Gaspar2014 - Spary Dryer

Chapter 8
Spray Drying: Scale-Up and Manufacturing
Filipe Gaspar, Joao Vicente, Filipe Neves and Jean-Rene Authelin
List of Abbreviations
A Area of the spray dryer chamber
API Active pharmaceutical ingredient
BD Bulk density of the product
CCF Central composite face-centered design
Cfeed Solids content in the feed
Cp Heat capacity coefficient
dD Droplet size
Dnoz Diameter of the nozzle orifice
DoE Design of experiments
Dv Diffusion coefficient in the gas phase
Dv50 Volumetric median particle size
Ffeed Flow rate of feed fed to the spray dryer
Fdrying Flow rate of drying nitrogen
m Mass flow rate
MFP Maximum free passage
MW Molecular weight
Pfeed Atomization pressure of the feed
Pout Vapor pressure at outlet temperature
Psat Saturation pressure of the solvent
PSD Pharmaceutical spray dryer
Pwb Vapor pressure at wet bulb temperature
QbD Quality by design
F. Gaspar ()
Particle Engineering Services, Hovione, Lisbon, Portugal
e-mail: fgaspar@hovione.com
J. Vicente · F. Neves
R&D Drug Product Development, Hovione, Loures, Portugal
J.-R. Authelin
Pharmaceutical Sciences Operation, Sanofi R&D, Vitry Sur Seine, France
© Controlled Release Society 2014 261
N. Shah et al. (eds.), Amorphous Solid Dispersions,
Advances in Delivery Science and Technology, DOI 10.1007/978-1-4939-1598-9_8
262 F. Gaspar et al.
Qloss Heat lost from the drying chamber walls

RS out Relative saturation at the outlet of the drying chamber
Tb Boiling temperature
Tcond Drying gas temperature at the exit of the condenser
Tdew Dew point temperature at the outlet of the drying chamber
Tfeed Temperature of the feed solution
Tg Glass transition temperature
Tin Drying gas temperature at the inlet of the drying chamber
Tout Drying gas temperature at the outlet of the drying chamber
Troom Room temperature
Twb Web bulb temperature
U Overall heat transmission coefficient
Vm Molecular volume
x Molar fraction in the liquid phase
y Molar fraction in the gas phase
σ Surface tension
μ Viscosity
ρl Liquid density
ρg Gas density
ρP Particle density
Hvap Vaporization heat
γ Activity coefficient
8.1 Technology Transfer and Scale-Up to Commercial Units
In the pharmaceutical industry, the production of spray-dried powders is still widely

based on the batch concept. The quantities required in the early stages of the develop-
ment are typically small but may increase by many orders of magnitude as the drug
candidate advances through the clinical phases and reaches the market. This requires
the scale-up to different units along the process. Pharmaceutical spray dryers are
available in a wide range of scales: from lab units where milligrams of material can
be produced to large commercial units capable of handling multiple tons of pow-
der per day. If necessary, a spray drying process can be scaled up directly from the
laboratory to a final production scale. However, the quantities required for clinical
trials are more efficiently produced in pilot or small production scales, where prod-
uct losses and scale-up risk are considerably lower. Therefore, these intermediate
production scales are commonly used during the development of the process. Never-
theless, it is important to mention that during scale-up, some quality attributes of the
product can change, and there is need to understand whether these changes are ac-
ceptable, and if so, desirable. For example, powder properties such as flowability and
compressibility can be improved significantly when moving from lab units to larger
ones. Changes in the particle size distribution, level of residual solvents, friability,
density, and compressibility of the powder may strongly influence the properties of
8 Spray Drying: Scale-Up and Manufacturing 263
the final tablet, viz. hardness, friability, disintegration time, and dissolution rate. A
careless scale-up strategy may lead to considerable losses of expensive materials and
ultimately jeopardize the timelines of a clinical program.
Despite its criticality, the scale-up of the spray drying process is still vastly empir-
ical and based on costly experiments, and their statistical interpretation. To minimize
the experimental burden of such an approach, recent efforts have focused on applying
mechanistic models and simulation tools to describe the process of spray drying. In
fact, mechanistic modeling and process simulation tools have been successfully used
in chemical and oil industries for more than half a century. This rational approach
has gained wide recognition, and pharmaceutical scientists are now making use of it
during development, scale-up, and manufacturing (Koulouris and Lagonikos 2002).
Nevertheless, pharmaceutical process development will require some sort of simu-
lation and experimental testing at small scale, and at least some level of verification
in a production environment.
The following sections highlight some important considerations regarding the
most critical decisions related to the spray drying process, viz. selection of scale,
atomizer, and key process parameters. Common challenges associated with the op-
eration of the process will also be addressed. At the end of the section, a scale-up
methodology based on scientific first principles, simulation models, and process
characterization techniques are presented.
8.1.1 Manufacturing Scales
The way droplets are dried within the drying chamber dictates the characteristics
of the final product. Several aspects like evaporation rate, particles trajectories,
residence time, and wall deposition are governed by the factors like atomization
device and conditions, design and positioning of the gas disperser, the dimensions
of the chamber, and the location of the atomizer and exhaust gas duct. Spray dryers
are available in multiple configurations, including cocurrent, countercurrent, mixed
flow, fountain, or fluidized spray drying mode. The pharmaceutical industry predom-
inantly uses the cocurrent mode since it minimizes the exposure of the product to
high temperatures which may be crucial when processing thermally labile products
or materials with a low glass transition temperature. The cocurrent mode is therefore
considered the most suitable for the majority of the pharmaceutical applications.
Spray drying is a continuous process capable of full automation, and can be
designed to meet any capacity required in the pharmaceutical industry. A process
can be run in a large size unit for as short as 30 min or 1 h, or can be run continuously
for many days. The selection of the right scale involves several considerations, but
ultimately, it is primarily driven by the targeted process throughput and batch size
requirements.
Fig. 8.1 Lab-, pilot-, and commercial-scale spray dryers at Hovione
8.1.1.1 Lab to Production Equipment
Laboratory scale spray dryers are particularly useful for producing small quantities
of prototype formulations in early stages of development. They can process small
quantities of solution (as low as 2 ml) with relatively high yield. On the other ex-
treme, the process can run continuously for hours or days providing the flexibility of
producing hundreds of grams or even a few kilograms of material. It is not surprising
that lab-scale units have been used to produce commercial quantities of very-low-
volume products. A typical feature of these small-scale systems is that the drying
chamber and cyclone are constructed in glass (Miller and Gil 2012), enabling the
privileged visualization of the drying and separation processes (see Fig. 8.1a). The
main limitation of the lab units is the powder properties of the resulting materials,
namely particle size. The small dimensions of the drying chambers limit the residence
time, and therefore, the droplets need to be small in order to be dried completely
before leaving the drying chamber, or they will collide with the walls. Therefore,
most small units produce powders with mean particle sizes below 10–20 μm, more
often between 3 and 10 μm. There are, however, lab-scale units (e.g., ProCepT
R&D Spray Dryer) that operate under laminar flow, allowing the drying of much
larger droplets (100 μm and larger). These units represent an excellent platform to
mimic the size and morphology of the industrial-scale spray-dried particles and can
be used at early stages of development to assess the criticality of particle size and
other powder attributes in the quality of the product.
Pilot- and commercial-scale spray dryers (see Fig. 8.1b, 8.1c) are suitable for a
wide range of batch sizes, ranging from less than 1 kg to several metric tons. They
share many similarities regarding the configuration, materials of construction (typi-
cally stainless steel), ability to handle most organic solvents, and level of automation.
Additional features such as cleaning-in-place or recirculation of the drying gas (close-
loop units) may be included. Some units can operate under vacuum (to minimize
risk of powder exposure of highly potent drugs) but commonly operate under slight
Fig. 8.2 Water evaporation capacity for different spray dryer units
positive pressure. Despite these many options, the main difference between scales is
the evaporation capacity and the throughput.
The evaporation capacity depends mainly on the drying gas flow rate, solvent,
temperature profile (inlet and outlet temperature), and heat loss of the spray drying
unit. Fig. 8.2 shows the water evaporation rate at different production scales and
nominal drying gas flow rates.
8.1.1.2 Considerations for the Selection of Scale
Scaling up a spray drying process offers in most cases a more energy-efficient process,
lower manpower input per kilogram produced, and greater flexibility in adjusting and
optimizing product attributes. The latter is particularly advantageous if the purpose
is to obtain large and denser particles for solid oral dosage forms. For inhalation,
however, the challenge is to maintain particle properties unchanged throughout the
scale-up processes. The selection of ideal scale for a given product is dictated by the
commercial demand projections and the real throughput of the process. To calculate
real throughput, one needs to know not only the throughput in the spray drying step
(and how it changes at different scales—see example in Fig. 8.2) but also which step
(e.g., solution preparation, spray drying, or secondary drying) is the bottleneck of
the process. Other variables with significant impact on the real throughput of the
process are batch size, cleaning regime, and the duration of the cleaning process.
8.1.2 Nozzle Selection
Once the scale of spray dryer is established, it is important to select an atomization

