Targeting fidelity of pharmaceutical systems models by optimization of precision on parameter estimates

Abstract

Quantitative models have gained momentum to drive the development of pharmaceutical processes. The assessment of the prediction fidelity of these models is key to provide interpretability of process phenomena and to enable decision-making. Evaluating parametric uncertainty is paramount when the focus is on systems models, which combine different sub-models together, and, thus, parameters related to previous units may strongly impact the prediction of one final output. A framework is proposed to assess reliability in model predictions, where the precision of parameter estimates is explicitly optimized to target pre-set tolerance requirements on process key performance indicators and product critical quality attributes. A direct compression systems model for the manufacturing of oral solid dosage products is used as a case study. Results show that the proposed methodology is effective at guaranteeing the target model fidelity and at quantifying the maximum acceptable uncertainty in the estimates of model parameters.