RESEARCH ARTICLE – Pharmaceutics, Drug Delivery and Pharmaceutical Technology

Application of Process Analytical Technology for Monitoring

Freeze-Drying of an Amorphous Protein Formulation: Use of
Complementary Tools for Real-Time Product Temperature
Measurements and Endpoint Detection
STEFAN C. SCHNEID,1 ROBERT E. JOHNSON,2 LAVINIA M. LEWIS,2 PETER STÄRTZEL,1 HENNING GIESELER1,3
1
Freeze Drying Focus Group, Division of Pharmaceutics, University of Erlangen-Nuremberg, Erlangen 91058, Germany
2
Global Biologics, Pfizer Inc., Chesterfield, Missouri 63017
3
GILYOS GmbH, Wuerzburg 97076, Germany

Received 8 November 2014; revised 2 January 2015; accepted 22 January 2015

Published online 17 February 2015 in Wiley Online Library (wileyonlinelibrary.com). DOI 10.1002/jps.24389

ABSTRACT: Process analytical technology (PAT) and quality by design have gained importance in all areas of pharmaceutical development
and manufacturing. One important method for monitoring of critical product attributes and process optimization in laboratory scale
freeze-drying is manometric temperature measurement (MTM). A drawback of this innovative technology is that problems are encountered
when processing high-concentrated amorphous materials, particularly protein formulations. In this study, a model solution of bovine serum
albumin and sucrose was lyophilized at both conservative and aggressive primary drying conditions. Different temperature sensors were
employed to monitor product temperatures. The residual moisture content at primary drying endpoints as indicated by temperature sensors
and batch PAT methods was quantified from extracted sample vials. The data from temperature probes were then used to recalculate critical
product parameters, and the results were compared with MTM data. The drying endpoints indicated by the temperature sensors were not
suitable for endpoint indication, in contrast to the batch methods endpoints. The accuracy of MTM Pice data was found to be influenced by
water reabsorption. Recalculation of Rp and Pice values based on data from temperature sensors and weighed vials was possible. Overall,
extensive information about critical product parameters could be obtained using data from complementary PAT tools.  C 2015 Wiley

Periodicals, Inc. and the American Pharmacists Association J Pharm Sci 104:1741–1749, 2015
Keywords: process analytical technology; quality by design; freeze-drying; protein formulation; process optimization; manometric tem-
perature measurements; water re-absorption; proteins; lyophilization; stability

INTRODUCTION because of the removal of ice from the side toward the center
also impairs the accuracy of pressure rise measurements, and is
The growing need for process optimization and process time
influenced by the vial size and fill depth. In many cases, reduc-
reduction in freeze-drying has led to cycle conditions at which
tion of Ap becomes significant in the last third of primary drying
the product temperature becomes closer to the critical formu-
and from this point compromises pressure rise measurements.
lation temperature, which is typically defined by the collapse
In the first two-thirds of primary drying that is the primary ap-
temperature for amorphous products. This, in turn, makes rig-
plication for MTM measurements, Ap reduction is not expected
orous process and product monitoring and application of novel
to be a significant factor for the configuration applied, whereas
process analytical technology (PAT) tools in the optimization
water reabsorption leads to falsely reduced pressure rise curves
and production process necessary.1
even in early primary drying.
Manometric temperature measurement (MTM) and the
Different alternative methods to MTM for endpoint detection
SMARTTM freeze-dryer system that is based on product feed-
of primary drying were employed in this study for comparison.
back from pressure rise experiments has already proven to be a
Thin wire thermocouples are the most common device for mon-
reliable tool for development of optimized freeze-drying cycles
itoring of product temperatures in laboratory-scale freeze dry-
within a single laboratory experiment.2–5 However, the data
ers, with better sensitivity for the thinner 36 gauge version if
from MTM technology can be compromised if high-concentrated
the placement in the vial is performed correctly (center-bottom
amorphous substances, particularly proteins, are lyophilized.
position). The thicker 20 gauge thermocouples are physically
Significant differences can be observed between product tem-
more robust, but placement is more difficult and contact to the
peratures at the bottom of the vial calculated from MTM data
ice is commonly lost at an earlier stage. Both thermocouples
and measured with different temperature sensors. This phe-
measure the temperature at the fusion point of two dissimi-
nomenon is likely because of water reabsorption effects of the
lar wires at the tip of the sensor. Temperature remote inter-
amorphous matrix during the pressure rise experiment.6 Re-
rogation system (TEMPRIS) sensors have been introduced as
duction of the product area Ap toward the end of primary drying
a wireless temperature measurement system that is suitable
for freeze dryers from laboratory scale to production, which
Correspondence to: Henning Gieseler (Telephone: +49-931-90705678;
Fax: +49-931-90705679; E-mail: info@gilyos.com) provides superior comparability of temperature data through-
Journal of Pharmaceutical Sciences, Vol. 104, 1741–1749 (2015) out development and transfer. Their operation results in mea-

C 2015 Wiley Periodicals, Inc. and the American Pharmacists Association surement of the temperature at the temperature-sensitive area

