International Journal of Pharmaceutics 495 (2015) 290–301

Contents lists available at ScienceDirect

International Journal of Pharmaceutics

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

Continuous manufacturing of extended release tablets via powder

mixing and direct compression
Tuomas Ervastia , Simo-Pekka Simonahoa , Jarkko Ketolainena , Peter Forsbergb ,
Magnus Franssonb , Håkan Wikströmb , Staffan Folestadb , Satu Lakiob , Pirjo Tajarobib ,
Susanna Abrahmsén-Alamib,*
a
University of Eastern Finland, Faculty of Health Sciences, School of Pharmacy, Kuopio, Finland
b
AstraZeneca R&D, Mölndal, Sweden

A R T I C L E I N F O A B S T R A C T

Article history: The aim of the current work was to explore continuous dry powder mixing and direct compression for
Received 9 June 2015 manufacturing of extended release (ER) matrix tablets. The study was span out with a challenging
Received in revised form 23 August 2015 formulation design comprising ibuprofen compositions with varying particle size and a relatively low
Accepted 24 August 2015
amount of the matrix former hydroxypropyl methylcellulose (HPMC). Standard grade HPMC (CR) was
Available online 28 August 2015
compared to a recently developed direct compressible grade (DC2). The work demonstrate that ER tablets
with desired quality attributes could be manufactured via integrated continuous mixing and direct
Keywords:
compression. The most robust tablet quality (weight, assay, tensile strength) was obtained using high
Continuous manufacturing
Direct compression
mixer speed and large particle size ibuprofen and HPMC DC2 due to good powder flow. At low mixer
Continuous mixing speed it was more difficult to achieve high quality low dose tablets. Notably, with HPMC DC2 the
Extended release processing conditions had a significant effect on drug release. Longer processing time and/or faster mixer
HPMC speed was needed to achieve robust release with compositions containing DC2 compared with those
Adaptive control containing CR. This work confirms the importance of balancing process parameters and material
properties to find consistent product quality. Also, adaptive control is proven a pivotal means for control
of continuous manufacturing systems.
ã 2015 Elsevier B.V. All rights reserved.

1. Introduction Pharmaceuticals, White Papers, 2014; Poechlauer et al., 2012;

In a vast majority of modern industries continuous manufactur-
ing has over the years gradually replaced traditional step-by-step
manufacturing of products. The start-to-finish production lines
generally provides improved efficiencies. In the pharmaceutical
industry on the other hand, where highly advanced medicinal
products are produced, adoption of new manufacturing technologies
is rather slow and batch processing continues to be the dominating
process platform. Several reasons have been adressed that underpins
this slow manufacturing evolution such as regulatory policies, poor
intellectual property incentives and lack of proven off-the-shelf
equipment (Plumb, 2005; Nicholson, 2014; Lee et al., 2015). During
recent years industry and academia leaders have taken significant
initiatives to progress continuous manufacturing (Mascia et al.,
2013; International Symposium on Continuous Manufacturing of
* Corresponding author at: AstraZeneca R&D Mölndal, S-431 83 Mölndal, Sweden.
E-mail address: susanna.abrahmsen-alami@astrazeneca.com
(S. Abrahmsén-Alami).
Buchholz, 2010). In addition, the regulatory and quality consider-
ations have recently been summarized and assessed emphasizing
general support from health agencies for implementation of
continuous manufacturing (Lee et al., 2015; Allison et al., 2015;
ICHQ8-Q11 guidances). Nevertheless, remaining gaps and challenges
are identified that must be adressed to facilitate wider adoption and
implementation of continuous manufacturing as a viable approach
in pharmaceutical industry.
For marketed drug products the most common dosage form is a
tablet. In tablet manufacturing, typical disadvantages of batch
processing are associated with scale-up, lack or insufficient real
time quality control, insufficient process understanding, and long
manufacturing cycles (Leuenberger, 2001a,b; Vervaet and Remon,
2005). Here, continuous manufacturing holds a great potential to
overcome some or all of these classical problems. For example,
scale-up of production volumes is achieved with the same
equipment through longer processing times rather than by scaling
process equipment geometries as with batch processing (Leuen-
berger, 2001a). Thereby, manufacturing facilities can be made
more compact and costs for heating/cooling, ventilation and other

