(W1230-01-04) A Hybrid Mechanistic Model for Predicting Milling Performance of Roller Compacted Ribbons

Purpose: Roller compaction is one of the commonly used manufacturing processes offering advantages of fast processing, easier scale up and the suitability for moisture and heat-sensitive drugs. However, due to lack of liquid addition in the process, the roller compacted granules after milling often demonstrate sub-optimal particle size and flowability which requires tuning the formulation composition or process parameters. The present study is aimed at developing a design space to better understand the milling performance of diverse formulations at different milling conditions. A mechanistic milling model was implemented to extract intrinsic milling parameters to establish predictive correlations on milling performance with formulation material properties and ribbon characteristics.

Methods: Roller compaction was performed on a Gerteis minicompactor under roll force of 4 kN/cm and 8 kN/cm with a 2 mm constant roll gap. A set of 9 diverse formulations were studied that included 3 placebo formulations of lactose (LAC), mannitol (MN) and dicalcium phosphate (DCPA) in microcrystalline cellulose at a 2:1 ratio respectively. Each placebo formulation was adjusted by 30% acetaminophen (ACM) or ibuprofen (IBU) to produce 6 drug formulation batches. The solid fractions (SF) of the ribbons were determined by volume displacement method using Geopyc. Various mechanical properties of the blends such as tensile strength and Young’s modulus were obtained on rectangular simulated ribbons compressed at various porosities. In addition, flat-faced tablets were prepared to obtain blend’s plasticity and brittleness parameters. Ribbons of each formulation thus produced at two different solid fractions were subjected to milling under 4 different settings - 1) using 1 mm screen at 60 rpm speed and 2) 150 rpm speed, 3) using 0.8 mm screen at 60 rpm speed within Gerteis mill chamber, respectively, and finally 4) using 1 mm screen at 1000 rpm speed on a comil. Particle sizing (PS) of all milled samples was carried out on Sympatec-Rodos by laser diffraction method. The PS data was fitted to a population balance model (PBM) to derive the breakage function (u) and breakage rate (s) in the milling process. Statistical regression analysis was then performed to predict u, s and mean size from the known information of material properties and ribbon SF.

Results: Judicious selection of formulations led to significant variation in ribbon SF that ranged from 55% to 85%. IBU due to its plastic nature produced greater SF than ACM for all the fillers, i.e., MN, LAC and DCPA. Similarly, IBU-based formulations had greater plasticity and lower brittleness than ACM-based formulations. Furthermore, greater roll force led to greater ribbon SF and mean particle size. It was found that milling in Gerteis at 60 rpm led to largest mean PS while in Comil produced smallest PS within the range 25-200 μm. The order of mean PS from Gerteis mill was at 60 rpm (1 mm screen) > 150 rpm (1 mm screen) > 60 rpm with 0.8 mm screen. Strong dependency for PBM parameters, u and s between both mills were observed for all formulations suggesting that milling performance on other setting can be predicted based on a given setting (Fig. 1). Finally, a regression model was developed to predict u and s from known material property and ribbon solid fractions that resulted in acceptable correlation (Fig. 2). In addition, the bimodality of PS was investigated in terms of Span value of the cumulative distribution which also led to reasonable prediction when the intrinsic material property and solid fraction information are known.

Conclusion: The study shows that ribbon solid fraction and intrinsic material properties of the formulation play critical role in controlling the milling outcome. Mechanistic characteristics such as breakage rate and breakage function can be predicted from formulation plasticity, brittleness, Young’s modulus, and solid fraction information to simulate cumulative particle size distribution profiles and probability of multimodality in the milled material’s particle size.


Fig. 1. Interdependence of breakage rate (u) and breakage function (s) at various mill settings.


Fig. 2. Predictability of u and s from blend properties and ribbon solid fraction.