Small-Volume Continuous Manufacturing of Merestinib. Part 2. Technology Transfer and cGMP Manufacturing

Article abstract of DOI:10.1021/acs.oprd.8b00442

Technology transfer of a small volume continuous (SVC) process and Current Good Manufacturing Practices (cGMP) manufacturing of merestinib are described. A hybrid batch-SVC campaign was completed at a contract manufacturing organization under cGMP. The decision process by which unit operations were selected for implementation in flow for the cGMP campaign is discussed. The hybrid process comprised a Suzuki-Miyaura cross-coupling reaction, a nitro-group hydrogenolysis, a continuous amide bond formation, and a continuous deprotection. A continuous crystallization using two mixed suspension, mixed product removal (MSMPR) crystallizers and a filtration with in situ dissolution were employed for purification between the two SVC steps. Impurity levels were monitored using both online process analytical technology (PAT) and offline measurements. The continuous processing steps operated uninterrupted for 18 days to yield the drug substance in solution at a throughput of 12.5 kg/day. Crystallization in batch mode afforded 183 kg of the drug substance in specification. Success of the campaign was attributed to robustness of the control strategy and to the multiyear partnership in continuous manufacturing between the development organization and the contract manufacturer. Key learnings are offered from the perspectives of both the development organization and the contract manufacturer.

Full text of DOI:10.1021/acs.oprd.8b00442

Subscriber access provided by UNIV OF TEXAS DALLAS
Full Paper
Small-volume continuous manufacturing of merestinib
part II: technology transfer and cGMP manufacturing
Brandon J. Reizman, Kevin Paul Cole, Molly Hess, Justin L. Burt, Todd D. Maloney, Martin D. Johnson,
Michael E. Laurila, Richard F. Cope, Carla Vanesa Luciani, Jonas Y. Buser, Bradley M. Campbell, Mindy
B. Forst, David Mitchell, Timothy Braden, Christopher K. Lippelt, Moussa Boukerche, Derek Starkey,
Richard D. Miller, Jing Chen, Baoquan Sun, Martin Kwok, Xin Zhang, Sam Tadayon, and Ping Huang
Org. Process Res. Dev., Just Accepted Manuscript • DOI: 10.1021/acs.oprd.8b00442 • Publication Date (Web): 12 Mar 2019
Downloaded from http://pubs.acs.org on March 13, 2019
Just Accepted
“Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted
online prior to technical editing, formatting for publication and author proofing. The American Chemical
Society provides “Just Accepted” as a service to the research community to expedite the dissemination
of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in
full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully
peer reviewed, but should not be considered the official version of record. They are citable by the
Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore,
the “Just Accepted” Web site may not include all articles that will be published in the journal. After
a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web
site and published as an ASAP article. Note that technical editing may introduce minor changes
to the manuscript text and/or graphics which could affect content, and all legal disclaimers and
ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or
consequences arising from the use of information contained in these “Just Accepted” manuscripts.
is published by the American Chemical Society. 1155 Sixteenth Street N.W.,
Washington, DC 20036
Published by American Chemical Society. Copyright © American Chemical Society.
However, no copyright claim is made to original U.S. Government works, or works
produced by employees of any Commonwealth realm Crown government in the course
of their duties.

Page 1 of 35
Organic Process Research & Development
1
2
3
4
5
6
7
8
9
Small-volume continuous manufacturing of
merestinib part II: technology transfer and cGMP
manufacturing
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Brandon J. Reizman,§ ‡* Kevin P. Cole,§ ‡* Molly Hess,§ Justin Burt,§ Todd D. Maloney,§ Martin D. Johnson,§ Michael E. Laurila,§
Richard F. Cope,§# Carla V. Luciani,§ Jonas Y. Buser,§ Bradley M. Campbell,§ Mindy B. Forst,§ David Mitchell,§Δ Timothy M.
Braden,§ Christopher K. Lippelt,§ψ Moussa Boukerche,§a Derek R. Starkey,§ Richard D. Miller,§
Jing Chen,† Baoquan Sun,† Martin Kwok,† Xin Zhang,† Sam Tadayon,† Ping Huang†
AUTHOR ADDRESS
§
Small Molecule Design and Development, Eli Lilly and Company, Indianapolis, IN 46285, USA
†
Shanghai SynTheAll Pharmaceutical Co., Ltd. (“STA”), No. 13 Building, #90 Delin Rd, WaiGaoQiao Free Trade Zone, Shanghai
200131, P.R. China
ACS Paragon Plus Environment
1

Organic Process Research & Development
Page 2 of 35
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
TOC GRAPHIC
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
ACS Paragon Plus Environment
2

Page 3 of 35
Organic Process Research & Development
1
2
3
4
5
6
7
8
9
ABSTRACT
Technology transfer of a small volume continuous (SVC) process and Current Good Manufacturing Practices (cGMP) manufacturing of
merestinib are described. A hybrid batch-SVC campaign was completed at a contract manufacturing organization under cGMP. The
decision process by which unit operations were selected for implementation in flow for the cGMP campaign is discussed. The hybrid
process comprised a Suzuki-Miyaura cross-coupling reaction, a nitro-group hydrogenolysis, a continuous amide bond formation, and a
continuous deprotection. A continuous crystallization using two mixed suspension, mixed product removal (MSMPR) crystallizers and a
filtration with in situ dissolution were employed for purification between the two SVC steps. Impurity levels were monitored using both
online process analytical technology (PAT) and offline measurements. The continuous processing steps operated uninterrupted for 18 days
to yield in-solution drug substance at a throughput of 12.5 kg/day. Crystallization in batch mode afforded 183 kg of in-specification drug
substance. Success of the campaign was attributed to robustness of the control strategy and to the multi-year partnership in continuous
manufacturing between the development organization and the contract manufacturer. Key learnings are offered from the perspectives of
both the development organization and the contract manufacturer.
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
KEYWORDS
continuous processing, small volume continuous, flow chemistry, continuous crystallization, scale-up, flow-NMR
INTRODUCTION
ACS Paragon Plus Environment
3