http://dx.doi.org/10.1016/j.ijpharm.2015.08.077
0378-5173/ ã 2015 Elsevier B.V. All rights reserved.
 T. Ervasti et al. / International Journal of Pharmaceutics 495 (2015) 290–301
costs can be reduced (Plumb, 2005; McKenzie et al., 2006; Spencer
et al., 2011). Finally, to enable appropriate control of continuous
manufacturing lines advanced on-line and in-line process and
material sensors will be a prerequisite and offer opportunities to
improved process understanding, product quality and yield
increase. In all, these advantages have raised a large interest in
the pharmaceutical industry for continuous manufacturing, albeit,
extensive experience with pharmaceutical applications is still
lacking.
The most straightforward example of a continuous tablet
manufacturing line is the integration of continuous dry powder
mixing and continuous direct compression. Here continuous
powder feeders have an essential role for the overall performance
of the continuous tablet manufacturing line. For example the tablet
quality will fluctuate if there is a large variability in the inflow
composition, even if the mixer and tablet press would perform
perfectly. Hence it is crucial to ensure that the feed rate of each
bulk material is controlled accurately (Marikh et al., 2005;
Pernenkil and Cooney, 2006; Portillo et al., 2008). Loss-in-weight
(LIW) feeders have been shown to be the most accurate material
feeding devices for dry powder based continuous pharmaceutical
manufacturing lines. Most of the research considering LIW feeders
has concentrated on the effect of powder properties and the feeder
design on the performance (Engisch and Muzzio, 2012). The effect
of screw design, charging container configuration, and interparti-
cle cohesion on solid flow has been studied by Hou et al. (2014).
They identified formation of three flow regimes in a complicated
manner in a screw feeder that must be considered to achieve
optimized operation.
Continuous mixing is currently a widely studied area (Pernenkil
and Cooney, 2006; Berthiaux et al., 2008; Portillo et al., 2008, 2009,
2010). Mechanistic insight into optimization of continuous dry
powder mixing has been reported by Gao et al. (2011) who
demonstrated improved performance by increasing blade speed
of a mixer while keeping a constant axial velocity. Portillo et al.
(2008) studied a continuous dry powder mixer, where the impeller
rotation, the operation angle, the number of blades and the angle of
blades could be adjusted. They found that decreasing the impeller
rotation rate and increasing the upward angle of the mixer gave
optimal processing conditions for that particular mixer. The mean
residence time, the time that the powder is inside the mixer, has been
one of the main variables affecting the mixing performance (Portillo
et al., 2008, 2009, 2010; Vanarase and Muzzio, 2011). When the
optimal mixer speed is studied, mixer properties, material properties
and total mass flow material have to be considered. While
engineering aspects of continuous dry powder mixing in general
have been extensively studied, also beyond pharmaceutical powders
(Pernenkil and Cooney, 2006; Berthiaux et al., 2008; Portillo et al.,
2008, 2009, 2010; Bridgewater, 2012), systematic studies comprising
material parameters are so far fewer. In addition, studies on the
performance and operational optimization of complete continuous
tablet manufacturing trains are also few. Vanarase and Muzzio
(2011) reported that mixing performance was largely dominated by
the material properties of the mixture, while in a later article they
found that the bulk density was the key material parameter affecting
the mean residence time (Vanarase et al., 2013). The importance of
incorporating powder material properties when assessing system
performance was also concluded by Boukouvala et al. (2012) in a
recent flowsheet modeling and sensitivity analysis of a whole
continuous tablet manufacturing line.
Previous studies considering continuous manufacturing of
tablets have focused on probing uniformity of product performance
only with regards to drug substance assay, drug distribution and
mechanical properties of tablets. In the present work the viability of
integrated continuous mixing and compaction processes for
manufacturing of extended-release matrix tablets is investigated.
291
The performance of tablets based on this formulation principle relies
not only on assuring assay and proper distribution of the drug
substance but also on the ability to provide extended and robust drug
release in-vivo. The latter depends mainly on the polymeric matrix
and its distribution in the composition which also needs to be
considered during continuous powder processing. Commercially
available hydroxypropyl methyl cellulose (HPMC) grades are mostly
designed for wet granulation and have a relatively narrow particle
size with a low mean particle size and therefore poor powder flow
characteristics. Those grades are therefore not suitable for high-
speed direct manufacturing processes. However, direct compress-
ible grades of HPMC has recently been launched and their
performance have been investigated (Mohamed et al., 2013; Rogers
et al., 2013; Heiman et al., 2015). Some investigations on direct
compression of hydrophilic matrices have been reported previous-
ly, but to the best of our knowledge, an integrated continuous
manufacturing via continuous dry powder mixing and continuous
direct compression has not yet been demonstrated. The overall aim
of the current work was to explore continuous mixing and direct
compression with a challenging formulation design with regards to
powder characteristics and composition. The design included low
dose compositions for which drug substance uniformity aspects
were tough and a relatively low amount of the matrix former
(HPMC) for which proper distribution was essential. Standard CR as
well as DC grade HPMC were included in the study. The processing
settings were varied broadly to stress the system further with the
aim to understand the critical relationship between raw materials
and final product performance (tablet mass variability, tablet tensile
strength, drug assay and drug release). The intent was both to enable
a deeper understanding of the critical relationship between raw
materials and final product performance but also to initially explore
how the feeder and compression data relate to final quality.
2. Materials and methods
2.1. Materials
Two different grades of hydroxypropyl methylcellulose (HPMC)
were used as matrix formers: standard wet granulation grade,
Methocel K100 Premium LV CR (CR, Dow Chemical Company,
Midland, Michigan, USA) and a new direct compressible grade,
Methocel K100 Premium LV DC2 (DC2, Dow Europe GmbH,
Bomlitz, Germany). The model active substance was ibuprofen
(Zhengzhou Debao Fine Chemical Co. Ltd., Henan, China). Two
different ibuprofen particles sizes (IbuPS) were used: (1) as
received from the supplier (large IbuPS) and (2) small IbuPS
prepared by ball milling (Retch S1, Haan, Germany). Mannitol
(Parteck M200, Merck KGaA, Darmstadt, Germany) was added as a
soluble filler and sodium stearyl fumarate (PRUV, Moehs,
Barcelona, Spain) as a lubricant.
2.2. Design of experiment
The study was performed as an experimental design with four
factors: HPMC particle size (HPMC PS), ibuprofen particle size
(IbuPS), ibuprofen load (Ibu%) and mixer speed. The design was a
full factorial with 19 runs. However, poor flow of the powder
blends were observed when low IbuPS and high Ibu% were
combined and the associated manufacturing difficulties led to the
exclusion of the factor combination. Thus a d-optimal design was
adopted; the process parameters are listed in Table 1. MODDE 10
(Umetrics MKS AB, Umeå, Sweden) was used both to construct the
experimental design and for evaluation of data. All models were
fitted with multiple linear regression (MLR).
For the center point of the design (N17, N18 and N19), a
1:1 mixture of the two HPMC grades was prepared. In addition, a
292 T. Ervasti et al. / International Journal of Pharmaceutics 495 (2015) 290–301

Table 1 concentration was changed within three levels (2%, 15% or 22%
D-optimal experimental design worksheet.
w/w) as well as the concentration of mannitol (44%, 51% or 64%, w/
Run Speed (rpm) HPMC PS (mm) Ibu PS (mm) Ibu (%) w). The total feed rate of powders was kept constant at 3.5 kg/h.
N1 300 77 30 2
N2 1200 77 30 2 2.3. Powder characterization
N3 300 120 30 2
N4 1200 120 30 2
The particle morphology of HPMC and ibuprofen particles were
N5 300 77 66 2
N6 1200 77 66 2
assessed by scanning electron microscopy (SEM, Quanta 200, FEI
N7 300 120 66 2 Company, Eindhoven, the Netherlands). Samples were mounted on
N8 1200 120 66 2 coal tape on stubs and sputter coated with gold using a Cressington
N9 300 77 66 22 108 auto sputter coater (Cressington Scientific Instruments Ltd.,
N10 1200 77 66 22
Watford, UK).
N11 300 120 66 22
N12 1200 120 66 22 The particle size distribution for the characterized materials
N13 300 77 30 15 was determined using a Malvern Mastersizer 2000 laser diffraction
N14 1200 77 49 22 spectrometer (Malvern Instruments Ltd., Malvern, UK) with an
N15 300 120 49 22
attached Scirocco 2000 dispersion unit (Malvern Instruments Ltd.).
N16 1200 120 30 15
N17 750 100 49 15
The test was performed in duplicate. The 10th, 50th and 90th
N18 750 100 49 15 percentile of the particle size distribution was reported, which
N19 750 100 49 15 means that 10% (or 50% or 90%) of the particles have a diameter
according to the d10 (or d50/d90) value or smaller. The air pressure
setting used was 0.1 bar, the feeder rate 50% and the obscuration
1:1 mixture of the two particle sizes of ibuprofen were used for the value was 0.4–2.6.
experiments (N14, N15, N17, N18 and N19). The compressibility of powder blends were measured using a
In all experiments, the concentrations of HPMC and PRUV were powder rheometer (FT4, Freeman Technology, Tewkesbury, UK).
constant (32% and 2% w/w, respectively), the ibuprofen Powder samples of 10–15 ml were introduced in a glass vessel with