nozzle appropriate to the scale. The purpose of the atomization stage is to produce
a fine mist (spray) from a liquid feed to substantially increase the liquid surface
area and improve the efficiency of heat and mass transfer. For example, 50 ml of
a solvent atomized in 800 million droplets of 50 μm creates a surface area about
6 m2 . By the generation of such high surface area, droplets dry fast, in the order of
seconds or fraction of a second depending on the drying conditions. Moreover, the
control of the atomization process dictates droplet size and consequently the particle
size.
Sprays may be produced in various ways, but essentially, all that is needed is a high
relative velocity between the liquid to be atomized and the surrounding gas. Some
atomizers accomplish this by discharging the liquid at high velocity into a nearly
stagnant gas. Notable examples include pressure nozzles and rotary atomizers. An
alternative approach is to expose the relatively slow-moving liquid to a high-velocity
gas stream. The latter method is generally known as two-fluid atomization. Indepen-
dent of the device, atomization is a complex phenomenon of inertial, shearing, and
surface tension forces, the balance of which determines the angle and penetration of
the spray as well as the density number, droplet velocity, and size distribution. All
these characteristics are markedly affected by the internal geometry of the atomizer,
the properties of the gaseous medium, and the physical properties of the liquid itself,
particularly its surface tension and viscosity.
Atomizers are generally classified according to the type of energy used. Rotary
atomizers (centrifugal energy) use high velocity discharge of liquid from the edge
of a wheel or disk. Two-fluid nozzles (kinetic energy) rely on the breakup of liquid
on impact with high-speed gas at the orifice. Pressure nozzles (pressure energy)
feature the discharge of liquid under pressure through an orifice, and ultrasonic
nozzles (acoustic energy) breakup of liquid is promoted through sonic excitation.
For each class of atomizer, there are several configurations and designs available to
handle the diversity of feed materials and to meet the specific spray-dried product
characteristics (Masters 2002). The liquid feed properties (viscosity, surface tension,
solids concentration) impact the atomization performance in all types of atomizers.
However, their sensitivity to each property depends on the particular type of nozzle.
8.1.2.1 Rotary Nozzles
For rotary nozzles, atomization is achieved by centrifugal energy transmitted to the

liquid stream by a disk or wheel rotating at high speed (from 10,000 to 50,000 rpm).
The liquid is fed into the center of a rotating wheel, moves to the edge of the wheel un-
der the centrifugal force, and is disintegrated at the wheel edge into droplets. A spray
angle of about 180◦ is best accommodated in large-diameter chambers (Mujumdar
2006). Rotary nozzles can be used to atomize slurries, suspensions, or solutions of
high viscosity. Besides the feed properties, the operating variables that influence
droplet size are feed flow, rotational velocity, wheel diameter, and design. Rotary
nozzles typically produce droplets of a wide range of sizes: from 20 to 200 μm
(Masters 2002).
8.1.2.2 Two-Fluid Nozzles
Two-fluid nozzles, also known as pneumatic nozzles, use a compressed gas to atomize
the liquid feed. There are different designs of nozzles on the market. The two major
groups of two-fluid nozzles are known as an external mixing nozzle and an internal
mixing nozzle.
External mixing nozzle is operated with low liquid pressure. The liquid feed is
provided through an inner duct while an atomization gas is fed by an external annular
opening around the liquid orifice. On the other hand, internal mixing nozzles take
advantage of gas expansion at the nozzle outlet. Part of the pressure energy applied
is used to scatter the liquid fragments within and beyond the nozzle orifice by the
sudden gas expansion (Walzel 2011).
Although an external mixing two-fluid nozzle is the most common in the lab-
and the pilot-scale spray dryers, an internal mixing nozzle is far more efficient in
regards to the gas to liquid ratio and, therefore, preferred for larger-scale spray dryers,
especially when small particle sizes (less than 10 μm) are required (Miller and Gil
2012). The main disadvantage is the air/nitrogen pressure required to overcome the
high pressure drop of this type of nozzle. Nevertheless, both internal and external
mixing nozzles produce droplet sizes within the range of 5–75 μm (Masters 2002) and
offer, probably, the best control over droplet size since feed and gas flow rates can be
controlled independently. However, the gas consumption and pressure required may
limit their use at industrial scales mainly when drying organic solvents. Therefore,
pneumatic nozzles are most suitable for small scales or when very small particle
sizes are required.
8.1.2.3 Pressure Nozzles
In pressure nozzles, atomization is achieved by converting pressure energy into ki-

netic energy. Often the design of pressure nozzles includes the inlet slots to impart
a swirling motion to the liquid at the swirl chamber entry and a convergent section
to accelerate the flow as it enters the orifice. The swirl motion of the liquid pushes
it to the wall and, consequently, the liquid is ejected from the orifice as a conical
sheet that spreads outwards due to centrifugal forces. These nozzles require the use
of high-pressure pumps as pressures can go up to 450 bar. Droplet size can be ma-
nipulated with the operating pressure, but feed flow is dependent on that pressure.
With increasing atomization pressure, the droplet size decreases and the feed flow
increases. This dependent manipulation of the droplet size is one of the major draw-
backs of this type of nozzles, i.e., in order to change droplet size at constant feed flow,
it is necessary to change the nozzle dimensions or design. Further, as they involve

the acceleration of the liquid feed, they are not suitable for high viscous feeds.
In large-scale spray dryers, these nozzles are used for production of medium
to large particles (30–200 μm). They also produce more uniform powders with a
narrower particle size distribution than pneumatic or rotary nozzles and, therefore,
are preferred for the production of powders for oral dosage forms.
8.1.2.4 Ultrasonic Nozzles
The principle of the ultrasonic nozzles is based on the usage of high-frequency sound
waves to atomize the feed and produce very narrow droplet size distributions with
low velocities. For ultrasonic nozzles, the feed delivery pump controls the liquid
flow. The ultrasonic nozzle generates a uniform droplet size distribution ranging
approximately from 20 to 100 μm. However, the frequency of vibration is specific
for a given nozzle, and so does the droplet size produced. In practice, in order to
change the droplet size, it is required to use a different nozzle. Higher-frequency
nozzles produce smaller droplets. The major drawback of these devices is that the
throughput is limited (typically up to 50 ml/min) which limits their applicability to
laboratorial- and pilot-scale units.
8.1.2.5 Considerations for the Selection of Nozzle
In the selection of the atomizer and atomization parameters, two general require-
ments should be considered: one is to provide the throughput that meets the required
powder production and the second is to generate a droplet size that provides for the
target particle size distribution. In the pharmaceutical industry, the most commonly
used nozzles are two-fluid and pressure nozzles, owing to their simplicity of use,
easy of cleaning, ability to handle wide variety of feeds and the reduced tendency
(when compared to rotary nozzles) to generate wall deposits. In most applications,
a pressure nozzle is preferred than a two-fluid nozzle, primarily because pressure
nozzle provides powders with a narrower particle size distribution. Exceptions in-
clude very fine powders or when feeds have very high viscosities or large suspended
particles which may block or damage the pressure nozzle. In the former case, the
greater flexibility to manipulate and control particle size in the fine range favors two-
fluid nozzles. Powders with higher densities are generally obtained from pressure
nozzles compared to two-fluid nozzles. This is associated with the degree of aera-
tion of feed during the atomization process. Low particle densities, if required, can
be obtained by optimizing the aeration effect (gas ejection or pressurization of the
feed).
Table 8.1 below summarizes the main characteristics of pressure and two-fluid
nozzles.
Table 8.1 Guidelines for nozzle performance

Pressure Nozzles Two-fluid nozzles
Pressure swirl External Mixing Internal Mixing
High atomization gas