Organic Process Research & Development
Page 4 of 35
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
Continuous manufacturing (CM) is a maturing technology that is finding increased usage globally among pharmaceutical
manufacturing organizations.1,2,3,4,5,6,7 The transition toward CM has been motivated by the need to enable faster, more exothermic, and/or
greener chemical transformations in pharmaceutical synthesis, by the ability to reduce process footprint using an emerging suite of
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
intensified unit operations, and by the increased process insights brought about by more extensive use of process analytical technology
st
(
PAT) in the control of product quality.8,9,10 As newer, 21 Century pharmaceutical manufacturing facilities are constructed, it is
anticipated that there will be motivation to reduce capital costs by replacing larger equipment for batch processing with leaner, more
modular equipment for CM.11
In the past 10–15 years, there has been a shift toward the broader use of contract manufacturing organizations (CMOs) in
pharmaceutical manufacturing, driven prominently by cost pressures in the pharmaceutical industry13,14,15 and by the emergence of low
12
16
cost cGMP manufacturing partners in India and in China. Assuming that the trend toward increased utilization of CM technology in
pharmaceutical syntheses continues, it is of growing importance that CMOs gain familiarity with CM technology. Since the manufacture
of early developmental products is frequently outsourced, it is advantageous to be able to conduct flow unit operations at CMOs in order
to install enabling chemistry or separations early in the product lifecycle. Having a CM-capable CMO network also helps sustain the global
supply chain of pharmaceutical compounds and intermediates, lessening risk that compounds constructed from transformations that are
unamenable to batch processing incur production delays when the demand for continuous processing technology exceeds internal
capacity.
As the reliance on CM increases, the success of chemistry, manufacturing, and controls development for pharmaceutical production
will be contingent on delivery of a control strategy that leverages the strengths of CM, while accounting for the challenges added by
process intensification. The U.S. Food and Drug Administration has commented on the role of CM in implementation of a control strategy,
17
citing the importance of understanding process dynamics in the ability to control quality attributes. These challenges associated with
the dynamics of the process include design for startup and shutdown operations, mitigation of process disturbances, and the need for
ACS Paragon Plus Environment
4

Page 5 of 35
Organic Process Research & Development
1
2
3
4
5
6
7
8
9
material diversion strategies. For manufacturing organizations entering into CM, addressing these technical challenges may introduce
additional operational challenges, including the need to manage and monitor richer sets of process data, the need for around-the-clock
campaign staffing, and the need for real-time or nearly real-time analytical feedback to confirm that the process is in a state of control.
Collaboration between development and manufacturing organizations is imperative to ensure successful technology transfer and product
quality throughout production.12
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
In Part I of this publication, Lilly described development of a process for the small volume continuous (SVC) manufacturing of
18
merestinib (1, LY2801653, Scheme 1). In the route presented in Part I (Scheme 1), the Suzuki-Miyaura cross-coupling between bromide 2
and pyrazole 3 generated nitro-compound 4 (Step 1), which underwent hydrogenolysis to afford the aniline 5 (Step 2). Addition of mixed
anhydride 6 to aniline 5 afforded the amide 7 (Step 3), which finally underwent deprotection in a heated flow reactor to afford 1 (Step 4).
OEt
N
OEt
N
N
N
Br
F
OEt
N
Step 1
Step 2
F
F
O
N
PdCl2[dtbpf]
K3PO4, water/THF
H2
+
O
O
THF, catalyst
NO2
N
N
B(pin)
N
N
NO2
HN
N
N
NH2
2
3
4
5
OEt
N
N
N
Step 3
+ 6
THF
Step 4
F
F
5
NH Cl
4
O
F
O
F
O
O
THF/water
O
O
N
N
N
N
N
N
N
N
H
H
7
1
F
Anhydride Formation
F
O
O
O
O
O
PivCl
Et3N, THF
HO
N
O
N
8
6
Scheme 1. Synthesis of merestinib.
ACS Paragon Plus Environment
5

Organic Process Research & Development
Page 6 of 35
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
Following a successful 20 kg process demonstration, a team of Lilly chemists and engineers was afforded approximately six months to
orchestrate technology transfer of the continuous process to a CMO for delivery of more than 125 kg cGMP drug substance (DS) suitable
for Phase 3 clinical trials. Given the abbreviated timeline for technology transfer, the Lilly team elected to adapt the demonstrated fully
continuous process to operate as a hybrid batch-continuous process at the CMO. Summarized in Table 1, the team’s decisions to convert
CM unit operations to batch mode were made based upon the intended process control strategy for the cGMP campaign and the
robustness of the unit operations at this stage of process development. Step 1, which was operated during the CM demonstration in a
repeating batch mode for the Suzuki-Miyaura cross-coupling reaction, was readily converted to a fully batch process, which reduced the
overall technological and logistical complexity of implementing the cGMP sequence at the CMO. Perceived challenges in delivering aniline
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
5
from the hydrogenolysis reaction in a continuous reactor within an abbreviated timeline contributed to the team’s decision to also
18
execute the Step 2 hydrogenolysis in batch mode. Discussed in Part I, these challenges associated with continuous hydrogenolysis
technology at the time included lack of development time to characterize catalyst bed life, the high dilution required to achieve a
homogeneous feed stream, the criticality of adjusting process conditions to control for purity, and occasional episodes of reactor fouling.
Operation in batch mode also decoupled the process cycle time for Step 2 from the throughput of the downstream continuous unit
operations, and thus the hydrogenolysis reaction time could be lengthened on the basis of in-process control (IPC) sampling.
Table 1. Chosen modes of operation for cGMP campaign.
Process Step
Implementation for cGMP Campaign
Rationale
Step 1
Batch
Ease of implementation of existing process in
batch mode
Step 2
Batch
Complexity of rapid transfer of continuous
reactor to CMO
Ability to extend reaction time in batch mode
based on IPC
ACS Paragon Plus Environment
6

Page 7 of 35
Organic Process Research & Development
1
2
3
4
5
6
7
8
9
Step 3
Step 4
Continuous
Demonstration of impurity control strategy
under cGMP
Continuous Reaction
Batch Crystallization
Batch Isolation
High temperature / pressure needed to enable
reaction.
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Insufficient capacity for CM crystallization at
CMO
Preference for clearly delineated batch size,
quantity
Critical to the process impurity control strategy was suppression in the DS of process intermediates 4 and 5, as well as the many
process impurities potentially introduced during hydrogenolysis. Design of the batch process for Step 2 was conducted with the goal of
controlling for indazole 4 and any hydrogenolysis impurities in the resulting aniline 5. Control of the aniline 5 was then dependent on the
process design in Step 3, namely assurance of near-complete consumption of 5 in the Step 3 reaction and precise control over the Step 3
crystallization to afford consistent rejection of any unreacted 5. The opportunity presented by CM to minimize process variability in the
unit operations in Step 3 motivated the Lilly team’s decision to operate this step fully in continuous mode. The Step 4 deprotection
19
reaction was designed to operate at high temperature and pressure and was also operated in continuous mode. For the DS
crystallization, impurity rejection was shown in the earlier process demonstration to be comparable between batch and continuous modes;
however the DS crystallization kinetics were slow such that long processing times were necessary to achieve high product yields. In lieu
of installing more and/or larger crystallization vessels for the CM process, the DS workup and crystallization were chosen to operate in
batch mode. This decision also allowed for clear delineation of DS batches from the hybrid batch/continuous process.
STA Pharmaceutical was selected for cGMP production of 1 using the hybrid process. In advance of this selection, Lilly had partnered
with STA for approximately three years to strengthen STA’s position in CM. This collaboration included visits of STA scientists to Lilly to
ACS Paragon Plus Environment
7

