectra of representative modifications oftalline and amorphous forms are placed on the left and right side, respectively. Theing low-dose model formulations.
deviations. In addition, new measurements needed for the repro-ducibility test significantly enlarged the dataset of processed 19FMAS NMR spectra up to 50 analyzed forms of atorvastatin in total.When we compare the original data set (Fig. 3) and the new datasetincluding multiple measurement of one specie (e.g. Fig. 4), theresulting pattern of the obtained Vi3(19F)–Vi4(19F) correlation plotsremains practically unchanged. This indicates that factor analysisof 19F MAS NMR spectra provides self-consistent results that arebasically independent on the number of processed experimentaldata and multiple presence of one amorphous form.
3.4. Two-component mixtures and detection limits
For several reasons, including stability and patenting issues, it isimportant to be able to ensure that, whichever solid-state form is
J. Brus et al. / International Journal of Pharmaceutics 409 (2011) 62–74 69
200 160 120 80 40 0
200 160 120 80 40 0
200 160 120 80 40 0
FV
A6
A4
A1
A0
A3
A2
A5 FX
FI
0.80.60.40.20.0-0.2
-0.4
-0.2
0.0
0.2
0.4
A0
A1
FI'
A2
A7
A5
A3
A8
A9
A10
A11A12A13
A4
FV
A14
A15
FX' FX''
FI''A16
A6
13C
13CV i
4
F btainea ous f
cspcwtwpXAatattatVd
sscrosittt
3a
pptAmcaq
ig. 8. Correlation plot of Vi3(13C) and Vi4(13C) coefficients. The coefficients were otorvastatin. The corresponding 13C CP/MAS NMR spectra of crystalline and amorph
hosen, this form is not transformed or contaminated by other solidtate forms. That is why an important question that often bothersharmaceutical companies concerns the detection limits of minoromponents in the mixtures of various polymorphs. In this contexte prepared and analyzed a wide range of two-component sys-
ems (crystalline–amorphous and amorphous–amorphous) in theeight ratio ranging from 1:99 to 99:1. It was found out that theresence of 1% of crystalline form of atorvastatin (e.g. Form I, Form) dispersed in the amorphous matrix of atorvastatin (e.g. Forms0–A6) can be easily detected. The recorded 19F MAS NMR spectras well as the calculated Vi3(19F) or Vi4(19F) coefficients differ fromhose obtained for pure forms (Fig. 5). On the other hand, the lowestmount of an amorphous form of atorvastatin (e.g. Forms A0–A6)hat can be clearly detected in the crystalline matrix of atorvas-atin (Forms I and X) is higher than in the previous case, and it ist least 5–10%. This is the lowest amount of the amorphous frac-ion of atorvastatin that induces distinct changes in Vi3(19F) andi4(19F) coefficients (Fig. 5). Smaller amounts are under the limit ofetection (LOD).
The situation is a bit more complicated in the mixtures con-isting of two amorphous forms of the API. It was found out that inome cases just the presence of approximately 5–10 wt.% of a minoromponent can induce changes in 19F MAS NMR spectra that can beecognized by the factor analysis. It is worth noting that the valuesf Vi4(19F) coefficients representing the two-component mixturesystematically move towards the center of the region that is lim-ted by Vi4(19F) scores of both pure forms of atorvastatin (Fig. 6). Ashe obtained scores almost linearly vary with the composition ofhe analyzed mixtures, the calibration curves describing particularwo-component systems can be also created.
.5. Effect of filler compounds in low-dose model formulations oftorvastatin
So far, the proposed method has been tested on the systems ofure atorvastatin without filler compounds. The real application inharmaceutical research, however, strongly depends on the abilityo detect, identify and distinguish different amorphous forms of the
PI in low-dose tablet formulations. That is why a large collection ofodels of low-dose formulations (mixtures with filler compounds)ontaining approximately 1–5 wt.% of atorvastatin was preparednd a new dataset of 19F MAS NMR spectra was created. Subse-uently, the recorded 19F MAS NMR spectra of these systems were
Vi3
d by the factor analysis of 13C CP/MAS NMR spectra of the representative forms oforms are placed on the left and right side, respectively.
