Development of a control strategy for a defluorinated analogue in the manufacturing process of casopitant mesylate

Article abstract of DOI:10.1021/op1000836

Casopitant mesylate was identified as part of the search for drugs with activity on the Central Nervous System (CNS) by GlaxoSmithKline. During late-phase development studies to develop the manufacturing process, a new impurity was found. This synthetic i

Full text of DOI:10.1021/op1000836

Organic Process Research & Development 2010, 14, 832–839
Development of a Control Strategy for a Defluorinated Analogue in the
Manufacturing Process of Casopitant Mesylate
Fernando Bravo,*,† Zadeo Cimarosti,† Francesco Tinazzi,† Damiano Castoldi,† Paul Stonestreet,‡ Annalisa Galgano,† and
Pieter Westerduin†
GlaxoSmithKline R&D, Chemical DeVelopment, Via Fleming, 4, Verona, Italy, and GlaxoSmithKline R&D, Chemical
DeVelopment, Old Powder Mills, Near Leigh, Tonbridge, Kent TN11 9AN, U.K.
Abstract:
controls, derived from current product and process understand-
ing, that assures process performance and product quality. The
commercial process to synthesise casopitant mesylate (1) is a
multistage convergent process, summarized in Scheme 1. The
mesylate salt 1 is obtained after 8 stages, including two
hydrogenation reactions (the first one using Rh/C as catalyst in
stage 1, the second using Pd/C as catalyst in stage 2) and one
reductive amination in stage 7.
In an advanced phase of the development of the compound,
a new impurity, the defluorinated analogue 6 (see Scheme 2),
was identified in isolated intermediate 2 at levels of ∼0.1%
a/a.
Given the potential risk for the quality of the drug substance
if this impurity is carried through to drug substance as is or
after synthetic transformations, the development of a control
strategy supported by appropriate process understanding was
therefore considered key to ensure appropriate and consistent
quality of the drug substance or drug product.
Casopitant mesylate was identified as part of the search for drugs
with activity on the Central Nervous System (CNS) by Glaxo-
SmithKline. During late-phase development studies to develop the
manufacturing process, a new impurity was found. This synthetic
impurity, a defluorinated analogue of the drug substance, is
discussed in detail to show the process development studies carried
out to ensure quality control for the final drug substance following
the principles of quality by design. The process understanding
gained, in combination with risk analysis, allowed the development
of a control strategy for enhanced level of quality assurance. This
Control Strategy allows moving the control of this impurity to the
point of origin instead of testing in the drug substance.
1. Introduction to Quality by Design and Process Back-
ground
Quality by design (QbD) is a new approach for process
development to ensure the patients’ needs and product perfor-
mance by which quality is not just tested in the final drug
substance or drug product, but it is built in within the process.
The principles of QbD were initially introduced by the FDA in
a seminal communication1 and have been increasingly adopted
by the pharmaceutical industry as a means to deepen the
scientific understanding to ensure quality control and improve
process and product robustness.2 A number of regulatory
guidelines to illustrate the QbD approach have followed the
initial communication: for example, ICH Q83 describes an
enhanced approach by the use of process understanding,
whereby process performance over a range of material attributes,
manufacturing process options, and process parameters is
considered; ICH Q94 discusses quality risk management tools
to perform a risk assessment and risk mitigation; ICH Q105
introduced the concept of control strategy, defined as a set of
For the reader’s benefit, a glossary section with the
definitions of the terms used within this text is included
in Appendix 1.
2. Discussion: The Control Strategy
The impact of the defluorinated impurity 6 on the quality of
the drug substance was analysed by conducting a risk assess-
ment, as indicated by ICH Q9,4 which identified appropriate
process understanding studies required for mitigation of the key
risks. To assess the impact on the process, we examined the
fate and origin of this new impurity. The impurity fate was
determined by reacting pure impurity 6 through stages 4-7.
With this approach, it was possible to conclude that the
defluorinated analogue of the drug substance (7 in Scheme 2)
was formed. More importantly, this impurity had the same
retention time of the drug substance 1 in the HPLC method
used for its release.
* To whom correspondence should be addressed. E-mail: fbravo@iciq.es.
Current address: Institute of Chemical Research of Catalonia (ICIQ); Av. Països
Catalans 16; 43007 Tarragona (Spain).
An ad-hoc, nonroutine analytical method for the separation
and quantification of this particular impurity 7 with respect to
the drug substance 1 was developed, which employed a chiral,
normal-phase HPLC coupled with an MS detector.6 This new
method allowed confirmation that impurity 7 was present in
batches of drug substance at levels of ∼0.1% w/w, that is, this
impurity is a drug substance-critical quality attribute (drug
† GlaxoSmithKline R&D, Verona, Italy.
‡ GlaxoSmithKline R&D, Kent, U.K.
(1) Pharmaceutical cGMPs for the 21st centurysA risk based approach
(initiative launched in 2002).
(2) We have recently published an application of the QbD principles to
the control of genotoxin impurities: Cimarosti, Z.; Bravo, F.;
Stonestreet, P.; Tinazzi, F.; Vecchi, O.; Camurri, G. Org. Proc. Res.
DeV. [Online early access]. DOI: 10.1021/op900242x. Published
Online: Nov 6, 2009.
