freeze-drying process monitoring using a cold plasma ... · modified. this was mainly due to the...
TRANSCRIPT
www.pda.org/bookstore
RESEARCH
Freeze-drying Process Monitoring Using A Cold Plasma
Ionization Device
Y. MAYERESSE,1 R. VEILLON,1 PH. SIBILLE,2 and C. NOMINE2
1Freeze-drying, Industrialization, GlaxoSmithKline Biologicals, Rue de l’Institut, 89, 1330 RIXENSART, BELGIUM
2Adixen Vacuum Technology, Avenue de Brogny, 98, BP2069, 74009 annecy Cedex, FRANCE
ABSTRACT: A cold plasma ionization device has been designed to monitor freeze-drying processes in situ by
monitoring lyophilization chamber moisture content. This plasma device, which consists of a probe that can be
mounted directly on the lyophilization chamber, depends upon the ionization of nitrogen and water molecules using
a radiofrequency generator and spectrometric signal collection. The study performed on this probe shows that it is
steam sterilizable, simple to integrate, reproducible, and sensitive. The limitations include suitable positioning in the
lyophilization chamber, calibration, and signal integration. Sensitivity was evaluated in relation to the quantity of
vials and the probe positioning, and correlation with existing methods, such as microbalance, was established. These
tests verified signal reproducibility through three freeze-drying cycles. Scaling-up studies demonstrated a similar
product signature for the same product using pilot-scale and larger-scale equipment. On an industrial scale, the
method efficiently monitored the freeze-drying cycle, but in a larger industrial freeze-dryer the signal was slightly
modified. This was mainly due to the positioning of the plasma device, in relation to the vapor flow pathway, which
is not necessarily homogeneous within the freeze-drying chamber. The plasma tool is a relevant method for
monitoring freeze-drying processes and may in the future allow the verification of current thermodynamic freeze-
drying models. This plasma technique may ultimately represent a process analytical technology (PAT) approach for
the freeze-drying process.
KEYWORDS: Freeze-drying, Process analytical technology, PAT, Cold plasma, Ionization, In-process monitoring
Introduction
Lyophilization, or freeze-drying, is the method of
choice for biologicals stabilization. This predominantly
three-phase process comprises formulation freezing, sub-
limation, and secondary drying. The first two phases are
performed below the glass transition temperature (Tg9),
and sublimation is also performed under vacuum to re-
move crystallized ice (1) (2). The product is warmed up
during secondary drying in order to desorb unfrozen and
sorbed water (3). Since its early development in the
1950s, freeze-drying has undergone many mechanical
and automatism improvements, including sterilization,
in-place cleaning, automatic loading and unloading, and
process monitoring. Nowadays, freeze-dryers are essen-
tial equipment and require operators to demonstrate
many competencies and skills.
As the complexity of biomolecules increases, the sta-
bilization needs become critical. Recent biotech prod-
ucts require well controlled stability, and today, the
freeze-drying process remains a unique stabilization
process (4) (5). In the case of vaccines, antigenicity is
linked to complex glycoprotein conformations or live
organism preservation. Therefore, vaccine freeze-dry-
ing covers the most stringent formulation and process
conditions. The complexity of the freeze-drying pro-
cess depends upon the freezing behavior of formula-
tions, thermodynamic thermal transfers, and gas flow
dynamics under vacuum (6) (7) (8).
As the equipment, products and processes are becom-
ing increasingly complex, the control of the freeze
drying process is critical. This is achieved by control-
ling the following parameters:
Corresponding authors. Tel 00 32 656 9752 fax 00 32 656
7349: Email Address: [email protected], Yves.
[email protected], [email protected]
160 PDA Journal of Pharmaceutical Science and Technology
www.pda.org/bookstore
● The freeze-dryer equipment (performance, size)
● The product (Tg9, formulation, volume, vial, stop-
per)
● The freeze-drying cycle (time, pressure, tempera-
ture)
During freeze-drying it is essential to monitor depen-
dent parameters such as sublimation endpoint and
product temperature. The sublimation endpoint is crit-
ical in two scenarios:
● If the secondary drying is launched before the
sublimation endpoint, the product temperature
may exceed Tg9, causing some vials to collapse
and an increased defect rate. This is a quality issue.
● If the secondary drying is delayed after the subli-
mation endpoint, the cycle will not be well opti-
mized and some freeze-drying capacity may be
wasted. This is a business issue.
With respect to product temperature measurement,
new freeze-dyers often have automatic loading de-
vices, which reduce the need for manual handling of
vials. This is beneficial, as operators remain a major
source of contamination. However, with automatic
loading devices, the placement of product temperature
probes is no longer possible, and product temperature
profiles are no longer available, even if product tem-
perature measurement is challenged. This results in a
lack of information for producers, quality assurance
officers, and authorities.
The aim of the process analytical technology (PAT)
initiative, which was launched by the Food and Drug
Administration in 2002, was to promote a better un-
derstanding and control of processes. As a result,
on-line monitoring devices have been developed. PAT
also aims to finalize the quality decision process dur-
ing the manufacturing process through the assessment
of critical process parameters, method development
and scaling up, capability assessment, and the deploy-
ment and transfer to industrial scale.
However, the deployment of PAT with respect to
freeze-drying faces the major barrier of lack of avail-
ability of compatible monitoring equipment. In fact,
the extreme conditions of the freeze-drying process
(vacuum, aseptic rules, and low temperature) render
many devices inappropriate. Therefore, the quest for
new monitoring devices to control the freeze-drying
process with a high degree of assurance has become
essential.
This section reviews the PAT tools currently used for
monitoring the freeze-drying process with respect to
the following criteria:
1. Global load monitoring. As freeze-drying is depen-
dent upon heat and mass thermal transfer, some
heterogenicities may limit control. It may be erro-
neous to rely on individual vial measurement to
control the whole load.
2. Compatibility with automatic loading/unloading
devices. The placement and removal of vials must
not be impaired.
3. Compatibility with cleaning in place(CIP)/stopper-
ing devices. No leads should compromise the
movement of shelves, CIP ramps, or nozzles.