Schneid et al., JOURNAL OF PHARMACEUTICAL SCIENCES 104:1741–1749, 2015 1741

1742 RESEARCH ARTICLE – Pharmaceutics, Drug Delivery and Pharmaceutical Technology
at the tip of the sensor that is larger than the sensing area
for the thermocouples, but also located at the center-bottom
of the vial. Because of their impact on the freezing behavior,
all invasive temperature sensors may not be indicative for the
noninstrumented vials especially in a manufacturing environ-
ment. Comparative pressure measurement relies on the gradi-
ent between two pressure sensors with different measurement
methods [Pirani sensor: pressure dependent on composition of
chamber atmosphere; capacitance manometer (CM): absolute
pressure, independent of atmosphere composition]. The Pirani
sensor records the elevated pressure (1.6×) compared with
the CM throughout primary drying, and the pressure drops to
the CM level when no more water is removed from the product.
The purpose of this investigation was to investigate and
compensate the lack of accuracy of MTM measurements dur-
ing primary drying by measuring product temperature data
with two different types of thermocouples and with novel wire-
less TEMPRIS sensors,7 and to use these data for derivative
calculation of critical product parameters. For this purpose, a
high-concentrated amorphous protein model formulation was
freeze-dried at conservative and at aggressive cycle conditions.
Sample vials were removed during primary drying to deter-
mine the residual moisture content remaining in the vials and
evaluate the accuracy of endpoint indications of the different
temperature sensors and process analyzers. The product and
process data provided by the monitoring systems were com-
pared, and recalculation of vapor pressure of ice and product
resistance based on temperature data was performed.
MATERIAL AND METHODS
A product solution containing 50 mg/mL bovine serum albumin
(>96%; Sigma–Aldrich, Munich, Germany), 25 mg/mL sucrose
(>99.5%; Sigma–Aldrich), and 0.107 mg/mL phosphate buffer
(Sigma–Aldrich) was used throughout this investigation. The
solution was adjusted to pH 7.4 with 2.5 N NaOH (Sigma–
Aldrich) prior to the final volume adjustment. Four-hundred
ninety three product vials (2 mL serum tubing; Schott forma
vitrum, Mainz, Germany) that were surrounded by one row of
empty “dummy” vials to mitigate radiation effects were filled
with 1 mL of the solution. Note that 33 vials including stoppers
in the center of the array were marked and accurately weighed
prior to the filling step (Sartorius analytical balance, accuracy
0.01 mg). The exact mass of the dispensed amount of solution
was then determined by reweighing the vials to account for
variations in the filling procedure. These vials were extracted
from the freeze dryer at various sampling points during the
process.
The array of vials was loaded onto the middle shelf of a
Lyostar II freeze dryer with an installed sample extractor door
(SP Scientific, Stone Ridge, New York) and lyophilized. All runs
were performed in the Auto-MTM mode that collects MTM data
periodically (60 min intervals, valve closure time 25 s) based
on a user predefined shelf temperature and chamber pressure
over time profile. Two different process designs were employed
for primary drying: a conservative cycle (low shelf tempera-
ture of −25°C) and an aggressive cycle (high shelf tempera-
ture of +25°C). The cycle steps for both conditions were se-
lected to imitate the lyophilization cycle previously reported by
Johnson et al.8 (Table 1). Aggressive primary drying of protein
formulations, even above the collapse temperature, has been
Schneid et al., JOURNAL OF PHARMACEUTICAL SCIENCES 104:1741–1749, 2015
Table 1. Freeze-Drying Cycle for Both Primary Drying Temperatures
Freezing 1° Drying 2° Drying
Ts (°C) −45 −20 −45 −25/+25 +40
Ramp rate (°C/min) 1 1 1 1 1
Time (min) 120 60 120 Variable 600
Pressure (mTorr) 100 100
described not to impair the protein structure and stability in
several cases.9,10 The transition to secondary drying was per-
formed manually once the end of primary drying was indicated
by comparative pressure measurement (decrease of the Pirani
signal to less than 10 mTorr above the level of the CM).
Three thin wire thermocouples (36 gauge; Omega Engi-
neering, Newport, Connecticut) and three regular thermocou-
ples (20 gauge; Omega Engineering) were placed into prod-
uct vials in edge and center positions. Additionally, four
wireless, battery-free TEMPRIS sensors (IQ Mobil Solutions,
Holzkirchen, Germany) were placed into vials in edge and
center positions. The sensor’s operation principle is excita-
tion by high-frequency radiation and evaluation of the ob-
tained response using the TEMPRIS Data Server software.
The TEMPRIS system has been previously described in the
literature.11,12 All temperature sensors were placed in center-
bottom position in the vial to maintain contact with the last re-
mainder of ice. Data of the different temperature sensors were
recorded to generate an alternative data set to compensate the
increasingly inaccurate MTM data caused by water reabsorp-
tion on the dried matrix containing high-protein contents.
The density of the product solution at room temperature
was determined to be 1.022 g/mL; the total solute concentra-
tion was 76 mg/mL. In combination with the fill weight per
vial measured for the sample vials, the mass of total solids and
the mass of remaining water in the vials removed during sam-
pling could be determined. The sample vials extracted during
