Journal of Pharmaceutical and Biomedical Analysis 149 (2018) 329334

Contents lists available at ScienceDirect

Journal of Pharmaceutical and Biomedical Analysis

journal homepage: www.elsevier.com/locate/jpba

Short communication

Raman spectroscopy as a PAT for pharmaceutical blending:

Advantages and disadvantages
Daniela Riolo a , Alessandro Piazza a , Ciro Cottini a , Margherita Serani a , Emilio Lutero a ,
Erika Cuoghi a , Lorena Gasparini a , Debora Botturi b , Iari Gabriel Marino c , Irene Aliatis c, ,
Danilo Bersani d , Pier Paolo Lottici d
a
Chiesi Farmaceutici S.p.A., R&D, Chemistry Manufacturing and Control, Via Palermo 26 A, 43122 Parma, Italy
b
SAIMP srl, Via Mangini28/D/1, 16031 Sori, Genova, Italy
c
OPTOTESYS, Via Giardino 3/A, 42016 Guastalla (RE), Italy
d
Universit di Parma, Dipartimento di Scienze Matematiche, Fisiche e Informatiche, Parco Area delle Scienze 7/A, 43124 Parma, Italy

a r t i c l e i n f o a b s t r a c t

Article history: Raman spectroscopy has been positively evaluated as a tool for the in-line and real-time monitoring of
Received 29 April 2017 powder blending processes and it has been proved to be effective in the determination of the endpoint
Received in revised form 7 November 2017 of the mixing, showing its potential role as process analytical technology (PAT). The aim of this study
Accepted 8 November 2017
is to show advantages and disadvantages of Raman spectroscopy with respect to the most traditional
Available online 8 November 2017
HPLC analysis. The spectroscopic results, obtained directly on raw powders, sampled from a two-axis
blender in real case conditions, were compared with the chromatographic data obtained on the same
Keywords:
samples. The formulation blend used for the experiment consists of active pharmaceutical ingredient
Raman spectroscopy
High performance liquid chromatography
(API, concentrations 6.0% and 0.5%), lactose and magnesium stearate (as excipients).
(HPLC) The rst step of the monitoring process was selecting the appropriate wavenumber region where the
Process analytical technologies (PAT) Raman signal of API is maximal and interference from the spectral features of excipients is minimal. Blend
Powder blending processes proles were created by plotting the area ratios of the Raman peak of API (AAPI ) at 1598 cm1 and the
Pharmaceuticals Raman bands of excipients (AEXC ), in the spectral range between 1560 and 1630 cm1 , as a function of
mixing time: the API content can be considered homogeneous when the time-dependent dispersion of
the area ratio is minimized. In order to achieve a representative sampling with Raman spectroscopy, each
sample was mapped in a motorized XY stage by a defocused laser beam of a micro-Raman apparatus.
Good correlation between the two techniques has been found only for the composition at 6.0% (w/w).
However, standard deviation analysis, applied to both HPLC and Raman data, showed that Raman results
are more substantial than HPLC ones, since Raman spectroscopy enables generating data rich blend
proles. In addition, the relative standard deviation calculated from a single map (30 points) turned out
to be representative of the degree of homogeneity for that blend time.
2017 Elsevier B.V. All rights reserved.

1. Introduction
In pharmaceutical industry, establishing the correct uniform
distribution of active ingredients and excipients is a signicant
challenge. The nal product uniformity has a crucial role, as it
affects the quality, safety and efcacy of the pharmaceutical drug.
At present, the most widely used method for dening drug content
uniformity of blend is the HPLC method, based on sampling of 13
dose of drug at different positions in the blender, using a sample
Corresponding author.
E-mail address: irene.aliatis@gmail.com (I. Aliatis).
thief (thief probe). As this off-line analysis is labour intensive and
time-consuming and there are possible errors associated with thief
probe (e.g., disruption of the mixture when the probe is inserted,
preferential sampling of free owing components, segregation,
etc.), at-line, on-line and mostly in-line spectroscopic techniques
have been suggested in literature. In this work, we focus our atten-
tion on Raman spectroscopy. Raman spectroscopy has already been
described for various in-process monitoring, such as blend homo-
geneity, granulation and drying processes, tablet evaluation and
coating monitoring, revealing to be a signicant tool for the imple-
mentation of PAT [15]. Since the last decade, transmission Raman
spectroscopy has been suited for the analysis of tablets and capsules
[612], while conventional Raman apparatuses in backscattering

