Development of a sampling method to validate the cleaning process of an automatic hard capsule filler

Introduction
The increase in demand, the use of preparations during drug shortage and the possibility of increasing batch sizes following the new Good Preparation Practices for Compounding necessitate the automation of compounding processes. Our facility is already equipped with an automatic capsule filler (INCAP SE, Bonapace) acquired for a clinical trial in 2021. In the context of scaling up automated production using this capsule filler, it is necessary to validate the cleaning process (CLEAN). Our aim is to develop a method for validating the CLEAN between two productions of different molecules.

Materials and Methods
The Good Manufacturing Practice recommend validating the CLEAN by swabbing and using a worst case scenario approach to select the molecule for monitoring, the sampling plan and the sampling methodology. For each molecule, cleanability, solubility in water and ethanol, toxicity and therapeutic activity were analysed. A score from 1 to 5 was assigned to each criterion, enabling a total score to be calculated for each molecule, the one with the highest score being the most complex to clean. Similarly, the sampling plan was evaluated based on the following criteria: difficulty of CLEAN, surface porosity, routine accumulation and direct contact with raw materials. The sampling methodology (number of swabs, solvent and volume used, extraction time) was studied and validated by measuring the recovery rate, which had to be greater than 75%. The residual quantities swabbed were measured using high-performance liquid chromatography coupled to a diode array detector.

Results
Four compounds intended for automated production were analysed. Prednisone represents the worst case senario with a score of 540 out of 3125, compared to 288 for sildenafil, and 30 for both caffeine and spironolactone. The dosing disk, hopper, discharge chute, capsule moulds and sealing slide are the most critical among the 18 selected points, involving 3 types of materials: plastic, cast iron and aluminum. For aluminum samples, we opted to extract prednisone using two swabs in 10 mL ethanol for 15 minutes. The average recovery rate over 12 measurements and 3 operators is 82 9.3 %, with repeatability and reproducibility covariance of 11 and 11.3% respectively.

Conclusion and discussion
The sampling method is validated for aluminum surfaces and must be extended to each material in the capsule filler. Conducting these samples will allow us to validate the CLEAN of the capsule filler. Finally, the developed methodology will also be used to validate the operators in the CLEAN steps and monitor the process.

See the poster