(6) Dams, R.; Bernabé, E.; Nicoletti, A.; Loda, C.; Martini, L.; Papini,
D. Quantitation of Defluorinated Analogue of Casopitant Mesylate
by Normal Phase Liquid Chromatography-Mass Spectrometry.
Manuscript in preparation.
(3) ICH Q8 Pharmaceutical Development.
(4) ICH Q9 Quality Risk Management.
(5) ICH Q10 Pharmaceutical Quality System.
832
•
Vol. 14, No. 4, 2010 / Organic Process Research & Development
10.1021/op1000836  2010 American Chemical Society
Published on Web 06/17/2010

Scheme 1. Some details of the commercial route for the synthesis of casopitant mesylate (1)
Scheme 2. Reaction pathways in the formation of the defluorinated impurity 6
substance-CQA, see glossary in Appendix 1). Moreover, this
level constituted a high risk for the drug substance to fail the
quality criteria, because at the point in time when this was
discovered, it was not possible to obtain a toxicological
qualification for this derivative 7, which meant that this impurity
would need to be controlled at levels below 0.15% w/w
(according to ICH guidelines).7 This risk impelled us into
applying, once again, the concepts of QbD, specifically, to gain
(7) ICH Q6a Specifications: Test Procedures and Acceptance Criteria for
New Drug Substances and New Drug Products: Chemical Substances.
Vol. 14, No. 4, 2010 / Organic Process Research & Development
•
833

an understanding of the scientific basis for the formation of the
defluorinated impurity 7 to be able to minimise it. In addition,
the application of QbD would facilitate the moving of the
attribute control upstream and controlling the impurity at its
point of origin, instead of testing it in the drug substance. The
rationale that supports bringing the control upstream is detailed
in the next sections, and it contributes to the simplification of
the problem because no modification of the registered method
for drug substance specifications would be required.
A root cause analysis was used to identify the source of the
formation of the impurity 6 in the isolated intermediate 2 and
the impurity 7 in the final drug substance 1. The details of this
analysis are reported in the Supporting Information. This
analysis brought us to the following possible explanations for
the presence of the defluorinated by products 6 and 7:
2.1. Introduction to the Elements of Control of the
Control Strategy. In general, each element of control can be
categorised into one of three control modes. For the particular
case of the drug substance-CQA 7, examples of all three control
modes were applied: (i) attribute controls, which include in-
process controls (IPCs), and specifications for starting materials,
intermediates, solvents, and drug substance, (ii) parametric
controls, which involve operation within proven acceptable
ranges (PARs) for parameters that have an impact on drug
substance-CQAs (referred to as critical process parameters,
CPPs, in ICH Q83), and (iii) procedural controls, which describe
operations linked to drug substance-CQAs, such as facilities
setup, equipment configuration, order of addition, reagent and
solvent choice, sequence of events, etc.
2.2. Initial Studies to Identify Potential Critical Process
Parameters. Process research studies were undertaken to
understand which parameters contribute to the formation of the
impurity 6 during the hydrogenation of the double bond or the
hydrogenolysis. Previous knowledge gained in the development
was taken into account to identify these parameters, and, in
particular, the information was derived from initial multivariate
studies (design of experiments, DoE). When these multifactorial
experiments were performed the impurity 6 had not been yet
identified, and considering that impurity 6 results from an over-
reduction reaction, it was decided to consider a similar over-
reduction reaction as a model (reduction of the ketone group
on the same substrate) and select those parameters that
influenced this model reaction. The parameters of interest and
their ranges are collected in Table 1. It is worth noting that
1. The source could be the starting materials used. The
impurity 6 in isolated intermediate 2 could result from
the presence of the corresponding defluorinated ana-
logue of the starting material 5 (8 in Scheme 2), which
would be transformed into 6, and then in turn into 7
through the process chemical transformations.
2. The defluorinated impurities 6 and/or 7 could be
generated in the process, in particular as a result of a
reduction reaction:
a. Impurity 6 in isolated intermediate 2 could be
formed during one or both the hydrogenations
performed in stages 1 and 2.
b. Additionally, we considered that a further source
for impurity 7 could be the reductive amination
occurring in stage 7.
Table 1. Parameters investigated in the initial DoE of stage 1
The first and last hypothesis (i.e., 1 and 2b) were discarded
on the basis of the following: the defluorinated impurity
(hypothesis 1) could not come from the starting material because
a marker of 8 was available, and it was confirmed that the
starting material 5 was free of 8 as contaminant. The defluori-
nation could not have occurred in the reductive amination
(hypothesis 2b) because, when the reaction mixture was left
for extended times, higher temperatures or higher equivalents
of reducing agent that were added increased levels of impurity
7 compared to standard operating conditions did not occur.
