International Journal of Pharmaceutics 280 (2004) 1–16

Freeze drying—principles and practice for

successful scale-up to manufacturing
S.C. Tsinontides∗ , P. Rajniak, D. Pham, W.A. Hunke, J. Placek, S.D. Reynolds
Merck Research Laboratories, Merck & Co. Inc., Sumneytown Pike WP78-204, West Point, PA 19486, USA
Received 13 October 2003; received in revised form 9 April 2004; accepted 9 April 2004

Available online 25 June 2004

Abstract
Freeze Drying involves transfer of heat and mass to and from the product under preparation, respectively, thus it is necessary to
scale these transport phenomena appropriately from pilot plant to manufacturing-scale units to maintain product quality attributes.
In this manuscript we describe the principal approach and tools utilized to successfully transfer the lyophilization process of a
labile pharmaceutical product from pilot plant to manufacturing. Based on pilot plant data, the lyophilization cycle was tested
during limited scale-up trials in manufacturing to identify parameter set-point values and test process parameter ranges. The
limited data from manufacturing were then used in a single-vial mathematical model to determine manufacturing lyophilizer
heat transfer coefficients, and subsequently evaluate the cycle robustness at scale-up operating conditions. The lyophilization
cycle was then successfully demonstrated at target parameter set-point values.
© 2004 Elsevier B.V. All rights reserved.
Keywords: Freeze drying; Lyophilization; Heat and mass transfer; Process scale-up; Pilot plant; Manufacturing; Pocket logger

1. Introduction at very low pressure to sublime the solvent and re-

move it from the formulation. Once the water is re-
Lyophilization is commonly used in the pharma- moved, the product vials are sealed under vacuum or
ceutical and biotechnology industries to improve the an inert gas head space (i.e., N2 , Ar). The resulting
stability of formulations. The active pharmaceutical highly porous cake has low moisture content and can
ingredient and accompanying excipients are first sol- be stored over extended periods of time at the desig-
ubilized in a solvent (usually water), and the solution nated storage conditions until its intended use.
is rendered sterile by filtering it through 0.2 ␮m or Over the past few decades, the investigation of the
equivalent sterilizing grade filters. The sterilized solu- fundamental physical phenomena occurring in each
tion is filled into vials, then loaded into a lyophilizer step of freeze drying has led to producing stable and
where the solution is frozen, and subsequently heated elegant freeze-dried pharmaceutical dosage forms. A
comprehensive review of the principles and practice
of freeze drying was published by Nail and Gatlin
Abbreviations: AVG, average value; N, number of samples (1992). Extensive work in studying the physical
tested to determine average value chemistry, and transport phenomena during freezing
∗ Corresponding author. Tel.: +1 215 652 2651;
and primary drying (MacKenzie, 1975; Pikal et al.,
fax: +1 215 652 4088.
E-mail address: stelios tsinontides@merck.com
1983a, 1984, 1990; Pikal, 1985, 1990a,b; Franks,
(S.C. Tsinontides). 1990) paved the way for designing appropriate formu-

0378-5173/$ – see front matter © 2004 Elsevier B.V. All rights reserved.
doi:10.1016/j.ijpharm.2004.04.018
2 S.C. Tsinontides et al. / International Journal of Pharmaceutics 280 (2004) 1–16

Nomenclature
A, B constants in Eq. (3)—values taken from Pikal (1985) defined in Table 4
C, D constants in Eq. (2) defined in Table 4
Dwin,e Effective pore diffusivity in the cake defined in Table 3 (m2 /s)
h overall heat transfer coefficient as defined in Eqs. (1) and (4), defined in Table 4 (W/m2 K)
hc heat transfer coefficient by direct conduction between shelf and glass vial, defined in Eq. (2)
hr heat transfer coefficient by radiative heat from upper and bottom shelves, defined in Eq. (2)
hg heat transfer coefficient by convective heat transfer from bottom shelf, defined in Eq. (3)
kdes Rate constant of desorption step defined in Table 3 (s−1 )
K, amax Langmuir equilibrium parameters defined in Table 3
P chamber pressure (Pa)
T temperature (◦ C)
Tice ice temperature is the maximum temperature of frozen amorphous solution during primary
drying (◦ C)
Tc collapse temperature of an amorphous system; considered to be equal to Tg for design and
scale-up purposes (◦ C)
Tg glass transition temperature of an amorphous system defined in Table 3 (◦ C)
Tg,solid , Kmix Gordon–Taylor parameters defined in Table 3

Greek symbols
λcake effective thermal conductivity of the cake defined in Table 3 (W/m K)