processed by the factor analysis. The resulting subspectra as wellas characteristic features of Vi3(19F)–Vi4(19F) correlation plot arequite comparable with the previous results keeping thus distinctvertical ordering of the amorphous forms and horizontal separa-tion of crystalline forms of atorvastatin (compare Fig. 7 and Fig. 4).The high degree of congruence between these two sets of exper-imental data is demonstrated on the correlation plot of Vi4(19F)scores obtained for pure forms of atorvastatin and the correspond-ing models of dosage formulations (Fig. 7, bottom chart). The onlyexception observed for Form A1 can be explained by extremelylow concentration of the API in the model system that was below0.5 wt.%.
3.6. Comparison with other experimental techniques
Although the above-introduced results have demonstrated theability of the proposed method to distinguish different amorphousforms of atorvastatin in pure state as well as in the mixtures withfiller compounds, the relation between the observed systematicvariation of Vi4(19F) coefficients and structural changes or molecu-lar arrangement in the amorphous phase was not discovered. Thatis why we tried to examine capability of 19F MAS NMR spectra toreflect these structural changes in detail. The evaluation processused here is based on the explicit comparison of 19F MAS NMRspectra with structurally highly receptive experimental data pro-vided by 13C CP/MAS NMR, FTIR and XRPD (Fig. 1). These techniqueswere selected because XRPD data reflect supermolecular structureand long-range periodic order; 13C CP/MAS NMR spectra provideinformation about short-range arrangement and overall molecu-lar conformation; and FTIR spectra are sensitive to the changes inmolecular structure of the investigated system. However, due to thestrong signal overlap and unacceptable experimental time requiredto record spectra with reasonable signal-to-noise ratio all the mea-surements were conducted only for the samples of atorvastatinwithout filler compounds.
In general, there is a limited number of procedures that allowindependent comparison of different spectroscopic data. As one ofthem we propose comparative factor analysis, the method based
on the correlation of Vij scores obtained by factor analysis of indi-vidual datasets. Using this procedure, the sets of 13C CP/MAS NMRspectra, FTIR spectra and XRPD patterns are separately processedat first. Subsequently, the obtained Vij scores are cross-correlatedto search for the clear relationships. Specifically, the factor analysis
7 al of Pharmaceutics 409 (2011) 62–74
oeptsmfpsatpXiia
4
4
ap2nhspoapsdiis
vsrFSaMt(oMa
4
noamEnsUtlo
0.40.30.20.10.0-0.1-0.2-0.3-0.4
-0.4
-0.2
0.0
0.2
0.4
A0
A1
A2
A7
A5
A3
A8A9
A10
A11
A12
A13
A4
FV
A14
A15
A16
A6
y=0,010+0,959x
R2=0,927
13C V
i4
19FV i
4
0.40.30.20.10.0-0.1-0.2-0.3-0.4
-0.4
-0.2
0.0
0.2
0.4
A0
A1
A2
A7
A5
A3
A8A9
A10
A11
A12
A13
A4
FV
A14
A15
A16
A6
y=0,008+1,006x
R2=0,872
FTIR Vi4
19FV i
4
0.40.20.0-0.2-0.4
-0.4
-0.2
0.0
0.2
0.4
A0
A1
A2
A7
A5
A3
A8A9
A10
A11
A12
A13
A4
FV
A14
A15
A16
A6
RTG Vi7
y=0,038+0,924x
R2=0,820
19FV i
4
0 J. Brus et al. / International Journ
f 13C CP/MAS NMR spectra resulted in nearly the same results asxtracted from 19F MAS NMR data and the obtained correlationlot of Vi3(13C) and Vi4(13C) coefficients provides clear distinc-ion of the prepared drug forms (Fig. 8). We must note that thecores derived form 19F MAS NMR spectra are in perfect agree-ent with the scores extracted form 13C CP/MAS NMR spectra. This
act is clearly demonstrated in Vi4(13C)–Vi4(19F) cross-correlationlot (Fig. 9, upper chart). Similarly we analyzed other spectro-copic data, and linear relationships were found also for Vi4(19F)nd Vi4(FTIR); and Vi4(19F) and Vi7(XRPD) coefficients (Fig. 9, cen-ral and bottom chart, respectively). Overall, we performed allossible cross-correlations including entire Vij(19F; 13C; FTIR andRPD) scores, but only those involving Vi4(19F) and those exhibit-
ng direct proportion are reported in this contribution. Supportingnformation including interpretation of the obtained subspectrand singular values can be found in Appendix A.
