This item was submitted to Loughborough's Research Repository by the author.

Items in Figshare are protected by copyright, with all rights reserved, unless otherwise indicated.

Digital design of end-to-end manufacturing process for mefenamic acid using

mechanistic modelling
PLEASE CITE THE PUBLISHED VERSION

https://dechema.de/en/ISIC_2021.html

VERSION

AM (Accepted Manuscript)

LICENCE

CC BY-NC-ND 4.0

REPOSITORY RECORD

Brown, Cameron J., Niall A. Mitchell, Frederik Doerr, John McGinty, Magdalene W. S. Chong, Murray
Robertson, Sara Ottoboni, et al.. 2022. “Digital Design of End-to-end Manufacturing Process for Mefenamic
Acid Using Mechanistic Modelling”. Loughborough University. https://hdl.handle.net/2134/19330640.v1.
Digital design of end-to-end manufacturing process for mefenamic
acid using mechanistic modelling
Cameron J. Brown, University of Strathclyde / EPSRC Future CMAC Research Hub,
Glasgow, UK; Niall A. Mitchell, Process Systems Enterprise Ltd, London, UK;
Frederik Doerr, University of Strathclyde / EPSRC Future CMAC Research Hub,
Glasgow, UK; John McGinty, University of Strathclyde / EPSRC Future CMAC
Research Hub, Glasgow, UK; Magdalene W. S. Chong, University of Strathclyde /
EPSRC Future CMAC Research Hub, Glasgow, UK; Murray Robertson, University of
Strathclyde / EPSRC Future CMAC Research Hub, Glasgow, UK; Sara Ottoboni,
University of Strathclyde / EPSRC Future CMAC Research Hub, Glasgow, UK; Wei
Li, Loughborough University / EPSRC Future CMAC Research Hub, Loughborough,
UK; Jan Sefcik, University of Strathclyde / EPSRC Future CMAC Research Hub,
Glasgow, UK; Alison Nordon University of Strathclyde / EPSRC Future CMAC
Research Hub, Glasgow, UK; Blair Johnston University of Strathclyde / EPSRC
Future CMAC Research Hub, Glasgow, UK; Chris J. Price University of Strathclyde /
EPSRC Future CMAC Research Hub, Glasgow, UK; Chris D. Rielly, Loughborough
University / EPSRC Future CMAC Research Hub, Loughborough, UK; Brahim
Benyahia, Loughborough University / EPSRC Future CMAC Research Hub,
Loughborough, UK; Alastair J. Florence, University of Strathclyde / EPSRC Future
CMAC Research Hub, Glasgow, UK

Digital twins built using mechanistic models are playing an increasingly significant role
in helping pharmaceutical industries develop robust and more economically efficient
manufacturing processes. Building a digital twin for an end-to-end drug manufacturing
process allows exploration of the individual and combined effects of numerous process
parameters during the active ingredient and drug product manufacturing stages on the
final drug product performance. This allows for the development of a more robust drug
manufacturing process and greater assurance of product quality while reducing
process development timelines and resources.
This presentation will outline the quantitative analysis performed on the manufacturing
process for mefenamic acid by building the end-to-end flowsheet model in gPROMS
FormulatedProducts, shown in Figure 1.
 Figure 1. gPROMS FormulatedProducts flowsheet model

Each element of the drug substance production step, including synthesis,

crystallization and isolation (filtration and washing) are validated individually using
process data:
1) First crystallisation process is telescoped directly from the output of the final
synthesis step. A combination of cooling and anti-solvent addition is used to generate
particles of appropriate purity and size. Kinetic parameter estimations were made
considering the impact of impurities on the various rate processes and
thermodynamics.
2) The first isolation process consists of filtration, a first wash with a mixture of
crystallization and wash solvent followed by two washes with pure wash solvent.
Parameter estimations are made to build a correlation between porosity and driving
force, and to evaluate the effect of heel formation on filtration time. Using the
experimental data generated during a DoE the validation of the isolation model is
achieved. Using CMAs established during crystallization, isolation process is
simulated and parameters are selected to be transferred to the semi-continuous
platform (CFD25). Global sensitivity analysis is used to determine how the variance of
the model output depends on the uncertainty of the input factors (e.g. PSD, solid
content, diffusion coefficient, porosity, particle sphericity, and mother liquor and input
stream composition).
3) The predicted CFD25 filtration and washing CQAs gives the final product
composition after the last wash step.This simulated washed cake composition is used
then to simulate the dissolution process solvent swapping into the second
crystallisation.
4) A Population Balance Model (PBM) is applied considering growth and dissolution,
as well as primary and secondary nucleation kinetics for the secondary crystallization
process. The kinetic parameters were fitted through well-designed batch cooling
crystallization experiments using seed operation. The mathematical model was
validated against data from batch cooling and direct nucleation control (DNC)
experiments, then used to design and optimize a multistage mixed-suspension mixed-
product removal (MSMPR) crystallizers. Sensitivity and uncertainty analysis were
performed to identify the Critical Process Parameters (CPP) which exhibit the highest
impact on the targeted Critical Quality Attributes (CQA), such as particle size and yield.
5) Final isolation process consists of filtration and two washes with pure wash solvent.
Parameter estimations are made to build a correlation between porosity and driving
force. The CMAs of the second crystallisation are used to (including the two sets of
PSD obtained from the second crystallisation simulation) simulate isolation process
and to identify the optimal conditions for a complete removal of the residual mother
liquor and impurities. Global sensitivity analysis and local sensitivity analysis were
performend to identify the CPP to use in the CFD25 platform, during the continuous
manufacturing.

Once each unit operation step in the production step is validated, a systems model
was configured to provide a quantitative representation of the end-to-end production
process including drug substance, drug product and product performance elements.
This end-to-end model was subsequently utilized to develop a quantitative
understanding of the effect of process disturbances, raw material variability, process
parameters, formulation parameters and model uncertainty on CQAs (e.g. tablet
properties) and manufacturability. Along with the analysis, the relative influence of
those factors (sensitivity indices) on the desired CQAs will also be presented.
Furthermore, the impact of batch and continuous operational elements of the process
will also be probed. This work aims at addressing existing gaps in the scientific
understanding of the effects of some of the API attributes on the desired KPIs during
this end-to-end flowsheet analysis.