Organic Process Research & Development
Page 8 of 35
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
learn and train on the use of CM equipment at both development and commercial manufacturing scale, and led to the development of
subject matter experts in flow chemistry and engineering at STA that communicated regularly with Lilly scientists on CM. Lilly also sited
multiple early-phase projects at STA that leveraged CM. Even with these prior successes, the installation of a multi-step continuous
process at STA was found to require even more dedication and collaboration between Lilly and STA to ensure that the project would be
staffed by trained personnel for several weeks of around-the-clock operation, including both preparation and campaign activities. Activities
also included installation of PAT and process automation when deemed most critical, configuration of new equipment, and development
of a control strategy and batch record consistent with STA’s quality system.
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
This publication summarizes the development activities at Lilly and STA that culminated in the organizations’ collective ability to
deliver cGMP material suitable for Phase 3 clinical trials. Technology transfer of a hybrid batch/continuous version of the process to STA
was completed to produce 183 kg of 1 under cGMP manufacturing, including operation of the two continuous steps for 18 days at a
throughput of 12.5 kg DS/day (prior to batch crystallization of the DS). Conclusions cite learnings from both Lilly and STA regarding the
successful transfer of a multi-step continuous process.
CONTROL STRATEGY DEVELOPMENT
The control strategy for the hybrid cGMP process considered development knowledge acquired before and during the SVC process
18
demonstration (discussed in Part I) in concert with the introduction of batch processes for Steps 1 and 2. As discussed in Part I, the
placement of the hydrogenolysis step in the synthetic route in Scheme 1 necessitated significant modifications to the process control
strategy in preparation for the cGMP campaign. Most noteworthy were the many impurities to be controlled at or prior to the DS.
Reduction of 4 to the aniline 5 was known to proceed through the pathway in Scheme 2, generating intermediates nitroso 9,
hydroxylamine 10, azoxy 11, azo 12, and hydrazo 13.20,21,22 Other process impurities generated during the hydrogenolysis or in forward
processing of hydrogenolysis intermediates to the amide bond formation step are listed in Table 2. Under certain processing conditions,
ACS Paragon Plus Environment
8

Page 9 of 35
Organic Process Research & Development
1
2
3
4
5
6
7
8
9
reduction leading to loss of the fluoro substituent (14) and/or the ethoxy-ethyl protecting group (15) on 5 were observed. Importantly,
process intermediate 5 and process impurity 15 were potential genotoxic impurities (GTIs) and were controlled to a target of less than 80
ppm in the DS for the hybrid cGMP campaign. Additional hydrogenolysis impurities included ethyl-adduct 16, which formed from reaction
of 5 with acetaldehyde from the cleaved ethoxy-ethyl protecting group, and hydroxybutyl adduct 17, which formed as a byproduct with
the use of tetrahydrofuran (THF) as the solvent. Forward processing of hydroxylamine 10 was observed to generate the hydroxamic acid
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
18. Likewise, forward processing of 14 afforded defluorinated amide impurity 19. Impurity 15 had the potential to undergo a second amide
bond formation reaction in Step 3 to generate either the bisamide 20 or the pivaloyl-adduct 21. Pivalamide 22 was a potential byproduct
from reaction of 5 with anhydride 6 or acid 8.
H2
R1 NO2
R1 NO
R1 NHOH
R1 NH2
Pd/C
EtO
4
9
10
5
N N
F
R =
1
R1
N
R1 NH
O
R1
N
O
N R1
HN R1
N R1
N
11
12
13
N
Scheme 2. Hydrogenolysis intermediates and process impurities.
Table 2. Other key process impurities.
F
O
O
W
N
N
N
X
O
Pyridone
Y
O
N
N
N
Z
t-Bu
EtO
EE
Piv
Process Impurity Formation Conditions
W
X
Y
Z
ACS Paragon Plus Environment
9

Organic Process Research & Development
Page 10 of 35
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
1
4
5
Defluorination of 5
EE
H
H
F
H
H
H
H
1
EE deprotection during
hydrogenolysis
16
Reductive amination
acetaldehyde)
(5 +
EE
F
H
CH CH
2
3
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
17
18
19
Reaction in presence of THF
Acylation of 10
EE
EE
F
F
H
F
F
F
H
OH
H
(CH ) OH
2
4
Pyridone
Pyridone
Pyridone
Pyridone
Piv
Amidation of 14
EE
20
21
Di-acylation of 15
Pyridone
Piv
H
Di-acylation of 15
H
22
Pivalamide formation
EE
H
Crystallization for impurity control at Step 2 was avoided because of the containment needed for isolation of the potential GTI 5.
Likewise, prior development work indicated that a single crystallization of the amide 7 would be sufficient to achieve impurity targets,
and thus it was preferred to use crystallization at Step 3 as the primary rejection point for all hydrogenolysis-derived impurities. This
necessitated extensive fate and purge experiments for Step 3, as well as the DS crystallization at Step 4. Impurity purge data from the
Step 3 and Step 4 crystallizations informed the Step 2 end-of-reaction targets for all hydrogenolysis-derived impurities and the nitro
intermediate 4. A set of preliminary batch crystallization studies explored the rejection of potential Step 2 and Step 3 process impurities.
These impurities were spiked into the Step 3 crystallization in batch mode in three experiments to confirm that the rejection of the
impurities was constant in the region of interest. The experiments included (i) a seeded isothermal batch crystallization at 20 °C, (ii) a
seeded isothermal batch crystallization at 50 °C, and (iii) a final seeded isothermal batch crystallization at 50 °C, wherein the impurity
levels were adjusted to satisfy proposed impurity tolerance levels and to account for any reduced rejection efficiency in the multi-
component mixture. The spiked impurity levels are provided in Figure 1a. As shown in Figure 1b, rejection efficiencies of the tested
ACS Paragon Plus Environment
10

