Continuous Direct Compression of a High-dose Drug

Evaluation of Process and Quality Attributes

Dr. Robin Meier1, B.Sc. Andreas Teske1, Diploma Engineer Daniel Bexte1,
Pharmacist Juliana Kotthoff2
1
L.B. Bohle Maschinen und Verfahren GmbH, Ennigerloh
2
Institute of Pharmaceutics and Biopharmaceutics, Heinrich Heine University Düsseldorf

Summary Introduction
Continuous production has now been adopted The continuous manufacturing of medicinal pro-
even in the pharmaceutical industry and is indis- ducts is now an indispensable, innovative process
pensable in current product developments. This in the pharmaceutical industry. Many publications
study focuses on the process of continuous direct in this field describe and claim a slow develop-
compression, with the aim of guaranteeing cons- ment from batch-based production to continuous
tant product quality over the whole production run. production, but reality is very different. Alongside
A production run over eight hours was carried out conferences taking place across the whole world
to manufacture a high-dose acetylsalicylic acid on an almost monthly basis dealing with this sub-
tablet. In-line, at-line and off-line analyses of the ject area closely, as well as research consortia de-
intermediate and final product were performed. dicated to this topic, there are now five products
Process and quality data were used to detect and which are manufactured continuously and which
analyze errors in the process and eliminate them are approved by the FDA (Food and Drug Adminis-
from future processes. Specific instances when tration) and EMA (European Medicines Agency).
the product quality was out of specification could
be fully explained with the aid of the recorded data. Some companies which are highly active in the
field and have placed several plants for conti-
nuous production in their production facilities
Keywords around the world, now even follow the maxim of
producing each new product with a QbD (Qua-
Continuous production, continuous direct com-
lity by Design) approach over continuous pro-
pression, QbCon, PAT, NIR
cesses. This shows that even the slow-to-mo-
dernize pharmaceutical industry is embracing
these new, better and safer methods of production.

Compared to batch-based manufacturing, the be-

nefits of continuous production are diverse, and
not every pharmaceutical production considers
the same benefits as a top priority. In a nutshell,
benefits range from the possibility of manufactu-
ring higher quality and therefore safer products,
to the possibility of substantial savings through
smaller GMP areas with less transport and sto-
rage, all the way up to greater production flexibility
in the life cycle of a product, to name just a few [1].

