Journal of Pharmaceutical and Biomedical Analysis 70 (2012) 301 309 [623188]

Journal of Pharmaceutical and Biomedical Analysis 70 (2012) 301– 309
Contents lists available at SciVerse ScienceDirect
Journal of Pharmaceutical and Biomedical Analysis
jou rn al h om epage: www.elsevier.com/locate/jpba
High-throughput NIR-chemometric methods for determination of drug content
and
pharmaceutical properties of indapamide powder blends for tabletting
Alina Porfirea, Lucia Rusb, Andreea Loredana Vonicac, Ioan Tomutaa,∗
aIuliu Hatieganu University of Medicine and Pharmacy, Department of Pharmaceutical Technology and Biopharmaceutics, 41 Victor Babes Street, 400023 Cluj-Napoca, Romania
bIuliu Hatieganu University of Medicine and Pharmacy, Department of Drug Analysis, 6 L. Pasteur Street, 400023 Cluj-Napoca, Romania
cS.C. Polipharma Industries, 550052 Sibiu, Romania
a r t i c l e i n f o
Article history:
Received
4 April 2012
Received
in revised form 18 July 2012
Accepted
22 July 2012
Available online 31 July 2012
Keywords:NIR spectroscopy
PLSParticle size
Powder
flow
Indapamidea b s t r a c t
This paper describes the development and application of NIR-chemometric methods for active content
assay and pharmaceutical characterization (granulometric analysis and flowability assessment) of inda-
pamide powder blends for tabletting. Indapamide powder blends were prepared and their NIR spectra
were recorded in reflectance mode. Partial least-squares (PLS) regression followed by leave-one-out
cross-validation was used to develop calibration models for predicting the indapamide content and
pharmaceutical properties. The method for indapamide assay was validated in terms of trueness, preci-
sion, accuracy. The near infrared based property predictions were compared with the reference method
results and no significant differences were found between the reference and predicted characteristics. The
developed NIR-chemometric methods can be useful tools for prediction of active content, granulometric
properties and parameters related to flowability of pharmaceutical powders.
© 2012 Elsevier B.V. All rights reserved.
1. Introduction
Powders are encountered in pharmaceutical industry and prac-
tice as a pharmaceutical dosage form as well as in the manufacture
of tablets, capsules and suspensions. Their properties and behav-
ior of finely divided solid materials are of considerable importance,
being determined by the chemical and physical properties of their
components and by the manner in which their components interact
[1]. As a result, the knowledge of chemical composition and phys-
ical behavior of powder blends is essential in the development of
solid dosage forms.
In the manufacture process of solid dosage forms, the uniform
mixing of drug and excipients is an essential step before proceed-
ing to other operations, since it is well known that inert excipients
can affect the characteristics, quality, stability, and even the perfor-
mance of the final product. Thus, blend homogeneity is crucial to
ensure uniformity of dosage units of the end product. Uniformity of
dosage units warrants that each unit in a batch has active substance
content within a narrow range around the label claim [2]. Problems
concerning the assurance of content uniformity of tablet dosage
units are due to three main effects: loss of powder homogeneity,
∗Corresponding author at: Department of Pharmaceutical Technology and Bio-
pharmaceutics,
Faculty of Pharmacy, 41 Victor Babes Street, 400012 Cluj-Napoca,
Romania.
Tel.: +40 264595770; fax: +40 264595770.
E-mail address: tomutaioan@umfcluj.ro (I. Tomuta).variability of tablet weight (either by changing the flow properties
of powder or by changing the weight of the die volume) and lack of
powder homogeneity. The uniformity of a powder blend is deter-
mined in practice by estimating the distribution of the drug, based
on its assay in representative samples from the blenders [3]. The
assay of active content is usually done through conventional HPLC
and UV spectrophotometry. Because particle segregation can occur
during the tabletting process, a tool able to detect particle segrega-
tion on line (in the hopper of the compression machine) would be
very useful to ensure uniformity of dosage units of the final prod-
uct. Besides assuring the uniformity, an accurate pharmaceutical
characterization of powder blends in terms of particle size, size dis-
tribution and flow properties is of equal importance. The methods
currently used for granulometric analysis are sieving, microscopy
or indirect measurements related to particle size such as sedimen-
tation rates, permeability and optical properties [1]. The powder
flow properties are evaluated by employing parameters such as
angle of repose, Carr’s index, Hausner ratio, and time of flow. In con-
clusion, a complete evaluation of powder characteristics requires a
variety of labor-intensive and time-consuming techniques [4].
Near-infrared (NIR) spectroscopy is an interesting alternative
for powder characterization regarding both chemical composition
and various physical or pharmaceutical properties. This technique
is offering rapid, non-invasive and non-destructive sample anal-
ysis, requiring little or no sample preparation, and has gained
wide acceptance in pharmaceutical industry [5,6] . NIR spectrum
of powder contains both chemical and physical properties and the
0731-7085/$ – see front matter © 2012 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.jpba.2012.07.026

302 A. Porfire et al. / Journal of Pharmaceutical and Biomedical Analysis 70 (2012) 301– 309
response is quick, so NIR spectroscopic technique can be used for
on-line monitoring of powder from point of view of powder homo-
geneity, powder flow properties and particle size distribution, in
order to assure consistent uniformity of dosage units. Moreover,
because the information provided is very complete, this technique
can detect the source for lack of uniformity of dosage units. A
lot of papers are reporting the determination of chemical com-
position of powders by NIR methods [7,8] , but few applications
have been performed on low dosage forms, due to low sensitivity
of NIR techniques [9–11] . Other papers are focusing on predic-
tion of pharmaceutical properties of powders based on their NIR
spectra. Sarraguc ¸ a and co-workers [12] reported the prediction of
particle size distribution of pharmaceutical powder blends having
paracetamol as active ingredient by the use of NIR spectra of pow-
der blends and other two data blocks (their flowability properties
and the mass fraction of the components present in the samples)
taken separately or together, in a multi-block strategy. Alcala et al.
