20

Lyophilization Process Validation

Christian Bindschaedler
Serono Laboratories S.A., Aubonne, Switzerland

I. INTRODUCTION

The comprehensive validation of freeze-dried products encompasses many

topics including the installation and operational qualification of the
lyophilizers, the bacterial challenge of the sterile filter, the media simulation
studies, the validation of the filling process, and the cleaning validation [1].
This chapter will be restricted to examining the validation of the lyophi-
lization process and several aspects of the manufacturing steps preceding
freeze-drying.
The basic purpose of carrying out a validation of the manufacturing
process is to establish documented evidence that provides a high degree of
assurance that the process consistently produces a product meeting its
predetermined specifications and quality attributes. Regardless of which
type of validation approach is used, the validation of the lyophilization
process includes two complementary aspects:
Examination of the freeze-drying parameters
Examination of product characteristics
While examination of the final product is essential to ensure that the
lyophilization process performs consistently and as intended, the monitoring
of freeze-drying parameters also ensures that they are maintained within an
acceptable range and provides an additional degree of assurance that the
process is under control.
Compared to many other pharmaceutical processes, freeze-drying is
intrinsically more complex. This is because process parameters and product

Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.

 characteristics are inherently interdependent. The freeze-drying parameters
affect the product being freeze-dried, but the converse is also true. A marked
impact on the course of freeze-drying may be caused by the chemical
composition of the solution and subsequently the thermodynamic and
structural properties of the frozen solution, the load of vials or ampoules,
the geometrical characteristics of the containers, and the fill volume. The
knowledge of the interrelationship between the operating parameters and
the product freeze-drying pattern is therefore an important requisite for
successful development and validation of new products. For good control of
the freeze-drying of formulated products, the following subjects should be
mastered:
The thermodynamic and structural properties of the frozen product
The effects of the programmable freeze-drying parameters on
dependent process variables
The effects of the dependent process variables on product character-
istics
The converse effects of product characteristics on dependent process
variables
Before concentrating on process parameters (Section III), we will briefly
outline the behavior of the product during the three separate but
interdependent stages of freeze-drying: freezing, sublimation (primary
drying), and desorption (secondary drying).

II. FREEZE-DRYING FUNDAMENTALS

A. Freezing

1. Thermodynamic Requirements
Freezing of the solution is required to prepare the product for lyophiliza-
tion. This part of the process is often most critical because the porous
structure of the final product will closely reflect that of the initial frozen
product. As a consequence, the freezing will affect the progress of primary
and secondary drying, as well as the properties of the final product.
The freezing of aqueous solutions occurs in a series of steps. As the
solution of drug and excipients is cooled, a temperature is reached where
pure ice crystals are nucleated. The crystals progressively grow as the
temperature decreases and a continuous network of interstitial phase
appears in-between, wherein all the solutes are concentrated. As the cooling
goes on, a temperature is reached where no further ice is generated at the

Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.

 expense of the interstitial phase. Depending on the composition of the
product, two typical situations may then be encountered:
When the formulated solution contains essentially saline or organic
solutes that crystallize easily, the interstitial phase will crystallize out
abruptly as an eutectic or a mixture of eutectics. The crystallization
results in an immediate hardening of the frozen system, which
becomes fully rigid. At this point, the system has reached its
maximum temperature for complete solidification (eutectic point,
Te), which is a basic parameter of the freeze-drying process.
When the formulated solution predominantly contains polyols, sugars,
or polymers, the interstitial phase does not usually crystallize out
upon cooling but increases progressively in viscosity as a glass-like
system. In the case where the interstitial phase has effectively
the structure of a glass, the frozen system becomes fully rigid once
the glass transition temperature (Tg) is reached. In contrast, some
amorphous systems may show no such definite transition, but they
eventually become very stiff at low temperature, as shown by
electrical resistance studies.
Regardless of the freezing behavior of the formulated solution, it is
essential to make sure that the temperature of the product is decreased
below the temperature where complete solidification is observed. If this
condition is not respected, the incompletely frozen interstitial phase will boil
or induce pellet partial melting or collapse during lyophilization.

2. Structural Requirements
Although the thermodynamic requirement of freezing below the solidifica-
tion temperature is compulsory, this condition is not always sufficient to
guarantee an easy and satisfactory lyophilization of the product. This is
because the structure of the frozen system is affected not only by
thermodynamic factors but by kinetic ones.
It is often observed that a very rapid cooling results in small ice
crystals, whereas a slow freezing favors the formation of large crystals. The
size of ice crystals can have a dramatic effect on the course of lyophilization.
A quenching of the formulated solution in liquid nitrogen will generate a
glass-like solid with minute embedded ice crystals, which may be very
difficult to dry. On the other hand, large ice crystals, resulting in large
interconnected pores, will create a structure favorable to sublimation. The
large pores offer little resistance to water vapor flow. Thereby the drying is
accelerated and the risk of product overheating at the end of primary drying
is minimized.

Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.

 However, in some cases the facilitation of primary drying via large ice
crystals is counterbalanced by a prolongation of the secondary drying time,
owing to the reduced surface area of large pores which limits the rate of
water desorption. In some other cases, slow freezing may favor degradation
of the active substance due to pH shift and high ionic strength. There is
therefore no universal freezing scheme for obtaining an ideal frozen
structure. For each formulation, the optimal conditions should be
approached by trial-and-error during development and optimization of
the freeze-drying cycle. Simple modifications may consist in varying the
slope of the freezing ramp, or in replacing the freezing ramp by a step-down
freezing including one or several intermediate plateaus. The benefit of these
improvements is a more homogeneous temperature distribution inside the
containers. This results in a more uniform pore size, and the final product
often shows a decreased between-samples variability for residual moisture.
Despite these precautions, a slow freezing process is not always
sufficient to ensure a dimensionally stable pellet of good cosmetic
appearance. In this respect, formulated solutions that produce interstitial
metastable glasses on cooling are sometimes a source of problems. In such
systems, devitrification followed by partial erratic recrystallization of the
excipient may occur during lyophilization, thereby generating a pellet of
poor powdery appearance.

3. Thermal Treatment
The remedy to this situation is to perform a thermal treatment* of the
frozen solution. This treatment consists of a controlled rewarming of the
solution until devitrification and recrystallization of the excipient occurs,
followed by a last freezing step below the solidification temperature. A
typical excipient justifiable of such a treatment on thermodynamic grounds
is mannitol. In the absence of proteins that maintain the structural integrity
of the pellet, mannitol solutions often yield cakes of poor appearance. The
induction of mannitol crystallization by rewarming around  25
C evades
this problem and allows one to obtain elegant pellets that are easy to
lyophilize and do not shrink.
In some cases, we have found that amorphous or vitreous solutions
that do not crystallize upon rewarming or during lyophilization also benefit
from a thermal treatment, but for structural reasons. For such excipients,
the rewarming step induces some reorganization of the structure of the

*The concept of thermal treatment was introduced by L. R. Rey in 1960 (Ann NY Acad Sci
85:513–534). See also Chapter 1 of this book.

Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.

 frozen solution involving complex mechanisms, one of which may be the
disappearance of small ice crystals via molecular diffusion. The reorganized
structure is easier to freeze-dry than the initial one, offers a good safety
margin with respect to collapse, and yields pellets of good cosmetic
appearance.

B. Primary Drying

1. Thermodynamic Requirements
Once the product is adequately frozen, the next step is the removal of ice,
i.e., primary drying. During primary drying, the rate of ice sublimation is
dependent on the amount of heat supplied to the product. The temperature
of the product equilibrates as a function of two opposite effects: the transfer
of heat from the shelf or from the gaseous atmosphere to the product, and
the cooling due to ice sublimation. As the ice–vapor interface (moving front)
moves toward the bottom of the containers, the rate of ice sublimation tends
to diminish because the nascent porous matrix in the upper part of the pellet
offers some resistance to vapor flow.
As a result of the lesser ice sublimation, the temperature of the product
increases progressively during primary drying. The maximum primary
drying product temperature is attained when the sublimation front reaches
the bottom of the frozen solution, i.e., when almost all of the ice has
disappeared. At this stage, it is essential to make sure that the maximum
temperature reached by the product remains consistently lower than the
incipient melting temperature of the eutectic Tim (for crystalline systems) or
the softening or collapse temperature Tc (for vitreous or amorphous
systems). If this temperature is exceeded, the crystalline systems will undergo
partial liquefaction and the glassy-like ones will yield pellets showing
bottom collapse. Inasmuch as they generate a heterogeneous product
structure, these defects can in extreme cases jeopardize the efficacy and the
stability of the drug product because the altered areas may display different
activity and degradation profiles.
In vitreous systems, it is to be noted that the collapse temperature, Tc,
can exceed the glass transition temperature Tg by several degrees or more
[2]. The retention of pellet structure below Tc arises from the fact that when
the frozen solution passes through the glass transition temperature, it
returns from a glass to a highly viscous amorphous material. It is only when
the viscosity of this material has decreased significantly that the fluidity of
the interstitial phase is sufficient to cause collapse.

Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.

 2. Determination of Product Primary Drying Time and
Maximum Temperature
Despite the practical importance of determining the point where primary
drying terminates, there is no easy or universally recognized method to do
this. Part of the difficulty arises from the fact that the boundary between
primary and secondary drying is not clear-cut. A reason for this is that
the upper part of the cake is subject to a limited water desorption whereas
the bottom of the cake is still undergoing sublimation. Another reason is
that all of the samples do not necessarily dry at the same rate [3] because
freezing–drying is inherently a statistical process in many respects.
Perhaps the best way to determine the end of primary drying, when the
lyophilizer design of the main valve allows it to, is to perform pressure rise
tests during primary drying [4]. When this testing is not possible, the end of
primary drying can be estimated from the freeze-drying graphs according to
several approximate methods.
For process review purposes, we determine three endpoints as depicted
in Figure 1:
Change in slope of the product temperature traces
Change in slope of the chamber pressure curve (not possible with
calibrated leaks)
Change in slope of the condenser temperature trace
The three endpoints generally show an acceptable agreement. The drying
times calculated from these three estimates may be somewhat shorter than
the real duration of primary drying, but they are reliable enough for
comparison purposes.
Although the measurement of product temperature with probes
inserted nearly to the bottom of the vials or ampoules is often questioned
for many reasons, we use this method routinely to evaluate the duration of
primary drying and the maximum primary drying temperature as shown in
Figure 1. During prevalidation trials, we have run a series of more or less
extreme freeze-drying cycles for sucrose and lactose formulations. The
pellets produced were examined for lyophilization-related defects such as
melt-back, bottom collapse, or shrinkage. In parallel, differential thermal
analysis (DTA) and electrical resistance measurements (see Chapter 1) were
carried out to determine the glass transition (Tg) and the softening
temperature (Tc) of the frozen solutions. The maximum primary drying
product temperature was tabulated for each freeze-drying cycle together
with the type, frequency, and severity of cosmetic defects noted in the
pellets. In many instances, the maximum product temperature determined
from the freeze-drying trace correlated quite well (within 1–2
C) with the

Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.

 Figure 1 Determination of maximum product primary drying temperature and of
product primary drying time. (1) For a given shelf temperature cycle (a), product
temperature traces (b) will be achieved. (2) At the end of primary drying, when all the
ice has sublimed, an inflection (c) is seen on each of the product temperature traces.
(3) The temperature recorded at point c is the maximum product primary drying
temperature. During primary drying, the temperature of the product is lower than or
equal to the maximum product primary drying temperature. The secondary drying
commences after point c. (4) The transition from primary to secondary drying
corresponds to a region where transitions are driven by statistical events (d).
Therefore, for different product samples, the time to reach the inflection (c) will vary.
(5) Based on product temperature traces, the product primary drying time can be
defined as the lapse of time from the moment when the vacuum is applied until point
c is reached, i.e., time segment e for probe number 1. (6) The product primary drying
time can also be determined from the chamber pressure curve (time segment f) or
from the condenser temperature trace (time segment g).

temperature range predicted from the cross-examination of pellet appear-

ance, and of DTA and resistivity curves. As the maximum product
temperature was exceeding Tg to approach Tc, minor defects (minimal
height shrinkage, minimal radial shrinkage, or minimal bottom collapse)
were progressively appearing in a few samples. However, after the softening
temperature Tc was passed, an increasing proportion of samples displayed
significant collapse at the base of the cake. For the above formulations,
the softening temperature Tc, which corresponds to the point where a
sharp decrease in resistivity is noted on electrical resistance curves (see

Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.

 Chapter 1), was therefore the upper temperature compatible with an
acceptable product.

C. Secondary Drying
As the sublimation front moves down to the bottom of the frozen solution,
it leaves behind a porous matrix made of solutes and bound water. It is
essential that the major part of this water is removed from the matrix to
avoid liquefaction and chemical degradation of the active material upon
storage. The desorption of the bound water is usually achieved by
progressively raising the temperature of the shelves up to the temperature
at which drying is completed.
During the transition from primary to secondary drying, care should
be taken with glass-like systems to slowly increase the temperature of the
shelves, so as to avoid collapse or ‘‘retrograde’’ collapse [2]. Collapse may
occur in the frozen part of the cake, if some containers still contain
significant amounts of ice while shelf temperature is raised. Retrograde
collapse takes place in an upward direction in that part of the cake above the
sublimation front that contains no more ice. The dried part of the cake is
characterized by a glass transition temperature that increases progressively
from the Tg of the frozen solution of the Tg of the final product as the
desorption progresses. If the shelf temperature is moved too fast, the glass
transition temperature at a given moisture content will be exceeded, thus
resulting in collapse in the dried part of the cake.
The time needed to complete desorption is highly dependent on
product formulation and drying temperature. Crystalline mannitol usually
requires only a short secondary drying, because the amount of bound water
is minimal. On the other hand, glass-like systems formulated with sucrose
or lactose necessitate a prolonged secondary drying.

III. INDEPENDENT VERSUS DEPENDENT

PROCESS PARAMETERS

Besides verifying that the observed values of programmable operating

parameters are within the expected range, part of the validation exercise
consists of examining the impact of the programmable variables and of their
variations on ‘‘dependent’’ process parameters. A control of the dependent
parameters and an understanding of how they are related to the
programmed variables is essential because the dependent parameters
critically affect product properties (Figure 2).

Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.

 Figure 2 Interrelationship between independent and dependent freeze-drying
parameters and product characteristics.

A. Independent Parameters
In freeze-drying runs, there are programmable variables that can be
maintained at fixed, predetermined levels, regardless of the actual
course of the lyophilization. Under normal conditions, the temperature
of the silicon fluid circulated inside the shelves (referred to hereafter as
shelf temperature), the time phasing of the temperature plateaus (soaks),
and the ramping rate are not affected by the characteristics and load of
lyophilized product. The shelf temperature and the observed heating and
cooling rates are usually in close agreement with the programmed settings,
unless the heating or cooling capacity of the thermoregulating unit is
exceeded.

B. Dependent Parameters
In contrast with programmable operating parameters, there are variables
that cannot be assigned fixed values during lyophilization because their level
is the result of the conjoint effect of several variables on the product being
lyophilized.

1. Condenser Temperature
An example of a dependent operating parameter is condenser temperature.
During primary drying, condenser temperature equilibrates at different
levels, depending on the amount of water being sublimed. For a given
formulation and at each time point, the temperature of the condenser is

Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.

 dictated by several variables including the temperature of the shelf, the
chamber pressure, the load of product, and the time evolved since the
commencement of primary drying.

2. Chamber Pressure
Lyophilization may be performed without pressure control, i.e., under
‘‘maximum vacuum.’’ In the absence of pressure control, the pressure
generated in the lyophilizing chamber is sustained only by ice sublimation
and/or water desorption. Under these conditions, the pressure is to be
regarded as a dependent variable, as its actual level is dictated by other
parameters such as shelf temperature, time, batch size, and product
characteristics.
In contrast with the above situation, chamber pressure becomes an
independent programmable variable when a calibrated leak, or successive
leaks of different pressure levels, is performed during the lyophilization. A
usual way to generate a calibrated leak is to repeatedly inject nitrogen into
the chamber so as to maintain the total pressure within a narrow range
of values, this independent of the course of ice sublimation or water
desorption. Another method for generating a constant pressure inside the
chamber is to open and close repeatedly the valve between the condenser
and the vacuum pumps, so as to maintain the water vapor pressure within
predefined limits.
With regard to process validation, it is essential to note that when the
chamber pressure is controlled, the resultant supply of heat to the product
and the progress of lyophilization depends on the combined effects of two
intensive independent variables: temperature and pressure. This is in
contrast with the maximum vacuum process whereby shelf temperature is
the only controllable intensive variable.

3. Product Temperature
On most industrial freeze-dryers, the temperature of the product is a
‘‘dependent’’ parameter that cannot be assigned fixed predetermined values.
Product temperature results from the conjoint effects of the programmable
operating variables, of the formulation being freeze-dried, of the design of
product containers, and of the engineering characteristics of the lyophilizer.
Although technically feasible, the keeping of a constant product tempera-
ture throughout primary drying necessitates a feedback system that
regulates the temperature of the shelves as a function of the temperature
of the probes inserted in the product or, alternately, as a function of the
manometric measurement of product temperature [5]. Such regulating
systems are usually restricted to research equipment.

Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.

 4. Product Drying Time
The primary drying time of the product, which can be defined as the time
elapsed from the moment when the vacuum is created in the chamber to
the disappearance of the last ice crystals in the product, is obviously a
dependent parameter as the rate of ice sublimation is controlled by the same
factors as product temperature. This is also the case for the secondary
drying time of the product, i.e., the time elapsed from the moment when
ice sublimation is complete to the end of the cycle. During this period, water
is desorbed from the product at a rate dictated both by technical factors
and by product characteristics.

5. Shelf Temperature and Pressure Versus Product Temperature

During Primary Drying
During primary drying, the temperature of the product is dependent on
shelf temperature and on chamber pressure. The higher the temperature
of the shelf, the higher the temperature of the product will be. An increase
in chamber pressure favors the thermal exchanges at the gas–product
interface and the thermal conductivity from the shelf to the tray. More heat
is transported to the product and this results in a rise of product
temperature.
The functional relationship between product temperature, on the
one hand, and shelf temperature and chamber pressure, on the other
hand, is affected by many factors including the size and design of the
lyophilizer, the characteristics of the product, and the time evolved since
the start of primary drying. With a sucrose formulation in vials, we
have observed a maximum primary drying product temperature rise of
þ5
C when the shelf temperature was varied from  15 to þ 30
C, whereas
a pressure variation from 30 to 250 microbars generated an increase of
around þ 2.5
C. With a lactose formulation in ampoules lyophilized in
a larger freeze-dryer equipped with a plate-type condenser, the effect
of pressure was found to be predominant: þ 6.5
C for a pressure move
from 50 to 300 microbars, versus þ 1
C for a shelf temperature move from
0 to 25
C.

6. Shelf Temperature and Pressure Versus Product

Primary Drying Time
An increase in shelf temperature will unambiguously accelerate the primary
drying of the product, unless it is excessive and promotes a slowing down of
water removal consecutive to collapse or melt-back.

Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.

 The effects of pressure variations, as they occur when pressure is
controlled independently via calibrated leaks, are more complicated [6–12].
In many cases, an increase in chamber pressure favors ice sublimation and is
reflected in a shorter primary drying time because of the improved thermal
exchanges and of the higher product temperature [6,10–12].
However, a slowing down of ice sublimation will be observed if the
total pressure in the chamber becomes too close to the pressure above the
sublimation interface [9,10,12]. Indeed, for an efficient removal of water
vapor from the containers, a sufficient pressure differential must exist
between the ice–vapor interface and the chamber. The total pressure above
the ice–vapor interface is approximately equal to the saturated vapor
pressure of ice at the temperature of the sublimation front, as the headspace
contains mostly water vapor [10–12]. Therefore, the pressure gradient that
promotes water removal will no longer exist if the pressure level of the
calibrated leak exceeds the saturated vapor pressure of ice at the target
product temperature.
While a too high pressure in the chamber will prevent water removal, a
high sublimation rate will usually be observed when the total pressure in the
chamber is approximately one-fourth to one-half of the saturated vapor
pressure over ice. At fixed product temperature, a higher vacuum in the
chamber would create a larger pressure gradient between the product and
the chamber, a condition that favors evaporation. However, in many
instances this advantage is not sufficient to offset the poor thermal
exchanges associated with high vacuum, especially when the product
containers stand on trays. In actuality, the driving force for sublimation
increases at relatively low pressures (up to 0.3–0.4 millibars) when chamber
pressure is raised because the increase in ice vapor pressure resulting from
the elevation of product temperature tends to be greater than the
corresponding increase in chamber pressure [10].

7. Maximum Vacuum Versus Pressure-Controlled Lyophilization

Compared to lyophilization under maximum vacuum, the implementation
of calibrated leaks during primary drying offers a number of advantages.
The primary drying of the product is shortened, and therefore a few hours
can be saved on the overall cycle time. In the same equipment, the batch size
can be varied within wide limits with minimal effects on cycle time and on
product properties. The transposition of the cycle to another freeze-drier
and the scale-up operations are facilitated, as the control of the pressure
level minimizes the effects of conceptual differences among lyophilizers.
Noteworthy differences in drying rate will be attenuated under constant
pressure conditions. Technology transfers are also made easier. If the source

Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.

 site cycle is found to be nonoptimal for the receiving site equipment, a small
correction of chamber pressure will readily bring the wished result, with no
need to adjust shelf temperature.
Although pressure-controlled lyophilization brings significant advan-
tages, a drawback of this technique is to risk promoting collapse in delicate
formulations. This is because the assigned pressure modifies the course of
sublimation and elevates product temperature in parts of the cycle where
little evaporation would occur under maximum vacuum.

IV. SURVEY OF CRITICAL PROCESS PARAMETERS

The selection of the process parameters that need to be quantified and vali-
dated requires a knowledge of the many variables that may have an effect on
the product. There is no universal validation framework, and the critical
parameters should be identified based on the profile of the freeze-drying
cycle, as well as on the characteristics, requirements, and release specifica-
tions of the product. The present section will be restricted to give examples
of process parameters that are frequently assessed in validation reports.

A. Shelf Loading Temperature

The shelf temperature during product loading should be specified and
controlled. It should be demonstrated that the product is stable during the
storage period in the freeze-dryer prior to freezing. The loading temperature
and the eventual holding time following completion of loading should be
compatible with obtaining a homogeneous product.
Large differences in temperature between the product loaded first, on
the top shelf, and the product loaded last, on the bottom shelf, may lead to
different freezing patterns. To verify the uniformity of product temperature,
thermocouples may be placed in containers located on different shelves in
order to record the cooling and freezing pattern in the various areas.
Another possibility for verifying that the structure of ice crystals is
homogeneous is to perform a mapping of the residual moisture of the
freeze-dried product in the various parts of the lyophilizer. If the loading
conditions are not satisfactory, this should be reflected in a significant trend
for moisture content from the top to the bottom shelf.

B. Shelf and Product Freezing Rates

The rate at which the product is frozen can have a significant impact
on product quality and should be controlled. In scale-up operations,

Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.

 comparable product freezing profiles should be maintained. In critical
situations, the range of acceptable rates may be investigated and validated
by varying the slope of the freezing ramp while keeping the rest of the cycle
constant.

C. Shelf and Product Freezing Temperatures

Considering the inherent temperature variations from sample to sample and
from location to location, the product should be frozen at least 5
C below
its solidification temperature [13] as determined by DTA and electrical
resistance measurements, and a safety margin of not less than 10
C should
be observed in terms of shelf temperature.

D. Freezing Time
The freezing duration should be long enough to ensure that the whole
volume of solution within a container and the contents of all containers are
completely frozen.

E. Shelf Temperature Profile

The shelf temperature during the soaks as well as the duration of the soaks
and of the ramps should comply with the settings and should be
reproducible to within a set range from batch to batch.

F. Shelf Temperature Ramping Rate

At the beginning of primary drying, the shelf heating rate should not be too
high to promote product melting at the base of the cake. At the end of
primary drying the ramping rate should not be too high so as to lead to
collapse or retrograde collapse.

G. Product Temperature During Primary Drying

Product temperature during primary drying should be consistently
maintained below the level where incipient melting or collapse is observed.

H. Pressure During Primary Drying

If the lyophilization is performed under maximum vacuum, the pressure in
the chamber reflects the amount of water vapor from the product in transit
to the condenser. The maximum pressure during primary drying should be

Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.

 tabulated and should be reproducible to within a set range while keeping the
batch size constant. If the chamber pressure is controlled by one or several
successive calibrated leaks to maintain specific pressure ranges, then these
ranges should be reproducible and comply to within a range with the
nominal settings.

I. Pressure During Secondary Drying

For effective desorption, the chamber pressure should be sufficiently low at
the terminal stage of secondary drying. A maximum acceptable pressure
should be specified.

J. Terminal Drying and Cycle Total Duration

The cycle duration, in particular the terminal drying time, should be
substantiated with regard to all product characteristics susceptible to
be affected by under- or overdying. A time range should be specified.
If the cycle allows a comfortable safety margin with respect to residual
moisture, then the cycle can be assigned a fixed duration, with a lower and
upper time tolerance. If not, the cycle may be stopped based on a pressure
rise test. This test consists of closing temporarily the main valve and
recording the pressure rise in the chamber. The lyophilization is
discontinued if the recorded pressure rise per minute does not exceed
a maximum specified value; otherwise the cycle is continued until the test is
passed. The maximum pressure rise value must be compatible with an
acceptable level of residual moisture in the product.
For products in vials, cycles of fixed duration are easily implemented,
as the vials can be automatically stoppered. However, with ampoules it may
be desirable to have cycles of variable duration for schedule-related reasons.
An elegant way to avoid variations in freeze-drying duration is to stop the
lyophilization after a fixed time and to store the ampoules under dry sterile
nitrogen in the lyophilizer until sealing is possible.
The critical segments of the freeze-drying cycle should never be
modified in order to satisfy production planning requirements. A possible
modification is to prolong the freezing soak. Once the product is fully frozen
below its solidification point, no structural changes can be anticipated
except perhaps a minimal grain growth. However, if a variable freezing time
is allowed, the absence of resultant effects on the product should be
substantiated and a range of acceptable freezing times defined.
An alternative is to prolong the last secondary drying step. However,
prolonging secondary drying may generate a risk of overdrying and
degradation of the active molecule. This point should be carefully assessed.

Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.

 This risk can be at least partially evaded by including in the cycle a terminal
soak with a decreased shelf temperature. For example, if secondary drying is
effected at 40
C, a subsequent segment at 25
C will act in many instances as
a conservation step. The basic reason for this is that at the beginning of
secondary drying the moisture decreases quickly to reach asymptotically
a minimum value that is highly dependent on the drying temperature.
Subsequently, providing that the step at 40
C is long enough, very little
additional water will be removed at 25
C no matter how long the drying is.
If such cycle variations are allowed, documented evidence should show that
they do not affect the product.

K. Partial Batch Size

Partial loading can affect both freezing and drying. When the freezing step is
performed rapidly to low temperatures, the rate of product cooling may be
dependent on product load. Therefore, if the freezing is critical, the effects of
batch size should be substantiated with regard to shelf cooling rate and
product freezing pattern.
When the chamber pressure is controlled by calibrated leaks, the
load of product in the freeze-dryer usually plays a minimal role, if any,
on the product drying rate. However, when lyophilization is achieved
under maximum vacuum, the batch size will have some effect on the
duration of primary and secondary drying. This is because the chamber
pressure is a direct function of the amount of water vapor in transit. As
the pressure has an effect on the thermal exchanges during primary
drying and on water desorption during secondary drying, partial loading
may affect the course of both drying stages. The scenario anticipated for
partial loading is a prolongation of product primary drying time (unless
there is intimate contact between product and shelves) and a shortening of
secondary drying. Therefore, it should be verified that the excess time
required to complete primary drying does not result in an increase of the
maximum product temperature above the incipient melting or the softening
point.
While lyophilizing ampoules on trays under maximum vacuum, we
have effectively observed a prolongation of primary drying by 20–30%
under partial load conditions. However, in no case was an increase in
product temperature noticeable, even when the transition ramp to
secondary drying was started before completion of product primary drying.
With regard to product properties, it should be shown that partial
batch size does not affect susceptible properties such as residual moisture,
reconstitution time, pH, or activity.

Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.

 L. Condenser Temperature
The condenser temperature should be reproducible to within a set range of
temperatures for the same batch size.
During primary drying, condenser temperature results from the rate
at which water vapor is being condensed. Condensation rate is dependent
on the rate at which water is being sublimed from the product. Effective
sublimation can only be achieved if a sufficient temperature difference,
which corresponds to a driving force, is maintained between the product and
the condenser.
For freeze-dryers equipped with a coil condenser cooled down to
 80
C, the increase in temperature consecutive to water condensation is
usually not critical even when the sublimation is intensive. In contrast, we
have observed in the past a self-accelerated drying effect while using a
lyophilizer of the older generation fitted with a plate condenser refrigerated
at  80
C. As the sublimation of ice was becoming intensive, the chamber
pressure was rising sharply and the condenser was showing a marked
temperature elevation. The pressure rise in turn generated an increase in
product temperature, thereby further accelerating the sublimation.
The outcome of this self-accelerating process was a strong increase in
condenser temperature (up to  40
C) and an elevation of product tempera-
ture by a few degrees (up to  30
C). However, the difference of temperature
between the product and the condenser temperature was still sufficient to
ensure a very rapid drying of the product, much faster than in freeze-dryers
equipped with an effective coil condenser cooled at  80
C. The observation
that the primary drying was not impaired is consistent with the exponential
decrease of the vapor pressure of ice as temperature is lowered. A sufficient
pressure gradient for freeze-drying is established even when the difference of
temperature between the product and the condenser is not very large.
Therefore, rather than the hindrance of product drying, the actual
risk associated with condenser overloading is the elevation of product
temperature, which can result in pellet collapse. For a given target cycle, this
eventuality can be ruled out by running a high-pressure/high-temperature
cycle that challenges the capacity of the condenser and the product to resist
a high sublimation rate. This test may be helpful to define a maximum
acceptable condenser temperature below which no pellet collapse is visible.
In contrast with the situation of primary drying, a low-temperature
condenser is a great asset during secondary drying for formulations
that must be dried to very low residual moisture. As the vapor pressure
over ice is 0.5 microbars at  80
C versus 11 microbars at  60
C, the
desorption of water can be achieved in a more complete way with a
condenser cooled at  80
C.

Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.

 V. PROSPECTIVE VALIDATION OF CYCLE LIMITS
A. Definition and Aims
Process development, optimization, and prevalidation trials usually yield a
cycle believed to be close enough to the optimal freeze-drying conditions.
The selected cycle is therefore supposed to represent the target working
conditions that will be applied for future production batches. The next step
in product development may be to implement a performance qualification of
the target process, i.e., to perform a batch-to-batch consistency study using
the target cycle. However, prior to carrying out this consistency study, it is
useful to demonstrate the robustness of the target cycle with respect to
deviations from the target conditions.
The purpose of prospective validation of cycle limits is to implement
planned deviations around the target cycle by changing the programmable
operating parameters in order to examine the impact of these variations on
dependent freeze-drying parameters and on product properties.
The validation of cycle limits recovers two different but complemen-
tary approaches: vertical validation and horizontal validation. Vertical vali-
dation, the most important approach, consists of varying the shelf
temperature around its target values. The chamber pressure may also be
varied if the pressure level is independent of shelf temperature, such as in the
case of calibrated leaks.
In horizontal validation, it is the time phasing of the segments at
constant temperature that is varied. In its simplest form, horizontal
validation consists of varying the duration of the terminal secondary
drying step and examining the effects of this variation on sensitive properties
such as residual moisture, bioactivity, or reconstitution time. If the freeze-
dried formulation exhibits a significant risk of collapse, it may be well worth
varying also the duration of the primary drying plateau.
In addition to horizontal and vertical validation, a complementary
challenge of process robustness may consist in changing the duration of the
shelf temperature ramps, i.e., the heating or cooling rate, while keeping the
initial and final temperatures unchanged. With delicate formulations, this
may be useful for critical parts of the cycle such as the freezing ramp or the
transition ramp from primary to secondary drying.

B. Choice of Process Limits

The process limits chosen for the validation of the cycle limits should be
wide enough to guarantee that the target cycle is sufficiently robust to
withstand the small process fluctuations that occur routinely in industrial
freeze-dryers. For example, if the programmed shelf temperature is 30
C

Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.

 and the maximum departure from the programmed values is 2
C, then the
actual shelf temperature may be as low as 28
C or as high as 32
C. In order
to guarantee that the range from 28
C to 32
C is effectively covered in terms
of actual temperatures, the planned variations around the target shelf
temperature should be not less than 4
C.
On the other hand, the process window defined by the area between
the lower and upper limits should correspond to a possible operating range.
Therefore, the limits should not be so wide as to produce an unacceptable
product. For example, the upper limit may be chosen as the temperature
where a few pellets start to display minimal collapse, whereas the lower limit
may correspond to a temperature producing a pellet of high residual
moisture, but still within specification.
For several validations, we have implemented planned deviations of
10–15
C around the target temperature during primary drying and of
5–10
C for secondary drying. The products obtained from these extreme
cycles complied with all specifications and the range of the temperature
variations was satisfactory for most practical purposes. When the freeze-
dried product can accommodate shelf temperature variations of 10
C or
more, the lyophilization cycle can often be transposed without modifications
to another lyophilizer.
It is a common observation that for a same shelf temperature a
product may dry at somewhat different rates in two different models of
lyophilizers, and that the temperature of the product during primary drying
may be marginally higher in one piece of equipment than the other.
Consequently, if the optimal shelf temperature is found to be 0
C during
primary drying for a given lyophilizer, the shelf temperature giving an
equivalent freeze-drying trace in terms of primary drying product tem-
perature and drying time may be 5
C in other equipment. If the robustness
of the process has been demonstrated adjustment of the shelf temperature by
a few degrees should pose no major problem during a technology transfer.