lations and lyophilization cycles. At the same time, tions. Based on the collective results from experiments
considerable advancements in modelling of freeze and model simulations, the final lyophilizer set points
drying have yielded mathematical models to describe and window of operation were determined. The pro-
the dynamic behavior of primary and secondary dry- cess parameter set-point values were then successfully
ing of pharmaceutical solutes (Liapis and Bruttini, demonstrated in a final run.
1994, 1995; Liapis et al., 1996; Sadikoglu and Liapis,
1997). The recent advancements in lyophilization
technology notwithstanding, issues of scale-up have 2. Experimental approach
attracted limited attention. Scale-up is a critical step
that determines timely product commercialization, The lyophilization cycle was developed from stud-
and regulatory guidances expect that the manufactur- ies in a laboratory-scale lyophilizer1 . Once the set
ing process be demonstrated at manufacturing-scale points were determined (e.g., temperature, pressure,
to produce uniform product within the batch, with duration of drying steps at different temperatures, rate
desired physical and chemical attributes (FDA of temperature increase), the cycle robustness was
Guideline, 1987). evaluated by varying temperature and pressure around
This manuscript presents a methodology utilized to set points to establish operating ranges. The optimized
successfully scale-up the lyophilization cycle of a la- cycle was then applied successfully to a pilot plant
bile pharmaceutical formulation from pilot plant to lyophilizer with the manufacture of clinical and sta-
manufacturing. A series of scale-up runs were exe- bility batches with no adjustment to cycle set-point
cuted to evaluate a proposed process window of op- values based on comparable lyophilizer dimensions.
eration. Data from the scale-up runs were then used
in a mathematical model to evaluate the robustness
of the lyophilization cycle to likely operating condi- 1 VirTis Benchmark 1000 lyophilizer.
 S.C. Tsinontides et al. / International Journal of Pharmaceutics 280 (2004) 1–16 3

Fig. 1. Representation of Pocket Logger placement in a frame to monitor product temperature of vials at different locations.

Upon transferring the process to Manufacturing, the lyophilization cycle and its operation was found to be
development scale-up batches to demonstrate the cy- robust. Its small size and portability allowed its use
cle were limited due to plant availability and active to probe vials at many locations on different shelves
raw material supplies. Initially, a small number of without influencing the rate of heat transfer to the
vials containing the formulation were placed into the product being monitored. Fig. 1 demonstrates how the
lyophilizers among placebo vials during the shake- Pocket Logger was placed among filled vials in a tray
down and qualification of the manufacturing lyophiliz- to monitor product temperature at different locations
ers. A placebo formulation made of lactose and su- inside the lyophilizers. Fig. 2 shows the placement of
crose had been developed to match the formulation the thermistor inside a product vial. Thermistors were
physical properties by closely matching Tg and solid placed in product vials carefully ∼3 mm above vial
content (Tg = −25 ◦ C, 25% solids content). bottom and centered for consistency of measurement
During the placebo trials Pocket Loggers2 were in the different vials and studies. The probed vials
evaluated to monitor product temperature during were frozen outside the lyophilizer at a similar rate as
lyophilization. Pocket Loggers are used widely to the product to secure the probe in place. The pocket
monitor temperature, humidity or other variables dur- loggers and probed vials were then placed into desig-
ing material shipment and other applications. In all nated locations inside the frames just prior to loading
of our studies we used Pocket Logger model XR-440 the frames into the lyophilizer.
(size 12.0 cm × 6.1 cm × 2.3 cm), which was quoted The accuracy of the thermistor probe used on the
to have a wide range of operating temperature (−40 Pocket Logger was evaluated against thermocouple
to 60 ◦ C), and good accuracy (±0.15 ◦ C). The pocket probes used during development in laboratory-scale
logger was tested at the extreme conditions of the lyophilizers (thickness of thermocouple probe was
∼0.25 mm). Fig. 3 compares the readings of ther-
mistor and thermocouple probes in a laboratory-scale
2 Pocket logger is a product of Pace Scientific Inc. In our studies lyophilizer study. The two probe readings were
we used Pocket logger model XR-440, and temperature thermistor slightly different toward the end of primary drying
model PT-907. with the thermistor probe indicating a faster rise of
4 S.C. Tsinontides et al. / International Journal of Pharmaceutics 280 (2004) 1–16

Table 1
Timeframe of technical transfer activities in manufacturing
Time frame Activity

Months 0–11 Placebo runs; vials containing active

formulation dispersed among placebo vials
for testing lyophilization equipment operation
Trial 1: preliminary scale-up run (1/2 placebo
1/2 active vials)
Month 15 Trial 2: scale-up #2
Trial 3: scale-up #3
Month 18 Trial 4: demonstration run

product temperature than the thermocouple probe. The

difference in the readings is attributed to the probe
sizes. The thermistor probe diameter is large and the
temperature reading represents that of a larger area
around it, hence somewhat less accurate in measuring
local temperature than a thermocouple. The somewhat
lower sensitivity of the thermistor to measure local
temperature was taken into account when interpreting
Fig. 2. Location of the thermistor probe inside a product vial. data generated at the manufacturing scale.
The technology transfer process for scale-up was
carried out in three phases, as shown in Table 1. The

80

Lyo Shelf Temperature

60
Thermocouple Probe

40 Thermistor Probe
Temperature (˚C)

20

-20

-40

-60
0 10 20 30 40 50 60 70 80
Time (hr)

Fig. 3. Product temperature during a lyophilization cycle measured by a thermocouple probe (∼0.25 mm diameter) and a data logger
thermistor probe (∼2.85 mm diameter).
 S.C. Tsinontides et al. / International Journal of Pharmaceutics 280 (2004) 1–16 5

