(162b) Impact of Raw Material and Blend Properties on the Screw Feeding, Continuous Blending and Tableting Unit Operation of an Integrated Continuous Direct Compression Platform

Continuous manufacturing will likely change the way how drug products are developed and commercialized. Although continuous processes simplify the development process by removing the need of multiple unit operations and scale up, they also bring in new challenges. Only limited resources are available early in the development stage and hence the formulation and process needs to be developed in a short time frame with only a limited quantity of active ingredient. This is particularly challenging for continuous processes because most of the equipment is designed at production scale. Drug product development for continuous processes requires therefore innovative strategies and simulation to maximize the information with a minimal amount of active. Understanding and modeling the impact of material and blend properties at each unit operation is therefore the key step to develop drug products of high quality with minimal resources.

Direct compression is by far the simplest and most cost efficient manufacturing route for oral solid dosage forms. Although direct compression is an inherently continuous technique, simple unit operations preceding tableting (i.e. weighing and blending) are historically performed batch wise. To enable continuous direct compression, the integration of continuous powder feeding units, continuous dry powder mixers and a tablet press is required.

This work aims to leverage raw material and blend properties with required settings for optimal feeding, blending and tableting in an integrated continuous direct compression platform. First, the raw material properties of a wide range of fillers for direct compression and active pharmaceutical ingredients (API) were determined. The characterization included determination of particle size and shape (laser diffraction, static image analysis, scanning electron microscopy), specific surface area, tapped, bulk and true density, flowability (drained angle of repose, dynamic angle of repose, flow function coefficient, basic flowability energy, specific energy, aeration index, permeability), compressibility, charge density, moisture content (loss-on-drying) and dynamic vapor sorption. Subsequently, principal component analysis was applied on the API and filler dataset. This allowed to select 6 APIs (acetaminophen micronized, powder and dense power, caffeine powder, metoprolol micronized and etravirine spray dried) and 5 DC fillers (Avicel PH 101 and 200, Emcompress, Pearlitol 100SD, Tablettose 80) with unique properties. These were combined together with magnesium stearate in 30 tertiary mixtures that were also characterized for their properties.

Continuous direct compression trials were performed with these formulations to evaluate the impact of blending (impeller speed and configuration), die filling (paddle wheel speed) and compression (main compression force) parameters on responses of the process (fill level, mean residence time and number of blade passes) and product (blend and content uniformity, tablet weight variability and tensile strength).

The hopper refill performance could be related to the flowability of the powder. Materials with a flow function coefficient below 2 (i.e. no powder flow) gradually reduced the apparent volume of the refill valve over time. At the feeding stage, the amount of material dispensed per screw revolution was related to the bulk and tapped density whereas the variability on the feeder flow rate was correlated with the flow function coefficient of the material. At the first or intensive mixing stage, the fluidization state of the powder bed depended on the process settings. The residence mass in the first continuous mixer was positively correlated with the bulk and tapped density of the blend when the mixer operated in the densely stirred powder bed flow regime. Additionally, more permeable blends resulted in lower residence masses as these blends fluidized more easily. At the second blending or lubrication stage, all blends were completely fluidized and as a result this response was not correlated with material properties. At the die filling stage of the tableting unit operation, the residence mass in the feed frame of the tablet press was a strong function of the bulk density of the blend (R²=0.97) as the powder bed was in a densely stirred regime during processing. In contrast, the fill depth required for achieving target tablet weight (175 mg) was negatively correlated with the bulk density (R²=-0.90). In contrast, tablet weight variability depended on the compressibility (R²=0.68), unconfined yield strength (R²=0.72) and flow function coefficient (R²=0.56) of the blends. Overall, blends characterized by a low density and poor flowability yielded the highest tablet weight variability at low paddle speed.

The tabletability was strongly dependent on the selected filler. The cellulose based fillers resulted overall in tablets with the highest tensile strength. Additionally, Avicel PH101 showed better tabletability compared to PH200 driven by the higher bonding area of the smaller particles. However, the cellulose based fillers exhibited a high lubricant sensitivity. As the cellulose particles were subjected to a higher number of paddle wheel revolutions, the tensile strength decreased with 45% for some blends. Pearlitol 100SD also resulted in tablets of good strength and showed only a limited lubrication sensitivity. However, high ejection forces were observed with pearlitol 100 SD and higher levels of magnesium stearate (1.25 % w/w) were required to limit the ejection force during tableting. Formulations consisting of Tablettose 80 or Emcompress fillers with the elastically deforming acetaminophens resulted in relatively weak tablets.

The database will be further completed with blend and content uniformity data. Subsequently, advanced multivariate models such as T-PLS will be applied to link the API and Filler properties to the process parameters and all responses of the continuous direct compression process. These models are valuable tools to select excipient properties as a function of API properties during drug product development for continuous direct compression processes.