drying were first allowed to warm up to room temperature in
a glove box (rH < 5%) and subsequently briefly vented with
dry air. After this step, the vials were weighed, and the mass
of remaining water was calculated. The accuracy of this gravi-
metric procedure is limited by the low mass of solids in the vial
(75 mg cake per vial) and the corresponding minimal mass
of water within the cake (1% moisture content is equivalent
to 0.76 mg). Therefore, samples of each run were additionally
analyzed using coulometric Karl Fischer titration in a water
vaporizer (Mitsubishi CA-06 with VA-06) to obtain supportive
data for residual moisture contents below 10%. The solid sam-
ple (50–70 mg) was transferred into the vaporizer unit where
the sample was heated to 140°C. Dry nitrogen gas (Linde, Pul-
lach, Germany) was used as a carrier to transfer the moisture
into the measurement cell. The endpoint indications of the var-
ious process analyzers were correlated to the moisture content
at the respective sampling point.
RESULTS AND DISCUSSION
Evaluation of Product Temperature Data
Comparison of process data obtained from one freeze-drying
cycle at aggressive and one at conservative conditions revealed
that the primary drying time (as indicated by the Pirani sig-
nal decrease to within 10 mTorr of the CM value) was reduced
DOI 10.1002/jps.24389
 RESEARCH ARTICLE – Pharmaceutics, Drug Delivery and Pharmaceutical Technology
Figure 1. Temperature data recorded during primary drying of a
representative run with +25°C primary drying temperature (a) and of
a representative run with −25°C primary drying temperature (b).
from approximately 45 h for the conservative conditions to less
than 10 h in the aggressive runs. Elevation of shelf tempera-
ture by 50°C resulted in an increase of product temperature of
approximately 10°C (−30°C compared with −20°C) as indicated
by thermocouples and TEMPRIS sensors. The product cakes
obtained at −25°C and at +25°C, however, showed comparable
and acceptable appearance. Only marginally higher shrinkage
was found for some of the aggressively dried cakes.
The temperature profiles recorded by the temperature sen-
sors during primary drying at aggressive and conservative con-
ditions are shown in Figure 1. As expected, temperature sensors
in edge vials consistently indicated the end of primary drying
earlier than sensors in center vials for all types of probes used.
Both types of thermocouples placed in center vials recorded
similar product temperatures during the steady state of pri-
mary drying, but the thicker 20 gauge TC’s lost contact to the
ice much earlier and increased over a broader time interval.
The TEMPRIS sensors in center vials showed systematically
about 1°C lower product temperatures than the thermocouples
for the conservative drying conditions, and the effects of atyp-
ical radiation on TEMPRIS sensors in edge vials were slightly
lower than for thermocouples, which is consistent with previous
observations.7 A detailed investigation of the endpoint indica-
tion and correlation to residual moisture contents is provided
later in the text.
DOI 10.1002/jps.24389
1743
Figure 2. Manometric temperature measurement product tempera-
ture at the sublimation front (Tp ) and at the vial bottom (Tb ) compared
for two runs with conservative and aggressive drying conditions.
The product temperature at the sublimation interface calcu-
lated from MTM measurements in early primary drying was
approximately 10°C higher for the aggressive conditions than
for the conservative conditions (Fig. 2). It is important to note
that because of the substantially reduced primary drying time
in the aggressive cycle, the number of reliable MTM measure-
ments (i.e., before 2/3 of the mass of ice has been removed) was
much lower than in the conservative process. Here, only four
reliable measurements compared with about 30 for the conser-
vative cycle conditions were obtained; the fifth MTM measure-
ment for aggressive conditions already showed a greatly re-
duced pressure increase. Surprisingly, the agreement of MTM
temperature data with thermocouples was much better in the
runs with +25°C primary drying temperature, especially after
the first 20% of primary drying had been completed (equivalent
to 7 h for conservative and 1.5 h for aggressive conditions; cf.
Fig. 1). The MTM calculation at conservative conditions started
to become inaccurate after only few hours, which was apparent
in continuously decreasing Tp and Tb values without further
temperature increase correlated to growing dry layer thickness
as observed for the different temperature sensors (Fig. 1b). In
contrast, the product temperature data from MTM and tem-
perature sensors in the +25°C runs were in good agreement
during the first 4 h of primary drying, even after the first tem-
perature sensors had already lost contact to the ice (Fig. 1a).
After approximately 5 h of primary drying at +25°C, the MTM
data also started to become inaccurate. This is most likely be-
cause of reduction of the product area Ap that is manifested in
partial freedom of ice in the cake near the vial walls, and vials
in edge position becoming free of ice.5 Note that Ap may also
initially increase in early primary drying as the ice is removed
faster at the sides of the vial; this effect requires further inves-
tigation. The temperature gradient calculated for the product
at the bottom of the vial and the product sublimation interface
using the MTM algorithm3 was much higher for the aggressive
drying conditions (cf. Fig. 2) and decreased with the progressing
