IPQ Inside the global regulatory dialogue

Uncertainties in Subvisible Particulates Need to Be Understood, FDA Tells Biotech CMC Strategy Forum

The uncertainties around subvisible particles (SVPs) may warrant their being controlled as a critical quality attribute, FDA Office of Biotechnology Products Division of Therapeutic Proteins (DTP) Deputy Director Barry Cherney stressed at a CASSS/FDA CMC Strategy Forum in Washington, D.C in January.

Drawing from the ICH Q8 guidance, Cherney underscored, as a “major point,” the applicability to SVPs of the principle that “if there is a theoretical risk of uncertain significance, in the absence of convincing data that the attribute is not critical, you should assume it is critical and needs to be controlled.”

The January forum – the latest in the series of CMC strategy forums held semiannually in the US on key biotech regulatory concerns – brought industry and agency experts together to discuss “a practical approach to the analysis and immunogenic potential of aggregates and particles.”

The uncertainties that present themselves around the analysis, significance and control of particulates in biotech products have been receiving considerable attention recently, and the CMC strategy forum continued the dialogue that has taken place at industry/regulator venues in the latter half of 2010.

The issues were explored at a session of the AAPS annual meeting in New Orleans in November and at a two and a half-day USP workshop held at its headquarters in Maryland in December. The USP workshop assessed the current science and regulation of particulates in biotech products as well as in aerosols and small molecule injectables and how the pharmacopeia can contribute in the guidance and standard-setting effort.

At the CMC strategy forum, Cherney focused on the “current regulatory considerations for the assessment of subvisible particles” – an area in the particulate debate where the analytical and clinical uncertainties are especially challenging and the science fast-moving.

Cherney offered insights on: ● the knowns and unknowns of SVPs and immunogenicity ● assessing the risks of SVP occurrence and the severity of the consequences ● risk reduction and control strategies, and ● current recommendations.

[The rest of this story is normally available only to subscribers. By special arrangement with IPQ, the full story is being made available to PDA members on a complimentary basis.  For IPQ subscription/license information, click here.]

In providing some background on the “knowns and unknowns,” Cherney stressed that coming to grips with the current level of understanding is necessary to build the base for the risk assessment process.  “If you don’t frame what the issues are, it is hard to have a discussion and do risk assessment,” he said.

Document How Certain And Uncertain You Are and Assess Risk

Explaining that for SVPs, “the types and amounts of protein aggregates that may induce an immune response for our products are not clear,” Cherney provided recommendations on what firms should do in assessing and reducing risk in the face of this uncertainty.

“I think one of the things you do,” he affirmed, “is that you document exactly how certain and uncertain you are about your product, and one of the ways to do that is to do risk assessment.”

At issue is the probability of occurrence of the SVPs and their detectability.

Cherney emphasized that understanding the occurrence requires “assessing the relevant manufacturing and storage conditions that you are using that could lead to increased amounts” of SVPs, adding that, “of course,” the analytical techniques need to be relevant.

An underlying theme in the biotech regulatory official’s remarks was that “if you are not employing these methods designed to detect and characterize subvisible particulates, you are exposing patients to unnecessary risk.  Ignoring a risk is not risk mitigation.”

The product characteristics and risk evaluation, in turn, determine whether monitoring and characterization needs to be done on an ongoing basis in addition to during the development program. “You should do these characterizations as part of your development program, but continued characterization activities following manufacturing changes, I think, are also important,” Cherney commented.

Clinical trials and post-marketing adverse event reporting are not sufficient to qualify the immunogenic significance of subvisible particles and eliminate the need for post-approval assessments and the improvement opportunities they provide, the regulator stressed. The understanding of risks evolves. “Risk assessments should be preformed on a periodic basis following approval, and any newly-identified risks should be mitigated in a timely fashion” as part of the continuous improvement effort.

From Uncertainty to Risk Reduction and Control

After exploring the occurrence and severity of immune responses, Cherney focused on the risk reduction and control strategies that are needed in the face of the uncertainities surrounding SVPs.