Therefore, it was concluded that either the hydrogenation
using Rh/C as catalyst in stage 1, the hydrogenolysis of the
Cbz group using Pd/C as catalyst in stage 2, or both8 contribute
to the formation of the defluorinated analogue 6, which is then
converted into 7 when performing stages 4-8. It is worth
mentioning that for procedural simplification, the hydrogena-
tions were performed sequentially, this is, without filtration of
the spent Rh/C catalyst once the formation of 4 was complete,
followed by addition of Pd/C to the reaction mixture and in
situ hydrogenolysis to form 5.
parameter
levels studied (L/M/H)
5% w/w Rh/C loading
(C in Figure 1)
0.76/2.29/3.82% w/w (based on
dry catalyst)
partial pressure of hydrogen
(A in Figure 1)
0.2/1.6/3.0 bar
volume of ethyl acetate
(B in Figure 1)
3/6.5/10 volume
(8) Activation of the strong C-F bond is a topic of growing interest, and
different metals have been reported to perform this transformation.
From a bibliographic search, it would seem more likely that Rh is the
responsible due to a higher number of hits, although both Rh and Pd-
based catalyst have been reported to conduct the defluorination
reaction. Indeed, there is a relatively higher number of examples of
defluorination of fluorobenzenes in the presence of Rh, in particular
employing homogeneous Rh complexes; see for instance: (a)
Aizenberg, M.; Milstein, D. Science 1994, 265, 359–361. (b) Aizen-
berg, M.; Milstein, D. J. Am. Chem. Soc. 1995, 117, 8674–8675. (c)
Braun, T.; Noveski, D.; Ahijado, M.; Wehmeier, F. Dalton Trans.
2007, 3820–3825. (d) Noveski, D.; Braun, T.; Stammler, A.; Stammler,
H.-G. Dalton Trans. 2004, 4106–4119. (e) Young, R. J.; Grushin, V. V.
Organometallics 1999, 18, 294–296. Less common is the use of
metallic Rh: ref 8e (Rh nanoparticles). (f) Freedman, L. D.; Doak,
G. O.; Petit, E. L. J. Am. Chem. Soc. 1955, 77, 4262–4263 (Rh/Al2O3).
(g) Stanger, K. J.; Angelici, R. J. J. Mol. Catal. A: Chem. 2004, 207,
59–68 (Rh supported on SiO2 and Pd-SiO2). In less extent, Pd-based
catalyst have also been successfully employed. (h) Ukisu, Y.;
Miyadera, T. J. Mol. Catal. A: Chem. 1997, 125, 135–142 (hetero-
geneous Pd/C). (i) Aramend´ıa, M. A.; Borau, V.; Garc´ıa, I. M.;
Jime´nez, C.; Marinas, A.; Marinas, J. M.; Urbano, F. J. C. R. Acad.
Sci. Paris, Ser. IIC, Chim. 2000, 3, 465–470. In contrast with the
literature precedents, the information collected by catalyst suppliers
suggests the use of Pd/C for defluorination, see: Handbook of
Pharmaceutical Catalysis; Johnson Matthey Catalysts: Royston, U.K.,
2009, p 37.
Once the source of impurity 7 was identified (hypothesis
2a), a series of experiments to understand the formation of the
impurity 6 (sections 2.2-2.5) and its relationship with impurity
7 (section 2.6) were undertaken. This would facilitate the
identification of a series of mechanisms of control that would
ensure the quality of the drug substance with respect to this
particular drug substance-CQA.
834
•
Vol. 14, No. 4, 2010 / Organic Process Research & Development

Figure 1. Half-normal plot for the Rh/C hydrogenation: over-reduced by product.
temperature is not included in this list because it was established
that it did not influence the model reaction.
The experimental design chosen was 2-level (one high and
one low value for each parameter of the study), full factorial
experimental design comprising two centre points (10 experi-
ments in total).
(partial pressure of hydrogen, catalyst loading, and temperature)
are all considered potential CPPs of the reaction. These forcing
experiments were performed to understand if only one catalyst or
the combination of both catalysts contributed to the formation of
impurity 6. In most cases, the spent Rh/C catalyst was filtered after
completion of stage 1 to differentiate both catalysts and their impact.
Table 2 collects the information gained with these experiments.
The experiments collected in Table 2 allowed the following
conclusions:
Figure 1 shows the analysis of the data obtained from these
experiments and provides an assessment of the impact of the
parameters on the over-reduction byproduct. The graphical
representation is obtained from the commercially available
software Design Expert Version 7.0.3, and it is called a half-
normal plot. This half-normal plot is used to assess the
significance of the individual parameters within the design. This
can be seen as the parameters which have a higher effect also
have a corresponding higher numerical x-axis value on the
graph, and therefore, they appear further to the right-hand side
of the half-normal plot.9 From this figure, it is deduced that the
most significant parameters impacting the over-reduction are
pressure (A in Figure 1), catalyst loading (C in Figure 1), and
their interaction (AC in Figure 1). Volume of ethyl acetate does
not have significant effect on the response, at least in the ranges
explored. The validity of this model is further addressed with
the same software by an analysis of variance (ANOVA, more
details are reported in the Supporting Information), which
indicated that the parameters of pressure and catalyst loading
have statistical significance on the response.
2.3. Forcing Experiments to Confirm Potential Critical
Process Parameters. Our initial risk assessment suggested
including partial pressure of hydrogen and catalyst loading as
factors in a subsequent forcing experiment because they do have
an impact on the formation of over-reduction by product. After
further risk assessment, it was decided to include the temperature
of the reaction in these forcing experiments because higher
temperatures could accelerate the kinetics of formation of the
defluorinated impurities. In a QbD context, these parameters
1. In all the experiments, the amount of impurity 9 (see
Scheme 2), the defluorinated analogue of non isolated
intermediate 4, is negligible or very small.