reduction in ice layer thickness, Lice .
The likely reason for the better accuracy of MTM data at ag-
gressive drying conditions observed for this formulation is the
higher absolute pressure increase and the corresponding re-
duced influence of water reabsorption during the pressure rise
Schneid et al., JOURNAL OF PHARMACEUTICAL SCIENCES 104:1741–1749, 2015
1744 RESEARCH ARTICLE – Pharmaceutics, Drug Delivery and Pharmaceutical Technology
Figure 3. Manometric temperature measurement vapor pressure of
ice at the sublimation interface (Pice ) compared for two runs with con-
servative and aggressive drying conditions.
experiment. The water vapor coming from the sublimation front
is partially bound and immobilized in the already dried product
matrix, so that the amount of vapor reaching the chamber is
reduced. As the dried cake fraction increases during primary
drying, the amount of water vapor that can potentially be ab-
sorbed may increase. This temperature-dependent difference
is observable in the vapor pressure of ice at the sublimation
front (Pice ) data at both conditions (Fig. 3). As described pre-
viously, the chamber pressure was controlled at 100 mTorr in
all runs. All derivative calculations by the MTM algorithm are
based on Pice and Rp that are fitted from the pressure rise data,
and the difference between Pice and chamber pressure (Pc ). The
water reabsorption phenomenon leads to a falsely reduced Pice
value as not all water vapor that is released from the subli-
mation front reaches the chamber atmosphere but is bound by
the amorphous product matrix, which results in errors in the
following calculations. As the total pressure rise and the fitted
Pice is much higher for the aggressive drying conditions (Fig. 3),
the error induced by water reabsorption is relatively lower,
and the product temperature calculation is more accurate
in the +25°C runs. Another important factor is the expo-
nential relationship between product temperature and vapor
pressure of ice, which leads to much higher Pice values at
high-product temperature and relatively lower reduction be-
cause of water reabsorption. In addition, the absorption of
water into the amorphous protein/sugar matrix is also ex-
pected to be temperature-dependent, with stronger effects at
low-product temperatures. Apparently, the water reabsorption
phenomenon occurs to a higher degree for proteins than for
disaccharides, as previous investigations for similar formula-
tions containing lower protein fractions and higher amounts
of sucrose did not observe a comparable dependence on pro-
cess conditions.8 The reabsorption of water at the conditions
encountered during freeze-drying is not identical to the sorp-
tion properties of the pure solids at room temperature. In addi-
tion, the moisture level within the dried cake prior to the start
of the pressure rise test may differ between the protein and
the sugar, leading to different capacities for water reabsorp-
tion. These characteristics may be investigated further in the
future.
Schneid et al., JOURNAL OF PHARMACEUTICAL SCIENCES 104:1741–1749, 2015
Figure 4. Manometric temperature measurement product resistance
compared for two runs with conservative and aggressive drying condi-
tions.
Evaluation of Product Resistance Data
The product resistance of the dried layer, Rp , has been estab-
lished as an additional critical product parameter for freeze-
drying.13–15 Rp has been shown to indicate microcollapse within
the cake structure at aggressive primary drying conditions as
well as effects of annealing and different freezing conditions.
If the very aggressive primary drying conditions used in this
study would have led to changes in the inner structure of the
cake, the Rp data would be expected to show a drop or pro-
ceed at a lower level.4,16 However, high-concentrated protein
formulations are often surprisingly robust against collapse at
aggressive freeze-drying conditions.17
In the current study, Rp data determined by MTM were
very comparable in all runs performed and showed no mas-
sive depression, which would, in turn, indicate collapse
(Fig. 4). Although the formulation contained significant
amounts of sucrose that by itself shows a collapse tempera-
ture of approximately −32°C, the high collapse temperature of
the protein component preserved the overall cake integrity and
prevented coordinated viscous flow that is linked to collapse of
the cake structure.17 Although the MTM Rp calculation may
be flawed because of water reabsorption during the pressure
rise test that could impair the curve fit, the comparable cake
appearance for product of all runs and the lack of shrinkage
supports the conclusions drawn from the MTM Rp data.
Endpoint Indications and Residual Moisture Levels
As described in the Material and Methods section, vials in cen-
ter position were removed from the chamber during the sec-
ond half of primary drying using a sample extractor. Typi-
cally, three weighed vials per sampling point were removed,
ventilated, and weighed. Three additional vials per sampling
point were removed for Karl Fischer measurements. Sampling
was conducted at five to nine time points per run (six cy-
cles in total, three per condition). The preweighed vials re-
maining within the freeze dryer were weighed after secondary
drying and compared with Karl Fischer results to validate
the weighing method. As the primary drying step was much
longer for the conservative conditions, samples were taken less
DOI 10.1002/jps.24389
 RESEARCH ARTICLE – Pharmaceutics, Drug Delivery and Pharmaceutical Technology 1745