C. Practical Considerations
The lyophilizer should be loaded to full or at least to half capacity in order
for the extreme cycles to be representative of future manufacturing
conditions. For this kind of validation, the selected lyophilizer may be
somewhat smaller than the ones intended to be used for commercial batches.
As the product is not to be released, the load of active product can be
completed with placebo if the rate of water evaporation from placebo
containers matches that of the active drug. The lyophilizer may be opened at
the end of freezing to assess the impact of freezing on product properties, or
at various stages of secondary drying to assay residual moisture. The latter

Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.

 operation is helpful to establish the kinetics of drying at various
temperatures and to evaluate the time required for secondary drying.
A pressure rise test may be effected each time the lyophilizer is opened in
order to correlate the pressure rise per unit of time with the residual
moisture in the pellets. Based on this correlation, a pressure rise specification
for stopping the cycle may be subsequently defined to guarantee a product
of acceptable moisture.
The products manufactured from extreme cycles may be placed on
real-time and accelerated stability studies to assess their degradation rate
compared to the target cycle. If the batches produced under extreme
conditions are shown to be stable on storage, this is a good assurance that a
small deviation from the target cycle, occasioned say by a compressor
failure, will not jeopardize the stability of the batch produced. A strong
point of prospective validation for extreme cycles is to facilitate sub-
sequently the management of cycle deviations. In this respect, the batch-
to-batch consistency approach is of little value to substantiate abnormal
cycles [13]. A validation package including three identical cycle runs will
obviously be of no use to assess a significant deviation in a critical part of
the cycle, and in the absence of justification for batch release the affected
batch should be either placed on stability or rejected.

D. Validation Program
Perhaps the ideal validation program of extreme cycles would consist of
varying the independent operating parameters one by one, in order to
determine how a punctual change impacts the freeze-drying trace and the
final product. Unfortunately, such an extensive testing is usually unreali-
zable, and therefore a minimum program must be defined.
If the lyophilization is performed under maximum vacuum, the
validation may simply consist of running a ‘‘high shelf temperature’’ cycle
and a ‘‘low shelf temperature’’ cycle as shown in Figure 3.
If parameters other than shelf temperature are believed to play a
critical role, an alternative may be to run, for example, a ‘‘low-temperature/
low-pressure/short-soaks’’ cycle followed by a ‘‘high-temperature/high-
pressure/long-soaks’’ cycle. The inconvenience is that when many changes
are performed simultaneously, the cumulation of these changes may not
always result in a ‘‘real worst case’’ because the various independent
parameters interact in a complex way.
When the freeze-drying cycle includes a calibrated leak during primary
drying, eight vertical deviations can theoretically occur, keeping the segment
duration unchanged. The shelf temperature can be normal and the pressure
too high or too low or the pressure can be normal and the shelf temperature

Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.

 Figure 3 Process window for shelf temperature. The surface area comprised
between the lower and upper limits corresponds to temperature range leading to an
acceptable product.

Table 1 Combination of Critical Cycles

Shelf temp. Pressure

High High
High Low
Low High
Low Low

too high or too low. The four other departures shown in Table 1 represent
four extreme conditions that are worth testing.
The high-temperature/low-pressure cycle is often the least critical
because it creates conditions favoring ice sublimation.
The low-temperature/low-pressure cycle theoretically results in a low
product temperature. However, the drying rate is low and there is a risk that
primary drying is not over when the temperature of the shelves is raised. In
this case, collapse or retrograde collapse may occur. In addition, final
product moisture may be above specification.

Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.

 The combination of high temperature and high pressure promotes a
significant elevation of product temperature, especially at the end of primary
drying, and this may result in collapse. The condenser temperature may also
increase significantly. However, these conditions lead to a high sublimation
rate and the cooling effect resulting from ice sublimation limits the rise of
product temperature.
In some cases, the low-temperature/high-pressure cycle can be
anticipated to be the most damaging to the product because it can create
conditions where the rate of ice sublimation is slowed down, thus resulting
in product overwarming, collapse, and high residual moisture.
The four deviations shown in Table 1 can be implemented only
during primary drying or, alternatively, during the whole cycle. A high
temperature during secondary drying will be a test of the resistance of
the active substance to thermal degradation. A high pressure during
secondary drying does not favor desorption, and this test may be used
to define a maximum acceptable pressure level for terminal secondary
drying.

VI. PERFORMANCE QUALIFICATION

OF THE TARGET CYCLE
A. Definition and Aims
The purpose of carrying out a performance qualification of the target cycle
is to demonstrate that the product can be manufactured in a reliable and
reproducible manner using the selected freeze-drying process. In order to
substantiate process reproducibility, it is commonly accepted that three
consecutive successful runs is adequate.
In addition to the qualification of the lyophilization process itself, it is
usual to collect liquid samples at various stages of the manufacturing. In this
way, the losses in active material or its degradation can be monitored
throughout the process.
As the batch-to-batch consistency study is usually the last step
preceding the release of the drug product onto the market, it is essential at
this time that the batches be manufactured as normal production runs. In
particular, production timing should be respected and the validation batches
should be full scale or at least representative of the size range of the
commercial batches.
The qualification of the target cycle may be carried out in the way
of a prospective validation if the batches produced are not for use in
humans. In this case, the risk of product contamination is not a critical
issue and the taking of in-process samples is facilitated. The obvious

Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.

 drawback of this approach is the sacrifice of active material. More
importantly, if a product load is completed with a placebo, the filtration
and filling operations will not be representative of a full-size batch, nor
will all of the areas within the lyophilizer be occupied by active product.
Especially at this stage of product development, the use of a placebo should
be carefully considered. Unless the active molecule is present in minute
amounts, the placebo formulation will never perfectly match the thermo-
dynamic, structural, and evaporative properties of the active product.
It is therefore often preferable to carry out the batch-to-batch
consistency study in a concurrent fashion [1,14], i.e., to manufacture full-
scale batches of active product under Good Manufacturing Practices
(GMP) conditions, using the standard manufacturing operating procedures.
The sterile batches produced are then placed on stability at various
temperatures. After a monitoring of product stability and a careful
evaluation of the validation data, they may be progressively released onto
the market, if the implemented manufacturing process gives full satisfaction.
The performance qualification study normally consists of two parts:
Careful control of in-process parameters and evaluation of the
performance of the equipment (process performance qualification)
The extensive and rigorous evaluation of product quality (product
performance qualification)
After each validation run, the batch records and the freeze-drying
graphic should be carefully examined. Furthermore, the validation program
may include the items developed hereafter.

B. Validation Program

1. Control of the Freeze-Drying Parameters

Shelf Temperature Profile. The shelf temperature during the

soaks, the duration of the soaks, and of the ramps should be consistent
with the settings and should be reproducible to within a set range from
batch to batch.
Dependent Operating Variables. The pressure in the chamber,
in particular the maximum pressure during primary drying, the mini-
mum pressure during secondary drying, and the maximum condenser
temperature, should be tabulated. If the batch size is kept constant, these
parameters should be reproducible for batches produced with the same
equipment.

Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.

 Product Temperature Trace. The traces of the different batches
should be compared and found in reasonable agreement. Using the method
shown in Figure 1, one can tabulate the maximum product temperature at
the end of primary drying and the product primary drying time (individual
probes and the mean of all).
The pressure curve and the condenser trace may also be used to
determine the primary drying time of the product (Figure 1).

2. QC Release Testing of the Final Product

A full quality control testing of the final product should be performed
according to the standard analytical operating procedures and release
specifications.

3. Visual Inspection of the Final Product

After each run, the full batch of product should be subject to visual
inspection, with particular attention paid to lyophilization-related defects
[13]. The various defects should be tabulated and the acceptance criteria
may state a maximum acceptable level for each defect.

4. Stability Studies
The batches should be placed on real-time and accelerated stability studies.

5. Extensive Analytical Testing of the Product for Uniformity

In contrast with routine QC where a very limited number of samples
are tested, a large number of product containers of each batch should be
tested for critical analytical properties such as assay, activity or bioactivity,
purity, degradation products, residual moisture, and reconstitution time.
The purpose of extensive testing is to show that each pellet is typical and
representative of the rest of the batch, which is a guarantee of product safety
and efficacy as the patients are usually injected with the contents of a single
container.
A random sample of product containers may be used for this testing.
Alternately, the product may originate from various defined areas within
the lyophilizer. The data gathered should be interpreted in terms of
descriptive statistics. For each analytical attribute, the mean, the standard
deviation, the percentiles, the extreme values, and the normality of the
distribution can be determined.
One-tailed or two-tailed confidence limits for the mean content,
activity, purity, or residual moisture may also be calculated. For each batch
produced, the 99.9999% confidence limits for the mean should be found to

Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.

 be within specifications. For a well-controlled process, 99.9999% one-tailed
limits may typically lie in the range of 50–75% of the upper specification
limit.
For each analytical property, the proportion of individual pellets lying
outside specification may be tabulated. This level should be minimal if the
capability of the process and of the lyophilizer is consistent with the
specification limits. If the distribution of pellet contents obeys the normal or
log-normal law, it is further possible to predict for the whole batch the
proportion of samples that can be expected to be outside the specifications.
Alternately, process capability indices can be calculated [15].