Page 11 of 35
Organic Process Research & Development
1
2
3
4
5
6
7
8
9
impurities did not vary significantly with temperature or with the change in initial concentration at 50 °C. Rejection efficiencies in excess
of 60% were observed for all impurities except 18, 19, 20, and 21.
Crystallization of an impurity-rich Step 3 reaction solution was repeated in continuous mode using a series of two 250 mL baffled
mixed suspension, mixed product removal (MSMPR) crystallizers. This laboratory-scale MSMPR crystallization system has been described
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
23
previously. The same three impurity-rich product solutions prepared for the batch crystallization studies were assessed in these studies.
The antisolvent (cyclohexane) and solution of the intermediate 7 were each fed to the first of the two MSMPR crystallizers. Isothermal
crystallization conditions, with both MSMPR crystallizers operating at 20 °C and with 4 h mean residence time, were originally selected
because a continuous cooling crystallization was known to introduce additional risk of solids fouling in the transfer lines between
crystallizers. As shown in Figure 1c, the isothermal procedure resulted in relatively poor rejection of many impurities in comparison to
batch mode. Upon increase of the temperature of the first MSMPR crystallizer to 50 °C (while maintaining a temperature of 20 °C in the
second crystallizer), impurity rejection was observed to greatly improve. Use of a combined antisolvent and cooling continuous
crystallization afforded high rejection efficiency for many significant process impurities, including >90% rejection of the aniline 5. Similarly
to the results in batch mode, poorer rejection efficiencies were observed for 18, 19, 20, and 21. Further repetition of the cooling
crystallization procedure with adjusted impurity levels revealed all rejection efficiencies to be nearly unchanged at the spiking level of
interest. (Note that impurities 13 and 18 were not included in this final experiment.) Based on these extensive impurity rejection studies,
it was decided for the cGMP campaign to operate the Step 3 continuous crystallization under non-isothermal conditions despite the added
risk of solids fouling in the transfer lines between the two MSMPR crystallizers. Re-dissolved Step 3 intermediate 7 obtained from the
spiking study was forward processed through the final deprotection step to confirm acceptable purity of the DS following a batch workup
and isolation procedure. Based on the collective results of these fate-and-purge studies, in-process limits for key impurities were
established for the cGMP campaign.
ACS Paragon Plus Environment
11

Organic Process Research & Development
Page 12 of 35
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Figure 1. Impurity rejection study for Step 3 crystallization development. (a) Impurity spiking levels relative to amide 7. (b) Percent
impurity rejection for seeded, isothermal batch crystallization. (c) Percent impurity rejection for continuous crystallization with the first
MSMPR crystallizer at 20 °C or 50 °C and the second MSMPR crystallizer at 20 °C.
ACS Paragon Plus Environment
12

Page 13 of 35
Organic Process Research & Development
1
2
3
4
5
6
7
8
9
Development for the hydrogenolysis focused on identification of batch processing conditions that would deliver aniline intermediate 5
with all Step 2 impurities controlled to below their target levels. Initial batch hydrogenolysis screening experiments were conducted at 250
mg scale with a Biotage® Endeavor using 5% Pd/Al O (Johnson Matthey® A-302099-5 and A-302011-5), 5% Pd/C (Johnson Matthey®
2
3
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
A503023-5 and A405028-5), and 5% Pd(S)/C (Johnson Matthey® A103023-5) catalysts. Promising screening conditions were then tested at
a 5 g scale in a Parr® autoclave (results in Table 3). The THF charge was limited to 50 mL for the scale-up experiments (compared to
approximately 150 mL for full solubility of 4 at room temperature); consequently the potential for formation of solvent-related impurities
was proportionally reduced relative to the continuous process. From Table 3, hydrogenolysis conditions employing 5% Pd/C (A503023-5)
were found to promote the greatest selectivity toward aniline 5 while offering acceptable control of defluorinated 14 and the
hydrogenolysis intermediates, with the exception of hydroxylamine 10. As a reaction intermediate, control of 10 could be facilitated in
batch mode by extending the Step 2 reaction time (if necessary) based on IPC end-of-reaction results.
Table 3. UHPLC analysis of Step 2 hydrogenolysis reaction conducted at 5 g scale in Parr® autoclave. All experiments were conducted at
8 bar and in 50 mL THF. N.D. = not detected by UHPLC (LOQ 0.05%).
Experiment
1
2
3
4
5
Catalyst
A503023-5 (Pd/C) A302011-5
Pd/Al O )
A302099-5
A302011-5
A302011-5
(
(Pd/Al O )
(Pd/Al O )
(Pd/Al O )
2
3
2 3
2 3
2 3
Catalyst Loading
Temperature
Product Purity
Impurities
5 wt%
50 °C
5 wt%
50 °C
5 wt%
30 °C
5 wt%
30 °C
1 wt%
30 °C
99.65%
99.10%
99.50%
99.50%
98.24%
1
0
0.05%
N.D.
0.09%
0.12%
N.D.
0.08%
N.D.
0.09%
N.D.
N.D.
11
N.D.
12
N.D.
N.D.
N.D.
0.38%
ACS Paragon Plus Environment
13

Organic Process Research & Development
Page 14 of 35
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
1
1
1
3
4
5
N.D.
N.D.
0.08%
0.10%
0.08%
N.D.
0.81%
0.07%
0.05%
0.09%
0.08%
0.16%
N.D.
0.10%
0.11%
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
For the 5% Pd/C catalyst system, the impact to changes in processing parameters was investigated via in-process monitoring with
1
19
the use of H and F NMR. Information on the NMR experimental configuration can be found in the Supporting Information, as well as in
previous publications that have described the in-process NMR method and its application to measurement of mass transfer rates.24,25
Manipulated conditions in the in-process NMR experiments included changes to catalyst loading, pressure, temperature, reaction
concentration, catalyst lot, reaction additives, oxygen levels, and agitation rate. Impurities 14 and 15 were found to be well-controlled
within the range of conditions tested. Impurities 12 and 13 formed more prominently under low mass transfer or low catalyst loading
conditions, where the primary pathway to aniline formation became disfavored. Residual sulfur, which could enter into Step 2 from the
palladium removal unit operation in Step 1, was confirmed as a catalyst poison and also resulted in formation of higher levels of 12 and 13
due to higher in situ content of hydroxylamine 10. Longer reaction times and lower concentrations of 4 led to more elevated levels of the
hydroxybutyl adduct impurity 17.
Using the NMR data, a kinetic model was developed that allowed for selection of suitable processing conditions that would control
the defluorinated impurity 14 and hydrogenolysis intermediates to acceptable levels. The model included catalyst deactivation kinetics and
was used to understand the feasible operating region, normal operating ranges and proven acceptable ranges, and to assess criticality. To
lessen the risk of generating impurities 12 and 13, the optimized cGMP process was designed to produce 5 in 30 kg batches at 45 °C and
8
bar, using 3 wt% Pd/C (5%, A503023-5). The optimized batch process afforded the aniline 5 in ≥99.7% purity in lab-scale
ACS Paragon Plus Environment
14