Table 2. Residual Moisture Content (mg Water Per Vial) at Onset,

Midpoint, and Offset for Thick Wire Thermocouples (20 Gauge), Thin
Wire Thermocouples (36 Gauge), TEMPRIS Probes, and Comparative
Pressure Measurements at +25°C and −25°C Primary Drying
Temperature

Average Average Average

+25°C Run TC 20 g TC 36 g TEMPRIS Pirani

RM onset 319 156 202 6

(mg/vial)
RM midpoint 80 92 71 4
(mg/vial)
RM offset 4 5 4 2
(mg/vial)

Average TC Average TC Average

− 25°C run 20 g 36 g TEMPRIS Pirani

RM onset 330 218 192 48

(mg/vial)
RM midpoint 199 113 112 8
(mg/vial)
RM offset 56 15 13 3
(mg/vial)

sible to determine the moisture content within about ±0.3% to

Figure 5. Residual Moisture content of sampled vials determined
by weighing and by Karl Fischer titration for a representative run at
+25°C primary drying temperature (a) and for a representative run at
−25°C primary drying temperature (b).
frequently than in the aggressive runs. The residual moisture
contents (RM) determined by weighing and by Karl Fischer
measurements for one representative run at +25°C and one
run at −25°C are shown in Figure 5.
Although the vials removed during the first two sampling
points, especially for the conservative cycle, showed high vari-
ation in the mass of water remaining in the product, the dis-
tribution became more homogeneous from the third sampling
point onwards. For the aggressive drying conditions, a very fast
drop in the RM can be observed with only little additional re-
moval of water after about 8 h of primary drying. The average
RM in the samples was less than 1% after only 8.3 h of pri-
mary drying, and the samples at the endpoint showed a RM of
0.13% measured with the Karl Fischer method. For the conser-
vative drying conditions, the residual moisture in late primary
drying was much higher, on average 10% after 38 h, and still
3%–5% after 46 h directly before the start of secondary drying.
The residual moisture content of the product after completion
of secondary drying was comparable for both primary drying
conditions, on average 0.1%. The weighing method proved to be
reliable within the measurement accuracy. As 1% of residual
moisture is equivalent to only 0.76 mg of water, it is only pos-
±0.5% by the weighing method applied, which was consistently
achieved with the weighing procedure used in this study.
To determine the moisture level that was prevalent at on-
set, midpoint, and offset of the product temperature increase of
the thermal probes’ endpoint indication, these time points were
calculated for each sensor using the method suggested recently
by Patel.18 Afterwards, the average times (n = 3) for onset,
midpoint, and offset were averaged for temperature sensors of
each type (two in center and one in edge position; for TEMPRIS,
data for the second sensor in edge position was omitted), and
the residual moisture at these points was interpolated from
the residual moisture curve (Fig. 5). The endpoint times for the
different temperature sensors were compared with those for
comparative pressure measurement. The results of all sensors
at both drying conditions are shown in Table 2 and in Figure 6.
The MTM Pice values that can generally be used as endpoint de-
tection were also calculated. However, it can be seen in Figure 3
that it is difficult to determine an onset or midpoint for the for-
mulation used in these experiments. Therefore, the MTM Pice
measurements were not used for the endpoint comparison.
The remaining mass of water in the vial for the different
endpoint indications (onset, midpoint, offset) was calculated
for the temperature sensors and for comparative pressure mea-
surements. The onset, midpoint, and offset time for each sensor
used were determined, and the average time was calculated for
each type of temperature sensor. The moisture level was cal-
culated based on the fitting curves shown in Figure 5. As a
distinct change of the rate of water removal is apparent in the
moisture of samples removed from the freeze dryer, a transi-
tion time point was defined, which marks the change from rapid
sublimation of ice to predominantly slower desorption from the
dried matrix. The water content values before this time point
were calculated based on the almost linearly decreasing curve,
whereas the water contents at later process times were calcu-
lated according to the slower desorption curve. Although this
procedure based on the definition of a transition time is only
an approximation, the results were overall in good agreement

DOI 10.1002/jps.24389 Schneid et al., JOURNAL OF PHARMACEUTICAL SCIENCES 104:1741–1749, 2015

1746 RESEARCH ARTICLE – Pharmaceutics, Drug Delivery and Pharmaceutical Technology

Comparison of MTM and Gravimetric Data

A direct comparison between the RM at the sampling points
determined by weighing and the residual moisture calculated
from MTM measurements revealed that MTM severely under-
estimated the mass of water removed from the cake for both
conservative and aggressive primary drying conditions. It is im-
portant to note that the MTM data shown here were acquired
using the Auto-MTM mode of the SMARTTM freeze dryer, which
is not applied during conventional optimization of freeze-drying
cycles, and records of MTM data during a preprogrammed pro-
cess. Therefore, MTM measurements were not terminated as
recommended once two-third of the total mass of water had
been removed, and the following measurements (after 11 h
process time for aggressive and 30 h process time for conser-
vative conditions) are not accurate and representative.
Figure 7 illustrates the integrated MTM mass during pri-
mary drying in comparison to the gravimetric results of re-
moved samples. Although the stagnation of MTM mass flow in
the late part of primary drying is because of the inaccuracies
in late primary drying associated with the method, significant
underestimation of mass flow is also present in the earlier stage
of primary drying. The MTM mass flow rate is calculated ac-
cording to Eq. (1) from the difference between vapor pressure
Figure 6. Residual moisture content at onset, midpoint, and offset
of various temperature sensors and of comparative pressure measure-
ments for both conservative and aggressive cycle conditions.