https://doi.org/10.1016/j.jpba.2017.11.030
0731-7085/ 2017 Elsevier B.V. All rights reserved.
330 D. Riolo et al. / Journal of Pharmaceutical and Biomedical Analysis 149 (2018) 329334
geometry were mostly applied for in situ analysis of powder blend-
ing. However, to the best of our knowledge, there are relatively few
studies describing the use of in-line backscattering Raman spec-
troscopy for this purpose [1316]. The major problem when dealing
with conventional backscatter Raman spectrometers is the laser
spot size (generally smaller than 500 m diameter), which enables
the measurement on small volume of sample. Tricks to overcome
the sub-sampling limitations of this technique are achieved in lit-
erature by continuously rotating the sample during measurement
[3,17], by scanning at several positions the sampling area [6,17] or
by employing defocused probes or wide area optics that produce
laser beam diameters of 37 mm [1820]. Allan et al. [13] describe
the rst use of a PhAT probe (Kaiser Optical Systems Inc.) for in-situ
Raman monitoring of powder mixing in a high shear blender. They
found that the sampling volume of a PhAT probe (3 mm spot size)
was approximately 1300 times larger than that of a non-contact
Raman probe (150 m spot size) and approximately 16700 times
larger than that of an immersion probe (60 m spot size). In all
the aforementioned studies, the results obtained by in-line Raman
spectroscopy were veried measuring the content of uniformity of
the blends by off-line HPLC analyses [15,16] or by other in-line tech-
niques, as near infrared spectroscopy (NIR) [13,14] and acoustic
emission spectroscopy (AES) [13]. No comparison between off-line
methods, analysing the same thief probe, is performed. The goal of
this work is to evaluate the Raman and HPLC results obtained on
the same thief probe, sampled from a two-axis blender in real case
conditions (i.e. three-component drug formulation, particle sizes,
loading of the blender, mixing speed) and compare the found end-
points of the powder blending process. Moreover, the response of
Raman spectroscopy on low dose of API (i.e. 0.5%) and its effect on
blending time is evaluated.
2. Materials and methods
2.1. Materials
The studied formulation was a powder blend of API and excipi-
ents with two nal API concentrations of 6.0 (600 g/10 mg blend)
and 0.5 (50 g/10 mg blend) % (w/w). The used excipients were lac-
tose (Molkerei MEGGLE Wasserburg GmbH & Co. KG, Wasserburg,
Germany) and magnesium stearate (Peter Greven GmbH & Co.KG,
Bad Munstereifel, Germany). The particle sizes range from less than
7 m (90% of API particles) and 15 m (90% of ne excipients par-
ticles) to more than 350 m (as coarse lactose that represents the
major component of the formulation). For condentiality reasons,
no more qualitative or quantitative information about the formu-
lation can be disclosed.
2.2. Mixing
After weighing the correct amounts, all the components were
transferred into the blender (Dyna-MIX , Willy A. Bachofen AG
Maschinenfabrik, Muttenz, Switzerland). A blender, based on a
three-dimensional movement, with two simultaneous rotation on
two axes X and Y, was used to prepare the samples. In the mixer
process each axis rotates for a given time with two different speeds.
At the end of each cycle the rotation direction changes. The process
is repeated until the end of the mixing. Different batch sizes can be
manufactured with this equipment using different drums, related
to the total amount of blend prepared: 22, 55 and 80 liters for batch
sizes of 8, 20 and 35 kg. For this experiment, two 8 kg batches size
were prepared. The blending process can be summarised as follows:
1) carrier: carrier excipients are mixed without API; 2) weighing
API step; 3) pre-dispersion phase: API is added to an aliquot of car-
rier (the mixing time is 54 and 62 min for the higher and the lower