© L.B. Bohle Maschinen und Verfahren GmbH 1

Similar to various manufacturing processes in Methods
the “batch-based world”, the continuous manu-
The continuous direct compression was execu-
facturing of solid medicinal forms also provides
ted on the QbCon®dc (direct compression) line at
various manufacturing methods. The most pro-
L.B. Bohle Maschinen + Verfahren GmbH
minent routes are the manufacture of (film-coa-
(figure 1). This line consists of the following unit
ted) tablets based on continuous dry granulation,
operations for a pure direct compression setup:
continuous wet granulation and drying or conti-
nuous direct compression. Two of the products
Material supply
continuously manufactured and approved so far
are produced via continuous direct compression, The containers with raw ingredients are connec-
which provides the same benefits as batch pro- ted via normal hose lines to vacuum conveyors
duction, i.e. lower costs through great savings, (Volkmann, Germany) which carry the materials
shorter process times and smaller required ma- to the dosing station. The vacuum conveyors
chine and therefore GmP areas. The formula- are responsible for the first filling and later cyc-
tion must naturally be suitable for this purpose lical refilling of the gravimetric powder feeders.
and have sufficient flowability and compactibility.
Powder dosing and mixture
Due to the significant nature of the subject, this
The materials are supplied by means of gravimetric
study presents the results of a continuous direct
powder feeders of a twin screw type (GZD 100.12
tabletting process which was performed over
and GZD 200.12, Gericke AG, Switzerland) to the
a production period of eight hours (one working
continuous blender (GCM 450, Gericke AG, Swit-
shift). Alongside the simplicity of implementing
zerland), which generates a consistently high mixing
a new formulation, it shows that a long process
quality and should eliminate any feeder fluctuations.
runtime is possible without significant problems.
In-line and at-line process and product analyses
In-line quality control of the powder blend
measure the product quality in real time and allow
a high degree of process control and adjustability. The powder leaves the continuous blender and
Off-line analyses performed later should confirm is transported over a surface which has a ne-
the consistent quality of manufactured medicinal ar-infrared probe underneath (abbreviation: NIR,
forms. To understand this study, it is also import- SentroPAT and SentroProbe, Sentronic, Ger-
ant that this is a pure feasibility study without a many). This is able to detect ASS concentra-
product-relevant background. The formulation tion in real time through prior multivariate PLS
and the API (active pharmaceutical ingredient) (partial least squares regression) modelling.
concentration were chosen arbitrarily, and no dis-
covery study for possible process parameters or The calibration of the PLS model was perfor-
even a design space was performed beforehand. med with fresh API-excipient mixes of known
concentration, which were passed over the NIR
Materials measuring station. The adopted PLS model
was used after calibration in the entire wave-
Acetylsalicylic acid was used as API in a direct
length range of 1100-2100 nm. The integration
compressible quality. It is a free-flowing pow-
time per spectrum was 9ms, and 100 measure-
der, which contains 90% acetylsalicylic acid
ments were averaged for an output spectrum.
and 10% corn starch (abbreviation: ASS, Shan-
dong Xinhua Pharmaceutical Co., Ltd, Chi-
na). This was kindly provided in large quanti-
ties and had a total 78% ratio in the formulation.
Microcrystalline cellulose (abbreviation: MCC,
Vivapur 102, JRS, Germany) was used as a
dry binder (21%) and sodium stearyl fumara-
te (PRUV, JRS, Germany) with 1% concen-
tration was used as a lubricant for tabletting.
© L.B. Bohle Maschinen und Verfahren GmbH 2
Figure 1: Production facility (QbCon® dc) used for the continuous direct tabletting in this study.
Tabletting and at-line quality control gular intervals to test these for friability and dis-
integration time according to the pharmacopoeia.
The powder subsequently flows into the filling hop-
per of the tablet press (XL 200, Korsch AG, Germa-
The disintegration time of the tablet was respec-
ny), which produced tablets with a target weight of
tively assessed, which disintegrated in the slo-
240 mg. The rotor consisted of 25 punch stations
west time of six test specimens. Ten tablets were
fitted with a set of round, biconvex 8 mm punches.
respectively taken at the start and towards the
A tablet tester (UTS 4.1, Kraemer Elektronik,
end of the process, which had to be tested re-
Germany) was attached to the tablet press and
garding their ASS-content off-line via UV spec-
connected to the entire control system. Mea-
troscopy at a maximum absorption of 229 nm.
surements of the weight, breaking force and
height of the tablets in regular intervals were
The QbCon®dc line was operated with a
sent back to the tablet press for feedback cont-
throughput of 27.5 kg/h, which resulted in do-
rol in order to maintain consistent product quality.
sing rates of 18.975 kg/h for ASS/corn starch,
8.25 kg/h for MCC and 0.275 kg/h for PRUV. The
Other:
blender was operated with a constant rotational
The automation platform (L.B. Bohle, Germany) speed of 120 rpm, and the rotor speed of the tablet
guarantees strict process monitoring and control press could be changed in the 70-80 rpm range.
of the entire process. Data recording and visu-
alisation operates via this system. Since the ma- The required minimum tablet breaking force had to
terial flows of the blender and tablet press can be > 50 N. With short pretests, the precompression
never be 100% the same, the rotor speed of the force was set to 8 kN and the main compression force to
press was readjusted manually by the opera- 21 kN as a start value. Changing process para-
tor at irregular intervals to preserve a consistent meters over time are permitted through the
filling level in the powder hopper of the press automation system if the need for readjust-
(manually, because in this study an automated ment is detected and initiated downstream of
fill-level sensor was not attached to the press). the tablet press by means of a testing system.
Off-line samples of the tablets were taken at re-
© L.B. Bohle Maschinen und Verfahren GmbH 3
Results and discussion the powder weight in the feeder. The feeder scales
have to be switched off during this refilling period,
Gravimetric powder feeders replace the initial
since feeder control would not work reliably due to
weighing and dispensing step of batch produc-
the quick change in weight. In this time, the feeder
tion and are an integral part of practically every
therefore dispenses in a way which is not compara-
continuous production technology. A uniformi-
bly reliable, like in the gravimetric status [3]. This can
ty of the finished dosage form can only be gua-
be detected in light fluctuations in the dosing rate,
ranteed by well-adjusted dosing systems [2].
which occur especially when the feeder is refilled.
A gravimetric powder feeder works (simplified) by
recording the dispensed weight over time at a high
Another characteristic of many powders is the
frequency via the integrated load cell. The differen-
higher rotational speed which is necessary when
tial over time indicates the present dosing rate. In
the feeder reservoir becomes emptier over time
case of deviations from the target value, the cont-
(bottom right in figure 2 - red border). This be-
rol system reacts by adjusting the screw speed to
havior is due to the compressibility of a pow-
maintain the target dosing rate. This control con-
der. If the compressibility is high, mass-pressure
cept can be refined even more with preset dosing
from above causes a compaction of the pow-
profiles. Figure 2 shows the behavior of the feeder
der in the lower layers of freshly filled-up reser-
for the MCC. In the top part of the figure, three
voir and therefore an increase of the bulk den-
curves are shown which represent the current
sity. More material is moved into the screws
dosing rate (blue), the screw speed (black) and
and simultaneously carried away per rotation.
the current weight of powder in the feeder (red).
To maintain the target dosing rate, the rotational
speed must decrease. During the gradual emp-
Since material is constantly dispensed, the ab-
tyruptly. To a great extent, the feeder dispensed
solute weight of the powder decreases over time.
the MCC correctly in this production process.
Procedures running automatically refill the feeder
via vacuum conveyors, which explains the cycli-
The fluctuations of the dosing rate at
cally decreasing and further increasing value for
the start of the production process

