Elcin Icten Gencer¹, Erin Shen¹, David Perez-Aguilar¹, Duc Nguyen¹, Saloni Daftardar¹, Shruti Gour¹, Mina Hamedi¹, Killian Ryan¹, John Chung¹, Fernando Alvarez-Nunez¹, Fabrice Schlegel¹, Pablo Rolandi¹, Ananya Chowdhury².

¹ Process Development, Amgen Inc.

² Siemens-PSE

Dry granulation via roller compaction is a key unit operation in the manufacture of immediate release tablets. One of the challenges for late-stage development is identifying optimal roller compaction processing parameters for scale up through numerous Design of Experiment (DoE) studies, which require manufacturing time, drug substance and analysis. In this work, a first-principles in silico model available in gPROMS was extended to guide selection of key DoE studies and reduce the number of large-scale runs needed to identify optimal processing parameters. The prediction accuracy of this model was demonstrated for the scale up of the roller-compaction process at a commercial manufacturing site, with prediction errors within a few percent. This case study demonstrates the possibility of using this “VirtualPactor” model as a tool to reduce the number of experiments required to configure commercial roller-compaction process parameters. Instead of carrying out trial-and-error experiments to define the inputs for a large-scale roller compactor, product teams can carry out a DoE study on a small-scale roller compactor, use the data from the study to calibrate the VirtualPactor model, and use the calibrated model to find the inputs for the large-scale compactor that satisfy the output specifications. This model is developed into a web application for use by domain experts in an effort to democratize access to advanced modeling techniques and enable Quality by Design during process scale-up and tech transfer.