Use of Manometric Temperature Measurement (MTM)

and SMARTTM Freeze Dryer Technology for Development

of an Optimized Freeze-Drying Cycle
HENNING GIESELER,1,3 TONY KRAMER,2 MICHAEL J. PIKAL1
1
School of Pharmacy, University of Connecticut, 69 North Eagleville Road, Storrs, Connecticut 06269
2
Parenteral Center of Emphasis (PCoE), Pfizer Inc., Eastern Point Road, Groton, Connecticut 06340
3
Department of Pharmaceutics, University of Erlangen, Cauerstrasse 4, Erlangen, Germany

Received 31 October 2006; revised 3 February 2007; accepted 17 February 2007

Published online in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/jps.20982

ABSTRACT: This report provides, for the first time, a summary of experiments using
SMARTTM Freeze Dryer technology during a 9 month testing period. A minimum ice
sublimation area of about 300 cm2 for the laboratory freeze dryer, with a chamber
volume 107.5 L, was found consistent with data obtained during previous experiments
with a smaller freeze dryer (52 L). Good reproducibility was found for cycle design with
different type of excipients, formulations, and vials used. SMARTTM primary drying end
point estimates were accurate in the majority of the experiments, but showed an over
prediction of primary cycle time when the product did not fully achieve steady state
conditions before the first MTM measurement was performed. Product resistance data
for 5% sucrose mixtures at varying fill depths were very reproducible. Product tem-
perature determined by SMARTTM was typically in good agreement with thermocouple
data through about 50% of primary drying time, with significant deviations occurring
near the end of primary drying, as expected, but showing a bias much earlier in primary
drying for high solid content formulations (16.6% Pfizer product) and polyvinylpyrro-
lidone (40 kDa) likely due to water ‘‘re-adsorption’’ by the amorphous product during the
MTM test. ß 2007 Wiley-Liss, Inc. and the American Pharmacists Association J Pharm Sci
96:3402–3418, 2007
Keywords: freeze drying/lyophilization; formulation; mathematical model; proces-
sing; proteins

INTRODUCTION ingredient (API) by initially freezing the solvent

(usually water) from the solution, followed by
Lyophilization evolved as a well established removal of ice through sublimation. However, it is
drying technology over the last few decades, but still commonly assumed that freeze drying cycles
many scientific principles of process design and can only be developed by a trial and error
control were not developed in the literature until procedure (i.e. testing the finished product). In
recent years. Freeze drying is still the method of many cases, process development consists of
choice to stabilize an active pharmaceutical simply taking an already ‘‘established’’ cycle in
the company that has worked well over the years
for a different product.
Correspondence to: Henning Gieseler (Telephone: þ49- An important objective of freeze drying process
9131-8529556; Fax: þ49-9131-8529545;
E-mail: [email protected])
design is the development of a cycle that is robust,
Journal of Pharmaceutical Sciences, Vol. 96, 3402–3418 (2007)
economical and can be easily transferred to all
ß 2007 Wiley-Liss, Inc. and the American Pharmacists Association freeze dryers irrespective of size and design.1 To

3402 JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 96, NO. 12, DECEMBER 2007
 FREEZE-DRYING CYCLE OPTIMIZATION 3403

be robust and economical, the product tempera- about the critical parameters and optimal process
ture during primary drying needs to be controlled, settings during a freeze drying run. Computer
which requires control of the heat input. Since modeling of the freeze drying process was reported
primary drying usually consumes the largest in the literature to facilitate this procedure.
fraction of the freeze drying cycle time, high However, this procedure uses theoretical models
product temperatures are favored to achieve high that require experimental input parameters to
sublimation rates which, in turn, lead to shorter delineate the drying behavior, and the model still
cycle time.2,3 However, to successfully develop an needs to be ‘‘verified’’ by a few laboratory
optimized freeze drying cycle, knowledge of the experiments.8 Finally, highly skilled manpower
‘‘critical’’ temperature of the formulation is is necessary to evaluate systematically all critical
necessary, which refers to the eutectic tempera- parameters for processing a given product, or even
ture (Teut) for crystalline and the collapse to use complex computer modeling software.
temperature (Tcol) or glass transition temperature Recently, the concept of a ‘‘smart’’ freeze dryer
(Tg’) for amorphous systems.4,5 The eutectic was introduced as a commercially available soft-
temperature is usually several degrees higher ware product, operating on a modified laboratory
than the collapse temperature and, in general, a scale freeze dryer (FTS LyostarTM II and
crystalline formulation is easier and faster to SMARTTM Freeze Dryer, FTS Systems, Stone
freeze dry.1 However, many of API’s used in freeze Ridge, NY). Two main advantages of this ‘‘expert
drying are proteins or peptides, which remain system concept’’ were reported: first, even users
amorphous during the process. To maintain the who have little or no knowledge about the freeze-
complex protein structure and minimize mobility, drying process will be able to develop an optimized
thereby avoiding significant degradation, stabi- freeze-drying run with only a single freeze-drying
lizers must be added to the formulation. Stabi- experiment. Second, this technology will instan-
lizers need to be in the same phase as the drug to taneously provide the user with data on heat and
interact with the drug and therefore to maintain mass transfer properties during this run which
the API activity during the process and during are usually determined by time consuming
storage.6 The amount of amorphous component is procedures.9 Such data might be important when
critical for process design since the collapse a cycle is scaled from the laboratory to pilot or
temperature of the amorphous phase is the upper production. Moreover, such a technology may
boundary for the product temperature. To avoid provide critical freeze-drying parameters which
collapse of the freeze-dried portion of the sample, might be used as a basis for a better under
the product temperature at the sublimation standing of the freeze drying process and therefore
interface (Tp) must remain below the collapse would address the FDA initiative to encourage
temperature, Tc. Negative effects associated with process analytical technologies.10
structural collapse of the product have been SMARTTM Freeze Dryer is based on mano-
reported recently and may be summarized as metric temperature measurements (MTM) which
follows: (a) an unacceptable appearance of the is a procedure to measure product temperature at
cake, (b) longer reconstitution times, (c) higher the sublimation interface by quickly (i.e. less than
residual moisture content of the final product, and a second) isolating the chamber from the con-
in some cases (d) increased degradation due to denser for a short period of time (i.e., about 25 s).
higher mobility in the system during collapse.4 The resulting pressure rise data are analyzed by
The design of a freeze drying cycle is (often) an algorithm which finally leads to direct mea-
difficult, even when Tc is known. The product surement of both the pressure at the sublimation
temperature (Tp) should be controlled close to but interface ( Pice) and the mass transfer resistance
not above Tc during the primary drying phase. Tp (Rp).7,9 Based on Pice and Rp, data for the product
depends on many factors, such as chamber temperature (Tp), the heat transfer into the
pressure, shelf temperature, resistance of the product (dQ/dt), the actual ice thickness (Lice),
dried product and heat transfer coefficient of the the temperature at the bottom of the vial (Tb), the
container.7 In addition, the product dry layer vial heat transfer coefficient (Kv) and sublimation
resistance (typically) increases over time, which rate (dm/dt) are instantaneously calculated by the
may lead to a significant increase of product software using additional algorithms based on
temperature at the end of primary drying.8 basic steady state heat and mass transfer theory.
Traditionally, numerous experiments in the Thus, SMARTTM Freeze Dryer technology uses
laboratory were necessary to gain information product ‘‘feed back’’ information to maintain Tp