Table 2
Lyophilization parameters during the technology transfer batches in manufacturing
Parameters Preliminary scale-up Trial 2 Set-point values Trial 3 Set-point values Trial 4 Demonstration run
set-point values
Primary drying
Temperature (◦ C) −21 −22 −20 −20
Pressure (Pa) 11.0 9.5 12.5 11.0
Secondary drying
Temperature (◦ C) 40 39 41 40
60 59 61 60
Pressure (Pa) 5.0 3.5 6.5 5.0

first phase consisted of performing placebo runs fol- ±1.5 Pa) had significant impact on the process because
lowed with an initial trial (Trial 1) to identify the pre- of high solid content in the product and consequently
liminary lyophilization cycle set points at manufactur- high resistance of the dried cake to the vapor flow. The
ing scale, shown in Table 2. In the second phase (Tri- results from Trials 2 and 3 were not sufficient to read-
als 2 and 3), the set point parameters were challenged ily identify final lyophilization parameter set points,
to determine a process operating window and the final hence a theoretical model was employed to evaluate
scale-up set-point values. In Trial 2, the pressure and lyophilization cycle robustness at manufacturing scale
shelf temperature set points were held below target and assist in determining the appropriate parameter
to test conditions of low heat transfer to the product. set points. The final parameter set points, shown in
In Trial 3, the cycle was tested under aggressive heat Table 2, were demonstrated in Trial 4. Fig. 4 shows
transfer conditions. The lyophilization cycle ramps the locations of product vials with thermistors during
and duration of the steps were not varied. Relatively the scale-up runs. Vials were probed throughout the
small changes in the operating conditions (±1 ◦ C and cabinet to obtain a comprehensive picture of product
drying within a lyophilization load. Vials at the center
and inside the shelf and around shelf periphery with
Trial 2 Trial 3 and without contact with metal frames were probed.
Frame 1 Frame 1 L1
Moisture testing was done on vials from similar loca-
tions at the end of lyophilization.
L2 L1 L3 L2 L3

3. Theoretical approach
Frame 2 N2 Frame 2
Because of the limited availability of experimen-
M2 M1, N1 M3 N2 N1 N3
tal data at manufacturing scale, a theoretical model
predicting primary drying in a single vial (Rajniak
et al., 1999) was used to fully explore the robust-
N3 ness of the lyophilization cycle and to set final pa-
Frame 3 Frame 3 K2
rameter set points. Model equations and boundary
conditions, which include energy balances in the
J1
frozen and dried regions of the cake, mass balances
in the dried cake and at the moving interface, and
K1 J2 J1 K1
adsorption–desorption equilibrium and rate expres-
Shelf Front Shelf Front
sions, are found elsewhere (Rajniak et al., 1999,
Fig. 4. Top view of thermistor and moisture mapping locations of 2000), and are based on those developed by Liapis
product vials in the lyophilizer for Trials 2 and 3 (locations are and coworkers (Liapis and Bruttini, 1994, 1995;
not limited to a single shelf). Liapis et al., 1996; Sadikoglu and Liapis, 1997). The
6 S.C. Tsinontides et al. / International Journal of Pharmaceutics 280 (2004) 1–16

Table 3
Model parameter values
Parameter description Value

Effective thermal Conductivity of the cake, λcake (W/m K) 0.04

Effective pore diffusivity in the cake, Dwin,e (m2 /s) 0.0011
Langmuir equilibrium parameters K = 355, amax = 1256 mole/m3
Rate constant of desorption step, kdes (s−1 ) 3.3 × 10−5
Gordon–Taylor parameters Tg,solid = 110 ◦ C, Kmix = 5.6
Glass transition temperature, Tg (◦ C) −25

model has a number of parameters specific to product shelves, and hg the convective heat transfer from the
physical properties, product package configuration, bottom shelf to the vial via the gas located between
and lyophilizer properties, and were calculated or es- vial bottom and shelf. The first two contributions are
timated based on readily available product/process in- independent of operating pressure, and the combined
formation. The values of these parameters are shown contribution of the two routes of heat transfer was
in Tables 3 and 4. The effective pore diffusivity for assumed to vary linearly with shelf temperature in the
water transport in the dried cake and the overall heat limited window of temperatures considered as follows
transfer coefficient between the shelf and bottom (Tsinontides et al., 2001):
of the vial were first obtained by fitting the model
hc + hr = C + DT (2)
predictions to laboratory experimental data. It was
assumed that the effective pore diffusivity would not T is the absolute temperature of the shelf, and C and
change upon scale-up to a manufacturing lyophilizer. D are empirical constants specific to the vial/shelf
The assumption follows from the equivalent freezing configuration. The empirical relation (2) is used to
step in all cycles and consequently the same structure express the temperature dependence of the “fictitious”
of the frozen and dried product was expected. How- radiation heat transfer coefficient hr (based on the lin-
ever, it was expected that heat transfer coefficients ear temperature driving force assumption; Peters and
could vary among different units, as a result of dif- Timmerhaus, 1981). The gas contribution of heat
ferent lyophilization design and size. Therefore, the transfer coefficient, hg , is dependent on the operating
scale-up data were used to determine heat transfer chamber pressure. hg is increasing with increasing
coefficients for modeling primary drying at the man- pressure, and its form was adopted from Pikal (1985).
ufacturing units. The overall heat transfer coefficient
AP
between the product and the shelf, h, was expressed hg = (3)
in the form proposed by Pikal (1985), as the sum of 1 + BP
three contributing factors: The value of constant A is not specific to the vial/shelf
h = h c + hr + hg (1) configuration, but the value of B is, and values of
A and B for tubing and molded vials are found in
hc denotes heat transfer conduction to the product from Pikal (1985). Using Eqs. (2) and (3), the heat transfer
the shelf through the glass vial/shelf contact points, coefficient takes the following expression in terms of
hr denotes the radiative heat from the top and bottom the independent operating variables, shelf temperature