strength, respectively); 4) pre-mixing phase: the remaining part of
the carrier is added to the pre-dispersion phase and mixed (the pro-
cess lasts 72 min for both strengths); 5) after that, a sieving step was
performed; 6) nal mix (the mixing lasts 60 min for both strengths;
this phase was not sampled and analysed).
The mixing kinetics of the two mixtures (i.e. 6.0 and 0.5%) was
monitored during the steps 3) and 4) of the mixing process.
A single compartment end-sampling thief probe was used to
collect samples (1030 mg) from nine locations in the mixture
(three at the top, three at the middle and three at the bottom of
the mixing-drums). Each sample was subdivided in two vials: one
for Raman analysis and one for HPLC analysis. A total of 207 drug
samples, dened by the API concentration, the regime of the mix-
ing process and the mixing time (in min), were measured by both
techniques. Samples are summarised in Table 1.
2.3. Raman analysis
A Jobin Yvon Horiba LabRam confocal Raman spectrometer
(Horiba, Kyoto, Japan), equipped with a Nd:YAG blue laser at
473.1 nm, and attached to an Olympus BX40 microscope, mounting
an objective with 4X magnication and NA = 0.20, was employed.
The blue laser was selected because of its higher efciency (4 )
and the better response of the detector to that wavelength. This
laser operates at 100 mW at source. No risk of thermal degrada-
tion of the drug occurred because, in order to obtain a large area of
measure, less sensitive to microscopic distributions of the powders,
the laser beam was defocused (the microscope stage was low-
ered of about 4.5 mm from the focus position). An area of 30 mm2
(6 5 points) was automatically scanned. The confocal hole was set
to 900 m, corresponding to the maximum aperture. The Raman
maps were measured on loose grains (the amount of the grains is
the quantity of powder taken with a little spatula from the vial).
The spectra were acquired on the map at steps of 1 mm, with an
exposure time of 1 s and 18 scans for spectrum. The spectral res-
olution was about 3.5 cm1 . For each sample, 30 different points
were measured. Data were acquired using LABSPEC 5.78.24 soft-
ware package (Jobin Yvon Horiba). Spectra were extracted in the
spectral range 8001800 cm1 , baseline corrected by a 4th degree
polynomial curve and normalized to the area of the Raman band of
API at 1598 cm1 .
2.4. HPLC analysis
The samples were transferred to glass vials, then dissolved into
volumetric asks with Water/Acetonitrile 40/60 v/v (milliQ Puried
water, Sigma-Aldrich, Milan, Italy) and analysed for their drug con-
tent using a previously developed HPLC method. Chromatographic
separation was performed on an Agilent 1200 series liquid chro-
matographic system (Agilent Technologies, Waldbronn, Germany),
equipped with a quaternary pump, a diode-array detector (DAD)
and a thermostated autosampler. The chromatographic run was
performed by using an Atlantis C18 column, 3 m, 150 3.9 mm
(Waters, Milford, Massachusetts) at a ow rate of 1.0 ml/min main-
tained at 40 C throughout the analyses. The mobile phase was
composed of Sodium dihydrogen phosphate 0.02 M (Sodium dihy-
drogen phosphate anidrous, Honeywell Fluka, Bucharest, Romania)
corrected at pH 3.0 with H3 PO4 85% buffer solution (Sigma-Aldrich
ACS Reagent, Sigma-Aldrich, Milan, Italy) (phase A) and Acetoni-
trile (Sigma-Aldrich Analytical grade, Sigma-Aldrich, Milan, Italy)
(phase B). Separation was achieved using a gradient from 40% to 10%
(phase A) after which the mobile phase was returned to the start-
ing conditions and re-equilibrated for 2 min. The total run time was
8 min. The injection volume was 50 l, and the autosampler tem-
perature was maintained at 25 C throughout the analyses. Each
sample is twice injected and the results are averaged.
 D. Riolo et al. / Journal of Pharmaceutical and Biomedical Analysis 149 (2018) 329334 331

