Sterile filtration QRM & Pupsit

Annex 1 is in the second revision cycle. The last revision still showed PUPSIT being included in Annex 1, although not as rigid as previously and a bit better worded. PDA provided comments on PUPSIT related sections of Annex 1 during both review cycles. We are currently awaiting response to comments and finalization of the Annex. No final release date has been provided.

Annex 1 was first published in 1971, a 16-page document, which still requires to be located as a whole document to see whether PUPSIT was mentioned. Revisions followed and the earliest version of PUPSIT can be found in Volume 4 Good Manufacturing Practices, Medicinal Products for Human and Veterinary Use, 1998. Int his document, page 73, paragraph 85 it states, “The integrity of the sterilized filter should be verified before use…”. However, PUPSIT was not enforced by the European regulatory authorities. It only was sporadically enforced starting 2007.

The foundation of the enforcement was a published EU GMP Guide Annexes Q&A of 2007, which states: “The filter-sterilization process may be physically stressful for the filter. For example, high temperatures during the process may cause the filter to distort, potentially leading to fluid pathways that allow the passage of particles greater than 0.2 μm in size. The performance of a filter can improve with use, as particles begin to block individual pathways and remove larger pathways that smaller particles could successfully navigate. For these reasons, filters should be tested both before use but after sterilization and again after use.” Based on the language in the 2007 Q&A statement and also statements made by European regulators, it became clear that the flawed filter masking is a major reason why PUPSIT is now being enforced.

The foulant used was OvaltineTM, a proteinaceous malt, cocoa extract, at specific concentrations, which has been found to mimic biologic solutions exceptionally well. The concentration of Ovaltine™ was 24 g/L at the worst-case trial scenario and 0.8 g/L for the more realistic case.

Typically, that indication can be detected during the initial filterability test, meaning the filter/solution trial which is used to size the filter system. The test will show whether the fluid is a high foulant fluid, meaning the test filter blocks rapidly and the total throughput is diminished.
Risk assessment of the filter manufacturing, handling, preparation, sterilization, usage, and controls can provide information on the potential for filter flaws that could contribute to masking.

One requires to evaluate the entire filtration system and design PUPSIT into the system (existing or new). There are certain prerequisites which require to be fulfilled to implement PUPSIT and these need to be reviewed and taken care of right from the start to avoid revising the filter system at a later stage. For example, in most cases PUPSIT will add to the complexity of the system due to the fact that filter wetting fluids require to be captured on the filtrate side, the filtrate side needs to be under atmospheric pressure, the filter may need to be dried after the test before use, vent filters on the filtrate side may need to integrity tested before the filtration process starts running etc. The details of the factors which require to be taken into account implementing PUPSIT can also be found in PDA Points to Consider of Pre-Use Post-Sterilization Integrity Testing (PUPSIT).

Throughout the PDA and BioPhorum SFQRM consortium activities, communication happened with European and other regulators. The consortium group made it a point to update the regulators on the progress of our workgroups and submitted the masking test protocol and publications for review. The very constructive interaction with the regulators made the consortium work as valuable and we hope this may be a good example case for future projects.

That question is still open. The process validation of sterilizing grade filters asks for a product bacteria challenge test under process conditions, meaning the maximum pressure conditions during the filtration process. When PUPSIT is used, the pressure of the product or water wetted filter goes up above the Bubble Point, which typically is higher than 50 psi. The question prevails, whether the product bacteria challenge test needs to be performed at that pressure conditions. Having said this, the integrity test phase is very brief, the pressure will be at the 50 psi level only very shortly, whereby the differential pressure during filtration is utilized at a much longer timeframe.