Efficient atomization at low
Feed flow > 5 kg/h consumption restricts the
gas-to-liquid ratios
(not suitable for lab units) production of small particles at
larger scales
Flexible angles: 40-70⁰ Narrow angles: ~20º Flexible angles: 20-60º
Eroded by suspensions Suited for suspensions Eroded by suspensions
Not suitable for viscous feeds Insensitive to viscosity Sensitive to viscosity
Weak manipulation of droplet High degree of control over droplet size
size (for the same throughput)
Favors flowability Fine powder with low densities
Particle size – 20 to 500 µm Particle size – 3 to 200 m
Narrow distribution Wide distribution;
span ~ 1.4-1.8 span ~2.0-2.4
8.1.3 Typical Challenges
Most of the causes of unplanned shutdowns and/or limited run times are related either
to (1) excessive buildup of material or the equipment walls, (2) improper atomization
of the feed, and/or (3) chemical stability constraints (feed solution and/or powder).
These issues and the ways to overcome (or account for) them are addressed in this
section.
8.1.3.1 Product Accumulation
Product accumulation on the walls of the equipment is one of the most common
occurrences when developing or during scale-up of a spray drying process. Product
may accumulate on the walls of the drying chamber, cyclones, conveying ducts, or at
the nozzle tip. Wall deposits are more commonly observed in the small scales since
radial distances from the atomizer and residence time are shorter. Product buildup
can be caused by several factors:
• Product stickiness; materials with a low glass transition temperature exhibit sticky
properties and tend to build up on the walls of the equipment. The level of buildup
is related to the content of solvent in the powder, the glass transition temperature
of wet product, and the drying temperature at which the product is exposed. In the
production of amorphous forms, this is particularly important since the deposits
Fig. 8.3 Typical problems associated with spray drying operations: a condensation in the bag filter,
b nozzle bearding, c dripping, and d stringing
on the equipment surfaces will occur when drying at temperatures close or above
the glass transition temperature. The problem can be mitigated by reducing the
outlet temperature (Tout ) or by operating at lower relative saturation (RS out ) at
the same outlet temperature. The latter is achieved by reducing the condenser
temperature (Tcond ) and/or the feed flow rate (Ffeed ). The stickiness tendency can
be predicted offline, for example, using a hot stage and exposing the product to
different temperatures.
• Solvent condensation; the gas inside a spray drying process equipment is partially
saturated with solvent and, therefore, is prone to condense if exposed to cooler
surfaces. If the surface is allowed to cool down below the dew point of the gas
stream, then condensation will occur and the powder will accumulate on those
wet surfaces (Fig. 8.3a illustrates an example where condensation occurred in the
filter bag). To prevent this, the spray dryer is typically insulated or heat jacketed
and the heating of the chamber prior to start of the atomization process is done in
a gradual manner and over an extended period of time (typically between 20 and
60 min).
• Bearding; during spray drying, some droplets entrapped in the eddies around the
nozzle may collide with the tip of the nozzle, dry on its surface and start building
up around it. (see Fig. 8.3b). This process is commonly referred as bearding. If left
unattended, the buildup can interfere with the spray formation and/or fall down
into the drying chamber, promoting further powder deposition and ultimately
clogging of the equipment. There are several approaches to overcome bearding.
The simplest way is the repositioning of the nozzle (by changing the depth of
the nozzle tip inside the chamber) to reduce or eliminate the droplet entrapment
into the eddies. A change of atomization conditions as well as the drying gas
flow may also be beneficial, though often there are narrow margins to manipulate
these parameters. An engineering alternative is to include an additional gas stream
concentric to the nozzle tip to prevent the collision of the droplets at the nozzle
level. Another option is to use a nozzle with an anti-bearding cap which provides
less deposition area for buildup to occur. More commonly, a combination of the
above approaches is needed to successfully overcome bearding.
• Too large droplets; if droplets are too large for the drying chamber, they will
touch the chamber wall before drying is complete. The solution is to decrease
the droplet size or increase the drying temperature. Production scale may need
to be changed if the target particle size cannot be achieved. This should be well
thought out early in the development since all units have their own requirements
regarding the maximum droplet size allowed.
8.1.3.2 Improper Atomization
Atomization is probably the most critical step involved in the spray drying process.
The problems that can be associated with poor atomization include:
1. Inadequate atomization settings; this is the most common cause of poor atomiza-
tion. Often in an attempt to increase particle size, atomization is set to maximize
droplet size either by using low operating pressures (for pressure nozzles) or by
using low atomization ratios (for two-fluid nozzles). However, there is a fine line
between the conditions that allow maximum droplet size and those that lead to
not fully developed spray.
2. Dripping; in pressure nozzles, the spray is fully developed only above a critical
pressure. Below the critical pressure, often between 10 and 30 bar, large droplets
may form which will fall into the drying chamber, which in turn may cause the
formation of a wet layer of product in the discharging pipe of the drying chamber
(see Fig. 8.3c). Dripping may also occur when restarting the operation after an
unexpected shutdown. To prevent dripping, pressure nozzle with check valves
that only allow the passage of liquid above a preset pressure can be employed.
Some suppliers offer integrated check valves that can be adjusted for different
pressures. Other option is the use of a compressed gas to purge the feed line once
the high-pressure pump is switched off.
3. Stringing; stringing is the term used to refer to a bad atomization that happens in
high evaporative systems. If a fine spray with high specific area is exposed to high
temperatures, string like particle formation may occur before the spray pattern
is completely developed. String-like particles are obtained and agglomeration
occurs, see Fig. 8.3d. In order to avoid stringing, it is required to delay the
onset of particle formation by operating at lower temperatures or by reducing
the concentration of the feed. Both approaches will reduce the throughput. If not
resolved, solvent system or even the type of atomization system may have to be
replaced.
4. Poor assembly of the nozzle; most nozzles require careful assembly of the internal
and external components. Incorrect installation will result in poor atomization and
often results in the shutdown of the process and loss of product. Nozzle assembly
procedure should be carefully followed as per instruction and thoroughly tested
before use.
8.1.3.3 Chemical Stability
The product purity profile is a very important attribute for any product. During spray
drying operation, chemical degradation may occur:
• When the product is dissolved in the solvent prior to spray drying. This is partic-
ularly true at commercial scale, as the holding time may be long. It is therefore
critical to define the solution storage conditions, especially the temperature and
duration.
• When the freshly spray-dried product is still hot and not fully dried. In particular,
some deposits of product may remain for a long time in the spray drying chamber,
exposed to hot temperature (the walls of the chamber typically exhibit temperature
close to the exit temperature of the drying gas).
Aforementioned chemical stability risks will be discussed below using a solid dis-
persion (API to excipient ratio of 1:4), where the solvent system used was a mixture
of dichloromethane to ethanol ratio (97.5:2.5, % w/w), as an illustration.
8.1.3.3.1 Degradation of the Feed Solution
The degradation kinetics of a feed solution, at three different temperatures, is shown

in Fig. 8.4. In order to determine accurate degradation rates, the timescale in this
study was much longer than typical holding time. Although the formation of multiple
impurities was observed, depicted in Fig. 8.4 is only the main (and most critical)
impurity, for which an upper limit of 0.1 % was set.
As shown in Fig. 8.4, the kinetics show a linear increase over time, more pro-
nounced at higher temperature. Using a simple Arrhenius equation, the activation
energy (Ea) can be calculated, from which a mathematical expression capable of
describing the growth of the impurity can be obtained. Equation can be readily used
to construct a two-dimensional plot of temperature versus hold time (see Fig. 8.5).
Obtaining this type of representations enables the selection of the most appropriate
feed solution temperature, in agreement with the target run time; for example, for a
Fig. 8.4 Main impurity formation over time, as a function of the feed solution temperature
Fig. 8.5 Maximum (hold

time vs. temperature),
considering the main impurity
formation
target spray drying time of about 24 h, a hold temperature range of 2–8 ◦ C keeps the
degradation level well below the 0.1 % limit. Additionally, in the case of emergency
(e.g., equipment failure, requiring full process stop in order to allow for maintenance),
Fig. 8.5 plot also shows that feed solution could be kept for more than 1 week (168 h)
in case hold temperature is lowered to − 10 ◦ C.
8.1.3.3.2 Degradation Inside the Spray Dryer Chamber
The spray-dried product accumulated in the chamber is exposed to harsh conditions

(solvent vapor and high temperature) for an extended period of time and, therefore, it
is expected to show a less favorable purity profile. Considering a worst-case scenario,
where the entire amount of powder deposited on the walls would suddenly fall into
the main product container, the entire batch could be jeopardized; therefore, it is
important to evaluate the degradation of the deposited material as a function of its age
(e.g., to define the optimal cleaning frequency of the spray dryer: too frequent would
Fig. 8.6 Arrhenius plot for

main impurity formation of
the powder (lab kinetic results
vs. spray drying data)
impact cycle time, whereas too rare would imply a risk of product contamination).
These types of studies can be conducted using a two-step approach (the same example
previously used to address the feed solution chemical stability is being considered):
• Acquisition of the degradation kinetics using a sample of freshly spray-dried
product. In this illustrative case, a product with 3 % of residual solvents (the
typical value obtained at normal operating conditions) was enclosed in containers
and kept at four different temperatures (25, 35, 45, and 65 ◦ C).
• Verification of the degradation kinetics at the real spray drying scale. In this
illustrative case, a spray drying run (using a PSD3 unit, with an inlet and outlet
drying gas temperatures of 95 and 50 ◦ C, respectively) was performed for about
1 h (with a feed rate of 50 kg/h of feed solution), in order to obtain some product
deposited on the walls; afterwards, the spray drying of pure solvent (at similar flow
rate and thermal profile) was maintained for 24 h, in order to keep the product on
the wall under normal degradation conditions. At the end of the trial, the powder
deposited on the walls of the chamber was manually recovered (scrapped) and
analyzed.
The results obtained for the current illustrative example are shown in Fig. 8.6.
As shown in Fig. 8.6, the rate of impurity growth observed in the laboratory is
typically comparable to the degradation observed in the powder deposited on the
spray dryer; considering the good alignment of the data, the degradation in real
manufacturing conditions can be estimated from lab data (in this case 0.25 %/day).
Therefore, by knowing the overall amount of deposited solids during a typical run
(in this case known to be less than 1 % of the batch size) and considering that the
drug product is typically homogenized and sieved, it is possible to anticipate the
maximum impact for a given time window (e.g., 0.01 % of degradation during 4
days) and define the cleaning frequency (e.g., every 4 days of production would be,
in this case, a good compromise).
8.1.4 Modeling Tools and Mechanistic Interpretation
The scale-up of spray drying processes has been primarily conducted based on actual
experimental data and experience mainly because the process, characterized by rapid
and simultaneous heat and mass transfer between the droplets and the drying gas,
is difficult to describe mathematically, and some of the parameters are often not
readily measured. Furthermore, the whole process is extremely dependent on the
feed properties and equipment scale and design. Despite this, fundamental modeling
approaches for process characterization have been proposed in the literature. Among
these are thermodynamic models to estimate the humidity of the exhaust air (Berman
et al. 1994), atomization models to predict droplet size (Lefebvre 1989; Senecal et al.
1999), or drying kinetics studies to anticipate the morphology of the particles (Larhrib
et al. 2003; Littringer et al. 2013).
In this section, several tools, which will serve as the basis for the scale-up method,
are presented: (1) a thermodynamic model to predict the outlet conditions of the spray
drying, viz. relative saturation and temperature of the outlet gas; (2) an atomization
model to predict the droplet size; and finally (3) a simplified model for droplet drying
to define the general trends of particle size and morphology.
8.1.4.1 Thermodynamic Modeling
The thermodynamic model allows the characterization of the process in terms of