Table 1
List of the investigated samples.

API concentration 6.0% (w/w) API concentration 0.5% (w/w)

Time Blender positiona Time Blender positiona

PRE-DISPERSION
T1 = 10 min 1 T, 2 T, 3 T, 1 M, 2 M, 3 M, 1 B, 2 B, 3 B T1 = 10 min 1 T, 2 T, 3 T, 1 M, 2 M, 3 M, 1 B, 2 B, 3 B
T2 = 20 min 1 T, 2 T, 3 T, 1 M, 2 M, 3 M, 1 B, 2 B, 3 B T2 = 20 min 1 T, 2 T, 3 T, 1 M, 2 M, 3 M, 1 B, 2 B, 3 B
T3 = 30 min 1 T, 2 T, 3 T, 1 M, 2 M, 3 M, 1 B, 2 B, 3 B T3 = 30 min 1 T, 2 T, 3 T, 1 M, 2 M, 3 M, 1 B, 2 B, 3 B
T4 = 40 min 1 T, 2 T, 3 T, 1 M, 2 M, 3 M, 1 B, 2 B, 3 B T4 = 40 min 1 T, 2 T, 3 T, 1 M, 2 M, 3 M, 1 B, 2 B, 3 B
T5 = 54 min 1 T, 2 T, 3 T, 1 M, 2 M, 3 M, 1 B, 2 B, 3 B T5 = 50 min 1 T, 2 T, 3 T, 1 M, 2 M, 3 M, 1 B, 2 B, 3 B
T6 = 62 min 1 T, 2 T, 3 T, 1 M, 2 M, 3 M, 1 B, 2 B, 3 B

PRE-MIXING
T1 = 12 min 1 T, 2 T, 3 T, 1 M, 2 M, 3 M, 1 B, 2 B, 3 B T1 = 12 min 1 T, 2 T, 3 T, 1 M, 2 M, 3 M, 1 B, 2 B, 3 B
T2 = 24 min 1 T, 2 T, 3 T, 1 M, 2 M, 3 M, 1 B, 2 B, 3 B T2 = 24 min 1 T, 2 T, 3 T, 1 M, 2 M, 3 M, 1 B, 2 B, 3 B
T3 = 36 min 1 T, 2 T, 3 T, 1 M, 2 M, 3 M, 1 B, 2 B, 3 B T3 = 36 min 1 T, 2 T, 3 T, 1 M, 2 M, 3 M, 1 B, 2 B, 3 B
T4 = 48 min 1 T, 2 T, 3 T, 1 M, 2 M, 3 M, 1 B, 2 B, 3 B T4 = 48 min 1 T, 2 T, 3 T, 1 M, 2 M, 3 M, 1 B, 2 B, 3 B
T5 = 60 min 1 T, 2 T, 3 T, 1 M, 2 M, 3 M, 1 B, 2 B, 3 B T5 = 60 min 1 T, 2 T, 3 T, 1 M, 2 M, 3 M, 1 B, 2 B, 3 B
T6 = 72 min 1 T, 2 T, 3 T, 1 M, 2 M, 3 M, 1 B, 2 B, 3 B T6 = 72 min 1 T, 2 T, 3 T, 1 M, 2 M, 3 M, 1 B, 2 B, 3 B
a
T stands for at the top, M stands at the middle and B stands at the bottom of the mixing bowl.

Since the analysed molecule is still under development, it is not

possible to completely describe the applied analytical chromato-
graphic conditions.

2.5. Statistics and correlation study

As the time required to obtain a homogeneous blend was

the response variable, a mathematical method that allows rapid
homogeneity detection by extracting suitable information from
the collected Raman data was elaborated. Data analysis has been
performed by Matlab (Mathworks Inc., Natick, MA, USA) and the
following values have been calculated:

