(M1430-08-47) Accurate Bulk Filling for an Automated and Flexible Drug Substance Manufacture

Purpose: Bulk filling is the final step in drug substance manufacturing; therefore, it is vital to ensure accuracy and consistency during this process to avoid high value product wastage. This step includes a 0.2 µm filtration of the drug substance prior to filling into biocontainers or bottles, configured in a single, serial, or parallel filtration format. Traditionally, bulk filling has been completed manually via full operator control of manifold assembly, filter preparation (e.g., manual venting), pump speed and product recovery. This practice has resulted in extended manufacturing shifts, increased risk of operator error and large product hold-ups in manifold assemblies. One solution to improve this issue is to transition to an automated process allowing faster production and higher consistency in batch filling. To optimize an automated bulk filling process, three major process elements must be considered. Firstly, the capacity and simplicity of the sterile flow path: it is important that the system has the capacity to fill the number of containers required, but also that this can be completed in a consistent manner mitigating the risk of operator error. Secondly, how the filling is controlled, as this will affect both the batch accuracy and the overall processing time. Thirdly, the hold-up within the flow path and the method by which the product is recovered to minimize wastage. Work has been completed using the Allegro™ Connect Bulk Fill System (BFS) to optimize an automated filtration and bulk filling process, featuring an in-situ pre-use post-sterilization integrity testing (PUPSIT) filtration design and a flexible, sterile distribution flow path configuration.

Methods: Simulant feeds were produced at viscosities of 1, 8, 20 and 35 cP; representing media, buffer, and drug substance feeds such as vaccines and high concentration monoclonal antibodies. Reverse Osmosis (RO) water was used for the 1 cP feed and a proprietary monoclonal antibody simulant was used at increasing concentrations for the viscous feeds. Four container configurations were studied, namely, biocontainer bag and bottle fills, each up to 120 containers and 320 containers, sized at nominal volumes from 1 L to 20 L. Biocontainer bags and bottles were each filled in sub-batches of six or sixteen, to 60 % and 80 % nominal volume, respectively. Filling was conducted by a four-step controller whereby the speed decreased as the container weight increased. This was pre-set and controlled by an automated recipe. Bulk filling trials were then completed to assess filling accuracy, flow path and filter integrity, and non-recoverable hold-up volumes to determine the optimal process configurations for the scoped applications.

Results: Prior to product commitment, the system completed in-situ automated pre-use leak testing at 400 mbar. Over the full study, all 56 assemblies passed leak testing. Additionally, filters sized from 1 in. to 20 in. (effective filter area: 0.04 to 2.08 m²) were installed to allow in-situ filtration prior to bulk filling. Filters were primed at 0.13 L/min, flushed with approximately 1.5× of the capsule hold-up volume with RO water and pre-use forward flow integrity tested (IT) in an automated and successive manner. Filters were installed in single, serial, or parallel configurations and all 43 filters passed pre-use and post-use IT. This assured the system was closed and integral prior to and after product commitment to bulk filling. The filling controller for each container size was optimized with the 1 cP feed and then used for subsequent viscous feed trials. Each container was successfully filled to the target weight with average maximum and minimum percentage set-point deviations noted at +1.19 % (7.2 g) for 1 L biocontainer bags and +0.03 % (4.5 g) for 20 L bottles filled with 1 cP feed. The maximum and minimum percentage set point deviation over all container types, viscous fluids, and system capacities were +1.83 % and 0.00 %. Draining for product recovery was completed on both the distribution and the filter set-ups of each configuration to assess the non-recoverable hold-up volume remaining in the flow path. Non-recoverable hold-up in the varied flow paths remained below 2.5 % of the initial batch volume, studied here from 20 L to 150 L.

Conclusion: This process optimization study shows the system can achieve consistent and accurate filling over the full scope of container sizes and bulk substance viscosities. This was also proven over the full range of 16 flow path configurations dependent on filter format (no filter, single, serial, or parallel), container type (bottle or biocontainer bag) and filling capacity (120 or 320 fills). The flexible process design and consistent product performance over a wide range of bulk filling applications suggests this system would suit high throughput multi-product facilities (e.g., CDMOs). Additionally, the automated nature of the system reduces the risk of operator error by limiting the number of interactions to assembly and disassembly of the manifold flow path and connected containers.


Average set-point deviation (%) for biocontainer bag and bottle filling to 60 % and 80 % nominal volume, respectively with water (~ 1 cP). Data shown was collected using 1 L, 2 L, 5 L, 10 L and 20 L containers. The ⊕ symbol indicates the average set point deviation and ‘n’ represents the sample number.


Average set-point deviation (%) for biocontainer bag and bottle filling to 60 % and 80 % nominal volume, respectively, with water (~ 1 cP) and three mAb simulant feeds at a viscosity of 8 cP, 20 cP and 35 cP. Data shown was collected using 1 L, 2 L, 5 L, 10 L and 20 L containers. The ⊕ symbol indicates the average set point deviation and ‘n’ represents the sample number.