Figure 2: Dosing rates, screw speed and static weight of the MCC feeder in the production time; bottom left:
Extract of the dosing rates in the initial period; bottom right: Extract of all three process parameters within a
short period.
© L.B. Bohle Maschinen und Verfahren GmbH 4
(blue border) will be discussed later on. ned by strong bridging in the dosing container
Figure 3 shows the dispensing of the ASS and where less powder has gone into the screws.
the PRUV, which show completely opposing be-
havior. The dosing rate of the ASS is constant The bridge had to be destroyed at a later stage so
over time and shows no anomalies, except for that the feeder continued working in a narrow rota-
the discussed short fluctuations occurring during tional speed range. Since the dispensing of the lu-
refilling routines. The constant rotational speed bricant can be critical for the subsequent tabletting
of the ASS feeder (yellow curve) even indicates stage, problems resulting from bad feeder perfor-
that the direct compressible quality of the ASS mance would be evident through high ejection for-
has a powder bed consolidation during the refil- ces, etc. in the tabletting process. As this was not the
ling process that is hardly noteworthy but is be- case, satisfactory dispensing and sufficient blen-
neficial to the stability of the dosing process. ding in the continuous blender could be presumed.

The dispensing of the PRUV was considered diffi- After continuous blending, there is the possibili-
cult for two reasons. The powder is quite cohesive, ty of automatically discharging the product which
adheres to surfaces, and does not have good flo- does not meet the previously defined specificati-
wability. Combined with the lower dosing rate of 275 ons. These previously defined specifications could
g/h, difficult dispensing is to be expected. In relati- be API concentration limits, over or under which
ve terms, greater fluctuations appear in the dosing further processing would certainly lead to tablets
rate (red curve) compared with MCC and ASS, but with an incorrect potency. This measurement was
the performance of the feeder is satisfactory, which carried out via NIR spectroscopy. Details on the
is not in the least due to the strong control of the measurement can be found in the methods sec-
screw speed. This increases from an initial value of tion. Since this trial is purely a feasibility study,
approx. 15% to values of 27% on average, sho- an automated discharging of non-conforming ma-
wing the typical, discussed fluctuations in this area terial was not in use. The NIR signal was simply
after the refilling stages. Starting from four hours used to monitor the API concentration over the
into the production time, the rotational speed sett- the entire production time.
les at approx. 22%. This behavior can be explai-