. Discussion
.1. Structural variability of amorphous atorvastatin
Several studies have revealed the existence of differentmorphous forms of pharmaceutical solids depending on thereparation technique (Graeser et al., 2008; Greco and Bogner,010; Craig et al., 1999; Heinz et al., 2008; Yu, 2001). So it isot surprising that the different amorphous forms of relativelyigh-molecular-weight atorvastatin containing several flexibleide-chains were simply prepared by different solvent-evaporationrocesses. Moreover, the differently prepared amorphous formsf atorvastatin, that is still under extensive consideration (Annd Sohn, 2009), exhibit significantly different physicochemicalroperties. Quite recently it has been reported that the intrinsic dis-olution rates of different amorphous forms of this API considerablyiffer from 0.183 to 0.252 mg min−1 cm−2 (Shete et al., 2010). That
s why the experimental approaches of exact structural character-zation of different solid state forms are still a subject of enormouscientific effort.
The differences between the prepared amorphous forms of ator-astatin can be hardly recognized using conventional physical andpectroscopic techniques. This fact is clearly demonstrated in theecorded experimental spectra (19F MAS NMR, 13C CP/MAS NMR,TIR and XRPD, Fig. 1) which exhibit only weak spectral variations.pecifically, the structural differences of the amorphous phase oftorvastatin have the smallest impact of on spectral pattern of 19FAS NMR spectra. This follows from the fact, that XRPD, FTIR spec-
roscopy and 13C CP/MAS NMR probe the whole molecular systemi.e. positions of all atoms, vibration modes of all chemical bondsr global structure of carbon skeleton, respectively), while the 19FAS NMR spectra are reflecting only the local structural changes
round the site occupied by the fluorine atom.
.2. Distinguishing of amorphous forms of atorvastatin
Recent studies meaning the ability of different analytical tech-iques to characterize differently prepared amorphous formsf pharmaceutical solids have shown that the most powerfulpproach is based on the combination of spectral methods withultivariate analysis (Graeser et al., 2008; Heinz et al., 2008).
specially Raman spectroscopy combined with principle compo-ent analysis (PCA) is being the most sensitive technique towards
tructural differences in amorphous samples (Heinz et al., 2008).nfortunately, the PCA of Raman spectra can be hardly used forhe low-dosage products of atorvastatin due to the strong over-ap of characteristic vibration bands of the API with the signalsf excipients and very low signal-to-noise ratio. That is why we
Fig. 9. Cross-correlation plots of Vi4(19F); Vi4(13C); Vi4(IR) and Vi7(XRPD) coefficientscharacterizing amorphous forms of atorvastatin. Upper chart: cross-correlation ofVi4(19F) and Vi4(13C); central chart: cross-correlation of Vi4(19F) and Vi4(FTIR); andbottom chart: cross-correlation of Vi4(19F) and Vi7(XRPD) scores.
applied factor analysis on highly sensitive 19F MAS NMR spectrathat can be recorded within short time and in reasonable quality
also for the low-dose formulations with atorvastatin. Specificallywe used SVD algorithm that has significant advantage over otherapproaches. Predominantly the applied algorithm produced notonly scores characterizing each analyzed sample, but also the sub-spectra separating specific spectral aspects which can be correlated
J. Brus et al. / International Journal of Pharmaceutics 409 (2011) 62–74 71
F f repra
wscBaoatt(
oipMtcpfpoatt
ig. 10. Selected 19F MAS NMR, 13C CP/MAS NMR, XRPD patterns and FTIR spectra occording to the calculated Vi4(19F) coefficients.
ith structural changes (Fig. 2). Then according to the obtainedubspectra it was easy to find the most suitable scores allowinglear distinction between various amorphous forms of atorvastatin.ecause the structural differences between amorphous forms oftorvastatin are relatively subtle then just the scores with a highrder can reflect these differences. In our particular case thosere 4th-order scores Vi4(19F) which in the most suitable correla-ion with 3rd-order scores Vi3(19F) allow the differences betweenhe amorphous forms to be clearly observed in vertical dimensionFig. 3).