drying gas flow (Fdrying ); relative saturation at the outlet of the drying chamber
(RS out ); inlet, outlet, and condenser temperatures (Tin , Tout , Tcond ); dew point of
wet drying gas (Tdew ); and feed flow (Ffeed ). A multivariate relationship of such
variables can be used to define the process operation range that respects the desired
outlet conditions (Tout , Tdew , and RS out ). The determination of such variables plays a
role in the development of the spray drying process in several distinct ways. Tout is
one of the most important process parameters since it has an impact on particle and
powder properties such as density, surface area, mechanical strength, and physical
stability. RS out is the main driving force in the drying process and, therefore, has
direct impact on the level of residual solvents in the final powder. The latter is of
particular importance when producing amorphous materials since the level of residual
solvent in the solids strongly affects its glass transition temperature (Tg ). Dew point
temperature (Tdew ) is also an important parameter to have in mind in order to prevent
solvent condensation in the equipment.
Essentially, the thermodynamic modeling consists in a set of equations that relate
process parameters through mass and heat balances and liquid–vapor equilibrium
equations. Below are the critical equations:
Fig. 8.7 Accuracy of the

thermodynamic model. (Data
from 386 batches, using
multiple solvent systems and
four scales of Hovione’s Niro
spray dryers)

B

Fdrying · Cpdrying · (Tin − Tout ) = Cpi · (Tout − Tfeed ) · ṁ1i
i=A

B

+ ΔHvap_i · ṁ2i + Qloss (8.1)
i=A
Qloss = U · A · (Tout − Troom ) (8.2)
yi · P = xi · γi · Pisat . (8.3)
The accuracy of the thermodynamic model can be significantly improved by

measuring experimentally the heat loss of a particular unit. Figure 8.7 depicts the
accuracy of the model by comparing experiment with model inlet temperatures.
The development of a thermodynamic model is of utmost importance for modeling
and scale-up purposes and provides an expeditious way to anticipate process condi-
tions at any scale. Figure 8.8 below shows the typical information that is obtained
through this model.
8.1.4.2 Droplet Size Estimation
The second stage of process modeling consists in applying a reliable correlation

between atomization parameters, liquid properties, and droplet size. Numerous ex-
perimental studies have been carried out, and several equations have been proposed to
relate droplet size to nozzle design, atomization energy, and physical and flow prop-
erties of the gas and liquids employed. Although some published models have proved
Fig. 8.8 Output of thermodynamic modeling for a process using pressure nozzle
good predictive capabilities, it is worth noting that these models were developed for
specific nozzle geometry, and extrapolation to other nozzles needs to be done with
care. An example of such correlations is shown in Eq. 8.4 below. The correlation was
developed based on data from 12 nozzles from Spraying Systems (Maximum Free
Passage, SK Series SprayDry®) with different geometric dimensions.
0.25 −0.450 −0.25
dD = 2.41σ 0.25 μ0.25 Ffeed Pfeed ρg . (8.4)
Figure 8.9 shows that despite many influences of process and formulation variables
on the particle formation process, droplet size is still the major factor controlling the
particle size, and it can be estimated by Eq. 8.4. The atomization model is a valuable
tool for scale-up since it can be used to select the nozzle that best suits the targets
for process throughput and particle size.
8.1.4.3 Particle Formation
To describe the particle formation process, several authors (Vehring 2008) emphasize
the usefulness of the Peclet number (ratio between droplet evaporation rate and
diffusional motion of the solutes) as a mean to predict the morphology of spray-
dried powders. For low Peclet numbers, the diffusion motion of the solutes is fast
compared to the velocity of the receding droplet surface and the droplet is allowed
Fig. 8.9 Dv50 is highly

correlated with the droplet
size produced during the
atomization process.
Trendline was forced to pass
through the origin
to shrink while solutes migrate to the droplet center. At a critical supersaturation

level, dense and solid particles are produced. For high Peclet numbers, on the other
hand, the evaporation predominates over diffusion and the surface becomes rapidly
enriched in solutes that precipitate. In these cases, an outer layer is formed almost
instantaneously at the droplet surface leading to hollow, light, and porous particles.
Often in pharmaceutical applications, viscous feeds are obtained as a result of
the formulations used (e.g., polymers, proteins, carbohydrates among other large
molecules) and particles tend to be hollow. However, the plasticity of the phar-
maceutical materials during drying has resulted in greater morphologies diversity
(Walton 2000). Due to the low diffusivity, most pharmaceutical particles form a
shell earlier during drying and the rate of evaporation decays gradually as the shell
becomes thicker. At this point, the shell mobility is determined not only by the dif-
fusion of the dissolved solids but also by their solubility and, more importantly, by
the mechanical properties of the formed shell (Vehring et al. 2007). If the drying rate
is high, the critical thickness, i.e., the thickness that assures the mechanical stability
of the shell, is reached very early in the drying process and the resulting particles
sustain the spherical form of the droplet. On the other extreme, if the drying is slow,
the thin shell formed in the early stages of drying will recede until its thickness is
stable enough to sustain the particle structure. So by controlling the evaporation rate
and drying time, one can exert control over particle morphology (see Fig. 8.10). The
mechanism of evaporation in a still gas, based on boundary-layer theory, can be jus-
tifiably applied to many spray drying conditions (Masters 2002) and used to estimate
the drying time of the droplets. While the constant evaporation rate is applied, the
drying time can be expressed by the following equations:
ρL 2
t= dD − dP2 (8.5)
8Dv (Pwb − Pout )
Fig. 8.10 Illustrative relationship between drying time and the sphericity of the spray-dried particles
0.00143T 1.75
Dv = 2 (8.6)
1/2 1/3 1/3
MWAB (v )A + (v )B
0.68
Tb
Twb = 137 log (Tout ) − 45 (8.7)
373.15
In Eq. 8.5, the vapor pressure at wet bulb temperature can be calculated from
the Antoine equation, while the solvents concentration (Pout ) in the drying gas can
be determined in the thermodynamic step described previously. For example, fast
evaporation and low drying times can be imposed by manipulating Tout and droplet
size (dD ) and used to promote the production of smooth spherical particles.
The other feature throughout the formation of particles is the creation of internal
pressure when droplets are dried at temperatures close or above the boiling point of
the solvent. In this case, the vapor pressure inside the particles is higher than the outer
surface and particles can, depending on the shell properties, inflate or break apart as
the result of the pressure gradient. When particles expand, then particle size becomes
a function of droplet size and outlet temperature (Fig. 8.11a). On the other hand, if
the material is friable, the particles tend to break apart and the degree of breakage
is typically more pronounced at higher temperatures. In those cases, particle size is
still dependent on the droplet size, but the outlet temperature or relative saturation
has a negative effect on particle size (Fig. 8.11b).
The tools presented throughout this section are intended to be used in all the
steps of the scale-up to assure the production of powders with the desired quality,
viz. particle size, particle morphology, and level of residual solvents. This enhanced
understanding and mechanistic thinking, in line with the quality by design (QbD)
initiative, will support the establishment of ample design spaces that are both unit
and scale independent.
Fig. 8.11 a Particle inflation observed above boiling point; b typical behavior of a friable material
8.1.5 Scale-Up Methodology
As discussed in the Sect. 8.1.1, the spray dryer scale may have a significant impact on
the properties of the spray-dried material and general scale-up procedures have to be
established to assure an uneventful transfer to larger scales. Before embarking on the
scale-up of any process, it is highly recommended to attain a stable and robust process
at lab scale. Only then one can clearly understand how key process parameters such
as temperature profile in the drying chamber (Tin and Tout ), condenser temperature
(Tcond ), drying gas and feed flow (Fdrying and Ffeed ), and atomization conditions should
be set at the larger-scale unit. The present methodology is based on the mechanistic
understanding described in the previous section and comprises three modeling steps:
thermodynamic, atomization, and particle formation.
8.1.5.1 Thermodynamic Step
Through thermodynamic modeling (see Sect. 8.1.4), one can calculate the relative
saturation of the drying gas. This combined with a small set of experiments at lab scale
provides the relationship between the level of residual solvents in the product and
relative saturation of the drying gas. This is particularly important when producing
amorphous materials since their glass transition temperature (Tg ) is affected by the
residual solvent since solvent acts as a plasticizing agent. Figure 8.12a illustrates how
these relationships can be used for scale-up purposes: If the target is to manufacture
a spray-dried (wet) powder at a Tg > 60 ◦ C, then the level of residual solvent in the
spray-dried material should not exceed 9 % w/w. At lab scale, this residual solvent
level was obtained when operating with a relative saturation of 8 %. Therefore, a
possible scale-up condition, which can be seen as a conservative or safe approach,
is to maintain at the larger-scale equipment a relative saturation at the exit of the
drying chamber (RS out ) at a similar level. If droplet size is maintained somewhat
Fig. 8.12 a Effect of the relative saturation of the drying gas on the level of residual solvents and
Tg and b effect of the scale on the desorption curves of a spray-dried product
Fig. 8.13 a Thermodynamic space at pilot scale and b projection to commercial scale
unchanged, the extended residence time in larger scales will provide a safety margin
to the assumed relationship between RS out and level of residual solvents. As can be
seen in Fig. 8.12b, the sorption curve approaches the equilibrium when increasing
the scale (equilibrium data is obtained with dynamic vapor sorption studies).
The multivariate thermodynamic relationship of the process parameters with RS out
can also be used to define the process operation range that respects the imposed
constraints. Apart from the limits imposed on the RS out , the constraints usually
include equipment limitations (e.g., maximum inlet temperature or minimum con-
denser temperature), product requirements (e.g., outlet temperature limited by the
product degradation profile, physical stability, or stickiness behavior), and other pro-
cess constraints (e.g., Fdrying limited by the gas disperser or flow requirements, Ffeed
limited to avoid high dew points). These theoretical relationships provide a bridge
between processes at different scales, as depicted in Fig. 8.13 below.
Fig. 8.14 a stable spray drying process, b product degradation in the cyclone, and c bad atomization
leading to product accumulation in the bottom of the drying chamber
This thermodynamic analysis is a very powerful tool often replacing the need
for experimentation at different scales. However, as mentioned before, product con-
straints should ideally be studied at lab scale to minimize development costs. Visual
degradation of material, bearding, heavy accumulation due to inadequate tempera-
ture profiles, or poor atomization among other are all easier to understand and to
solve at lab scale (see Fig. 8.14).
Some of the most common studies performed at lab scale include:
• Stability of the drug in the feed. This can be critical to the success of the scale-
up. In larger scales, product may be held in solution for many hours or days,
and therefore, the kinetics of degradation at different feed temperatures should
be known. In some cases, the feed temperature needs to be reduced to prevent
impurity growth.
• Process yield (expect > 80 % for a sample of more than 5 g). The reasons for a
low yield are very diverse (e.g., product stickiness or bad atomization) and should
be solved at small scale before scale-up.
• Ability/ease of the secondary drying step. At lab scale, drying profiles and
physical/chemical stability of the product can and should be evaluated.
• Process edge of failure by spray drying, for example, at elevated relative saturation
(RS out ) can be very useful to understand the limits of the process and provide extra
information for the scale-up process.
• When stable process conditions are found, it is recommended to run the process
for an extended period to monitor the robustness of the process. Note, for example,
that some processing issues (e.g., nozzle bearding or heavy product accumulation
in the equipment walls) may not be obvious in very short tests.
Fig. 8.15 a Feed solution characterization regarding the impact of Cfeed and Tfeed on feed viscosity
and b simulation of four nozzles MFP (maximum free passage) SK series SprayDry nozzles from
Spraying Systems; the core/orifice codes are –80/16, –70/20, –65/21, and –65/17
8.1.5.2 Atomization Step
Once the thermodynamic conditions have been established, there is a need to select
the nozzle that best suits the targets of droplet size and process throughput. The
most common nozzles used at lab scale are external two-fluid nozzles. However,
during scale-up, there is typically the opportunity to improve powder properties by
switching to a pressure nozzle. This results from the greater ability to produce and
dry larger droplets in the larger drying chambers of the commercial units.
Feed properties like viscosity, density, and surface tension are well known to af-
fect droplet size (and hence particle size). Therefore, for an accurate estimation of
droplet size, it is recommended to characterize the solution regarding those proper-
ties. Frequently, all of these properties are dependent on the feed temperature (Tfeed )
and solids content (Cfeed ).
Cfeed is a critical parameter which impacts the process viability and product qual-
ity in several manners, viz. process throughput, particle/powder density, or feed
viscosity. Economic considerations of the process favor high Cfeed , but concentra-
tions close to the saturation point should be further studied to minimize the risk of
product precipitation. Further, the use of high concentrations may lead to highly
viscous feeds which may be difficult to atomize (Fig. 8.15a illustrates a typical re-
lationship between feed concentration and temperature with feed viscosity). Tfeed is
rarely manipulated to obtain target powder properties. Nevertheless, it may affect
solution stability and also influences viscosity and solubility. A strict control of Tfeed
is required, namely when operating close to solubility limits.
After determination of the feed properties, correlations like the one shown in the
Sect. 8.1.4 (Eq. 8.4) are commonly used to predict the droplet size.
In the example given in Fig. 8.15b, the ranges of interest were 52–80 μm for
droplet size and 14–21 kg/h for the feed flow rate. The most adequate nozzle, ac-
cording to the feed properties measured and simulations performed, was the 65/21.
The 65/21 nozzle fulfills both criteria within the typical operating conditions of
pressure 30–100 bar.
8.1.5.3 Particle Formation Step
In order to establish a target for droplet size, it is required to study the effect of the
drying condition on droplet drying and consequent particle formation. In the particle
formation step, which is by far the most complex physical mechanism to describe,
two approaches may be considered to estimate the final particle size.
The first approach assumes a constant shrinking ratio, i.e., a characteristic ratio
between droplet size and particle size. When there is no information available about
the product drying behavior, a pragmatic approach can consider a general shrinking
ratio of about 3.3 (as seen in Fig. 8.9, most spray drying products follow the general
trend of droplet size, independently of their nature and drying conditions).
A second approach is to use experimentally found shrinking ratio. This requires
some prior knowledge of product/drying behavior. For example, the shrinking ratio
observed at a smaller scale (and using similar drying conditions) can be used to
model particle size at larger scale. In this approach, there is a need to measure or
estimate the apparent density of the particles (measured, for example, by mercury
intrusion porosimetry—Fig. 8.16), which can then be used to estimate particle size
using a mass balance to the solids in the droplet and particle (Eq. 8.8).