Fig. 1. The direct compression continuous manufacturing set-up used in the experiments. (A) and (B) The whole continuous manufacturing line. (C) A timeline describing the
process phases. (D) Modulomix.
 T. Ervasti et al. / International Journal of Pharmaceutics 495 (2015) 290–301
a splitting device. Pre-conditioning was done with a rotating blade
in order to have a standardized initial packing condition for all
samples. After pre-conditioning the surplus powder was removed
by splitting the glass vessel and the blade exchanged for a vented
piston. A normal stress of 1–15 kPa was applied on the powder bed
using a vented piston while the change in volume was measured.
The measurement was made in triplicate.
2.4. Process description
The direct compression continuous manufacturing set-up used
in the study is described in Fig. 1. It depicts that the feeding, mixing,
and compressing parts are all integrated into a complete
continuous tablet manufacturing line. Two kinds of LIW feeders
(K-Tron, Types K-ML-D5-KT20 and K-CL-SFS-KT20, Niederlenz,
Switzerland) were used to feed the raw materials into the
continuous mixer (Modulomix, Hosokawa Micron, Doetinchem,
The Netherlands). After mixing, the powder mixture was guided
(flow controlled by gravity) into the hopper of the tablet press
(PTK-PR1000, PTK CO., Ltd, Incheon, Korea). This very simple set-up
enabled conducting experiments without any conveyors, thus the
risk of segregation was minimized. The feeders and the mixer were
adjusted and controlled by an in-house software, and data from the
feeders were acquired for later analysis.
In every run, the set points of lubricant and HPMC feeders were
kept constant at 70 g/h (2% w/w) and 1120 g/h (32% w/w),
respectively. The set point of mannitol feeder had three different
levels: 1540 g/h, 1785 g/h and 2240 g/h. Also for ibuprofen feeder,
three different set point values were used: 70 g/h, 525 g/h and
770 g/h. With these set point values, the desired ibuprofen content,
2, 15 and 22% (w/w), were reached in the final compositions.
The mass flow of each excipients and ibuprofen was acquired
from the feeders twice a second. To evaluate the feeding accuracy,
an estimate for the ibuprofen concentration of the total mass flow
fed into the mixer was defined. The estimated ibuprofen
concentration (ECIBU) was calculated with the following equation:
MFIBU ðtÞ
ECIBU ðtÞ ¼ ; ð1Þ
MFIBU ðtÞ þ MFHPMC ðtÞ þ MFMANN ðtÞ þ MFPRUV ðtÞ
where MF describes the mass flow of material x at time point t. The
ECIBU was compared to label claim (LC) (2%, 15% or 22% w/w) to
estimate how accurately the desired ibuprofen level was main-
tained during feeding.
The continuous dry powder mixer consists of a horizontal
cylindrical process chamber in which a horizontal agitator with
paddles is rotating (Fig. 1D). The paddles are fitted on a shaft and
are rotating at a fixed distance from the wall of the process
chamber. The speed of the rotor can be adjusted. The powder is
moved forward and backward in the chamber by paddles to
achieve a more efficient mixing process. Although high shear
forces can be applied to the mixture, attrition effects by impact
forces will be limited. The mixer has two powder inlets and one
outlet. In this study, the first inlet was used for the excipients and
ibuprofen while the second inlet was used to feed the lubricant. In
the current study, trials were performed with a mixer speeds of
300 rpm and 1200 rpm. In the center point of the design 750 rpm
was used. The mixer speed can be represented by the Froude
number (Fr) which describes the ratio of gravity and centrifugal
forces. Fr is calculated using Eq. (2).
Rv2
Fr ¼ ; ð2Þ
g Transmission Raman measurements were performed on three
where R describes radius of the impeller (m), v is angular velocity
of the impeller (rad/s) and g is gravity (m/s2). By calculating Fr
mixers can be classified into three categories according to their
293
mixing methods: Frs less than 1 indicate that gravitational forces
are predominant (dense phase flow, thrust mixers, free-fall
mixers); Frs slightly above 1 indicate fluidized motion of material
in the mixer (fluid bed mixers); and in the mixers with Frs
considerably above 1 centrifugal forces are predominant (centrif-
ugal and intensive mixers) (Gericke, 1993; Marikh et al., 2005;
Marikh et al., 2006; Berthiaux et al., 2008). With the mixer used in
this study the mixer speed of 300 rpm corresponds to a Fr of 6.4,
and 1200 rpm to a Fr of 103.
The tablets were compacted with a turret speed of 48 rpm using
eight 7 mm concave punches to produce tablets with a target
weight of 150 mg. The force feeder of the tablet press fed the
powder into the dies with a rotating speed of 96 rpm. The tablet
press control cabinet was used to store the compression data.
During the compaction process, the average compression force (CF)
of all of the punches for each rotation was recorded. The
compression force was measured from the lower main compres-
sion roll. The tablet press settings (compression force and punch
displacement) were adjusted with the first experiment (N9).
Thereafter the settings were kept at the same levels for all of runs.
With these settings a compression force between 5 and 9 kN was
obtained. The compaction data was analyzed using the Matlab
R2012a software (MathWorks, Natick, MA, USA).
The timeline of the process is described in Fig. 1C. To start
processing, the feeders and the mixer were first turned on (time
point 0) and the powder mixture was collected into the hopper of
the tablet press for 12 min (apart fron run N9 where 10 min was
used) to serve adequate fill level for the press. The complete
continuous processing was started at 12 min when tablet press was
turned on. The duration of each run was totally 32 min (12 min
hopper filling + 20 min continuous direct compression).
2.5. Tablet analysis
The weight and tensile strength of the tablets were measured
(n = 10) with a MultiCheck Turbo III (Erweka GmbH, Heusenstamm,
Germany). In addition, tablet weight variations were calculated.
Dissolution testing of drug release from tablets (n = 3–6) was
carried out in a USP dissolution apparatus II (Hanson research SR8-
Plus, Chatsworth, CA, USA) using ultraviolet (UV) detection with
fibre optics (Varian Cary 50 Bio, C Technologies Inc., Bridgewater,
NJ, USA). A fibre-optic probe was inserted directly into each vessel
and the UV absorption measured in situ. The USP II method was
modified by incorporating a quadrangular stationary basket above
the paddles, where tablets were placed. Dissolution testing was
done at 37  C in 1000 ml phosphate buffer pH 7.4 (I = 0.1) with a
paddle speed of 50 rpm. Dissolution profiles were adjusted to 0%
released at time point 0, noise-filtered (Savitzky–Golay, order
0 and window 5) and then normalized to 100% released based on
the level at the end of the dissolution. The number of individual
profiles were 1–3 for every batch for the 32 min process time point
(used in the DoE) and 3 for batches N5, N6, N11, N12, N14 and
N15 where dissolution was studied for several process time points.
The averages were used for the DoE. The time to reach 80% released
(T80) was linearly interpolated between data points encompassing
80% released.
2.6. Assay and content uniformity
Tablets from all batches were sampled every two minutes from
the start of the tablet press with a sampling time of 10–20 s.
tablets per sample with an Accura transmission Raman system
(Horiba Scientific, Lille, France) equipped with a 300 mW, 785 nm
diode laser, and an xy-stage for automatic sample presentation of
up to 32 tablets. The incident laser beam diameter was 4.3 mm,
294 T. Ervasti et al. / International Journal of Pharmaceutics 495 (2015) 290–301
spectral range captured 135–2240 cm1. All tablets were mea-
sured for 60 s, using 4 acquisitions, i.e., a total acquisition time of
4 min.
Since the Raman measurements are estimates of the concen-
tration of active in the tablet, the models were fitted using a
concentration measure (% w/w ibuprofen) as the response
obtained from the dissolution testing, eg the end of the dissolution
when all material was dissolved. An orthogonal projections to
latent structures (OPLS) model was fitted in Simca 13.0.2 (Umetrics
MKS AB, Umeå, Sweden) with the acquired Raman spectra (n = 744,
range 400–1700 cm1), pre-treated with 1st derivatives window
91 cubic fitting followed by standard normal variate (SNV) and
centering, as the X matrix and the nominal content (2, 15 or 22% w/
w) as the Y vector. The first PLS component gave a model with an R2
of 0.99. Based on this model, the tablets at each nominal level that
spanned from the lowest to the highest predicted were selected for
dissolution analysis. The value of 100% released ibuprofen was
used instead of a regular content reference method. In this way, a
large variation of content could be fed into the final models for each
nominal level where reference data from the dissolution analysis
would be used as the Y vector. The dissolution data (% released
ibuprofen) was converted into concentration (% w/w ibuprofen) by
multiplying with the nominal content in mg (3, 22.5 and 33) and
dividing by the individual tablet weight. In addition, reference
measurements on tablets used for the regular dissolution testing
were also included in the Y vector.
The final calibration models for each ibuprofen level were as
follows: (1) 2% level: 1 + 3 PLS components (i.e., one predictive and
three orthogonal components) with a root-mean-square error of
estimation (RMSEE) of 0.11% ibuprofen (n = 28), (2) 15% level: 1 + 2
components with an RMSEE of 0.21% ibuprofen (n = 25) and (3) 22%
Fig. 2. SEM images of (A) HPMC CR, standard wet granulation grade. (B) HPMC DC2, direct compression grade. (C) Original ibuprofen.(D) Milled ibuprofen.
level: 1 + 3 components with an RMSEE of 0.79% ibuprofen. R2 for
these three models were respectively 0.94, 0.98 and 0.91. The
acquired transmission Raman spectra were then predicted using
these models. The predicted content in% w/w ibuprofen was
converted into% label claim (LC) by multiplying with the individual
tablet weight and dividing by the nominal content in mg. Assay
was calculated for each batch and process time point (every two
min) as the average of the three predicted tablet contents. Content
uniformity was estimated by pooling tablets from the four last time
points (n = 12) and then calculating the relative standard deviation
(RSD).
3. Results and discussion
3.1. Scanning electron microscopy of raw materials
Scanning electron microscope images of different grades of
HPMC and ibuprofen are represented in Fig. 2. Between the HPMC
grades not only the particle size but also particle shape differed
(Fig. 2A and B). The standard CR grade was a mixture of fibrous and
irregularly shaped flat particles while the direct compressible
grade (DC2) particles were agglomerated and rounder suggesting
better flow properties for the latter. Fig. 2C and D revealed that
milling of ibuprofen had clearly increased the fraction of fine
particle sized material.
3.2. Particle size distribution of raw materials
The results from particle size measurements were consistent
with observations from SEM images. The particle size of CR was
smaller than that of DC2, and the milled ibuprofen had smaller
 T. Ervasti et al. / International Journal of Pharmaceutics 495 (2015) 290–301 295

