Академический Документы
Профессиональный Документы
Культура Документы
h i g h l i g h t s g r a p h i c a l a b s t r a c t
a r t i c l e i n f o a b s t r a c t
Article history: This paper describes the development and validation of a new multivariate calibration method based on
Received 1 November 2013 diffuse reectance mid infrared spectroscopy for direct and simultaneous determination of three veter-
Received in revised form 8 January 2014 inary pharmaceutical drugs, pyrantel pamoate, praziquantel and febantel, in commercial tablets. The best
Accepted 14 January 2014
synergy interval partial least squares (siPLS) model was obtained by selecting three spectral regions,
Available online 24 January 2014
37153150, 28652583, and 22981733 cm 1, preprocessed by rst derivative and SavitzkyGolay
smoothing followed by mean centering. This model was built with ve latent variables and provided root
Keywords:
mean square errors of prediction (RMSEP) equal or lower than 0.69 mg per 100 mg of powder for the
Multi-analyte determination
Quality inspection control
three analytes. The method was validated according the appropriate regulations through the estimate
MIR of gures of merit, such as trueness, precision, linearity, analytical sensitivity, bias and residual prediction
siPLS deviation (RPD). Then, it was applied to three different veterinary pharmaceutical formulations found in
Multivariate analytical validation the Brazilian market, in a situation of multi-product calibration, since the excipient composition of these
Veterinary pharmaceutical formulations commercial products, which was not known a priori, was modeled by an experimental design that
scanned the likely content range of the possible constituents. The results were veried with high perfor-
mance liquid chromatography with diode array detection (HPLCDAD) and high performance liquid chro-
matographytandem mass spectrometry (HPLCMS/MS) and were in agreement with the predicted
values at 95% condence level. The developed method presented the advantages of being simple, rapid,
solvent free, and about ten times faster than the HPLC ones.
2014 Elsevier B.V. All rights reserved.
Corresponding author. Address: Departamento de Farmcia, Universidade Federal do Paran, Av. Pref. Lothrio Meissner, 632, 80210-170 Curitiba, PR, Brazil. Tel.: +55 41
33604094; fax: +55 41 33604106.
E-mail address: pontarolo@ufpr.br (R. Pontarolo).
http://dx.doi.org/10.1016/j.saa.2014.01.078
1386-1425/ 2014 Elsevier B.V. All rights reserved.
M.S. Piantavini et al. / Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 125 (2014) 396403 397
Materials and methods randomly mixed with the samples from the rst design, in order
to obtain a robust model.
Reagents and samples
Procedure
The PP, PZ and FB standards were obtained from SigmaAldrich
(St. Louis, MO, USA) and stored protected from light. The choice The powder samples were prepared, manually homogenized
excipients, colloidal silicon dioxide, sodium lauryl sulphate, micro- and directly measured. The spectra were recorded from 4000
crystalline cellulose, butylated hydroxytoluene, talc and corn to 400 cm 1 as the average of 64 scans and with a resolution
starch, were purchased from certied suppliers and used without of 4 cm 1. Six replicates of samples from the central point were
further purication. Acetonitrile and methanol were of HPLC grade. also obtained for evaluating repeatability. These results were
Ultrapure water was obtained using a Milli-Q purication system compared with similar replicates obtained on another day by
from Millipore (Bedford, MA, USA). Powder samples were prepared a different analyst for estimating intermediate precision. Twenty
by weighing with an analytical balance (0.0001 g), according to an spectra of the empty cell were also recorded for estimating the
experimental design. All the three analyzed commercial veterinary instrumental noise.
formulations have the same composition of API: 144 mg of PP,
50 mg of PZ and 150 mg of FB per tablet. Their compositions of Analysis of real samples
excipients are not publicly available.
Tablets of three different manufacturers available in Brazil were
purchased in Curitiba, PR. The contents of 20 tablets of each formu-
Apparatus and software
lation were crushed and mixed into a homogeneous powder. The
MIR spectra of these samples were acquired in the same conditions
The spectral dataset was acquired from a Bruker Alpha FT-IR
described in Section 2.4, in six replicates. These same commercial
spectrometer, equipped with a diffuse reectance accessory, and
samples were also analyzed by HPLCDAD and HPLCMS/MS, as
under controlled temperature (20.0 0.2 C) and humidity
described in Section 2.6.