Table 4
Heat transfer coefficients for manufacturing lyophilizers
Trial (#) Shelf temperature Chamber pressure Constant C Constant D Heat transfer
(◦ C) (Pa) (W/m2 K) (W/m2 K2 ) coefficienta , h (W/m2 K)
2 −22 9.5 −267.4 1.08 11.5
3 −20 11.5 −267.4 1.08 15.0
a Calculated from Eq. (4) using A = 1.04 W/m2 K Pa and B = 0.027 Pa−1 (Pikal, 1985).
 S.C. Tsinontides et al. / International Journal of Pharmaceutics 280 (2004) 1–16 7

-15

K1; Edge Vial

-20
-22 ˚C, 9.5 Pa

-25
Temperature (˚C)

-30
Tg'=-25 ˚C

-35
N1; Center Vial

-40
Shelf Tg' J1: Active K1: Active

L1: Active L2: Active L3: Active M2: Active

-45
M3: Active N1: Active N2: Active N3: Active

-50
0 10 20 30 40 50 60
Time (h)

Fig. 5. Pocket logger thermistor (product) temperature during primary drying for Trial 2 in manufacturing at conservative primary drying
conditions.

-10
L1; Edge Vial

-15

-20 ˚C, 12.5 Pa

-20

-25
Temperature (˚C)

-30

K2; Center Vial Tg'=-25 ˚C

-35

-40

Shelf K2: Active

-45
Tg' L1: Active

-50
0 10 20 30 40 50 60
Time (h)

Fig. 6. Pocket logger thermistor (product) temperature during primary drying for Trial 3 in manufacturing at agressive primary drying
conditions.
8 S.C. Tsinontides et al. / International Journal of Pharmaceutics 280 (2004) 1–16

and chamber pressure. were conducted with primary drying at ±1 ◦ C from

AP −21 ◦ C and ±1.5 Pa from respective pressure set
h = C + DT + (4) points.
1 + BP
Figs. 5 and 6 show primary drying temperature
Eq. (4) contains four constants whose evaluation re- profiles of product vials from Trials 2 and 3, re-
quires, in principle, at least four experiments at differ- spectively (probe locations shown in Fig. 4). The
ent operating conditions. However, in the present case, conditions of primary drying and the formulation
only two experiments were performed. Since constants glass transition temperature, Tg (−25 ◦ C), are shown
A and B are dependent on the operating pressure and on the figures. Fig. 5 shows that the product inside
vial type and not as much on the properties of the the manufacturing lyophilizer dried at two generally
lyophilizer shelves, their respective values were bor- distinct rates. Vials in the interior of the shelves
rowed from Pikal (1985) and listed in Table 4. The dried slower than vials close to the shelf periphery
values of constants C and D were then determined by (edge vials). The profiles for center and edge vials
fitting the model predictions to experimental temper- are marked in Fig. 5 (N1 and K1 for center and
ature transient of centered vials from Trials #2 and edge, respectively) with the remaining vials drying
#3 (shown in Figs. 5 and 6). Table 4 shows the cal- at intermediate rates. The marked variation in the
culated overall heat transfer coefficients at the oper- product temperature transients, and hence apparent
ating conditions of Trial 2 and 3. As expected, at the product drying rates, were typical in all trials and
more aggressive conditions (Trial 3), the overall ef- depended mostly on vial location on a shelf. Fig. 6
fective heat transfer coefficient is higher (15.0 versus and subsequent figures (Figs. 7, 12–14) show repre-
11.5 W/m2 K). sentative center and edge vial product temperature
trends to demonstrate the range of drying rates within
a lyophilizer. The variability of temperature transients
4. Results and discussion (and hence of drying rates) during primary drying
within a lyophilization load underlines the importance
4.1. Experimental results of spatial variability of heat transfer in large units.
A lyophilization cycle can not be too aggressive be-
The pilot plant lyophilization cycle is shown in cause product at the periphery of a lyophilizer shelf
Fig. 3. Primary drying was conducted at −20 ◦ C could collapse due to aggressive heating (by melting
and 11.0 Pa, and secondary drying at 40 and 60 ◦ C the frozen solution), nor too conservative because
with pressure held at 5.0 Pa. The final product was product at shelf centers could collapse upon increas-
demonstrated to have desired physical and chemi- ing shelf temperature due to incomplete primary
cal attributes (i.e., white to off-white cake appear- drying.
ance, moisture content of less than 2.5% (w/w), low The thermal results in Figs. 5 and 6, along with the
levels of degradates) in the pilot plant lyophilizer; physical appearance and moisture results of the final
hence the cycle was determined to be appropriate product (Fig. 8) showed that the lower temperature and
for scale-up. The first scale-up batch (Trial 1) was pressure during primary drying in Trial 2 resulted in
conducted using the pilot plant demonstrated cycle some minor product partial collapse upon increasing
set points. This was the first time a manufacturing shelf temperature. Approximately 5% of product vials
lyophilizer was run with significant amount of product in Trial 2, predominantly located near shelf centers,
in the lyophilizer (1/2 of the lyophilizer had product had signs of frozen solution melt at the bottom center
and 1/2 had placebo formulation). The final product of the cakes. The minor partial collapse was evident
attributes were satisfactory, and upon review of the by a different color and texture, crescent-shaped, area
temperature trends of product vials, the preliminary at the lower part of cakes. The recorded temperature
set points of the lyophilization cycle were set, shown of center vials (e.g., vial N1 in Fig. 5) showed that the
in Table 2. The primary drying set-point temperature product located close to shelf centers did not complete
was decreased from −20 to −21 ◦ C, but the other set primary drying. Several thermistors showed product
points were not changed. Therefore, Trials 2 and 3 temperature at or below the glass transition tempera-
 S.C. Tsinontides et al. / International Journal of Pharmaceutics 280 (2004) 1–16 9