2. The Rh/C catalyst also promotes the formation of
minor amounts (∼5% a/a) of compound 3 (i.e., a minor
amount of the product of stage 2 is formed after
completion of stage 1). The levels of 3 after stage 1
increased to ∼20% a/a in the case of the stressed
conditions of entry 4.
3. Entry 1 represents the standard operating conditions,
without intermediate filtration of the spent Rh/C
catalyst once stage 1 is completed. Quite surprisingly,
these conditions gave the highest levels of impurity 6
of the experiments performed.
4. Under the same conditions of entry 1, the inclusion of
an intermediate filtration of Rh/C (entry 2) gave only
small levels of the defluorinated impurity 6.
5. Stressed hydrogenation conditions in stage 2 (Pd/C
hydrogenolysis, entry 3) did not contribute to an increased
formation of the impurity 6 compared to entry 2.
6. Stressed hydrogenation conditions in stage 1 (Rh/C
hydrogenation, entry 4) resulted in the formation of
impurity 6 at much higher levels compared to entry 2.
From the comparison of entries 2 and 3, it could be deduced
that Pd/C does not contribute to the formation of impurity 6,
only Rh/C is responsible for its formation. Furthermore, it is
also apparent that Rh/C performs the defluorination on the
substrate 3 (the product of stage 2) and not on substrates 4 or
5. To further confirm this hypothesis, a couple of experiments
with compound 3 were performed, as collected in Table 3.
(9) The y-axis of the graph is defined as “cumulative probability, a measure
of the % of individuals expected to fall below a given level”. The
x-axis is defined as “a measure of the magnitude of the effect”.
Vol. 14, No. 4, 2010 / Organic Process Research & Development
•
835

Table 2. Stressed experiments on stage 1 and stage 2
filtration of spent Rh/C
level of 6 in non isolated
intermediate 3
entry
conditions for stage 1
at end of stage 1
conditions for stage 2
1
2
3
4
TVa
no
TVa
0.26% a/a
0.03% a/a
0.02% a/a
0.21% a/a
TVa
yes
yes
yes
TVa
TVa
stressedb
TVa
stressedb
a TV refers to target value conditions: partial pressure of hydrogen ) 2.5 bar; T ) 25 °C; loading of 5% Rh on Charcoal ) 0.0229 wt; loading of 5% Pd on charcoal )
0.0302 wt (based on dry catalyst). b Stressed conditions are as follows: partial pressure of hydrogen ) 4.0 bar; T ) 40 °C; loading of 5% Rh on Charcoal ) 0.0358 wt;
loading of 5% Pd on Charcoal ) 0.0469 wt (based on dry catalyst).
Table 3. Stressed experiments performed on 3 as substrate
demonstrated impact on the quality of the drug substance, or a
potential impact identified with the use of strict risk assessments
tools, suitable ranges of operation (PARs) are needed. Our
approach to define and verify the PARs for these parameters
was to perform verification experiments, run at 1 L scale. This
scale was considered adequate to account for any scale
dependency, on the basis that agitation conditions at equivalent
power-per-unit volume (P/V) ratio were used in vessel with the
same geometric configuration as the commercial vessel. The
verification experiments were based on a statistical methodology
whereby two sets of conditions are run, namely, forcing and
mild conditions and where all parameters are modified at once
in the forcing and mild conditions. The aim is to maximise or
minimise the effect of the formation of the impurities, which
takes into account inherently the interaction among parameters.
The results of these experiments are presented in Table 4. These
tests were performed without the intermediate filtration of the
spent Rh/C catalyst after completion of stage 1 to have a worst
case scenario.
for Rh/C and Pd/C hydrogenations
level of 6 in non isolated
entry
conditions
intermediate 3
1
2
none (initial value)
stresseda Rh/C
0.02% a/a
0.59% a/a
hydrogenation conditions
3
stresseda Pd/C
0.02% a/a
hydrogenation conditions
a Stressed conditions are as follows: partial pressure of hydrogen ) 4.0 bar; T
) 40 °C; loading of 5% Rh on charcoal ) 0.0382 wt; loading of 5% Pd on
charcoal ) 0.0464 wt.
These experiments confirm that Pd/C does not contribute to
the formation of the defluorinated impurity 6 because levels in
entry 3 (stressed Pd/C hydrogenation conditions) are equivalent
to the initial value. It is worth noting that, when the starting
material 5 is reacted with Pd/C, no defluorination is observed,
only debenzylation occurs, the resulting deprotected dihydro-
pyridone is purged in the crystallization and does not contami-
nate the intermediate 2.
The experiment in entry 2 confirms the role of the Rh/C
catalyst in the defluorination on the substrate 3.
All the previous information allowed generating a mecha-
nistic scheme for the formation of defluorinated impurities,
which is summarised graphically in Scheme 2.
One logical conclusion that can be reached by observing
the pathway for the formation of the defluorinated impurity 6
is to include a filtration for the spent Rh/C catalyst after
completion of stage 1. This unit operation, although introducing
a minor complexity in the procedure, would have a considerable
impact in the quality of isolated intermediate 2 because the
conversion of 3 into 6 would be minimised, and hence improve
the quality of the final drug substance 1. The filtration constitutes
the first element of control (procedural control) of the control
strategy.