DOI 10.1002/jps JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 96, NO. 12, DECEMBER 2007
3404 GIESELER, KRAMER, AND PIKAL

below a (predefined) ‘‘critical temperature’’ of the addresses the capability of SMARTTM to develop
product but high enough to reduce cycle time to a an ‘‘optimized cycle’’, i.e. the capability to keep the
minimum. To do so, the system ‘‘balances’’ the product temperature within a target temperature,
appropriate amount of heat by adjusting shelf the agreement of thermocouple and SMARTTM
temperature and chamber pressure. A modifica- calculated temperature data, and the reliability of
tion of the MTM procedure to optimize the the primary drying endpoint prediction.
secondary drying phase by determining the water
desorption rate during secondary drying was also
reported in the literature. For detailed informa-
tion about the MTM procedure and the SMARTTM MATERIALS AND METHODS
Freeze Dryer concept, the reader is referred to the
literature.7,9 Materials
The goal of this paper is to report on the
Sucrose, trehalose, lactose, glycine, bovine serum
experimental results obtained during a prelimin-
albumin (BSA), and polyvinylpyrrolidone (PVP,
ary evaluation of the commercially available
Mr: 40 kDa), were purchased from Sigma Chemi-
SMARTTM Freeze Dryer technology. The design
cal Company (St. Louis, MO) and used without
of experiments was planned in two consecutive
further purification. Two Pfizer (Pfizer, Inc.,
steps: a first step or ‘‘basic test’’ to evaluate
Groton, CT) formulations were used from the
parameters required to perform accurate MTM
actual production line: formulation # 1 (‘‘Pfizer I’’)
measurements when using SMARTTM software in
which consists of an API (small molecule) with
combination with a new freeze dryer (i.e.,
citrate buffer in Water for Injection with a total
different model than the one originally14–16 used).
solid content of 16.6% (remains fully amorphous)
Since the original experiments on MTM technol-
and formulation # 2 (‘‘Pfizer II’’) with a 4:1 ratio
ogy were initially performed on a much smaller
of PEG to oligonucleotide at a total solid
lab dryer (FTS Durastop), determination of the
content of 6% (partially crystalline formulation).
effective chamber volume of the larger scale
All reagents were analytical grade. Water was
laboratory freeze dryer (FTS Lyostar) and, more
purified by demineralization. Vials used for freeze
important, the minimum product load in the
drying were 5, 10 or 20 mL serum tubing vials of
drying chamber necessary to perform SMARTTM
20 mm finish from West Pharmaceutical Services
runs with this dryer were investigated. In the
Inc., (Lionville, PA) or Wheaton Science Products
second step, SMARTTM Freeze Dryer performance
(Millville, NJ). The reader is referred to Table 1 for
is evaluated on an individual ‘‘case study’’ basis by
details. Stoppers were gray butyl stoppers (West
using a variety of different products and formula-
Pharmaceutical Company). Values for collapse
tions, different vial sizes and different process
temperatures (Tc) of the excipients used during
conditions. These ‘‘case studies’’ are used to reveal
the experiments were either determined by
the merits and demerits of this new technology, in
freeze-dry microscopy (FDM), using a standard
particular with a focus on SMARTTM as a
procedure (‘‘a’’)11 or were taken from the literature
potential process analytical technology tool for
(‘‘b’’).12,13Tc’s were used as follows:
258C for
the future. The paper addresses the following
‘‘Pfizer I’’ and ‘‘Pfizer II’’ (‘‘a’’),
348C for sucrose
questions: (a) reproducibility of cycle design and
(‘‘a’’),
298C for trehalose (‘‘b’’),
338C for lactose
data obtained from a run with an emphasis on
(‘‘a’’), and
248C for PVP, 40 kDa (‘‘b’’).
product resistance, (b) agreement between pri-
mary drying endpoint predicted by MTM and
alternative end point determinations (e.g. Pirani), Table 1. Overview of Vials Used in This Study
and (c) consistency of the product temperature
predicted by SMARTTM during primary drying. It Type Av (cm2) Ap (cm2) Manufacturer
is important to note that this study does not focus 5 mL (I) 3.65 2.91 West Pharmaceuticals
on other drying options (e.g. secondary drying) 5 mL (II) 6.39 5.44 Wheaton
available in the software package. 10 mL (I) 4.71 3.89 Wheaton
For the experiments performed, the first 10 mL (II) 3.95 3.21 Wheaton
criterion for a ‘‘successful run’’ was the appear- 20 mL 6.83 5.74 West Pharmaceuticals
ance of the final product, i.e. the absence of visible
Av is the outer cross-sectional area, Ap the inner cross-
shrinkage or full collapse of the cake structure sectional area of each type of vial. Values are either measured
(‘‘product quality criteria’’). The second criteria or taken from the literature.

JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 96, NO. 12, DECEMBER 2007 DOI 10.1002/jps
 FREEZE-DRYING CYCLE OPTIMIZATION 3405

Determination of the Effective Chamber Volume temperature data (Tb) and temperatures calcu-
lated for the bottom of the vial by the MTM
Since the chamber volume (Vc) is an important
method (Tb-MTM).14 Note that ‘‘center’’ vials does
input parameter for the MTM equation, it is
not necessarily mean in the center of the shelf,
necessary to obtain this value for a new freeze
but does mean tightly packed in a hexagonal
dryer. The exact volume of 5 vials (20 mL, gray
formation with a minimum of two rows of
tert-butyl stoppers) was determined from the
surrounding product vials. Accuracy of the
weight difference of the ‘‘empty vials þ stopper’’
temperature reading for the thermocouples used
and ‘‘vials þ stopper, filled with water at 258C’’.
was within0.58C (Omega Engineering). All
The five dry and empty vials were then placed onto
thermocouples were calibrated over the range of
the middle shelf of the freeze dryer, followed by

40 to 208C, using an Omega CL27 series
reduction of the distance to the upper shelf.
calibrator with ramp function and a resolution
Stoppers and vials must be carefully compressed
of 0.18C (Omega Engineering).
between the shelves to immobilize the stoppers.
After evacuation of the chamber (50 mTorr) and
separation of the chamber from the condenser, the Freeze Drying Procedure and Additional
top shelf was retracted and the stoppers were SMARTTM Freeze Dryer Input Parameters
allowed to pop-out of the vial neck, resulting in a
Freeze drying was performed with a laboratory
pressure increase in the chamber. The MKS
scale freeze dryer (Lyostar IITM, FTS Systems,
Baratron pressure gauge was calibrated to ensure
Stone Ridge, NY) with a total shelf area (Ashelf-tot)
accurate pressure measurements. The chamber
of 0.43 m2 and with SMARTTM Freeze Dryer
volume, Vc, was then calculated, using the ideal
software installed (FTS Systems). All excipient
gas law (Eq. 1):
solutions used in this paper were prepared by
pV VV weight/weight ratio (w/w) and filtered (0.22 mm,
Vc ¼ (1)
Dp Millipore1, Billerica, MA) prior to filling. The two
where pV (mTorr) is the pressure in the vials, Pfizer formulations were prepared by weight/
which was 742 mTorr (reported by the local volume ratio (w/v). A number of selected vials
weather information at the day of the experi- were weighed before and after filling to determine
ment), VV (m3) is the total gas volume of the vials the effective fill weight of the solution, which is a
and Dp (mTorr) is the effective pressure increase required input parameter in the software. Fill
in the chamber after all stoppers popped out of the volume and number of vials were varied according
vial neck. A chamber volume (or even more to the experimental design. Details for each run
precise: MTM gas volume) for the ‘‘Lyostar II’’ are provided in the individual sections. All vials
laboratory freeze dryer could be determined to be were then loaded onto stainless steel (bottomless)
107.5 L. Note that the MTM valve is located at a trays with stainless steel rail-guards. For each
much lower point in the duct in the freeze dryer run, all vials were arranged in a tightly packed
used for this test due to additional equipment hexagonal formation. All vials loaded were either
modifications. The authors would expect a slightly loaded on the middle shelf of the freeze dryer
lower value for a Lyostar II freeze dryer delivered (partial load), or all three shelves were filled with
from the factory (about 104 L). vials (full load), and the bottom of the tray was
removed (rail-guard still in place). Note that
different number of vials represented different ice
sublimation areas. Radiation heat shields were
Thermocouple Placement and Calibration
used for all experiments performed. The heat
T-type copper-constantan thermocouples (0.01 mm shields included one row of empty (dummy) vials
diameter, Omega Engineering, Newport, CT) were to minimize radiation heat transfer from the
used in all experiments to determine the product freeze dryer chamber wall and the door, as well as
temperature. All thermocouples were placed in the aluminum foil attached to the inside of the
center of the vial touching the vial bottom. The chamber door to minimize radiation from the
thermocouple product temperature (Tb) was deter- door. The option for primary as well as secondary
mined in vials located at different spots on the shelf drying was set to ‘‘1’’. The nature of drug product
(center vials and front row vials facing the chamber was either set to ‘‘small molecule’’ or ‘‘proteins’’
door). Only temperature data from center vials dependent upon the molecular weight for the
were used to directly compare thermocouple excipient used. A leak rate test was performed

