Despite the promise of continuous processing taking over more of drug manufacturing, some obstacles must be overcome. Ash Kumar, a pharmaceutical engineer at Ghent University in Belgium and his colleagues recently asked: Why is batch processing still dominating the biologics landscape?
The opinion of these researchers is that wider adoption will depend on a list of technical, management, and regulatory improvements, including the “application of detailed modeling and expert systems to support the development and regulatory requirements.”
Other experts agree. For example, Nima Yazdanpanah, principal at Procegence, a modeling and simulation service provider, calls modeling and simulation “crucial” in the transition to continuous processing. In particular, he points out that modeling will be needed to design new equipment and processes that can run in a continuous-manufacturing mode. Plus, he mentions the value of modeling in process integration, system dynamic and control strategies development, and a list of other areas.
The most needed kinds of modeling in the transition to continuous processing, says Yazdanpanah, are “varied—depending on objectives and questions of interest.” For example, he notes that mechanistic modeling and simulation, or hybrid methods, are crucial in developing new continuous-manufacturing sets of equipment that should be able to work in a system seamlessly, such as perfusion bioreactors, simulated moving bed chromatographic columns, or continuous viral inactivation.
The most noteworthy applications, according to Yazdanpanah, include residence time distribution modeling and system dynamics in response to a set of disturbances, advanced process control, and dynamic correlation between critical material attributes, critical process parameters, and critical quality attributes. In addition, modeling and simulation applications support quality-by-design in process development. Here, the design space and control spaces are developed by knowledge-based approaches to ensure consistent quality of manufactured products by a robust process.
Ultimately, scientists and engineers must decide how and where to put models to work. “There is no generalized modeling recipe to be developed and implemented for all applications, scales or modalities,” Yazdanpanah says. He adds that the mechanistic models should follow the SMART rule, which stands for specific, measurable, achievable, realistic, and trackable/testable.
“Some models are over complicated, expensive, not accurate, or over simplified—that cannot be useful for real industrial applications,” Yazdanpanah says. “For some cases, you should not start any activity before running an in-silico model first, but don’t waste your time or money on simulating everything.”
