freeze-drying process monitoring using a cold plasma ... · modiﬁed. this was mainly due to the...

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


● 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


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


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


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).



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


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.



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


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).



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


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)



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


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.



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


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.


freeze-drying process monitoring using a cold plasma ... · modiﬁed. this was mainly due to the...

Documents