Making the point that the uncertainties about the risks may elevate SVPs into the critical attribute/control category, he stressed that control does not mean setting a specification.

“What ICH says about critical quality attributes is that they are controlled within an appropriate limit range or distribution to ensure the desired product quality.” However, that control “does not necessarily mean that attribute needs to be specified.”

Action limits, for example, Cherney said, can be used for SVPs to allow for a more thorough investigation when an out-of-trend event is observed. Issues such as the nature of the particulate and the amount of protein “will inform your risk assessments and you can plug it in and say, ‘well, is this a risk or not, and what do we do with it?’”

Algorithms that indicate when additional testing is warranted can be applied, as well as additional sampling to address the assay variability problem and validation to better assure the attribute is adequately controlled by the process.

Cherney concluded his talk on addressing SVPs by offering a series of concrete recommendations for: ● characterization ● testing in the micron and submicron ranges ● continuous process improvement, and ● approval and post-approval.

A robust characterization of the subvisible content of the drug product should be conducted at release, on stability and in use. In turn, continuous process improvement should entail the use of new technologies “particularly in areas where we know there is an analytical gap” – for example, in the 0.1 to 1 micron range.

In general, for approval and post-approval, the biotech official said, FDA has been suggesting that companies “conduct a robust risk assessment of the potential impact that subvisible particulates have on the quality, safety and efficacy of your product, and propose a strategy that provides adequate control of perceived risks they may pose.”

Differentiating Micron and Submicron Particulates is Useful

In responding to a question on the vagaries of SVP terminology, Cherney referenced the earlier talk by Amgen’s Linda Nahri on SVP testing in which she explained the value of differentiating the issues involved in submicron particulates from those in the micron and bigger range.

The FDA official stressed the “usefulness” of this distinction in addressing the different capabilities of the analytical methods in the two contexts.

In the micron range, there “are good analytical techniques” to do quantitation, while the focus in the submicron range is on qualitative assessment, he noted, adding that the latter qualitative techniques should be applied during development.

Cherney pointed out that there are techniques coming on line that will be useful, particularly when stressing the product.

“As you go from very [large] to the smaller subvisible particulates as we are now defining them, you see a great increase in the numbers.” He noted that some of the techniques are showing 200,000 particles in the 2-10 micron range. For some products, patients may be receiving more than microgram quantities chronically. The question, Cherney said, is “what is the risk of immunogenicity there?”

Another participant asked what FDA’s view was of the use of filters to eliminate SVPs.

“If you have in-line filters – and some of our products do – I think most of the risks go away provided you show that that filter gets rid of the particulates of interest,” Cherney responded. While you can “validate those things,” the issue might be around the impact of the 0.2 micron filter on the submicron region particulates.

“There are qualitative tests to look at those things right now, but you have to do the full risk assessment,” he stressed. “Certainly filtering is going to reduce the risk. Will it eliminate it? I suspect not. But there is no zero-risk. I think we recognize that.”



What I am going to do today is to talk about the Division of Therapeutic Proteins’ current thinking on the process.  These things are not written in stone.  The intention is to have a robust discussion of the issues.  We have not talked throughout all of FDA, so this is not an entire FDA [perspective].  But I think many people within FDA, including the Office of Blood, generally have similar types of expectations.

As a general outline I would like to:

● give a little background about the subvisible particulates and immunogenicity.  If you don’t frame what the issues are, it is hard to have a discussion and do risk assessment.  So, I would like to frame what we know about these things and what we do not know

● then discuss the risk assessments in terms of the incidents, the frequency of occurrence that may be due to product characteristics or clinical characteristics, and then risk assessments regarding the severity of the consequences

● then some of the risk reduction control strategies

● and then finally some of our current recommendations.

Knows and Unknowns

First, what we know and what we don’t know about immunogenicity and sub-visible particulates:

For this discussion I would like to mention that I am talking about sub-visible particulates.  I am talking about the range from 0.1-100 microns.

What we do know?

We know that animal data have demonstrated that protein aggregates can greatly enhance the immune response to the protein of interest.  Various types of aggregates may do it.  We know aggregates are capable of doing it in animal models, and indeed we have clinical data indicating that protein aggregates also can enhance immune responses in patients.  A lot of information from growth hormone shows that aggregates can increase immune responses.

