The quality-by-design (QbD) initiative of the regulatory authorities grants biopharmaceutical manufacturers a higher flexibility once a design space has been defined based on an extensive process understanding. In addition, this process understanding opens up the path to a much more rational way of process development and production facility operation. Two possibilities to create process understanding are high-resolution screening experiments based on Design-of-Experiments (DoE) and mechanistic modeling. The first requires a high sample amount, relies on the user’s expertise in preselecting relevant parameters, and is finally restricted to the calibrated experimental range. Mechanistic modeling relies on a much smaller set of experimental data and on sophisticated mathematical methods, allowing to extrapolate to processes outside the calibrated range, and to mathematically describe the relation of critical process parameters (CPP) and critical quality attributes (CQA) This contribution discusses the challenges, pitfalls, and best practices in using mechanistic modeling to fulfil the latest QbD requirements. Conventional experimental and DoE-based QbD studies start with the preselection of the most relevant process parameters affecting process performance and product quality. In contrast, the model-based QbD presented in this work, does not need a preselection of parameters, eliminating a time consuming and uncertain decision process from the workflow. Next, a mechanistic chromatography model is calibrated based on 3-4 experiments, and the quality of the model parameters is assessed by means of parameter confidence intervals and parameter correlations. Subsequently, the model parameters uncertainties and the typically observed variabilities of the process parameters are used to simulate a virtual sensitivity analysis. The probability of occurrence and the impact of an out of specification run resulting from this analysis are used to compute a CQA impact score and to cluster the process parameters into non-relevant, key, and critical parameters. Finally, a fine Latin Hypercube sampling is carried out for the key and critical parameters with respect to the key performance indicators and quality attributes. This high dimensional data set is used to define a process design space. This talk will present the technological workflow of model-based QbD by means of an industrial study of purifying an antibody from a mixture containing multiple product related impurities.

Model-based quality by design: Challenges, pitfalls and best practices

255th National Meeting (2018)

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.