[13] monitored both physical (particle size distribution and bulk
density) and chemical (moisture content) properties during a wet
granulation process at a production plant by the use of a portable
NIR spectrometer for on-line spectral acquiring on a noninvasive
approach, and used chemometric algorithms to characterize the
formulation and obtain a better knowledge of the granulation pro-
cess. Other authors [14] used NIR spectra of flowing powders to
evaluate the consistency of powder flow, based on the plots of
1/noise versus time as well as PCA score plots, and to evaluate drug
concentration in powder mixtures using the same NIR spectra. The
same group quantified the powder flow by recording the in-line
spectra of powder being discharged from a container and the results
are further utilized as a tool for distinguishing the difference in flow
behavior between the cohesive and surface modified or dry coated
powders [15] .
Indapamide is an oral antihypertensive diuretic agent indicated
for the treatment of hypertension and edema. A 2.5 mg formula-
tion is used world-wide for the treatment of mild to moderate
hypertension and low dose 1.5 mg formulation of indapamide
in a sustained-release (SR) coated tablet was recently developed
to maximize the efficacy/safety ratio following international rec-
ommendations favoring low dose antihypertensive therapy in
hypertension [16,17] .
This work describes the development of NIR-chemometric
methods for chemical and pharmaceutical characterization of
low-dosage indapamide powder-blends for tabletting. The chem-
ical characterization consisted in determination of API content
of the blends using a validated NIR-chemometric method. The
pharmaceutical characterization was done through determina-
tion of particle size and dispersion, Carr’s index, Hausner ratio
and time of flow using NIR-chemometric methods. These pro-
posed methods can be used as tools for detection of particle
segregation of powder-blends during tablet manufacturing pro-
cess.
2. Materials and methods
2.1. Materials
Indapamide (micronized powder, average particle size 15 /H9262m)
was purchased from PharmaZell, Germany. Lactose (Tablettose
80, average particle size 177 /H9262m) was provided by Meggle,
Germany. Microcrystalline cellulose (average particle size 100 /H9262m)
and sodium starch glycolate were obtained from JRS Pharma,
Germany. Colloidal silicon dioxide (Aerosil) was supplied by Rohm-
Pharma Polymers, Germany. Polyvinylpyrolidone was from BASF,
Germany. Magnesium stearate was purchased from Union Derivan,
Spain.2.2.
Sample preparation
For calibration purpose, powder blends for indapamide tablets
were prepared. Briefly, indapamide, lactose (63.75%, w/w), micro-
crystalline cellulose (21.67%, w/w), sodium starch glycolate (5.00%,
w/w), polyvinylpyrrolidone (6.00%, w/w) and colloidal silicon diox-
ide (0.75%, w/w) were mixed using a planetary mixer (PRS type,
Erweka, Germany) for 5 min. Magnesium stearate (0.75%, w/w) was
then added and the mixing was continued for 1 more minute. Sub-
sequently, the powder blend was separated into three particle size
classes: 0–100, 100–200 and 200–300 /H9262m.
The mixture composition was designed for a tablet weight of
approximately 120 mg and a usual amount of active ingredient
inside each tablet of 2.5 mg (2.08%, w/w). This formulation will be
further considered as the 100% active content formulation.
2.3. Calibration and validation protocol
2.3.1. Indapamide assay
For the calibration and validation of NIR method for indapamide
assay, powder blends were prepared. The protocol included batches
and days as sources of variability for calibration and validation
steps.
The calibration set used included 5 different formulations
containing 1.67, 1.88, 2.08, 2.29 and 2.5% (w/w) of indapamide
(corresponding to 80, 90, 100, 110 and 120% indapamide content
formulations). Three independent batches were manufactured per
formulation type in three different days.
In order to validate the NIR method for indapamide assay, the
same formulations as for the calibration set, corresponding to 80,
100 and 120% indapamide content, were manufactured. Four repli-
cates were prepared for each concentration level, in three different
days, resulting in a total of 36 validation samples.
2.3.2. Particle size and flow characteristics
In order to build calibration models for prediction of particle
size and powder flow characteristics, blends consisting in parti-
cles belonging to known size classes were mixed according with
a D-optimal experimental design with three factors and five levels
developed in Modde 9.0 software (Umetrics, Sweden). In this exper-
imental design, the input variables were granulometric classes and
the levels were the mixing ratios. The matrix of experimental design
is shown in Table 1.