2.4. Risk Assessment and Definition of Proven Accept-
able Ranges. The previous information also confirms that, when
Rh/C catalyst is used, the parameters of the reaction (i.e., partial
pressure of hydrogen, Rh/C catalyst loading and temperature)
do have an impact on the level of defluorinated impurity 6 that
is formed and are, therefore, confirmed as CPPs. Furthermore,
a thorough risk assessment (see below) identified low volumes
of reaction as one potential CPP; the rationale is that substrate
5 has low solubility in the ethyl acetate used as reaction solvent,
and a slow reaction leading to long contact time between the
substrate/products and the catalyst/hydrogen system can take
place under these conditions, which would in turn favor over-
reduction pathways. For such parameters, which show a
Table 4. Verification experiments to set PARs for
parameters in Rh/C hydrogenation
defluorinated
impurity 6 in isolated
entry
conditions
intermediate 2
1. (verification forcing) P ) 2.5a + 0.3 bar of H2
T ) 25a + 12 °C
0.05% a/a
cat loading ) 0.0229a + 0.0027 wt
V ) 2.0a - 0.6 vol
2. (verification mild)
P ) 2.5a - 0.3 bar of H2
T ) 25a - 12 °C
0.03% a/a
cat loading ) 0.0229a - 0.0027 wt
V ) 2.0a + 0.6 vol
a These are the target value conditions.
Because the derived intermediate 2 contained the impurity
6 at levels lower than 0.15% a/a (this level of impurity 6 in
intermediate 2 is considered appropriate as explained in section
2.6), the ranges obtained for these parameters are considered
valid and therefore constitute their PARs. Working inside the
PARs for these parameters collected in Table 5 represents a
second element of control (parametric control).
2.5. Final Risk Assessment: Definition of Quality Critical
Process Parameters and Quality Process Parameters. Glaxo-
SmithKline has its own internal QbD methodology, whereby a
risk assessment process and a risk quantification tool (failure
mode and effects analysis, FMEA, see ICH Q9)4 are used to
discern between two types of CPPs. On the basis of the
following criteria, these parameters are classified:
1. Quality critical process parameter (QCPP) is a param-
eter that influences a drug substance-CQA and has a
836
•
Vol. 14, No. 4, 2010 / Organic Process Research & Development

Table 5. PARs for parameters in Rh/C hydrogenation
a. Partial pressure of hydrogen: TV ( 12%.
b. Temperature of the Rh/C hydrogenation: TV (
48%.
c. Rh/C catalyst loading: TV ( 10.5%.
d. Volume of solvent for the Rh/C hydrogenation:
gTV - 30%.
parameter
lower PAR
upper PAR
partial pressure of hydrogen
temperature of the
Rh/C hydrogenation
Rh/C catalyst loading
volume of solvent for
the Rh/C hydrogenation
2.2 bar
13 °C
2.8 bar
37 °C
0.0202 wt
1.4 vol
0.0256 wt
nonea
3. A procedural control, whereby filtration of the spent
Rh/C catalyst, once stage 1 is complete, is performed.
This control certainly helps in reducing the levels of
impurity 6 in intermediate 2, but because the previous
PARs had been determined without intermediate filtra-
tion of the spent catalyst, it is not strictly necessary to
obtain drug substance of adequate quality, and we
could keep the flexibility of removing the filtration to
simplify the process if required for operational reasons
at a later stage. Indeed, this demonstrated an advanta-
geous outcome of the QbD paradigm, which allows
the introduction of regulatory flexibility, once a sound
scientific rationale has been found.
4. A further level of attribute control was introduced in
our control strategy on the basis of risk assessment,
which is to avoid that the levels of 6 in intermediate
2 could be higher by unexpected contamination of the
starting material 5 with its defluorinated analogue 8.
Although all batches of the starting material 5 analysed
to date were free of impurity 8, we set a restrictive
specification limit for this impurity of 0.05% a/a.
a On the basis of experiments, the volume of solvent impacts only because of
limited solubility of that gives slower reaction and favors over-reduction
pathways, and therefore, only a lower PAR limit is required.
5
high risk to fall outside the design space (i.e., the
PAR). QCPPs typically require tight controls.
2. Quality process parameter (QPP) is a parameter that
influences a drug substance-CQA and has a low risk
to fall outside the design space (i.e., the PAR).
The FMEA gives a numeric assessment of the risk derived
from a multidisciplinary assessment, where functional experts
from organic chemistry, engineering, analytical sciences, and
plant operations are present to evaluate the risk on the basis of
existing scientific understanding and process knowledge. In this
case, the risk that the drug substance would fail to comply with
the quality requirements for the drug substance-CQA 7 was
considered low for all parameters, and therefore, no QCPPs are
found; only QPPs remain.
2.6. Correlation of Impurity 6 with Drug Substance-
Critical Quality Attribute 7. Spiking experiments demon-
strated that the relationship of the impurity 7 in the drug
substance with respect to impurity 6 in intermediate 2 is
approximately 1:1 (see Table 6). These data are derived from
batches produced at scale (100 kg scale for entries 2-6). This
result constitutes another element of the control strategy: to
obtain drug substance 1 with levels of 7 of no more than (NMT)
0.15% w/w, it is possible to move the control upstream by
placing a specification for impurity 6 in isolated intermediate 2
of NMT 0.15% a/a (attribute control).