6. Uniformity of the Product from Various Areas Within the Lyophilizer

This item is discussed in Section VIII.

7. Analytical Testing of In-Process Samples

Liquid samples should be taken at the critical stages of the process,
e.g., bulk material, formulated solution before and after filtration, filling
solution, filled vials or ampoules at the beginning, middle, and end of filling.
When traces of material are to be assayed, care should be taken that the
collecting test tubes or containers do not adsorb significant amounts of drug
or leach out impurities in the formulated solution.
If they are unstable, the samples of each given step may be analyzed as
soon as collected. Alternately, if their stability is sufficient, the different
liquid samples may be stored in the refrigerator until the filling is complete
and they may be tested at the same time. If the freezing induces no
degradation, a third possibility is to freeze the liquid samples and to analyze
them at the same time as the samples of finished product.
Whenever feasible, the simultaneous analysis of samples from the
different sampling points offers a major advantage, that of enabling a more
accurate determination of the losses or variation in purity occurring during
manufacture. In actuality, for many analytical methods, the interassay
variability tends to be larger than the intraassay variability. As a
consequence, when samples from the different production steps are analyzed
separately, the results may not be fully comparable, and this may prompt a
conclusion of losses or degradation during manufacture while in fact the
product is stable.
As an example, the validation program based on liquid sample testing
may include the following items:
1. Sterile filter compatibility testing. It should be demonstrated that
filters of the same surface area as for commercial batches are

Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.

 compatible with the formulation being filtered [16]. The filter
should not induce significant degradation of the active product or
leach out undesirable substances. The retention of active material
inside the filter, in other words the loss on filtration, should be
quantified and assessed statistically.
2. Formulation and filling. The acceptability of the formulation and
filling processes should be evaluated with regard to the level of
oxidation or other degradation products. The maximum level
of degradation products should be consistently maintained below
the specification limit for the final product, but the acceptance
criterion may also be a maximum admissible purity decrease over
a given step or over the whole production process.
Special attention should be paid to the compatibility of the
plastic tubing that connects the filling tank to the dispensing
needles, although this can be the object of a separate validation
work. The active material may be adsorbed onto or may diffuse
into the tubing, or the tubing may leach out activators or
plasticizers into the formulated solution. With sensitive chromato-
graphic methods, these impurities may generate extra peaks that
appear in the retention time range where real degradation
products are quantified, thus leading to an underestimation of
actual product purity.
3. Assessment of the overall manufacturing process. When the
freeze-dried product is analyzed together with in-process samples,
the overall losses in active material throughout the production can
be calculated by comparing the mass or the activity of the starting
bulk material (or that of the formulated bulk before filtration)
with the mass or activity recovery in the freeze-dried product. An
estimate of activity variations on freeze-drying is supplied by the
comparison of the activity of the filling solution (filling tank or
filled containers) with that of the final product.
The losses on manufacture or the variation of purity can be
conveniently assessed via two-way analysis of variance (batch versus step)
and post hoc comparisons. The acceptance criterion may be that the
variation observed, if found significant at p ¼ 0.05, should not exceed a
predefined maximum admissible level.
To conclude this section, it is worth noting that the process
qualification based on liquid samples, although necessary, is not sufficient
to qualify all aspects of the stability of the formulated solution. For example,
it may happen that the formulated solution has to be stored in the filling tank
for a few hours longer than usual. The behavior of the solution during this

Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.

 extra time cannot be safely predicted from the above in-process data. In
order to release the batch produced, it is necessary to have a documented
study about the stability of the filling solution at different temperatures, over
an adequate period of time, e.g., 24 h.
A third aspect of product stability pertains to product uniformity
across the shelves (Section VIII.B).

VII. RETROSPECTIVE VALIDATION

OF THE FREEZE-DRYING PROCESS

A. Definition and Aims

Retrospective validation is based on the examination of the batch record
data of a large number of production batches once the product has entered
the market place. As commercial batches are usually produced under
admittedly constant operating conditions, the retrospective analysis of
freeze-drying data will supply valuable information about the reproduci-
bility of the process, especially if it includes batches produced over a period
of several years. However, unless significant deviations have occurred
during the production runs, retrospective validation will supply little
information about the robustness of the production process. This approach
is therefore complementary to the prospective validation of the cycle limits
that challenge the robustness of the freeze-drying process. Although
retrospective validation proceeds from the same philosophy as concurrent
validation, it supplies a more complete picture of the reproducibility and
reliability of the process. In contrast with concurrent validation, the
retrospective approach supplies documented evidence as to the capability of
the process to accommodate factors such as the replacement of lyophilizer
components, the maintenance performed in the production unit, or the
run-to-run fluctuations that occur in the daily operation of industrial freeze-
dryers. Retrospective validation is also a useful tool to substantiate that the
change of lots of excipients, vials, or stoppers has no effect on the product.
A retrospective validation of the freeze-drying process will usually
consist of two major areas:
Examination of the reproducibility of the freeze-drying parameters, to
show that the process and the equipment can be maintained within a
state of control
Examination of the reproducibility of product analytical properties,
especially those sensitive to lyophilization, to show that acceptable
product characteristics can be maintained
The results of the commercial stability program may also be included.

Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.

 B. Validation Program
1. Freeze-Drying Data
The retrospective study of freeze-drying data may include the following
parameters for each batch reviewed:
Batch size (percentage of full capacity)
Temperature of shelves during the successive freeze-drying steps
Duration of the soaks and of the temperature ramps
Total cycle time
Maximum pressure in chamber during primary drying, or the pressure
level of the calibrated leak
Minimum pressure at the end of secondary drying
Maximum temperature of the condenser
Temperature range of the product at the end of freezing
Maximum primary drying product temperature
Product primary drying time
The data tabulated can be used to calculate the mean values of the
various parameters and the standard deviation among the batches. The
temperature of the shelves and the duration of the different segments
should be reproducible to within defined limits. The extreme values
observed and compatible with an acceptable product can be used to define
an operating range for each parameter, e.g., for the shelf temperature or
the duration of any segment. The highest and lowest pressure values, the
highest condenser temperature values, and the highest product tempera-
tures should also be assessed with regard to residual moisture, pellet
appearance, and product activity in order to show that they did not affect
the product.
The freeze-drying data are also suitable for trend analysis. The
reproducibility of the freeze-drying process can be followed over years and
the database may be used to trace back the drift of a specific parameter from
its initial values.
When the freeze-drying cycle entails one or several segments of
a variable duration, statistical testing can be used to compare the
analytical properties of the batches with different segment times. For
example, if the duration of the freezing step is variable, the batches may
be categorized in two groups, ‘‘short freezing’’ and ‘‘long freezing’’. A t test
for independent samples can then be performed to verify that the freezing
time has no significant effect on properties such as residual moisture,
bioactivity, or pH.
If lyophilizers of different types are used to lyophilize a drug product
according to the same freeze-drying cycle, useful information about process

Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.

 robustness can sometimes be gained by comparing the values of the
dependent parameters. The analytical properties, in particular the activity
and the moisture of product batches manufactured in different lyophilizers,
also deserve comparison.
When different batch sizes are manufactured within the same
lyophilizer, one can assess the effects of loading on the dependent operating
variables. Partial and full-size batches should be comparable for pellet
cosmetic appearance and for analytical properties such as bioactivity or
residual moisture.
The deviations to the normal freeze-drying process that occurred
during the period reviewed should be examined for their impact on product
properties. The deviation and investigation reports and the stability studies
performed on the deviating batches may be referenced.

2. QC Release Data and Visual Inspection

Evaluation of the analytical data should demonstrate that the product
consistently achieves its predicted levels of quality, activity, and purity. The
visual inspection data should also be assessed, at least for lyophilization-
related defects such as collapse or melt-back.
The critical analytical data should be tabulated and analyzed in terms
of descriptive statistics (mean, coefficient of variation, extrema), control
charts, and trend analysis [17]. If the data of several years are included,
yearly means may be calculated, and the significance of the variations from
year to year may be investigated by analysis of variance to evaluate the
reproducibility of the process.
For each analytical test, the number and frequency of nonconforming
batches should be tabulated. An explanation for the failure should be
provided, in such a way as to distinguish between process- and nonprocess-
related failures.
Based on a large number of batches, the mean content in active
material in the final drug product, or its bioactivity, should be found
consistent with the claimed content or potency. The extent of the losses
throughout manufacturing should be assessed statistically.
Another opportunity of the retrospective validation is to examine the
effects on the documented process changes, based on a large number of
batches. For example, if the type or the size of the sterile filter has been
changed, it may be useful to perform a t test to compare the potency or the
drug content of batches produced before and after the change in order to
verify that the filters behave similarly for losses on filtration. If during a
period the ampoules or vials were provided by an alternative supplier, it may

Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.

 be worth verifying that the glass qualities are equivalent by comparing the
pH ranges of the final reconstituted products.