2.4. NIR equipment and software
Near infrared spectra were recorded using a Fourier-transform
NIRS analyzer (Antaris II, ThermoElectron Scientific, USA) in
Reflectance Sampling configuration, equipped with an indium gal-
lium arsenide (InGaAs) detector. Since the powder samples are not
homogeneous, the device is equipped with a system for rotation
of samples during the measurements so that obtained spectrum
is representative for the sample and to ensure reproducibility of
the measurements. Each reflectance spectrum was recorded using
OMNIC software (Thermo Scientific, USA) by integrating 32 scans,
over the range from 11,000 to 4000 cm−1, with a resolution of
8 cm−1.
2.5. Reference methods
Indapamide assay was performed on powder blends using
a reference HPLC method. Accurately weighted samples were
extracted with 5 ml methanol in an ultrasonic bath for 10 min and
the obtained suspension was centrifuged for 5 min at 5000 rpm.
Aliquots of the clear supernatants were diluted with the mobile

A. Porfire et al. / Journal of Pharmaceutical and Biomedical Analysis 70 (2012) 301– 309 303
Table 1
Composition
of calibration set for powder flow characteristics, according to D-
optimal
experimental design.
Exp. name X1 X2 X3
N1 33.33 33.33 33.33
N2 71.43
14.29 14.29
N3 40.00 40.00 20.00
N4
44.44 44.44 11.11
N5
14.29 71.43 14.29
N6
45.45 45.45 9.09
N7 57.14
14.29 28.57
N8 14.29 57.14 28.57
N9
22.22 55.56 22.22
N10
45.45 27.27 27.27
N11 28.57 14.29 57.14
N12
14.29 28.57 57.14
N13 30.77
38.46 30.77
N14 14.29 14.29 71.43
N15 45.45
9.09 45.45
N16
36.36 18.18 45.45
N17 18.18
36.36 45.45
N18 9.09 45.45 45.45
N19
33.33 33.33 33.33
N20
33.33 33.33 33.33
N21
33.33 33.33 33.33
N22 33.33 33.33 33.33
X1, % of particle size class 0–100 /H9262m; X2, % of particle size class 100–200 /H9262m; X3, %
of particle size class 200–300 /H9262m.
phase in 5 ml volumetric flasks. Aliquots of the obtained solu-
tions (20 /H9262l) were then analyzed by HPLC (Agilent, USA) with
UV-detection. Data recording and processing were done by Chem-
Station for LC software. Separation was carried out at 30◦C on a
Zorbax SB-C8 analytical column (150 mm × 4.6 mm, 5 /H9262m parti-
cle size), with a mobile phase containing phosphoric acid solution
(0.1%, pH 3.0):acetonitrile (45:55, v/v) at a flow rate of 1.5 ml min−1.
Detection was performed at 240 nm. Under the given chromato-
graphic conditions the retention time of indapamide was 1.9 min.
Active content in the powder blend was determined using the linear
regression of standard indapamide in the range of 0.5–2.5 /H9262g ml−1.
For particle size characterization of indapamide powder blends,
the samples were sieved using a set of sieves with different aper-
tures (100, 200, 300 /H9262m). The mean particle size and polydispersity
index of powder blends were calculated according to a method
previously described [18] .
The characterization of powder blends in terms of flow prop-
erties was done according to European Pharmacopoeia methods
[19] . The determined characteristics were Carr’s index, Hausner
ratio and time of flow.
2.6. Data processing
Different spectra pre-processing methods were used in order
to enhance the information searched for the study, to decrease
the influence of the side information contained in the spectra and
finally to construct the calibration methods. The pre-processing
methods were applied in combination with the whole spectra or
different spectral regions. Development of model for indapamide
assay was based on using two combinations of data pre-treatment
methods: first derivative (FD) followed by standard normal variate
(SNV) and first derivative followed by multiplicative scatter correc-
tion (MSC). In order to construct the calibration model for particle
size characterization, the spectral pre-treatments tested were con-
stant offset elimination (COE) and SNV. Similarly, some spectral
pre-treatment methods were used in order to develop the calibra-
tion model for flow property determination, namely: straight line
subtraction (SLS), FD followed by SLS and FD followed by SNV.
The OPUS 6.5 (Bruker Optics, Germany) software package
has been used for multivariate regressions, with no spectraTable 2
Statistical
parameters and number of principal components for different models
proposed
for indapamide assay, without data pre-treatment as well as after different
spectra
pretreatments.
Model (a) (b) (c)
Pre-treatment None FD + SNV FD + MSC
Spectral range selected (cm−1) 9000–5950, 5760–5340
Number of PLS factors 8 5 5
R20.916 0.984 0.986
RMSECV (%, w/w) 0.116 0.078 0.076
RMSEE (%, w/w) 0.077 0.054 0.056
Bias −0.0256
−0.0018 0.0025
pre-treatment as well as after applying pre-processing methods.
This software permits models’ validation by full cross-validation.
In this procedure, iterative calibrations were performed by remov-
ing in turn each standard from the training set and then predicting
the excluded sample with that calibration [20] .
The QUANT package of the OPUS 6.5 software was used for
the development of prediction models based on partial least
squares (PLS) regression. The models were validated by full cross-
validation, by removing in turn each standard from the training set
and then predicting the excluded sample with that model.
Modde 9.0 software was used to find correlations between frac-
tions used to obtain synthetic powder blend and particle size and
flow characteristics of obtained powder blend.
