Empirical Modelling to support Scale up of Primary Pharmaceutical Processes

Abstract

Abstract Historically, design spaces registered for a pharmaceutical process in a new drug application or similar document were derived from lab scale experimentation. Recently, regulatory agencies such as the Food and Drug Administration (FDA) and the European Medicine Agency (EMA) started questioning the validity of these design spaces at commercial scale. More specifically, can it be ensured that stretching a critical process parameter has the same effect over different scales? We propose a strategy deploying multivariate analysis to enhance understanding of scale dependent process effects in primary processes complementary to mechanistic understanding in order to provide general support to scale up activities. In the future, this analysis will be based on data generated from a design space verification lab, which provides reactor vessels of two significantly different scales (1 and 16 L) with minimal differences in vessel geometry, soft sensors and spectroscopic data objects. Maximizing process comparability across scales should allow a rational evaluation and management of process variability risks associated with scale up. Ultimately, the proposed strategy should enable GSK to decide whether (1) a design space is scale independent or (2) scaling factors can be identified such that the design space is valid over intended scales or (3) scale effects are present and should be incorporated in the overall design space. This manuscript firstly describes a preliminary study to demonstrate the added value of empirical modelling when analyzing process data over different scales and is based on historical data not generated within the dedicated design space verification lab. This comprehensive analysis included typically critical process parameters, process parameters with small natural variability and scale dependent parameters. Secondly, for situations where scale effects are significant and must be incorporated to ensure validity of the design space at commercial scale, a scale up model was developed. For our case study, the proposed design space based on experiments at both scales was smaller than the design space based on the lab scale data only (Figure 5).