4555%). The spectra were obtained in the absorbance mode by
using OPUS (OPtical User Software) for windows, version 6.0, from Chromatographic analysis
Brucker Optik (Bremen, Germany). Data were handled using OPUS,
MATLAB software, version 7.13 (The Math-Works, Natick, USA), The validation by HPLCDAD was adapted from a published
and PLS Toolbox, version 6.5 (Eigenvector Technologies, Manson, method14 and was carried out with an Agilent 1100 HPLC System
USA). (Wilmington, NC, USA) that consisted of a G1311A quaternary
pump, a G1379A degasser, a G1329A autosampler, a G1316A
Experimental design column oven and a G1315B DAD detector. The analytes were
separated on an XBridge C18 250 4,6 mm (5 lm particle size)
Fifty six powder samples were prepared according to a CCD column coupled with an XBridge C18 20 4.6 mm (5 lm particle
(arotability = 1.6818; aortogonality = 1.2872) with three factors, PP, PZ size) guard column. The mobile phase was water (pH 3.5, adjusted
and FB, and 8 levels plus a central point, as depicted in Fig. 2. with 85% phosphoric acid) and acetonitrile in gradient mode. A
The total mass of each sample was xed in 100.0 mg. The range ow rate of 1.2 mL min 1 and detection at 215 and 312 nm were
of each factor was varied from about 80.0% to 120.0% of its nominal used. The injection volume was 10 lL, and the column temperature
content in the analyzed formulations (Section 2.1). This range was was maintained at 40 C.
chosen in order to cover from 90.0% to 110.0% of the API contents, The analysis by HPLCMS/MS was performed as described by
which are the acceptable limits commonly established by the phar- Pontes et al. [22] and was carried out with an Agilent 1200 HPLC
macopoeias. The central point of CCD corresponds to 28.8 mg of PP, system (Santa Clara, CA, USA) coupled with a triple quadrupole
10.0 mg of PZ and 30.0 mg of FB per 100 mg of sample. Each sample API 3200 from Applied Biosystems MDS Sciex Instruments (Foster
was completed to 100 mg with a mixture of excipients, which was City, CA, USA). The analytical separations were achieved on an
prepared based on another experimental design. This design con- XBridge C8 50 2.1 mm (5 lm particle size) column coupled with
sisted of three factors that were varied in certain ranges: the two an XBridge C8 10 2.1 mm (5 lm particle size) guard column
major excipients, microcrystalline cellulose and corn starch, and maintained at 20 C. The mobile phase consisted of water/acetoni-
a mixture of the minor excipients, colloidal silicon dioxide, sodium trile (15:85 v/v) containing 0.1% formic acid and 3 mmol L 1 of
lauryl sulphate, butylated hydroxytoluene and talc. Nevertheless, ammonium formate. The isocratic ow rate was 200 lL min 1
this design was not shown here for reasons of commercial interest. and the injection volume was 20 lL. The electrospray ionization
The excipient mixtures obtained from the second design were source was operated in the positive mode under the following
working conditions: ion spray voltage of 5000 V; source tempera-
ture of 450 C; nebulizer and dryer gas (nitrogen) of 45 and 40 psi,
respectively; collision activated dissociation gas (CAD) of 4 psi and
curtain gas (CUR) of 10 psi. Quantication was performed in multi-
ple reaction monitoring (MRM) mode, maintaining a dwell time of
150 ms.
In both methods, a mass equivalent to one tablet was dissolved
in acetonitrile/methanol (50:50, v/v), sonicated and ltered
through a quantitative lter paper (J Prolab, 0.28 lm pore size).
Then, the ltered solution was diluted in a volumetric ask con-
taining the mobile phase, according to the appropriate concentra-
tion. All samples were prepared under low light exposure and
ltered through a 0.22 lm PVDF syringe lter prior to injection.