VIII. UNIFORMITY OF THE PRODUCT WITHIN

THE LYOPHILIZER
A. Definition and Aims
The study of the properties of the lyophilized product collected from
various locations within the lyophilizing chamber, according to a
predefined sampling plan, is a useful technique, especially in areas such as
troubleshooting, product development, cycle optimization, validation of
cycle limits, and batch-to-batch consistency. The main purposes of carrying
out a ‘‘mapping’’ of the product within the lyophilizer are the following:
To assess the equivalence of the characteristics of the product arising
from the various areas in the chamber
To evaluate the ability of the manufacturing process to yield a
homogeneous product
To evaluate the capability of the lyophilizer to yield a homogeneous
product
Various product characteristics can be assessed via product mapping.
For example, considering protein drug products, typical attributes to be
investigated for uniformity may include residual moisture, pellet appear-
ance, immunoassay, protein content, and various impurities or degradation
products that may form during filling or freeze-drying. Product nonunifor-
mity may typically be promoted by two factors:
An inadequate filling process and/or holding step in the lyophilizer
prior to freeze-drying
A poor freeze-drying cycle or an ineffective lyophilizer
Although it is conceivable during drug development to remove
product samples just prior to freezing or at the end of freezing in order to
investigate the effects of filling and storage in the lyophilizer, it is often easier
to sample the freeze-dried product. Compared to liquid samples, the testing
of freeze-dried samples offers several advantages.
The freeze-dried samples are representative of the finished product to
be released onto the market.
Providing the product is reconstituted in a small volume of diluent, the
resulting solution is more concentrated than the formulated solution
before freeze-drying. It is therefore easier to assay minute amounts

Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.

 of drug or impurities and the analytical methods used for QC release
can be implemented.
Once the lyophilizer is unloaded, the removing of stoppered vials is
compatible with good manufacturing practice and can therefore be
implemented in the frame of a concurrent validation.
The disadvantage of collecting samples only after the freeze-drying is
that the results may be more difficult to interpret in the case where the
product is not of the same characteristics from any position within
the lyophilizer, as the differences observed reflect the overall effect of the
successive manufacturing steps.

B. Testing of Product Uniformity Across the Shelves

A change of product characteristics across the shelves can have several
causes. During filling, the trays supporting the vials or ampoules are usually
loaded by small blocks onto the shelves of the freeze-dryer, which are
precooled at a defined temperature. Therefore, the product filled first (top
shelf) is in contact with the containers and stored for a longer time inside the
freeze-dryer than the product filled last (bottom shelf). Conversely, the
product filled last remains for a longer time in the filling tank, which may or
may not be refrigerated. In addition, the plastic tubing connecting the filling
tank to the dispensing needles may interact with the filling solution and this
may result in an uneven distribution of the drug substance within the
product containers filled first and last.
With sensitive products, it is therefore useful to verify that the
expected level of product purity and activity can be consistently maintained
throughout the filling and the product loading prior to freeze-drying.
A simple example of mapping is shown in Figure 4A for a
performance qualification run for a recombinant drug product filled and
released by ‘‘mass.’’ Each vial contains only a few micrograms of active
protein, and therefore it is important to show that the protein content of the
vials is uniform across the shelves of the freeze-dryer. Although no protein
material was expected to disappear during lyophilization, the concern here
was that protein may be lost during the filling process.
To demonstrate that this was not the case, the shaded trays shown in
Figure 4A were sampled and for each tray 3  2 vials were assayed for
protein content. The protein content results were analyzed with a two-cell
analysis-of-variance model including a factor, the left/right positioning, and
a covariate, the shelf number. In order to increase the power of the statistical
testing, the shelf number was handled as a covariate and not as a factor,
based on the assumption that the filling was progressing at a constant rate.

Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.

 Figure 4 Sampling plans for the determination of product uniformity within the
lyophilizer. (A) Uniformity of protein content for a product in vials. (B, C)
Uniformity of residual moisture for a product in ampoules. The lyophilizer has eight
shelves and each shelf holds 15 trays of product containers. Arabic numerals refer to
the tray number and show the order in which the trays are loaded during filling.
Roman numerals relate to the sealing sequences. The trays are unloaded by blocks
12, starting from tray 120. As the trays are ‘‘sealed’’ in twos, the first two trays
subject to sealing are identified by the number I, the last two trays by the number VI.
The trays from which samples are collected for analysis are shaded.

Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.

 In this particular case, the comparison of vials located on the left and on the
right side served as a control because no differences of the mean protein
content were anticipated. The homogeneity of the protein content across the
shelves was demonstrated by the nonsignificance of the slope of the
regression line of the protein content versus the shelf number. If
the regression mean square had been found to be much larger than the
within-mean square (which is a measure of the within-tray variability of
protein contents with respect to the content predicted for each tray by the
regression line), then there would have been a significant variation trend of
protein content across the shelves. The acceptance criterion stated that if the
slope was found to be significant at p ¼ 0.05, then the variation in protein
content from the top to the bottom shelf had to be no more than a given
percentage in order for the process to be considered as acceptable.
Sampling layouts similar to the one shown in Figure 4A are
appropriate when product heterogeneity reflects a change or a degradation
occurring during filling or storage in the lyophilizer prior to freezing. For
example, it is reasonable to anticipate a potential increase in oxidation
products during filling, reflected as a trend across the shelves, but there is no
reason to expect different amounts of oxidized products in product
containers coming from two neighboring trays.

C. Testing of the Equivalence of the Product from All Areas

Within the Lyophilizer
This situation is more complex when the product property under
examination is affected by the freeze-drying process and the equipment. In
this case, it should be demonstrated that lyophilization process and the
lyophilizer yield a product that consistently achieves its predicted level of
quality wherever it comes from in the lyophilizing chamber.
A typical case is that of residual moisture. Pellets from a same batch
show inherent differences in moisture content. Residual moisture is
dependent on a multitude of factors including the location of the samples
in the chamber, their positioning in the middle or at the edge of the trays,
and the consistency of the contact between the trays and the shelves.
When looking at process steps, a first potential cause of product
heterogeneity is the difference of temperature among the product containers
stored in the lyophilizer prior to freezing. The next critical step is freezing
because in many freeze-dryers the circulation of silicon fluid inside the
shelves is of limited efficiency. Thus, depending on the way the silicon fluid is
circulated, all of the shelves will not necessarily cool down at the same rate,
especially under full-load conditions. Furthermore, the corner of the shelves
corresponding to the fluid inlet will undergo a faster cooling than the corner

Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.

 corresponding to the outlet. The temperature differences will be transmitted
to the product, thereby generating variations in the ice crystal structure,
with repercussions upon drying. During primary and secondary drying, the
rate of water removal from the samples will also be affected by their location
with respect to the main valve, and with respect to the inlet and outlet of the
silicon fluid.
In extreme cases, the variations in residual moisture may be large
enough to have dramatic repercussions on batch stability. For example, the
product located in the upper part of the chamber may be found stable
whereas the product from the lower part is unstable. It is therefore essential
to demonstrate that the batches produced have uniform residual moisture,
or at least that the differences observed are too small to jeopardize product
stability.
Two of the sampling layouts developed to assess the moisture
uniformity of a sucrose-based recombinant drug product in ampoules are
shown in Figure 4B and C. When the lyophilization was over, the trays of
ampoules were unloaded progressively by blocks of 12, starting from the
bottom shelf, and the vacuum was restored in the chamber in the meantime
between each unloading operation. This is a standard operating procedure
in order to avoid ampoule pickup of excessive amounts of moisture prior
to sealing.
Because the validation batches were produced under GMP conditions,
it was not possible to recover the opened ampoules during the lyophilizer
unloading. The ampoules were therefore collected only after sealing. As the
trays of ampoules were ‘‘sealed’’ in twos, the ampoules selected for analysis
arose from pairs of trays, and not from individual trays (see Figure 4B
and C). The sealing of the ampoules of a pair of trays took about 2.6 min.
Thus, the ampoules of the first two trays were exposed to laminated air for a
maximum period of 2.6 min and those of the last two trays for 13–15.6 min.
Apart from the compliance with GMPs, the taking of samples after
sealing offered a major advantage, that of supplying sealed samples fully
representative of routine production. Nevertheless this procedure had a
drawback: it did not allow one to distinguish readily between variations in
moisture due to sealing and variations resulting from the position of the
samples within the lyophilizer.
In order to overcome this problem, the moisture results were analyzed
by analysis of variance. The following factors were examined for their effect
on the dependent variable ‘‘moisture’’:

Shelf (1 to 8)
Rear–front positioning (rear, front, and, for some layouts, middle)
Left–right positioning (left, right)

Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.