It depends on the complexity of the filter system before and after and how this system is used. A single-use process system, which may be configured in its design to allow PUPSIT may still be a functionally closed system, without major impact on the filtrate side of the system. However, in the single-use case one would probably see a product wet PUPSIT being run with additional tests of the vent filters, on the filtrate side. This would mean that the time of product contact within the filter matrix is prolonged, which would raise the question, what effect does it have on unspecific adsorption, fouling of the filter, microbial growth etc. The wetting fluid will not be used, but we need to understand that we have a stop within the filtration process, which typically is validated. Question may be raised, whether the PUPSIT activity has an effect on the filter, filter matrix, retentivity, leachable levels, unspecific adsorption, microbial growth or size influence

There is no recommended wetting solution, as both wetting methods have its disadvantage. The wetting method is application, filter system design and end-user dependent. More about the advantages and disadvantages of the wetting methods can be found in PDA Points to Consider of Pre-Use Post-Sterilization Integrity Testing (PUPSIT).

There has to be some type of measurement of the filtration process; in this case where flow is not tested and the feed pressure stays constant, it would be filtration time. If the filtration time increases, one can be assured that the filter fouls and becomes blocked, meaning the risk of masking increases. Having said this, is required for terminal sterilizing grade filters, before the filling system and one would have to measure flow to feed the fill line.
In addition, in the initial trial for to size a filter system for a new application, filterability trials would be performed. These trials determine how fast the filter blocks and according to these trials a filter system is sized for the specific application. The filterability trial creates a good indication whether the fluid is high foulant or not and how the filtration system requires to be configured and sized to avoid any elevated blockage of the filter.

The PUPSIT enforcement happened when the Q&A of 2007 stated that there may be a risk of masking of a flaw within the filter membrane. The consortium performed masking trial tests and determined that filter flaw masking is possible, although to do so the filter blockage has to be above 80%. This means the consortium created scientific data, which can be utilized to run trial work with any fluid to be filtered and determine whether the fluid is high foulant or not and how the filtration system needs to be sized to stay well below an elevated blocking rate. All taking into consideration, we now have test points to assess the risk of masking and therefore can gauge whether PUPSIT would be advisable or not, since PUPSIT has its own risks attached

Ultimately the regulators require to recognize that the filter user is in control of their filtration process; meaning the sterilization of the filter is validated and controlled, the fluid properties and process conditions are validated and controlled, the handling and installation is controlled and so on. The enduser has all the need to be in control to assure that the filter’s filtrate is sterile and will remain so during the filling process. PUPSIT is a measurement tool, with risk attached and may weaken the previously design robustness of a controlled process. The regulators need to be aware about this and rather determine whether an

Again, one requires to be in control of the process and the sterile filtration activity. To gain this control, one requires to process validate the filtration system within the specific application under specific process conditions. Once done, certain test methods, like the post-use integrity test are helpful tools to see that the filter used has been integral. Moreover, training is a key element, which has to be implemented as any process validation exercise would be void without knowledgeable end-users.

Certainly prevention, as detection is too late and shows that relies on measuring quality instead of implementing quality. Prevention is key and a supporting tool may be detection. Prevention means one is in control of the process, which requires to be the modus operandi.

Absolutely, as PUPSIT is not without risk and that risk requires to be taken into considerations. The data created by the SFQRM consortium help and support such risk assessment, as masking risks were determined, risk assessments of filter manufacturing, transportation, storage, installation etc. were all evaluated and can be judged in comparison to the risks PUPSIT presents to a filtration system. The years of discussions about and evaluations of PUPSIT showed that the ideal is to perform an objective risk assessment instead of generically implement PUPSIT.

Specific PUPSIT assemblies, configurations, and procedures should be included in a risk assessment to determine if and to what extent they pose such a risk to product sterility and what controls can be employed to reduce that risk and what studies can be used to validate the effectiveness of those controls.
It should be noted that PUPSIT usually involves the installation and manipulation of fittings downstream of the sterilizing filter which if not designed, maintained, or performed properly could compromise the down-stream sterile pathway. Because the PUPSIT assembly and procedure provide additional process complexity and variables – that can pose a risk to the sterile product pathway and product sterility – the assembly and resulting activities should be included in each of the PUPSIT relevant APS studies.