Fig. 2. Experimental design (CCD) used for varying PP, PZ and FB contents. All the injections were repeated three times and each sample
Calibration (circles) and validation (down triangles) samples. was analyzed in triplicate.
M.S. Piantavini et al. / Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 125 (2014) 396403 399
Results and discussion regions that may most contribute for predictive models. For each
analyte the main bands were identied [46]. In the spectrum of
Selection of excipients and experimental design PP (Fig. 3a), it is possible to observe a broad band with peaks at
3067, 2969 and 2892 cm 1, which are related to CAH, phenolic
The oral absorption and therapeutic efcacy of any drug de- OAH and carboxylic OAH stretching vibrations, respectively; a
pends on its aqueous solubility and gastrointestinal permeability. strong peak at 1659 cm 1, related to the aromatic carboxylic
The Biopharmaceutics Classication System (BCS) [43] is a scien- C@O stretching; a peak at 1612 cm 1, related to the cyclic imine
tic framework used to predict bioavailability of drugs based on C@N bond; a peak at 1150 cm 1, related to the CAN stretching of
the properties of solubility and permeability [44]. Furthermore, tertiary amine; and a peak at 713 cm 1, related to the out-of-plane
BCS has been used as criterion for excipient selection of pharma- CAH bending of tiophene ring [47]. In the spectrum of PZ (Fig. 3b),
ceutical formulation considering pharmacotechnical aspects it can be observed a small band with two peaks at 3291 and
which contribute to the drug dissolution and consequent absorp- 3234 cm 1, related to the rst overtones of the C@O stretching
tion [45]. According to BCS, drug substances are divided in four vibrations; two strong peaks at 2898 and 2848 cm 1, related to
classes: high solubility and high permeability (I); low solubility the stretching vibrations of ACAH bound to tertiary amines in
and high permeability (II); high solubility and low permeability the lactam ring; a strong band between about 1700 and
(III); and low solubility and low permeability (IV). In agreement 1600 cm 1 and centered at 1666 cm 1, related to the stretching
with this classication, PP and PZ are classied as Class II drugs vibrations of the two amide carbonyl groups; and a strong band
[45] and therefore their bioavailability are limited by the dissolu- centered around 1450 cm 1, characteristic of di-substituted
tion rate. Classication of FB was not available, and thus, only PP amides. In the spectrum of FB (Fig. 3c), it can be observed a broad
and PZ data were considered to select the excipients. Considering band with a peak at 3209 cm 1, corresponding to the carbamate
this context and a specic knowledge of the most common excip- NAH stretchings; two peaks at 2977 and 2825 cm 1, correspond-
ients used in veterinary formulations, six excipients were chosen. ing to the CAH stretching from the OACH3 groups; strong peaks
The two major excipients were corn starch, used as diluent and at 1769 and 1687 cm 1, related to carbamate C@O stretchings;
lubricant, and microcrystalline cellulose, used as diluent, lubricant and a peak centered at 1514 nm, related to carbamate NAH
and disintegrant. The four minor excipients were talc, used to pre- bending.
vent sticking during manufacturing, sodium lauryl sulphate, used
as lubricant tensoactive and agglutinant, colloidal silicon dioxide, siPLS model
used as absorbent, and butylated hydroxytoluene, used as antiox-
idant. Mixtures of excipients were prepared according to an The spectra of all 56 prepared samples (Fig. 4) were divided into
experimental design with three factors, as already mentioned in 36 for the calibration set and 20 for the validation set, according to
Section 2.3. The content ranges of these factors were varied an experimental design (Fig. 2). The validation samples were se-
20% around the expected values (not shown). Then, the excipient lected in order to test the predictive ability of the model by ensur-
mixtures were randomly combined with the samples from the ing a homogeneous and representative distribution of the three
CDD (Fig. 2). This procedure aims to obtain a representative and analytes within the analytical ranges. Preliminary tested models
robust model that is able to predict samples with composition indicated that no signicant differences were observed between
of excipients varying in the studied range. In addition, the random the results of PLS1 and PLS2. Thus, for simplicity reasons PLS2
combination between the experimental designs of the analytes was chosen. In PLS2, in contrast to PLS1, all the analytes are pre-
and excipients aims to avoid chance correlations, also contribut- dicted simultaneously from the same set of loadings. The use of
ing for a robust model [26]. preprocessing techniques for PLS models developed from reec-
tance infrared spectra of powder samples is almost mandatory,
MIR PP, PZ and FB spectra due to the presence of non-linear baseline deviations, caused by
the multiplicative light scattering. In this work, the most used
MIR spectra of pure PP, PZ and FB powder samples are shown in techniques multiplicative scatter correction (MSC), standard nor-
Fig. 3. The interpretation of these spectra helps in highlighting the mal variate (SNV), and rst derivative with smoothing [48] were
tested. The lowest root mean square error of cross validation
(RMSECV) was obtained with rst derivative followed by
Fig. 3. Diffuse reectance MIR spectra of pure (a) PP, (b) PZ and (c) FB powder
samples. Fig. 4. MIR spectra of all the 56 prepared samples.