DOI 10.1002/jps JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 96, NO. 12, DECEMBER 2007
3406 GIESELER, KRAMER, AND PIKAL

prior to all SMARTTM runs since a low leak rate is where N is the number of product vials, Ap is the
mandatory to perform valid pressure rise mea- vial cross sectional area (cm2), V is the chamber
surements. The highest acceptable leak rate for a volume (L), Ts is the shelf temperature setpoint
SMARTTM run is reported to be 30 mTorr/h (or and (Rp þ Rs) the sum of product and stopper
7.810
6 mTorr/L/s for a Lyostar II with a 107.5 L resistance (cm2 Torr h/g). The ‘‘NA’’ term, is the
chamber volume) by the manufacturer (FTS total area of sublimation interface, and the ‘‘NA/
Systems). During the experiments performed in V’’ term is the ice area to chamber volume ratio or
this study the typical leak rate was determined to ‘‘volume normalized area’’. During a pressure rise
be within 8–12 mTorr/h (2.110
6 to 3.110
6 experiment, the initial exponential part of the
mTorr/L/s) for this laboratory scale freeze dryer. pressure rise curve or ‘‘resistance dominated
phase’’ must be short relative to the total duration
of the measurement to allow sufficient develop-
Manometric Temperature Measurements, MTM ment of the second, ‘‘temperature dominated
phase.’’14 In general, very low Q values would
MTM measurements were performed at 60 min
lead to a significant lower product temperature
intervals during primary drying, and pressure
prediction by the MTM procedure. Note that even
data were collected at a rate of 10 points per
under full load conditions the Q value will
second with a total time of 25 s for one
decrease over time during primary drying since
measurement. Nonlinear regression analysis
the product resistance (typically) increases.
was used to fit the MTM equation to the data,
Errors may arise from the assumption that
which was instantaneously performed by
‘‘NA’’ remains constant over the total primary
SMARTTM software after each measurement.
drying time. This assumption is dependent on the
The MTM equation employed in the software to
drying rate and therefore properties related to
evaluate the vapor pressure of ice at the
drying behavior (e.g. concentration, fill volume,
sublimation interface ( Pice) and the product
formulation type). For example, this assumption
resistance (Rp) was previously reported in the
was found to be incorrect after about 60% of the
literature.9 Note that the temperature gradient in
total primary drying for typical amorphous
the product (DT) is calculated and simultaneously
products and about 80% of the total primary
obtained by the MTM fit, rather than set to a fixed
drying time for typical crystalline materials.9 In
value of 1 K.14 The resistance of stoppers does
^ on total general, when the first vials (typically edge vials)
change with pressure but the impact of Rs
finish primary drying (i.e., finish ice sublimation)
resistance is negligible. Therefore, the fitted total
^ þ Rs)
^ was treated as area normal- the bias between a predefined ‘‘N’’ and the actual
resistance (Rp
^ number of vials containing ice (‘‘Nc’’) starts to
ized product dry layer resistance (Rp).
increase.
The commercial SMARTTM Freeze Dryer tech-
nology was developed for a new laboratory scale
RESULTS AND DISCUSSION
freeze dryer (Lyostar II) which is different in size
and geometry from the initially tested smaller
Basic Investigation—Minimum Product Load
unit (Durastop). Here, for the Lyostar II, the
for SMARTTM Runs
‘‘effective gas volume’’ for a pressure rise test is
A minimum ice sublimation area (Asub) was determined to be 107.5 L as reported above. The
reported14 to be necessary for accurate MTM ratio ice sublimation area to chamber volume
and therefore product temperature measure- ‘‘Asub:Veff’’ for the smaller lab freeze dryer is 2.9.
ments. For a small lab scale freeze dryer (FTS This ratio should be essentially identical for the
Durastop) with an effective chamber volume of bigger freeze dryer and independent of the type of
52 L, a minimum Q value of 0.2 or ice sublimation product. To verify this assumption, five different
area requirement of 150 cm2 was recently found.14 experiments were performed using a 75 mg/mL
Here, 5% glycine samples were used to evaluate trehalose solution with 1 mL fill volume in 5 mL (I)
Asub during the reported testing procedure. The Q tubing vials. A decreasing number of vials were
value is given from the first term of the MTM used on the middle shelf to characterize the
equation:14 impact of a decreasing ice sublimation area on the
  MTM analysis (308 vials—Asub: 896 cm2, 150
3:461 N Ap Ts vials—Asub: 437 cm2, 100 vials—Asub: 291 cm2, 75
Q (2)
VðRp þ Rs Þ vials—Asub: 218 cm2 and 50 vials Asub: 146 cm2).

JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 96, NO. 12, DECEMBER 2007 DOI 10.1002/jps
 FREEZE-DRYING CYCLE OPTIMIZATION 3407

After performing SMARTTM Freeze Dryer runs, ments when using 50, 75 and 100 vials, or a
thermocouple and Rp data were used to calculate calculated Q value of 0.3 or lower (Fig. 1). The
Q values according to Eq. (2) for every MTM product temperature ‘‘error’’ for 308 and 150
measurement during the individual cycles. product vials, however, does not deviate by more
As expected, a decrease in the number of vials than 18C with a Q value of 0.4 or greater. Note that
leads to a significant increase of the ‘‘resistance- all Q values were calculated only for the initial
dominated part’’ which, in turn, leads to an 2/3rd of primary drying for each run. It is clear
increase of the difference between product tem- from Figure 1 that the temperature error does not
perature measured by thermocouples and product depend on the Q value alone as the number of vials
temperature calculations by MTM (Fig. 1). is also a factor at Q values around 0.2. This, in
Analysis of the pressure rise raw data for the turn, can either be caused by the approach to the
MTM measurement for each run reveals that the end of primary drying when a significant fraction
‘‘temperature-dominated’’ part decreases from of vials are finished with ice sublimation. The
18 s (308 vials) to 13 s (150 vials), 9 s (100 vials), displacement of the curve for 50 vials to the right
5 s (75 vials) and 3 s (50 vials), respectively (data (i.e. a slightly lower Q value, Fig. 1) was found to
not shown). The second MTM measurement (i.e. be a direct result of the sensitivity of the MTM
total time for the product in primary drying of 2 h) ‘‘Rp þ Rs’’ prediction (tendency to over predict
generally provides more representative data ‘‘Rp þ Rs’’) when only a few product vials were
since the ‘‘first guess’’ shelf temperature setting present in the drying chamber.
selected by SMARTTM is typically low (i.e. a One might argue that a ‘‘maximum temperature
‘‘conservative’’ guess), and therefore the pressure error’’ of only 18C might be too restrictive,
increase during the first pressure rise test is low particularly since the thermocouple temperature
(i.e. not typical of the rest of the cycle). The measurements are also subject to some error.
temperature difference between thermocouple However, using at least 100 5 mL vials, with a total
and MTM data increases significantly (>28C) ice sublimation area of about 300 cm2, generally
after the second or third pressure rise measure- provides a Q value in excess of 0.3 and insures

