(445c) Multivariate Analysis of Historical Process Data for Monitoring, Control and Scale-up

A quality by design (QbD) initiative requires a good foundation of organized data. Without this foundation, there could be no developments in use of prior knowledge or implementation of quality by design approach or in realizing full potential of Process Analytical Technology (PAT) work. Over viewing an historical process data presents many challenges including that of preprocessing of data, large amount of missing data, and large number of collinear measurements etc. Use of multivariate data analysis such as hierarchical PLS provides a useful tool for process understanding and continuous process improvement, and overcomes challenges faced in analyzing process data. Organization of data into hierarchical fashion not only facilitates data management, but also provides convenience in analyzing data. Present work gives examples from formulation, granulation, tablet press and analytical results data. Current work demonstrate use of multivariate methods, at the lower base level, to analysis local models for comparison of formulation, equipment setup and process parameters that provides a more detailed understanding at local level of the unit operations i.e. granulation and tablet press. Present work also demonstrate at the top level, how to relate quality of product with formulation, and process parameters and provide an overview of the entire process, and reflect the most dominant patterns and trends. Application of local models for investigating scale-up of unit operation, and use of global models for entire process transfer is discussed.