The Annex 1 commenting team encouraged to take this example out of Annex 1, as a) there may be many opinions of what defines a small batch and b) even a small batch of a high foulant fluid could have a masking effect, which has been the main reason for the encouragement of the PUPSIT implementation. Small or large batch more importantly is to have control over the filtration process to be assured a sterile effluent is achieved throughout the filtration process.

Masking of the second filter within a redundant filtration set-up, the filter closest to the filling line will be most probably voided by the first filter separating any foulants which could cause masking on filter two. However, redundant filtration means that either of the filters are assigned to establish a sterile filtrate, which means if the second filter fails the post-use test, the first filter will be tested and if it passes it is assumed the filtrate has been sterile. This means that filter one, the one which carries the brunt of the foulant retention, is crucial and therefore requires to be pre-use, post sterilization integrity tested. It may be argued that filter two had no masking propensity and therefore does need to undergo PUPSIT. That argument has to be made and the outcome is unknown for now.

The discussion we had with European regulators was positive. The scientific approach has been appreciated. The last revision showed a tendency of a softer tone and the inclusion of risk assessment wording. However, we do not know what the final outcome of the Annex 1 will look like.

The update discussions we had were very constructive and the feed-back positive. The proactive approach and continuous involvement of the European regulators was appreciated.

Since both filters are validated as sterilizing grade filters, achieving a sterile filtrate as individual filters and the secondary filter is used as insurance filter, presumably one has to perform PUPSIT on both. This certainly raises the complexity of the filter system set-up and we have seen companies move from redundant filtration to a single sterilizing grade filter design due to the complexity to test a redundant filtration system.

We have not investigated true masking with the data mining activity but checked whether there is an influence on the Bubble Point measurement before and after the product bacteria challenge test. We found that a Bubble Point shift only happened in 5 of 518 data sets and the fluids used in these 5 outliers were high foulant fluids. However, we do not know whether these fluids would have masked a flaw or not, as all filters used were integral pre- and post-use.

No, we have not, as this was not part of the scope. It may be the case that a solvent flush as the wetting fluid removes the masking, but that requires to be investigated on an application by application basis.

TR66 states that PUPSIT is less important for filters that have been gamma sterilized, as filter flaws as experienced with steam sterilized filters do not occur. It is true gamma sterilization is much less stressful on the filter membrane and device than steam sterilization. However, this has not been into consideration by regulators and the request for PUPSIT has been seen in both instances, steam and gamma sterilization.

The tests were performed with product and water and in instances with water for the pre-use test and product-wet with the post-use test and a correlation factor utilized to compare to the water wet test results. In a case where water wet is used.

If contaminants are flushed out of a filter the flow rate would recover, which did not happen. Typically, when contaminants are entrapped in the filters matrix it will not be possible to recover the filter with a water wetting flush, but one has to use a rigorous cleaning cycle using chemicals and water at elevated temperatures. Besides, if the foulant used could be flushed out of the filter matrix, all filters would have failed the test, instead of the detection of few filters pass the post-use test.

There are quite a few factors that could promote the masking of a filter by the feed stream.
• The feed stream itself creates implication, meaning what product is filtered, for example Methylcellulose is known to foul filters rapidly
• The type of foulant in the feed stream, the composition and behavior of the foulant, for example can the foulant create a collapsible, dense filter cake on top of the membrane
• The filtration process conditions, like differential pressure can have an effect on masking as higher differential pressure could compact the filter cake.
• The membrane constitution may have an effect on masking, as some of the membrane materials are highly adsorptive and foul faster than others.
• The filter membrane combination can have an effect, as a heterogenous membrane combination with 0.45/0.2 micron layers, may have an advantage to protect the 0.2 micron membrane.