Figure 1. Relationship between the total deviation of Tb-MTM from Tb and the corre-
sponding Q-value. Q values were calculated from SMARTTM resistance data during a
run up until 2/3rd of primary drying. Runs were performed with 75 mg/mL trehalose
solutions, a 1 mL fill volume and vial type 5 mL (1). Symbols represent: ~ ¼ 50 vials,
¼ 75 vials, ^ ¼ 100 vials, * ¼ 150 vials, and & ¼ 308 vials.

DOI 10.1002/jps JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 96, NO. 12, DECEMBER 2007
3408 GIESELER, KRAMER, AND PIKAL

temperature errors are minimal. The minimum drying conditions show good reproducibility of the
ratio ‘‘Asub:Veff’’ is then calculated to be 2.7 for the ‘‘Ts, t’’ profile (i.e. cycle recipe) selected by the
Lyostar II, which is essentially the same ratio SMARTTM Freeze Dryer for duplicate runs, in
found for the smaller freeze dryer. Clearly, if particular for those runs performed with trehalose
possible, one should use more than hundred 5 mL and the ‘‘Pfizer II’’ formulation (Tab. 2). Note that
(I) vials to insure the maximum temperature SMARTTM might include an additional step (e.g.
accuracy. step # 3, 5% sucrose solution, 220  10 mL (II)
vials, 1 mL, Tab. 2) when performing runs under
identical product and process (setting) conditions.
SMARTTM Case Study I: Reproducibility of
However, even when the additional step was
Primary Drying Cycle Adjustments
included, the impact on the product temperature/
Duplicate SMARTTM Freeze Drying experiments time profile was insignificant since the additional
were conducted for a 5% sucrose, 7.5% trehalose step was short (60 min) and shelf temperature
and a 6% ‘‘Pfizer II’’ partially crystalline formula- deviates by not more than 28C from the recipe
tion to determine the reproducibility of the shelf obtained for the first run. Note that an increase of
temperature (Ts) and time (t), recommended for 58C in shelf temperature results in only about a
primary drying by the software. The freezing and 18C increase in product temperature.3 Further, it
secondary drying conditions for these cycles were should be emphasized that it is the product
controlled using the ‘‘built in’’ expert system temperature history that is most critical in
algorithms parameterized by user input informa- assessing the reproducibility of a freeze drying
tion (e.g. fill depth, collapse temperature, and process.
nature of product: crystalline or amorphous). For The SMARTTM freeze dryer software does not
the runs performed, the recommended primary allow shelf temperature alterations after 2/3rd of

Table 2. Primary Drying Conditions Selected by the Smart Freeze Dryer for Duplicate Runs of 5% sucrose
(Tc setting:
348C), 7.5% Trehalose (Tc setting:
298C) and a 6% ‘‘Pfizer II’’ formulation (Tc setting:
258C)

Shelf Temperature (Ts) Setpoint During Primary Drying

SMART Cycle Time (t) at Shelf Temperature, Ts

Product (Run #),
Number of Vials, Fill Volume Step # 1 Step # 2 Step # 3 Step # 4
5% sucrose (# 1), Ts:
37.08C, Ts:
15.28C, Ts:
17.88C, —
220  10 mL (II) vials, 1 mL t: 58 min, t: 58 min, t: 956 min,
Pc: 57 mTorr Pc: 57 mTorr Pc: 57 mTorr
5% sucrose (# 2), Ts:
37.0 8C, Ts:
13.8 8C, Ts:
15.88C, Ts:
17.08C,
220  10 mL (II) vials, 1 mL t: 58 min, t: 58 min, t: 60 min, t:840 min,
Pc: 57 mTorr Pc: 57 mTorr Pc: 57 mTorr Pc: 57 mTorr
5% sucrose, Ts:
37.08C, Ts:
24.18C, Ts:
21.58C, —
112  20 mL vials, 3 mL t: 57 min, t: 57 min, t: 1440 min,
Pc: 57 mTorr Pc: 57 mTorr Pc: 57 mTorr
5% sucrose, Ts:
37.08C, Ts:
18.98C, Ts:
23.08C, Ts:
14.18C,
336  20 mL vials, 3 mL t: 57 min, t: 177 min, t: 897 min, t: 471 min,
Pc: 57 mTorr Pc: 57 mTorr Pc: 57 mTorr Pc: 57 mTorr
7.5% trehalose (# 1), Ts:
32.08C, Ts:
14.28C, Ts:
12.18C, —
308  5 mL (I) vials, 1 mL t: 77 min, t: 92 min, t: 495 min,
Pc: 71 mTorr Pc: 71 mTorr Pc: 65 mTorr
7.5% trehalose (# 2), Ts:
32.08C, Ts:
13.88C, Ts:
12.18C, —
308  5 mL (I) vials, 1 mL t: 77 min, t: 92 min, t: 495 min,
Pc: 71 mTorr Pc: 71 mTorr Pc: 65 mTorr
6% ‘‘Pfizer II’’ formulation (# 1) Ts:
28.38C, Ts:
9.48C, Ts: 22.58C, Ts: 14.28C,
161  10 mL (I) vials, 6.8 mL t: 57 min, t: 118 min, t: 238 min, t: 1316 min,
Pc: 85 mTorr Pc: 85 mTorr Pc: 85 mTorr Pc: 85 mTorr
6% ‘‘Pfizer II’’ formulation (# 2) Ts:
28.38C, Ts:
9.48C, Ts: 22.58C, Ts: 14.28C,
161  10 mL (I) vials, 6.8 mL t: 57 min, t: 118 min, t: 238 min, t: 1316 min,
Pc: 85 mTorr Pc: 85 mTorr Pc: 85 mTorr Pc: 85 mTorr

JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 96, NO. 12, DECEMBER 2007 DOI 10.1002/jps
 FREEZE-DRYING CYCLE OPTIMIZATION 3409

the estimated primary drying time has expired.9 amorphous materials (>15%) in combination with
However, when the product temperature exceeds a high fill depth where potential water re-
the target temperature in the early phase of adsorption effect by the dry amorphous layer
primary drying due to an overestimate of Ts, the cannot be ignored. Water re-sorption by the dried
software decreases instantaneously the shelf layer during the MTM measurement means not
temperature to maintain the target temperature all the water vapor produced by sublimation
(e.g. 5% sucrose, 220  10 mL (II) vials, run #2 or reaches the chamber and is available to increase
5% sucrose 336  20 mL vials: decrease of Ts in chamber pressure, thereby leading to a MTM
two consecutive steps). In particular for sucrose temperature that is too low. This effect is
formulations, an over prediction of Ts was currently being further investigated in an attempt
observed only in the second step and may be a to expand SMARTTM applications.
result of not fully developed steady state condi-
tions. In contrast, the product temperature may
increase by 1–38C (depending upon type of
SMARTTM Case Study II: Primary Drying Estimate
product, fill depth and product resistance) near
and Actual End Point Detection
the end of primary drying for some products, even
at a given (constant) shelf temperature. Here, a An algorithm based on steady state heat and mass
decrease in Ts as primary drying proceeds is transfer principles is used in the SMARTTM
necessary to maintain the product below the freeze dryer software to provide a rough estimate
target temperature.3,9 This alteration in Ts is for the total primary drying time for a given
found in the process for the 6% ‘‘Pfizer II’’ product.3 This predicted cycle time, which is
formulation, where the product temperature at calculated from the first MTM measurement
Ts exceeds the defined target temperature. In performed during a freeze-drying run, would
response, the system reduced Ts in the final phase be expected to agree with total cycle time as
to 14.28C. Note that even with different loading indicated by thermocouples (TCs) which are
conditions (5% sucrose, 112 vials and full placed bottom center, with end point determined
load ¼ 336 vials), cycle reproducibility is very by comparative pressure measurements (Pirani
consistent, in particular in the choice of shelf and Capacitance Manometer), and with end
temperature. Table 3 gives an example of a simple point determined by the MTM procedure late in
‘‘automatic cycle’’ that was derived from the cycle the process. This MTM procedure involves a
proposed by SMARTTM. Here the initial low comparison between the vapor pressure of ice
temperature step from SMARTTM was omitted determined by MTM with the chamber pressure,
as unnecessary, but as expected, the product with a small difference between these pressures
temperature history was nearly identical to that indicated the end point of primary drying.
obtained from the SMARTTM run. The agreement between this initial estimate of
In a few cases, SMARTTM recommends an primary drying time by the SMARTTM software
increase in Ts over two or even three steps is and endpoint determined by Pirani, TC, or
near the end of ice sublimation (data not shown). MTM is very good for the runs performed with
These case studies involved highly concentrated glycine, lactose, and PVP (Tab. 4). During our

Table 3. Comparison of transferability of a SMARTTM Cycle Recipe to a Simple Run Using an Automated Program

Ts Setpoint During Primary Drying, Endpoint by,

Cycle Time (t) at Shelf Temperature, Ts (a) Pirani,

Product Number of Vials,
Fill Volume Step # 1 Step # 2 Step # 3 (b) av. Tp
TM
5% sucrose (SMART run) Ts:
37.08C, Ts:
15.28C, Ts:
17.88C, (a) 792 min, (b) 780 min
220  10 mL (II) vials, 1 mL t: 58 min, t: 58 min, t: 956 min,
Pc: 57 mTorr Pc: 57 mTorr Pc: 57 mTorr
5% sucrose (auto run) Ts:
18.08C, — — (a) 810 min, (b) 800 min
220  10 mL (II) vials, 1 mL t: 840 min, Pc: 57 mTorr — —
A SMARTTM cycle with 5% sucrose was performed at the longest shelf temperature/time setting from this run taken and as a ‘‘one
step’’ setting in the automated cycle. The automated cycle was calculated by deducting the intermediate steps (i.e.
37.08C for 58 min)
in the SMARTTM cycle

DOI 10.1002/jps JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 96, NO. 12, DECEMBER 2007
3410 GIESELER, KRAMER, AND PIKAL

Table 4. Comparison of SMARTTM Primary Drying Time Estimate and Effective Cycle Time Determined by MTM,
Pirani and Average Thermocouple (TC) Readings for Various Products

Product SMARTTM Endpoint Endpoint by Endpoint by: Endpoint by:

(User Input Parameters) Estimate (h) MTM (h) PIRANI (h) av. TC (h)
5% glycine, 22.9 23.6 22.5 22.7
116  20 mL vials
5 mL fill, Tc:
308C
10% glycine, 15.9 13.0 13.6 13.2
112  20 mL vials
2 mL fill, Tc:
288C
5% lactose, 16.9 17.3 18.8 18.6
145  20 mL
3 mL fill, Tc:
318C
3% PVP, 20.0 19.0 19.5 18.6
112  20 mL vials
5 mL fill, Tc:
248C
5% sucrose, 41.8 27.2 27.1 26.8
112  20 mL vials (29.9)a
3 mL fill, Tc:
348C
16.6% Pfizer I formulation, 50.8 33.0 32.0 31.2
161  10 mL (I) vials (30.7)a
6.75 mL fill, Tc:
258C
5% sucrose þ 5% BSA, 28.5 54.6b 53.5b 27.9
178  10 mL (II) vials
5 mL fill, Tc:
238C
Note that the MTM endpoint is taken when Pice by MTM within 5 mTorr of Pc. Time (h) represent the time interval, starting from
the onset of primary drying (shelf ramping).
a
SMARTTM endpoint estimate calculated from the second MTM measurement performed during primary drying. Note that the
software calculates as a matter of routine the estimate of the primary drying endpoint from the first MTM measurement taken in
primary drying.
b
Secondary drying effect at the end of primary drying which extends the effective primary drying time.

experiments, the criteria for an endpoint by Pirani drying time. However, a simple extension of
or by MTM is a maximum difference of 5 mTorr the time before the first MTM measurement
from the MKS Baratron reading.9 To insure (i.e. by performing MTM measurements in 90
accurate results, a leak rate test with an empty min intervals) for these particular products
chamber was performed prior to all SMARTTM should improve the initial estimate of primary
runs. drying time.
Some atypical results were found for the More difficult to interpret is the result obtained
primary drying time estimate for the 5% for the 5% sucrose þ 5% BSA binary mixture
sucrose and 16.6% ‘‘Pfizer I’’ formulations (Tab. (Fig. 2). Here total cycle times measured by MTM
4). Here, the initial SMARTTM estimate is much and Pirani are almost twice of what both the
longer than the other estimates of primary SMARTTM software and the thermocouple data
drying time. Note that both runs show very predict for cycle time from the first MTM
good agreement between Pirani, MTM and TC measurement (Tab. 4). After about 25 h, the
data. Calculation of the primary drying time increase of the average thermocouple reading
prediction from the second MTM measurement indicates a loss of contact to the ice in the product,
gives a primary drying time estimate of 29.9 h for which is accompanied by a sharp decrease in the
the 5% sucrose and 30.7 h for the ‘‘Pfizer I’’ Pirani reading. The Pirani gauge (as well as Pice
formulation, which are in good agreement with data by MTM) then continues to read much
the effective cycle time obtained from the other higher than the MKS Baratron for an additional
methods. This result suggests that the product is 25 h, suggesting significant water flux and
not in steady state condition at the time of therefore, presence of ice. The initial phase of
the first MTM measurement which, in turn, is a this pressure drop might be related to presence of
requirement for accurate prediction of primary insulated ice slabs in the product (i.e., isolated

JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 96, NO. 12, DECEMBER 2007 DOI 10.1002/jps
 FREEZE-DRYING CYCLE OPTIMIZATION 3411

Figure 2. SMARTTM freeze drying cycle generated for a 5% sucrose þ 5% BSA binary
mixture. Upper plot: & ¼ shelf temperature setpoint and ^ ¼ average thermocouple
(n ¼ 3) reading over time, lower plot: * ¼ pressure reading by the Pirani gauge,
~ ¼ pressure reading by MKS Baratron, * ¼ Pice data from MTM fit. Note that the
primary drying endpoint by Pirani and MTM is different from an endpoint indicated by
thermocouple data.