65
61 ˚C, 6.5 Pa

55

45
41 ˚C, 6.5 Pa

35 L1; Edge Vial

Temperature (˚C)

25
K2; Center Vial
15

-5

Shelf K2: Active L1: Active

-15

-25
53 57 61 65 69
Time (h)

Fig. 7. Pocket logger thermistor (product) temperature during secondary drying for Trial 3 in manufacturing at aggressive secondary drying
conditions.

3.0

Max. 2.4
2.5
Max. 2.3
Max. 2.3
Max. 2.2

2.0 Max. 1.9 Max. 1.8 Min. 2.1

Min. 2.0
Moisture, % w/w

Min. 1.8
Min 1.6
1.5 Min. 1.4
Min 1.6

1.0
AVG=1.7 AVG=1.7 AVG=1.6 AVG=2.0 AVG=2.2 AVG=2.2

0.5
N=122 N=179 N=182 N=9 N=10 N=7

0.0
Trial 2 Trial 3 Trial 4 (Dem. Run) Pilot Plant 1 Pilot Plant 2 Pilot Plant 3

Fig. 8. Final product moisture in manufacturing scale-up (Trials 2–4) and pilot plant batches.
10 S.C. Tsinontides et al. / International Journal of Pharmaceutics 280 (2004) 1–16

ture of the product,3 Tg , at the end of the primary dry- transfer characteristics of the lyophilization units, an
ing step (center vials in Fig. 5). If the temperature of issue addressed later in the manuscript.
the product is much lower than shelf temperature and
below Tg , then the sublimation of water was likely not 4.2. Theoretical results—lyophilization cycle
complete in the entire cake. Increase of shelf tempera- robustness evaluation
ture to proceed to secondary drying caused melting of
portions of frozen solutions still undergoing sublima- The primary drying set points could not be read-
tion. The location of the collapsed region at the bottom ily determined because of product partial collapse ob-
center of the cake was indicative of incomplete pri- served in Trial 2. Availability of manufacturing time
mary drying. Partial collapse of cakes at the upper and and of large amounts of costly materials to continue
outer surfaces is characteristic of aggressive heating. scale-up studies was limited, and the determination of
In Trial 3 all product vials completed primary drying primary drying set points could not be done experi-
before the end of step, as demonstrated in Fig. 6. The mentally. Furthermore, the shelf temperature of man-
recorded product temperature at all locations spanning ufacturing lyophilizers oscillated around the set point
from shelf centers (vial K2) to shelf edges (vial L1) during primary drying (evident by the measured prod-
were well above Tg and close to the shelf temperature uct temperatures in Figs. 5 and 6). Hence, the available
by end of primary drying step. experimental data were incorporated into a mathemat-
Fig. 7 shows the respective recorded temperatures ical model to evaluate lyophilization cycle robustness
of product at center and edge vials during secondary and determine final parameter set points.
drying in Trial 3. As expected, similar variability of Fig. 9 shows the ice temperature (frozen solution
heat transfer within the lyophilizer shelves was ob- temperature) trend during primary drying at the ex-
served at the secondary drying temperatures. The vials perimental conditions of scale-up Trials 2 and 3. The
at shelf centers approached shelf temperatures (more simulation determines the increase of ice temperature
effectively heated), but vials at the shelf periphery re- and the duration of primary drying for a given target
mained somewhat at lower temperature. The impact fill volume (height of the frozen solution). Collapse is
of secondary drying conditions on the product was de- predicted if the ice temperature exceeds the collapse
termined based on the final moisture results, shown in temperature during primary drying, or the predicted
Fig. 8. Fig. 8 shows the average moisture (AVG), the primary drying duration exceeds the actual duration
number of samples tested (N), and range of moisture of the step used in manufacturing. The simulations
values for each scale-up and pivotal pilot plant batches. for Trial 3 conditions showed that primary drying was
The average moisture results of scale-up batches were completed successfully with maximum ice tempera-
comparable, ranging between 1.6 and 1.7%. The mi- ture of −26.2 ◦ C in 42.9 h, less than the experimental
nor partial collapse observed in ∼5% of product in allocated time of 52 h. On the contrary, the simula-
Trial 2 did not affect the average moisture value due to tions for Trial 2 conditions predicted a maximum ice
the larger number of vials tested. However, the limited temperature of −28.4 ◦ C after 52.4 h. The results in
number of partially collapsed product caused greater Fig. 9 are in agreement with the experimental prod-
variability in moisture values. Overall, the moisture of uct temperature trends in Figs. 5 and 6, which showed
product from manufacturing was lower than that from that primary drying was not completed in Trial 2, and
pilot plant (∼1.6% compared to ∼2.1%). Secondary thus the limited amount of collapsed product and high
drying conditions for batches in the pilot plant and product moisture values.
manufacturing were very similar, thus product from Fig. 10 shows the ice temperature trend during
both scales was expected to have similar moisture con- primary drying at the Trial 3 operating conditions.
tent. The difference is attributed to the different heat The amplitude and frequency of shelf temperature
oscillations were closely matched to those observed
3 The collapse temperature, T , is usually determined to be
during manufacturing with temperature oscillating
c
slightly higher than the Tg of the frozen solution. For process
+1.3 ◦ C/−0.6 ◦ C from set-point value at a frequency
development considerations, a conservative approach has been of ∼0.6 h−1 . The oscillations shifted the average shelf
adopted to consider Tc equal to Tg . temperature to a slightly higher value from set-point
 S.C. Tsinontides et al. / International Journal of Pharmaceutics 280 (2004) 1–16 11