with the moisture content of sample vials and successfully rep-

resented the characteristics of both drying profiles.
Figure 6 and Table 2 provide an overview of the results for
the endpoint comparison. As the mass of water remaining in
the vial far exceeded the mass of the cake (76 mg dry prod-
uct per vial), the residual moisture content in the scope of the
endpoint indication comparison is expressed in milligram of
water per vial to avoid misinterpretation of results. The re-
sults clearly show that all temperature sensors started to lose
contact with the ice long before the end of primary drying.
The moisture levels for each sensor type were comparable for
both conditions, with higher water content for the 20 gauge
thermocouples (320 mg), and comparable levels for thin wire
thermocouples and TEMPRIS sensors (200 mg). In contrast,
the decrease of the Pirani signal (onset) only began at a mois-
ture level of 48 mg water for the conservative and of even
6 mg for the aggressive conditions. At the offset endpoint in-
dication (sensor reading equal to shelf temperature), the mois-
ture content was very uniform at a low level for the aggres-
sive drying conditions for all temperature sensor types. In
contrast, the moisture content was still comparably high for
the conservative drying conditions, especially for the thick wire
thermocouple type (56 mg moisture per vial compared with 15
and 13 mg for thin wire thermocouples and TEMPRIS sen-
sors, respectively). The comparative pressure endpoint indica-
tion consistently correlated to a much lower moisture level,
with remaining water of less than 10 mg at the midpoint
for both drying conditions. This observation supports the su-
periority of comparative pressure measurements for drying
endpoint indication compared with temperature sensors. In Figure 7. Comparison of the amount of water remaining in the prod-
addition, noninvasive measurement of the chamber composi- uct calculated by MTM and from weighing of sampled vials (gravi-
tion is not affected by differences in drying properties because metric) for a representative run at +25°C primary drying temperature
of the sensor itself, as is the case for invasive temperature (a) and for a representative run at −25°C primary drying temperature
probes. (b).

Schneid et al., JOURNAL OF PHARMACEUTICAL SCIENCES 104:1741–1749, 2015 DOI 10.1002/jps.24389

 RESEARCH ARTICLE – Pharmaceutics, Drug Delivery and Pharmaceutical Technology 1747

of ice (Pice ) and chamber pressure (Pc ), the product area (Ap ),
and the sum of resistances of product (Rp ) and stopper (Rs )3,19 :

(Pice − Pc )Ap
dm/dt = (1)
(Rp + Rs )

On the basis of Eq. (1), the observed error could theoretically

be caused by falsely reduced Pice values, erroneously elevated
Rp results, changes in the product area (i.e., area in which ice
remains), or by a combination of these factors. Changes of Ap
can mostly be regarded as negligible in the first half of primary
drying in the setup used, as the ice is primarily removed from
the top to the bottom of the cake (height 6.5 mm) within hexag-
onally packed vials with small diameter (14 mm), leading to a
practically constant product area in each vial in early primary
drying. The product resistance was quite similar for both pri-
mary drying conditions and showed a plateau behavior, with
a minor decrease toward the end for the conservative drying
conditions. In contrast, Pice is potentially strongly influenced
by water reabsorption in the dried layer during the pressure
rise measurement. To further delineate the influence of the
individual factors on the observed problems with mass flow in-
tegration and to establish a basis for correction of flawed MTM
data, the gravimetric data and measurements of temperature
probes were combined to recalculate critical product attributes
for comparison.

Derivative Calculation of Critical Product Parameters: Vapor

Pressure of Ice
The vapor pressure of ice at the sublimation interface (Pice ) was
calculated at the respective time of sampling based on the av-
erage product temperature of the center-positioned thin wire
thermocouples and TEMPRIS probes before the sensors lost
contact to the ice. Data from the thicker 20 gauge thermocou-
ples was omitted because of the elevated product temperatures
during primary drying and the earlier loss of contact with ice.
To allow a valid comparison with MTM data, the slightly lower
product temperature at the sublimation interface, Tp , was cal-
culated from the directly measured temperature at the vial
bottom, Tb . The difference between Tp and Tb is the temper-
ature gradient over the ice layer (T), which was calculated
according to Tang et al.5 using Eq. (2), and subtracted from the
recorded temperature to obtain the temperature at the subli-
mation interface. Note that no MTM data were used to cal-
culate the temperature gradient in order to obtain completely
autonomous data for comparison.
 
24.7Lice (Pice − Pc )/(Rp + Rs ) − 0.0102Lice (Ts − Tb )
T =
1 − 0.0102Lice
The remaining ice layer thickness, Lice , was calculated from the
gravimetric data, and the shelf temperature, Ts , was introduced
from the data recorded by the freeze dryer. Pice −Pc signifies the
difference between ice vapor pressure and chamber pressure.
The resulting product temperature at the sublimation interface
Tp was introduced into Eq. (3) to calculate the vapor pressure
of ice for comparison to MTM measurements.
Figure 8. Comparison of the vapor pressure of ice at the sublimation
front calculated by MTM and from weighing of sampled vials (gravi-
metric) for a representative run at +25°C primary drying temperature
(a) and for a representative run at −25°C primary drying temperature
(b).
The course of the vapor pressure of ice calculated by MTM and
from temperature sensors at both aggressive and conservative
conditions is shown in Figure 8. It is observable that Pice de-
termined using MTM proceeds at a lower level and also starts
to decrease early. The Pice values calculated from temperature
data of thin wire thermocouples and TEMPRIS sensors in cen-
ter vials were initially in good agreement with the MTM data,
but continuously increased throughout primary drying that is
usually related to elevation of Rp with growing dry layer thick-
ness. Toward the end of primary drying, the temperature sen-
(2) sors lost contact to the ice, resulting in nonrepresentative Pice
values. The discrepancy between the Pice values derived from
temperature sensors and MTM data supports the hypothesis
that Pice calculated by MTM is depressed because of water re-
absorption effects during the measurement, and thereby causes
reduced integrated mass flow and underprediction of product
temperature.
Derivative Calculation of Critical Product Parameters: Product
Resistance

