Anticipation of scale up issues in pharmaceutical development

Proceedings of European Congress of Chemical Engineering (ECCE-6)

Copenhagen, 16-20 September 2007

Anticipation of scale up issues in pharmaceutical

development
a
F.L. Muller, J.M Latimerb

a
AstraZeneca, Macclesfield Works, SK10 2NA Macclesfield, UK
b
AstraZeneca, Avlon Works,BS10 7ZE , Bristol, uk

Abstract

We present a methodology to evaluate the robustness of pharmaceutical batch

processes on scaling up. The Scale Up Risk Evaluation (SURE) is applied on new
processes that are part way through their development. During a SURE study one first
collates current process understanding and then extrapolates this understanding to
evaluate scenarios at larger scale. The output of the SURE study is a ranked list of
scenarios that development teams use to prioritise further development.

Keywords: Batch Process, Process Development,

Scale up, Risk Assessment, Robustness & Reliability.

1. Introduction
The pharmaceutical industry is currently expanding the role of process engineers from their
traditional roles in chemical manufacturing and capital projects. Significant process
engineering populations can now be found in process development, which has typically been
the domain of the synthetic and development chemists. Some drivers for this trend are the
regulatory push for increased process understanding, and the frequency at which scale up
issues occur in processes moving to commercial scale (Sherlock & Brewis 2006).

From a process development point of view, the life cycle of an active pharmaceutical
ingredient (API) consists of a series of campaigns executed at scales starting at 10’s of grams
for discovery, then kg scale for early clinical work then 10-1000 kg for late stage clinical
work culminating in commercial manufacture. Although engineers do contribute to kg scale
campaigns, typically, significant engineering contribution is made to campaigns for 10’s Kg
onwards when development teams are formed, drawing on chemistry, analytical, pilot plant
and engineering.

The aim of a development team is to invent and scale up a synthesis in order to generate an
API for use in clinical trials. The pilot manufacture typically takes place in multi-purpose
batch reactors. As outlined by J Double et al. (2005), batch reactors have a degree of
flexibility unsurpassed when executing partially developed often multiphase (typically S-L)
processes.
 F. Muller et al.

The development team starts by identifying the “Route”; i.e. fixing starting materials and
intermediates. The process for a given Route is separated into stages; a stage covers either the
conversion of starting material or intermediates, or in case of a “Pure’s” stage the (re-)
crystallisation of the API.

Carey et.al. (2006) give an analysis of typical reactions used in pharmaceutical synthesis and
conclude that most pharmaceutical routes are “simple constructions from complex
fragments”, so called modular convergent synthesis. A typical pharmaceutical route consists
of 2-4 in-house stages forming intermediates, a stage to form the Crude API and a Pures stage
to achieve the desired quality of the API. (the Pures stage may also be combined with the
Crude stage).

For each stage a process1 is developed. This involves identification of solvent(s) and reagents
conditions (T, P) and the order in which operations are executed. Once a suitable process is
identified, it is optimized and scaled up to pilot scale in order to generate the required
quantity of API. The “process description” document describes the required operations of a
stage in detail.

A typical process consists of 20-40 operations; e.g. charge, heat, hold, cool filter, wash and
dry. The responsibilities of the engineers are typically related to (i) scale up of operations
(ii) minimization of environmental impact and hazard, and (iii) configuration of the flexible
(pilot) plant for the processing of a stage.

The extent to which engineers need to address scale up of operations depends very strongly
on the interaction between the chemist and the engineer. For example if a chemist working on
a process considers a mixture to be “thick” he may decide to dilute the process with solvent,
thus resolving the issue. If an engineer had been present he might have concluded that mixing
was not sufficient and suggested using an alternative agitator, thus also solving the issue. In
reality, a good chemist will have “chemi-neered” out the majority of the scale up issues
before an engineer gets involved.

Unfortunately, the chemist’s solution is not always robust with regards to scale up. Internal
studies with regard to “Right-First-Time” indicated that a significant proportion of processes
run at pilot scale gave unforeseen results; i.e. a scale up incident had occurred2. The severity
of these scale up incidents ranged from simple deviations (more added, longer times etc) that
still gave “in-spec material”, to material that had to be reworked or even product that was
totally unusable.

