Biopharmaceutical products are filtered numerous times in the course of their manufacture to control bioburden, remove precipitates
and separate solid contaminants (e.g.. fines from chromatography resins or diatomaceous earth flushed from depth filters).
In most cases, sterilizing-grade filters are used for product-stream filtration, although growing concerns about cost-of-goods
is resulting in increased use of bioburden reduction filters for these steps.
The following paragraphs describe the key characteristics one should look for when selecting a product-stream filter.
Validated 0.2 µm membranes.
As is the case with buffer filtration, the choice between a sterilizing-grade and a bio-burden-reduction filter often depends
on the unit operation to which the filtration is coupled. For example, sterilizing-grade membranes are required for the filtration
of bulk drug substance and are usually desired for filtration steps performed before any product hold. However, bioburden
reduction membranes maybe sufficient and more economical for many intermediate filtration steps (e.g., before a chromatography
column) and therefore their use can result in significant cost savings. As mentioned previously, regardless of the choice
of filter, it is important to choose a membrane that is validated for retention of bacteria and that the retention can be
correlated to an in-process integrity test.
Product-stream filtration is the highest value normal flow filtration operation in any biopharmaceutical process. Filter failures
which occur during product-stream filtration require time-consuming and costly investigations and may result in lost product.
As a result, normal flow filters used for product-stream filtration must
be constructed to withstand a broad range of operating conditions with respect to temperature, pressure and pH. Furthermore,
product-stream filters should be 100% tested by the manufacturer and should include instructions for integrity testing at
Product streams are some of the most challenging filtration steps and filter performance can vary widely because of the strong
effects of filter-fluid interactions. Therefore, it is important to select a filter that is optimized to provide high capacity,
regardless of the protein type, concentration, and formulation buffer. Furthermore, many product-stream filtrations are coupled
to relatively low-flow unit operations (e.g.. centrifugation, cell harvest or chromatography column loading), thereby making
differences in membrane permeability less important when determining the required filtration area.
Product streams contain a drug substance that will eventually be administered to a human patient. Therefore, product-stream
filters must be constructed of components that are proven to be biologically safe. Biological safety is demonstrated by the
performance of the USP (88) Class VI Plastics Test for Biological Reactivity and the burden of obtaining this information rests with the filter manufacturer
who should include test results as part of a validation package. In addition, vendors should be able to provide information
regarding filter effluent quality in terms of total organic carbon, buffering capacity, non-volatile residue, bacterial endotoxins,
and particle or fiber shedding.
Low protein binding.
Membranes and other materials used to construct filter cartridges and capsules should not bind proteins or preservatives that
are in the fluid because this may lead to product-loss or decreased shelf-life. Figure 5 shows the amount of several proteins
bound by microporous membranes cast from several common polymers. Most modern membranes are cast from PES or PVDF to ensure
minimal product loss.
Figure 5: Protein binding on various membrane materials.
Survey of filters for product-stream filtration.
GE Healthcare performed a study to evaluate the most common filters that are used for product-stream filtration. In this study,
two protein-containing feedstreams (1% bovine serum albumin and a monoclonal antibody purified on Protein-A chromatography
and subjected to low pH viral inactivation) were used to challenge a panel of membrane filters from Pall, Millipore, Sartorius,
and GE Healthcare. The specifics of the feedstreams are shown in Table 1.
Table I: Feed characteristics for product-stream filtration experiments.
Solutions were filtered at a constant pressure of 10 psid and the volume filtered as a function of time was recorded until
the flow rate had decayed by at least 50% or until the solution was exhausted. Based on the test results, estimations of the
number of equivalent 10-in. filter cartridges were made for a 2000-L batch filtered in 1 h. Results are presented in Figure
Figure 6: Comparative performance of multilayer sterilizing-grade filters for product stream (shorter bars represent better
In addition, experiments were run to compare the performance of ULTA Pure HC (sterilizing-grade) and ULTA Prime CG (bio-burden
reduction). As shown in Figure7,ULTA Prime CG provides equal or better filter capacity which, when coupled with a lower per-filter
cost, can translate to significant cost savings for applications where sterile-effluent is not required.
FIgure 7: Comparative performance ULTA Pure HC and ULTA Prime CG for product-stream filtration (shorter bars represent better