islands of ice in the dry layer), and ice removal transfer relationships relevant to freeze-drying.
may not be complete until well into secondary All equations were previously reported in the
drying. In this case, the end point of primary literature.7,9,14–16
drying is truly ambiguous, but if defined in the Reproducibility of product resistance is impor-
usual manner as the point at which all ice has tant to confirm reproducibility of pore structure
been removed, the longer time would represent and therefore reproducibility of freezing. Typical
the end point. dry layer resistance for a lyophile can be described
using a simple empirical relationship (Eq. 3)
between the product resistance and the dry layer
thickness.17,18 It is important to note that four
SMARTTM Case Study III: Reproducibility of ‘‘Types’’ of Rp increase with dry layer thickness
Product Resistance, Rp have been reported, where the data were obtained
by a microbalance measurement technique.8 Most
The MTM procedure was used to evaluate the systems are consistent with Eq. 3,
vapor pressure of ice ( Pice) and the resistance of
A1 l
the product and stopper (Rp þ Rs) by fit of the R^p ¼ Rp ð0Þ þ (3)
1 þ A2 l
MTM equation to the pressure rise data. It is
important to note that only these two values were where R^p is the normalized product resistance, l is
used to subsequently calculate all additional the dry layer thickness, Rp(0) is the initial product
process parameters (i.e. Tp, Tb-MTM, Kv, dm/dt, resistance at l ¼ 0, and A1 and A2 are constants
Lice, etc.) by using steady state heat and mass corresponding to the fit. It should be noted that

DOI 10.1002/jps JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 96, NO. 12, DECEMBER 2007
3412 GIESELER, KRAMER, AND PIKAL

Figure 3. Product resistance (Rp) as a function of dry layer thickness (l) for various fill
depths (given as Lice,t ¼ 0) of 5% sucrose solutions: * ¼ Lice,t ¼ 0: 0.2 cm (run # 1),
* ¼ Lice,t ¼ 0: 0.2 cm (run # 2), & ¼ Lice,t¼0: 0.34 cm (run # 1), & ¼ Lice,t ¼ 0: 0.34 cm
(run # 2), ~ ¼ Lice,t ¼ 0: 0.58 cm, ^ ¼ Lice,t ¼ 0: 1.02 cm. Rp data of a 20% sucrose solution
(filled stars, Lice,t ¼ 0: 0.28 cm) illustrates the impact of total solid content on Rp.

dry layer resistance potentially also depends on resistances, relative to the corresponding data
fill volume and solute concentration. obtained at lower Lice, t ¼ 0 values. The slope of the
Figure 3 illustrates the dependence of resis- Rp changes after about l ¼ 0.4 cm for the system of
tance on dry layer thickness for different 5% Lice, t ¼ 0 of 0.58 cm and suggests that a vial with a
sucrose formulations. Rp data are plotted up until very high fill depth may show a different pore
2/3rd of the end of primary drying. Very good structure in different parts of the product
reproducibility in terms of product resistance is which, in turn, can arise from heterogeneity
found for the two systems where duplicate runs in the freezing behavior. For 5% sucrose and
were made (i.e. 5% sucrose with initial ice layer Lice, t ¼ 0 ¼ 1.02 cm the ice crystals and resulting
thicknesses, Lice,t ¼ 0, of 0.2 cm and 0.34 cm). pores are evidently smaller near the top of the
Calculation of Lice,t ¼ 0 was performed from the fill cake, leading to a slightly higher resistance at
weight and inner cross-sectional area of the vial as small dry layer thicknesses.
reported before.9 For the duplicate runs, product In a second set of experiments, Rp was
resistance shows a non zero intercept at l ¼ 0 cm evaluated for the effect of varying concentration.
and increases almost linearly with increasing dry Figure 3 shows the product resistance for a 20%
layer thickness (‘‘Type I’’ behavior). Rp data of the sucrose run (Lice, t ¼ 0 of 0.28 cm), where the
5% sucrose run at Lice,t ¼ 0 ¼ 0.58 cm perfectly resistance is higher than found for 5% sucrose at
matches the resistance data at lower fill depth. corresponding dry layer thicknesses, as expected.
However, the ‘‘Type’’ of curve may be assumed The results for product resistance for varying
between ‘‘III’’ and ‘‘IV’’ which could indicate some concentration could be very useful in optimizing
microcollapse.8 Note that pure sucrose shows the formulation to be freeze-dried. To give an
some cake shrinkage, even when freeze dried well example, suppose that the desired dosage is 500 mg
below the collapse temperature.19 An Increase of of a compound with the option of freeze drying
Lice, t ¼ 0 from 0.58 to 1.02 cm results in higher with a 50 mg/mL solution using a 10 mL fill, or a

JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 96, NO. 12, DECEMBER 2007 DOI 10.1002/jps
 FREEZE-DRYING CYCLE OPTIMIZATION 3413

200 mg/mL solution with a 2.5 mL fill. Assuming (upper plot, right) where Rp is high at the
the two formulations had identical container size beginning relative to the rest of the data, then
and comparable product temperature at the ice- drops after the second MTM measurement and
interface during primary drying, the 50 mg/mL then steadily increases (with some scattering)
formulation with a 10 mL fill volume would take over the process time. In this experiment, the
approximately three times longer to complete ice lactose forms a dense ‘‘skin’’ on top of the cake
sublimation. Such evaluations would be impor- which acts as a barrier to the water vapor in the
tant in estimating the primary drying endpoint initial phase of drying (‘‘Type II’’ resistance
for a change in formulation characteristics (e.g. behavior). As drying proceeds, cracks develop
concentration, fill height, etc.). and the product resistance drops to a more
‘‘representative’’ resistance value in this early
stage of drying. Note that high freezing rates or
SMARTTM Case Study IV: Rp Data and Correlation high fill volumes have also been found to generate
to Product Drying Behavior a similar resistance pattern.20 An ‘‘ideal’’ behavior
(‘‘Type I’’) is found for a 5% glycine solution (upper
Divergence from the ‘‘ideal’’ model behavior plot, left), even at a high initial fill depth (Lice,t ¼ 0:
described above can also provide important 0.99 cm). Note, however, for this sample as well as
information about the drying behavior of a with two other examples in Figure 4, a very sharp
substance or formulation. For instance, Figure 4 increase in slope is observed near the end of the
shows resistance data for a 10% lactose solution curve. This sharp increase is likely an artifact

Figure 4. Product resistance (Rp) as a function of dry layer thickness (l) for different
materials and cycle conditions: ~ ¼ 5% glycine, 20 mL vial, 5 mL fill, Lice,t ¼ 0: 0.99 cm,
* ¼ 10% lactose, 5 mL (I) vial, 2 mL fill, Lice,t ¼ 0: 0.58 cm, ^ ¼ 5% PVP, 20 mL vial, 5 mL
fill, Lice,t ¼ 0: 0.97 cm, ¼ 16.6% ‘‘Pfizer I’’ product, 10 mL vial (I), 6.75 mL fill, Lice,t ¼ 0:
1.9 cm.

