Introduction
The challenges the pharmaceutical industry is facing are manifold. They include high research and development costs, long development timelines, complex regulatory requirements, supply chain vulnerabilities, an enormous market competition and an increasing scrutiny on drug prices. Leveraging digital strategies such as artificial intelligence, machine learning, data analytics and blockchain technology holds significant promise in addressing many of
these difficulties.
This article focuses on the challenges of the pharmaceutical industry that are closely linked to the production process, covering engineering, scale-up, technology transfer and operations. Digital, modeling- and simulation-based approaches are to provide answers to prominent hurdles in the following areas:
Engineering and equipment design: Selecting and designing equipment that meets the specific requirements of the process and the regulatory standards of the pharma industry is complex. It goes hand in hand with optimizing the layout of the facility, to provide a safe production environment and enable an efficient workflow. The interdependencies of the core process and the utilities are complex, options for future expansions must also be taken into consideration.
Effective technology transfer: Bringing biotechnological production from the laboratory to an industrial scale is an extremely challenging task, as upscaling the process to production scale involves numerous technological challenges and substantial commercial risks. Technology transfer requires a planned approach with the appropriate documentation, data and information covering all aspects of development, production, and quality control, considering the regulatory requirements (1).
Process optimization and quality control: Due to the high complexity of pharmaceutical manufacturing, it is not a simple task to maintain optimal process efficiency and product quality. The availability of data plays a significant role in process optimization.
Operational efficiency and workforce training: The implementation of robust process control systems, the management of supply chains, and a continuous training of the workforce to keep up with new technologies and regulations are ongoing operational challenges.
The development of comprehensive digital solutions is crucial for boosting the entire value creation process for pharmaceuticals. In the context of a holistic view on the interrelations of product, production process, and plant, the application of model-based technologies can provide support in drug development, process scale-up and manufacturing. Agility is increased and time to market reduced. Nonetheless, for this to materialize, it is not only essential to have the right software solutions, but also to break down silos and bridge gaps across disciplines. Smartly applied digitalization is crucial for the entire value chain, requiring a holistic approach. In order to leverage the full potential of available digital technologies, such as simulation or data analytics, these technologies need to be combined in a smart way. The prerequisite: a harmonization of data that results from all aspects and phases of the life cycle of a drug.
Integrated engineering
The use of digital platforms in the pharmaceutical industry allows for the collection of vast amounts of data from various sources. During production, metrological instruments (e.g.: PAT technology) generate data on the plant and product’s condition, and GMP-required records are kept digitally in the eBR (electronic batch record) to ensure product quality. Operational data supplemented by engineering data (e.g., 3D data, extensive information from piping and instrumentation diagrams (P&IDs), component specifications, and electrical planning data) provide the basis for efficient digital applications or strategies for maintenance, energy efficiency, or production optimization.
However, the integration of data derived from process monitoring and engineering is only possible as the data is accessible and complete throughout all engineering disciplines and phases (concept, basic and detail engineering). An integrated software landscape allows to combine data input of all project partners in real time and deliver harmonized datasets. This leads to high transparency, efficiency, quality, and execution speed. Furthermore, the result of such a consistent data integration is a virtual representation that comprises the production plant. This includes data regarding building layout, the respective equipment, utilities, modular structures, piping, and technical building equipment, and the production process itself. With its individual steps, required equipment, critical process parameters (CPPs) and recipes to engineer control and simulation systems a physical and procedural model of the plant is developed. Naturally, such a digital twin is useful for various applications.
During large and complex pharma investment projects, many project partners and multiple disciplines such as process engineering, 3D design, electrical engineering, automation, and qualification are involved. Engineering, process scaling, and the application of the requirements of current good manufacturing practice (cGMP) are just a few of the many aspects that must be covered. By digital, integrated engineering, data silos are avoided and uniform, harmonized data are generated. State-of-the-art software technology (e.g., computer-aided design software for creating P&IDs, 3D models, or electrical wiring diagrams) is part of one common software landscape for all project partners. Specific product life cycle management solutions cover all engineering workflows and user-oriented frontends and enable data input, data management and data use.
By merging the real-time data input of all the involved disciplines, a harmonized data set representing the digital twin of the plant and process is generated. This goes hand-in-hand with 100 % transparency for all project partners at all times, which significantly reduces project risks.
The effects are particularly beneficial for change management. Changes of components effect all disciplines and require massive effort when they occur in later engineering stages. Integrated software landscapes allow an improved change management, as consequences of changes become transparent before the change is approved, and all project partners can follow up on the change in real time.
Finally, this harmonized data serves as the basis for applications in the field of augmented or virtual reality applications, at it contains all the information on 3D dimensions, the equipment (tag numbers, spare parts) and its location on the manufacturing site.
The role of production simulations
The integration of production simulation tools into the software environment allows for a comprehensive understanding of complex systems and enables the examination of different scenarios in a virtual environment, facilitating informed decision-making processes. The benefits for plant development, process planning, optimization, or production planning are significant.
During process planning, production simulations are a tool to clarify questions concerning equipment occupancy, setup times, shift schedules and media consumption over time, as well as to determine theoretical process capacities. In the context of plant development, simulation supports the dimensioning of process equipment and pipelines.
Simulation software can be used in all phases of the project – as a valuable engineering tool, as well as a comprehensive tool for production planning. A virtual image of the production processes – a digital twin – is created throughout the project from initial estimates to minute-by-minute resolution of production. It includes all relevant features and behaviors of the real plant and comprehensively covers its whole lifecycle – from feasibility study throughout all the engineering phases to operations. Thus, warehouse capacities, personnel requirements, and possible shift schedules are known long before commissioning. In addition, a virtual operation of production highlights any automation challenges at an early stage, paving the way for efficient production planning.