Trial 3 (Tshelf = - 20˚ C, P= 12.5 Pa)

-20
Trial 2 (Tshelf = - 22˚ C, P= 9.5 Pa)

Tg'= - 25˚C
-25
Simulated Ice Temperature for Trial 3
Temperature (˚ C)

-30
(42.9 h, - 26.2˚C )

(52.4 h, - 28.4˚C)
-35
Simulated Ice Temperature for Trial 2

Trial 2&3 Experimental PD Duration = 52 h

-40

0 10 20 30 40 50 60
Time (h)

Fig. 9. Simulations of primary drying duration and frozen solution temperature (ice temperature), Tice , for Trials 2 and 3.

making Trial 3 conditions worst-case scenario to sim- plitude than shelf temperature as a result of the heat
ulate. The predicted ice temperature oscillations in transfer resistance limitations. Furthermore, as shown
Fig. 10 matched closely the experimental measure- in Fig. 10 the predicted duration of primary drying
ments of product oscillations in Figs. 5 and 6. The with the oscillations (42.0 h) was smaller than with-
frozen product temperature oscillated at smaller am- out the oscillations (42.9 h) due to the slight upward

-15

Trial 3 Oscillating Shelf Temperature (Tshelf = - 20 ˚C, P= 12.5 Pa)

-20

Tg'= - 25˚ C
-25
Temperature (˚C)

-30
(42.0 h, - 26.2˚C)

Ice temperature for oscillating Tshelf

-35

Trial 3 Experimental PD Duration = 52 h

-40

0 10 20 30 40 50
Time ( h )

Fig. 10. Evaluation of primary drying robustness with oscillating shelf temperature for Trial 3.
12 S.C. Tsinontides et al. / International Journal of Pharmaceutics 280 (2004) 1–16

shift of the averaged shelf temperature. The maxi- perature oscillation at −21 ◦ C with pressure setting
mum ice temperature with the oscillations present at 9.5 Pa. Primary drying duration at these conserva-
was −25.5 ◦ C, and thus primary drying was expected tive heat transfer conditions was predicted to be 46.7 h
to be completed without affecting product quality with maximum frozen solution temperature reaching
attributes. −27.0 ◦ C. At these conditions, the duration of pri-
Upon a complete evaluation of all process manu- mary drying was well within the actual set duration of
facturing data from scale-up trials, the solution tar- 52 h, thus considered to be in the safe operating win-
get fill volume was decreased by ∼2% (from 5.9 to dow. The middle curve predicted the duration of pri-
5.8 ml). However, the lyophilization cycle set points mary drying and maximum ice temperature at the pro-
were tested at the scale-up trials’ target fill volume posed operating conditions for primary drying, −20 ◦ C
to simulate worst-case scenario in terms of filling ca- and 11.0 Pa. Primary drying was predicted to require
pability. Furthermore, primary drying was simulated about 42.3 h with maximum ice temperature reaching
with shelf oscillations at extreme conditions of oper- −25.9 ◦ C, below the maximum allowable temperature
ation. Fig. 11 compares the results of such simula- of −25 ◦ C. Additional simulations were performed at
tions to those at the final primary drying set points of different target fills. In all three cases, the duration of
−20 ◦ C and 11.0 Pa. The top curve corresponds to a primary drying was predicted to be about ∼1 h shorter
shelf temperature oscillation at −19 ◦ C with pressure with an ice temperature decrease of ∼0.1–0.2 ◦ C at
setting at 12.5 Pa. Primary drying duration at such ag- the target fill volume of 5.8 ml.
gressive heat transfer conditions was predicted to be
only 38.5 h with frozen solution temperature reaching 4.3. Final lyophilization cycle demonstration in
−24.9 ◦ C. Operation at this condition would be risky manufacturing
with possibility that some limited number of prod-
uct vials could show partial collapse. This condition Based on the experimental data and theoretical pre-
was thus considered the upper bound of the operat- dictions, the primary drying set points were set to
ing range for temperature and pressure during primary −20 ◦ C and 11.0 Pa. These conditions were robust to
drying. The bottom curve corresponds to a shelf tem- likely deviations from operating set points (±1 ◦ C;

-20

Ice temperature for oscillating Tshelf= - 19˚C an d P = 12.5 Pa

Tg'= - 25˚ C
-25
Temperature (˚C)

-30

Ice temperature for oscillating Tshelf= - 21˚C an d P = 9.5 Pa

-35
Ice temperature for oscillating Tshelf= - 20˚C a nd P = 11 Pa

-40 Experimental PD Duration = 52 h

0 10 20 30 40 50
Time (h)

Fig. 11. Evaluation of primary drying robustness at different oscillating shelf temperatures and chamber pressures.
 S.C. Tsinontides et al. / International Journal of Pharmaceutics 280 (2004) 1–16 13

-10
L1; Edge Vial

-15

-20 ˚C, 11.0 Pa

-20

-25
Temperature (˚C)

-30
H1; Center Vial Tg'=-25 ˚C
-35

-40

-45 Shelf Tg' H1: Active L1: Active

-50
0 10 20 30 40 50 60
Time (h)

Fig. 12. Pocket logger thermistor (product) temperature during primary drying for the demonstration run (Trial 4) in manufacturing at final
set points.