1) the ratios between the areas of the Raman band of API at

1598 cm1 and the Raman bands of excipients in the spec-
tral range between 1560 and 1630 cm1 (AR = AAPI /AEXC ): i.e. 30
ratios (AR ) for each map and 270 ratios (AR ) for each sample
(nine different samplings from the mixing drum), for a total of
6210 points of analysis;
2) for each Raman map, the averaged ratio (<AR >) and the relative
standard deviation (RSD30 ) of the 30 AR values;
3) from the nine RSD30 , their averaged value for each time (Mean
RSD30 in the graphs);
4) for each mixing time, the grand mean (Mean in the graphs)
of the nine averaged values of <AR > and the relative standard
deviation (RSD9 ) of the nine <AR >. These data will be used for
the subsequent comparison with HPLC;
5) the averaged ratio (<AR >270 ) and the relative standard deviation
(RSD270 ) of all the 270 AR , thus disregarding the nine different
samplings into the blender.
On the chromatographic data side, the area of API, averaged over
two injections, has been measured and calibrated by the internal
standard to obtain the corresponding API amount. Then, the assay
(<AC >) has been calculated. For each mixing time, the averaged
value (Mean) and the relative standard deviation (RSD9 ) of the nine
assays have been calculated.
In summary, we have a total of 207 Raman maps and 207 chro-
matograms (414 considering the twice injection). Each Raman map,
composed of 30 spectra, is thus equivalent to a chromatographic
run. Blend proles are created by plotting the aforementioned sta-
tistical parameters as a function of mixing time. In particular, in
order to directly compare Raman and HPLC results, the Mean and
RSD9 values from both techniques have been used.
Fig. 1. Raman spectra of API (a), lactose (b) and Mg stearate (c). The gray areas show
the spectral ranges of interest for this study.
3. Results and discussion
3.1. Raman method
The present work is a comparative study between Raman and
HPLC results, obtained from a blending process of a pharmaceutical
drug in real case conditions (i.e. three-component drug formulation,
particle sizes, loading of the blender, mixing speed), to monitor the
mixing kinetics.
3.1.1. Standards
The rst step of the monitoring process was selecting an
appropriate wavenumber region where the Raman signal of the
component of interest (i.e. API) is maximal and interference from
other components (i.e. excipients) is minimal. In Fig. 1 individ-
ual Raman spectra of API, lactose and Mg stearate are illustrated.
The most intense band for the API was centred at 1598 cm1 ; this
peak originates from the unsaturated carbon bonds (aromatic rings)
present in the drug molecule. As no interference from the excipients
was observed in this region, this spectral feature (more specically
the spectral range between 1560 and 1630 cm1 ) was selected as
the focal point of this study. The effect of mixing on the Raman sig-
nal was evaluated by measuring the ratio between this peak area
and the multi-peak area in the spectral range 12001525 cm1 , rel-
ative to excipients, which includes the Raman bands at 1263, 1330,
332 D. Riolo et al. / Journal of Pharmaceutical and Biomedical Analysis 149 (2018) 329334
Fig. 2. Comparison between Raman (left) and HPLC results (right) calculated for the blend with API concentration of 6.0% (w/w). (Top) Raman ratios and HPLC assays; (bottom)
RSD9 curves. The pre-dispersion times are labelled on the x-axis.
1340, 1348, 1360, 1380, 1415, 1455 and 1470 cm1 of lactose and
the Raman bands at 1297, 1435 and 1460 cm1 of Mg stearate.
3.1.2. Overcoming of the sub-sampling limit
By specications, the blend is composed of powders with differ-
ent particle sizes. Fig. S1(a) clearly shows that the powder consists
of large crystals of lactose (ca. 250 m) covered by smaller grains of
API. By conventional micro-Raman spectroscopy, it has been pos-
sible to measure selectively the spectra on both the covered and
uncovered surface of the same crystal: API spectrum was present
in the former, whereas on the latter only lactose spectral features
were found. Nevertheless, this feature represents an obstacle for
the application of Raman spectroscopy to powder analysis. In this
study, to overcome this sub-sampling limitation, two solutions
have been combined: employing a defocused laser probe and mea-