ddroplet ρparticle
3 (8.8)
dparticle ρdroplet × Cfeed
Besides particle size, the control of the particle morphology may also be critical
for some applications. Surface area, particle density, and roughness or porosity
are all known to affect the performance of spray-dried powders. Therefore, it is
common during the development of a spray-dried product to explore the process to
produce powders with distinct characteristics. This step of powder optimization can
be conducted at pilot scale where a good compromise is achieved in terms of the
range of particle/powders that can be obtained and the ability to target these ranges
throughout the remaining of the scale-up process.
Particle properties are related to the drying kinetics of the droplets inside the drying
chamber and are primarily dependent on the mechanical and chemical properties of
the spray-dried material. These interactions are very complex to model, and therefore,
process development is typically ruled by some general trends, as described below
(see Fig. 8.17). The parameters that most influence particle morphology are Tout
(or RS out ) and Cfeed . Note that by adjusting the droplet size produced during the
atomization process, one can control the particle size in such a way that size and
morphology become almost independent of each other.
Fig. 8.16 Mercury intrusion in HPMCAS spray-dried particles manufactured at a Niro PSD4
• Increase T_out to promote the production

of smooth spherical particles
• Increase C_Feed to promote the

production of dense particles
• Decrease C_ feed to promote the

production of shriveled particles
(increased surface area)
• Size and morphology are roughly

independent
Fig. 8.17 General guidelines to produce particles of different morphologies
8.1.6 Process Intensification
When moving to a full commercial scale, one of the main goals is to increase the
process throughput while maintaining or improving the attributes of the powder. In
other words, the goal is to increase Ffeed and keep constant the RS out and droplet size,
already optimized at a smaller scale. Droplet size and throughput can be controlled by
nozzle selection and atomization conditions, while RS out is controlled by Tout , Ffeed ,
and Tcond . This is achieved by using the thermodynamic and atomization models
Fig. 8.18 Simulation of a nozzle performance and b relative saturation for three possible scenarios
described before. Figure 8.18 describes an illustrative example of how this can be
simulated.
As depicted in the simulations of Fig. 8.18a, several nozzles can produce droplets
within approximately the same size range but at very different feed flow rates. By
changing from nozzle A to nozzle B or C, droplet size can be maintained while the
process throughput is increased. With the feed flow defined, then, as can be seen in
Fig. 8.18b, the heat requirements can be determined by the thermodynamic modeling
in such a way that outlet temperature is defined in order to have approximately similar
levels of relative saturation of those optimized in previous stages. It is schematized
in Fig. 8.18; three different scenarios that, according to the developed models, have
similar conditions of droplet size and relative saturation. It is expected that the three
scenarios result in the production of powders with equivalent quality.
8.1.7 Conclusions
A scale-up method was proposed for spray drying processes. The method encom-
passes three steps: thermodynamics, atomization, and particle formation. A scale-up
approach based on keeping the relative saturation of the drying gas at constant level
can be derived using only lab-scale data to define the thermodynamic design space.
The atomization condition and nozzle selection are decided based on target process
throughput and droplet size considerations. The methodology presented here reduces
the risk of failure during the initial scale-up of the process. The process can then be
Table 8.2 Summary of the scale-up methodology

Feed properties
Tfeed Use the Tfeed defined at lab scale. Keep a strict control over Tfeed ,
mainly if operating close to the solubility limit
Cfeed Preferably defined at lab scale. May need adjustments for adjusting
powder properties or for process intensification
Ffeed Work within the thermodynamic space defined by the target through-
put and drying capacity of the equipment
Drying gas variables
Fdrying Use the nominal flow of the equipment and adjust for drying gas
density to keep pressure drop through gas disperser at nominal level
Tout and Tcond Work within the thermodynamic space. Use initially a conservative
scale-up approach; keep RS out similar to the lab scale. Adjust at larger
scale based on desorption/Tg data
Atomization variables
Nozzle/Pfeed Pressure nozzle preferred for most applications. Very small particles
(e.g., for inhalation powders), feeds with large suspended parti-
cles and very viscous feeds, may require other atomization systems,
namely two-fluid nozzle. Use droplet size correlations to select the
most suited nozzle and atomization conditions
Tin is a dependent variable that is limited by equipment constraints
gradually intensified to improve throughput while keeping material attributes roughly

unchanged. Table 8.2 summarizes the methodology suggested.
8.2 Development of a Manufacturing Process of a Spray-Dried