Page 15 of 35
Organic Process Research & Development
1
2
3
4
5
6
7
8
9
demonstrations, with process impurity 14 controlled to ≤0.05 area% and all hydrogenolysis reaction intermediates controlled to the limit of
quantitation (LOQ) of 0.05 UHPLC area%.
The aniline intermediate 5 was to be controlled by ensuring sufficient conversion in the Step 3 amide bond formation reaction,
such that any unreacted 5 would be purged to an acceptable level in the Step 3 crystallization. Kinetics for the consumption of aniline 5
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
26
18
in the amide bond formation reaction were determined in batch mode. In comparison to the non-GMP CM demonstration campaign,
the robustness of the Step 3 reaction in suppressing the impurity 5 was improved by increasing the target feed stoichiometry of the
anhydride 6 to 1.10 equiv. (justified by the crystallization fate-and-purge results for the acid 8 and the anhydride 6) and by extending the
reaction time in the PFR to 140 min. The cGMP Step 3 reactor was divided into two stages: a 9 L segment and a 5 L segment,
corresponding to 90 min and 50 min residence time, respectively, at the intended cGMP campaign flow rates. Acceptable progress of the
reaction after 90 min was assessed using online HPLC, installed between the two PFR stages. At 60 °C, a reaction time of 90 min was
sufficient to achieve less than 0.1 HPLC area% of 5 in the product stream. The second, 5 L stage of the Step 3 reactor was sized to
account for a lag period of 50 min to allow time for material diversion in response to the HPLC analysis prior to the crystallization.
The robustness of the Step 4 deprotection reaction was evaluated in laboratory flow reactor experiments using a 2.5 mL PFR
heated inside of a gas chromatography oven. A univariate study of reaction temperatures demonstrated that acceptable product purity of
≥
99.8% could be achieved between 150–185 °C. Hydrolysis of the DS 1 to aniline impurity 15 was also monitored in these flow reactor
robustness experiments, and it was found that at 170 °C the impurities 5 and 15 would be controlled to acceptable levels for reaction
times of less than 5 h. To afford the targeted crystal form in batch mode, the workup strategy from Part I was modified to include
concentration of the DS 1 product solution, followed by the introduction of cyclohexane to induce a layer separation in the THF/water
system. THF and DMSO were then added to the organic solution, and the solution was concentrated to afford 1 in the THF/DMSO solvent
18
identified in Part I. This workup strategy contributed to removal of the reaction byproduct acetaldehyde to <50 ppm in the DS. A
ACS Paragon Plus Environment
15

Organic Process Research & Development
Page 16 of 35
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
seeded, cooling crystallization with the antisolvent ethanol afforded the desired crystal form. The robustness of this workup procedure was
assessed by imposing univariate process disturbances and assessing the multi-component product mixture for composition by
quantitative NMR. The robustness of the DS crystallization in delivery of the target crystal form was also evaluated through a series of
experiments that tested the seed point, cooling rate, and perturbations to the solution composition including high and low levels of water,
DMSO, cyclohexane and the DS 1.
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
PAT Strategy and Batch Record Development
In addition to kinetic and thermodynamic process understanding that had been gained through development, the control of product
quality was to be ensured based on an analytical strategy and a batch record strategy that verified the process to be operating in the
intended state of control. Performance in Steps 1 and 2 was evaluated based on offline UHPLC, LC-MS, and ICP-MS analysis of the
product of each step. The analytical strategy for the continuous steps of the process was reliant on a combination of online process
analytical technology (PAT) measurements and offline measurements. Online HPLC was installed after the first stage of the Step 3 reactor
to monitor conversion of aniline 5 to amide 7 every 30 min. Using the reaction kinetic model and the measured rejection efficiency of the
aniline 5, an acceptable operating range for the process was established as a function of magnitude and duration of a disturbance in the
Step 3 reaction purity. The established operating range then informed decisions based on the Step 3 online HPLC results that would
instruct the operator to: (i) continue processing, (ii) adjust the feed rates of 5 and 6, or (iii) divert the reactor outflow prior to the
crystallization. These protocols for process manipulations in response to analytical testing were included in the batch record, along with
acceptable ranges for mass flow rates, temperatures, pressures, and liquid fill levels that were monitored on an hourly basis. Following the
in-process crystallization of the amide 7, the re-dissolved solution of 7 was analyzed offline every 8 h by LC-MS to confirm that potential
GTIs 5 and 15 were both below their respective in-process limits. Because of the risk of formation of 15 from hydrolysis under the thermal
ACS Paragon Plus Environment
16

Page 17 of 35
Organic Process Research & Development
1
2
3
4
5
6
7
8
9
deprotection conditions, the Step 4 reaction effluent was also monitored offline every 8 h by LC-MS to confirm that both aniline impurities
were below their in-process limits prior to the DS crystallization.
To accommodate for process dynamics, the batch record for the continuous process was divided by unit operation into separate
procedures for sequential and repeating operations. Sequential operations included procedures for startup, shutdown, filtration and
dissolution, and process pauses and restarts. Once sequential operations were completed, operators transitioned to a “state-of-control”
section of the batch record, where they were instructed to repeat a set of operations at a pre-defined frequency. Tolerances on operating
conditions in the batch record were specified within less than the acceptable ranges for the process, such that minor departures from the
target range could be corrected without the need for material diversion. Departures from the target operating range with the potential to
impact product quality or process safety were explicitly referenced in the batch record. These types of occurrences would have directed
operators to a process pause procedure that identified any requirements for material diversion. A limitation of this batch record strategy
was that the paper batch record had to be subdivided so that operators could perform multiple tasks at or near the same time on different
unit operations. Implementation of an electronic batch record that would alert operators of when tasks were due to be performed would
have greatly simplified the activity of batch record maintenance.
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
CM Equipment Transfer
The equipment build for the continuous process leveraged both existing equipment at STA and newly-fabricated equipment sourced
through D&M Continuous Solutions, LLC (Indianapolis, IN). The design and testing of the equipment from D&M was performed with close
involvement of the Lilly team. This saved considerable time and provided the team with confidence that the equipment would operate as
intended when transferred externally. Automated sequences were configured and tested at Lilly before electronic transfer of the software
to STA and installation on local programmable logic controllers (PLCs). Along with the equipment transfer, D&M and Lilly personnel
traveled to the Jinshan, China, site to assist the STA team with equipment installation and basic training. STA purchased the Lilly-
ACS Paragon Plus Environment
17