suring several positions (maps) of the same powder sample. The
defocusing yields a larger sampling volume, while the registration
of maps of spectra allows a statistical analysis intra-sample (e.g.
RSD30 ) and a sound base for statistical analysis inter-samples of
different mixing times (e.g. RSD9 ).
3.1.3. Intra-map variability
In Fig. S1(b), the Raman spectra belonging to two maps obtained
on samples of powder from blends containing API at different con-
centrations (i.e., 6.0 and 0.5%), at the same mixing time (10 min),
are displayed. As expected, the intensity of the characteristic peak
of API is lower in the low dose spectrum (about 2.5 times smaller).
The proles, normalized on the API peak at 1598 cm1 , show that
the Raman bands are different on each point of the map, i.e. the
map has an intrinsic variability, but the aforementioned statistical
analysis allows extracting meaningful parameters.
3.1.4. Raman blend proles
The averaged area ratios <AR > and the relative standard devia-
tions (RSD9 ), calculated from the nine Raman maps for each mixing
time, for both the steps of mixing, were plotted as a function of
the mixing time. The average is a merit factor of the API concen-
tration in the powder, whereas the RSD represents the normalized
trend of the homogeneity in the mixing. Variability in the area ratios
and RSD values will minimize when a uniform blend is reached. In
Figs. 2 and 3 (left), the times, shown on the x-axis, belong to the
pre-dispersion step. The <AR > values are spread out and randomly
placed along the y-axis for each time. The positions of the different
samplings in the blender do not affect the results.
As concerns the higher strength (Fig. 2, top), the curve of the
mean values of <AR > shows an abrupt increase in the rst 20 min of
blending, from 20 to 40 min it reaches a plateau and then it starts
to grow up less rapidly. The initial upward trend in the mean val-
ues could be explained by the mixing process: when the degree of
mixing increases, API grains become more detectable by the Raman
probe, because they fully cover the surface of the excipients grains
and it seems that the concentration grows up. The RSD of the aver-
aged area ratios <AR > is shown in Fig. 2 (bottom): high RSD values
indicate a non-uniform blend. The API content can be considered
homogeneous after 20 min, since the time-dependent dispersion of
the <AR > is minimized. In Fig. 3 (top), an almost constant trend in
the <AR > curve is detected for the samples with a lower strength of
API. The RSD curve is shown in Fig. 3 (bottom).
In Fig. S2, the RSD of the 30 AR values calculated for each Raman
map, labelled as RSD30 , are displayed for both strengths, together
with the Mean RSD30 and the RSD270 values for each time of mixing.
The curves show similar trends. This means that a single sampling,
made by 30 points of analysis, is already statistically representative
and could be sufcient for the Raman investigation.
As regards the samples belonging to further mixing steps, i.e. the
pre-mixing, no interesting trends were found. For both strengths,
the points are randomly located. This could be explained by the
different API concentration of the blend in the two steps: in pre-
dispersion, only a fraction of the nal carrier amount is present
(roughly 25%), whereas all the amount of carrier is added in pre-
mixing. The dilution of the API concentration, produced when all
the amount of excipients is added and blended into the mixture,
negatively affects the signal-to-noise ratio in the Raman data. For
this reason the Raman data from the pre-mixing step are not shown.
 D. Riolo et al. / Journal of Pharmaceutical and Biomedical Analysis 149 (2018) 329334
Fig. 3. Comparison between Raman (left) and HPLC results (right) calculated for the blend with API concentration of 0.5% (w/w). (Top) Raman ratios and HPLC assays; (bottom)
RSD9 curves. The pre-dispersion times are labelled on the x-axis.
3.2. Comparison between Raman and HPLC methods
Comparing the results obtained by experimental techniques, as
Raman spectroscopy, with the traditional chromatographic meth-
ods is a common routine in pharmaceutical researches. Raman
spectroscopy has revealed to be a useful analysis for in-process
monitoring, such as blend homogeneity, granulation and drying
processes, tablet evaluation and coating control. In this study the