The control strategy is summarised graphically in Figure 2.
3. Conclusion
By following the principles of QbD, we have derived a
control strategy based on the above-mentioned four elements
of control. The application of the QbD principles has allowed
us understanding the basic science behind the formation of the
impurity, introducing a degree of regulatory flexibility (e.g.,
working inside the conditions of the PARs for QPPs; desirable,
but optional filtration of the spent Rh/C catalyst), and moving
the control upstream, which constitutes a simplification of the
analytical procedure.
Table 6. Levels of impurity 7 in the final drug substance 1
as a function of level of impurity 6 in intermediate 2
impurity 6
present in
intermediate 2
impurity 7
present in
drug substance 1
4. Experimental Section
entry
General Procedure for the Hydrogenation and Hydro-
genolysis: Synthesis of 2-(4-Fluoro-2-methylphenyl)-4-pip-
eridinone (3). In a clean and N2/vacuum-purged hydrogenation
reactor, phenylmethyl 2-(R/S)-(4-fluoro-2-methylphenyl)-4-oxo-
3,4-dihydro-1(2H)-pyridinecarboxylate (5) (50 g, 1 wt) is
introduced. A catalytic amount of 5% Rh on Charcoal (1.145
g, 0.0229 wt based on the dry catalyst) is added into the reactor,
and then the reactor is purged with three cycles of N2/venting.
Ethyl acetate (100 mL, 2 vol) is added, the reactor is purged
once more with 3 cycles of N2/vacuum, followed by 5 cycles
of H2/vacuum, and then it is pressurized to 2.5 bar of H2. The
reaction is stirred for ca. 1 h at 25 °C until complete conversion
of the starting material into non isolated intermediate 4. The
catalyst is filtered and the spent catalyst is washed with ethyl
acetate (2 × 15 mL, 2 × 0.3 vol). The filtered catalyst waste is
disposed appropriately. The reactor is purged with 3 cycles of
N2/vacuum, and 5% Pd on charcoal (1.51 g, 0.0302 wt based
on the dry catalyst) is added over the reaction mixture. The
1
2
3
4
5
6
0.18% a/a
0.10% a/a
0.06% a/a
0.04% a/a
0.05% a/a
0.10% a/a
0.12% w/w
0.06% w/w
0.06% w/w
0.03% w/w
0.04% w/w
0.08% w/w
2.7. Summary of the Control Strategy for Defluorinated
Drug Substance-CQA 7. From all the above information, an
overall control strategy for the control of the defluorinated drug
substance-CQA 7 is derived, which comprises the following
elements of control:
1. Attribute control, whereby a specification limit of
NMT 0.15% w/w of impurity 7 in drug substance 1 is
replaced with a specification limit of NMT 0.15% a/a
of impurity 6 in isolated intermediate 2.
2. A parameter control point, whereby the process is run
within the PARs determined for the QPPs, as follows:
Vol. 14, No. 4, 2010 / Organic Process Research & Development
•
837

Figure 2. Overall control strategy for defluorinated impurity 7 (drug substance-CQA).
reactor is sealed, purged with 3 cycles of N2/vacuum, followed
by 5 cycles of H2/vacuum, and finally pressurized at 2.5 bar of
H2. The stirring is kept at 25 °C, and the headspace is vented
with cycles of H2/vacuum and repressurized with H2 regularly
throughout the reaction in order to remove the CO2 until
complete conversion into 3 (ca. 1 h) The mixture is filtered
and the solution of 3 in EtOAc is collected. The reactor and
the spent catalyst are washed with ethyl acetate (1 × 50 mL -
1 × 1 vol -; then 2 × 25 - 2 × 0.5 vol). Finally, the filtered
catalyst waste is disposed appropriately.
Compound 3 can be precipitated as the (rac)-camphorsul-
fonate salt as follows: 2-propanol (100 mL, 2 vol) is added and
the solution is concentrated to 175 mL (2.5 vol). The resulting
solution is diluted with more 2-propanol (225 mL, 4.5 vol),
and then racemic camphorsulfonic acid (34.0 g, 0.68wt)
dissolved in 2-propanol (100 mL, 2 vol) is added at room
temperature in 5 min, and the resulting solution is stirred for
10 min. The solution is concentrated to 175 mL (3.5 vol),
toluene (550 mL, 11 vol) was added and the mixture concen-
trated to 175 mL (3.5 vol); the racemic camphorsulfonate salt
of 3 precipitates from the medium. More toluene is added (325
mL, 6.5 vol), and the slurry was stirred overnight (ca. 14 h) at
room temperature. The solid was isolated by filtration, and
washed three times with toluene (3 × 50 mL, 3 × 1 vol), dried
in a vacuum oven at 40 °C for 15 h to give the racemic
camphorsulfonate salt of 3 (40.51 g, 74% th).