particle size than the original material (Table 2). Mixtures of the
different grades had an intermediate particle size.

3.3. Feeder data

Data from previous experiments suggests that the feeders need

to run for some time before reaching steady state (results not
shown). Feed rates and powder flow properties are known to affect
this time but in most cases 10 min have been found sufficient to
reach the equilibrium. Also, in the current study the largest
variation in the ibuprofen concentration was noted during the first
10 min, in particular with low dose ibuprofen (2% w/w) feed rate
(Fig. 3). With high dose ibuprofen feeding (15 and 22% w/w) the
settling time for the feeders was in all cases less than 5 min, and the
feeding was overall very accurate (Fig. 3A). With low dose
ibuprofen feeding (2% w/w) the settling time was notably longer
(Fig. 3B). After 10 min the ibuprofen concentration was mainly
between 90 and 110% from LC, but for some runs it took about
17 min before the ibuprofen concentration was settled (N5).
The mean values of mass flow were very close to the set point
values for HPMC and mannitol feeders and the RSD value was
below 3% and 1%, respectively. Fig. 4 shows the mean values of
ibuprofen mass flow values (grey bars) and the calculated RSD
values (black bars) of the mass flow for each run. The mean values
were close the set point values for all runs but RSD values were
quite high (4–11%) when the mass flow was 70 g/h. The reason for
this was the fact that the standard deviation was independent of
the mass flow, i.e., it was at the same absolute level for the high and
low mass flow values (data not shown) and if the mean value
decreased then RSD increased. However, the desired ibuprofen Fig. 3. Time to reach the steady state of continuous process. Estimated
content in the tablets was reached even though it for some runs concentration (ECIBU) relative to label concentration (LCIBU) of the powder before
took a longer time to reach. The mean mass flow values of the PRUV mixing, calculated from the feeder data. (A) Experiments (N9–N19) with high Ibu%
(15 and 22%) and (B) Experiments (N1–N8) with low Ibu% (2%). Calculated time to
feeder gave similar results. RSD values varied between 2 and 13%
reach steady state marked with arrows and deviation (3SD) with grey dashed
but the mean values were very close to the set point value. In horizontal lines. Grey solid horizontal line describes 100%.
general, if low RSD values are essential then the feeder mass flow
must be increased until the desired level of RSD is reached. that the flowability of most of the powder blends was better than
that of Avicel PH-102.
3.4. Flowability of powder blends
3.5. Compression data
The powder mixtures N3, N4, N7 and N8 had the lowest
compressibility values (9–10%), which meant that these powder The stability of compression force was evaluated visually by
mixtures had the best flowability (Fig. 5). All these powder plotting it as a function of process time (Fig. 6). After the tablet
mixtures contained DC2 and low ibu% (2% w/w). The powder press was started it took about 4 min before the compression force
mixtures N9, N10, N13 and N14 had the highest compressibility reached a stable level. A clear drop in the compression force data of
values (14–16%) and consequently the worst powder flow N9 was noted at about 22 min, which was due to lack of powder in
properties. These powder mixtures contained CR combined with the hopper of the tablet press. N9 was the first batch that was
an ibu% of 15–22% (w/w). Results from MLR modelling confirmed manufactured and the powder was collected into the hopper only
that HPMC PS was the most important factor for the powder flow for 10 min. In other runs the collection time was increased to
followed by Ibu%, IbuPS and mixer speed (Table 3). Coefficient plots 12 min. For this reason run N9 was excluded from further data
are provided in the Supplementary material (Fig. S1). evaluation.
Sun (2010) suggested that microcrystalline cellulose Avicel PH- The tablet press settings were determined using the first
102 could be used as a reference material when setting acceptance experiment N9 which had large ibuPS and HPMC PS and high ibu%.
limits for powder flow in tabletting. A powder with flow properties Since the bulk density, particle size and other powder properties
inferior to that of Avicel PH-102 likely exhibits flow problems were different for all of the experiments (except for centre points),
during high speed tabletting and should be avoided. The the compression force, weight and also the tensile strengths of the
compressibility for Avicel PH-102 was around 15%, which indicated tablets were at a different levels for every experiment. This is why

Table 2
Particle size of the raw materials.

HPMC CR (mm) HPMC DC2 (mm) HPMC mix (mm) Ibu original (mm) Ibu milled (mm) Ibu mix (mm)
d10 26.8 50.1 36.2 15.9 7.6 11.2
d50 76.5 119.7 100.2 66.4 29.7 49.1
d90 198.5 285.0 243.3 247.1 85.4 194.5
296 T. Ervasti et al. / International Journal of Pharmaceutics 495 (2015) 290–301

Fig. 4. Sensitivity of feeder mass flow. The calculated mean values (grey bars) of mass flow from the feeders and their RSD values (black bars) for ibuprofen. Horizontal lines
describe set points for the feeders.

Fig. 5. Flowability of the powder blends measured as compressibility after continuous mixing. Horizontal line describes the compressibility of Avicel PH 102.

Table 3
Model summary for models at 32 min process time R2 is the explained variance and Q2 is the cross-validated explained variance.