Roughly 30% of the scale up incidents occurred during reaction operations, 30% involved
crystallisation (slower, or unwanted nucleation) and 40% were related to other work up
operations. From this data it was estimated that about 1% of process operations led to a scale
up incident. It was also noted that scale up incidents are more likely to occur if the mixture in
the batch vessel is multi phase.

1
In this paper a “process” refers to the sequence of operations that is required to transform
the input material and reagents for a stage into the product of that stage.
2
A scale up incident is defined as an event during a development manufacture that had not
been expected based on the experimental laboratory data; e.g. an unexpected nucleation or
the formation of a new by-product.
 Anticipation of scale up issues in pharmaceutical development 3

2. Scale Up Risk Evaluation (SURE)

In order to help chemists and engineers identify the 1 in a 100 operation that is going to cause
a problem a Scale Up Risk Evaluation (SURE) methodology was developed. This
methodology aims to anticipate scale up issues early on in the development of a new
compound. Note: the SURE study is thus not a problem resolution technique like those
proposed by Keppner and Tregoe (1981) and the Britest consortium (“Driving Force
Analysis” Britest 2007, Sharratt and Borland 2003)

The principal of SURE is based on that of a hazard and operability study (HazOp): for each
operation in a process recipe various scale up/down scenarios are identified, where a scenario
is defined as an unplanned change in conditions. The potential impact of a scenario on the
process is then assessed as a “threat” or an “opportunity” and subsequently scored on:

(i) The likelihood the scenario will occur on a change in scale

(ii) The likelihood the scenario will affect product quality or process operability.

The output of a SURE study is a risk matrix, and a summary table of the threats and
opportunities that provide development drivers for the development team. The key elements
of a sure study are:

Participants
The evaluation of a ‘Process’ works most effectively with a cross functional team. The
minimum team requirements are a process owner (typically a chemist), an engineer and a
study facilitator. The facilitator brings a number of skills to the SURE study: a consistent
approach, knowledge of the tool, an independent eye and experience of other projects. The
inclusion of too many team members or observers can detract from or slow down the study.

Timing
The study can be performed at anytime in the development of a ‘Process’ but is best done
once both reaction and work up have basic definition. .

Excel tool
The application of a rigorous approach allows a consistent format to be applied and the output
can then be correlated against other studies. An Excel based tool is used to capture the
‘Process’ and SURE information.

The study is performed using a number of elements: description/capture of current

knowledge, possible challenge scenarios are tested, the impact of any meaningful scenarios
are measured.

Capturing understanding
The sequence of operations in the process description are described and recorded in order.
The operation to add a reactive reagent for example would be described by documenting: (a)
reagent name, concentration, amount (actual or by reference to other key material) and time
scale (rate) (b) the complexity of the reaction mass, e.g. rheology, the number of phases
(present, formed or destroyed). (c) the intended change or mechanism (A>B reaction), (d)
physical properties (boiling / melting point) and (e) experimental data (heat data/ reaction
profile).

A more simple operation, e.g. to heat or cool, would require much less description.
 F. Muller et al.

Risk Evaluation
For each step possible ‘Scenarios’ are generated Figure 2.1: Example of a risk matrix for a stage
with 12 operations
using classic (HAZOP type) guidewords; e’g longer
time, higher temperature. Any scenario that
generates a meaningful outcome is documented, 4,5,9 4,6 10,12

H
noting the mechanism involved and the nature of

Impact on conditions
impact (e.g. yield, quality, operability). The
scenarios are then weighted on two scales: (i) the
likelihood to change on scale and (ii) the potential

M
4,7,9 2,7,7 1,2,3,3,9
impact on operability or quality of product. The risk
are indicated as low, medium or high. The score is a
subjective one, which represents the view of the
team. If the assessment of the risk is difficult 3,8,12* 5,7,11 12*

L
because information is lacking, it scores
automatically as “High” to drive the development
team to generate of understanding in this area. The L M H
distribution of risk over various scenarios is not Impact on quality and operability
reported in this paper. * Scenario represents an opportunity

Output generation
The excel tool automatically populates a pictorial risk matrix (see figure 2.1) and a results
summary table. The range of risks identified can vary significantly but the later in the
development lifecycle and the more development resource applied the lower the balance of
risk profile should become.