All the performed subsequent tests, including analysis of vari-us two-component mixtures as well as estimation of uncertaintyn the calculated scores, confirmed sufficient reliability of the pro-osed approach. Even though the differences in the recorded 19FAS NMR spectra are visually barely perceptible, utilizing the fac-
or analysis the different amorphous forms of atorvastatin can belearly distinguished via the calculated scores. Moreover, the pro-osed method allows recognizing the mixtures of the preparedorms of atorvastatin with the detection limit of the crystalline
hase impurities in amorphous phase less than 1%. The detectionf minor fraction of amorphous phase impurities in the matrix ofnother amorphous form of atorvastatin is, however, much morericky. The 19F MAS NMR signals of both components are so similarhat neither the computer decomposition of 19F MAS NMR spectraesentative modifications of atorvastatin. The spectra are assorted in the same order
does allow clear identification and quantification of these mixtures.The factor analysis, however, is sufficiently powerful to provide rea-sonable results also in this case, and almost invisible changes in theshape of 19F MAS NMR spectra of the mixtures can be converted tothe consistent changes of the corresponding Vi4(19F) scores with thelowest detection limit being about 5–10%. Admittedly this detec-tion limit is not reached generally, because the differences betweenVi4(19F) scores of the pure modifications of atorvastatin formingthe mixture must be sufficiently large. For instance, this precondi-tion is fulfilled for the systems consisting of Forms A3 and A6; orForms V and A5 (Fig. 4). In the opposite cases, when the differencesbetween the scores of pure substances are considerably smaller, thedetection limits of minor components correspondingly increase.
From our experience built on the extensive testing, the best wayhow to utilize the proposed procedure starts with the creation of areference database of 19F MAS NMR spectra of pure forms of a givenAPI. Such database should include representative amorphous aswell as crystalline modifications. Approximately 10–15 forms of theparticular API seem to be enough. When a new sample is analyzed,
its 19F MAS NMR spectrum can be added into the already madedataset. Subsequently performed factor analysis then characterizesthis sample via its Vij(19F) scores. The reference database shouldbe kept constant or nearly constant for all analysis of unknownsamples.
72 J. Brus et al. / International Journal of Pharmaceutics 409 (2011) 62–74
F e setss re hig
bfitsfm1
spse
faN1ftttfc2
4
ortc
ig. 11. Subspectra S4(13C), S4(FTIR) and S7(XRPD) calculated from the representativignals assigned to the molecular sites exhibiting substantial structural variability a
Similarly, the reference database of different polymorphs muste created also for the API in dosage formulations. Although theller compounds are thought to be inert, weak interactions withhe molecules of the API cannot be excluded. In our case we noticedlight systematic deviation in Vij(19F) scores obtained for pureorms of atorvastatin from those obtained for the corresponding
ixtures with excipient (Fig. 7). Close inspection of the recorded9F MAS NMR spectra revealed slight systematic shift of 19F NMRignals in model mixtures towards higher frequencies. With highrobability, this phenomenon can be attributed to the changes inample susceptibility, because in dosage forms the dielectric prop-rties of the samples are dominated by the excipient.
Despite these problems the proposed method provides routine,ast and reliable tool to distinguish different amorphous forms oftorvastatin even in low-dose tablet formulation. Applying 19F MASMR spectroscopy the high-quality spectra can be recorded within0–30 min. Consequently the total time to identify an unknownorm of atorvastatin in the dosage form is less than 1 h. Thereforehe proposed method is considerably timesaving compared withraditionally used 13C CP/MAS NMR. As recently demonstrated evenhe one-day acquisition of 13C CP/MAS NMR spectra of low-doseormulations of levosimendan (<1%) did not allow the clear identifi-ation of the present crystalline modification (Virtanen and Maunu,010).