−6144.96 +24.01849
The second critical product parameter, product resistance, was
Pice = e Tp
(3) calculated from a combination of Pice derived from product

DOI 10.1002/jps.24389 Schneid et al., JOURNAL OF PHARMACEUTICAL SCIENCES 104:1741–1749, 2015

1748 RESEARCH ARTICLE – Pharmaceutics, Drug Delivery and Pharmaceutical Technology
Figure 9. Comparison of the product resistance calculated by MTM
and from weighing of sampled vials (gravimetric) for a representative
run at +25°C primary drying temperature (a) and for a representative
run at −25°C primary drying temperature (b).
temperature measurements in center vials, and dm/dt deter-
mined gravimetrically to allow comparison to MTM Rp data
using Eq. (4).4 :
Ap (Pice − Pc )
(Rp + Rs ) = (4)
dm/dt
Although both this calculation method and the MTM algo-
rithm determine the sum of “Rp + Rs ” (“product resistance”
imposed by the dry layer, and “stopper resistance” imposed by
the flow path through the stopper opening) simultaneously in-
stead of only Rp , the Rs term is much smaller (on the order of
0.1 Torr cm2 h/g), and can be neglected for the following con-
siderations. A comparison of the Rp data generated with the
different methods for runs at aggressive and conservative con-
ditions is shown in Figure 9. For both drying conditions, the
gravimetrically calculated Rp values were in good agreement
with MTM data for the first 20%–30% of primary drying. Dur-
ing this time, the product resistance proceeds at comparable
levels for aggressive and conservative primary drying. During
the later part of primary drying, a significant increase to more
Schneid et al., JOURNAL OF PHARMACEUTICAL SCIENCES 104:1741–1749, 2015
than 10 Torr cm2 h/g was observed, especially in the Rp data
for conservative drying.
It is not clear whether this behavior represents a real in-
crease of product resistance with dry layer thickness (which
cannot be observed in the MTM procedure), or whether the
calculation based on weighing of few sampled vials and tem-
perature monitoring of three selected vials per sensor type is
simply prone to errors in the late part of primary drying. The
calculation is strongly influenced by slight changes in product
temperature as this term is represented logarithmically in the
Pice calculation that is then used for determination of Rp . Ad-
ditionally, significant fluctuations in the dm/dt values are gen-
erated by sampling vials from different positions in the array,
all showing different individual drying behavior. If the large
elevation of Rp for conservative primary drying compared with
aggressive conditions indicated by this approach was correct,
significant errors would be present in the MTM data for the
conservative runs, leading to erroneous assumptions of compa-
rable cake structure. Most likely, the results of both approaches
are biased to some extent, with an underprediction for MTM
because of water reabsorption, and an overprediction for the
product temperature sensor-based method. The cake appear-
ance supports the conclusions drawn from the MTM data as no
significant differences could be found between cakes processed
at the two drying conditions.
Overall, it can be assumed that the data sets of Pice and
Rp are interrelated for each calculation method, and partially
compensate their respective individual effects on the mass flow
rate. For instance, a hypothetical twofold increase of Pice with
unchanged Rp would lead to a much higher mass flow rate and
an overprediction of mass flow rate to a higher extent than the
40% observed in MTM data. In contrast, calculation of dm/dt
based on Rp calculated from a gravimetric approach in com-
bination with MTM Pice data would result in a reduction of
total mass flow by approximately 50%, leading to an integrated
value of only 30% of the originally present water. This model
calculation shows that at least for conservative conditions, it is
not possible to use isolated data sets (i.e., Pice or Rp ) from dif-
ferent sources for calculation of a corrected mass flow profile.
Although the variation is smaller for the aggressive primary
drying conditions, the basic concerns against mixing of data
from different process analyzers for calculation of derivative
product parameters apply identically.
CONCLUSIONS
The freeze-drying process of a high-concentrated protein formu-
lation was monitored using complimentary process analyzers
for two alternative process designs. With regard to detection
of the endpoint of primary drying, comparative pressure mea-
surements showed the lowest residual moisture content at the
onset endpoint indication, and also the lowest value at the offset
indication. All temperature sensors lost contact to the ice much
earlier relative to the drying endpoint indicated by comparative
pressure measurements. Especially for the thicker 20 gauge
thermocouples, the endpoint detection was found to be poorly
representative. The TEMPRIS sensors appear preferable over
the thermocouples because of their wireless operation and load-
ing. The selected shelf temperature had substantial effects on
the endpoint indication of the different process monitors and
the respective residual moisture level. The MTM algorithm
DOI 10.1002/jps.24389