Organic Process Research & Development
Page 18 of 35
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
designed equipment from D&M and was responsible for equipment qualification. Lilly provided on-site support at STA for equipment
installation, 10 days of solvent trials, and throughout the duration of the CM portion of the hybrid cGMP campaign.
A flow chart illustrating the CM unit operations in the hybrid cGMP campaign is provided in Figure 2. Select equipment photographs
for Step 3 are provided in Figure 3, and additional photographs and operating procedures can be found in the Supporting Information. The
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
18
sequencing of unit operations for Steps 3 and 4 was the same as in the non-GMP CM laboratory demonstration campaign. Select
changes were made to the process and equipment design for both steps to address findings from the process demonstration. The
continuous process was operated to produce 15 kg/day of 7 from the Step 3 reaction and 12.5 kg/day of 1 from the Step 4 reaction.
Figure 2. Flow chart illustrating the CM portion of the cGMP production campaign.
ACS Paragon Plus Environment
18

Page 19 of 35
Organic Process Research & Development
1
2
3
4
5
6
7
8
9
Figure 3. CM unit operations for cGMP campaign Step 3. (a) two-stage PFR (submerged in heated bath) with online HPLC; (b) MSMPR
crystallizers; (c) filters.
Feed solutions containing aniline 5 and anhydride 6 in THF were dynamically mixed using a series of magnetically-driven stir bars in a
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
27
tube and charged to the series of PFRs, which were submerged in a heated water bath operating at 60 °C and 1.7 bar. To improve
control for stoichiometry at Step 3, Coriolis mass flow controllers (Micro Motion) were installed and configured with the peristaltic pumps
(Masterflex L/S) and a PLC to enable feedback control of the feed rates of aniline 5, anhydride 6, and the antisolvent cyclohexane. The
installation of feedback controllers offered a significant improvement in flow rate regulation compared to previous Lilly processes
employing similar peristaltic pumps which, without a controller, required routine calibration to prevent drift over time. Product solution
exiting the PFRs was transferred through a heated transfer line at 50 °C into the first of two 50 L MSMPR crystallizers, which operated at
50 °C and 15 °C, respectively. Cyclohexane was added to the first MSMPR crystallizer. Slurry was intermittently transferred from the first
to the second MSMPR crystallizer and then was intermittently transferred to one of two parallel agitated 100 L jacketed filters. The
frequency and volume of transfers was such that both MSMPR crystallizers operated with a mean residence time of 4 h. Accumulated
slurry on either filter was continually stirred and cooled to below 20 °C. Sustained cooling ensured impurity control consistent with that
demonstrated in the laboratory two-stage MSMPR crystallizer system and resulted in greater yield of 7 compared to the earlier non-GMP
CM process demonstration. After 8 h of accumulation, the slurry was filtered, washed with a 50:50 mixture of THF/cyclohexane, and then
dissolved in situ in a solution of 85:15 v/v THF/water with 5.0 mol% NH Cl. Unlike the previous non-cGMP process demonstration, all
4
filtration operations in the cGMP campaign were performed manually, and vacuum (rather than pressure) was used to deliquor the wet
cake. The dissolved solution was heated to 35 °C to maintain solubility of the intermediate 7 and transferred to one of two jacketed surge
vessels.
ACS Paragon Plus Environment
19

Organic Process Research & Development
Page 20 of 35
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
At Step 4, feed solution containing the intermediate 7 was pumped and warmed to greater than 100 °C in a series of three heat
exchangers. This solution then continuously flowed into a 7.2 L Hastelloy C276 PFR. This PFR was maintained at 25 bar and was heated
3
to 170 °C within a 1 m electric oven. Prior to use, the PFR was tested with pressurized nitrogen to confirm that it was sealed with no
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
leaks, and pressure relief valves were installed that would divert material to waste in the event of over-pressurization to 100 bar
(
significantly less than the rated pressure of the PFR and connections). The reactor temperature and pressure were continuously recorded
using a PLC and monitored as part of the hourly process inspection. Reaction solution was passed through the reactor for a residence
time of 38 min, cooled in flow through a series of heat exchangers, and then collected in drums that were supplied as feed to the DS
batch workup and crystallization.
cGMP CAMPAIGN
The synthetic route used in the cGMP campaign for Steps 1 and 2 and the preparation of anhydride 6 is summarized in Scheme 3.
Scale-up for the Suzuki cross-coupling reaction proceeded without event, but use of sodium diethyldithiocarbamate for metal treatment
proved challenging. The workup sequence included washing the organic layer with carbamate solution followed by layer separation, polish
filtration to remove precipitated metal complex, partial concentration by solvent distillation, and antisolvent and cooling crystallization. It
was found that two of the three isolated batches of indazole 4 contained high residual sulfur levels, which proved detrimental to the
subsequent hydrogenolysis during use testing. Similar sulfur contamination had only been observed during development if the
concentration step was taken too far and the product nucleated and crystallized in an uncontrolled manner. Recrystallization of 4 was then
conducted, which removed sulfur contamination to acceptable levels. Overall, the Suzuki coupling afforded 181 kg of indazole 4 in 87%
yield.
ACS Paragon Plus Environment
20

Page 21 of 35
Organic Process Research & Development
1
2
3
4
5
6
7
8
9
OEt
OEt
1. PdCl [dtbpf]
2
N N
N N
K PO , water/THF
3
4
60 ºC, 87-94%
F
F
2. Recrystallization
Pd/C, H2, THF
45 ºC, 95-100%
O
O
THF/water, 94-96%
2
+ 3
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
NO2
NH2
N
N
N
N
4
5
F
F
O
O
PivCl, Et3N, THF
O
O
O
10-20 ºC, 90-95%
HO
N
O
N
6
8
Scheme 3. Batch portion of synthetic sequence used in cGMP manufacturing campaign for Steps 1 and 2 and preparation of the
anhydride.
Hydrogenolysis of 4 to aniline 5 was accomplished in six equal sized batches, affording a total of 161 kg of 5 in 98% assay yield and
>
99.7% purity. These batches were combined in pairs and concentrated to the desired ~30 wt% strength to afford three batches, each
traceable to the three different feed lots of the starting material 2. The three batches were stored in drums until use in the continuous
process. UHPLC analysis of the three concentrated feedstocks showed the aniline 5 purity to be >99.7% and that all named impurities
were controlled through Step 2 to below the 0.05% UHPLC LOQ.
Production of the anhydride 6 was completed in two batches, affording 156 kg in 92% averaged yield. As a change from the CM
demonstration, anhydride 6 was crystallized and re-dissolved prior use in the Step 3 process. The revised isolation procedure successfully
alleviated the risk of fouling at Step 3—caused in development by precipitation of residual TEA·HCl—and offered advantages including
improved stability of the anhydride and control of residual pivaloyl chloride. After the anhydride was formed, TEA·HCl was removed by
filtration and the THF solution was partially concentrated under vacuum. Heptane was then added to drive the anhydride 6 out of
ACS Paragon Plus Environment
21