Amy Rosenberg will discuss later a lot of the data on human growth hormone and other data from protein therapeutics that indicates that these aggregates can play a role in enhancing immune responses (IPQ “In the News” Jan. 18).

Immune responses to therapeutic products are an undesired phenomenon – it is not something that we would really like for our products.  They may be expected.  By that I mean it may be already identified in package inserts that there is a certain level of antibodies or antibodies formed.  But there may also be adverse events that are unexpected. I think the case of red cell aplasia with some of the erythropoetins was an unexpected event when it first occurred, although now I would guess we would call it a rare but expected event.

The important thing is that these responses may impact safety and efficacy.  The development of nonresponsiveness to therapy is not uncommon to therapeutic proteins.  Many times the hint is that these are immune responses, but the cause is seldom thoroughly investigated….

The immune responses will be product- and indication-dependent.  There is no such thing as ‘one size fits all.’  Every product will be different.  Even for monoclonal antibodies – very similar structures, but the responses are different.

When we turn to the product characteristics, one of the highest risks for inducing neutralizing antibody appears to be associated with highly repetitive arrays of native proteins.  Those potentially include sub-visible particles with sizes ranging between 0.1-10 microns in our therapeutic proteins.  We know these particles are present in every therapeutic protein.  The content, in fact, in the smaller size range is much higher than the sub-visible content in what we test for in the USP <788> testing.

We also know that these can dramatically increase under certain conditions, as in stress.  The use of some piston pumps and rotary pumps can sometimes shed little nanoparticles of metal, and those can potentially allow for seeding of protein particulates.

We know that current testing methods employed by most manufacturers do not monitor all size ranges of aggregates.  They report different values and have difficulty distinguishing one protein from other types of particulates.

What do we know that we don’t know?

Some of the product/patient attributes that are responsible for undesired immune resposes are really unclear – we might say a lot of the product attributes are.  We do understand some of the factors that influence this – such things as foreign sequences, product-related impurities, route of administration, and dosing schedules can all impact immunogenicity.  But the impact of specific product attributes and how they synergize together with  respect to patient factors for any given product is not really clear.

In terms of sub-visible particles, the types and amounts of protein aggregates that may induce an immune response for our products are not clear.

One of the most important points that we hear all the time is there is really no definitive data indicating the low levels of protein sub-visible particulates that occur in therapeutic products impact immune responses.  We do know that there is a correlation with particulates where there are large numbers of them, but the low numbers that are typically observed…we don’t know that there are any immune responses for those…. There is some uncertainly about what those levels are and what those types are.

What do you do in the case of uncertainty?

In this case it is really difficult to directly assess the impact of specific attributes on clinical performance.  When you do a clinical trial, you have a product, and you have all these different attributes – you have heterogeneous attributes, so you are not looking at one specific thing.  It is very hard to inject into people a product that has a high content of sub-visible particulates as an experiment.  It is inappropriate.

But, you can also argue that the lack of direct scientific knowledge regarding a product attribute and its relationship to safety and efficacy does not reduce risk, it simply ignores it.

The totality of the data from various sources indicates there is a theoretical risk that therapeutic protein aggregates can enhance immunogenicity.  This risk may not translate into risk for your specific product, but in many cases there is going to be a considerable amount of uncertainty surrounding the impact of sub-visible particulates on the immunogenicity of a therapeutic protein.

What do you do in the case of uncertainty?  I think one of the things is that you document exactly how certain and uncertain you are about your product.  One of the ways to do that is to do risk assessment.

Assessing the Risks

Product characteristics:

In talking about the occurrence of an immune response, one of the important characteristics is the product characteristics.  We know homology to an endogenous protein is an important characteristic.  Foreign sequences are innately immunogenic because the body hasn’t been exposed to them, hasn’t developed tolerance to that protein, so the immune response is generally pretty energetic.

Protein aggregates, we know, can induce immune responses. What we don’t know are the types and amounts of those things that are required to do that.