3. Results and discussion
3.1. Indapamide assay
The first aim of our research was to develop a NIR-chemometric
method suitable for the direct quantification of indapamide in
powder blends for tabletting. For this purpose, a protocol consist-
ing of preparing and analyzing 15 synthetic powder mixtures has
been followed. The drug content of indapamide formulations is
currently determined by HPLC technique, but this is quite time-
consuming in the case of high volume production. In this context,
we have developed and validated a NIR-chemometric method for
rapid quantification of indapamide in powder blends, which could
be a viable alternative for the conventional method since it is more
rapid than conventional technique and does not require any sample
preparation.
3.1.1. Spectra investigation
The NIR reflectance spectrum for 3 powder blends at 1.67%,
2.08% and 2.5% (w/w) indapamide content is presented in Fig. 1. As
seen in this figure, the intensive spectral peaks of indapamide are
mainly in the region of 9000–4000 cm−1. As a result, model devel-
opment for indapamide assay and future sample analysis were
based on the spectral regions 9000–5950 cm−1, 5760–5340 cm−1
and 4780–4475 cm−1. The spectral regions used in analysis were
selected before applying any spectral pre-treatment method.
3.1.2. Model development and validation
Model development for indapamide assay consisted in checking
several spectral pre-treatment methods (None, SNV, FD, MSC, SLS,
MMN, FD + SNV, FD + MSC and FD + SLS) in association with different
previously selected spectral regions. Among these methods, only
3 (None, FD + SNV and FD + MSC) were selected and presented in
Table 2, as they had best results. PLS regression was performed with
the calibration set and cross-validation was carried out for model
validation. The predictive ability of the chosen model was checked
on independent samples during method validation step.
The first step in model development was to investigate the
most adequate pre-processing method. This investigation and the

304 A. Porfire et al. / Journal of Pharmaceutical and Biomedical Analysis 70 (2012) 301– 309
Fig. 1. NIR absorption spectra of powder blends at 1.67%, 2.08% and 2.5% (w/w) indapamide content.
selection of the model with the biggest predictive potential was
mainly based upon the choice of the number of factors (princi-
pal components), and the calculation of root mean square error
of cross validation (RMSECV), root mean square error of estimation
(RMSEE), bias and R2. A large amount of models were generated
after applying different spectra pre-processing methods in com-
bination with different spectral regions. Table 2 shows the main
characteristics of the potentially interesting models.
Based on the analysis of different spectral regions, the mod-
els generated with the use of spectral regions 9000–5950 and
5760–5340 cm−1had best results. Among the models presented
in Table 2, it can be seen that the (b) and (c) models have better
prediction ability than (a) model. The (a) model is using too many
factors to obtain good prediction values and has very low R2and
higher values for RMSECV, RMSEE and bias than (b) and (c) models.
For the latter two models, the optimum number of PLS factors is
the same, and the values of the most relevant statistical parame-
ters were very close, so it was impossible to select the most fitted
model for indapamide assay based on these criteria.
3.1.3. Validation of the method
For validation purpose, independent batches comprising 4 repli-
cates of indapamide powder blends at 3 different active content
levels (1.67, 2.08, and 2.5%, w/w) were prepared and analyzed in 3
different days, resulting in a total of 36 spectra.
The predictive performance for models (b) and (c) was evalu-
ated with accuracy profiles computed on the external validation
results, in order to select the most fitted model for indapamide
assay. The accuracy profile has the advantage of taking into account
the total error, which is the sum of the trueness (systematic error)
and precision (random error), and meets the ICH Q2 (R1) guideline
requirements [21,22] .
Fig. 2(a) shows the accuracy profile for indapamide assay, based
on the validation results obtained with the developed NIR (b)
model, FD followed by SNV spectra-pretreatments, in the spectral
region 9000–5950 and 5760–5340 cm−1. Similarly, Fig. 2(b) shows
the accuracy profile for indapamide assay, based on the validation
results obtained with the developed NIR (c) model, FD followed
by MSC spectra-pretreatments, in the spectral region 9000–5950
and 5760–5340 cm−1. The acceptance limits were set at ±5%, as
required for the determination of active content in pharmaceu-
tical formulations. The ˇ-expectation tolerance limits should be
included in the acceptance limits. Seen from Fig. 2(a) and (b),the
ˇ-expectation tolerance limits are fully included within the
±5% acceptance limits, so it can be concluded that both mod-
els will provide results with adequate accuracy for indapamide
assay, whatever the active content of powder blends over the range
1.67–2.5% (w/w). The (c) model is however best fitted for our pur-
pose, because the ˇ-expectation tolerance limits for (b) model are
closer to the acceptance limits than those of the (c) model, espe-
cially between 1.67 and 2.08% indapamide content. As a result, the
Fig. 2. (a and b) Accuracy profile based on the validation results of the (b) and (c)
model.
The plain line is the relative bias, the dashed lines are the ˇ-expectation
tolerance
limits (ˇ = 95%) and the dotted curves are the acceptance limits (±5%).

A. Porfire et al. / Journal of Pharmaceutical and Biomedical Analysis 70 (2012) 301– 309 305
Fig. 3. Linear profile of NIR model for indapamide assay. The dashed limits on the
graph
correspond to the accuracy profile and the dotted curves represent the accep-
tance
limits at ±5% expressed in percentage of active content. The continuous line
is
the identity line y = x.
(c) model was chosen and further validated to be used for the deter-
mination of indapamide content in powder blends by NIR method.
