(T1530-02-14) Scaling-Up from the Bench: Model Based Approaches to Dry Granulation

Purpose: Bench-scale development offers a rapid, material sparing approach to assessing the impact of drug substance form, formulation, and tablet architecture on manufacturability, stability, & performance of the drug product. Roller compaction for dry granulation is a key unit operation for drug product development, and the ability to use a compaction simulator to predict target process conditions for kilo-scale roller compaction using gram quantities of material is useful and economically advantageous. Traditional empirical approaches rely on range-finding studies using process-scale equipment to both determine a target ribbon solid fraction and the process parameters to achieve it. This typically requires hundreds of grams of material, but predictive modeling can achieve the same end-result with less than a gram of material, enabling a new era of bench to CTM scaling. A calibrated roller compaction model with an analytical solution ^[1] to Johanson’s theory[^2] is presented (Eqn 1). The only material related inputs required by this new approach are compressibility factor (K) and material bulk solid fraction (SF_Bulk), which can be determined experimentally using the compaction simulator and material sparing techniques. Roll Force=((〖SF>〗_Target/〖SF〗_Bulk )^K (W×D)×π/2∙(Γ(K-0.5))/(Γ(K)))/√((2⁄S)/D) (Eqn 1) In this work two techniques, in-die and out-of-die analyses, were evaluated for determining K and SF_Bulk using a STYL’One Nano compaction simulator. These measured material properties were leveraged in the proposed roller compaction model to predict process parameters to achieve a target ribbon solid fraction. The accuracy of the prediction was evaluated as a function of the experimental technique and material properties.

Methods: Materials Highly ductile PROSOLV^® EASYtab SP (JRS Pharma) and highly brittle Kollitab^TM DC 87 L (BASF) Equipment: STYL’One Nano (Medel Pharm) compaction simulator, Mini-Polygran (Gerteis) roller compactor, Hand Mill (Gerteis) mill, GeoPyc (Micromeritics) envelope density tester, , and ST50 (Sotax)Tablet Tester. Compressibility per USP < 1062 > was examined by compressing two replicates of 100 mg compacts at compression stresses from 7 – 150 MPa at 1 mm/s speed (V profile) using a 6.35 mm round, flat faced tooling on the STYL’One Nano. In-die force-displacement data and out of die physical measurements of tablets were utilized along with absolute density values to determine solid fraction. Linear regression was employed for determining Compressibility Factor (K) and material bulk solid fraction (SF_Bulk) using the following equation.
ln (SF)= 1/K ln(CS)+ln(SF_Bulk ) (Eqn 2)
Figure 1 Assessment of K and SF_Bulk
Figure 2 In-Die Compressibility (orange) vs Out of Die Compressibility (Blue)
Granulation Bench-scale: Slugs were produced at approximately 0.80 MPa tensile strength using the compaction simulator. Solid fraction was determined from slug geometry, weight, and absolute density. Slugs were milled in the Hand Mill with a 1mm screen. Roller Compaction: Ribbons and granules were produced on the Mini-Polygran. Solid fraction of ribbons was measured via GeoPyc and absolute density. Predictive Modeling: Using the material properties from compaction simulator, roll forces were predicted using equation (1) to achieve  target solid fraction.

Results: Prediction accuracy was assessed by comparing the measured solid fraction of ribbons to the target. Figure 3 compares prediction accuracy using in-die and out-of-die techniques for brittle (high Compressibility Factor, K) and ductile (low Compressibility Factor) materials. As a surrogate for granule tensile strength, SF within ± 0.05 of target is considered acceptable. This translates to a roll force nominally controlled within ± 0.8 kN/cm. Predictions from out-of-die analysis resulted in SF within ± 0.05 of the target and in-die analysis predictions were less accurate, resulting in lower SF suggesting the model underestimated the roll force. The prediction was more accurate for materials with low Compressibility Factor, but the impact of experimental technique on prediction accuracy was consistent between materials. Figure 3: Prediction accuracy as a function of material properties

Conclusion: The use of predictive modeling coupled with material-sparing approaches expedites scaleup from bench-scale screening to kilo-scale roller compaction with only grams of material. This approach demonstrates how experimental derivation of Compressibility Factor, K and SF_Bulk can be coupled with fundamental models of roller compaction to accurately predict process parameters to achieve target product quality attributes. Results show that the out-of-die technique is more accurate than the in-die technique for determining K and SF_Bulk, but it does require additional time and material, in comparison. The use of this approach was demonstrated for both brittle and ductile materials, showing that accuracy does depend secondarily on material properties, but is acceptable within the range of properties representative of typical pharmaceutical formulations. Future work will focus on optimizing in-die measurement accuracy for more accurate prediction of K and SF_Bulk and demonstrating the technique with additional materials.

References: [1] Sousa, Ricardo, Pedro Cardoso Valente, Mostaf Nakach, Lionel Bardet, Marie Wacquet, Noel Midoux, and Jean-Rene Authelin. "Roller compaction scale-up made simple: an approximate analytical solution to Johanson's rolling theory." Journal of Pharmaceutical Sciences 109, no. 8 (2020): 2536-2543.
[2] Johanson, J. R. "A rolling theory for granular solids." (1965): 842-848.

Figure 1 Assessment of K and SF_Bulk.

Figure 2 In-Die Compressibility (orange) vs Out of Die Compressibility (Blue)

Figure 3: Prediction accuracy as a function of material properties