4. Compliance with aseptic handling. There should be
no source of contamination within the materials or
during positioning.
5. Steam sterilization. The device must sustain re-
peated steam sterilization at a minimum of 123 °C
and 2 bars for a duration of 3 h.
6. Leakage control. Placement of the device should
not induce freeze-dryer leakage. It should also sup-
port at least a 5-m bar vacuum, and measurement
should be independent of equipment leak rate.
7. Simple integration into an industrial freeze-dryer.
The device should be installed to current existing
ports using triclamp flanges, and the data acquisi-
tion signal should be compatible with 21CFR
Scada / recorders.
8. Calibration or correlation to primary methods (see
Table I).
Few types of equipment fulfill all the success criteria
(9) (10) (11) (12) (13) (14) (15) (16) (17) (18). For
example, some can monitor whole loads but are hardly
steam sterilizable or compatible with leak control. The
moisture probe presented by Roy et al. (19) was prom-
ising, but failed to reach production scale as it could
not sustain steam sterilization. The last two devices in
161Vol. 61, No. 3, May–June 2007
www.pda.org/bookstore
TABLE I
Technology Comparison
Success Factor
Monitor
Global
Load
Automatic
Loading
CIP 1
Stoppering
Device
Aseptic
Handling
Steam
Sterilizable
Leak Rate
Control
Simple
Integration Calibration
Temperature probe NO NO NO 1/2 YES YES 1/2 YES
Wireless product
temperature probe NO NO NO 1/2 YES YES 1/2 YES
Conductivity probe NO NO NO NO YES YES YES YES
Microbalance NO NO NO NO NO 1/2 NO YES
FTNIR product probes NO NO NO NO YES YES NO YES
Pirani/capacitive differential
pressure control YES YES YES YES 1/2 YES YES YES
Moisture probe YES YES YES YES NO YES YES YES
Pressure rise measurement YES YES YES YES YES NO YES NO
Mass spectrometry YES YES YES YES NO NO NO
TDLAS YES YES YES YES YES YES NO
Cold plasma YES YES YES YES YES YES YES
FTNIR: Fourier transform infrared.
TDLAS: Tunable diode laser absorption spectroscopy.
162
PD
AJo
urn
alo
fP
harm
aceutic
alS
cie
nce
and
Techno
log
y
www.pda.org/bookstore
Table I could potentially be valuable for freeze-drying
monitoring. However, a full cost-benefit analysis
needs to be undertaken for each technique.
In the present study, we evaluate the cold plasma
technique for monitoring freeze-drying on a pilot,
scaling-up and industrial scale, with respect to com-
parison with other methods (microbalance, Pirani),
reproducibility, product signature in scaling-up, and
industrial lots monitoring.
Materials and Methods
Cold Plasma
This equipment is based on the inductive coupled
plasma/optical emission spectroscopy (ICP/OES)
technique, which is normally used in chemical and gas
analysis (Lyotrack, Adixen, France). Although this
type of instrument used to be larger, recent improve-
ments have allowed size reduction. It is now similar in
size to a pressure gauge. The device is composed of a
quartz tube that directly contacts the lyophilization
chamber, and it is connected to the freeze-dryer ports
via an ISO-KF25 or triclamp flange. All parts of the
device that make contact with the internal atmosphere
of the freeze-dryer are constructed from stainless steel,
quartz, or FPM (fluoroelastomere gasket), and can
withstand sterilization conditions at 2,3 or 2.3 bars and
135 °C. A plasma sensor measures the ratio of water
vapor to nitrogen under vacuum conditions (labeled as
humidity in the software). The principle of the plasma
sensor is portrayed in Scheme 1. A radio frequency
source with a low power consumption (,15 W) cre-
ates a cold plasma in the quartz tube under vacuum
below 3 mbar. The plasma remains active in the vac-
uum range from 3 to 0.005 mbar, which is compatible
with the pressure range utilized in lyophilization pro-
cesses. The light emitted by the plasma is character-
istic of the gas present in the plasma and can be
collected with an optical fiber and diffracted with an
optical spectrometer. The optical spectrum is analyzed
by the plasma sensor software, and the curve of hu-
midity is displayed in real time. The humidity signal
from the plasma ranges between 0 (no water vapor)
and 1 (saturated with water vapor). Scheme 2 summa-
rizes the integration of the device to the computer and
recorder. Humidity values may also be transferred
from the sensor via a 0–10-V analog output. This
signal may be further integrated to a 21 CFR digital
recorder.
The software delivers a graphical display of the
plasma signal curve versus time (Scheme 2). The
primary drying endpoint can be determined by signal
treatment using an adjustable threshold or signal first
derivative.
Microbalance
A microbalance was used as a reference method for
sublimation monitoring in pilot freeze-dryers (20).
This allows the direct measurement of mass transfer
Scheme 1
Cold plasma sensor principle.
Ethernet com.
Freeze
Dryer
Cold Plasma
Lyotrack PC + soft.
Power supply
21CFR recorder
Freeze-dryerSCADA
Ethernet com.
Freeze
Dryer
Freeze
Dryer
Cold Plasma
Lyotrack PC + soft.
Power supply
21CFR recorder
Freeze-dryerSCADA
Scheme 2
Integration scheme and software main panel with
plasma signal graph.
163Vol. 61, No. 3, May–June 2007
www.pda.org/bookstore
compared with Lyotrack, which measures vapor satu-
ration in the chamber; the product probe monitors the
heat and mass transfer balance. The microbalance is
placed directly on the freeze-dryer shelf (CWS-40
Martin Christ, Germany). The microbalance servomo-
tor, which is controlled by a personal computer inter-
face with Labview® platform software, lifted vials
every 5 min for weight loss monitoring and then
replaced them on the shelf. The data were treated
using Statistica 7.1 software (Statsoft, USA).
Freeze-dryer Load
In the pilot study, formulations were prepared with
sucrose 2%, 5%, 10% w/w and mannitol 4% w/w
(Merck, Germany) and water for injection (WFI).