added. Once the reaction started (the yellow-brown colour of
the solution disappeared and the solvent started refluxing), the
remaining 2-bromotoluene (27.4 mL, 1.37 vol) was added
dropwise (within 20 min) keeping a gentle reflux. The reaction
mixture was stirred at reflux for 30 min then cooled to room
temperature, giving the Grignard solution. Contemporaneously
and in a separate flask, benzyl chloroformate (33.2 mL, 1.66
vol) was added dropwise (within 15 min) to a solution of
4-methoxypyridine (20 g, 1 wt) in dry THF (200 mL, 10 vol)
at 0 °C. After the addition was completed, the white suspension
obtained was stirred at 0 °C for 30 min. The Grignard solution
was added dropwise (within 20 min) over the suspension at 0
°C and the reaction mixture was stirred at 0 °C for 1 h. Then
a 20% w/w solution of NH4Cl (80 mL, 4 vol) was added at 0
°C and the mixture was stirred at 20 °C for 5 min. More water
(50 mL) was added to enhance the separation of the phases.
The two layers were separated and the organic layer was treated
with a 95:5 mixture of 20% w/w solution of NH4Cl/20%
solution of HCl (80 mL, 4 vol). The aqueous phase was
discarded, and the organic layer was concentrated, and 2-pro-
panol was added (this operation was repeated twice). The solid
formed was collected and washed with 2-propanol. The solid
was dried under vacuum in the oven at 50 °C overnight, giving
8 (47.48 g, 80.6% th).
1H NMR (400 MHz, DMSO-d6): δ ppm 8.22 (dd, 1H),
7.07-7.35 (m, 9H), 5.82 (d, 1H), 5.37 (d, 1H), 5.13-5.23 (m,
2H), 3.29 (dd, 1H), 2.32-2.39 (m, 1H), 2.27 (s, 3H). HRMS
(ES+) Calcd. for C20H19NO3 [M + H]+: 322.1443; found
322.1449.
8 (256.2 g, 1 wt) was charged in an hydrogenation reactor
and reacted with hydrogen as described in the general procedure
for hydrogenation and hydrogenolysis; after removal of the
solvent, crude 6 was obtained (150 g). Compound 6 can be
precipitated as the (rac)-camphorsulfonate salt as follows: To
50 g of the crude obtained after the hydrogenation/hydrogenoly-
sis, 2-propanol was added (120 mL) and then concentrated
under vacuum. This operation was repeated. Then more
1H NMR (600 MHz, DMSO-d6): δ ppm 9.51 (br.s., 1H),
9.32 (br.s., 1H), 7.70 (dd, 1H), 7.19 (m, 1H), 7.16 (dd, 1H),
4.98 (d, 1H), 3.62 (m, 2H), 2.95 (dd, 1H), 2.90 (m, 1H), 2.89
(d, 1H), 2.63 (m, 1H), 2.55 (m, 1H), 2.53 (m, 1H), 2.40 (d,
1H), 2.37 (s, 3H), 2.24 (m, 1H), 1.94 (t, 1H), 1.85 (m, 1H),
1.80 (d, 1H), 1.28 (m, 2H), 1.03 (s, 3H), 0.74 (s, 3H). MS: m/z
208 [M + H]+, as free base.
Synthesis of the Defluorinated Impurity 2-(2-Methylphe-
nyl)-4-piperidinone (6). To a suspension of Mg (6.2 g, 0.31
wt) and iodine (120 mg, 0.006 wt) in dry THF (140 mL, 7 vol)
at 60 °C, a portion of 2-bromotoluene (1.0 mL, 0.05 vol) was
838
•
Vol. 14, No. 4, 2010 / Organic Process Research & Development

2-propanol (425 mL, 8.5 vol) was added followed by (rac)-
camphorsulfonic acid (49.2 g, 0.8 equiv). The mixture was
stirred for 2 h, and then 180 mL of 2-propanol were added.
The solid was collected by filtration and washed with 2-propanol
(2 × 85 mL), cyclohexane/2-propanol 1/1 (170 mL) and
cyclohexane/2-propanol 3/1 (255 mL). The solid was dried
under high vacuum at room temperature, obtaining of the
racemic camphorsulfonate salt of 6 (52.47 g, 46.8% th).
1H NMR (400 MHz, DMSO-d6): δ ppm 9.42 (br.s, 2H),
7.66 (dd, 1H), 7.31-7.37 (m, 2H), 7.28 (dt, 1H), 4.99 (dd, 1H),
3.54-3.68 (m, 2H), 2.90-2.99 (m, 2H), 2.90 (d, 1H), 2.59-2.69
(m, 1H), 2.50-2.59 (m, 2H), 2.41 (d, 1H), 2.36 (s, 3H),
2.20-2.28 (m, 1H), 1.94 (t, 1H), 1.80-1.91 (m, 1H), 1.80 (d,
1H), 1.23-1.34 (m, 2H), 1.03 (s, 3H), 0.74 (s, 3H). HRMS
(ES+) Calcd for C12H15NO [M + H]+: 190.1232; found
190.1235.
high risk to the process falling outside the design space or
proven acceptable ranges in the unit operation or stage inputs,
stage outputs, device etc.
Quality Attribute in the Unit Operation or Stage Inputs,
Stage Outputs, Device etc measurable property of inputs and
outputs that (as determined by Risk Assessment) present a low
risk to the process falling outside the design space or proven
acceptable range.