Response R2|Q2 Significant factors in order of importancea

Compressibility 0.97|0.94 HPMC PS, Ibu%, IbuPS, IbuPS  Ibu%, Speed
Compaction force 0.91|0.79 Speed  Speedb ,HPMC PS  Ibu%, Speed, HPMC PS, Ibu%
Tablet weight (mean) 0.87|0.72 Speed  Speed, Ibu%, HPMC PS  Ibu%, Speed, HPMC PS
Tablet weight (sd) 0.84|0.67 Speed  Speed, HPMC PS, Speed, IbuPS, Speed  HPMC PS
Tensile strength 0.90|0.81 Ibu%, HPMC PS, HPMC PS  Ibu%, Speed
Assay No model
Dissolution, time to 80% 0.99|0.97 Ibu%  Ibu%, Ibu%, HPMC PS, HPMC PS  Ibu%, Speed
Content uniformity 0.81|0.70 Ibu%  Ibu%, Ibu%, IbuPS
a
Positive factors in bold.
b
Due to the nature of the model, the square term Speed  Speed is identical to HPMC PS  HPMC PS.
 T. Ervasti et al. / International Journal of Pharmaceutics 495 (2015) 290–301
Fig. 6. Robustness of continuous tablet manufacturing in terms of mean compression forces as a function of time. (A) and B) represent small Ibu PS. (C) and (D) large Ibu PS. (A)
and (C) HPMC CR. (B) and (D) HPMC DC2. Ibuprofen content is 2% (w/w) for all of the experiments. Black lines correspond with a mixing speed of 1200 rpm and grey lines
300 rpm.
the actual values cannot be compared with each other between the
different experiments.
The compression force data was quite smooth and steady when
tableting DC2 (Fig. 6B and D). This was mostly due to good
flowability of the powder and thus a more efficient fill of the die. In
addition, compression force was approximately 2 kN lower when
compressing DC2 with small ibu% (2%) (N3, N4, N7 and N8) than
with CR grade. These experiments had also the best flowability
(Fig. 5). Furthermore, when compressing DC2 there was more
empty space (air) between the particles in the die compared to CR.
It is known that smaller spherical particles can fill the restricted
space more efficiently (Virtanen et al., 2010). However, the similar
behavior might have been seen also here with HPMC particles even
though they are not spherical. In any case higher amount of
material reached the die when using CR grade which was also
observed in compression force, weights and tensile strengths of
the tablets.
IbuPS did not seem to have a significant effect on the
compression force (Fig. 6). It was also noted that when ibu%
was higher HPMC PS had no clear effect on the compression force.
It seemed that the higher ibu% diminished the effect of HPMC PS.
HPMC grades had different compaction properties (CR had better
compactibility (Rogers et al., 2013)), while the deformation
properties for all of ibuprofens were the same since the small
particle size was produced from the large particle size by milling.
Particle size can have an effect on deformation properties but with
particle size ranges used in this study it most probably did not have
significant effect.
When CR grade was used (N1, N2, N5, N6, N13 and N14) the
compression force decreased as a function of time. This was
297
more pronounced for experiments with low mixer speed. For
experiment N2 this decrease was quite subtle. A corresponding
decrease was also observed in the tablet weight and tensile
strength values discussed later. This decrease might be related
to compaction in the hopper or in the force feeder. In general,
when using faster mixer speed the compression data was
smoother (Fig. 6A and C) due to more homogeneous blend and
therefore more steady fill of the die during the tableting
process.
According to the MLR model, either a square term for HPMC PS
(HPMCpsHPMCps) or speed (speed  speed) was the most
significant parameter influencing the compression (Table 3).
Coefficient plots are provided in the Supplementary material
(Fig. S1). To be able to determine the correct square term, more
experiments would be needed. As can be seen from Fig. 6 both
factors (mixer speed and HPMC PS) had a significant role when
assessing compression data. In addition, ibu% was a significant
factor in the model. As stated before, experiments with low ibu%
had better flowability which most probably played a role in
tableting as well. Furthermore, experiments that had poor
flowability (N10, N13 and N14) had more fluctuation in the
compression data. It is known that poor flowability affects the die
filling during the tableting process (Lakio et al., 2010). In addition,
different particle sizes and shapes as well as density of the
materials can affect the die filling and induce segregation. By
adjusting the process parameters, for instance the mixer speed, the
fluctuations in the compression force profile can be minimized
leading better product quality. In more general terms, by adjusting
the process parameters to fit the material properties the product
quality can be improved.
298 T. Ervasti et al. / International Journal of Pharmaceutics 495 (2015) 290–301
3.6. Tablet analysis
3.6.1. Tablet weight and weight variation
In general, the tablet weight was stable for all batches within a
timeframe of 4–5 min after the start of the tableting. When the
tablet weight data was analyzed, it was observed to be divided into
two groups on the basis of HPMmiC PS. The weight of tablets
prepared using CR grade was higher but the weight decreased as
the process proceeded. Rogers et al. (2013) found that tablet
weigths were higher when DC grade was used. However they were
manufacturing 1000 mg tablets, so the die filling phenomenon was
different. The die in this study was much smaller since the tablet
weigth was only 150 mg. With DC2 the tablet weight was lower but
the stability was overall better (Fig. 7A). Higher mixer speed also
resulted in higher tablet weights, which could also be noted when
looking at model parameters obtained with MLR modelling
(Table 3 and Table S1), e.g., speed  speed and HPMCpsIbu%
were the most significanct factors. This was an expected outcome,
because higher mixer speed improved flow properties, thus die
filling was more efficient.
The tablet weight variation was generally noted to be rather low
within each run (Fig. 7A). The difference between minimum and
maximum weights was always less than 7.3 mg (mean 3 mg)
during steady state (except for N9 where the fill level of the hopper
was too low as aforementioned). However, the weight variation
was found to be slightly lower when DC2 was used as reported also
by Rogers et al. (2013). This can be explained by the better flow
properties of the DC2 grade. This observation was strengthened by
the MLR model where HPMC PS was the most important parameter
for both compressibility and tablet weight variation (Table 3 and
Table S1).
3.6.2. Tensile strength of tablets
The tensile strength of the tablets showed differences between
different grades of HPMC (Fig. 7B). In accordance with previous
work the tensile strength of tablets manufactured with CR grade
was higher (Rogers et al., 2013). In addition, the variation over time
was notable (decrease of 0.4–0.6 MPa) when using CR grade. This
decrease was also seen in the compression forces (Fig. 6). A
combination of low Ibu% and CR grade gave the highest tensile
strength tablets. With DC2, the tablet tensile strengths were
overall lower and showed no significant variation over time. The
MLR model also predicted HPMC PS to be the most important
factor (smaller particle size leading to higher tensile strength),
which supported these findings. However, it is good to realize the
small changes seen in the tensile strength data might not be
relevant for the overall tablet performance, since the tensile
strength was sufficiently good for all runs. In addition, due to tablet
Fig. 7. Robustness of continuous tablet manufacturing in terms of (A) the average weight of tablets (different grades of HPMC and low and high Ibu%), (B) the tensile strength
of the tablets (HPMC CR and HPMC DC2 represent the mean value for all experiments with HPMC CR and HPMC DC2, respectively), and C. ibuprofen assay (Ibu PS 30 and Ibu PS
66 represent average calculated from all experiment with ibuprofen particle size of 30 mm and, and Ibu PS 66 of 66 mm (except N5 and N7), respectively. Circles describe the