400 M.S. Piantavini et al. / Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 125 (2014) 396403
Savitsky-Golay smoothing (15 points in lter and rst order poly- RMSEC (root mean square of calibration) and, mainly, RMSEP (root
nomial t) and mean centering. The best PLS model was selected mean square of validation). The estimated RMSEP for the three
by random subsets (with 6 splits) cross validation and with 5 latent analytes were between 0.69 and 0.34 mg/100 mg. The trueness
variables (LV). Considering this number of LV, the minimum rec- can also be evaluated through the relative errors of prediction for
ommended number of calibration and validation samples for infra- each sample. The observed relative prediction errors for the valida-
red multivariate quantitative models was assured [49]. tion samples were between +3.9% and 2.6% for PP, +5.3% and
Since full spectrum PLS models use to include uninformative 3.9% for PZ, and +5.1% and 4.4% for FB. These results were con-
wavenumbers that may negatively affect them by increasing the sidered acceptable, since between 60% and 70% of the validation
errors [31], variable selection by siPLS2 was employed for improv- samples for each analyte presented relative errors below 2%, the
ing the predictive ability of the model. The spectra were divided in acceptable limits of trueness commonly adopted for univariate
8, 12, 20, 30 and 50 intervals combined in up to 5 subintervals. The methods in the pharmaceutical industry [50]. The precision was
best results, as shown in Fig. 5, were obtained with spectra divided assessed at two levels, based on the relative standard deviation
in 12 intervals and combined in 5 subintervals. This siPLS2 model (RSD) for six replicates. The RSD for repeatability (intra-run preci-
also used 5 LV and accounted for 97.87% of the total variance in the sion) were between 1.5% and 1.8%, while for intermediate precision
X block and 97.99% in the Y block. The average (three analytes) (inter-run) they were between 2.2% and 3.3%. All of these values
RMSECV of 1.21 mg/100 mg for the PLS2 model was signicantly are in accordance with the Brazilian regulations [37], which pre-
decreased to 0.74 mg/100 mg in this siPLS2 model. The RMSECV scribes a maximum RSD of 5%. The results of trueness and precision
for each analyte is shown in Table 1. The selected intervals (in light allow assuring that the method was considered accurate. The line-
gray in Fig. 5) correspond to three spectral regions, between 3715 arity of multivariate methods can be evaluated through the plot of
3150 cm 1, 28652583 cm 1, and 22981733 cm 1. These selected the reference versus predicted values, as shown in (Fig. 6ac) for
regions include specic absorptions of each analyte, such as a small PP, PZ and FB. The correlation coefcients (r) were 0.99 for PP
and broad band (2130 and 1790 cm 1) centered at 1923 cm 1 of and FB, and 0.96 for PZ. The linearity of this method was assured
PP; small peaks of C@O stretching rst overtones in the region of by verifying the random behaviors of the residuals distributions
33003220 cm 1, and a peak at 2848 cm 1, related to a stretching for the three analytes, as can be seen in Fig. 6df. Considering
vibration of ACAH bound to a tertiary amine of PZ; and a strong the linearity and accuracy evaluations, the working ranges of
band centered at 3209 cm 1, characteristic of carbamates, and a this method were established from 19.00 to 39.0 mg/100 mg for
carbonyl stretching at 1769 cm 1 of FB (but not the carbonyl vibra- PP, 6.70 to 13.40 mg/100 mg for PZ, and 20.00 to 40.0 mg/100 mg
tions from the other analytes). It should be noted that the entire for FB.