.3. Structural receptivity of 19F MAS NMR spectra
In general, the observed systematic variations of the scoresbtained by multivariate analysis of 19F MAS NMR spectra canesult from many reasons including changes in crystallinity, struc-ural transformations, presence of impurities etc. In order to findlear relationships between the molecular structure of atorvas-
of experimental data of various modifications of atorvastatin. The “high-amplitude”hlighted.
tatin and the observed systematic variation of Vi4(19F) scores, theobtained experimental data were compared with the results ofother spectroscopic techniques. Specifically, we cross-correlatedVi4(19F) scores with the coefficients obtained by the factor anal-ysis of 13C CP/MAS NMR, FTIR and XRPD experimental data. A bitsurprisingly we found almost perfect linear dependences betweenVi4(19F) coefficients and Vi4(13C), Vi4(IR), and Vi7(XRPD) scores.In addition, the absolute values of slopes of these dependencesare near to 1 (Fig. 9). This indicates that these scores are ofthe same significance and the corresponding subspectra (Si4(19F),Si4(13C), Si4(FTIR), Si7(XRPD)) are strongly interrelated, describingthus inherent properties of the amorphous phase of atorvastatin.
In general, 19F MAS NMR, 13C CP/MAS NMR as well as FTIRspectroscopy all provide site-specific information on the localmolecular structure. Therefore it is not surprising that given struc-tural changes of the amorphous phase can be found at the samelevel, i.e. in the subspectra of the same order. On the other hand,the information which is in the case of NMR and FTIR measure-ments reproduced in the 4th-order subspectra, is shifted up to the7th-order subspectrum of XRPD dataset. With high probability thisphenomenon can be explained by the fact that XRPD is primarilysensitive on structural changes of crystalline solids. Consequentlyone can assume that Si2 to Si6(XRPD) subspectra reflect complexrelationships between different crystalline forms of atorvastatin.Weak changes in amorphous phase are then reflected in the higher-order subspectra Si7 to Si9(XRPD) (Appendix A, A3a and A3b).
4.4. Factor analysis and molecular structure of atorvastatin
As the obtained 19F MAS NMR spectra can be only hardlyinterpreted in the terms of molecular arrangement of amorphousatorvastatin, the site-specific information on molecular structure
al of P
mstttNoa1tofoleonrtto
ssmaaiocpoatbyttfgibta1drbmmaspba
5
ciefor
An, S.-G., Sohn, Y.-T., 2009. Crystal forms of atorvastatin. Arch. Pharm. Res. 32,
J. Brus et al. / International Journ
ust be deduced from complementary 13C CP/MAS NMR, and FTIRpectra and XRPD patterns (Fig. 10). In the 19F MAS NMR spec-ra assorted in the descending order according to Vi4(19F) scoreshe NMR signals exhibit gradual narrowing and symmetrization ashe left-hand shoulder disappear. In the corresponding 13C CP/MASMR spectra these changes are associated with the lost of res-lution of methyl signals (ca. 20 ppm), signals of –CH units inromatic rings (ca. 120 ppm) and CO–NH– carbonyl signals (ca.60 ppm). The same molecular sites exhibit weak but still percep-ible spectral changes also in the corresponding FTIR spectra. Thesebservations thus indicate that structural variability of amorphousorms of atorvastatin is associated with the conformational changesccurring predominantly around these molecular sites. Besides theocal structural changes the recorded XRPD data indicate consid-rable changes in the molecular long-range periodic arrangementf the amorphous phase. In the same order as 19F MAS NMR sig-als become symmetrical and narrow, the sharp residual X-rayeflections gradually diminish. This indicates that primary struc-ural motifs of long-range periodic arrangement of the molecules ofhe API in amorphous phase play considerable role in the formationf amorphous forms of atorvastatin (Fig. 10).