 RESEARCH ARTICLE – Pharmaceutics, Drug Delivery and Pharmaceutical Technology
underestimated the mass flow rate significantly throughout
primary drying, which is expected to be associated with water
reabsorption during pressure rise tests. Both Pice and Rp values
calculated from gravimetric data in combination with temper-
ature probe data were significantly higher than the data from
MTM. The difference was especially high for conservative con-
ditions and mitigated for aggressive conditions. Compensation
of biased MTM data by using a combination of critical process
parameters provided by MTM and temperature sensors was
generally possible, but very sensitive to minor measurement
inaccuracies. Further studies are required to develop a reliable
method for improvement of MTM data for high-concentrated
amorphous systems at different process conditions.
REFERENCES
1. Food and Drug Administration. 2004. Guidance for industry: PAT—A
framework for innovative pharmaceutical development, manufacturing
and quality assurance.
2. Gieseler H, Kramer T, Pikal MJ. 2007. Use of manometric tem-
perature measurement (MTM) and SMART freeze dryer technology
for development of an optimized freeze-drying cycle. J Pharm Sci
96(12):3402–3418.
3. Tang X, Nail SL, Pikal MJ. 2006. Evaluation of manometric temper-
ature measurement, a process analytical technology tool for freeze-
drying: Part I, product temperature measurement. AAPS Pharm-
SciTech 7(1):E14.
4. Tang XC, Nail SL, Pikal MJ. 2006. Evaluation of manometric tem-
perature measurement, a process analytical technology tool for freeze-
drying: Part II measurement of dry-layer resistance. AAPS Pharm-
SciTech 7(4):93.
5. Tang XC, Nail SL, Pikal MJ. 2005. Freeze-drying process design
by manometric temperature measurement: Design of a smart freeze-
dryer. Pharm Res 22(4):685–700.
6. Gieseler H, Mather L, Pikal MJ. 2006. SMART Freeze-Dryer tech-
nology: Impact of water re-adsorption of polyvinylpyrrolidone (PVP) on
Manometric Temperature Measurements (MTM) during primary dry-
ing. Proc. 5th World meeting on pharmaceutics and pharmaceutical
technology. Geneva, Switzerland.
7. Schneid S, Gieseler H. 2008. Evaluation of a new wireless tem-
perature remote interrogation system (TEMPRIS) to measure prod-
uct temperature during freeze drying. AAPS PharmSciTech 9(3):729–
739.
DOI 10.1002/jps.24389
1749
8. Johnson RE, Oldroyd ME, Ahmed SS, Gieseler H, Lewis LM. 2010.
Use of manometric temperature measurements (MTM) to characterize
the freeze-drying behavior of amorphous protein formulations. J Pharm
Sci 99(6):2863–2873.
9. Schersch K, Betz O, Garidel P, Muehlau S, Bassarab S, Winter G.
2010. Systematic investigation of the effect of lyophilizate collapse on
pharmaceutically relevant proteins I: Stability after freeze-drying. J
Pharm Sci 99(5):2256–2278.
10. Schersch K, Betz O, Garidel P, Muehlau S, Bassarab S, Winter G.
2012. Systematic investigation of the effect of lyophilizate collapse on
pharmaceutically relevant proteins, part 2: Stability during storage at
elevated temperatures. J Pharm Sci 101(7):2288–2306.
11. Schneid S, Gieseler H. 2008. Evaluation of a new wireless tem-
perature remote interrogation system (TEMPRIS) to measure product
temperature during freeze drying. AAPS PharmSciTech 9:729–739.
12. Schneid S, Gieseler H. 2008. A new generation of battery-free wire-
less temperature probes as an alternative to thermocouples for vial
freeze drying. Proc. CPPR freeze drying of pharmaceuticals and biolog-
icals conference. Breckenridge, Colorado.
13. Overcashier DE, Patapoff TW, Hsu CC. 1999. Lyophilization of pro-
tein formulations in vials: Investigation of the relationship between re-
sistance to vapor flow during primary drying and small-scale product
collapse. J Pharm Sci 88(7):688–695.
14. Kuu WY, Hardwick LM, Akers MJ. 2006. Rapid determination of
dry layer mass transfer resistance for various pharmaceutical formu-
lations during primary drying using product temperature profiles. Int
J Pharm 313(1–2):99–113.
15. Schneid S, Gieseler H. 2010. PAT in freeze drying: Monitoring of
product resistance using non-invasive NIR-spectroscopic TDLAS mea-
surements Proc. 7th World meeting on pharmaceutics, biopharmaceu-
tics and pharmaceutical technology. Valletta, Malta.
16. Schneid S, Gieseler H. 2008. Effect of concentration, vial size and
fill depth on product resistance of sucrose solutions during freeze dry-
ing. Proc. 6th World meeting on pharmaceutics, biopharmaceutics and
pharmaceutical technology. Barcelona, Spain.
17. Meister E, Gieseler H. 2008. Freeze-dry microscopy of protein/sugar
mixtures: Drying behavior, interpretation of collapse temperatures and
a comparison to corresponding glass transition Data. J Pharm Sci
98:3072–3087.
18. Patel SM. 2011. PAT for freeze-drying: Determination of end point
of primary drying. SP Scientific Webinar. Accessed on October 20, 2011,
at: www.spscientific.com/Webinars/Archives/.
19. Tang XC, Nail SL, Pikal MJ. 2006. Evaluation of manometric tem-
perature measurement (MTM), a process analytical technology tool in
freeze drying, part III: Heat and mass transfer measurement. AAPS
PharmSciTech 7(4):97.
Schneid et al., JOURNAL OF PHARMACEUTICAL SCIENCES 104:1741–1749, 2015