spectral region with wavenumbers below 1700 cm 1 led to a de- The gures of merit selectivity (SEL), sensitivity (SEN), analyti-
crease in the prediction ability of the model (Fig. 5). cal sensitivity (c), limits of detection (LOD) and quantitation
Finally, an attempt to improve the model by outlier detection (LOQ) were estimated based on the concept of NAS, as described
was carried out based on the identication of samples with ex- in the appropriate Refs. [25,41,42]. The interpretation of the SEL
treme leverages, large residuals in the spectral data or large resid- concept for multivariate methods is different from univariate ones
uals in the analytical concentration values, all with 95% condence and has no practical interest for quality control purposes. The SEL
intervals. Nevertheless, no outlier was detected. denition is only useful within a certain group of samples of sim-
ilar qualitative composition and for this method its estimate indi-
Validation of the multivariate model cate that about 21%, 30% and 20% of the analytical signal were used
for predicting PP, PZ and FB, respectively. Since the pure SEN is not
Once the model was developed, the analytical validation is appropriate for comparison with other methods, their values were
essential for quality control purposes. All the multivariate gures divided by the estimative of the instrumental noise (e = 7.3.10 4)
of merit estimated for this validation are listed in Table 1. The aver- and the more useful c was estimated. The inverse of c indicated
age trueness was evaluated through the parameters RMSECV, that the method was able to discriminate minimum content
Fig. 5. Results of siPLS2. The mean spectrum of 36 calibration samples and the average (three analytes) RSMECV values for each interval. The numbers at the squares bottom
indicate the number of LV for each interval and the dashed line indicates the RMSECV for the full spectrum model. The selected spectral regions are in light gray.
M.S. Piantavini et al. / Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 125 (2014) 396403 401
Table 1
Parameters estimated for validating the developed siPLS2-MIR method.
Fig. 6. Plots of reference versus predicted values for the calibration (circles) and validation (down triangles) samples of (a) PP, (b) PZ and (c) FB. PLS residuals for the
predictions of (d) PP, (e) PZ and (f) FB.
differences between 0.02 and 0.03 mg/100 mg of the analytes, con- analyzed in six replicates by the developed mid infrared method
sidering the random instrumental noise as the only source of er- and in triplicates by both the chromatographic methods. The pre-
rors. Though not necessary for this kind of method, LOD and LOQ dicted mean values and their standard deviations are shown in Ta-
were also estimated based on the e. ble 2. According to non-paired t tests with 7 degrees of freedom,
The bias should be estimated only for the validation samples there were no signicant differences between the predictions of
and the values in Table 1, along with their standard deviation, were the three methods for all the analytes in the formulations #1 and
used in t tests with 20 degrees of freedom and at 95% condence #3, and for PP and FB in the formulation #2, at 99% condence level
level. The estimated t values were all below the critical t value (all the estimated t values below the critical t = 3.499). The results
(2.086), demonstrating the absence of systematic errors in the for predicting PZ in the formulation #2 were not in agreement be-
model. The RPD51 is the ratio of natural variation in the samples tween the MIR and the chromatographic methods. Clearly, the pre-
to the size of probable errors occurring during the prediction, diction of the MIR method was discrepant. A possible explanation
and it is more useful for comparing models on different data sets for this result is the exclusive presence of a palatabilizing agent in
or in absolute terms. It was calculated for the calibration and val- the formulation #2, which is indicated in its package, but not iden-
idation sets and the minimum RPD was estimated as 3.0 for the PP tied. Since this substance was not included in the model and
prediction in the validation set, which is above 2.4, the lower limit could absorb overlapping selective PZ peaks, it would interfere spe-
desirable for good calibration equations [51]. cically in the quantication of PZ. For overcoming this problem, it
would be necessary to identify this substance and include it in the
Analysis of commercial tablets experimental design. As we are not able to identify this palatabiliz-
ing substance, an alternative adopted was to try to predict PZ in the
The developed method was applied to the simultaneous quan- formulation #2 by testing sub-models built with each one of the
tication of PP, PZ and FB in three different commercial formula- three continuous spectral regions selected by siPLS (Fig. 5). This
tions (tablets), and the results were veried by independent was successfully achieved by selecting only the region between
HPLCDAD and HPLCMS/MS methods. The samples were 3715 and 3150 cm 1, where PZ presents specic vibrations related
402 M.S. Piantavini et al. / Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 125 (2014) 396403
Table 2
Mean values and standard deviations for the simultaneous determination of PP, PZ and FB in three commercial formulations by the HPLCDAD (n = 3), HPLC/MSMS (n = 3) and
the proposed MIR spectroscopy (n = 6) methods.