Table 3 shows the ICH Q2 (R1) validation criteria of the devel-
oped method for indapamide assay. As seen in the accuracy profile
and in Table 3, the relative bias has values between −0.0032% and
0.0053% for all the concentration levels in the validation set.
The precision of the method was evaluated by calculating
two parameters: repeatability (intra-assay precision) and inter-
mediate precision (repeatability over different days) at the three
indapamide content levels taken in the validation protocol. Both
parameters had satisfactory values for all LPC concentration lev-
els. The repeatability and intermediate precision are both good for
all studied samples. The best repeatability values were obtained
at the highest indapamide content of powder blends, 2.5% (w/w),
while the best intermediate precision was obtained at 1.67% (w/w)
indapamide content.
As shown in Fig. 2(b) and Table 3, the accuracy of the method
is good whatever the indapamide content of powder blends, as the
ˇ-expectation tolerance limits do not exceed the ±5% acceptance
limits. The largest relative tolerance limits were (−4.3%, 4.7%), at the
highest indapamide content of the powder blends in the validation
batches.
The linear profile of the prediction model is shown in Fig. 3. The
linear model was represented by plotting the calculated concen-
trations of the validation samples as a function of the introduced
concentrations. As seen in the figure, the R2value is 0.994 and the
slope is very close to 1, confirming the linearity of the model for
indapamide assay.
According to data presented in Figs. 2 and 3 and Table 3,
the NIR-chemometric method using (c) model has reproducibil-
ity and satisfactory accuracy profile and linearity profile. After
analyzing the statistical parameters it can be concluded that the
NIR-chemometric method ((c) model) is linear and sufficiently pre-
cise and accurate for the assay of indapamide in powder blends for
tabletting.
3.1.4.
Application of the method
The results obtained for validation of NIR method ((c) model)
indicated that this method could be used for the determination of
indapamide content in powder blends with active content rang-
ing from 1.67 to 2.5% (w/w). Once validated, NIR method has been
applied for active content assay in 6 control powder blends contain-
ing 2.08% (w/w) indapamide, which is the expected indapamide
content in real powder blends used for tabletting. The reference
HPLC method has also been used for indapamide assay in the con-
trol samples. The NIR predicted values for indapamide content incontrol
samples were compared with values obtained by the refer-
ence HPLC method, in terms of active content recovery. The average
recovery was 99.92% for NIR method and 100.35% for HPLC method.
Student’s t test has been used for comparison of the two methods.
Results obtained did not show any statistical difference (p > 0.05)
between the results predicted by NIR model and reference values
obtained using HPLC method.
3.2. Particle size and flow characteristics
The second objective of our research was to develop a calibration
model for the prediction of particle size and flow characteristics
of powder blends. For this purpose, 22 powder blends with inda-
pamide were prepared according with a D-optimal experimental
design. These mixtures were characterized in terms of mean par-
ticle size, polydispersity index, Carr’s index, Hausner ratio and the
time of flow, using reference methods. The results are presented in
Table 4.
Modde 9.0 software was used to find correlations between frac-
tions used to obtain synthetic powder blends and the searched
responses, particle size and flow characteristics of obtained powder
blends. The influence of fractions used to obtain synthetic powder
mixtures on their particle size and flow characteristics is shown in
Fig. 4.
The results showed a very good correlation between fractions
used in the preparation of synthetic mixtures and particle size or
flow characteristics, respectively, as the R2values were greater than
0.9 and Q2values were higher than 0.8 for all models. Thus, a high
amount of fraction 0–100 /H9262m led to obtaining synthetic powder
blends with small mean diameter, large polydispersity index and
poor flow properties. On the other hand, a large amount of frac-
tion 200–300 /H9262m resulted in obtaining synthetic powder blends
with high mean diameter, small polydispersity index and good flow
properties.
3.2.1. Spectra investigation
The near infrared spectra of the three fractions (0–100, 100–200
and 200–300 /H9262m) used for the preparation of indapamide powder
blends in the calibration set are shown in Fig. 5. As shown in the fig-
ure, significant differences between the spectra of these fractions
are present especially in the range 7000–4000 cm−1. These differ-
ences were related to the granulometry of powder blend and not to
different indapamide concentration of the three fractions, because
indapamide is adsorbed on the surface of excipients during the mix-
ing of excipients and drug. If the spectral changes in Fig. 5 would
be related to different indapamide concentrations in the fractions,
the highest absorption would appear in the spectra of the granulo-
metric class 0–100 /H9262m, because indapamide particles belong to this
granulometric class. As a consequence, the region 7000–4000 cm−1
was further used for calibration of models.
3.2.2. Multivariate calibration for particle size and flow
characteristics
Different mathematical models based on multivariate calibra-
tion were applied to find direct correlation between NIR spectrum
of the powder blends in the calibration set and their particle
size characteristics. Table 5 shows the main characteristics of the
developed models for each particle fraction. When no spectra
pre-treatment method was applied, poor correlation between NIR
spectra and particle size characteristics was obtained (data not
shown). The best spectra pre-processing methods for all parti-
cle fractions seemed to be constant offset elimination (COE) and
straight line substraction (SLS). For the fraction with lower parti-
cle size, the best models were generated when using the spectral
region 10,000–4000 cm−1and SLS pre-processing method, having
lowest RMSECV and bias among the studied methods and the best

306 A. Porfire et al. / Journal of Pharmaceutical and Biomedical Analysis 70 (2012) 301– 309
Table
3
Validation
results of NIR method for indapamide assay.
