Mechanistic modelling in pharmaceutical product and process development: A review of distributed and discrete approaches

Pharmaceutical product and process development is transitioning from traditional heuristics-based approaches to a Quality-by-Design (QbD) methodology, emphasising systematic process design and understanding of critical parameters. While Design of Experiments (DoE) is key for identifying critical process parameters, it has limitations in scalability and potential over-fitting. Detailed mechanistic or first-principles modelling, using distributed or discrete approaches, offers a promising tool for understanding complex, heterogeneous systems. This paper reviews the roles, opportunities, and challenges of detailed mechanistic modelling in pharmaceutical product and process development. The role of mechanistic models is first discussed from strategic, business, and regulatory perspectives. The workflow of mechanistic modelling is then described, consisting of model selection, calibration, validation, and maintenance. Case studies of key unit operation developments, such as wet granulation and fluidised bed system, are reviewed, highlighting process characteristics, model requirements, and application challenges. Proper model development and experimental design are essential to avoid pitfalls, such as limited applicability or excessive data requirements. Despite rising interest in machine-learning approaches, mechanistic modelling aligns well with data-driven methods, offering high-resolution process understanding and enabling optimal development with fewer experiments. This approach surpasses conventional trial-and-error methods, providing deeper insights and innovative solutions for pharmaceutical processes.