 In addition, the pairs of trays were classified and six groups according
to their sealing rank, I to VI (Figure 4B and C), and the sealing rank was
handled as a covariate in the analysis-of-variance model. No interaction
terms between the main factors or between the covariate and the main
factors were defined in the model.
The sampling plan shown in Figure 4B led to a 32-cell orthogonal
model and that of Figure 4C to a 24-cell nonorthogonal model. In both
models each cell was confounded with a pair of trays. The residual variance,
i.e., that between cells, was therefore representative of moisture differences
among pairs of trays, whereas the within-cell variance accounted for
ampoule-to-ampoule variations within a same pair of trays.
When analyzing the results of each batch, the residual to within-mean
square ratio was calculated to determine whether the use of trays had a
significant impact on moisture uniformity. In most runs, this F ratio was
nonsignificant. This demonstrated that the contact between the trays and
the shelves was consistent enough to ensure a homogeneous product when
running the target cycle, owing to the relatively slow freezing process and to
the time allowed for secondary drying.
For each run, the significance of the factors shelf, rear–front
positioning, and left–right positioning, as well as the significance of the
regression term associated with the sealing, were assessed with respect to
the within þ residual mean square. This analysis was complemented by
the partitioning of the total variance into components so as to quantify the
contribution of each of the following factors to moisture variability:
Shelves
Rear–front positioning
Left–right positioning
Sealing
Tray-to-tray variations
Ampoule-to-ampoule variations (differences among ampoules from a
same pair of trays þ analytical error)
Based on an overall analysis of seven runs, the contribution of the
shelves to moisture variability was found to be only 3.5% (Figure 5). The
drying was best under the butterfly valve. It was marginally less efficient for
the shelves most remote from the butterfly valve. Virtually no differences
were noted when comparing the front, middle, and back rows of trays.
The contribution to product variability of left–right positioning was 0.3%.
The average moisture content of the samples located on the right side was
found to be 272 mg/amp versus 279 mg/amp for the left side. This was
consistent with the circulation of silicon fluid inside the shelves from the
right to the left side.

Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.

 Figure 5 Distribution of product residual moisture within the lyophilizer.

Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.

 The moisture variability resulting from the uptake of moisture during
sealing was found to be 6.6% on the average. The increase in moisture
upon sealing was calculated (regression coefficient) to be approximately 7 mg
per sealing rank. Therefore, the sealing induced a mean moisture increase of
24 mg, and in the worst case the ampoules were calculated to pick up 42 mg
water (sealing rank 6).
With regard to the upper specification limit for moisture, this increase
was marginal, and it was calculated that there was no risk to breach the
specification limit at the relative humidity of the clean room, assuming a
cumulation of worst cases.
Finally, for the various runs the moisture variability arising from the
trays varied between 0 and 14%, whilst the ampoule-to-ampoule variations
represented the major part of the total variance, i.e., 65–95%. Although the
differences among ampoules were the major sources of moisture variation,
the absolute magnitude of these variations was too small to jeopardize
product homogeneity and stability. Indeed, for the batches produced, the
coefficient of variation of moisture content (including all sources of
variation) did not exceed 15%. As the mean moisture of the final product
was approximately half of the maximum allowed content, there was
therefore no risk to produce pellets above specification for moisture content.
From these results it was concluded that the lyophilizer was
appropriate for the sucrose formulation under validation, as it resulted in
minimal moisture variations from the different areas of the chamber. The
unloading of the trays by blocks of 12 was also an adequate practice
considering the specification range for relative humidity in the clean room.

IX. CONCLUSIONS

The validation of lyophilization cycles is a complicated issue because the

process parameters and the characteristics of the product are closely
interrelated. The process affects the final product, but the characteristics of
the product undergoing lyophilization impact the dependent operating
parameters, the freeze-drying pattern, and dictate the basic requirements for
a successful process. This interdependence limits the opportunities of using
placebo formulations and most of the validation runs must be performed
with active product. The high cost and limited availability of some
materials, as well as the need to validate the process under conditions that
are representative of routine production, justify that part of the validation is
performed in a concurrent fashion. The concurrent validation should then
be completed by a retrospective review of the data accumulated for

Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.

 commercial batches, so as to track the reproducibility, the reliability, and
the trends of the process over a longer period of time.
In the next decade, the increasing demand for high-technology
lyophilized products may prompt pharmaceutical companies to optimize
and shorten their freeze-drying processes, so as to keep the size of clean
room facilities and the manufacturing costs at a reasonable level. In some
cases, the redesigned freeze-drying cycles will probably be more extreme or
use alternative techniques such as pressure ramping or cyclic pressure
lyophilization [18], and it will be of the utmost importance to ensure that the
new processes are robust. In this respect, perhaps a promising breakthrough
is the emergence of mass spectrometers that can be mounted on the chamber
or between the chamber and the condenser. Besides supplying quantitative
data about gases in the chamber and vapor components in transit from the
product, these spectrometers enable a straightforward evaluation of
the effects of a variation of pressure or temperature in terms of water
removal from the product. As increasingly sophisticated products and
processes emerge, such equipment will prove increasingly useful for cycle
optimization, validation, and process compliance.

REFERENCES

1. EH Trappler. Validation of lyophilized products. In: IR Berry and RA Nash,

eds. Pharmaceutical Process Validation. 2nd ed. New York: Marcel Dekker,
1993, pp 445–477.
2. AP MacKenzie. Collapse during freeze-drying—qualitative and quantitative
aspects. In: SA Goldblith, L Rey and WW Rothmayr, eds. Freeze-Drying and
Advanced Food Technology. New York: Academic Press, 1974, pp 277–307.
3. ML Roy, MJ Pikal. Process control in freeze-drying: determination of the end
point of sublimation drying by an electronic moisture sensor. J Parenter Sci
Technol 43:60–66, 1989.
4. K Murgatroyd. The freeze-drying process. In: P Cameron, ed. Good
Pharmaceutical Freeze-Drying Practice. Buffalo Grove, IL: Interpharm Press,
1997, pp 1–58.
5. N Milton, MJ Pikal, ML Roy, SL Nail. Evaluation of manometric temperature
measurement as a method of monitoring product temperature during
lyophilization. J Pharm Sci Technol 51:7–16, 1997.
6. SL Nail. The effect of chamber pressure on the heat transfer in freeze-drying of
parenteral formulations. J Parenter Drug Assoc 34:358–368, 1980.
7. TA Jennings. Effect of pressure on the sublimation rate of ice. J Parenter Sci
Technol 40:95–97, 1986.
8. RG Livesey, TWG Rowe. A discussion of the effect of chamber pressure on heat
and mass transfer in freeze-drying. J Parenter Sci Technol 41:169–171, 1987.

Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.

 9. TA Jennings. Discussion of primary drying during lyophilization. J Parenter
Sci Technol 42:118–121, 1988.
10. MJ Pikal. Freeze-drying of proteins. I. Process design. Pharm Technol Int
37–43, January 1991.
11. MJ Pikal, ML Roy, S Shah, Mass and heat transfer in vial freeze-drying of
pharmaceuticals: role of the vial. J Pharm Sci. 73:1224–1237, 1984.
12. MJ Pikal. Use of laboratory data in freeze-drying process design: heat and
mass transfer coefficients and the computer simulation of freeze-drying.
J Parenter Sci Technol 39:115–139, 1985.
13. FDA Guide to Inspections of Lyophilization of Parenterals. Food and Drug
Administration, Rockville, MD, July 1993.
14. E Trappler. Lyophilization. In: MJ Groves and R Murty, eds. Aseptic
Pharmaceutical Manufacturing II. Buffalo Grove, IL: Interpharm Press, 1995,
pp 291–309.
15. DC Montgomery, Introduction to Statistical Quality Control. New York: John
Wiley & Sons, 1996, pp 430–474.
16. V Kumar, R Murty. Validation of aseptic process. In: MJ Groves and R Murty,
eds. Aseptic Pharmaceutical Manufacturing II. Buffalo Grove, IL: Interpharm
Press, 1995, pp 101–116.
17. DC Montgomery. Introduction to Statistical Quality Control, New York:
John Wiley & Sons, 1996, pp 179–312.
18. JD Mellor. Fundamentals of Freeze-Drying. London: Academic Press, 1978.

Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.