Concentration level
(% indapamide)Mean indapamide
content
(%)Trueness Precision Accuracy
Relative
bias (%)Recovery
(%)Repeatability
(RSD%)Intermediate
precision (RSD%)Relative tolerance
limits
(%)Tolerance limits
(/H9262g
ml−1)
1.67 1.67 0.0015 100.09 1.17 1.16 [−3.3, 3.4] [1.61, 1.72]
2.08 2.08 −0.0032 99.85 1.34 1.17 [−3.4, 3.1] [2.01, 2.15]
2.50
2.51 0.0053 100.21 1.15 1.49 [−4.3, 4.7] [2.39, 2.62]
Table 4
Particle
size and flow characteristics of powder blends with indapamide. Each result is the mean of 3 determinations.
Exp. name Mean particle size (/H9262m) Polydispersity index Carr’s index Hausner ratio Time of flow (s)
N1 150.0 54.7 12.6 1.13 27.0
N2 92.0 78.8 17.5 1.20 40.3
N3 130.0
47.9 14.2 1.16 29.1
N4
116.7 57.4 14.4 1.17 32.0
N5
150.0 35.8 10.5 1.10 22.8
N6
113.6 56.9 14.2 1.17 32.1
N7 112.4
72.9 15.0 1.18 33.5
N8 164.3 39.1 10.9 1.12 22.4
N9
150.0 44.7 12.0 1.13 24.1
N10
131.8 63.5 12.9 1.15 29.6
N11
178.6 49.6 10.4 1.11 21.8
N12 192.9
38.0 8.8 1.09 18.9
N13 150.0 52.6 12.2 1.13 25.8
N14 207.1
35.3 8.7 1.08 17.8
N15 150.0 63.9 12.2 1.14 27.4
N16
159.1 56.9 12.2 1.13 25.6
N17
177.3 42.5 10.9 1.12 21.7
N18
186.4 34.7 9.7 1.10 19.8
N19 150.0 54.7 12.3 1.14 27.2
N20 150.0 54.7 12.8 1.15 24.2
N21 150.0
54.7 12.4 1.14 26.6
N22 150.0 54.7 12.8 1.15 26.5
prediction ability. The calibration model was based on PLS regres-
sion and the optimum number of factors was find to be 7. For the
other two fractions, the calibration models were based upon the
NIR spectrum of the sample in the region 10,000–6900 cm−1and
the use of COE spectra pre-processing method. The optimum num-
ber of PLS factors was 9 for the fraction of 100–200 /H9262m particle size
and 8 for the fraction of 200–300 /H9262m particle size.
Multivariate calibration was applied as well in order to find cor-
relations between NIR spectra of powder blends in the calibration
set and their flow properties. The predicted characteristics were
Carr’s index, Hausner ratio and the time of flow. The main character-
istics of the proposed models for the prediction of flow properties of
indapamide powder blends are shown in Table 6. As shown in this
table, the full spectral range 10,000–4000 cm−1was used for model
development and PLS regression has been applied. The best NIR
model for prediction of Carr’s index was based on pre-treatment
of spectra with FD followed by SLS method. The model used 5 PLS
factors, had a R2value of 0.930 and a RMSECV value of 0.72. The NIR
model chosen for prediction of Hausner ratio was built using the
combination of two spectra pre-treatment methods, FD and SNV. In
these conditions, a number of 6 principal components were foundto
be optimum. The model was characterized by a R2value of 0.925
and a RMSECV value of 0.0114. Ultimately, the best model for pre-
diction of the time of flow was based upon pre-treatment of NIR
spectra with SLS method. The prediction model had 8 PLS factors,
a R2value of 0.950 and a RMSECV of 1.17.
3.2.3.
Application of the methods
The NIR methods developed were implemented in order to pre-
dict particle size and flow characteristics of nine control samples,
indapamide powder blends. The samples were characterized in
terms of particle size, polydispersity index, Carr’s index, Hausner
ratio and time of flow using European Pharmacopoeia reference
methods. The same characteristics were predicted using the
previously developed NIR-chemometric methods. The measured
versus predicted characteristics of control samples are presented
in Table 7. As shown in the table, the relative error (%) of NIR
methods did not exceed 10% for all the parameters. The best
prediction results were obtained for Hausner ratio, the errors
being under 5% for its prediction. A good similarity was found
between the results of the proposed NIR chemometric method
and reference methods, as no significant difference (p > 0.05)
Table 5
Statistical
parameters and number of principal components for the models proposed for prediction of particle size.