Type I glass vials (Forma Vitrum, Switzerland) of 3 or
9 mL volume were filled with 2 mL of the prepared
solution and stoppered with bromobutyl stoppers (Hel-
voet, Belgium).
Freeze-dryer
Four freeze-dryers were used in our study:
● Pilot freeze-dryer, D25 Dry-winner (Heto, Den-
mark)
● Pilot freeze-dryer, Epsilon D25 (Martin Christ,
Germany)
● Scaling-up freeze-dryer, Finn Aqua FCM30D
(GEA, Germany)
● Industrial freeze-dryer GT500 (GEA, Germany)
The pilot freeze-dryers used were mounted with a
Pirani ACG controller (BOC Edwards, USA) and a
capacitive pressure gauge (MKS, USA). Data acqui-
sition was performed with a DX200P recorder
(Yokogawa, Japan). The data were treated using Sta-
tistica 7.1 software (Statsoft, USA).
Freeze-drying Cycle
The basic freeze-drying cycle was defined as follow-
ing: Freezing below –40 °C for at least 1 h; sublima-
tion at 0.080 mbar (106 mtorr), with a temperature
ramp of 3 h to –21 °C and holding –21 °C for at least
18 h.
Results
Use of the Cold Plasma Sensor
Figure 1 presents a typical response of the cold plasma
during a freeze-drying cycle. The plasma device was
placed on top of a 3.5-m3 chamber and the primary
drying was undertaken at constant pressure and tem-
perature over 21 h. As the vacuum was pulled down,
the signal increased rapidly over 40 min to reach vapor
saturation inside the lyophilization chamber. It re-
mained saturated for 12 h and then it decreased ac-
cording to a sigmoid-shaped curve. After 21 h of
primary drying, the signal became stabilized, indicat-
ing that most of the vapor had been replaced by
nitrogen, and the shelf temperature was gently in-
creased. The sensor signal responded to this tempera-
ture increase. By 27 h the response had re-stabilized at
3.5% of relative saturation, possibly indicative of sub-
Plasma signal during a freeze-drying cycle
00:00 06:00 12:00 18:00 24:00 30:00 36:00
relative time
0.0
0.2
0.4
0.6
0.8
1.0
Pla
sm
a s
ignal [
arb
.unit]
Figure 1
Cold plasma sensor response during a freeze-drying cycle. The sensor was positioned on top of the chamber.
The load was composed of 11,000 vials (eleven thousand).
164 PDA Journal of Pharmaceutical Science and Technology
www.pda.org/bookstore
limation end-point. A correlation to other end point
assessment methods will be proposed later.
At 28 h, secondary drying was carried out with a shelf
temperature rise and pressure dropdown. The pressure
was pulled down to maximal vacuum in 4 h. At lower
pressure (0.015 mbar), the relative composition of gas
is changed. The sensor response detected this pressure
drop, and the baseline signal shifted to 8.5% of rela-
tive saturation. When the shelf temperature was in-
creased after 31 h, moisture was desorbed from the
cakes and vapor flow increased. The sensor response
increased to 16% of relative saturation and ultimately
stabilized at 11% after 5 additional hours.
Positioning the Sensor
As the positioning of the sensor on the freeze-dryer
may have an impact on the signal, three different
positions were tested:
● on the lyophilization chamber port
● on the condenser port
● between chamber and condenser
The same freeze-drying cycle and load was used
throughout (200 vials, filled volume 0.5 mL with
sucrose 5%). A pilot freeze-dryer equipped with
multiple ports was used for this study. The three
different responses of plasma are plotted on Figure
2. When the cold plasma sensor was located on the
chamber, the response signal was higher and more
stable than when placed between the chamber and
condenser. If the plasma sensor was placed on the
tubing between the chamber and condenser, the
signal was high, but was more influenced by
changes in pressure. The measure was performed
using a dynamic flow with controlled pressure. As
long as the pressure was regulated by nitrogen in-
jection, the gas content changed sensitively. There-
after, the plasma response may be affected.
At the end of sublimation, the signal remained
higher in the chamber than at the in-between posi-
tion. When the device was settled on the condenser,
no response to vapor was observed, indicating that
this is not a suitable position for monitoring the
lyophilization process. Setting on either the cham-
ber or an intermediate position, therefore, seems
the best location, but larger-scale testing and differ-
Plasma response in function to position
chamber response
betw een chamber response
condenser response0:00 2:00 4:00 6:00 8:00 10:00 12:00 14:00
Relative time [hh:mm]
0.0
0.2
0.4
0.6
0.8
1.0
Pla
sm
a r
esponse [re
lativ
e to s
atu
ratio
n]
Figure 2
Plasma response in relation to positioning (1 chamber, - between chamber and condenser, ‚ condenser).
165Vol. 61, No. 3, May–June 2007
www.pda.org/bookstore
ent geometries will be necessary to optimize posi-
tioning.
Limit of Sensitivity and Correlation with Microbalance
The cold plasma sensor limit of sensitivity was estab-
lished in the pilot study, when the minimal load that
the plasma device could monitor was evaluated. In the
same set of experiments, the plasma system was com-
pared to other equipment used for freeze-drying mon-
itoring. In the first instance, a microbalance was used
to correlate the plasma sensor signal with a single vial.
In order to track most of the vapor leaving the vial, the
device was positioned on the tubing between chamber
and condenser.
A design of experiment (DOE) plan was initiated
using eight different freeze-drying cycles. The input
process parameters were chamber pressure (10.10-6
bar; 200.10-6 bar), shelf temperature (–31°C; –11°C),
and sucrose concentration (WFI 100%; sucrose 10%
w/w). The studied responses were microbalance sub-
limation endpoint and plasma sensor sublimation end-
point.
The experiments were conducted according to a full
factorial plan using three factors with two levels. The
sublimation endpoints were determined by the first
derivative of the signals (Figure 3). When only one
vial was sublimed, the two methods for sublimation
endpoint evaluation were well correlated; the R2 factor
is above 0.99 (Figure 4).