3. Analysis of SURE study output

The output from 16 SURE studies, resulting in 591 scenarios was analysed. For all operations
we counted the number of times a specific scenario occurs. Rather than listing each specific
scenario (i.e. higher, lower Temperature), they where classed into a number of groups (e.g
temperature; see table 3.1). The data suggests that 62% of the scenarios relate to time,
temperature, more/less, agitation, physical properties and the presence of a new compound.
Hulshof (2000) identified the main cause of scale up problems to be related to mass transfer
and mixing, followed by longer processing times and heat transfer. Clearly we identified the
same issues, though not in the same order of
importance. In addition this work demonstrates that Table 3.1: Number of times a scenario is
physical properties (or lack there of!) and new encountered
materials appearing in the plant (O2, water, Fe, Ni) scenario type Total number
are also considerable sources of concern. none 81 13.7%
Time 78 13.2%
In contradiction to Hulshof(2000) we find that heat more/less 63 10.7%
Temperature 61 10.3%
& mass transfer, raw material quality and reaction
PhysProps 55 9.3%
profiles figured low in the scenarios generated (This
Agitation 54 9.1%
data set does not contain a hydrogenation though)
New Compound 43 7.3%
This probably reflects the degree to which these Crystallisation 33 5.6%
issues are “chemineerd” out. Chemists, very much Concentration 33 5.6%
aware of potential scale up and hazard issues, avoid Heat transfer 24 4.1%
complex rheologies, ensure heat generation/removal Particle 19 3.2%
rates are low and use reactive gasses only as a last Other Scenarios 18 3.0%
resort (other then hydrogen for which purpose built Omit 13 2.2%
equipment may be available on site). Mass transfer 12 2.0%
RM quality 2 0.3%
Profile 2 0.3%
 Anticipation of scale up issues in pharmaceutical development 5

Table 3.2: Mechanisms affected by the key scenarios

Time (13%): Temperature (11%)
• Reaction (degradation, by-product formation, • Reaction (Decomposition, hydrolysis, impurity
conversion/yield) formation)
• Reaction profile • Solids (melting, solidification of liquids)
• Particle (Super-saturation, nucleation, growth, • Crystallisation (Faster, thicker mixtures)
attrition and form)
• Gas evolution and foaming
• Distillation
More/less Species, Shear(10%) Agitation (9%)
• Reaction (Rate, Concentration, Degradation,) • Solid suspension
• Filtration (compression) • Emulsification
• Phase split, emulsion • Blending, reaction selectivity
• Others: Gas evolution, evaporation, solubility, • Inclusion
Impurities to next stage
Physical properties (9%) New Compounds (7%)
• Crystal (form, hydrate, nucleation) • Ingress (O2, MOC, tap water vs distilled)
• Oiling out
• Solubility
• Density

In table 3.2 we list what mechanism was affected for each of the main scenario types. In
addition to Hulshof’s original list other sources of significant risk are: (i) a lack of
understanding in reaction mechanisms, (ii) crystallisation & nucleation and (iii) agitation
requirements. Of these three, nucleation is the most difficult to scale up. Especially as one is
not always aware the system is supersaturated….

Finally in table 3.3 we show the distribution of the generated scenarios over the various
process operations. Clearly the charging operations generate by far the most scenarios. This
reflect the fact that many operations include a charge (e.g. reactions, phase separations,
filtration washes, drown out crystallisations). Similarly, the “hold” operation is typically
executed in order to complete reaction or crystallisation; both operations that carry a high
level of risk.

Heating and cooling, as well as typical “work up” operations like “wash”, “Distill” and
“phase separate” account also for a significant part of the scenarios. More detailed analysis
revealed that the risk these operations pose falls in a few categories:

• Wash (filtrations): Insufficient removal of impurities, impurities precipitate, residual

wash liquor concentration is high, degradation (hydrolysis)
• Distill: super-saturation, nucleation too early, wrong final composition, degradation
• Phase sep: Wrong split, phase inversion, emulsification
• Heat/Cool: unexpected nucleation, degradation, over reaction (more impurities), more
evaporation

4. Participant’s feedback

Clearly, a robust approach like SURE requires a significant investment of time from the
development team. To ensure the SURE process develops and made best use of people’s time,
feedback on the methodology is always discussed at the end of a SURE study. Typically, both
chemists and engineers are enthused by the process. For many engineers it was the first time
they discussed all aspects of a process in detail, which gave them a much better insight in
what is planned, and why it is required.
 F. Muller et al.

