(762e) Delivering a Design Space for a Continuous Drug Substance Manufacturing Process Using Simulation

On behalf of: M. Berry, P. Shapland, H. Robinson, P. Clements, M Hughes, S. Ukuser, P. Hamilton, I. Areri, C. Clarke, L. Wong, G. Breen, A. Ochen, M. Andersson and A. Richards.

A Design Space for a telescoped, continuous process for the manufacture of a GSK drug substance has been established using simulation. The continuous process involves five chemical transformations and 13 associated process parameters. In order to define proposed ranges for these parameters, enabling flexible operation during commercial manufacturing, a development strategy based on reaction kinetics and simulation was applied. In a departure from the sequential DOE and empirical modeling approach typically used, the mechanistic approach described provides a new development paradigm that is particularly appropriate for a highly telescoped continuous manufacturing process. As a result, the number of experiments required to assess the impact of operation within the design space for specific impurities was reduced significantly.

Kinetic parameters were determined from detailed laboratory batch experiments, along with evidence for the proposed reaction mechanisms for both the intended chemistry and undesired impurity formation. The resulting kinetic parameters and mechanistic description of the chemistry were then used in process models that also simulate the behaviour of the flow reactors used. This was supported by experimental measurements of the residence time distribution of the equipment train. Simulation was then used to identify the combinations of parameter settings that are most forcing with respect to specific impurities. These conditions were subsequently confirmed as being controlling of quality using a laboratory scale mimic of the intended commercial scale manufacturing equipment.

The model correctly predicted the most forcing conditions in the majority of cases. In addition, it was able to correctly identify those impurities at greatest risk of failing to meet the proposed drug substance specification, allowing process parameter ranges to be adjusted to mitigate this risk. For example, temperature setpoints for specific reactors were increased and successfully implemented in the laboratory as a direct result of simulation using a multivariate optimisation. Following an appropriate level of verification, the model can be used as an input to future technical risk assessments, for example prior to operating at manufacturing scale. This case study highlights the benefits of a mechanistic modeling approach but some limitations and future opportunities are also identified.