Cold Plasma Sensor Comparison with the
Pirani-capacitive Measurement and Product Probe
The cold plasma response was compared to the Pirani-
capacitive differential pressure measurement and the
product temperature probe, other common devices for
freeze-drying monitoring on a larger scale. The objec-
tive was to compare the sublimation endpoint with the
three methods for freeze drying 1000 vials. The ref-
erence methods used were
● individual vial monitoring with a product temper-
ature probe (Pt100)
● global load monitoring with a Pirani-capacitive
differential pressure measurement
After 12.5 h of primary drying, the three signals
simultaneously began to evolve. As the product tem-
perature started to rise from –35 °C to the shelf tem-
Sublimation end point time D.O.E.Pattern
Shelf temperature
Chamber pressure
Formulation dilution µbalance endpoint plasma endpoint
--- -1 -1 -1 20:47 21:07
+-- 1 -1 -1 15:06 13:12
-+- -1 1 -1 27:50 27:50
++- 1 1 -1 12:28 11:45
--+ -1 -1 1 16:55 16:48
+-+ 1 -1 1 12:36 12:45
-++ -1 1 1 22:05 22:18
+++ 1 1 1 11:45 12:14
Figure 3
Results obtained with the parallel testing of plasma sensor and microbalance.
Sublimation endpoint w ith Plasma and microbalance
Plasma end point = -0.0296+1.0296*x; 0.95 Pred.Int.
00:4200:8100:21
Sublimation endpoint w ith microbalance
12:00
18:00
24:00
Sublim
atio
n e
ndpoin
t w
ith P
lasm
a
Figure 4
Sublimation endpoint of one single vial determined
by plasma sensor and microbalance; (o) Individual
sublimation endpoint, (—) sublimation endpoint
regression, plasma sensor endpoint 5 – 0.0296 1
1.0296 x microbalance end point, R25 0.99, (----)
95% confidence interval of regression.
166 PDA Journal of Pharmaceutical Science and Technology
www.pda.org/bookstore
perature (–20 °C), both the Pirani and plasma sensor
signals decreased. After 15 h 20 min, the product
temperature reached the shelf temperature, indicating
that the monitored vial had no more sublimation en-
dotherm and the sublimation endpoint had been
achieved for this vial. The product probe signal
matched the plasma sensor signal. However, this cor-
relation was not extensively studied because product
temperature is dependent on temperature probe posi-
tioning inside the vial, and the probe may monitor heat
transfer rather than mass transfer.
The Pirani signal exhibited a small sigmoid decrease
up to 14.5 h, but the signal was perturbed by the
nitrogen injections for pressure control, suggesting
that this system is not ideal for the measurement of
signal stabilization.
The plasma signal marked a deep sigmoid decrease
from 1 to 0.1 arbitrary unit (au) after 15 h, where 1 au
indicates that the atmosphere of freeze-dryer is only
composed by water vapor and 0 au only composed by
nitrogen. This signal continued to decrease until 18 h,
when the signal derivative was less than 0.001 au per
hour (Figure 5).
Reproducibility Between Batches
Another evaluation test was performed on a larger
scale, utilizing up to 11,000 vials (3 mL) per load. The
plasma sensor was positioned on the top of the cham-
ber. The objective was to test signal reproducibility
after three consecutive placebo runs under the follow-
ing conditions:
● same equipment
● same load (vial number, formulation, filling vol-
ume)
● same freeze-drying cycle
As shown in Figure 6, the plasma responses were very
similar from one cycle to another. The sublimation
endpoint and the secondary drying signal were nearly
the same between the three cycles. The plasma signal
-35
15
65
115
165
12:00 13:00 14:00 15:00 16:00 17:00 18:00
time [hh:mm]
tem
pera
ture
[°C
]
pir
an
i-cap
acit
ive [
arb
.un
it]
0.00
0.10
0.20
0.30
0.40
0.50
0.60
0.70
0.80
0.90
1.00
Pla
sm
a s
ign
al [a
rb.u
nit
.]
pirani relative signal shelf temperature product temp plasma signal
Figure 5
Sublimation endpoint determination with three methods: Plasma sensor (1), Product temperature (-), Pirani-
capacitive relative signal (E).
Plasma signal reproductibility over 3 lots
00:00:00 06:00:00 12:00:00 18:00:00 24:00:00 30:00:00 36:00:00
Freeze-drying relative time
0.0
0.2
0.4
0.6
0.8
1.0
Pla
sm
a s
ignal [
rela
tive to s
atu
ratio
n]
Lyotrack lot1
Lyotrack lot2
Lyotrack lot3
Figure 6
Evaluation of plasma sensor consistency during
three consecutive lots under the same conditions
(product, lot size, cycle, freeze-dryer). Lot 1 (—),Lot 2 (- -), Lot 3 (. . .).
167Vol. 61, No. 3, May–June 2007
www.pda.org/bookstore
showed consistency between these three consecutive
lots.
Use of Plasma in Scale-up Freeze-dryers
This later study focused on the scaling-up of the
lyophilization process. Scaling-up is a sensitive issue
because freeze-dryer geometry may vary from one
piece of equipment to another. In this study, the iden-
tical placebo product was scaled up from pilot (210
vials in a 0.2-m3 chamber) to scaling-up freeze-dryer
(11,600 vials in a 3.5-m3 chamber). In both tests, the
plasma sensor was positioned in the lyophilization
chamber. As seen in Figure 7, although though the
signal was relatively higher for the pilot freeze-dryer,
the plasma responses between the two lot sizes exhib-
ited a very similar shape. For the two cycles, the
setting and the product were the same. The sublima-
tion endpoint was similar between the two lot scales,
but the sublimation lasted for 1 more hour with the
pilot freeze-dryer. The secondary maximal signal,
which corresponds to the sorbed moisture release,
appeared at approximately the same time, with less
than a 15-min delay for a scaling-up size.