A promising future topic is predictive decision support for the operators, in which simulations run continuously in the background while a plant is in operation and decisions can be made based on predictive production data. A highly accurate simulation model is the basis for online production monitoring. All data from the IT/OT systems must be integrated so that the simulation software (in the present case: INOSIM Foresight) can access it: real-time production data, data from the respective batch systems and from the process control systems, or EMS data. This poses challenges in the field of automation technology that can only be overcome with cross-disciplinary expertise and a close interlocking and integration of process technology, engineering, digitalization, and automation.
Process development and scale-up
Of course, digital data recording and documentation is essential for product and process development, and modeling and computer simulation techniques are increasingly being used. Clinical development requires a scale-up for the GMP-compliant manufacturing of larger product quantities. The process, initially developed on lab scale, is mapped digitally and provides the basis for technology transfer.
In the course of product and process development, process conditions are specified using quality by design (QbD) principles, generating valuable information. The critical process parameters (CPPs) are defined, which have a decisive influence on the critical quality attributes (CQAs) of the product. This implies that it is reasonable to develop and optimize the product or process and at the same time derive valuable information for the engineering of the scale-up (pilot) plant and for the production scale plant for industrial manufacturing.
Online Optimization – Model predictive control
Model-based technologies are decisive for simulations and many optimization measures. On account of the QbD approach, DoE-based bioprocess models have been applied in process development for over a decade. In the manufacturing stage, however, process models are rarely applied, even though the exploit of the power of mathematical models was recommended by the FDA in the process analytical technology (PAT) guideline (2004) (2).
Following this guideline, future submissions may include new control strategies.
Model predictive control (MPC) is an up-and-coming method for ensuring precise control over critical process parameters in biopharmaceutical manufacturing systems (3).
The question is if the same process models that are generated for understanding and scaling of the process – the bioprocess models of the QbD/DoE approach – can also be used for operational support.
Using process models to improve operational efficiency
The process models can also be used in several ways to improve process efficiency. Typically, the operating strategy of a bioreactor is defined based on the execution of a design of experiments (DoE). From the obtained results, the process settings corresponding to the “best” outcome are determined and then used during production. The selected time-profiles of the process settings, such as substrate, or inductor feed, are prescribed for the entire duration of the batch production. A more detailed investigation of the process (e.g., variation of feed rates during batch execution to find potentially better process input profiles) is practically infeasible on the real equipment since it would be too time- and resource-demanding.
However, by integrating the available process models into a 2-step optimization strategy (see Figure 1), process efficiency can be improved considerably: First, an (offline) optimization problem is formulated and solved. The outcome of this optimization are the time-profiles of all the process inputs (i.e., temperature, substrate- and inductor feed) and process outputs (i.e., optimal biomass- and product concentration).
In contrast to the DoE, the shape of the input profiles is allowed to change during the batch execution. This approach offers the advantage that the objective function of the optimization problem can be systematically specified. For example, the amount of product at a pre-defined processing time should be maximized. The process model is one of the constraints of the optimization problem. Furthermore, the profiles of the process settings need to be constrained to the validity region of the process models. The offline optimization that maximizes the product at a pre-defined processing time can be formulated as
with u(t) denoting the time profiles of the process settings that result in the optimal process output y. The constraints umin , umax , ∆umin and ∆umax are defined by the operating space covered by the DoE, as well as process and equipment limitations.
Secondly, an online optimization based on model predictive control (MPC) is proposed. MPC is an advanced process control algorithm that employs a process model to predict and optimize the future behavior of the system (4). The aim of the control loop is to keep track of the reference trajectories created during the first step (see above). Process disturbances (e.g., heating system or pump failure) or model uncertainties can lead to deviations of the real process trajectories from the computed, optimal ones.
The MPC will adjust the process inputs in real-time in order to keep the process output y close to the pre-computed reference.
The application of the latter approach requires the real-time measurement of the product concentration. Therefore, process analytical technology (PAT) or soft sensors are needed to provide this information.
In case the installation of such equipment is not feasible, the first step, i.e., the offline optimization, can still be used to achieve significant improvements in process operation – at (almost) no cost, since the process model is often in place already. Furthermore, the available process model can be used as a basis for developing a soft sensor to provide the necessary real-time information for the MPC.
Conclusions
Addressing the complexity of modern biopharmaceutical production processes, it is necessary to break up silos between development and manufacturing. Using a comprehensive picture of the interdependencies of plant, process, and product and an early understanding of their interactions, improved engineering results and a whole range of further advantages are obtained. Model-based technologies allow the sharing of knowledge needed for scale-up, support the implementation of the QbD approach, and the application of advanced process control techniques.
Figure 1. Two-step optimization strategy for the systematic improvement of conventional bioreactor operating procedures.
References and notes
- Annex 4, WHO guidelines on technology transfer in pharmaceutical manufacturing https://cdn.who.int/media/docs/default-source/medicines/norms-and-standards/guidelines/production/trs1044-annex4-technology-transfer-in-pharmaceutical-manufacturing.pdf?sfvrsn=9a174f1a_1&download=true
- Guidance for Industry PAT — A Framework for Innovative Pharmaceutical Development, Manufacturing, and Quality Assurance; U.S. Department of Health and Human Services, Food and Drug Administration, September 2004. https://www.fda.gov/media/71012/download
- Eslami, T. & Jungbauer, A. (2024). Control strategy for biopharmaceutical production by model predictive control. Biotechnology Progress
- Rawlings, J. B., & Mayne, D. Q. (2009). Model Predictive Control: Theory and Design. Nob Hill Publishing.