Table 3.3: Distributions of the scenarios types over typical batch processing operations

l
ta
p

Sc e P
Se

To
ry k
l is
R er

ec

w
e

en
te

al

ng
sf
e

nd

nd
rg

ct

ch

flo
h
l

st
as

ita

R
til

t
d

an

re
as

lt e
ha

oo

ea

ea

ha

ra

ra
scenenario type

ol

EO
is

pH

2
Ph

Ag
W

Tr

Fi

G
C

C
none 36 6 9 6 11 5 1 2 1 1 1 2 81 14%
Time 27 11 5 11 1 3 8 5 2 1 1 1 1 1 78 13%
more/less 32 1 3 9 8 1 1 1 2 2 1 2 63 11%
Temperature 26 7 7 1 3 5 4 2 3 2 1 61 10%
PhysProps 24 3 3 15 2 1 2 2 2 1 55 9%
Agitation 29 11 2 3 1 2 2 1 2 1 54 9%
New Compound 19 5 3 3 4 2 1 1 1 4 43 7%
Crystallisation 9 2 6 5 2 3 1 1 1 3 33 6%
Concentration 14 2 3 6 3 1 1 2 1 33 6%
Heat transfer 9 2 9 2 1 1 24 4%
Particle 4 3 4 2 1 2 2 1 19 3%
Other Scenarios 6 6 1 3 1 1 18 3%
Omit 5 5 1 1 1 13 2%
Mass transfer 3 2 1 1 1 1 2 1 12 2%
RM quality 1 1 2 0%
Profile 1 1 2 0%
Grand Total 245 61 40 38 38 36 35 25 16 12 11 10 9 7 6 1 1 591
41% 10% 7% 6% 6% 6% 6% 4% 3% 2% 2% 2% 2% 1% 1% 0% 0% 100%

Chemists are key contributors to the discussions, providing most of the information. They
generally find the methodology useful as it gives them additional areas to focus on and
increases the development team’s confidence in the process

5. Conclusions

This paper presents a new methodology to assess the scale up risk of batch processes under
development: Scale Up Risk Evaluation (SURE). This methodology is has been applied
successfully to 16 processes in AstraZeneca providing the development teams with a ranked
list of scenarios that could be a possible threat to robust scale up.

The various threats to robustness can be evaluated by looking at the nature of the scale up
scenarios generated. The most frequent deviations from lab conditions relate to changes in
time, temperature, agitation, physical properties and the number of species. The reason these
deviations generate potential scale up risk lies in the mechanism affected. The paper presents
a whole range of important mechanisms. The paper highlights two key sources of risk:
(i) (unwanted) reactions and (ii) the unpredictable nature of nucleation of a new phase.
Surprisingly heat and mass transfer do not rank high. The view is that experienced
development chemists avoid these issues during process design.

Another interesting point revealed by this study is the fact that most deviations of anticipated
behaviour are expected when charging additional material to the batch; This reflects the fact
that compositional changes that occur on a charge lead to reaction and nucleation. Typical
work up operations (wash, distill, filter) also generate a significant number of potential scale
up issues.

The SURE methodology is generally seen as a valuable addition to the development

community as it addresses process robustness in a rigorous way. An additional benefit is that
after a SURE study, the process is generally better understood by the development team as a
whole.
 Anticipation of scale up issues in pharmaceutical development 7

6. Aknowledgements

The authors like to thank M. Wernersson and P. Crafts for the SURE studies they facilitated
and L Bell and S Knight for the compilation of the “Right-First-Time” data.

References

Britest website (2007): www.britest.co.uk/description_of_tools.php

Carey J.S., Laffan D., Thomson C., Williams M.T. (2006). Org.biomol. Chem., 4, 2337
Double, J.M., Gourlay, B. and Atherton, J.H. (2005), Survey of process intensification
equipment requirements, 7th World Congress of Chemical. Engineering Proceedings,
O156–003 (IChemE, ISBN 0 85295 494 8).
Hulshof L.A (2000). Organic Process Research and Development, July 10-12 2000,
Montreal, Canada
Kepner C H, Tregoe B B (1981). The New Rational Manager. Princeton Research Press
Sharratt PN and Borland JN, (2003). AIChE Annual meeting, San Fransisco, November,
Paper 145b.
Sherlock J.P., Brewis N (2006). AIChemE Process Development, Symposium, June 11-14 ,
2006, Palm Spring CA, USA