Label claim (mg/tablet) HPLCDAD (mg/tablet) HPLC/MSMS (mg/tablet) Multivariate model (mg/tablet)
PP PZ FB PP PZ FB PP PZ FB PP PZ FB
144a 50 150 157.4 5.0 49.7 0.3 155.3 2.1 155.3 4.4 44.4 1.8 151.4 3.1 158.9 1.1 48.5 1.8 155.7 1.9
144b 50 150 153.8 6.2 49.8 0.2 158.3 3.0 154.7 7.7 45.1 0.5 154.8 2.6 160.2 1.3 70.4 1.2 158.6 1.3
144c 50 150 144.9 0.6 52.8 1.3 148.8 2.0 145.9 0.5 54.2 1.4 142.5 2.9 143.7 2.0 53.5 1.3 146.9 2.3
a
Commercial tablet brand #01.
b
Commercial tablet brand #02.
c
Commercial tablet brand #03.
to the rst overtones of the C@O stretching vibrations (Fig. 3b). [8] G.A. Forcier, R.F. Mushinsky, R.L. Wagner Jr., J. Pharm. Sci. 60 (1971) 111113.
[9] M.S. Piantavini, F.L.D. Pontes, L.B. Cerqueira, P.G. Peralta-Zamora, R. Pontarolo,
This sub-model used the same previous preprocessing, 4 LV and ac-
Journal of Analytical Chemistry (2014). in press.
counted for 91.78% of the variance in X and 91.84% in Y. It pre- [10] European Pharmacopoeia, sixth ed., Strasbourg, France, 2007.
dicted 51.0 2.9 mg/tablet of PZ in formulation #2, a value with [11] M.M. Ghoneim, M.M. Mabrouk, A. Tawk, J. Pharm. Biomed. Anal. 30 (2002)
no signicant differences from the results obtained by HPLC 13111318.
[12] W.J. Allender, J. Chromatogr. Sci. 26 (1988) 470472.
DAD, 49.8 0.2 mg/tablet, and HPLCMSMS, 45.1 0.5 mg/tablet [13] A.P. Argekar, S.V. Raj, S.U. Kapadia, Talanta 44 (1997) 19591965.
(Table 2), as veried by a t test at 99% condence level. The sub- [14] W. Bialecka, A. Kulik, Acta Pol. Pharm. 67 (2010) 463468.
models built with the other two spectral regions provided poor [15] J. Li, Y. Wang, A. Fenwick, T.A. Clayton, Y.Y. Lau, C. Legido-Quigley, J.C. Lindon, J.
Utzinger, E. Holmes, J. Pharm. Biomed. Anal. 45 (2007) 263267.
predictions for PZ in this formulation, indicating that they may [16] J.Z. Sabbatini, J. AOAC Int. 92 (2009) 2633.
be prone to interference of the palatabilizing substance. [17] V.A. Thorpe, J. Chromatogr. Sci. 26 (1988) 545550.
[18] S. Babic, D. Mutavdzic-Pavlovic, D. Asperger, M. Perisa, M. Zrncic, A.J. Horvat,
M. Kastelan-Macan, Anal. Bioanal. Chem. 398 (2010) 11851194.
Conclusions [19] G. Balizs, J. Chromatogr. B 727 (1999) 167177.