These are just a few factors, which require to be observed. However, all of these factors are tested
during the filterability trials run to size the filtration system. The filterability trials will allow a judgment
of the masking potential due to the fluid stream and filtration conditions.

During the Test Process and Results of Potential Masking of Sterilizing Grade Filters study trials, different foulant concentrations were used and the masking of the filter flaws were detected above 80% blockage rate. This gives an indication that the filter matrix requires to be blocked at a high level before masking of minor membrane flaws is happening.

There is not a big difference in risks, neither complexity, as both water and product wet PUPSIT cause a multitude of actions that need to be taken to perform the test. The water and product wet PUPSIT have their distinct test activity needs. For example, when performing the water wet PUPSIT one requires to dry the membrane before use to avoid product dilution. When performing the product wet test, one has to make sure the wetting conditions allow the use as least product as possible to wet the entire filter membrane and the product may requires to be discarded after the test. Both wetting methods include a higher level of complexity than not performing PUPSIT.

The foulant used was OvaltineTM, a proteinaceous malt, cocoa extract, at specific concentrations, which has been found to mimic biologic solutions exceptionally well. The concentration of Ovaltine™ was 24 g/L at the worst-case trial scenario and 0.8 g/L for the more realistic case.

PUPSIT does not specify any sterilization method, therefore gamma sterilization of the filter before the pre-use integrity test is fine. Moreover, gamma sterilized filters should never be steam sterilized or autoclaved as it may damage the membrane.

No, we have not, as we utilized one foulant (Ovaltine at two different concentrations) only. A database of different fluids with fouling propensities would be beneficial but requires to be collected over time.

That varies from end-user to end-user and application to application, therefore one would not be able
to define it.

When the product wet integrity test is used, the filter will be flushed with product being the wetting fluid, which means the fluid utilized at that point can be considered the filter flush volume to remove any leachables. The product used for the filter wetting process would be discarded and not being used as a filtered product. After PUPSIT with the product, the filter can be used without another filter
flush.

0.8 g/L, it has been found a very good mimic for biologic solutions.

No, 24 g/L is a worst case foulant concentration and it showed in the Phase 1 trials when 10” filter cartridges blocked rapidly. As an example, the typical total throughput on a 10” filter with 0.8 g/L Ovaltine can be 90-100 kg, whereby the total throughput at 24 g/L was around 3-4 kg.

No, with such a high blocking rate one would risk the need of a sterilizing grade filter exchange before the batch is filtered, which would be an unacceptable event. Typically, the blocking rate is much lower as the filters are sized with safety margins to avoid any blocking of the filter.

Initial surveys showed that the blockage rate in terminal sterilizing grade filtration is around 20-30%.

The four filter suppliers tried to collect as many borderline flawed 10” filters as possible and 24 filters were all we could collect in 6 months of production.

We were required to gain more masking test results to enhance the data base, especially also using the foulant concentration of 0.8 g/L. Since it was very difficult to gain flawed 10” filter elements, we resolved to use 47 mm disc filters flawed on purpose via laser drilling.

10 microns.

In Phase 1 of the masking studies with the 10” filter cartridges, also the diffusive flow test was used, and these test results require to be still published.

When a new application requires the implementation of a sterilizing grade filter, that fluid has to undergo a filterability trial to determine the appropriate filter unit and size. The filterability trial will be the test to gauge whether the fluid stream is a high fouling fluid or not and the results of this test can be used for risk assessment purposes.

The wetting fluids should not create membrane swelling. If membrane swelling occurs it would indicate a chemical incompatibility of the membrane with the fluids, which is undesirable and has to be avoided. For that reason, chemical compatibility is tested during the filter process validation.

Both tests would work, it depends on the filter size whether one uses diffusive flow or Bubble Point test. In the case of the 47 mm disc trials only the Bubble Point test could be used.

Very representative as a multitude of different filter materials and configurations were tested. Since the four major filter suppliers were involved, the trial work covered the majority of the filter types used in the industry.