Detail structural information can be directly derived from theubspectra S4(13C), S4(FTIR) and S7(XRPD) (Fig. 11). The signalsignificantly enhanced in their relative amplitudes clearly reflectolecular sites or typical interatomic distances that are most
ffected by the structural changes in the amorphous phase oftorvastatin. For instance, in S4(13C) subspectrum such relativencrease in signal intensity is clearly apparent for the signals res-nating at ca. 160, 120 and 20 ppm. Oppositely, the signals ata. 45, 70 and 180 ppm are relatively suppressed. This confirmsrevious statement that structural changes in amorphous phasef atorvastatin are basically connected with local rearrangementround amide carbonyl and neighboring methyl groups, whilehe hydroxyl-substituted side-chains and carboxyl unit seem toe structurally unaffected. The great advantage of factor anal-sis is clearly demonstrated in S4(FTIR) subspectrum in whichhe relative signal enhancement is observed not only for vibra-ion bands of CO–NH groups (ca. 3200 and 1700 cm−1) but alsoor methyl units (ca. 3000 cm−1). Moreover, the above-discussedlobal molecular rearrangement of the molecules of atorvastatinn amorphous phase and the formation of partially ordered assem-lies is clearly apparent in S7(XRPD) subspectrum. As reflected byhe high-amplitude diffraction peaks at 8◦ and 20◦ 2�, the aver-ge correlation length typical for these assemblies is about 4 and1 nm, respectively. Roughly estimated, the mentioned correlationistances can be attributed to some structural motifs of aromaticings probably involved in �–� interactions. Consequently it cane supposed that the amorphous phase of atorvastatin consist ofolecular assembles which build-up is associated with rearrange-ent of amide groups and stacking of aromatic rings. The individual
morphous forms of atorvastatin then can differ in the extent ofuch arrangement. However, detail interpretation of the observedhenomena that can be derived from the application of pair distri-ution function transforms is behind the scope of this contributionnd it is currently under investigation.
. Conclusion
Due to the major success of fluorinated compounds in medicinalhemistry, it may be predicted that the number of fluorine contain-
ng drugs on the market will continue to increase. Similarly we canxpect growing interest in the formulation of APIs in amorphousorms including development of solid solutions and dispersionsf APIs in polymer matrix for which, however, traditional high-esolution spectral data can be hardly obtained. Consequently anyharmaceutics 409 (2011) 62–74 73
attempt to develop a method of detail structural characterizationof these amorphous low-dose formulations of APIs is worth of theeffort.
In this contribution it was demonstrated that 19F MAS NMRspectroscopy combined with factor analysis (SVD algorithm) offersa fast, routine and reliable tool to identify disordered and amor-phous forms of atorvastatin (representative of fluorine-containingAPIs) in both substance itself and low-dose tablet formulations.Specifically, it was shown that even relatively poorly-resolved 19FMAS NMR spectra can be used to detect subtle differences betweenvarious amorphous forms of atorvastatin. Applying the factor anal-ysis on 19F MAS NMR spectra, the sample-specific coefficients(scores, Vij(19F)) characterizing every product can be obtained. Ina graphical representation these scores provide clear identifica-tion and distinction of different amorphous forms of atorvastatin.The proposed method also offers the ability to identify and quan-tify various mixtures of different modifications of atorvastatin.Predominantly, however, the proposed method allows the fastcharacterization of low-dose pharmaceutical products. Applying19F MAS NMR spectroscopy the acceptable spectra of tablet formu-lations containing ca. 1–5% of atorvastatin can be recorded within10–30 min. Consequently the total time to identify an unknownform of atorvastatin in the dosage form is less than 1 h. In con-clusion, the proposed method has at least the ability to revealdeviations from the expected output and thus to unambiguouslyidentify undesired products of manufacture.