Other product-related variants like deamidation or different forms of sialic acid also have implications that these are other product attributes that might enhance immunogenicity.

Of course, you do not know how any of these things synergize with each other, but process-related impurities – host cell proteins and DNAs that can act as adjuvants and microbe-associated molecular patterns, or as they have been called in the past, pathogen-associated – these things we know have some role in inducing immune responses.

Clinical characteristics:

Given the context of clinical uncertainty in sub-visible particulates, what can you do?  I think the risk is influenced by the probability of occurrence of these particulates and the detectability of these things.  In order to really understand the occurrence you must assess the relevant manufacturing and storage conditions that you are using that could lead to increased amounts of sub-visible particulates.  Of course, you have to use relevant analytical techniques to look at these things.

If you are not employing these methods designed to detect and characterize sub-visible particulates, you are exposing patients to unnecessary risk.  Ignoring a risk is not risk mitigation.

Whether monitoring is part of your development program or characterization activities are performed routinely should be based on your product characteristics and your risk evaluation.  You notice I put development and characterization. I think that, yes, you should do these characterizations as part of your development program, but continued characterization activities following manufacturing changes, I think, are also important.

Comments on an industry perspective on SVP monitoring:

What I have just stated seems a little bit at odds – it doesn’t seem we are on the same page with an industry perspective that was published recently by Singh et al. [JPS Vol. 99 (2010)].  I took a couple of the comments [from] there and would like to further expand on our response to some of these comments:

First of all, that these particles are qualified in clinical trials and presumably therefore do not need to be monitored post-approval.

They are qualified in a clinical trial in one sense, since the products are approved based on a favorable safety and efficacy profile of the product that you used in the clinical studies.  But if you do not characterize that clinical trial material for sub-visible particulates, then you are relying solely on the process to control these things.  They are at unknown levels, and they may or may not impact the immunogenicity profile of your product, and you do not know what happens when you make changes to manufacturing.

Another issue that was mentioned is that particulates are present in marketed products that remain safe and efficacious despite the lack of monitoring of small SVP less than 10 micron.

I think that misses a point inherent with approval.  Approval does not mean that the safety and efficacy profile is optimal.  Most protein products are immunogenic. If you look at the labels you will see they have immune responses – there are neutralizing antibodies for many of these products.  You look at the labels – non-responders that can develop over time for therapeutic proteins.

What we do say is that you have a favorable safety and efficacy product.  There can be continuous improvement.  You can potentially reduce the adverse events with products by manufacturing and careful control of the product.

Nor does this approach guarantee that sub-visible content is comparable following manufacturing changes without additional assessment and monitoring.  If you do not have a baseline of what was used in clinical studies, it is difficult to figure out what you do with a product post-manufacturing changes.

One of the arguments that can be made is ‘we have post-marketing adverse event reporting and we don’t see anything there.’  But this is not a sensitive tool for monitoring non-responders that may result from immunogenicity.  We have a lot of products that develop non-responders….  The baseline is so high it is hard to see things.

Even for novel, unexpected adverse events, like pure red cell aplasia, it was a laboratory in France that first identified that issue – not the adverse event reporting system.  That was a dramatic response.  Hopefully we have improved our surveillance programs to pick up these things, but the issue is that you should not rely on your surveillance program to detect these things.

I think a fundamental principle is that when a new risk is identified, that risk, as it pertains to a specific product, should be assessed to the extent possible.  As I said, this is fundamental.  And risk assessments should be performed on a periodic basis following approval, and any newly-identified risks should be mitigated in a timely fashion.

This is the same issue that came up in a recent PDA conference on viruses.  People have identified that bovine serum may contain VC virus.  That was known.  It was known that one manufacturer had a problem many years ago, and people had studied cows and found that this virus is prevalent.  Yet nobody went back and said ‘we are using this stuff, we are not looking at this, is this a problem for us?’  You have to look at the literature, keep up to date  and see where you have a risk that you have not identified and are not controlling appropriately.  It is all part of continuous improvement, which is a big initiative with FDA.

