(654b) In-Depth Understanding of the Impact of Material Properties on the Performance of Jet Milling of Active Pharmaceutical Ingredients

Milling of active pharmaceutical ingredients (APIs) is a critical unit operation to enhance bioavailability, especially of poorly soluble drug substances. Among possible options for milling, air jet milling has the advantages of self-classifying, non-contaminating, and non-degrading operation with a narrow particle size distribution [1]. Within the framework of Quality-by-Design, both experimental and modeling work have been performed to understand the impact of process settings, e.g., mass and gas flow rates, on the milling performance. So far, different types of mechanistic models have been proposed, e.g., population balance modeling (PBM) [1], computational fluid dynamics and discrete element methods (CFD-DEM) [2], and energy-based modeling [3]. While energy-based modeling is much easier and computationally less demanding than other mechanistic models, it cannot consider the impact of material properties on the performance [4]. An in-depth understanding of the impact of material properties on performance is quite relevant since possible experiment numbers using pure APIs are very limited in the design phase.

This study presents the analytical results of spiral jet milling performance varying APIs and process settings. Before the experiments, the material characterization of each API was performed in terms of mechanical properties, e.g., Young’s modulus and Poisson’s ratio, and energy parameters, e.g., elastic recovery and specific work of compaction. Four different APIs with multiple grades were milled by a Hosokawa Alpine spiral jet mill 50AS. For each API, full-factorial experimental designs were performed changing the mass feed rate and gas feed rate as the factors. Particle size distributions were then characterized by laser diffraction measurements, where the 10th (dv10), 50th (dv50), and 90th (dv90) percentile of the cumulative volume distribution were computed. As indicators of the milling performance, the ratios of dv10, dv50, and dv90 for unmilled APIs to those for milled APIs were calculated.

The experimental results showed that Young’s modulus influenced the milling performance. The ratios of dv10 for unmilled to milled APIs became larger, i.e., higher performance when Young’s modulus of APIs was higher. While the preferability of less elastic materials could explain this observation, it was also observed that there was collinearity between Young’s modulus and the initial particle size in the training data. Regarding the impact of process settings, multi-linear regression models showed that gas flow rate had a higher influence on the milling performance than mass feed rate. The effects of material properties and process settings could be generalized through advanced analysis, e.g., partial least squares (PLS) regression and the PBM calibration.

In conclusion, an in-depth understanding of spiral jet milling was possible through extensive experiments and model-based analysis. The findings can be utilized to design the spiral jet milling process for new APIs with less experimental effort. More detailed and robust insights can be obtained through further detailed statistical analysis, as well as by linking experimental results with the PBM.

[1] S.S. Bhonsale, Bard Stokbroekx, Jan Van Impe. Assessment of the parameter identifiability of population balance models for air jet mills, Computers & Chemical Engineering, 143 (5), 107056 (2020).
[2] S. Bnà, R. Ponzini, M. Cestari, C. Cavazzoni, C. Cottini, A. Benassi. Investigation of particle dynamics and classification mechanism in a spiral jet mill through computational fluid dynamics and discrete element methods, Powder Technology, 364, 746-773 (2020).
[3] R. MacDonald, D. Rowe, E. Martin, L. Gorringe. The spiral jet mill cut size equation, Powder Technology, 299, 26-40 (2016).
[4] S. Naik, B. Chaudhuri. Quantifying dry milling in pharmaceutical processing: a review on experimental and modeling approaches, Journal of Pharmaceutical Science, 104 (8), 2401-2413 (2015).