Vent filters are typically very robust filters and do not experience fouling. For that reason, the filters are commonly not pre-use integrity tested but post-use. As there is no filter fouling happening, masking is void and therefore the post-use test would detect any potentially flawed filter. It is to the end-user whether the end-user wants to use a pre-use test or not as it is a commercial risk and not a patient safety risk.

Yes, the results are published.

That may be a task that can be considered in the future, utilizing the knowledge and experience of the filter end-users.

It can be, but ideally it is in front of the sterilizing filter since one does not know what happens to the fluid within the process step between bioburden reducing and sterile filter.

A post-test to pre-test bubble point ratio was calculated for each filter and the distribution of these ratios examined to determine whether there were cases of elevation of the BP due to bacterial retention restricting flow through the membrane to the point where masking of a filter defect could occur; that is, a defective filter could pass integrity testing due to the restricted flow and apparent reduction in filter pore size due to the bacteria retained during the validation work. Ratios were averaged across all tests for the same test fluid. The mean average ratio was 1.00 ± 0.15, indicating that on the average.

So far, I have heard only the enforcement in terminal sterilizing grade filtration, not in other filtration applications or positions.

Phase 1 was done in the PDA Training & Research Institute, Phase 2 with the membrane discs was performed in the filter supplier application labs, due to time constraints.

The fluids used in the few events of a Bubble Point shift were high foulant fluids, therefore considered as fluids with risk for masking.

Potentially, but that still requires appropriate testing to verify the 0.45 micron membrane protects the 0.2 micron membrane and does not foul itself.

Such membranes were used in the data mining results, as the data mining looked at a large database of product bacteria challenge tests, which typically use a low Bubble Point filter membrane. The masking studies did not consider such membranes as membranes with flaws were needed and used.

The diffusive flow test found one of the masked filters as failed post-use. That same filter passed Bubble Point test.

The laser drilling does not allow smaller holes to be drilled and it is important to create a detectable filter flaw level.

Not confident at all, as there is no such fluid to represent all the fluids within the industry. However, it is a very good representation of biological fluids.

The filters underwent a thorough wetting procedure to assure wetting was appropriate. Typically, a wetting problem creates a gross failure and one would rewet to see the failure diminishing or being voided.

Any defect, smaller or larger, is considered a defect because the wetting fluid is expelled at a differential pressure level lower than the minimum allowable Bubble Point. Therefore, any defect would be detected, except when the defect is masked.

Yes, it would at least be a good step to use the filterability trial work to determine whether a fluid is a high or low fouling fluid. It creates a good indication of whether or not there is a propensity of masking.

Everything in this question describes the filterability trial work performed within this particular end-users application, therefore an “own filterability” trial has been performed and is typically performed. What we described in the “generic” masking study is a masking trial, not a filterability trial.

Yes, it is as a flaw created by the pre-use sterilization process is typically a larger, more catastrophic event. We did not want to determine how the filter behaves with a catastrophic flaw but wanted to know whether a minor flaw can be detected post-use or whether such a flaw may be masked. Therefore, the masking trial work utilized very specific filters with flaws that were minor and potentially not as large as one find when the steam sterilization handling went wrong and caused major damage.

Any effect of the product to the filter or vice versa requires to be evaluated and validated. So yes, the residence time of the product within the filter matrix during the tests of the vent filters requires to be taken into considerations within the process validation of the filter.

The filter should not be subjected to a higher differential pressure at a specific temperature setting than the maximum allowable operating pressure given by the filter supplier. This guidance would be a good level of assurance that the filter is not damaged. However, when the filter is dried after the water wetting cycle and PUPSIT the Bubble Point of the membrane needs to be exceeded, which
would represent a high differential pressure condition through the larger pore area and that for a prolonged drying time. One needs to find out whether this drying cycle potentially damages the filter or not, by running tests within the process validation of the filter.