Use of Plasma Sensor for Secondary Drying
Monitoring
The secondary drying behavior of different formula-
tions was compared under the same process condi-
tions, using an amorphous sucrose formulation and a
crystalline formulation of mannitol with some manni-
tol hydrate (a complementary study by X-ray diffrac-
tion showed some mannitol hydrate formation).
With the sucrose formulation, the plasma sensor signal
showed a fast release of moisture upon desorption
(Figure 8). Once the shelf reached the target temper-
ature, the sorbed moisture was released and equilib-
rium was reached asymptotically. However, for crys-
talline mannitol the plasma sensor signal did not
increase immediately, even though the target shelf
temperature was reached (Figure 8). In the case of
crystalline mannitol, the structure retained more mois-
ture upon desorption, and the plasma sensor was able
to track this behavior. The last placebo test was con-
ducted on a full industrial scale, involving over
100,000 3-mL vials and a 3-day cycle. The plasma
sensor was set at the top of the approximately 15-m3
freeze-dryer chamber. As shown in Figure 9, the
Plasma signal during scaling up studies
(same product / same cycle)
( 2 freeze dryer size / 2 lot size)
00:00:00 12:00:00 24:00:00 36:00:00
Freeze-drying relative time
0.0
0.2
0.4
0.6
0.8
1.0
Pla
sm
a s
ignal [
rela
tive to s
atu
ratio
n]
lyotrack lot scale up
lyotrack lot pilot
Figure 7
The cold plasma sensor signal during the scal-
ing-up study with the same placebo. The scaling-up
consisted of moving from pilot freeze-dryer (—) to
scaling-up freeze-dryer (- -).
Mannitol Secondary Drying
0:00:00 1:00:00 2:00:00 3:00:00 4:00:00 5:00:00 6:00:00
Relative time [hh:mm:ss]
0.56
0.58
0.60
0.62
0.64
0.66
0.68
0.70
Pla
sm
a s
ignal [
arb
.unit]
Sucrose secondary drying
00:00:00 01:00:00 02:00:00 03:00:00 04:00:00 05:00:00 06:00:00
relative time
0.06
0.07
0.08
0.09
0.10
0.11
0.12
0.13
0.14
0.15
0.16
0.17
Pla
sm
a S
ignal [
arb
.unit]
Figure 8
Plasma sensor signal during secondary drying study with two formulations. The cycles were the same for
mannitol formulation (E) and sucrose formulation (h).
168 PDA Journal of Pharmaceutical Science and Technology
www.pda.org/bookstore
plasma device remained sensitive under these condi-
tions. When sublimation commenced, the signal rap-
idly reached saturation. After 20 h the signal started to
decrease in a sigmoid shape, but increased again when
the signal was close to stabilization. At this stage the
shelf temperature was reasonably increased to accel-
erate the sublimation endpoint. The plasma signal
slowed down as the sublimation of remaining vials
stopped. At the sublimation endpoint the signal stabi-
lized at a higher level than during scaling-up tests. At
the beginning of secondary drying, the pressure was
reduced to full vacuum, and the sensor signal in-
creased because of proportional changes in chamber
gas content. When the shelf temperature was raised,
the plasma sensor responded to the release of sorbed
moisture. After 6 h the signal started to return to its
initial level during secondary drying, indicating the
potential endpoint of secondary drying.
The cold plasma sensor demonstrated an ability to
monitor industrial-sized lots. The persisting high
moisture signal at the end of primary drying suggests
that top chamber moisture remains high and that po-
sitioning would be best suited in the vapor flow path-
way.
Discussion
In order to become accepted as a suitable in-process
monitoring device, the cold plasma system has to
demonstrate sensitivity, reproducibility, and correla-
tion to existing methods. In this series of studies, a
cold plasma system for the monitoring and validation
of freeze-drying cycles was assessed at the research
and development level, during scaling-up, and on an
industrial scale.
As a research and development tool, this system pro-
vides a greater understanding of the kinetics of the
freeze-drying cycle. The main application for this
technology is an accurate determination of the subli-
mation endpoint, based on the quantity of remaining
water molecules in the freeze-drying chamber. One of
the challenges of freeze-drying validation is to prove
consistency throughout process parameter testing and
product quality attributes. The process parameters are
either direct (shelf temperature, chamber pressure,
moisture content) or indirect (product temperature).
With the cold plasma sensor, a direct parameter is
monitored for the full load of the freeze-dryer. In
comparison, product probes only measure an indirect
parameter (the endothermic effect generated by subli-
mation) on an individual vial within the load.
Within a freeze-dryer chamber, there are two ways to
control pressure: either by injecting nitrogen to com-
pensate the vacuum generated by the pumps, or by
directly controlling the vacuum pumps with a valve.
With nitrogen injection control, when the pressure
decreases, the quantity of nitrogen is reduced in rela-
tion to the water molecules. The cold plasma sensor
signal is therefore sensitive to changes in pressure, due
to the relative quantitative measurement of water ver-
sus nitrogen inside the freeze-dryer. Secondary drying
is often associated with a drop in pressure and subse-
quent modification in the water-to-nitrogen ratio, as
the pressure control is achieved through the injection
of nitrogen. The plasma signal should therefore be
normalized in relation to total pressure in order to
avoid this baseline signal shift during secondary drying.
The cold plasma device also allows gas dynamics
inside a freeze-dryer to be evaluated. As shown in
Figure 7, under non-saturated conditions, there is a
clear difference between pilot and scale-up freeze-
dyers with respect to nitrogen injection. As nitrogen
molecules are heavier than water molecules and have
a different thermal conductivity (nitrogen 0.146 z W z
m-1z °C-1, water 0.603 W z m-1
z °C-1), the ratio of
gases present in the freeze-dryer will affect the heat
transfer between the shelf and the vial at low pressure.
The gas composition will in turn influence the subli-
mation speed, resulting in a different cycle time. This
explains why it is not equivalent to control the freeze-
dryer pressure by closing and opening the vacuum
Lyotrack during a placebo test at full scale
00:00 12:00 24:00 36:00 48:00
relative time
0.0
0.2
0.4
0.6
0.8
1.0
lyotr
ack full
scale
shelf temperature profile
Figure 9
Cold plasma sensor signal during placebo testing at
full industrial scale.