Dispersion Under a Quality by Design Approach
8.2.1 Methodology Overview
Pharmaceutical QbD is a systematic scientific risk-based holistic and proactive ap-

proach to pharmaceutical development that begins with predefined objectives that
address product and process understanding. Successful product development relies
on consistent application of a proven methodology. The key steps are the same, ir-
respective of the product or formulation being developed. One proven methodology
is described within this chapter. The framework is shown in Fig. 8.19, while a short
description of the main steps is given below.
Target Product Profile and Critical Quality Attribute The target drug profile consists
of prospective and dynamic summary of the characteristics of a drug that should be
achieved in order to reproducibly deliver the therapeutic benefit; the target product
profile (TPP) sets an important number of performance parameters that will be the
Fig. 8.19 Quality by design framework (main stages)
basis of the critical quality attribute’s (CQA’s) definition, that is, the attributes of
the drug that must be kept within appropriate limits in order to ensure the desired
product quality.
Risk Assessment (Development Phase) For each CQA, an analysis of the potential
critical process parameters (pCPPs) and potential critical material attributes (pC-
MAs) is conducted. The aim is to evaluate, in each process step, which operating
parameters or raw materials have the potential to impact a CQA, within the known
ranges, and therefore should be monitored or controlled, in order to ensure the de-
sired quality. Since the number of parameters is usually high, a risk assessment,
based on prior knowledge of product/process, is used to rank the parameters in terms
of perceived criticality; the ultimate goal is to keep the development process as lean
as possible, by focusing the studies on those parameters and material attributes with
a higher likelihood of having a critical impact.
Process Development The output of the previous risk assessment is a qualitative
match between CQAs and pCPPs/pCMAs. To confirm the dependences and quantify
the effects, a process development stage is conducted. If a statistical approach is
followed, a sequence of design of experiments (DoE) is usually performed with dif-
ferent objectives: screening, optimization, and robustness studies. This development
stage constitutes the core of the QbD methodology since most of the specific process
knowledge is generated during this stage. Although not mandatory, a model, either
statistical and/or mechanistic, is a usual outcome of this stage. Process analytical
tools can also to be considered at this stage; based on the need to improve, the CQA’s
monitoring as the process is scaled up.
Design Space and Normal Operating Range Once the impacts of the
pCPPs/pCMAs are quantified on the CQAs, a feasible operating space can be defined.
This space, known as the design space, will consider all the interactions between op-
erating parameters and material attributes and will often be multidimensional. The
normal operating range (NOR) is established within the design space, and can be
thought of as the ranges where the process typically operates.
Risk Assessment (Manufacturing) After defining the design space and NOR, an
exhaustive analysis of the process is conducted at the manufacturing scale. In this
study, a failure mode effect analysis (FMEA) of all manufacturing aspects are re-
viewed, challenging the equipment operating ranges and procedures against the
process knowledge gathered in the previous steps. The purpose of this study is to
understand and quantify the risk of failure and to define actions to minimize it.
Criticality Analysis By knowing the feasible operating regions, the design space,
and after evaluating the equipment/procedures at the manufacturing scale, directly
evaluating the practical NORs, a final criticality analysis will take place in order to
identify parameters and/or material attributes that will require tight monitoring or
control; for example, all those for which the corresponding NORs are close to the
boundaries of the design space.
Process Control Strategy Once the criticality around a process parameter and/or
raw material attribute is confirmed, adequate control strategies will be set in place.
The ultimate goal is to assure that operation is always taking place within the design
space, therefore assuring the quality of the final product. For this purpose, and
considering the dependence of a control strategy on a given monitoring capability,
the final implementation of process analytical tools may be carried out at this stage.
The subsequent steps of this methodology are mainly focused on documentation
aspects associated with the filing process and, given the purpose of this current article,
will not be further discussed. This work focuses on the steps highlighted in Fig. 8.19,
which will be discussed in detail in the sections below.
8.2.2 Case Study Overview
Two particular difficulties are generally recognized during the formulation of solid
dispersions by spray drying: the need to dissolve both the drug and the polymer in a
common solvent system and the need to prevent phase separation during the removal
of the solvent. The selection/optimization of formulations will not be addressed in this
chapter as the corresponding approaches/methodologies are extensively discussed
throughout this book by other authors; the formulation presented in this case study,
optimized in previous stages of development, consisted of (1) a binary solvent system
(methylene chloride and ethanol, 95/5 % w/w), (2) hypromellose phthalate (HPMCP)
mixed with the drug at a ratio of 4:1, and (3) a solids concentration of 9 % w/w.
The solution is prepared in a reactor with a mechanical stirrer and a thermal circuit
for temperature control. After complete dissolution, the solution is fed to the spray
dryer. Droplet size is controlled by the liquid feed flow and by the type of atomizer and
atomization conditions (pressure nozzles were used during this work). Tout is used to
define the morphology of the particles and assure an efficient drying. The particles
obtained were separated from the drying gas through a cyclone. The unit is operated in
closed-loop mode, i.e., with recirculation of the drying gas. The solvent was removed
by a condenser temperature within the gas recycling unit. Finally, the spray-dried
material is collected and transferred to a double cone dryer for a secondary drying
operation to fulfill the applicable limits for residual organic solvents.
8.2.3 Target Product Profile and Critical Quality Attributes
As introduced in Sect. 8.2.1, the roadmap of any QbD approach starts with the Target
Product Profile (TPP) definition; this summary of drug characteristics (e.g., phar-
macokinetic properties and stability) will serve as the basis for a set of performance
parameters (e.g., immediate release drug: 80 % in ≤ 30 min, 36-month shelf life at
room temperature, respectively) that, in turn, will be linked to a set of Critical Quality
Attributes (CQAs; e.g., shelf life will depend on the amount of residual solvents due
to its impact on chemical stability; release profile will depend on particle size for
some drugs due to its impact on dissolution).
As one may expect, not all attributes of a drug will be classified as critical; Q6A
(Conference and Harmonisation 1999) offers guidance on this matter, by distinguish-
ing groups of attributes that should always be classified as critical regardless of the
drug’s end use (e.g., identification, assay, purity) from others whose classification
(critical, noncritical) will depend on the final product nature (see decision trees in
Q6A Conference and Harmonisation 1999). In order to conduct this exercise, it is
therefore important to have the full picture of all manufacturing processes that are
associated with a given product since, as depicted in Fig. 8.20, some of the attributes
of an intermediate manufacturing step (e.g., density of the bulk powder) are estab-
lished as critical due to their importance for subsequent manufacturing steps (e.g.,
by affecting tablet hardness during downstream operations).
For the sake of simplicity, the current case study will only focus in one of the
manufacturing steps (the spray drying of the bulk powder, as previously introduced in
Sect. 8.2.2) and, within this one, only show the application of the QbD methodology
for two of the CQAs: particle size (Dv50) and bulk density (BD). The considered
specifications for these CQAs (31–57 μm for particle size and 0.100–0.200 g/ml for
BD) were set based on the interaction/interdependence with the downstream process
Fig. 8.20 Interactions between upstream and downstream processes during CQA definition
where, as introduced before, these attributes reveal a critical importance during

manufacturing of the final oral dosage form.
8.2.4 Risk Assessment
The number of parameters involved in any spray drying process is relatively large,
and evaluating the impact of each one, in all CQAs, would be difficult to manage
(both from a cost and time perspective); therefore, one of the main goals of the
risk assessment, as previously introduced in Sect. 8.2.1, is to reduce the number of
pCPPs that will be studied in subsequent stages of process development. As shown
Fig. 8.21, the procedure adopted considers the ranking, for each CQA, of all process
parameters, according to the perception of criticality (that relates to the mechanistic
understanding and relevant manufacturing experience with similar products).
Once a ranking of perception of criticalilty is obtained, the process parameters
can then be divided into three groups: a first group (Tout , Pfeed , Dnoz , Tcond ), that is
considered to have a potentially relevant impact on the CQAs and, therefore, will
be studied in detail in order to establish the design space; a second group (Fdrying ,
Cfeed ), considered to have a lower potential of criticality—these parameters will not
be optimized (they will be fixed at preestablished set point values based on prior
knowledge of the process or similar processes) but will be used during the final
evaluation of the process robustness (in order to confirm the assessment made); a
Fig. 8.21 Ranking of pCPP per each CQA
third group (the least ranked parameters), that are expected to have no impact on the
process and that will be excluded from further studies. The rational for exclusion of
these parameters needs to be appropriately justified in a QbD application.
8.2.5 Process Modeling
The outcome of the previous risk assessment is a qualitative match, based on per-
ceptions, which needs to be confirmed and quantified during a subsequent process
development stage. A mechanistic description of the process, as introduced in
Sect. 8.1.5, can be used as a powerful tool to establish the design space with mini-
mum need for experimentation at final scale. However, this type of description is not
always readily available, and the use of a statistical approach constitutes a pragmatic
alternative. The limitation of the statistical approach is its reliance on scale-dependent
experimentation and the difficulty in extrapolating relationships to other scales or
equipment. Selection of the most adequate approach (mechanistic or statistical) de-
pends, therefore, on the (1) existence/reliability of the mechanistic understanding,
(2) availability of material for experimentation, and (3) flexibility required on the
design space. In this section, a statistical approach will be illustrated in order to
quantify relationships between process parameters and product attributes; the three
involved steps are described in Fig. 8.22.
The objectives of a screening stage, used to confirm the most significant factors,
are to determine the ranges of process parameters to be investigated and to reduce the
Fig. 8.22 Statistical modeling approach: structured sequence of design of experiments
number of parameters to be further studied. Since an accurate quantification of the