Critical Process Parameter (CPP) process parameter
whose variability has an impact on a critical quality attribute
and therefore should be monitored or controlled to ensure the
process produces the desired quality [ICH Q8].
Design Space multidimensional combination and interaction
of input variables (e.g., material attributes) and process param-
eters that have been demonstrated to provide assurance of
quality; working within the design space is not considered as a
change, and movement out of the design space is considered
to be a change and would normally initiate a regulatory post
approval change process. Design Space is proposed by the
applicant and is subject to regulatory assessment and approval
[ICH Q8].
Quality Critical Process Parameter process parameter that
influences a critical quality attribute and (as determined by risk
assessment) presents a high risk to the process falling outside
the design space or proven acceptable ranges.
Quality Process Parameter process parameter that influ-
ences a critical quality attribute but (following a risk assessment)
presents a low risk of the process falling outside the design
space or proven acceptable ranges.
Control Strategy (planned) set of controls, derived from
(current) product and process understanding that assures process
performance and product quality; the controls can include
parameters and attributes related to drug substance and drug
product materials and components, facility and equipment
operating conditions, in-process controls, finished product
specifications, and the associated methods and frequency of
monitoring and control. (ICH Q10 definition).
Proven Acceptable Range (PAR) upper and lower limits
for process parameter or attribute values between which the
parameter or attribute is known to produce a process output
(e.g., intermediate, API or DP) that meets the CQAs; the PAR
may or may not represent the point of failure, and the PAR for
a given process parameter or attribute may be dependent upon
the PAR values for one or more other process parameters or
attributes (e.g., multivariate).
Synthesis of the Defluorinated Impurity (2R,4S)-4-(4-
Acetyl-1-piperazinyl)-N-{(1R)-1-[3,5-bis(trifluoromethyl)phe-
nyl]ethyl}-N-methyl-2-(2-methylphenyl)-1-piperidinecarbox-
amide (7). Prepared as described in the published literature.10
Compound 7 can be precipitated as the methanesulfonate
salt as described in the published literature.10a
1H NMR (600 MHz, DMSO-d6): δ ppm 9.65 (br.s., 1H),
8.01 (s, 1H), 7.70-7.76 (m, 2H), 7.20 (d, 1H), 7.10 (d, 1H),
7.07 (t, 1H), 7.03 (t, 1H), 5.36 (q, 1H), 4.40-4.49 (m, 1H),
4.24 (dd, 1H), 3.94-4.04 (m, 1H), 3.53-3.62 (m, 1H),
3.43-3.52 (m, 3H), 3.39 (t, 1H), 3.10-3.22 (m, 1H), 2.93-3.03
(m, 1H), 2.84-2.94 (m, 1H), 2.80 (t, 1H), 2.75 (s, 3H), 2.36
(s, 3H), 2.33 (s, 3H), 2.13-2.19 (m, 1H), 2.07-2.13 (m, 1H),
2.02 (s, 3H), 1.85-1.96 (m, 1H), 1.64-1.74 (m, 1H), 1.46 (d,
3H). HRMS (ES+) Calcd for C30H36N4O2F6 [M + H]+:
599.2821; found 599.2805.
Acknowledgment
We are grateful to Orsola Vecchi, Simon Bryant, Riet Dams,
Giulio Camurri, Stefano Provera, Lucilla Turco, Ornella Cur-
curuto, Jill Trewartha, Paola Russo, Anna Nicoletti, Luca
Martini, Matteo Gonzi, Vern De Biasi, Vance Novack, Damiano
Papini, and Luca Mantilli for helpful discussions and to Matt
Kersey for critical reading of the manuscript.
Appendix 1: Glossary
Drug Product Critical Quality Attributes or Drug
Substance Critical Quality Attributes measurable properties
of drug product or API that are critical to ensuring patient safety
and efficacy. The property must be within a predetermined range
to ensure product quality. A property which is measured outside
the range indicates a batch failure.
Supporting Information Available
Critical Quality Attributes measurable properties of inputs
and outputs that (as determined by risk assessment) present a
Detailed description of the root cause analysis mentioned
in paragraph 2 by using the Ishikawa diagram and the ANOVA
analysis of the DOE reported in paragraph 2.2. This material is
available free of charge via the Internet at http://pubs.acs.org.
(10) (a) Alvaro, G.; Di Fabio, R.; Maragni, P.; Tampieri, M.; Tranquillini,
M. E. PCT Int. Appl. WO 2002/032867, 2002; Chem. Abstr. 2002,
136, 325568. (b) Bientinesi, I.; Cimarosti, Z.; Guercio, G.; Leroi, C.;
Perboni, A. PCT Int. Appl. WO 2007/048642, 2007; Chem. Abstr.
2007, 146, 461973. (c) Guercio, G.; Bacchi, S.; Perboni, A.; Leroi,
C.; Tinazzi, F.; Bientinesi, I.; Hourdin, M.; Goodyear, M.; Curti, S.;
Provera, S.; Cimarosti, Z. Org. Process Res. DeV. 2009, 13, 1100–
1110.
Received for review March 26, 2010.
OP1000836
Vol. 14, No. 4, 2010 / Organic Process Research & Development
•
839