Organic Process Research & Development
Page 22 of 35
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
solution. The solid anhydride was filtered and stored as a solid. Prior to use in the continuous flow sequence, the solid was re-dissolved in
THF and polish filtered to remove low levels of additional TEA·HCl.
A summary of processing Steps 3 and 4 is provided in Scheme 4, and a summary of feed flowrates maintained throughout Step 3
processing is provided in Figure 4. Online HPLC results for the Step 3 reaction are provided in Figure 5. Owing in part to improvements
made to the Step 3 reaction robustness that included adjusting the feed stoichiometry, extending the PFR residence time, and installing
feedback mass flow control, the aniline intermediate 5 was controlled to <0.05 HPLC area% throughout processing. The purity of the
intermediate 7 was generally maintained at ≥99.0 HPLC area%, with the remaining species detected by HPLC being the anhydride 6 and
the acid 8. The process operated uninterrupted and without the need for material diversion, aside from material diversions at process
startup and shutdown. These diversions were specified in the batch record based upon a fixed number of reactor turnovers and
collectively amounted to loss of less than 1% of intermediate 7. Consistent with established acceptable ranges for product quality, the STA
team elected to slow the overall process flowrate on the final day of operation for convenience so that process shutdown procedures could
be performed during daytime hours. The continuous crystallization of 7 also operated without interruption for 17 days. Subsurface solids
coating persisted throughout the campaign in both MSMPR crystallizers but did not impair operation.
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
OEt
1. 85/15 (v/v) THF/H O
2
0.05 equiv. NH Cl
170 ºC, 33-60 min
4
N N
1
. THF
2. Cyclohexane
3. Crystallization
DMSO/Ethanol
60 ºC, 140-220 min
F
2. Continuous Crystallization
Cyclohexane
O
F
O
O
5
+ 6
1
95%
85%
N
N
N
N
H
7
Scheme 4. Synthetic sequence used in cGMP manufacturing campaign for Steps 3 and 4.
ACS Paragon Plus Environment
22

Page 23 of 35
Organic Process Research & Development
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Figure 4. Flow rates of aniline 5 and anhydride 6 for the cGMP campaign (as measured by Coriolis mass flow meter). Flow rates were
intentionally reduced during the evening of day 15 to allow shutdown to occur during daytime hours. Vertical line denotes change in
intermediate 5 feed solution potency, leading to flow rate correction.
Figure 5. Online HPLC measurement of Step 3 reaction purity for cGMP campaign. Aniline 5 was not detected during the campaign (LOQ
=
0.05 area%).
The manual filtration, washing, and dilution operation was the most labor intensive and time consuming aspect of the continuous
process, occupying approximately 5 h of a single operator’s time per 8 h cycle. The process was slowed by the use of vacuum in lieu of
ACS Paragon Plus Environment
23

Organic Process Research & Development
Page 24 of 35
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
pressure filtration, resulting in an average filtration time of greater than 1 h. Shortly after process start-up, the filtration operation was
paused when it was discovered that the THF/water solution of 7 became turbid upon heating to 35 °C, as shown in Figure 6. Closer
observation revealed that this turbidity was caused by a slight phase separation of the THF/water mixture. It was hypothesized that this
phase separation could have been a result of residual cyclohexane retained in the wet cake after vacuum filtration. No impact to quality
was observed, and processing continued as intended. The temperature-dependent phase separation was observed consistently throughout
the campaign.
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Figure 6. Evolution of dissolved compound 7 solution turbidity with increasing temperature.
All solutions of the product 7 were sampled during the transfer to Step 4 and analyzed offline by LC/MS. These analyses showed the
aniline 5 to have been controlled exceptionally well throughout the Step 3 process with a concentration not exceeding 10 ppm with
respect to the amide product 7. The overall purity of the intermediate 7 was consistently found to be greater than 99.9 area% by HPLC,
with the primary byproduct being the DS 1 (the deprotected analog of 7). Defluorinated impurity 19 was occasionally detected at 0.05
area%. The flow rate for the thermal deprotection reaction was adjusted infrequently throughout the campaign within proven acceptable
ranges to accommodate the supply of 7 solution from Step 3. As expected, reducing the flowrate to the Step 4 PFR resulted in
observation of higher levels of 15 (formed via amide hydrolysis of 1). Importantly, analysis of 50 L product drums by LC/MS confirmed that
ACS Paragon Plus Environment
24

Page 25 of 35
Organic Process Research & Development
1
2
3
4
5
6
7
8
9
potential GTIs 5 and 15 were consistently controlled to less than 80 ppm in the reactor effluent (Figure 7). Based on solution assay, the
collective yield for the continuous process was 94%. Approximately 14.2 kg/day of the intermediate 7 was afforded from Step 3 (95% yield)
and 12.5 kg/day of the product 1 was afforded from Step 4 (>99% yield).
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Figure 7. LC/MS analysis of Step 4 reaction effluent product drums showing impurities 5 and 15, relative to 1 (LOQ = 10 ppm).
Product solution from the Step 4 flow process was purified in batch mode following the revised workup and crystallization procedure,
which afforded the targeted crystal form in 85% yield. Batch workup was completed in four approximately equal portions. Two of these
batches were completed before the conclusion of continuous operation of Steps 3 and 4, an illustration of process intensification afforded
by continuous operation. In total the process afforded 183 kg DS, with the Step 3 and 4 hybrid process requiring 25 days to complete.
Product purity was measured at ≥99.95% by HPLC, and impurities 5 and 15 were controlled to less than 10 ppm in the DS. All other
targets for registration stability were satisfied for all DS batches, including residual solvent levels, metals, and crystal form (Table 4).
The overall process yield (Steps 1–4) was 69%. After the cGMP campaign, a residence time distribution model of the CM process train
was constructed to simulate material flow into and out of the reactors, MSMPR crystallizers, filters, and surge vessels. The model was
used to follow the progression of feed lots of the aniline 5 and anhydride 6 throughout the CM process and to calculate the source
material composition of the four product batches.
ACS Paragon Plus Environment
25