effects is not crucial at this stage, low-resolution experimental designs can be consid-
ered. Hence, few experiments per studied factor are typically required. The second
level of the experimental plan consists of the optimization of the process conditions
over the most promising subregion found during the screening stage (considering
only the parameters that have previously shown statistical significance). Since the
ability of capturing interactions between process parameters may greatly influence
the outcome of this stage, high-resolution experimental designs are advised. Finally,
the robustness evaluation is conducted in order to determine the sensitivity of a CQAs
toward small changes in some process parameters. Typically, the robustness evalua-
tion is centered at the target operating point (defined during the optimization stage),
by considering variations on parameters that have not been studied in detail.
8.2.5.1 Screening Stage
Among the pCPPs of the spray drying step, the risk assessment identified feed concen-
tration (Cfeed ), feed pressure (Pfeed ), outlet temperature (Tout ), condenser temperature
(Tcond ), and nozzle orifice diameter (Dnoz ) as the highest-ranked ones.
For the screening stage, a DoE with resolution ≥ 4 is recommended in order to
retrieve unconfounded relationships between first-order terms. The structure of the
DoE is shown in Fig. 8.23. In order to study a broader range of process throughputs,
six star points were added to a 24−1 fractional factorial.
The results obtained, shown in Fig. 8.24, reveal that the response of BD seems to
be well described by a linear function of Tout . For particle size, although a reasonable
model has been obtained, a linear structure seems to be insufficient to explain the
observed variance (as denoted by the relatively low R2 and Q2 ).
Additionally, this screening phase shows that two factors (Tcond and Dnoz ) were
found to have no statistical significance (within the tested ranges) and, therefore,
should be subsequently excluded from the optimization stage. Finally, the screening
model also showed that a great share of the evaluated operational ranges is feasible
(considering the target ranges for Dv50 and BD).
Input variable Min Center Max
P_feed (bar) 50 85 120

T_out (ºC) 25 55 85
T_cond (ºC) -10 -5 0
D_noz (mm) 1.01 0.91 0.78
Fig. 8.23 Screening study: DoE structure (left side) and ranges of variation studied (right side)
p-value
D50 BD
P_feed 6.72x10-3 0.280
T_out 8.93x10-4 5.78x10-10
D_nozzle 0.846 0.381
T_cond 0.742 0.322
Particle size Bulk density
Fig. 8.24 Screening study: model adequacy (left side) and statistical significance data (right side)
8.2.5.2 Optimization Stage
During the optimization stage, the operating ranges were narrowed down and an ex-
perimental design (central composite face-centered design) was considered in order
to support nonlinear interaction models. Only Pfeed and Tout were selected as input
variables at this stage, in agreement with the conclusions taken before. The center
point assumed during the screening study was maintained since product attributes
obtained at 85 bar and 55 ◦ C were close to the target values. The DoE selected to sup-
port the optimization stage, as well as the considered ranges for the input variables,
are shown in Fig. 8.25.
As shown in Fig. 8.26, the quadratic interactions improved the accuracy of the
model for particle size prediction. For powder BD, the optimization study confirmed
the suitability of a linear model (only function of Tout ).
P_feed (bar) 70 85 100
T_out (ºC) 40 55 70
D_ nozzle = 0.91mm; T_cond = Cte = -5ºC
Fig. 8.25 Optimization study: DoE structure (left side) and ranges of variation studied (right side)
D50 ( ( BD (g/ml)
Constant 141.2 0.449
P_feed (bar) -1.46 ---
-5.44 x
T_out (ºC) -1.35
10-3
(bar
P_feed ·
ºC --- ---
T_out
)
(P_feed) 2 (bar)2 7.64 x 10-3 ---
(T_out) 2 (ºC)2 1.47 x 10-2 ---
R2 (-) 0.87 0.97
Q2 (-) 0.79 0.96
Fig. 8.26 Model predictions: Dv50 (left bottom) and BD (left top), coefficients, and adequacy data
8.2.5.3 Robustness Stage
The DoE considered in this stage is shown in Fig. 8.27. The selected ranges for
the input variables translate possible deviations around the parameters settings. The
optimal values for Pfeed and Tout were determined through the optimization model,
considering the target values for product attributes.
No reliable model was obtained for particle size prediction (a desired outcome,
considering the nature of this study), which is in agreement with the small response
of Dv50 observed in Fig. 8.28 (between 44 and 48 μm). For BD, however, it was still
P_feed (bar) P_opt -5 P_opt P_opt +5

T_out (ºC) T_opt -2 T_opt T_opt +2
F_drying
1300 1400 1500
(kg/h)
C_feed (%
8.5 9.0 9.5
w/w)
D_noz = 0.96 mm; T_cond = -5ºC
Fig. 8.27 Robustness study: DoE structure (left side) and ranges of variation studied (right side)
Fig. 8.28 Robustness study: results against targets in place (left side) and statistical significance
(right side)
possible to obtain a model as a function of Tout (denoting an undesirably high sen-

sitivity), confirming the strong dependence predicted by the screening/optimization
models. Based on the results of Fig. 8.28, it can be concluded that (1) the process is
robust towards the parameters Fdrying and Cfeed (thus validating the risk assessment)
and that (2) although all obtained results are within the specification limits, Tout will
need to be carefully monitored and controlled during the control strategy definition
(due to its high potential of criticality via a high process sensitivity).
8.2.6 Design Space
The design space was defined as the multidimensional combination of the pCPPs
where, according to the optimization model, all CQAs’ values are obtained within
the applicable target ranges. Although considering the intersection of the feasible
operating ranges for each CQA, the final design space was mainly constrained by
the BD target range, as shown in Fig. 8.29.
Fig. 8.29 Design spaces for Dv50, BD, and joint Dv50 and BD targets
Fig. 8.30 DoE structure for design space validation
Regarding the validation of the design space, there is no universally recommended

procedure for verification of a model, and there is limited literature or regulatory
guidance that addresses the extent of verification required to justify a design space
(Hallow et al. 2010). In the present work, eight additional runs were performed inside
the design space and confronted with the predictions of the optimization model. A
32 DoE was centered in the optimized operation point (see Fig. 8.30). The results
obtained showed very small errors (the root mean squared error was 0.06 μm for
particle size and 0.003 g/ml for BD).
Nevertheless, this type of assessment can be misleading, as the magnitude of
the model prediction errors (small or large, acceptable or unacceptable) can only
be properly judged when confronted with the target specification ranges, during
an uncertainty evaluation exercise (García-Muñoz et al. 2010; Peterson 2008). In
fact, all models will have a corresponding error distribution, causing the boarders
of the design space to be a less safe operating region. These error distributions are
the sum of all the variability that cannot be explained or controlled (e.g., sampling
errors, analytical method variability, influence of unknown factors, etc.) and should
Fig. 8.31 Design space boundaries as a function of the imposed confidence level
be accounted during definition of the design space limits. In this work, the error
distribution was calculated using the residuals of the optimization model and the
eight runs performed for the design space validation. A normal distribution was
fitted to the data (17 points in total) through a nonlinear parametric regression. The
error distribution was then used to estimate error intervals at different confidence
levels. By accounting the errors intervals on the limits of the CQAs’ target ranges,
it was possible to establish the design space in a true risk-based approach where
broader operating ranges for the pCPPs imply greater risks of excursions outside the
CQAs’ target ranges. As shown in Fig. 8.31, the 90 % confidence boundaries lead to
a significant reduction of the original design space. However, if the operating point is
properly selected (Pfeed = 90 bar and Tout = 55 ◦ C), a temperature variation of ±3 ◦ C
and a pressure variation ± 30 bar will have a very low likelihood of threatening
the target CQAs’ ranges (as the 99.7 % confidence boundary is not crossed), an
observation that is in agreement with the outcome of the robustness study.
8.2.7 Mechanistic Understanding
The design space reported in the previous section was built based on statistical
models; however, the ultimate goal of any QbD framework should be the enhanced
understanding of the process, and this is ultimately achieved when a mechanistic
interpretation of the underlying phenomena is derived. Theoretical description of
the process of particle formation comprises two sequential steps: the atomization
Fig. 8.32 Atomization model predictions: Influence of nozzle orifice diameter on droplet size and
throughput, Pfeed, ranged from 30 to 100 bar in both nozzles
process to describe the liquid breakup and drying kinetics to explain the final particle
morphology.
In the present work, droplet size was calculated using the atomization model
introduced in Sect. 8.1.5, and the obtained simulations (Fig. 8.32) were used to sup-
port the final nozzle selection, considering all relevant production goals (i.e., process
throughput and target particle size). As forecasted by the theoretical simulations (and
confirmed by the screening statistical studies of Sect. 8.2.5), the nozzle orifice di-
ameter has little impact on droplet size. However, it is one of the most important
factors that determine the process throughput; changing the diameter of the orifice
is actually the most expeditious way to control the process throughput and keep the
droplet size roughly unchanged.
Regarding the particle morphology, the maximum structural stability of the spray-
dried particle is in a spherical form; however, the interactions between several process
parameters, which are related to the mass and heat transfer of the droplet, can result in
different particle morphologies. From the theoretical point of view, the evaporation
begins as soon as the droplets are ejected from the nozzle. Consequently, the droplet
T_out = 25ºC Dv50 = T_out = 55ºC Dv50 = 44 T_out = 85ºC Dv50 =