169Vol. 61, No. 3, May–June 2007
www.pda.org/bookstore
pump valves or by injecting a gas. This phenomenon
has a practical application for the study of pressure
control and its impact on the freeze-drying process
between different freeze-dryers. Indeed, the cold
plasma system can help to fine-tune a good regulation
of nitrogen inside the freeze-dryer.
The signal response of the system in relation to probe
positioning confirms that water vapor repartition in-
side a freeze-dryer is not homogeneous during the
cycle. With respect to chamber position, the signal
slope shows a smaller decrease when it is located in
the tubing between the chamber and condenser. This
may be explained by the difference in volume between
these two parts of the freeze-dryer and by the prox-
imity of the coils to the separation tubing. At the
beginning of the cycle, the first water molecules to
escape from the vials travel randomly inside the cham-
ber and bounce off the walls and shelves until they
reach the condenser coils. As the shelf temperature
increases, the system becomes saturated with water
molecules and a gradient of water molecules is created
between the condenser, in-between tubing, and cham-
ber. This gradient of water molecules creates a diffu-
sion flow from the chamber towards the condenser.
Each water molecule escaping from a vial is directly
taken by the flow of the others and pushed towards the
condenser, where the cold coils induce condensation.
This is confirmed by the low values of water inside the
condenser. When the primary drying endpoint is
reached, the flow of water molecules ceases and a
sharp reduction in water molecules inside the tubing is
noted. Equilibrium is finally reached when a quantity
of molecules remain evaporating in the chamber (de-
sorption phase).
Following the law of probability, the water molecules
finding their way towards the condenser through the
tubing, generally represent 1/30 of the chamber sur-
face. Upon entering the tubing they diffuse towards
the cold coils. As the coils have a lower temperature,
a partial water pressure difference is created locally.
An interesting application of the cold plasma sensor is
to undertake humidity mapping of the freeze-dryer. If
measurements are made at different points during the
freeze-drying cycle, these data can be used for a
numerical analysis of mass transfer during the process
leading to a more precise modelization for the kinetics
of mass transfer in larger freeze-dryers. Indeed, many
models for freeze-drying have focused on the mass
balance within the vial (21):
Qin 5 Qout
Qin 5 Avz K
vz ~Ts 2 Tp!
Qout 5 DHsub z ~dm/dt!
Avz K
vz ~Ts 2 Tp! 5 DHsub z ~dm/dt! (1)
where
Qin 5 heat flow (cal/h or W)
Av
5 internal surface of vial (m2)
Ts 5 shelf temperature (K)
Tp 5 product temperature (K)
Qout 5 mass flow (cal/h or W)
DHsub 5 enthalpy of sublimation (2,839,000 J/Kg)
(dm/dt) 5 mass transfer rate 5 sublimation rate
(Kg/h)
(dm/dt) 5 Av z dL/dt, with dL/dt 5 the position of the
subliming front as function of time
Kv
5 heat transfer coefficient (W z m-2z K-1)
Freeze-drying modelling should consider the global
pathway of moisture from the vial to the condenser,
and some authors have proposed the Knudsen formula
to express the sublimation speed Gsub: (22):
Gsub 5 kPintS M1/ 2
2pRTD , 0 , k , 1 (2)
where
k is the coefficient of evaporation
Pint is the saturated vapor pressure of ice
M is the molecular weight of water vapor
R is the gas constant
Tint is the absolute temperature of sublimating ice
The effect of the condenser may be represented as
follows:
Gsub 5 k~Pint 2 P9c!S M1/ 2
2pRTD l . l (3)
170 PDA Journal of Pharmaceutical Science and Technology
www.pda.org/bookstore
where
P9c is the pressure in the condenser
l is the distance from the sublimating ice to the
condensing ice
l is the mean free path of vapor molecule between
two collisions
The cold plasma sensor may allow the direct relative
measurement of Pint and P9c, and the evaluation of the
Knudsen coefficient of evaporation and Gsub. These
data may help characterize the gas flow kinetics within
a freeze-drying unit. The sensor may also be an ap-
propriate tool for modelling the vapor flow within a
freeze-dryer.
On a pilot scale, the cold plasma system was shown to
be sensitive to the sublimation of a single vial con-
taining a volume of 0.5 mL. This is of particular
interest, as the lyophilization process depends upon
individual vial behavior when each vial is being pro-
cessed individually. During the thermodynamic pro-
cess, some vials may be delayed and lead to hetero-
geneity, which is a major concern. This is controlled
through the cycle adjustment of stationary phases. The
cold plasma sensor could therefore be applicable to
homogeneity validation studies. The cold plasma de-
vice was also sensitive throughout the range of cham-
ber pressure (from 10.10-6 to 200.10-6 bar), shelf tem-
perature (from –31 to –11 °C), and formulation
resistance to water flow (from null with pure water to
high value with sucrose 10%). Because the sensor can
measure the water vapor content in freeze-drying from
1 to 100,000 vials, it can be used to develop a freeze-
drying cycle.
The correlation with existing methods of lyophiliza-
tion monitoring was achieved by comparing the cold
plasma sensor to microbalance and Pirani-capacitive
systems. The correlation of the plasma sensor and
microbalance was established with a single vial using
a DOE plan comprising eight tests. For the sublima-
tion endpoint monitoring, a correlation coefficient
above 0.99 was obtained between microbalance and
plasma sensor (Figure 10). The experimental design
showed that the main interactions with sublimation
speed were:
y 5 b0 1 ~b1x1 1 b2x2 1 b3x3!
1 ~b12x1x2 1 b13x1x3 1 b23x2x3! 1 ~b123x1x2x3! (4)
The experiment illustrated in Figure 3 demonstrated a
good correlation for endpoint determination between
microbalance and the plasma sensor. When an in-
depth analysis of the parameters was performed, the
main influencing parameters were found to be similar
between microbalance and plasma sensor. Shelf tem-
perature had the major impact on sublimation time,
followed by pressure and formulation concentration.
