(39c) Model-Based Design of Experiments (MBDoE) for Drug Product Development

Design of experiments (DoE) [1] is an important statistical tool that is used extensively in pharmaceutical drug development for the identification of safe operation range, critical process parameters (CPPs) and their effects on critical quality attributes (CQAs), as part of the quality by design (QbD) approach. By manipulating multiple operating conditions simultaneously – all possible combinations (full factorial design) or a portion of the possible combinations (fractional factorial design) – statistical DoE helps identify important interactions for model development and, subsequently, parameter estimate improvement. However, these DoE approaches mainly rely on polynomial model structures postulated to relate operating conditions to process outcomes for experimental design instead of leveraging pre-existing process knowledge, resulting in additional, often expensive, experimentation aimed at re-learning already known process physics. Model-based design of experiments (MBDoE) [2] is a branch of DoE in which the existing physical understanding of a process is leveraged to identify experimental conditions that help gather broader process understanding for further model refinement or parameter precision improvement through potentially lesser and intentional experimental effort.

This work presents an MBDoE workflow that leverages existing process knowledge and extensive experimental information from past campaigns to guide experimental design for the development of a new medicine. Specifically, the workflow assumes a priori knowledge about material and process to product relationship (an existing model) and identifies experimental design candidates that, given an experimental budget, are most important for refining the process parameters to accurately explain current system behavior [3, 4]; thus, preventing the design of expansive experimental campaigns such as full factorial design that may not significantly improve upon existing process understanding. The effectiveness of the workflow is demonstrated through optimal MBDoE case studies focused on parameter precision improvement for drug product unit operations such as roller compaction, tablet compression, and film coating. Additionally, this work demonstrates how the workflow can be used to identify the most important sampling times in dynamic processes to reduce over-sampling without losing information gained; a residence time distribution (RTD) identification exercise is used for illustration. Finally, the approach is augmented to deal with a hybrid process model [5] that combines a physical model with a partial least squares (PLS) model.

References

[1] Fisher, R. A., Genetiker, S., & Généticien, S. (1966). The design of experiments (Vol. 21). Edinburgh: Oliver and Boyd.

[2] Franceschini, G., & Macchietto, S. (2008). Model-based design of experiments for parameter precision: State of the art. Chemical Engineering Science, 63(19), 4846-4872.

[3] Kusumo, K. P., Kuriyan, K., García-Muñoz, S., Shah, N., & Chachuat, B. (2021). Continuous-Effort Approach to Model-Based Experimental Designs. In Computer Aided Chemical Engineering (Vol. 50, pp. 867-873). Elsevier.

[4] Kusumo, K. P., Kuriyan, K., Vaidyaraman, S., García-Muñoz, S., Shah, N., & Chachuat, B. (2022). Risk mitigation in model-based experiment design: a continuous-effort approach to optimal campaigns. Computers & Chemical Engineering, 159, 107680.

[5] Sen, M., & García Muñoz, S. (2021). Development and implementation of a hybrid scale up model for a batch high shear wet granulation operation. AIChE Journal, 67(5), e17183.