43 59
Fig. 8.33 SEM pictographs: Influence of Tout on particle size, at the same conditions (Pfeed = 85
bar; Dnoz = 0.96 mm; Tcond = −5 ◦ C)
surface rapidly becomes enriched in solutes. At a certain supersaturation level, pre-

cipitation will occur and a rigid particle will be formed with a given particle diameter.
From this standpoint, it is expectable that particle size closely follows droplet size.
However, at low temperatures, the mechanisms of solvent evaporation are slower,
allowing more time for particle surface to deform, shrink, or collapse. On the other
hand, at temperatures close or above the boiling point of the solvents, the increased
internal vapor pressure may lead the particles to inflate. Both phenomena could be
observed during the current work, being more notorious when the experimental data
is grouped in two sets: drying temperatures below and above boiling point. Two
models for particle size were built with different sensitivities towards Tout . Below
the boiling point, particle size is only dependent on the droplet size; above boiling
point, particle size is dependent on both droplet size and Tout . Both models of the
particle size showed good correlation coefficients (see Fig. 8.11a), and, together with
the different morphologies shown in Fig. 8.33, it is therefore possible to consolidate
and bridge the current mechanistic interpretation with the statistical model derived
for particle size in Sect. 8.2.5.
Since the atomization process is not affected by the drying condition (Tout , Tcond ,
and Fdrying ), the droplet size distribution produced by a given nozzle is mainly depen-
dent on Pfeed . Hence, since the same applies to the mass distribution of the droplets (at
constant Cfeed ), particles produced at high temperatures have lower densities. Thus,
powders dried at low temperatures have relatively high BD, because the individual
particles are shriveled and not very porous; on the other hand, powders dried at high
temperatures have relatively low BD, because the individual particles preserve their
inflated state and are, consequently, very porous (Langrish et al. 2006). Both inter-
pretations are well aligned with the statistical model derived in Sect. 8.2.5 for BD,
where Tout shows up as the most important operating parameter.
Therefore, through the development of mechanistic analyses, it is therefore pos-
sible not only to predict and control particle size but also to understand the main
physical phenomena that rule particle morphology and powder BD.
8.3 Conclusions
In the current work, a QbD methodology was applied during the development of a
spray drying process for the manufacture of a pharmaceutical solid dispersion. The
risk assessment stage (where mechanistic knowledge and past experience play an
important role) was found to be of major importance in order to keep development
as lean as possible (by focusing on the most important parameters). Although not
mandatory, the development of predictive models is strongly advisable to establish
a reliable design space where model uncertainty should be addressed. Statistical
approaches are a pragmatic way of establishing these relationships; however, a fun-
damental mechanistic understanding will always be advantageous as this one portrays
the physical principles that rule the process and, therefore, enables more general con-
siderations. For the current spray drying process, the combination of statistical and
mechanistic information enabled to conclude that (1) particle size depends mainly
on the droplet size and outlet drying temperature, (2) nozzle diameter dictates the
throughput and does not affect particle size significantly, and (3) BD can be controlled
just by manipulation of the outlet drying temperature due to its strong influence on
the porosity of the particles.
References
Berman J, Pierce P, Page PE (1994) Scale-up of a spray dry tablet granulation process:
thermodynamic considerations. Drug Dev Ind Pharm 20(5):731–755
Conference I, Harmonisation (1999) Specifications: test procedures and a cceptance criteria for new
drug substances and new drug products: Q6A, (October)
García-Muñoz S, Dolph S, Ward HW (2010) Handling uncertainty in the establishment of a design
space for the manufacture of a pharmaceutical product. Comput Chem Eng 34(7):1098–1107
Hallow DM, Mudryk BM, Braem AD, Tabora JE (2010) An example of utilizing mechanistic and
empirical modeling in quality by design. J Pharm Innov 5(4):193–203
Koulouris A, Lagonikos PT (2002) The role of process simulation in pharmaceutical process
development and product commercialization. Pharm Eng 22(1):1–8
Langrish TAG, Marquez N, Kota K (2006) An investigation and quantitative assessment of particle
shape in milk powders from a laboratory-scale spray dryer. Drying Technol 24(12):1619–1630
Larhrib H, Peter G, Marriott C, Prime D (2003) The influence of carrier and drug morphology on
drug delivery from dry powder formulations. Int J Pharm 257(1–2):283–296
Lefebvre AH (1989) Properties of sprays. Part Part Syst Char 6:176–186
Littringer EM, Paus R, Mescher A, Schroettner H, Walzel P, Urbanetz NA (2013) The morphology
of spray dried mannitol particles—the vital importanceof droplet size. Powder Technol
239:162–174
Masters K (2002) Spray drying in practice. In: SprayDry consult, International ApS, Denmark
Miller DA, Gil M (2012) Spray drying technology. In: Robert O. Williams III, Alan B. Watts and
Dave A. Miller Formulating poorly water soluble drugs. Springer, New York pp 363–442
Mujumdar AS (2006) Handbook of industrial drying, 3rd edn. CRC Press, Taylor Group, Boca
Raton, USA
Peterson JJ (2008) A Bayesian approach to the ICH Q8 definition of design space. J Biopharm Stat
18(5):959–975
Senecal PK, Schmidt DP, Nouar I, Rutland CJ, Reitz RD, Corradini ML (1999) Modeling
high-speed viscous liquid sheet atomization, Int. J. Multiphase Flow. 25
Vehring R (2008) Expert review pharmaceutical particle engineering via spray drying. Pharm Res
25(5):999–1022
Vehring R, Foss WR, Lechuga-ballesteros D (2007) Particle formation in spray drying. J Aerosol
Sci 38(7):728–746
Walton DE (2000) The morphology of spray-dried particles a qualitative view. Drying Technol
18(9):1943–1986
Walzel P (2011) Influence of the spray method on product quality and morphology in spray drying.
Chemical Engineering & Technology 34(7):1039–1048

Gaspar2014 - Spary Dryer

Uploaded by

Copyright:

Available Formats

Gaspar2014 - Spary Dryer

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Gaspar2014 - Spary Dryer

Uploaded by

Copyright:

Available Formats

Chapter 8

Spray Drying: Scale-Up and Manufacturing

Filipe Gaspar, Joao Vicente, Filipe Neves and Jean-Rene Authelin

Qloss Heat lost from the drying chamber walls

8.1 Technology Transfer and Scale-Up to Commercial Units

In the pharmaceutical industry, the production of spray-dried powders is still widely

8.1.1 Manufacturing Scales

Fig. 8.1 Lab-, pilot-, and commercial-scale spray dryers at Hovione

8.1.1.1 Lab to Production Equipment

8.1.1.2 Considerations for the Selection of Scale

8.1.2 Nozzle Selection

Once the scale of spray dryer is established, it is important to select an atomization

8.1.2.1 Rotary Nozzles

For rotary nozzles, atomization is achieved by centrifugal energy transmitted to the

8.1.2.2 Two-Fluid Nozzles

8.1.2.3 Pressure Nozzles

In pressure nozzles, atomization is achieved by converting pressure energy into ki-

it is necessary to change the nozzle dimensions or design. Further, as they involve

8.1.2.4 Ultrasonic Nozzles

8.1.2.5 Considerations for the Selection of Nozzle

Table 8.1 Guidelines for nozzle performance

High atomization gas

8.1.3 Typical Challenges

8.1.3.1 Product Accumulation

8.1.3.2 Improper Atomization

8.1.3.3 Chemical Stability

8.1.3.3.1 Degradation of the Feed Solution

The degradation kinetics of a feed solution, at three different temperatures, is shown

Fig. 8.5 Maximum (hold

8.1.3.3.2 Degradation Inside the Spray Dryer Chamber

The spray-dried product accumulated in the chamber is exposed to harsh conditions

Fig. 8.6 Arrhenius plot for

8.1.4 Modeling Tools and Mechanistic Interpretation

8.1.4.1 Thermodynamic Modeling

The thermodynamic model allows the characterization of the process in terms of

Fig. 8.7 Accuracy of the

Qloss = U · A · (Tout − Troom ) (8.2)

The accuracy of the thermodynamic model can be significantly improved by

8.1.4.2 Droplet Size Estimation

The second stage of process modeling consists in applying a reliable correlation

8.1.4.3 Particle Formation

Fig. 8.9 Dv50 is highly

to shrink while solutes migrate to the droplet center. At a critical supersaturation

8.1.5 Scale-Up Methodology

8.1.5.1 Thermodynamic Step

8.1.5.2 Atomization Step

8.1.5.3 Particle Formation Step

• Increase T_out to promote the production

• Increase C_Feed to promote the

• Decrease C_ feed to promote the

• Size and morphology are roughly

Fig. 8.17 General guidelines to produce particles of different morphologies

8.1.6 Process Intensification

Table 8.2 Summary of the scale-up methodology

gradually intensified to improve throughput while keeping material attributes roughly

8.2 Development of a Manufacturing Process of a Spray-Dried

Pharmaceutical QbD is a systematic scientific risk-based holistic and proactive ap-

Fig. 8.19 Quality by design framework (main stages)