DOI 10.1002/jps JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 96, NO. 12, DECEMBER 2007
3414 GIESELER, KRAMER, AND PIKAL

arising from heterogeneity in drying. That is, at during ice sublimation. Since Tp is calculated
this point, a significant number of vials have from Pice, which is obtained by non-linear regres-
finished primary drying, and the resistance data sion analysis, inaccuracies in the MTM fit impact
begin to be subject to a systematic error. the calculated value for Tp. As a first approxima-
Another divergence from the model equation is tion, the temperature at the bottom of the vial
found for a 5% PVP solution (Fig. 4, lower plot, (Tb-MTM) is calculated based on the assumption
left). Rp increases over time up until a product dry that all heat (dQ/dt, in cal/h per vial) provided
layer thickness of about 0.5 cm, followed by a by the shelf comes through the bottom of the vial
sharp decline of the resistance data and subse- (Eq. 4):
quent increase before 2/3rd of primary drying is
completed. Note that this ‘‘pattern’’ was found for dQ=dtl
Tb
MTM ¼ þ Tp (4)
PVP concentrations ranging from 3% to 10% (data Av Kice
not shown). While the increase in Rp shown by the
last point is likely an artifact, as discussed above, Where l is the ice thickness (cm), Kice is the
the decrease of Rp between 0.5 and 0.7 cm dry thermal conductivity of ice (cal/K cm s, known), Av
layer thickness is likely real and a result of is the outer cross-sectional area of the vial (cm2,
microcollapse of the structure, thereby decreasing known) and Tp is the product temperature (K)
the resistance to vapor flow through the product. at the sublimation interface as calculated by
Examination of the final product revealed no MTM. Tp, however, is for the majority of runs
obvious shrinkage in the cake, but a distinct approximately 1–38C lower than the temperature
‘‘splitting’’ of the PVP cake (i.e. a big gap measured at the bottom of a vial with a thermo-
separating the cake in two halves) was observed couple.7 Since MTM measures a temperature
in every single vial of the batch. which is representative for the vials which are
Clear shrinkage was found in the product when running on the coldest product temperature in
freeze drying a 16.6% ‘‘Pfizer I’’ amorphous an array of vials,14 the temperature measured
formulation with a 6.75 mL fill in a 10 mL (I) with a thermocouple (Tb) touching the center
vial (Fig. 4, lower plot, right). After an initial bottom of a vial which is located in the center of a
plateau phase (up until l ¼ 0.2 cm) the resistance batch is a reasonable value for a comparison with
decreases to about 3.25 cm2Torrh/g and Tb-MTM.
increases sharply around 0.4 cm dry layer Figure 5 shows the temperature profile of Tp, Tb-
thickness. Examination of the final product MTM and thermocouple data (center vials) over
revealed clear microcollapse of the structure time for a 5% sucrose solution. Here, very good
which led to a change in the product morphology agreement is found between thermocouple and
and thus resistance to mass flow. The resistance MTM data until about the 17 h mark. At longer
behavior for the ‘‘Pfizer I’’ product run is times, the MTM data are biased slightly to low
reproducible (data not shown) and might indicate temperatures, indicating the onset of significant
a shelf temperature setting in primary drying errors in the MTM measurements and/or choice of
which was too high for the product. More likely, a non-representative vial for thermocouple mea-
however, is that the resistance behavior at dry surement. In general, product temperature deter-
layer thicknesses beyond about 0.2 cm is seriously mined by MTM was reported to be accurate until
impacted by the water re-sorption effect, and approximately 2/3rd of the initial ice thickness.7
the resistance data are subject to serious sys- However, we found this conclusion to be valid only
tematic error. This effect, likely an issue for when great care is used in placement of thermo-
amorphous solutes at very high concentration, couples in the product. We experienced practical
was previously reported to be a limitation of difficulties in obtaining ‘‘representative’’ thermo-
SMARTTM technology.9 couple data, in particular when thermocouples are
set in place by ‘‘untrained’’ users. Observed
deviations between single thermocouples and
calculated Tb-MTM data in the early phase of
SMARTTM Case Study V: Product Temperature
primary drying can often be attributed to errors
Calculations by SMARTTM
associated with placing thermocouples. Possible
It is important to verify that the product reasons for a difference between MTM and
temperature at the sublimation interface (Tp) thermocouple temperature measurements are:
does not exceed the critical product temperature (a) Failure to place the thermocouple in the center

JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 96, NO. 12, DECEMBER 2007 DOI 10.1002/jps
 FREEZE-DRYING CYCLE OPTIMIZATION 3415

Figure 5. Five percent sucrose run (240 vials, 5 mL (I), 2 mL fill volume, Tc setpoint:

328C, target temperature:
358C). The solid line represents the shelf temperature
selected by SMARTTM Freeze Dryer, open circles the product temperature of center
vials, measured at the bottom center, filled diamonds the corresponding product
temperature at the bottom of the vial calculated by MTM and open squares the product
temperature at the sublimation interface. MTM data are plotted for the initial 2/3rd of
primary drying.

of the vial with the tip touching the bottom center. During some experiments a significant devia-
In theory, this section is the last area where ice tion between Tb and Tb-MTM can be observed even
will remain at the end of primary drying, in the early phase of primary drying and when
assuming only a minor heat transfer between thermocouples are accurately placed. For these
vials and therefore a ‘‘horizontal’’ drying front. (b) runs, an ‘‘erroneous’’ prediction of product tem-
Stoppers used with ‘‘thermocouple vials’’ were not perature must result from the MTM procedure.
carefully set in the vial neck to ensure equal From the experiments performed during our tests,
‘‘openings’’ and therefore representative mass flux these deviations were mainly found for (a) high
from the thermocouple vials. A thermocouple vial solid contents (>15%), (b) a very high fill depth,
where the stopper is set too deep will run at a and (c) solutions of pure PVP when using
higher temperature compared to the surrounding concentrations above 3%. Whereas case (a) and
vials (and vice versa). (c) For most of the sterile- (b) may represent actual practice, case (c) is not
filtered solutions, there is a freezing bias between likely to constitute a problem as PVP is not
thermocouple vials and all other vials. Vials commonly used in parenterals. However, PVP
containing a thermocouple run at lower product may serve as a good ‘‘model’’ to theoretically study
temperature compared to the rest of the batch, this effect in more detail in the future. At this
because they nucleate at higher temperature, point of investigation we attribute this deviation
thereby producing bigger pores and less resis- to a ‘‘water re-adsorption’’ effect, which increases
tance to vapor flow. This effect is typically large in in significance as the total amount of amorphous
a manufacturing environment where particulate material increases due either to high solids
contamination, serving as heterogeneous ice content or a large dry layer thickness. Also, the
nucleation sites, is minimal, but may or may effect is likely to be greater for those materials
not be significant in a laboratory. where the diffusion coefficient of water is high,

DOI 10.1002/jps JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 96, NO. 12, DECEMBER 2007
3416 GIESELER, KRAMER, AND PIKAL

such as one might expect for polymers like PVP. the first few hours of primary drying. Further
Also, PVP, is known to be hydrophilic, which investigations to clearly delineate the ‘‘re-adsorp-
would support a ‘‘re-adsorption’’ effect of polar tion’’ effect are necessary in the future and will be
water molecules in the in the product.21 part of a subsequent publication.
Figure 6 illustrates a SMARTTM cycle of 5%
PVP and a Lice,t ¼ 0 value of about 1 cm. Note that
the first two Tb-MTM predictions do agree well with CONCLUSIONS
the corresponding thermocouple data. SMARTTM
Freeze Dryer is adjusting for the measured The results obtained during a preliminary eva-
temperature being below the target by increasing luation of the SMARTTM Freeze Dryer concept
the shelf temperature in three steps to Ts ¼ 208C clearly demonstrated that this technology is (for
to increase Tp. Since there is an upper boundary the majority of formulations and concentrations
limit for a Ts setting integrated into the software, used in freeze drying) a useful tool for develop-
Ts will not be elevated further. During this run, Tb ment of a lyophilization cycle during a single
data by thermocouples are very close to the freeze drying run. However, the quality of the
collapse temperature reported for PVP, whereas cycle optimization is dependent upon the accuracy
Tp by MTM is within or lower than the target of the ‘‘software input parameters’’ which must be
temperature. Note that the product only showed provided by the user (e.g. collapse temperature,
minor shrinkage (<10%) after this run and very vial cross sectional area). These input parameters
good appearance (i.e. acceptable visual ‘‘product may greatly affect the MTM analysis (and there-
quality’’). Thus, the ‘‘process optimization’’ infor- fore Pice and Rp values) and/or process design. To
mation provided by this SMARTTM run is verify that product temperature calculated by
acceptable, even though the temperature and the SMARTTM system during a first run of an
product resistance data are not reliable beyond unknown product is valid, the use of thermo-

Figure 6. Five percent PVP run (112 vials, 20 mL, 5 mL fill volume, Tc setpoint:

248C, target temperature:
278C). The solid line represents the shelf temperature
selected by SMARTTM Freeze Dryer, open circles the product temperature of center
vials, measured at the bottom center, filled diamonds the corresponding product
temperature at the bottom of the vial calculated by MTM and open squares the product
temperature at the sublimation interface. MTM data are plotted for the initial 2/3rd of
primary drying.

JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 96, NO. 12, DECEMBER 2007 DOI 10.1002/jps
 FREEZE-DRYING CYCLE OPTIMIZATION 3417

couples is still recommended. This ‘‘verification’’ is REFERENCES

mandatory when using highly concentrated pro-
duct solutions and/or high fill volumes. To obtain 1. Schwegman JJ, Hardwick LM, Akers MJ. 2005.
reliable data from thermocouples, correct hand- Practical formulation and process development of
ling and placement is necessary which, in turn, freeze-dried products. Pharm Dev Technol 10:151–
might be difficult for an unexperienced user. In 173.
addition, SMARTTM requires a minimum sub- 2. Pikal MJ. 1990. Freeze-drying of proteins. Part I.
Process design. BioPharm 3:18–25.
limation area of the product of about 300 cm2 in a
3. Tang XC, Pikal MJ. 2004. Design of freeze-drying
laboratory freeze dryer of 107.5 L to accurately
processes for pharmaceuticals: Practical advice.
obtain Rp and Pice data. The ratio: Asub/Veff Pharm Res 21:191–200.
confirms a consistency of area to chamber volume 4. Pikal MJ, Shah S. 1990. The collapse temperature
when compared with other scaled systems. How- in freeze drying: Dependence on measurement
ever, the minimum sublimation area requirement methodology and rate of water removal from the
can also be seen as limitation of this technology glassy phase. Int J Pharm 62:165–186.
when only a few milligram of an expensive drug 5. Carpenter JF, Pikal MJ, Chang BS, Randolph TW.
are available to perform a SMARTTM freeze 1997. Rational design of stable lyophilized protein
drying runs, at least unless a satisfactory placebo formulations: Some practical advice. Pharm Res
can be developed. Very good reproducibility for 14:969–975.
6. Pikal MJ. 1990. Freeze-drying of proteins. Part II.
cycle design, product resistance calculations and
Formulation selection. BioPharm 3:26–29.
primary drying end point detection for various low
7. Milton N, Pikal MJ, Roy ML, Nail SL. 1997. Evalua-
molecular weight excipients and two formulations tion of manometric temperature measurement as a
from Pfizer were found. These results form the method of monitoring product temperature during
basis for this technology to become a reliable PAT lyophilization. PDA J Pharm Sci Technol 51:7–16.
tool in the near future. Product temperature 8. Pikal MJ. 1985. Use of laboratory data in freeze
determined by MTM is generally in very good drying process design: Heat and mass transfer
agreement with thermocouple data for all crystal- coefficients and the computer simulation of freeze
line (glycine, ‘‘Pfizer I’’ formulation) and most of drying. J Parenteral Sci Technol 39:115–138.
the amorphous materials (sugars), at least 9. Tang XC, Nail SL, Pikal MJ. 2005. Freeze-drying
throughout the first half of primary drying and process design by manometric temperature mea-
surement: Design of a smart freeze-dryer. Pharm
often throughout 2/3rd of primary drying. A
Res 22:685–700.
temperature error found for PVP even in the
10. Hussain AS. 2001. Emerging science issues in phar-
early stage of primary drying indicating a maceutical manufacturing: Process analytical tech-
potential water re-adsorption tendency which nologies. FDA Sci Board Meet 16:2001.
can lead to deviations much earlier than the 11. Fonseca F, Passot S, Cunin O, Marin M. 2004.
reported ‘‘2/3 rule’’. However, if evidence of water Collapse temperature of freeze dried Lactobacillus
re-adsorption is found during a cycle for a given bulgaricus suspensions and protective media. Bio-
formulation, a reduction of the fill depth and/or technol Prog 20:229–238.
assessment of initial MTMs can lead to a more 12. Adams GDJ, Ramsay R. 1996. Optimizing the lyo-
consistent process design. philization cycle and the consequences of collapse
on the pharmaceutical acceptability of Erwinia L-
Asparaginase. J Pharm Sci 8606:1301–1305.
13. Wang W. 2000. Lyophilization and development of
ACKNOWLEDGMENTS solid protein pharmaceuticals. Int J Pharm 203:1–60.
14. Tang XC, Nail SL, Pikal MJ. 2006. Evaluation of
The authors wish to thank Leslie Mather and manometric temperature measurement (MTM), a
Hung Lee from FTS Systems for logistical, tech- process analytical technology tool for freeze drying.
nical and software support during the testing Part I. Product temperature measurement. AAPS
Pharm Sci Tech 7:14.
phase. In addition, Larry Gatlin, Evgenyi Sha-
15. Tang XC, Nail SL, Pikal MJ. 2006. Evaluation of
laev, and Bill Petre from the Pfizer Parenteral manometric temperature measurement (MTM), a
Center of Emphasis for their efforts in data ana- process analytical technology tool for freeze drying.
lysis, the University of Connecticut School of Part II. Measurement of dry layer resistance. AAPS
Pharmacy, Regeneron and Amgen for participat- Pharm Sci Technol 7:93.
ing in the beta site testing and the performance of 16. Tang XC, Nail SL, Pikal MJ. 2006. Evaluation of
SMARTTM freeze drying runs. manometric temperature measurement (MTM), a

DOI 10.1002/jps JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 96, NO. 12, DECEMBER 2007
3418 GIESELER, KRAMER, AND PIKAL

process analytical technology tool for freeze drying. 19. Rambhatla S, Obert JP, Luthra S, Bhugra C, Pikal
Part III. Heat transfer coefficient measurement. MJ. 2005. Cake shrinkage during freeze drying: A
AAPS Pharm Sci Technol 7:39. combined experimental and theoretical study.
17. Pikal MJ, Costatino HR. 2004. Lyophilization of Pharm Dev Technol 1:33–40.
biopharmaceuticals. In: Borchard RT, Middaught 20. Gieseler H. 2004. Product Morphology and Drying
CR, Series CR, editors. Biotechnology: Pharmaceu- Behavior delineated by a new Freeze-Drying Micro-
tical aspects Vol. II:. Arlington, VA: AAPS Press. balance. Doctoral Dissertation, University of Erlan-
ISBN: 0-9711767-6-0. gen-Nuremberg, Erlangen, Germany.
18. Rambhatla S, Ramot R, Bhugra C, Pikal MJ. 2004. 21. Zeng XM, Martin GP, Marriott C. 2001. Effects of
Heat and mass transfer scale-up issues during freeze molecular weight of polyvinylpyrrolidone on the
drying. II. Control and characterization of the degree glass transition and crystallization of co-lyophilized
of supercooling. AAPS Pharm Sci Technol 5:58. sucrose. Int J Pharm 218:63–73.

JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 96, NO. 12, DECEMBER 2007 DOI 10.1002/jps