mixing degree of a pharmaceutical blend, with two different con-
centrations of API, has been measured from the Raman spectra
acquired on 207 thief probes. The goal of the work is to compare
the end-point of the powder blending process found by Raman with
that obtained by HPLC results, measuring the same thief probes.
The HPLC assays for the pre-dispersion step are shown in
Figs. 2 and 3 (right). In the case of 6.0% (w/w) API concentration
and pre-dispersion step, the RSD values calculated from the HPLC
assays show an abrupt decrease in the rst 20 min, as observed
by Raman. A signicant correlation (R2 = 0.90) between Raman and
HPLC curves is displayed (see Fig. S3). Similar methods have been
implemented by Vergote et al. in 2004 [16], who evaluated the
effect of mixing on a binary mixture obtained by a laboratory sim-
ulated blending process. Allan et al. in 2013 [13] performed in situ
Raman measurements to produce mixing proles in real time and
showed the advantages of Raman method over NIR spectrometry
when evaluating the mixing regimes for multi-component samples,
because of the superior Raman chemical specicity.
In contrast, (see Fig. 3), the agreement between Raman and HPLC
data for the blend with low dose (API concentration of 0.5% (w/w))
is weak: Raman area ratios are scattered and a meaningless trend
is found, whereas HPLC data display an upward trend. Only the
RSD curve from the HPLC data clearly shows the mixing end point.
Better results on low dose blend were found by Hausman et al.
in 2005 [15]. They evaluated the effect of blending time on a 1%
low dose blend of azimilide dihydrochloride. The authors found
that azimilide can be detected in the proposed model blend at lev-
els of approximately 0.1% near the detection limit, thanks to the
very good signal-to-noise ratio shown by this molecule. Low dose
azimilide blend uniformity was assessed by on-line Raman spec-
333
troscopy, as well as HPLC analysis of blend thief samples and tablets.
Moreover, they demonstrate how results from univariate Raman
analysis were consistent with a multivariate analysis.
4. Conclusions
The goal of this study was to verify if Raman spectroscopy
could be an effective substitute for HPLC methods in the monitor-
ing of the mixing kinetics of a pharmaceutical drug. Advantages
and disadvantages of the Raman technique have been found. A
good correlation between Raman and HPLC results was found for
the blend with API concentration of 6.0% (w/w) in pre-dispersion
regime, where both independent process-monitoring techniques
agree. The blend variability minimizes within 20 min. Moreover,
both techniques show no difference for the sampling position, sug-
gesting that a single sampling for time could be satisfactory for
the analysis. In particular, Raman spectroscopy allows generating
data rich blend proles, due to the ability to collect a large num-
ber of spectra from a single map. In addition, the Raman blend
proles were rapidly generated, compared to the lengthy time to
complete a blend prole with HPLC analysis. Although Raman spec-
troscopy yields important benets, in terms of time-consuming
and the possibility for in-line monitoring without operator, at the
present time it cannot substitute the HPLC analysis, because in all
the other investigated cases (i.e. low dose formulation and pre-
mixing regimes) the Raman curves are not reliable.
Acknowledgements
This research did not receive any specic grant from funding
agencies in the public, commercial, or not-for-prot sectors.
Appendix A. Supplementary data
Supplementary data associated with this article can be found, in
the online version, at https://doi.org/10.1016/j.jpba.2017.11.030.
334 D. Riolo et al. / Journal of Pharmaceutical and Biomedical Analysis 149 (2018) 329334
References
[1] J. Rantanen, Process analytical applications of Raman spectroscopy, J. Pharm.
Pharmacol. 59 (2007) 171177, http://dx.doi.org/10.1211/jpp.59.2.0004.
[2] T. De Beer, A. Burggraeve, M. Fonteyne, L. Saerens, J.P. Remon, C. Vervaet, Near
infrared and Raman spectroscopy for the in-process monitoring of
pharmaceutical production processes, Int. J. Pharm. 417 (2011) 3247, http://
dx.doi.org/10.1016/j.ijpharm.2010.12.012.
[3] H. Wikstrm, S. Romero-Torres, S. Wongweragiat, J.A.S. Williams, E.R. Grant,