Particle size 0–100 /H9262m 100–200 /H9262m 200–300 /H9262m
Pre-treatment COE SLS COE SLS COE SLS
Spectral
range selected (cm−1) 10,000–4000 10,000–6900 10,000–6900
Number
of PLS factors 7 7 9 8 8 7
R20.903 0.905 0.904 0.900 0.927 0.923
RMSECV
6.57 6.51 8.31 8.17 5.91 6.09
RMSEE 3.17
3.78 3.39 3.74 3.38 3.30
Bias
0.045 0.005 0.016 0.089 0.094 0.041

A. Porfire et al. / Journal of Pharmaceutical and Biomedical Analysis 70 (2012) 301– 309 307
(b) (a)
(d) (c)
(e) -20-1001020
X2*X3 X1*X2 X3 X2 X1-50510
X1
X2
X3
X1*X2
X2*X3
-2.0-1.00.01.02.0
X2*X3 X1*X3 X1*X2 X3 X2 X1-0.020-0.0100.0000.0100.0200.030
X2*X3 X1*X3 X1*X2 X3 X2 X1
-4-2024
X2*X3 X1*X3 X1*X2 X3 X2 X1
Fig. 4. (a–e) The influence of fractions used to obtain synthetic powder mixtures on their particle size (Y1), polydispersity index (Y2), Carr’s index (Y3), Hausner ratio (Y4) and
time
of flow (Y5). The input variables are as follows: X1, fraction 0–100 /H9262m; X2, fraction 100–200 /H9262m; X3, fraction 200–300 /H9262m.
Table 6
Statistical
parameters and number of principal components for the models proposed for prediction of powder flow characteristics.
Parameter for flow characterization Carr’s index Hausner ratio Time of flow
Pre-treatment FD + SLS FD + SNV FD + SLS FD + SNV SLS FD + SNV
Spectral
range selected (cm−1) 10,000–4000 10,000–4000 10,000–4000
Number
of PLS factors 5 4 6 6 8 5
R20.930 0.912 0.912 0.925 0.950 0.915
RMSECV
0.72 0.809 0.0123 0.0114 1.16 1.51
RMSEE 0.624
0.735 0.0103 0.0097 0.82 1.29
Bias 0.0067 0.0058 0.0042 0.0030 0.0022 0.0248
Table 7
Particle
size and flow characteristics of nine control powder blends (2.08% (w/w) indapamide), predicted using NIR method and determined by reference methods.
Particle size (/H9262m) Polydispersity index Carr’s index Hausner ratio Time of flow (s)
Predicted using NIR method (mean ± SD) 149.9 ± 46.1 59.9 ± 19.5 12.0 ± 2.8 1.14 ± 0.04 26.9 ± 9.4
Range of relative error (%) [−7.14, 6.75] [−7.85, 8.77] [−9.92, 6.69] [−0.86, 4.84] [−8.37, 9.92]
Determined
using reference method (mean ± SD) 153.0 ± 49.7 57.0 ± 18.8 12.2 ± 2.7 1.13 ± 0.04 27.1 ± 10.5
t-Value 0.1396
0.3252 0.1794 0.5222 0.0336
p-Value
0.8907 0.7493 0.8598 0.6086 0.9735
Relative error (%) was calculated as 100 ×(NIR − HPLC)/HPLC. p-Value greater than 0.05 indicates two means are similar.

308 A. Porfire et al. / Journal of Pharmaceutical and Biomedical Analysis 70 (2012) 301– 309
Fig. 5. NIR absorption spectra of the three fractions (0–100, 100–200 and 200–300 /H9262m) used for the preparation of indapamide powder blends in the calibration set.
between reference and predicted values was found for all studied
parameters. Therefore, the developed NIR methods are useful tools
for physical characterization of powder blends before tabletting.
4. Conclusions
In this work NIR methods based on PLS regression were devel-
oped for active content assay, granulometric analysis and assessing
flowability of indapamide powder blends.
Firstly, a NIR model appropriate for determination of inda-
pamide content in powder blends for tabletting (2.08%, w/w) has
been developed. The model was validated in terms of trueness,
precision and accuracy, for an active content ranging from 1.67 to
2.5%. Furthermore, it was shown that the proposed method has
a good similarity with reference HPLC method currently used for
indapamide assay.
Secondly, pharmaceutical characterization of powder blends
was done through NIR methods proposed for prediction of gran-
ulometric properties (mean particle size and polydispersity index)
and powder flow characteristics (Carr’s index, Hausner ratio and
time of flow). These methods were developed based on the finding
that a very good correlation exists between granulometric fractions
used in the preparation of powder blends and particle size or flow
characteristics, respectively. Comparing the NIR predicted values
of these parameters with those obtained using reference methods,
the differences were not significant and the prediction errors were
under 10%.
In conclusion, the overall results of this paper demonstrate that
near infrared spectroscopy methods are appropriate and advanta-
geous for predicting both chemical and pharmaceutical properties
of powder blends.
Acknowledgment
This work was supported by CNCSIS-UEFISCSU Romania, project
number PNII – IDEI ID 1350/2008.
References
[1] R.E. O’Connor, J.B. Schwartz, L.A. Felton, Powders, in: Remington. The Sci-
ence and Practice of Pharmacy, 21st edition, Lippincott Williams and Wilkins,
Philadelphia, 2005, pp. 702–719.
[2] European Pharmacopoeia 7.2, 2.9.40 Uniformity of dosage units, 2011.[3] C.V. Liew, A.D. Karande, P.W.S. Heng, In-line quantification of drug and excip-
ients in cohesive powder blends by near infrared spectroscopy, Int. J. Pharm.
386 (2010) 138–148.
[4] M. Blanco, M. Alcala, J.M. Gonzalez, E. Torrasz, A process analytical technology
approach based on near infrared spectroscopy: tablet hardness, content uni-
formity, and dissolution test measurements of intact tablets, J. Pharm. Sci. 95
(2006) 2137–2144.