The data created by SFQRM are a good scientific database to discuss the necessity of PUPSIT within specific applications, what type of risk may be involved for not or for implementing PUPSIT, and what needs to be done when implementing PUPSIT. All data from all workgroups create a good basis for evaluating one’s own processes, but also to discuss these processes including the findings of the SFQRM group.

According to Annex 1 the sterilizing grade filter needs to be tested, therefore it is the last filter in front of the filling needle. When the complexity of PUPSIT requires the filter to be further away from the filling needle then it needs to be so, as PUPSIT is required according to Annex 1.

When the risk of masking is minimal, one has to balance the risk of masking and PUPSIT implementation. Ultimately one needs to be in control of the filtration process and if this control is jeopardized by a high degree of complexity without any patient safety enhancement it would make no sense to implement PUPSIT.

Any process is different, from application to application, but also whether it is a hard-piped process or a single-use process. The depiction in the Points-to-Consider document are just examples and a PUPSIT implementation requires to be evaluated in detail for the different process layouts and applications.

The pros and cons of water, buffer, or product wet are described within the Points-to-Consider document in detail. There is no preference given by regulators.

No

Filter process validation is performed ideally with the product to be filtered under the process conditions of the filtration activity at the end-user. If there is an elevated temperature the bacteria challenge test requires to be performed under this elevated temperature, as long as the organisms are viable at that temperature. A viability test under these conditions is necessary to determine
whether the organisms are viable or not. If the temperature is too high and the organism shows mortality, one has evidence that the organism does not survive. At that point one reduces the temperature to reach viability and runs the product bacteria challenge test at these modified conditions.

It will not just be the evaluation of the masking risk, which creates a decision whether PUPSIT should be implemented or not, but also the risk elevation when PUPSIT is implemented. The risk balance requires to be evaluated and a decision made from that risk balance.

The failure rate may vary by connector type and also end-user training. It is difficult to put a generic number on the failure rates, as it may vary from site to site. A failed connector can be detected by leaks at or by the connector.

Any precipitation creates a higher fouling fluid, as seen in the past with animal-derived serum, which typically had precipitates and fouled filters rapidly. So, when proteins precipitate the fouling rate is typically higher.

Filters that are steam sterilized in-situ do not need to be flushed as the steam would flush out any potential leachables. Gamma sterilized system may need to be flushed with a small portion of the product if the first milliliter going directly into a vial. This would mean one would discard the first vials filled but just continue the filling process. PUPIST would mean one flushes the filter and discards the flush volume, runs the test, and requires multiple vents filters and additional connections (increased complexity), and flush the filter system again. The time and additional activities make the PUPSIT more complex.

Typically, the filter integrity test is an automated test, which does not necessarily reduce the risk, as downstream manipulation requires to be performed also with an automated test device. To divert the wetting fluid, vent the downstream filtrate side, potentially dry the filter or flush additional product to the downstream side all requires to be performed. Automation does not take this burden away. In addition, the risk of microbial ingress due to the increased complexity of the downstream side will not be voided by automation.

Yes, the filter requires to be fully wetted with product to perform the PUPSIT accurately. After the PUPSIT it would be advisable to flush the filter with product, as one does not know what happened to the product within the filter matrix during the test time of the sterilizing grade liquid filter and the vent filters afterward. So, it is better to discard that product of the filter matrix.

The so-called blow-through theory, when organisms supposedly are pushed through a membrane by elevated pressure has yet to be proven. It has been determined that organisms like mycoplasma can be pushed through a membrane at elevated pressures, but typically one would not find mycoplasma as a contaminant during terminal filtration. However, if there is the risk of a native organism being
pushed through the membrane matrix product bacteria challenge tests need to be performed to determine whether this happens or not. Similar to the potential damage of the filter by a drying cycle at elevated differential pressure, the risk requires to be tested.