L.S. Taylor, On-line content uniformity determination of tablets using
low-resolution Raman spectroscopy, Appl. Spectrosc. 60 (2006) 672681,
http://dx.doi.org/10.1366/000370206777670684.
[4] P. Matousek, A.W. Parker, Bulk Raman analysis of pharmaceutical tablets,
Appl. Spectrosc. 60 (2006) 13531357.
[5] J. Kim, J. Noh, H. Chung, Y.A. Woo, M.S. Kemper, Y. Lee, Direct, non-destructive
quantitative measurement of an active pharmaceutical ingredient in an intact
capsule formulation using Raman spectroscopy, Anal. Chim. Acta 598 (2007)
280285, http://dx.doi.org/10.1016/j.aca.2007.07.049.
[6] G.J. Vergote, C. Vervaet, J.P. Remon, T. Haemers, F. Verpoort, Near-infrared
FT-Raman spectroscopy as a rapid analytical tool for the determination of
diltiazem hydrochloride in tablets, Eur. J. Pharm. Sci. 16 (2002) 6367, http://
dx.doi.org/10.1016/S0928-0987(02)00058-1.
[7] J. Johansson, A. Sparen, O. Svensson, S. Folestad, M. Claybourn, Quantitative
transmission raman spectroscopy of pharmaceutical tablets and capsules,
Appl. Spectrosc. 61 (2007) 12111218, http://dx.doi.org/10.1366/
000370207782597085.
[8] P. Matousek, A.W. Parker, Recent advances in linear and nonlinear Raman
spectroscopy I, J. Raman Spectrosc. 38 (2007) 563567, http://dx.doi.org/10.
1002/jrs.1688.
[9] C. Eliasson, N.A. Macleod, L.C. Jayes, F.C. Clarke, S.V. Hammond, M.R. Smith, P.
Matousek, Non-invasive quantitative assessment of the content of
pharmaceutical capsules using transmission Raman spectroscopy, J. Pharm.
Biomed. Anal. 47 (2008) 221229, http://dx.doi.org/10.1016/j.jpba.2008.01.
013.
[10] A. Sparn, J. Johansson, O. Svensson, S. Folestad, M. Claybourn, Transmission
Raman spectroscopy for quantitative analysis of pharmaceutical solids, Am.
Pharmaceut. Rev. 12 (2009) 6271, http://dx.doi.org/10.1211/jpp.59.2.0005.
[11] M. Fransson, J. Johansson, A. Sparn, O. Svensson, Comparison of multivariate
methods for quantitative determination with transmission Raman
spectroscopy in pharmaceutical formulations, J. Chemometr. 24 (2010)
674680, http://dx.doi.org/10.1002/cem.1330.
[12] K. Buckley, P. Matousek, Recent advances in the application of transmission
Raman spectroscopy to pharmaceutical analysis, J. Pharm. Biomed. Anal. 55
(2011) 645652, http://dx.doi.org/10.1016/j.jpba.2010.10.029.
[13] P. Allan, L.J. Bellamy, A. Nordon, D. Littlejohn, J. Andrews, P. Dallin, In situ
monitoring of powder blending by non-invasive Raman spectrometry with
wide area illumination, J. Pharm. Biomed. Anal. 76 (2013) 2835, http://dx.
doi.org/10.1016/j.jpba.2012.12.003.
[14] T.R.M. De Beer, C. Bodson, B. Dejaegher, B. Walczak, P. Vercruysse, A.
Burggraeve, A. Lemos, L. Delattre, Y. Vander Heyden, J.P. Remon, C. Vervaet,
W.R.G. Baeyens, Raman spectroscopy as a process analytical technology (PAT)
tool for the in-line monitoring and understanding of a powder blending
process, J. Pharm. Biomed. Anal. 48 (2008) 772779, http://dx.doi.org/10.
1016/j.jpba.2008.07.023.
[15] D.S. Hausman, R.T. Cambron, A. Sakr, Application of Raman spectroscopy for
on-line monitoring of low dose blend uniformity, Int. J. Pharm. 298 (2005)
8090, http://dx.doi.org/10.1016/j.ijpharm.2005.04.011.
[16] G.J. Vergote, T.R.M. De Beer, C. Vervaet, J.P. Remon, W.R.G. Baeyens, N. Diericx,
F. Verpoort, In-line monitoring of a pharmaceutical blending process using
FT-Raman spectroscopy, Eur. J. Pharm. Sci. 21 (2004) 479485, http://dx.doi.
org/10.1016/j.ejps.2003.11.005.
[17] J. Johansson, S. Pettersson, S. Folestad, Characterization of different laser
irradiation methods for quantitative Raman tablet assessment, J. Pharm.
Biomed. Anal. 39 (2005) 510516, http://dx.doi.org/10.1016/j.jpba.2005.04.
029.
[18] M. Kim, H. Chung, Y. Woo, M. Kemper, New reliable Raman collection system
using the wide area illumination (WAI) scheme combined with the
synchronous intensity correction standard for the analysis of pharmaceutical
tablets, Anal. Chim. Acta 579 (2006) 209216, http://dx.doi.org/10.1016/j.aca.
2006.07.036.
[19] H. Wikstrom, I.R. Lewis, L.S. Taylor, Comparison of sampling techniques for
in-line monitoring using Raman spectroscopy, Appl. Spectrosc. 59 (2005)
934941.
[20] M.V. Schulmerich, W.F. Finney, R.A. Fredricks, M.D. Morris, Subsurface Raman
spectroscopy and mapping using a globally illuminated non-confocal
beroptic array probe in the presence of Raman photon migration, Appl.
Spectrosc. 60 (2006) 109114.