398
B. Igne and E. W. Ciurczak
design a process that maximizes the yield while minimizing impurities and ensuring
the targeted bioavailability for the API.
Process analytical technologies have been used in the development of these routes
to ensure that the chemists and engineers can identify and optimize critical parameters
to develop robust processes [3, 34, 35]. In this section, examples of the use of NIR
spectroscopy for reaction and purification monitoring and control are presented.
18.4.1.1 Reaction Monitoring and Control
Gaining understanding of the reaction under consideration is the primary value
proposition that NIRS provides to the development of drug substance manufacturing.
Chemists and engineers want to develop the most robust processes and will often
sample the reaction for analysis by HPLC or NMR, which can take several hours and
potentially be affected by sampling and quenching. Having the ability to track the
reactions in-line and in real-time can provide additional insights into the state of the
reaction in the vessel, its dynamics, and end-point. This is particularly relevant for
the monitoring of intermediates and final product in the early phases of development
or when sampling is difficult (such as for hydrogenation reactions performed at high
temperature and pressure). When relevant, NIRS may be used for reaction control,
making decisions on the process to ensure the quality of the product for the patient.
An example of reaction monitoring was published by Blanco et al. [36] Authors
followed the esterification of myristic acid by isopropanol using multivariate curve
resolution. Figure 18.2 shows the relative trends of the component parameters. In
simple reactions like this, PCA may also be used but when intermediates are formed,
the variance described by the principal components may be distorted by the appearance and disappearance of species. This could affect the score trends and impair the
ability of PCA to adequately track the reaction. Multivariate curve resolution, with
its constraints for specificity can allow a better understanding of the reaction [37].
An example of the deployment of NIRS to reduce the safety risks to operators
was published by Wiss et al. [38]. The authors used NIRS with a transmittance
immersion probe to monitor and control a highly exothermic reaction during the
formation of a Grignard reagent. After building calibration models, the system was
used to quantitatively track the reagents as a function of reaction time. Figure 18.3
shows the evolution of the reaction components. The authors subsequently used
NIRS to automatically control the feed rate of a reagent and limit the safety risks
associated with the highly reactive process.
A very similar example was described for the monitoring and control of a distillation process [4]. The authors built quantitative models for the API concentration
and the solvent composition and fed-back that information to the distillation system
controlling the temperature and reflux ratio of the column when product of a variable
extraction was continuously fed for distillation. The results showed far better control
of the process when the distillation was optimized to account for the variability of
the incoming material. Figure 18.4 shows the process variability of API %w/w, with
B. Igne and E. W. Ciurczak
design a process that maximizes the yield while minimizing impurities and ensuring
the targeted bioavailability for the API.
Process analytical technologies have been used in the development of these routes
to ensure that the chemists and engineers can identify and optimize critical parameters
to develop robust processes [3, 34, 35]. In this section, examples of the use of NIR
spectroscopy for reaction and purification monitoring and control are presented.
18.4.1.1 Reaction Monitoring and Control
Gaining understanding of the reaction under consideration is the primary value
proposition that NIRS provides to the development of drug substance manufacturing.
Chemists and engineers want to develop the most robust processes and will often
sample the reaction for analysis by HPLC or NMR, which can take several hours and
potentially be affected by sampling and quenching. Having the ability to track the
reactions in-line and in real-time can provide additional insights into the state of the
reaction in the vessel, its dynamics, and end-point. This is particularly relevant for
the monitoring of intermediates and final product in the early phases of development
or when sampling is difficult (such as for hydrogenation reactions performed at high
temperature and pressure). When relevant, NIRS may be used for reaction control,
making decisions on the process to ensure the quality of the product for the patient.
An example of reaction monitoring was published by Blanco et al. [36] Authors
followed the esterification of myristic acid by isopropanol using multivariate curve
resolution. Figure 18.2 shows the relative trends of the component parameters. In
simple reactions like this, PCA may also be used but when intermediates are formed,
the variance described by the principal components may be distorted by the appearance and disappearance of species. This could affect the score trends and impair the
ability of PCA to adequately track the reaction. Multivariate curve resolution, with
its constraints for specificity can allow a better understanding of the reaction [37].
An example of the deployment of NIRS to reduce the safety risks to operators
was published by Wiss et al. [38]. The authors used NIRS with a transmittance
immersion probe to monitor and control a highly exothermic reaction during the
formation of a Grignard reagent. After building calibration models, the system was
used to quantitatively track the reagents as a function of reaction time. Figure 18.3
shows the evolution of the reaction components. The authors subsequently used
NIRS to automatically control the feed rate of a reagent and limit the safety risks
associated with the highly reactive process.
A very similar example was described for the monitoring and control of a distillation process [4]. The authors built quantitative models for the API concentration
and the solvent composition and fed-back that information to the distillation system
controlling the temperature and reflux ratio of the column when product of a variable
extraction was continuously fed for distillation. The results showed far better control
of the process when the distillation was optimized to account for the variability of
the incoming material. Figure 18.4 shows the process variability of API %w/w, with