[20] P.S. Bonato, A.R. de Oliveira, F.J. de Santana, B.J. Fernandes, V.L. Lanchote, A.E.
Gonzalez, H.H. Garcia, O.M. Takayanagui, J. Pharm. Biomed. Anal. 44 (2007)
A diffuse reectance MIR spectroscopy method for the simulta- 558563.
neous determination of three API, pyrantel pamoate, praziquantel [21] R.M. Lima, M.A. Ferreira, T.M. Ponte, M.P. Marques, O.M. Takayanagui, H.H.
and febantel, in veterinary formulations was successfully devel- Garcia, E.B. Coelho, P.S. Bonato, V.L. Lanchote, J. Chromatogr. B 877 (2009)
30833088.
oped and validated through the estimative of multivariate gures [22] F.L.D. Pontes, R. Pontarolo, F.R. Campos, J.C. Gasparetto, M.A. Cardoso, M.S.
of merit. The use of siPLS2 allowed improving the model predic- Piantavini, A.C.L.B. Trindade, Asian J. Pharm. Clin. Res. 6 (2013) 191200.
tions through the empirical selection of the most discriminant [23] M.H. Ferreira, J.F.F. Gomes, M.M. Sena, J. Braz. Chem. Soc. 20 (2009) 1680
1686.
spectral ranges. For real samples, the developed method provided
[24] W.F.C. Rocha, A.L. Rosa, J.A. Martins, R.J. Poppi, Determination and validation of
results statistically similar to those obtained by two different HPLC nimesulide in pharmaceutical formulation by near infrared spectroscopy, J.
methods in almost of all the cases and presented the advantages of Braz. Chem. Soc. 21 (2010) 19291936.
being more rapid (about ten times faster), less expensive, no need [25] M.A.M. Silva, M.H. Ferreira, J.W.B. Braga, M.M. Sena, Talanta 89 (2012) 342
351.
for reagents or solvents and no generation of chemical waste. This [26] D. Xiang, J. Berry, S. Buntz, P. Gargiulo, J. Cheney, Y. Joshi, B. Wabuyele, H. Wu,
model was considered suitable for routine quality control determi- M. Hamed, A.S. Hussain, M.A. Khan, J. Pharm. Sci. 98 (2009) 11551166.
nations in laboratories of industries and handling pharmacies. An [27] A.A. Bunaciu, H.Y. Aboul-Enein, S. Fleschin, Appl. Spectrosc. Rev. 45 (2010)
206219.
aspect that should be highlighted is that this model was developed [28] A.A. Kandhro, A.H. Laghari, S.A. Mahesar, R. Saleem, A. Nelofar, S.T. Khan, S.T.H.
without direct access to the excipient composition of the commer- Sherazi, Spectrochim. Acta Part A 115 (2013) 5156.
cial formulations. The employed experimental design searched to [29] S. Mazurek, R. Szostak, Vib. Spectrosc. 57 (2011) 157162.
[30] A.L.H. Mller, R.S. Picoloto, M.F. Ferro, F.E.B. da Silva, E.I. Mller, .M. Flores,
incorporate the most likely substances present in the excipients Drug Test. Anal. 4 (2012) 500506.
in the adequate content ranges, in a situation of multi-product cal- [31] F.E.B. Silva, M.F. Ferro, G. Parisotto, E.I. Muller, E.M.M. Flores, J. Pharm.
ibration. Since practically all the multivariate infrared (NIR and Biomed. Anal. 49 (2009) 800805.
[32] D.M. Haaland, E.V. Thomas, Anal. Chem. 60 (1988) 11931202.
MIR) models for direct API determinations have been developed [33] M. Blanco, M. Alcal, Eur. J. Pharm. Sci. 27 (2006) 280286.
for a specic product, aiming at the industrial quality control, the [34] E. Micklander, K. Kjeldahl, M. Egbo, L. Norgaard, J. Near Infrared Spectrosc. 14
strategy of multi-product calibration by incorporating the compo- (2006) 395402.