Extensive testing of the method based on the comparison with13C CP/MAS NMR, FTIR and XRPD data confirmed that the computedVij(19F) scores can be considered as structure-related parametersdescribing changes in molecular arrangement of atorvastatin. Thisindicates that 19F MAS NMR spectra can in principle reflect changesin amorphous phase of organic (fluorine-containing) compounds insimilar extent as usually provided by the more informative tech-niques like 13C CP/MAS NMR, FTIR and XRPD. We note that isotropicNMR shifts are by far the easiest solid-state NMR parameter to mea-sure. That is why the proposed method that does not need NMRsignal assignment can be applied to other high-sensitive nuclei like31P, 23Na, 11B etc. While absent in filler compounds these nucleiare common components of many pharmaceutical compounds.Our work thus extends the application of solid-state NMR spec-troscopy in the field of qualitative and quantitative characterizationof amorphous pharmaceutical solids in low-dose formulations. Theexcellent results presented here have encouraged further work totest the general applicability of the proposed method to describestructural motifs in amorphous phase pharmaceutical solids, whichis underway.
Acknowledgement
Funding from Ministry of Education, Youth and Sports of theCzech Republic (research program 2B08021) is acknowledged.
Appendix A. Supplementary data
Supplementary data associated with this article can be found, inthe online version, at doi:10.1016/j.ijpharm.2011.02.030.
References
933–936.Aronhime, J., Lidon-Hadas, R., Niddam, V., Lifshitz, R., 2003. Atorvastatin hemi-
calcium form VII. US Patent 6605636 (12.8.2003).Aronhime, J., Lifshitz, R., Niddam, V., Lidon-Hadas, R., Wizel, S., 2006. Novel crystal
forms of atorvastatin hemi-calciun and processes for their preparation as wellas novel processes for preparing other forms. US Patent 7144916 (5.12.2006).
7 al of P
A
A
B
B
B
B
B
C
C
E
G
G
G
G
H
4 J. Brus et al. / International Journ
so, Y., Yoshioka, S., Miyazaki, T., Kawanishi, T., 2009. Feasibility of 19F-NMR forassessing the molecular mobility of flufenamic acid in solid dispersions. Chem.Pharm. Bull. 57, 61–64.
yalon, A., Levinger, M., Royblat, S., Niddam, V., Lifshitz, R., Aronhime, J., 2008.Polymorphic form of atorvastatin calcium. US Patent 7411075 (12.8.2008).
auer, J., Spanton, S., Henry, R., Quick, J., Dziki, W., Porter, W., Morris, J., 2001. Rito-navir: an extraordinary example of conformational polymorphism. Pharm. Res.18, 859–866.
lagden, N., de Matas, M., Gavan, P.T., York, P., 2007. Crystal engineering of activepharmaceutical ingredients to improve solubility and dissolution rate. Adv. DrugDeliv. Rev. 59, 617–630.
riggs, C.A., Jennings, R.A., Wade, R., Harasawa, K., Ichikawa, S., Minohara, K., Nask-agawa, S., 1999. Warner Lambert Co. US Patent 5969156 (19.10.1999).
rus, J., 2000. Heating of samples induced by fast magic-angle spinning. Solid StateNucl. Magn. Reson. 16, 151–160.
rus, J., Jegorov, A., 2004. Through-bonds and through-space solid-state NMR cor-relations at natural isotopic abundance: signal assignment and structural studyof simvastatin. J. Phys. Chem. A 108, 3955–3964.
laydenaq, N.J., Lehnertb, R.J., Tumocka, S., 1997. Factor analysis of time domain NMRdata: crystallinity of poly(tetrafluoroethylene). Anal. Chim. Acta 344, 261–269.
raig, D.Q.M., Royall, P.G., Kett, V.L., Hopton, M.L., 1999. The relevance of the amor-phous state to pharmaceutical dosage forms: glassy drugs and freeze driedsystem. Int. J. Pharm. 179, 179–207.
ads, C.D., Furnish, C.M., Noda, I., Juhlin, K.D., Cooper, D.A., Morrall, S.W., 2004.Molecular factor analysis applied to collections of NMR spectra. Anal. Chem.76, 1982–1990.
ogulapati, V.P.R., Chavakula, R., Bandami, M., Gorantla, S.R., 2004. Crystalline formof atorvastatin hemi-calcium. US Patent 20090240064 (24.09.2004).