highest and the lowest average values.
press setting every experiment had different level for the tensile
strength.
3.6.3. Assay and content uniformity
For most of the runs the assay stabilized relatively fast to an
acceptable level. There was only a very low variation in the assay
between the different batches and the results could not be
modeled. This means that the assay was robust no matter which
material properties and process parameters were used in the
experiments.
Interestingly, in the experiments with small IbuPS, assay results
were below the label claim throughout the whole processing cycle
(Fig. 7C). When the feeder data was analyzed more closely, it was
noticed that in every experiment done with small IbuPS there was a
transient decrease below the set point in the feed rate of ibuprofen
right after the process was started. From previous experience this
can be stated to be normal for the feeders used, e.g., that the
feedrate under- or overshoots during the first few minutes of the
process before reaching the proper speed of the feed screw.
However, it was surprising that with small IbuPS there was always
undershooting.
The assay appeared to be more stable when large IbuPS was
used unless the Ibu% was low. This could to some extent be related
back to feeder data, which indicated a slight overshooting of
ibuprofen feed rate in some but not all experiments. With low Ibu%
the assay was harder to control and the effect of particle size on
feed rate was found to be more pronounced. The dotted line in
Fig. 7C is the mean between experiments N5 and N7 which both
had large IbuPS, low Ibu% and were mixed at low mixer speed
(300 rpm). The highest level reached was 140% which for a 150 mg
tablet with Ibu% of 2% would equal to 4.2 mg ibuprofen instead of
the desired 3 mg. The high Ibu% obtained for these runs (N5 and
N7) could not be explained by the feeder data.
Fig. 7C emphasizes the importance of letting the feeding system
reach steady state before collecting powder for compression. In
these experimental runs the powder was collected already from
the starting of the line, before the feeders had reached stable feed
rates. Thus the powder fed into the hopper of the press during the
first ten minutes had varying Ibu%. When the tablet press was
started, the rotating discs in the force feeder still mixed the powder
and smoothened out the content variation, which caused
spreading of the incorrect content over larger powder volume
and thus increased the time to reach right Ibu% into tablets.
3.6.4. Dissolution
A MLR model was constructed based on the time to reach
80 percent release for tablets taken out 20 min after start of
compaction (32 min total processing time). The Ibu%, HPMC PS and
 T. Ervasti et al. / International Journal of Pharmaceutics 495 (2015) 290–301
mixer speed were the factors that had the largest effect on T80
(Table 3 and Table S1). The slower release observed at higher drug
load probably related to differences in matrix hydration and erosion
as a result of the larger relative amount of poorly soluble components
in the matrix (Tajarobi et al., 2009a). The tendency for faster release
from tablets containing DC2 was most probably related to the
inability of some formulations to form a homogeneous gel layer at
the tablet/dissolution media interface. Similar observations have
been made previously for DC grades at compositions when the
amount of HPMC was close to the percolation threshold (Heiman
et al., 2015). In the current study the amount of HPMC was set to 32%
w/w in all formulations. The review of Carabello (2010) nicely
summarize factors affecting drug release from HPMC matrices in the
light of the percolation theory. A general observation made in the
review is that higher polymer percolation thresholds will be
obtained for higher relative particle sizes of the polymer, owing to
the lower ability of the coarser particles to percolate the matrix
(Mohamed et al., 2015). The exact position of the polymer
percolation threshold also depends on the components in the
formulation (fillers, drug substance etc.). For HPMC CR model
formulations, using the same grade as in our study, the percolation
threshold has been identified to be in a concentration range between
30 and 35% (w/w) for binary HPMC based compositions with an
easily soluble additive (mannitol) (Tajarobi et al., 2009b).
The effect of processing time on the release behavior was
investigated in more detail for some of the experiments processed
at low (N5, N11, N15) and high mixer speed (N6, N12, N14). In Fig. 8
the release rate (T80) of ibuprofen is presented at different time
points of manufacturing. Generally, the processing time was found
to have an effect on the release rate. The tablets collected at early
processing times released ibuprofen significantly faster than
tablets collected at the end of the process. Notably this effect
was more pronounced when DC2 was used, e.g., it took longer time
to reach steady state and a robust drug release behavior. By
increasing the mixer speed the time to steady state was shortened.
The dissolution curves for tablets that gave robust and slow release
had a shape indicating first order release characteristics whereas
tablets that were less robust were characterized by burst and
release profiles with a shape indicating more influence of zero
order, erosional, release. These results were in some sense opposite
to the results obtained for tablet weight, assay and tensile strength
where the most robust condition and the shortest time to
equilibrium was obtained using DC2. This result highlights the
importance of balancing material and processing aspects and the
need to include several quality attributes in the evaluation.
It should be pointed out that apart from N5 and N6 the assays
never deviated more than 10% from the target. For N5 a clear
Fig. 8. Effect of raw materials and mixer speed on the robustness of continuous tablet manufacturing in terms of in vitro dissolution. T80 release (circles represent minimum
and maximum values) with a mixing speed of (A) 1200 rpm and (B) 300 rpm. Solid lines represent HPMC DC2, dashed lines HPMC CR, black lines large particle size of Ibu and
grey lines intermediate particle size of Ibu.
299
change (decrease) of assay over time was observed. Notably though
N5 and N6 were the formulations where T80 varied the least over
time. Hence the dissolution variation over time was most probably
not linked to variation in assay. Also, looking back at the HPMC
feeder data the conclusion was that there was no variabilities that
could explain the observed behaviour (feeder RSD <3%).
Clearly in the beginning of compaction process the capacity of
the tablet to control the release was poor. It should be noted that
the tablets that failed to provide sustained release did not break
under the dissolution process, but rather disintegrated. Thus,
DC2 in particular were not capable to form a strong homogeneous
gel layer at the tablet/dissolution media interface. This was most
probably not an effect of the different porosity or density of the
tablets since these factors generally have been found to have very
low effect on the release from HPMC based matrices, especially at
the relatively low drug loadings of the present study (Timmins
et al., 1992; Velasco et al., 1999). Also the fact that the tensile
strength for compositions containing DC2 was rather constant
with time support this statement (Fig. 7B).
A homogeneous distribution of the matrix former (HPMC) is
required to provide a robust release performance. Although the
bulk densities of HPMC CR and DC2 does not differe significantly
(Rogers et al., 2013) our data clearly indicate that when the HPMC
PS was small (HPMC CR) and the mixer speed was high this target
was easier to reach. Therefore, when using HPMC DC2 to gain the
advantage of improved powder flow properties and more robust
processing conditions, it is important to assure that the time