±1.5 Pa) to ensure complete primary drying for all with the earlier scale-up batches, center vials attained
product vials within the 52 h step duration and keep higher temperature than edge vials during secondary
the ice temperature below the glass transition temper- drying. The moisture results from the demonstration
ature of −25 ◦ C. The simulation results in Figs. 9–11 run are shown in Fig. 8, with an average value of 1.6%,
showed that the most influential operating parameter in line with the other scale-up trials. The product had
for heat transfer to the product vials was the shelf tem- no visible signs of collapse, and met all physical and
perature with chamber pressure having a lesser effect. chemical product specification attributes.
Figs. 12 and 13 show representative product tem-
perature trends from Trial 4 (demonstration run) shelf 4.4. Pilot plant versus manufacturing lyophilizers
center and edge vials during primary and secondary
drying, respectively. Fig. 12 shows that all vials com- Fig. 8 moisture results showed that the product from
pleted primary drying well in advance of the com- manufacturing had lower moisture from the pilot plant
pletion of the step, demonstrated by the trend of the batches despite the similarity of the lyophilization cy-
center vial (vial H1). The temperature of the frozen cles applied at the two facilities. Therefore, an ex-
solution of center vial H1 started to rise faster af- perimental active batch was manufactured in the pilot
ter ∼38 h, indicating completion of primary drying at plant and the product temperature was monitored us-
about 38–40 h. The experimental completion of pri- ing Pocket Loggers, as was done in manufacturing.
mary drying is in good agreement with the theroretical The product temperature results from secondary dry-
predictions in Fig. 11. The duration for primary dry- ing are shown in Fig. 14, and show that the product
ing was predicted to be ∼41 h, when adjusted for a fill temperature during the two steps of secondary dry-
volume of 5.8 ml. Fig. 13 shows the secondary dry- ing in the pilot plant unit was lower than that attained
ing product temperature trends for the same vials. As in a manufacturing lyophilizer. The difference is at-
14 S.C. Tsinontides et al. / International Journal of Pharmaceutics 280 (2004) 1–16

70
H1; Center Vial
60 ˚C, 5.0 Pa
60

50

40 ˚C, 5.0 Pa
40
L1; Edge Vial
Temperature (˚C)

30

20

10

-10
Shelf H1: Active L1: Active

-20

-30
53 57 61 65 69
Time (h)

Fig. 13. Pocket logger thermistor (product) temperature during secondary drying for the demonstration run (Trial 4) in manufacturing at
final set points.

70
H1; Center Vial
60 ˚C, 5.0 Pa
60

50

40 ˚C, 5.0 Pa
40

L1; Edge Vial

Temperature (˚C)

30

20

10

-10

-20 Shelf L1: Active H1: Active

-30
53 57 61 65 69
Time (h)

Fig. 14. Pocket logger thermistor (product) temperature during secondary drying for a batch manufactured in the pilot plant at final set points.
 S.C. Tsinontides et al. / International Journal of Pharmaceutics 280 (2004) 1–16 15