[35] Y. Zheng, X. Lai, S.W. Bruun, H. Ipsen, J.N. Larsen, H. Lowenstein, I. Sondergaard,
sition of several formulations is suitable for the future develop-
S. Jacobsen, J. Pharm. Biomed. Anal. 46 (2008) 592596.
ment of multivariate methods for public quality inspection [36] G.E.P. Box, J.S. Hunter, Ann. Math. Stat. 28 (1957) 195241.
control by the ofcial organisms. [37] Agncia Nacional de Vigilncia Sanitria, (ANVISA), Guia para Validao de
Mtodos Analticos e Bioanalticos, Resoluo RE no 899, Brazil, 2003.
[38] International Conference on Harmonisation, Tripartite Guideline Q2A Text
Acknowledgements on Validation of Analytical Procedures, Fed. Regist., 60 FR 11260, USA, 1995.
[39] International Conference on Harmonisation, Tripartite Guideline Q2B
Validation of Analytical Procedures: Methodology, Fed. Regist., 62 FR 27464,
M.S.P thanks CAPES-REUNI and CNPq for fellowships.
USA, 1997.
[40] M.H. Ferreira, J.W.B. Braga, M.M. Sena, Microchem. J. 109 (2013) 158164.
References [41] A.C. Olivieri, N.M. Faber, J. Ferr, R. Boqu, J.H. Kalivas, H. Mark, Pure Appl.
Chem. 78 (2006) 633661.
[42] P. Valderrama, J.W.B. Braga, R.J. Poppi, Quim. Nova 32 (2009) 12781287.
[1] Q.A. McKellar, F. Jackson, Trends Parasitology 20 (2004) 456461.
[43] TRSL, Therapeutic Systems Research Laboratories. BCS (Biopharmaceutics
[2] P. Taweethavonsawat, S. Chungpivat, P. Satranarakun, R.J. Traub, R. Schaper,
Classication System), <http://www.tsrlinc.com/resources/services/>,
Parasitol. Res. 106 (2010) 533537.
(accessed October 2013).
[3] S.C. Su, H.H. Chou, P.C. Chang, C.H. Liu, S.S. Chou, J. Food Drug Anal. 12 (2004)
[44] G.L. Amidon, H. Lennernas, V.P. Shah, J.R. Crison, Pharm. Res. 12 (1995) 413
244253.
420.
[4] A. Davis, D.H. Wegner, Bull. World Health Organ. 57 (1979) 767771.
[45] M. Martinez, L. Augsburger, T. Johnston, W.W. Jones, Adv. Drug Deliv. Rev. 54
[5] S.F. Andrade, Manual de Teraputica Veterinria, 2 ed., Roca, So Paulo, 2002.
(2002) 805824.
[6] A.A. Fasanmade, A.O. Akanni, A.A. Olaniyi, A.A. Fasanmade, F. Tayo, Biopharm.
[46] R.M. Silverstein, F.X. Webster, D.J. Kiemle, Spectrometric Identication of
Drug Dispos. 15 (1994) 527534.
Organic Compounds, seventh ed., John Wiley & Sons, New York, 2005.
[7] R. Jain, N. Jadon, K. Radhapyari, Talanta 70 (2006) 383386.
M.S. Piantavini et al. / Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 125 (2014) 396403 403
[47] F. Hegelund, R.W. Larsen, M.H. Palmer, J. Mol. Spectrosc. 247 (2008) [50] L. Huber, Validation and Qualication in Analytical Laboratories, second ed.,
100114. LabCompliance, New York, USA, 2007.
[48] A. Rinnan, F.v.d. Berg, S.B. Engelsen, TrAC Trends Anal. Chem. 28 (2009) [51] P. Williams, Implementation of Near-infrared Technology, in: P. Williams, K.
12011222. Norris (Eds.), Near-infrared Technology in the Agricultural and Food
[49] Standard Practices for Infrared Multivariate Quantitative Analysis E1655-05, Industries, second ed., American Association of Cereal Chemists Inc, St. Paul,
ASTM International, West Conshohocken, USA, 2012. USA, 2001.