[5] D. Christensen, M. Allesø, I. Rosenkrands, J. Rantanen, C. Foged, E.M. Agger, P.
Andersen, H.M. Nielsen, NIR transmission spectroscopy for rapid determination
of lipid and lyoprotector content in liposomal vaccine adjuvant system CAF01,
Eur. J. Pharm. Biopharm. 70 (2008) 914–920.
[6] M. Blanco, A. Peguerro, Influence of physical factors on the accuracy of calibra-
tion models for NIR spectroscopy, J. Pharm. Biomed. Anal. 52 (2010) 59–65.
[7] W. Li, A. Bashai-Woldu, J. Ballard, M. Johnson, M. Agresta, H. Rasmussen, S. Hu, J.
Cunningham, D. Winstead, Applications of NIR in early stage formulation devel-
opment. Part I. Semi-quantitative blend uniformity analyses by reflectance NIR
without calibration models, Int. J. Pharm. 340 (2007) 97–103.
[8] I. Tomuta, R. Iovanov, A.L. Vonica, S.E. Leucuta, High-throughput NIR-
chemometric method for meloxicam assay from powder blends for tabletting,
Sci. Pharm. 79 (2011) 885–898.
[9] P. Chalus, Y. Roggo, S. Walter, M. Ulmschneider, Near-infrared determination
of active substance content in intact low-dosage tablets, Talanta 66 (2005)
1294–1302.
[10] M. Blanco, M. Castillo, A. Peinado, R. Beneyto, Determination of low-
analyte concentrations by near-infrared spectroscopy: effect of spectral
pre-treatments and estimation of multivariate detection limits, Anal. Chim.
Acta 581 (2007) 318–323.
[11] E. Ziemons, J. Mantanus, P. Lebrun, E. Rozet, B. Evrard, Ph. Hubert,
Acetaminophen determination in low-dose pharmaceutical syrup by NIR spec-
troscopy, J. Pharm. Biomed. Anal. 53 (2010) 510–516.
[12] M.C. Sarraguc ¸ a, A.V. Cruz, H.R. Amaral, P.C. Costa, J.A. Lopes, Comparison of
different chemometric and analytical methods for the prediction of particle
size distribution in pharmaceutical powders, Anal. Bioanal. Chem. 399 (2011)
2137–2147.
[13] M. Alcala, M. Blanco, M. Bautista, J.M. Gonzalez, On-line monitoring of a gran-
ulation process by NIR spectroscopy, J. Pharm. Sci. 99 (2010) 336–345.
[14] J. Ropero, L. Beach, M. Alcala, R. Rentas, R.N. Dave, R.J. Romanach, Near-infrared
spectroscopy for the in-line characterization of powder voiding. Part I: Devel-
opment of the methodology, J. Pharm. Innov. 4 (2009) 187–197.
[15] L. Beach, J. Ropero, A. Mujumdar, M. Alcala, R.J. Romanach, R.N. Dave, Near-
infrared spectroscopy for the in-line characterization of powder voiding. Part
II: Quantification of enhanced flow properties of surface modified active phar-
maceutical ingredients, J. Pharm. Innov. 5 (2010) 1–13.
[16] M. Chaffman, R.C. Heel, R.N. Brodgen, Indapamide. A review of its pharmaco-
dynamic properties and therapeutic efficacy in hypertension, Drugs 28 (1984)
189–235.
[17] P. Schiavi, R. Jochemsen, D. Guez, Pharmacokinetics of sustained and immediate
release formulations of indapamide after single and repeated oral administra-
tion in healthy volunteers, Fundam. Clin. Pharmacol. 14 (2000) 139–146.
[18] J. Staniforth, Particle-size analysis, in: M.E. Aulton (Ed.), Pharmaceutics. The Sci-
ence of Dosage Form Design, 2nd edition, Churchill Livingstone, Philadelphia,
2002, pp. 152–165.
[19] European Pharmacopoeia 7.2, 2.9.36 Powder flow, 2011.
[20] A. Espinosa-Mansilla, F. Salinas, I. De Orbe Paya, Simultaneous determination
of sulfadiazine, doxycycline, furaltadone and trimethoprim by partial least
squares multivariate calibration, Anal. Chim. Acta 313 (1995) 103–112.

A. Porfire et al. / Journal of Pharmaceutical and Biomedical Analysis 70 (2012) 301– 309 309
[21] Ph. Hubert, J.J. Nguyen-Huu, B. Boulanger, E. Chapuzet, P. Chiap, N. Cohen,
P.A. Compagnon, W. Dewé, M. Feinberg, M. Lallier, M. Laurentie, N. Mercier,
G. Muzard, C. Nivet, L. Valat, Harmonization of strategies for the validation of
quantitative analytical procedures: a SFSTP proposal – Part I, J. Pharm. Biomed.
Anal. 36 (2004) 579–586.[22] Ph. Hubert, J.J. Nguyen-Huu, B. Boulanger, E. Chapuzet, P. Chiap, N. Cohen, P.A.
Compagnon, W. Dewé, M. Feinberg, M. Lallier, M. Laurentie, N. Mercier, G.
Muzard, C. Nivet, L. Valat, E. Rozet, Harmonization of strategies for the valida-
tion of quantitative analytical procedures: a SFSTP proposal – Part II, J. Pharm.
Biomed. Anal. 45 (2007) 70–81.