Organic Process Research & Development
Page 26 of 35
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
18
Table 4. Summary of non-cGMP CM demonstration campaign at Lilly and CM cGMP campaign at STA. DS for the cGMP campaign was
isolated in batch mode.
CM Demonstration
~100 h
CM cGMP
17 daysc
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Continuous operation time per step
CM throughput
5–10 kg DS/day
20 kg (61%)a
< 10 ppm
12.5 kg DS/day
183 kg (69%)
< 10 ppm
< 10 ppm
≤ 0.05%
DS isolated (overall yield)
Aniline 5 in DS
Deprotected Aniline 15 in DS
Total impurities in DS
65–77 ppmb
≤ 0.1%
Crystal form
Compliant
Compliant
Compliant
Compliant
Residual metals
Compliant
Residual solvents
Compliant
a-Some material was reserved for development activities.
b-Attributed to variable control of Step 3 reaction stoichiometry, and decision not to isolate anhydride 6 leading to fouling event in the
non-GMP laboratory demonstration.
c-Steps 3 and 4 were offset by one day.
KEY LEARNINGS
Perspective from STA
Collaborations in technology transfer from Lilly to STA have significantly sped up the CM capability build at STA. With the onsite
training and direct equipment transfer, STA has rapidly accumulated experience in continuous process development and manufacturing.
ACS Paragon Plus Environment
26

Page 27 of 35
Organic Process Research & Development
1
2
3
4
5
6
7
8
9
Currently STA develops CM processes and technologies independently and has scaled up a number of unit operations in its Jinshan plant
in China. It is critical that process knowledge and operational details are well communicated to achieve successful manufacturing at STA
or other CMOs. STA has been involved in continuous process development for early phase projects, which has led to more experience in
managing continuous cGMP manufacturing campaigns. Such opportunities also exist for other CMOs.
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
STA is still developing in its use of automation technology, and consequently many dedicated chemists, engineers, and operators
were required in order to meet the staffing requirements during the continuous cGMP campaign. Manually intensive responsibilities
included batch record maintenance, the filtration and dissolution unit operation, feed/product solution can changes, and analytical
support. Division of the batch record into several parts (managed in parallel) implied that multiple operators were needed to monitor the
process at a given time while the process operated in a state-of-control. For future campaigns, it is of primary importance to develop
additional automated equipment as well as PAT to minimize manpower requirements and improve the feasibility for multi-step operations
in CM.
Perspective from Lilly
CM technology has matured and now presents an opportunity for CMOs. Capabilities in DS CM may allow for differentiation among
CMOs for processes that involve unstable intermediates, operation under hazardous reaction conditions, containment in fume hoods for
cytotoxic compounds, and/or risk mitigation through process control. It is important that pharmaceutical companies investing in CM
continue to provide CMOs with opportunities to develop CM capabilities and expertise, so that a greater number of individuals can be
trained in the operation of continuous facilities. The nature of continuous processing is such that teams are best composed from
individuals of diverse backgrounds (e.g. process and analytical chemists, chemical engineers, automation engineers, and technicians).
Training and personnel retention at the CMO are potential barriers to externalization of flow technology and present a risk to building
long-term partnerships between pharmaceutical organizations and CMOs. The core pharmaceutical companies invested in CM can lessen
ACS Paragon Plus Environment
27

Organic Process Research & Development
Page 28 of 35
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
this risk by standardizing the continuous equipment sets that are used in manufacturing. CMOs can absolve some of this risk by
appointing a team that repeatedly engages in CM projects, so that learnings are more efficiently transferred from one project to the next.
Even with increased training, technology transfer and cGMP operation of continuous processes at CMOs will be challenging in the
near future. Both process knowledge and equipment operational knowledge must be communicated effectively in order to achieve
technical success. Development organizations may stress the technical aspects of the operation and risk overlooking other factors that are
equally critical to success. Efficient implementation of CM currently requires around-the-clock activity from plant engineers and operators
in monitoring online measurements, sampling, and replacing feed and product solutions. Close attention must be paid to how continuous
projects are staffed, and PAT and automation should be utilized whenever advantageous. In this process, repeated tasks such as filtration
operations, feed can recharges, routine monitoring of process conditions, and decisions to divert or forward process material based on
online HPLC results all presented opportunities for automation. If engaged, this level of automation would have significantly lessened
operational resources, lessened risk for inadvertent operational errors, and increased the likelihood of identifying process failures before
they became a risk to quality or safety.
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
At present, both the development and operation of CM processes generally require more resources and more strategic planning than
their batch analogs. In addition to the aforementioned staffing demands, the general unfamiliarity of manufacturing organizations with the
continuous processing equipment (compared to the knowledge of batch vessels) adds complexity to the campaign preparation and
operation. Furthermore, the simultaneous operation of multiple synthetic steps and unit operations requires that the development
organization be prepared with a more holistic understanding of the process control strategy, and likewise that the manufacturing
organization be prepared to identify and correct for disturbances in less time than would usually be afforded in batch processing. This
implies that the analytical organization at the CMO needs to also accommodate around-the-clock activity, and that more opportunities for
PAT and/or parametric control be leveraged to lessen the burden on analytical support.
ACS Paragon Plus Environment
28

Page 29 of 35
Organic Process Research & Development
1
2
3
4
5
6
7
8
9
This CM process achieved significant technical success and milestones for Lilly and STA. This campaign joins the work previously
1
reported by Lilly as one of the few multi-step continuous cGMP processes in the literature and is among the first of its kind to be
performed at a CMO. This process ran longer and afforded more DS than the previous Lilly process. Experience gained in the non-GMP
process demonstration was critical to the identification of equipment, process design, and process control elements that were taken
forward to the cGMP campaign. The Lilly team is grateful to the persistence shown by the STA team over the course of process
installation, qualification, and operation, and for the transparency granted by STA to Lilly team members throughout the campaign.
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
CONCLUSIONS
This report has documented the externalization of a hybrid batch/CM process to afford 183 kg of in-specification merestinib DS under
cGMP manufacturing. A control strategy for the cGMP process was developed, based first on learnings and data collected from a 20 kg
CM demonstration campaign and then on additional development research performed in both batch and continuous modes. While the
control strategy examined all quality attributes, this report has focused especially on the control of process impurities through collective
efforts in reaction analysis and modeling, crystallization rejection studies, and engineering of the CM unit operations. PAT, offline
analytical, and batch record strategies were developed and implemented in support of the control strategy during continuous operation.
Technology transfer of the hybrid process and control strategy from the development organization to the CMO occurred over six months.
This timeline was accelerated by knowledge transfer on CM that had occurred during collaboration between the two organizations several
years prior to engagement on this campaign. Perspectives offered by both organizations have highlighted the importance of staffing
decisions, PAT, and automation in making multi-step DS CM sustainable in the future. Overall, the success of this campaign can be
attributed to the robustness of the control strategy and to collaborative efforts shared between both organizations.
ASSOCIATED CONTENT
ACS Paragon Plus Environment
29