(T1430-11-61) Predictive Residence Time Distribution Models for Continuous Powder Blending

Purpose: Continuous manufacturing is an emerging technology with potential to increase the efficiency, flexibility, agility, and robustness of pharmaceutical manufacturing. Regulatory authorities have taken multiple steps to support the implementation of continuous manufacturing [1], including the finalization of the guidance on Continuous Manufacturing of Drug Substances and Drug Products from the International Council for Harmonization (ICH) last year [2]. Knowledge of process dynamics is essential for identification and mitigation of risks to product quality for continuous processes. In this regard, characterizing the residence time distribution (RTD) of a flow process offers powerful insight into the process. The RTD can be used to trace the flow and dispersion of material through a continuous process. Development of a traceability algorithm for the RTD, also known as an RTD model, enables prediction of how disturbances propagate through the process, allowing for isolation/diversion of out-of-specification material. Given that RTD models used for feed-forward process control may be considered medium-impact, properly accounting for the effects of material properties and process parameters in such models is critical to ensure product quality in continuous manufacturing processes.

Methods: Micronized acetaminophen was used as the model active pharmaceutical ingredient (API), and the bulk material was a mixture of lactose monohydrate and microcrystalline cellulose. Three loss-in-weight feeders fed API and excipients into a continuous tubular blender containing a configurable shaft. A near infrared (NIR) spectrometer was positioned below the blender outlet to monitor the composition of the blend in real time. For this purpose, off-line NIR calibration models were established using the same powder chute/NIR probe assembly as the in-line measurements. NIR spectra were collected roughly once every second during blending runs and API content of the exiting blends was estimated based on the appropriate NIR calibration model for each blend formulation. To verify the accuracy of the NIR models, blend samples were taken before and after all runs, as well as once every 15 seconds for select runs, and their compositions were confirmed off-line via high-performance liquid chromatography.

Results: Predictive models for the RTD of an API in a commercial scale continuous powder blender were developed at two different API loadings (5% and 30% w/w), incorporating the effects of total flow rate, blender speed, and bulk material properties. The predictive RTD models developed for each API loading were then validated against independent blending runs to assess model performance (Figure 1). Although reasonable predictability was observed for both API loadings, the 5% API model exhibited better predictive performance. Additionally, a series of challenge runs involving different particle sizes of API were performed to test the robustness and accuracy of the RTD model in different particle size ranges (d₅₀s of 9.0 and 130 µm). The RTD model exhibited better predictive performance for blends made using smaller API particle size than blends made using larger API particle size.

Conclusion: Understanding how bulk material properties and process parameters affect the RTD of a continuous process is critical for process understanding, enabling better informed process monitoring and control techniques. The level of understanding demonstrated when employing RTD models should be commensurate with how the RTD is used as part of the control strategy. Additionally, RTD model assumptions should be justified with appropriate considerations for the limitations of the model.

References: 1. Lee, S.L., et al., Modernizing Pharmaceutical Manufacturing: from Batch to Continuous Production. Journal of Pharmaceutical Innovation, 2015. 10(3): p. 191-199.
2. ICH, Continuous Manufacturing of Drug Substances and Drug Products Q13. 2022, International Conference on Harmonization of Technical Requirements for Registration of Pharmaceuticals for Human Use.

Acknowledgements: The authors would like to acknowledge Jessica Zombek from GEA for her input on and execution of the commercial scale blending runs, as well as Andrew Birkmire from GEA for helping to coordinate the collaboration.
Disclaimer: This abstract reflects the views of the authors and should not be construed to represent FDA’s views or policies.


Figure 1. Workflow diagram describing the development and validation of RTD models for API in commercial scale continuous powder blending.