Figure 3: Dosing rates and screw speeds of the feeders for ASS and PRUV in the production time.
© L.B. Bohle Maschinen und Verfahren GmbH 5
Figure 4 shows the relative progress of API this case, lasted over several minutes and were
concentration over the entire production time. combined with a such a high incorrectly dispen-
The target content of the tablets and there- sed MCC amounts that the continuous blender
fore also of the powder was at 70.2% ASS. was no longer capable of eliminating these errors.

Two things are apparent. On the one hand, the ex- In comparison, fluctuations repeatedly ap-
cessive ASS concentration decreases in the first pear in the average dosing rate in the la-
40 minutes. After this period, the measured values ter stage of the MCC dosing process.
remain at a constant level under normal fluctuati- But these fluctuations are always limited to very
ons. On the other hand, the ASS content remains small time-periods and fluctuate towards too litt-
on average at 68.2 ± 1.2% and therefore under le as well as towards too much dispensed pow-
the envisioned 70.2%. For the case of the decrea- der. These errors can be eliminated by a con-
sing ASS concentration in the first 40 minutes, the tinuous blender. From the moment the highly
blue-bordered part of figure 2 can be consulted. inaccurate dispensing of MCC stops, the ASS
concentration measured via NIR spectrosco-
This image shows an extract of the dosing ra- py settles on a constantly maintained value.
tes curve of the MCC. It can be noted that the
MCC was initially under-dispensed over long In the case of the insufficient concentration of
periods. The under-dispensing of the MCC ASS over the remaining production period, there
is responsible for the temporarily excessi- is certainly the possibility that the PLS model for
ve concentration of ASS in the entire mixture. the NIR method was not calibrated with sufficient
data or in a sufficiently robust manner. However,
A continuous blender is capable of eliminating hig- off-line measured tablet content shows something
her-frequency feeder-errors up to a certain cut-off different at the start and at the end of the produc-
frequency, which, roughly speaking, is in the range tion run. Both values are in the range of the in-line
of the powder’s mean residence time in the blen- measured ASS concentration, which suggests that
der. However, the feeder-errors, which occurred in the PLS model for this application was sufficient-

Figure 4: In-line measured ASS concentration after the continuous powder blender and ASS concentration in
the tablets at the start and around the end of the production period; for tablets n= 10,
average ± standard deviation.
© L.B. Bohle Maschinen und Verfahren GmbH 6
ly precise. An evaluation of the raw ASS/starch rences in the concentration measured via NIR.
powder and the UV-spectroscopic measurement,
however, allowed to draw the conclusion that the The rotor speed was adjusted by the operator ma-
directly compressible ASS no longer correspon- nually, since the material flows from the blender
ded to the specified content after its shelf-life, and the tablet press are never completely synchro-
but amounted to approx. 87%, explaining an API nised, i.e. at the same level. Tablet compression is
concentration which was consistently too low. a volumetrically controlled process. Through a pre-
vious evaluation of the dependence of tablet break-
The continuous tabletting process proceeded wit- ing force from the dwell time of the main compressi-
hout any issues during the whole production time on, it was identified that a range between a 70 and
and constantly delivered (except for content va- 80 rpm rotor speed creates a design space where
riations at the start) tablets consistent with the tablets of the same quality are formed. In fact, du-
specified quality criteria. Particular attention has ring the process it was merely necessary to alter
to be given to figure 5, in which the process pa- the rotor speed between 73 and 76 rpm to main-
rameters pre-compression and main-compres- tain a constant fill-level of the tablet press hopper.
sion force as well as rotor speed are illustrated The quality attributes of the tablets were mea-
over the whole production period. Start-up and sured at-line, with the tablet-press convey-
shut-down processes of the tablet press - where ing a quantity of 10 tablets to the tablet tes-
the filling shoe is filled or runs empty are visible ting device automatically via the product
at the beginning and at the end of the diagram. diverter at the outlet approx. every 5 minutes.
These two stages can be recognised by the in-
creasing or decreasing compression forces. Ap- As a result of these findings, the process para-
parently, the various concentrations of ASS at meters were automatically adjusted to main-
the start of the process have no huge influence tain constant values for the tablet weight and
on the compressibility and on the measured com- breaking force. For instance, a slightly falling
pression forces, which, on the one hand, could be pre-compression force over the entire pro-
due to the similar bulk densities of the materials cess run can be identified in figure 5. The re-
and, on the other hand, to the fairly small diffe- sults of the measured tablet weights, heights

