Mechanistic models, or science-based digital twins, use process understanding to describe and predict the quantitative behavior of pharmaceutical processes for risk assessment, control strategy design and evaluation, scale up/tech transfer, and design space exploration. In contrast to traditional DoE-based approaches, mechanistic models offer a structured, science-based manner to determine the truly critical parameters, rather than relying on heuristics-based reasoning and sparse DoEs. They also allow a significant reduction in data requirements for model calibration thereby making the approach suitable for routine application.

While mechanistic model-based approaches are becoming more common in pharmaceutical applications, the required and actual validity of these models is not always clear, which can result in a reluctance to apply these tools in a regulatory environment. However, techniques are well-established in the process systems engineering field to quantify uncertainty in a calibrated mechanistic model and assess how that uncertainty propagates to the quality attribute and process performance predictions made by the model.

In this webinar, we will compare and contrast mechanistic model-based approaches to QbD with traditional DoE-based approaches. The concept of model quality will be presented, and the verification and validation framework will be presented, alongside techniques to quantify model uncertainty. A standard workflow will be introduced to validate a mechanistic model (calibration and blind testing), characterize parametric uncertainty, and analyze the effects of that uncertainty in the resulting model predictions. A simple continuous direct compression process will be used to illustrate these concepts, where feeder and blender models are calibrated and the impact of the uncertainty in those parameters is explored.

Using a qualified mechanistic model, virtual DoEs can be performed in silico as a model-based risk assessment.

Finally, we will touch upon further improvements to qualifying model validity through the use of Bayesian approaches to obtain more accurate confidence regions for estimated model parameters.

What this webinar covers

• Mechanistic-model based approaches to Quality-by-Design
• Importance of model and data quality
• Standard workflow for model qualification, illustrated using a continuous direct compression example
• Virtual design space exploration
• Hybrid mechanistic and data-driven modelling
• Digitalization across the R&D and operations lifecycle

Presenter(s)

Dana Barrasso, Process Systems Enterprise

Dana Barrasso is a Principal Consultant in the Formulated Products business unit at Process Systems Enterprise. As PSE’s Strategy Director for Pharmaceuticals, she is the technical lead for collaborative and strategic activities with industry, academia, and regulatory bodies in the synthetic pharmaceuticals sector. Her responsibilities include acting as the technical director for the Systems-based Pharmaceutics Alliance, a collaborative project between Eli Lilly, GSK, Pfizer, Roche, Sanofi, Bayer and PSE aimed at developing the tools and workflows that enable systems-based approaches to drug product and process design. With a PhD from Rutgers University, Dana’s background is in solids process modeling, with a focus on wet granulation processes. With this experience, she leads the development and application of PSE’s offerings in drug product manufacture, including continuous direct compression, wet and dry granulation, and tableting.

Sean Bermingham, Process Systems Enterprise

Sean Bermingham joined PSE in 2000 and was PSE’s Global Head of Consulting for 8 years before forming the Formulated Products business in 2010. As head of this division, he is responsible for strategic business development, software development and services delivery for the pharma, food, consumer good, specialty & agrochemical industries. Sean is a leading figure in the development and adoption of mechanistic model-based tools for increasing R&D efficiency and managing risk in areas ranging from clinical trials to tech transfer and operations. He is the driver behind QbD 2.0, which focuses on improving the efficiency of QbD by reintroducing the “sound science” element of the original QbD definition. He is one of the founders of and the lead for the Systems-based Pharmaceutics Alliance (2013-present) with Bayer, Eli Lilly, GSK, Pfizer, Roche and Sanofi. Sean was also the lead for the £20.4m ADDoPT “Digital Design” Project (2015-2019) with partners including AZ, BMS, GSK & Pfizer. Sean is a Chemical Engineer with an MSc and a PhD from Delft University of Technology and an Executive MBA from Imperial College London.