Assessing process systems models for pharmaceutical development

Quantitative models are useful tools to accelerate the development of pharmaceutical processes. The assessment of the predictive capability of such models is fundamental to enhance their usage in a systematic way, particularly when the focus is on flowsheet models. In this study, we propose a systematic procedure that combines techniques that are typically used in different modeling contexts – namely, global sensitivity analysis, model-based design of experiments, and data reduction by means of multivariate statistical methods, with the advantage of enhancing readability and interpretation. The methodology is effective for the assessment of model fidelity and can support practitioners in the development of pharmaceutical processes. A direct compression systems model for manufacturing oral solid dosage products is used as a case study. Results show that just a subset of model parameters require precise estimation to meet the target critical quality attribute, and that calibrating flowsheet models on a unit-by-unit basis may be unnecessary when the focus is on one final quality attribute.