These conclusions about formulation are linked to the
chosen range (0–10% sucrose). However, if the sugar
concentrations were much higher (e.g., 40% sucrose),
the influence would become predominant. It is un-
likely that a product above 20% of dry material would
be freeze-dried.
As expected for a thermodynamically closed system
such as a freeze-dryer, the analysis of different mon-
itoring devices shows the strongest correlation be-
tween shelf temperature and pressure. The advantage
of microbalance in monitoring a freeze-drying cycle is
the continuous measurement of the weight decrease
inside vials. But as the plasma sensor is correlated
with microbalance, the knowledge of endpoint for
primary drying is sufficient to develop a robust freeze-
drying cycle. The comparison of the plasma sensor
with the Pirani-capacitance system confirmed a corre-
lation, but the plasma sensor was far more sensitive to
the water decrease inside the chamber. This correla-
tion was hardly achieved because of the Pirani gauge’s
signal lack of accuracy.
The plasma system can also determine the primary
drying endpoint, and can help to establish the temper-
ature slope between primary and secondary drying in
order to avoid a huge exhaust of water during this
Factor Name Sublimation time factor [h]
Micro-balance
Plasma sensor
β0 Intercept 17.44 17.25
β1 temperature –4.46 –4.77
β2 pressure 1.09 1.28
β3 dilution –1.61 –1.23
β12 Temperature x pressure
–1.96 –1.78
β13 Temperature x dilution
0.80 1.24
β23 Pressure x dilution –0.01 –0.04
β123 Temperature x pressure x dilution
0.46 0.27
Figure 10
Estimation of factor interactions with microbal-
ance and plasma sensor using a full factorial plan.
171Vol. 61, No. 3, May–June 2007
www.pda.org/bookstore
phase. If the ramp is too sharp, the glass transition of
the product may be exceeded, resulting in cosmetic or
stability problems. When we observed the plasma
signal in the sublimation ending region, the system
seemed to be sensitive to remaining sublimating vials
in the load. Indeed, even a small increase in shelf
temperature was detected by the plasma sensor. The
first derivative of the signal is a good way to control
process stabilization. Figure 11 illustrates the evolu-
tion of first derivative of the plasma signal. The pro-
posed threshold of 0.01 relative units per hour was
empiric and was determined through microbalance
individual vial testing. On a larger scale the threshold
should be adapted in relation to product specification
and visual defect rate.
Reproducibility was assessed through three placebo
tests under equal conditions and critical process pa-
rameters. As seen in Figure 6, a good signal superpo-
sition between the three cycles was recorded, indicat-
ing a good batch-to-batch reproducibility of the
plasma sensor signal and its potential for consistency
evaluation.
With respect to the product, even slightly different
formulations have characteristic fingerprints with the
cold plasma system. This system could therefore be
potentially used to assess the identity of the product
being freeze-dried. These data confirm that the plasma
sensor is eligible in the PAT toolbox for freeze-drying
as qualitative equipment.
One of the main benefits of plasma sensors is the
ability to be mounted on almost any kind of freeze-
dryers, irrespective of size. Indeed, scaling-up can be
achieved (Figure 7), and only a few more hours are
needed at the end of primary drying to assure a good
transfer of the cycle towards the industrial freeze-
dryer.
The kinetics of desorption during secondary drying
can also be monitored using the plasma device (Figure
8). This is of particular interest when freeze-drying
formulations containing hydrate crystal excipients. In-
deed, our study with mannitol confirmed that plasma
sensor was useful in monitoring hydrate form desorb-
tion once the activation energy has passed.
It is unlikely that product temperature probes will be
placed on industrial freeze-dryers with Sterilization In
Place (SIP), CIP and automatic loading and unloading.
Therefore, to be fully compatible with production
freeze-dryers, a monitoring device will require simple
integration and to be steam-sterilizable. The installa-
tion of the plasma device is similar to a capacitive
pressure probe with triclamp joints, and the output
signal can be directly integrated to a 21CFR digital
recorder. The plasma probe is steam-sterilizable; the
parts in contact to the lyophilization chamber are
stainless steel and a quartz tube. Cold plasma ioniza-
tion is therefore an efficient alternative technique to
product temperature probes.
Information provided by cold plasma sensors can be
useful from a PAT perspective, as the freeze-drying
cycle can now be followed with respect to indirect
(shelf temperature, pressure and time) and direct (wa-
ter quantity) parameters, thus allowing batch compar-
isons. Integration from a PAT perspective has the
advantage of global load monitoring and the ability to
Cold plasma signal and first derivative
0
0.2
0.4
0.6
0.8
1
00:0000:8100:21
Relative time [hh:mm]
Lyo
trac
k s
ign
al
[rela
tive
un
it]
-0.5
-0.4
-0.3
-0.2
-0.1
0
Fir
st d
eri
va
tiv
e [u
nit
/h]
Figure 11
Cold plasma sensor signal (__) and signal first derivative to time (‚) when the sublimation endpoint is reached.
172 PDA Journal of Pharmaceutical Science and Technology
www.pda.org/bookstore
evaluate product quality on the basis of in-depth pro-
cess parameters monitoring.
The plasma sensor is already established as a qualita-
tive system, but by using a calibration procedure it
will be possible to transform the obtained values into
a measure of water quantity within the freeze-dryer. A
custom-built calibration bench was developed to re-
create the atmosphere present inside a freeze-dryer.
Plasma calibration involved searching for the maxi-
mum intensity of optical peaks corresponding to water
vapor and nitrogen. A floating calibration was per-
formed to ensure that two different units could be
collected at the end the same signal. Using this bench,
the percentage of water vapor to nitrogen has been
adjusted from 0 to 100% over total pressure ranges
from 0.005 to 0.5 mbar. However, the cold plasma
sensor does not give any absolute measurements, and
further investigations are required in order to find a
fixed reference of moisture level. Furthermore, this
calibration bench working under vacuum is complex
to operate, and the calibration procedure will certainly
require further developments.