tributed to different heat transfer characteristics of the cycle at manufacturing was determined and demon-
lyophilizers, with the manufacturing lyophilizer in the strated experimentally. Freeze drying cycle scale-up
present case having a ‘more efficient’ heat transfer co- must be based on equivalent drying rates and extent of
efficient than the pilot plant unit at secondary drying drying at the different scales. Monitoring product tem-
conditions. Such differences between lyophilizers of perature inside the lyophilizers during development
different scales are case-specific and can not be known and technical transfer activities is one methodology to
a priori. The above results underscore the importance ensure successful scale-up. Scaling-up shelf tempera-
of accounting for different heat transfer characteristics ture and chamber pressure set-points may not be ad-
of lyophilizer units as the product is moved through equate, since different units might have different heat
development to manufacturing. Freeze drying cycle characteristics, irrespective of size, thus yielding dif-
transfer must be based on equivalent drying rates and ferent rates of heat transfer to the product.
extent of drying at the different scales, especially when
product final moisture content is critical. This could be
achieved by following the drying process using prod- Acknowledgements
uct temperature (or by monitoring the moisture con-
tent of the effluent gas from the lyophilizer chamber to The successful transfer of the lyophilization cycle
the condenser) during the development and technical of this product to manufacturing would not have been
transfer activities. Scaling-up shelf temperature and possible without the close collaboration between the
chamber pressure set points might not be sufficient, Pharmaceutical R&D and Manufacturing Technical
since different units can have different heat character- Teams. Especially, we would like to thank A. Kanike
istics, irrespective of size, thus yielding different rates and K. Illig of Pharmaceutical Development for their
of heat transfer to the product. early participation in the process development efforts.
We thank J. Bak of Manufacturing Validation for sug-
gesting the possible use of Pocket Loggers during the
5. Conclusions technical transfer of the lyophilization cycle to the
manufacturing units. In addition, the following col-
Appropriate scale-up of a freeze drying process in leagues, K. Ford and D. Cross (Pharmaceutical De-
a cost effective and efficient manner involves smart velopment), M. Akar, P. Coppens, R. Hanes, D. Long,
use of experimental tools to monitor the drying pro- D. O’Connell, A. Orella, and B. Phillips (Manufac-
cess of product in limited experiments at manufactur- turing Technical Operations) were invaluable for their
ing conditions. Use of modeling can tremendously en- participation and efforts in the planning of scale-up
hance the possibility of success by evaluating the ro- activities and collecting the thermal data.
bustness of the developed manufacturing cycle around
target set points. In this manuscript the methodology
for scaling-up and transferring a lyophilization pro- References
cess of a labile pharmaceutical product from pilot
plant to manufacturing was described. Experimental FDA Guideline on General Principles of Process Validation, May
data were collected during limited scale-up trials in 1987. Current Good Manufacturing Practices Regulations for
manufacturing using Pocket Loggers to determine the Finished Pharmaceuticals. 21 CFR Parts 210 and 211.
Franks, F., 1990. Freeze-drying: from empiricism to predictability.
lyophilization set points and process operating ranges. Cryo-Lett. 11, 93–110.
The experimental data were used to calculate the ap- Liapis, A.I., Bruttini, R., 1994. A Theory for the primary and
propriate heat transfer coefficients of manufacturing secondary drying stages of the freeze-drying of pharmaceutical
lyophilizers using a single-vial freeze drying model. crystalline and amorphous solutes: comparison between
The model was then used to evaluate the robustness of experimental data and theory. Separat. Technol. 4, 144–155.
the lyophilization cycle at different operating condi- Liapis, A.I., Bruttini, R., 1995. Freeze-drying of pharmaceutical
crystalline and amorphous solutes in vials: dynamic
tions, including changes in shelf temperature, chamber multi-dimensional models of the primary and secondary drying
pressure, and vial fill volume. Based on the combined stages and qualitative features of the moving interface. Drying
experimental and theoretical work, the freeze drying Technol. 13, 43–72.
16 S.C. Tsinontides et al. / International Journal of Pharmaceutics 280 (2004) 1–16

Liapis, A.I., Pikal, M.J., Bruttini, R., 1996. Research and simulation of freeze drying. J. Parent. Sci. Technol. 39, 115–
development needs and opportunities in freeze drying. Drying 135.
Technol. 14, 1265–1300. Pikal, M.J., 1990a. Freeze drying of proteins. Part I: process
MacKenzie, A.P., 1975. Collapse during freeze drying—qualitative design. Biopharmaceuticals 3, 18–27.
and quantitative aspects. In: Goldblith, S.A., Rey, L., Rothmayr, Pikal, M.J., 1990b. Freeze drying of proteins. Part II: formulation
W.W. (Eds.), Freeze Drying and Advanced Food Technology. selection. Biopharmaceuticals 3, 26–30.
Academic Press, London, pp. 277–307. Pikal, M.J., Shah, S., Roy, M.L., Putman, R., 1990. The secondary
Nail, S.L., Gatlin, G.A., 1992. Freeze drying: principles and drying stage of freeze drying: drying kinetics as a function of
practice. In: Avis, K.E., Lieberman, H.A., Lachman, L. (Eds.), temperature and chamber pressure. Int. J. Pharm. 60, 203–217.
Pharmaceutical Dosage forms: Parenteral Medications, vol. 2, Sadikoglu, H., Liapis, A.I., 1997. Mathematical modelling of the
Marcell Dekker Inc., New York, pp. 163–233. primary and secondary drying stages of bulk solution. Drying
Peters, M.S., Timmerhaus, K.D., 1981. Plant Design and Technol. 15, 791–810.
Economics for Chemical Engineers. McGraw-Hill, Tokyo, Rajniak, P., Placek, J., Reynolds, S.D., Hunke, W.A., 1999.
pp. 635–637. Mathematical Modeling of Primary Drying. AAPS Annual
Pikal, M.J., Shah, S., Senior, D., Lang, J.E., 1983a. Physical Meeting, paper 2881, 14–18 November, New Orleans.
chemistry of freeze drying: measurement of sublimation rates Rajniak, P., Placek, J., Tsinontides, S.C., Reynolds, S.D., 2000. In:
for frozen aqueous solutions by a microbalance technique. J. Proceedings of AIChE Annual Meeting, Mathematical Models
Pharm. Sci. 72, 635–650. for scale-up of Lyophilization. Paper 284j, 16 November, Los
Pikal, M.J., Roy, M.L., Shah, S., 1984. Heat and mass transfer Angeles.
in vial freeze drying of pharmaceuticals: role of the vial. J. Tsinontides, S.C., Pham, D., Rajniak, P., Hunke, W.A., Placek,
Pharm. Sci. 73, 1224–1237. J., Reynolds, S.D., 2001. In: Proceedings of AIChE Annual
Pikal, M.J., 1985. Use of laboratory data in freeze drying process Meeting, Lyophilization—From Pilot Plant to Manufacturing
design: heat and mass transfer coefficients and computer Using Theory and Experiment. Paper 48g, 9 November, Reno.