raeser, K.A., Strachan, C.J., Patterson, J.E., Gordon, K.C., Rades, T., 2008. Physico-chemical properties and stability of two differently prepared amorphous formsof simvastatin. Cryst. Growth Des. 8, 128–135.
reco, K., Bogner, R., 2010. Crystallization of amorphous indomethacin during dis-solution: effect of processing and annealing. Mol. Pharm. (on-line publication:
13.5.2010).riffin, J.M., Martin, D.R., Brown, S.P., 2007. Distinguishing anhydrous and hydrousforms of an active pharmaceutical ingredient in a tablet formulation using solid-state NMR spectroscopy. Angew. Chem. Int. Ed. 46, 8036–8038.
arris, R.K., 2006. NMR studies of organic polymorphs & solvates. Analyst 131,351–373.
harmaceutics 409 (2011) 62–74
Harris, R.K., Cadars, R., Emsley, L., Yates, J.R., Pickard, C.J., Jettid, R.K.R., Griesserd,U.J., 2007. NMR crystallography of oxybuprocaine hydrochloride, ModificationII◦ . Phys. Chem. Chem. Phys. 9, 360–368.
Harris, R.K., 2007. Applications of solid-state NMR to pharmaceutical polymorphismand related matters. J. Pharm. Pharmacol. 59, 225–239.
Heinz, A., Strachan, C.J., Atassi, F., Gordon, K.C., Rades, T., 2008. Characterizing anamorphous system exhibiting trace crystallinity: a case study with saquinavir.Cryst. Growth Des. 8, 119–127.
Heinz, A., Gordon, K.C., McGoverin, M.C., Rades, T., Strachan, C.J., 2009. Understand-ing the solid-state forms of fenofibrate – a spectroscopic and computationalstudy. Eur. J. Pharm. Biopharm. 71, 100–108.
Henry, E.R., Hofrichter, J., 1992. Singular value decomposition: application to anal-ysis of experimental data. Methods Enzymol. 210, 129–192.
Husak, M., Jegorov, A., Brus, J., Beek, van W., Patison, P., Christensen, M., Favre-Nicolin, V., Maixner, J., 2008. Metergoline II: structure solution from powderdiffraction data with preferred orientation and from microcrystal. Struct. Chem.19, 517–525.
Husak, M., Kratochvil, B., Jegorov, A., Brus, J., Maixner, J., Rohlícek, J., 2010. Simvas-tatin: structure solution of two new low-temperature phases from synchrotronpowder diffraction and ss-NMR. Struct. Chem. 21, 511–518.
Koonst, J.M., Ellis, P.D., 1995. Applicability of factor analysis in solid state NMR. Anal.Chem. 67, 4309–4315.
Levine, B.K., 1998. Chemometrics. Anal. Chem. 70, 209–228 (Review).Purser, S., Moore, P.R., Swallow, S., Gouverneur, V., 2008. Fluorine in medicinal
chemistry. Chem. Soc. Rev. 37, 320–330.Shete, G., Puri, V., Kumar, L., Bansal, A.K., 2010. Solid state characterization of com-
mercial crystalline and amorphous atorvastatin calcium samples. J. Pharm. Sci.Technol. 11, 598–609.
Tessler, L., Aronhime, J., Lifshitz, R., Maidan-Hanoch, D., Hasson, N., 2008. Novelcrystal forms of atorvastatin hemi-calcium and processes for their prepara-tion as well as novel processes for preparing other forms. US Patent 7456297(25.11.2008).
Virtanen, T., Maunu, S.L., 2010. Quantitation of a polymorphic mixture of an active
pharmaceutical ingredient with solid state 13C CPMAS NMR spectroscopy. Int. J.Pharm. 394, 18–25.Yu, L., 2001. Amorphous pharmaceutical solids: preparation, characterization andstabilization. Adv. Drug Deliv. Rev. 48, 27–42.
Zakrzewski, A., Zakrzewski, M., 2006. Solid State Characterization of Pharmaceuti-cals. Pergamon.