Conclusions
The cold plasma sensor appears to be a promising tool
for monitoring the freeze-drying process. Its main
interest lies in the ability to control freeze-drying at
each steps of industrialization from robust cycle de-
sign to scaling-up testing, homogeneity validation, and
industrial consistency lots. The main advantages of
this method are the sensitivity to a single vial, global
monitoring, and the noninvasive aspect. Other benefits
include compatibility with automatic loading, aseptic
handling, and sterilization. Reproducibility under the
same conditions and correlation with other monitoring
methods were established. The main limitation of the
plasma sensor is the requirement for adequate posi-
tioning within the vapor flow pathway in larger freeze-
dryers. Calibration will also require complex equip-
ment capable of generating a vapor-saturated chamber
under vacuum. The perspective is the evaluation of
cold plasma ageing in the production environment. In
the future, the sensor signal should be normalized in
relation to total chamber pressure in order to gain an
easier signal interpretation. Nevertheless, the cold
plasma technology may fulfill the gap for freeze-
drying monitoring equipment in the PAT field of ac-
tivity.
Acknowledgments
Many thanks to Adixen and GSK Bio development
teams for their transversal collaboration.
References
1. MacKenzie, A. P. The physico-chemical basis for
the freeze-drying process. Dev. Biol. Stand. 1977,
36, 51–67; and the International Symposium on
Freeze-drying of Biological Products, Washing-
ton, D.C., 1976.
2. Pikal, M. J.; Shah, S. The collapse temperature in
freeze drying: Dependence on measurement meth-
odology and rate of water removal from the glassy
phase. Int. J. Pharm. 1990, 62 (2–3), 165–186.
3. Pikal, M. J. Intravial distribution of moisture dur-
ing the secondary drying stage of freeze drying.
PDA J. Pharm. Sci. Technol. 1997, 51 (1), 17–24.
4. Jiang, S.; Nail, S. L. Effect of process conditions
on recovery of protein activity after freezing and
freeze drying. Eur. J. Pharm. Biopharm. 1998, 45
(3), 249–257.
5. Griebenow, K. Lyophilization-induced reversible
changes in the secondary structure of proteins.
Proc. Natl. Acad. Sci. U.S.A. 1995, 92 (24),
10969–10976.
6. Livesey, R. G. A discussion of the effect of cham-
ber pressure on heat and mass transfer in freeze-
drying. J. Parenter. Sci. and Technol. 1987, 41
(5), 169–17.
7. Overcachier, E. Lyophilization of protein formu-
lation in vials: Investigation of the relationship
between resistance to vapor flow during primary
drying and small scale product collapse. J. Pharm.
Sci. 1999, 88 (7), 688–695.
8. Greiff, D. Factors affecting the statistical param-
eters and patterns of distribution of residual mois-
tures in arrays of samples following lyophiliza-
tion. J. Parenter. Sci. Technol. 1989, 44 (3), 118–
129.
9. Jennings, T. A. Calorimetric monitoring of lyoph-
ilization. PDA J. Pharm. Sci. Technol. 1995, 49
(6), 272–282.
173Vol. 61, No. 3, May–June 2007
www.pda.org/bookstore
10. Gleseler, H. Influence of vial packing density
on continuously measured drying rate during
freeze drying using microbalance technique”,
4th World Meeting on Pharmaceutics,
Friedrich-Alexander University of Erlangen-
Nurenburg, 2002.
11. Roth, C. Continuous measurement of drying rate
of crystalline and amorphous systems during
freeze drying using in situ microbalance tech-
nique. J. Pharm. Sci. 2001, 90 (9), 1345–1355.
12. Xiang, J.; Hey, J. M.; Liedtke, V.; Wang, D. Q.
Investigation of freeze-drying sublimation rates
using a freeze-drying microbalance technique. Int.
J. Pharm. 2004, 279 (1-2), 95–105.
13. Brulls, M.; Folestad, S.; Sparen, A.; Rasmuson, A.
In-situ near-infrared spectroscopy monitoring of
the lyophilization process. Pharm. Res. 2003, 20
(3), 494–499.
14. Remmele, R. L.; Stushnoff, C.; Carpenter, J. F.
Real-time in situ monitoring of lysozyme during
lyophilization using infrared spectroscopy: Dehy-
dration stress in the presence of sucrose. Pharm.
Res. 1997, 14 (11), 1548–1555.
15. Tang, X. Nail, S. L.; Pikal, M. J. Evaluation of
manometric temperature measurement (MTM), a
process analytical technology tool for freeze-drying:
Part III, heat and mass transfer measurement.
AAPS PharmSciTech. 2006, 7 (7), 97.
16. Chase, D. R. Monitoring and control of the lyoph-
ilization process using a mass flow controller.
Pharmaceutical Engineering 1998, 18 (1).
17. Monteiro Marques, J. P.; Le Loch, C.; Wolff, E.;
Rutledge, D. N. Monitoring freeze-drying by low
resolution NMR: Determination of sublimation
endpoint. J. Food Sci. 1991, 56 (6), 1707–1728.
18. Schelenz, G.; Engel J.; Rupprecht, H. Sublimation
during lyophilization detected by temperature
profile and X-ray technique. Int. J. Pharm. 1994,
113 (2), 133–140.
19. Roy, M. L.; Pikal, M. J. Process control in freeze-
drying: Determination of the end point of subli-
mation drying by an electronic moisture sensor. J.
Parenter. Sci. Technol. 1989, 43 (2), 60–66.
20. Mayeresse, Y.; Veillon, R. On the way to rational
freeze-drying. Am. Pharm. Rev. 2006, 9 (7), 37–42.
21. Pikal, M. J.; Roy, M. L.; Shah, S. Heat and mass
transfer in vial freeze-drying of pharmaceuticals:
Role of the vial. J. Pharm. Sci. 1984, 73 (9),
1224–1237.
22. Mellor, J. D. Fundamentals of Freeze-drying. Ac-
ademic Press: London, 1978.
174 PDA Journal of Pharmaceutical Science and Technology