allowed to reach steady state, also with regards to release
robustness, is sufficiently long and/or that the mixer speed is
sufficiently high. In order to more rapidly provide a robust release
from tablets containing DC2, a higher content of matrix former
may be needed. Also when processing extended release matrix
tablets in a continuous mixing and direct compression set-up it is
essential to control the matrix former distribution during
manufacturing by suitable in-line tools.
3.7. Key aspects for continuous manufacturing (of ER matrix tablets)
When manufacturing ER matrix tablets via continuous mixing
and compression it is important to let the feeders reach steady
state before collecting powder for compression. In the current
study the largest variability in the Ibu% was noted during the first
10 min. The standard deviation was independent of the mass flow,
which meant that when using low mass flow of ibuprofen (70 g/h)
the RSD was quite high (4–11%). The findings were in accordance
with Pernenkil and Cooney (2006) who stated that a continuous
mixer can only reduce the short-term fluctuations introduced by
300 T. Ervasti et al. / International Journal of Pharmaceutics 495 (2015) 290–301
Fig. 9. Example of continuous manufacturing adaptive control based on the experimental data. (A) A coefficient plot based on MLR model for tensile strength of tablets. (B)
The effect of changing the variable (HPMC particle size) [time point 1] during continuous manufacturing and how adjusting another process parameter (mixer speed) [time
point 2] will reach the desirable tensile strength again. See more accurate description of the process in the text.
the feeding mechanism. Hence, the design of the feeding system
takes a critical role in design of continuous mixers. The demand
posed on the mixer performance can be greatly reduced by
designing a feeding system that is consistent and accurate.
The continuous mixer (Modulomix) was tested with a
challenging design where the mixer speed was varied in a large
range (300–1200 rpm) and also the inflow rate of powders was
very low (3.5 kg/h). In our study the high mixer speed gave
generally better results than the low mixer speed. For example a
robust drug release was achieved faster with high mixer speed. Our
results are not in accordance with Williams and Rahman (1970),
Portillo et al. (2009) and Martínez et al. (2013) who achieved
improved content uniformity at lower rotation rates due to the
longer residence times. The difference between our results and
Martínez et al. (2013) might be due to the much higher feed rates
(75 kg/h) used in their study.
Flow properties of the raw materials had a big influence on
manufacturability/processability of the ER matrix tablets. The
flowability had an influence on the flowability of powders from
feeders and how the powder mixtures filled the tableting dies.
Overall performance of the experiments with good powder flow
properties was significantly better than experiments with poor
flowability.
The time to reach the steady state for tablet weights was
approximately 20 min for all of the batches. When considering
tensile strength the steady state was reached in a few minutes for
HPMC DC2, but it took approximately 20 min for HPMC CR. For
assay it took approximately 20 min to reach steady state when
using small particle size of ibuprofen but only few minutes when
using large particle size. In addition some batches (N5 and N7) it
took longer to reach the steady state, in fact they did not reach it
during the process time used in the study. Nevertheless, one have
to remember that the system was intentionally provoked to get
significant differences between the batches. In general it could be
stated that steady state can be reached in 20 min regardless of the
measured responses.
Continuous manufacturing enables use of different formulation
parameters (e.g., particle size) and process parameters (e.g., mixer
speed) and still achieve the same final outcome. For example, in
this study the assay results could not be modeled due to the very
low variation between the different batches. This meant that the
final outcome (assay) was robust no matter which material
properties and process parameters were used in the experiments.
In addition, the continuous process can be controlled in real-time
and the parameters can be changed if needed inside the pre-
determined design space. To demonstrate the possibility to control
the continuous manufacturing process, an example of adaptive
control is represented in Fig. 9. Based on the MLR model for tensile
strength of tablets, the following situation was simulated: a
continuous process is run at 300 rpm mixer speed with 15%
ibuprofen content and using HPMC which has d50 of 55 mm. If for
example the supplier of HPMC changes and d50 of HPMC increases
from 55 mm to 75 mm, the tensile strength of tablets will decrease
under lower warning limit (Fig. 9B time point 1). However by
adjusting the mixer speed, in this case increasing it to 1200 rpm
(Fig. 9B time point 2), the tensile strengths can be raised back to
desired level (approximately 1.8 MPa). With batch processing it is
more challenging to change the process in real-time and previously
described adaptive control is not possible. Thus continuous
manufacturing can save a significant amount of time and resources
and produce high quality outcome when the material properties
and process parameters have been balanced in a right manner.
The continuous process can be affected positively by changing
the process parameters even though the material properties are
challenging. For example poor flowability, due to small particle
size, can be overcome by increasing the mixer speed.
4. Conclusions
Continuous mixing followed by direct compression can be
utilized to manufacture extended released hydrophilic matrix
tablets. The most robust tablet quality (weight, assay and tensile
strength) was obtained when using high mixer speed and large
particle size for both HPMC and ibuprofen regardless of the
ibuprofen load. This was due to good flowability of the powders
and efficient mixing. With low dose compositions at low mixer
speed it was more difficult to achieve high quality tablets due to
poor flowability and insufficient mixing of the powders. HPMC
 T. Ervasti et al. / International Journal of Pharmaceutics 495 (2015) 290–301
with large particle size (DC2) and processing conditions had a
significant effect on release. Longer processing time and/or faster
mixer speed was needed in order to achieve robust release with
compositions containing DC2. On the other hand, when using the
smaller HPMC PS (CR), the release of ibuprofen was more
controlled. In addition, the present work show that the process
data (flow from the feeders and compression force) can be used to
evaluate the process and to better understand the critical
attributes that have an effect on process quality. This thorough
investigation revealed the importance of balancing the process
parameters and the material properties and to evaluate the
potential for adaptive control of the whole continuous process.
Acknowledgements
TE, SPS and JK gratefully acknowledge the financial support as
well as the provision of equipment by the PROMIS Centre consortium
(supported by the Finnish Funding Agency for Technology and
Innovation (Tekes), the Regional Council of Pohjois-Savo, the North
Savo Centre for Economic Development, Transportation and
Environment, and participating industrial partners). Lars Johnson
(AZ) is acknowledged for taking SEM images. Johan Breitholtz and
Fredrik Winge are acknowledged for their assistance in setting up the
dissolution method. Peter van der Wel and Bert Dekens (Hosokawa
Micron B.V.) are acknowledged for providing the drawing and
information about the continuous Modulomix mixer.
Appendix A. Supplementary data
Supplementary data associated with this article can be found, in
the online version, at http://dx.doi.org/10.1016/j.