Figure 5: Important process parameters of the tabletting process in the production time.
© L.B. Bohle Maschinen und Verfahren GmbH 7
Figure 6: At-line measured critical quality attributes of the tablets; n= 10, average ± standard deviation.
and breaking forces are displayed in figure 6. for friability and the highest measured values for
the disintegration time (slowest disintegration of
The average weight of the measured tablets in n=6 tablets). In each case, the friability coinci-
that period amounted to 240.5 ± 1.3 mg, which ded with the value of < 1% required by the Ph.
corresponds to the required weight. The highest Eur., which is also very important for a downs-
deviation of a series of measurements was 1.6%, tream coating process. Since the aim was to ob-
which repeatedly shows that the tabletting pro- tain fast disintegrating tablets, which meet the
cess guarantees a constant tablet quality except requirements of the Ph. Eur. for non-coated ta-
for the initial deviations discussed. There were blets and disintegrate in < 15 min, the obtained
no previous requirements for the tablet height, values of < 60 s in almost each case are more
except for the fact that it had to be as consistent than sufficient to fulfil this quality requirement.
as possible over time, which could be confirmed
based on the data. An average height of 4.21 ±
0.02 mm was obtained over time, which corre-
sponds to a relative standard deviation of 0.45%.
This is a value which is absolutely satisfactory.

The same is true for the breaking force of the tab-

lets, which needed to be > 50 N; this was found to
be a sufficient strength for a later coating process.

The required breaking force was obtained in

each case in the 8h production time, in which
no individual measurement from which the ave-
rage values were calculated was under 50 N.
Other quality attributes of the tablets were ex-
amined off-line after the end of production.
Figure 7 shows the average measured values
© L.B. Bohle Maschinen und Verfahren GmbH 8
Figure 7: Off-line measured critical quality attributes of the tablets; n= 1 for friability and n= 6 for disintegration
time (value of the last disintegrated tablet shown).

Conclusion nufacturing of extended release tablets via powder

mixing and direct compression, Int. J. Pharm.,
The discussed data shows a successful continu- 495 (2015) 290-301.
ous production by means of direct tabletting over
eight hours. Not only were all the critical quality [3] R. Meier, J. Harting, J. Happel, P. Kleinebudde,
attributes maintained within their specifications Implementation of microwave sensors in continuous
over almost the entire production time, but it was powder feeding – a novel tool to bridge refill phases,
also possible to detect inaccurate intermediate pharmind, 79 (4) (2017) 576-582.
and final products via process-analytical tech-
nologies and process parameters. Even initial
products with a concentration that was too low
could be detected via the NIR method. The pre-
sent results show that continuous direct tablet-
ting for suitable formulations can represent easy
access to the world of continuous production.

L.B. Bohle Maschinen + Verfahren GmbH

Literature Industriestr. 18
D-59320 Ennigerloh, Germany
[1] J. Rantanen, J. Khinast, The future of pharmaceuti-
cal manufacturing sciences, J. Pharm. Sci., 104 (2015)
+49 2524 - 93 23 0
3612-3638.
[email protected]
www.lbbohle.de
[2] T. Ervasti, S.P. Simonaho, J. Ketolainen, P. Fors-
www.lbbohle.com
berg, M. Fransson, H. Wikström, S. Folestad, S. La-
www.continuous-production.com
kio, P. Tajarobi, S. Abrahmsén-Alami, Continuous ma-

© L.B. Bohle Maschinen und Verfahren GmbH 9