ijpharm.2015.08.077.
References
Allison, G., Cain, Y.T., Cooney, C., Garcia, T., Gooen, T., Holte, O., Jagota, N., Komas, B.,
Korakianiti, E., Kourti, D., Madurawe, R., Morefield, E., Montgomery, F., Nasr, M.,
Randolph, W., Robert, J.-L., Rudd, D., Zezza, D., 2015. Regulatory and quality
considerations for continuous manufacturing, May 20–21, 2014 continuous
manufacturing symposium. J. Pharm. Sci. 104, 803–812.
Berthiaux, H., Marikh, K., Gatumel, C., 2008. Continuous mixing of powder mixtures
with pharmaceutical process constraints. Chem. Eng. Process 47, 2315–2322.
Boukouvala, F., Niotis, V., Ramachandran, R., Muzzio, F.J., Ierapetritou, M.G., 2012. An
integrated approach for dynamic flowsheet modeling and sensitivity analysis of
a continuous tablet manufacturing process. Comput. Chem. Eng. 42, 30–47.
Bridgewater, J., 2012. Mixing of powders and granular materials by mechanical
means—a perspective. Particuology 10, 397–427.
Buchholz, S., 2010. Future manufacturing approaches in the chemical and
pharmaceutical industry. Chem. Eng. Process 49, 993–995.
Carabello, I., 2010. Factors affecting drug release from hydroxypropyl
methylcellulose matrix systems in the light of classical and percolation theories.
Cell 7 (11), 1291–1301.
Engisch, W.E., Muzzio, F.J., 2012. Method for characterization of loss-in-weight
feeder equipment. Powder Technol. 228, 395–403.
Gao, Y., Vanarase, A., Muzzio, F., Ierapetritou, M., 2011. Characterizing continuous
powder mixing using residence time distribution. Chem. Eng. Sci. 66, 417–425.
Gericke, H., 1993. Different methods of batch and continuous mixing of solids. Bulk
Solids Handl. 13, 2–11.
Heiman, J., Tajarobi, F., Gururajan, B., Juppo, A., Abrahmsén-Alami, S., 2015. Roller
compaction of hydrophilic extended release tablets- Combined effects of
processing variables and drug/ matrix former particle size. AAPS Pharm. Sci.
Technol. 16 (2), 267–277.
Hou, Q.F., Dong, K.J., Yu, A.B., 2014. DEM study of the flow of cohesive particles in a
screw feeder. Powder Technol. 256, 529–539.
International Symposium on Continuous Manufacturing of Pharmaceuticals, White
Papers, May 20-21, 2014. MIT US. https://iscmp.mit.edu/white-papers.
Lakio, S., Siiriä, S., Räikkönen, H., Airaksinen, S., Närvänen, T., Antikainen, O.,
Yliruusi, J., 2010. New insights into segregation during tableting. Int. J. Pharm.
397, 19–26.
Lee, S.L., O’Connor, T.F., Yang, X., Cruz, C.N., Chatterjee, S., Madurawe, R.D., Moore, C.
M.V., Yu, L.X., Woodcock, J., 2015. Modernizing pharmaceutical manufacturing:
from batch to continuous production. J. Pharm. Innovations (March), 1–9.
301
Leuenberger, H., 2001a. New trends in the production of pharmaceutical granules:
the classical batch concept and the problem of scale-up. Eur. J. Pharm.
Biopharm. 52, 279–288.
Leuenberger, H., 2001b. New trends in the production of pharmaceutical granules:
batch versus continuous processing. Eur. J. Pharm. Biopharm. 52, 289–296.
Marikh, K., Berthiaux, H., Mizonov, V., Barantseva, E., 2005. Experimental study of
the stirring conditions taking place in a pilot plant continuous mixer of
particulate solids. Powder Technol. 157, 138–143.
Marikh, K., Berthiaux, H., Mizonov, V., Barantseva, E., Ponomarev, D., 2006. Flow
analysis and Markov chain modelling to quantify the agitation effect in a
continuous powder mixer. Chem. Eng. Res. Des. 84, 1059–1074.
Martínez, L., Peinado, A., Liesum, L., Betz, G., 2013. Use of near-infrared spectroscopy
to quantify drug content on a continuous blending process: influence of mass
flow and rotation speed variations. Eur. J. Pharm. Biopharm. 84, 606–615.
Mascia, S., Heider, P.L., Zhang, H., Lakerveld, R., Benyahia, B., Barton, P.I., Trout, B.L.,
2013. End-to-end continuous manufacturing of pharmaceuticals: integrated
synthesis, purification, and final dosage formation. Angew. Chem. Int. Ed. 52,
12359–12363.
McKenzie, P., Kiang, S., Tom, J., Rubin, A.E., Futran, M., 2006. Can pharmaceutical
process development become high tech? AIChE J. 52, 3990–3994.
Mohamed, F.A.A., Roberts, M., Seton, L., Ford, J.L., Levina, M., Rajabi-Siahboomi, A.R.,
2013. Production of extended release mini-tablets using directly compressible
grades of HPMC. Drug Dev. Ind. Pharm. 39, 1690–1697.
Mohamed, F.A.A., Roberts, M., Seton, L., Ford, L.J., Levina, M., Rajabi-Siahboomi, A.R.,
2015. The effect of HPMC particle size on the drug release rate and the
percolation threshold in extended release tablets. Drug Dev. Ind. Pharm. 41, 70–
78.
Nicholson, P.I.I.W., 2014. Making do in making drugs: innovation policy and
pharmaceutical manufacturing. B.C.L Rev. 55, 491–562. http://dx.doi.org/
10.2139/ssrn.2311682.
Pernenkil, L., Cooney, C.L., 2006. A review on the continuous blending of powders.
Chem. Eng. Sci. 61, 720–742.
Plumb, K., 2005. Continuous processing in the pharmaceutical industry. Changing
the mind set. Chem. Eng. Res. Des. 83, 730–738.
Poechlauer, P., Manley, J., Broxterman, R., Gregertsen, B., Ridemark, M., 2012.
Continuous processing in the manufacture of active pharmaceutical ingredients
and finished dosage forms: an industry perspective. Org. Process Res. Dev. 16,
1586–1590.
Portillo, P.M., Ierapetritou, M.G., Muzzio, F.J., 2008. Characterization of continuous
convective powder mixing process. Powder Technol. 182, 368–378.
Portillo, P.M., Ierapetritou, M.G., Muzzio, F.J., 2009. Effects of rotation rate, mixing
angle, and cohesion in two continuous powder mixers—a statistical approach.
Powder Technol. 194, 217–227.
Portillo, P.M., Vanarase, A.U., Ingram, A., Seville, J.K., Ierapetritou, M.G., Muzzio, F.J.,
2010. Investigation of the effect of impeller rotation rate powder flow rate, and
cohesion on powder flow behavior in a continuous blender using PEPT. Chem.
Eng. Sci. 65, 5658–5668.
Rogers, T.L., Hewlett, K.O., Theuerkauf, J., Balwinski, K.M., 2013. Assessing how the
physical properties and enhanced powder flow of HPMC affect process control
during direct compression of matrix tablets. Tablets Capsules (October), 14–22.
Spencer, D.S., Gerogiorgis, D.I., Ramachandran, R., Evans, J.M.B., Barton, P.I., Trout, B.
L., 2011. Economic analysis of integrated continuous and batch pharmaceutical
manufacturing: a case study. Ind. Eng. Chem. Res. 50, 10083–10092.
Sun, C.C., 2010. Setting the bar for powder flow properties in successful high speed
tableting. Powder Technol. 201, 106–108.
Tajarobi, F., Abrahmsén-Alami, S., Carlsson, A., Larsson, A., 2009a. Simultaneous
probing of swelling, erosion and dissolution by NMR-microimaging—effect of
solubility of additives on HPMC matrix tablets. Eur. J. Pharm. Sci. 37 (2), 89–97.
Tajarobi, F., Abrahmsén-Alami, S., Hansen, M., Larsson, A., 2009b. The impact of dose
and solubility of additives on the release from HPMC matrix tablets—identifying
critical conditions. Pharm. Res. 26 (6), 1496–1504.
Timmins, P., Delargy, A.M., Minchom, C.M., Howard, J.R., 1992. Influence of some
process variables on product properties for a hydrophilic matrix controlled
release tablet. Eur. J. Pharm. Biopharm. 38, 113–118.
Vanarase, A.U., Muzzio, F.J., 2011. Effect of operating conditions and design
parameters in a continuous powder mixer. Powder Technol. 208, 26–36.
Vanarase, A.U., Osorio, J.G., Muzzio, F.J., 2013. Effects of powder flow and shear
environment on the performance of continuous mixing of pharmaceutical
powders. Powder Technol. 246, 63–72.
Velasco, M.V., Ford, J.L., Rowe, P., Rajabi-Siahhoomi, A.R., 1999. Influence of drug:
hydroxypropyl methyl cellulose ratio, drug and polymer particle size and
compression force on the release of diclofenac sodium from HPMC tablets. J.
Control. Release 57, 75–85.
Vervaet, C., Remon, J.P., 2005. Continuous granulation in the pharmaceutical
industry. Chem. Eng. Sci. 60, 3949–3957.
Virtanen, S., Antikainen, O., Räikkönen, H., Yliruusi, J., 2010. Granule size
distribution of tablets. J. Pharm. Sci. 99, 2061–2069.
Williams, J.C., Rahman, M.A., 1970. The continuous mixing of particulate solids. J.
Soc. Cosmet. Chem. 21, 3–36.