The third statement that I found was that sub-visible particles constitute sub-microgram levels of proteins that are well below what is detected or reported for other attributes.

It is clearly that it is not the relative sensitivity of the assay that is important. It is the understanding of the type and amounts of these sub-visible particles that are necessary to trigger an immune response.

I think the real intent is vaccine doses that lie in the microgram and milligram range, so these techniques are really more sensitive than necessary.  But I would argue that while sub-microgram levels of sub-visible particulates may be administered at one time, the amount necessary to actually trigger an immune response is really still not clear, and the impact on immunogenicity by multiple dosing is really unknown.

Under some conditions – either shaking or shedding or high-dose products – microgram levels of sub-visible particulates may be present, but undetected.  Size Exclusion HPLC (SE-HPLC) is not a sensitive method for looking at sub-visible particulates or large aggregates.  It is insensitive, in fact.  You will not see microgram quantities.  It is incapable of seeing that, just because of the amounts that you load onto those columns.

Risk of occurrence of an immune response:

I want to talk a little bit about the occurrence of immune response.

[Lilly Senior Research Scientist] Holly Smith gave a very nice presentation at the Breckinridge, Colorado meeting in the summer.  I took most of the next slides from her, since she did such a nice job of presenting these.

You can see one of the clinical characteristics, dosing frequency, from single to acute to chronic to intermediate – all increasing, presumably, the risk of an immune response:

Click to Enlarge

Click to Enlarge

Similar for dose concentration, route of administration, clearance, patient immune status, whether the product itself is an immune modulator – and the probability then, from low to unknown to high.

One of the things that was not on the [Smith] slides was the endogenous protein level.  If the body is exposed to high levels, then it develops tolerance. But when trace amounts of that protein are present, it may be silent to the immune system.  We think that may be the case for TPO – that the endogenous levels are so low that you don’t actually build tolerance to that protein.  So when you expose patients to higher levels of that protein, they can develop an immune response because they are not tolerized to that protein.

Risk of severity of response:

In terms of severity of the response:

Click to Enlarge

Is it an endogenous homolog? That is clearly one of the issues.  Is that homolog redundant or unique biology? And again we are talking about from either low severity or high severity in the red.  Impact on autoimmune – is it tolerizable?  Can you dose through any neutralizing response?  Is the intended disease life-threatening or not life-threatening? I guess that may be reversed.  I would think that life-threatening diseases are more important and more risky than non life-threatening diseases.  Treatment options, potential cross-linking, and impact from less serious to more serious.

I would add here – and I believe that Holly did – that these are always guarded, and that if you have a product where you get a neutralizing response, even though it may not knock out the endogenous homolog, it may prevent the efficacy of the product.  It is a serious consequence no matter what to the patients themselves.  We would like to avoid those things.  We would like to make our products as safe and as effective as possible.

Risk Reduction and Control Strategies

I think this is a major point:  If there is a theoretical risk of uncertain significance, in the absence of convincing data that the attribute is not critical, you should assume it is critical and needs to be controlled.

What ICH says about critical quality attributes is that they are controlled within an appropriate limit range or distribution to ensure the desired product quality.  That control does not necessarily mean that attribute needs to be specified.  Not all critical quality attributes are specified, although there is certainly a tendency for us to try to do that.

Action limits, for example, can be used for sub-visible particulates to allow for a more thorough investigation when an out-of-trend event is observed.  What is the nature of the particulate?  How much protein is present there?   [These] will inform your risk assessments and you can plug it in and say, ‘well, is this a risk or not, and what do we do with it?’

There are other methods and other approaches, too, if you are confident the attribute stays within appropriate limits.  Can other information you have allow you to predict what the levels are?  For example, if you know that light obscuration at 10 microns will predict what is going to occur at lower values, if you show a nice correlation between those different size ranges, then maybe you put an action limit on light obscuration at 10 microns, where you start to do an investigation and additional characterization of things that are in a smaller range because you know it may be a problem.  Action limits can be a way to go around and do some things.

The other thing is for much of our testing we have algorithms where if you get a certain result then you do additional testing.  Some of these methods are highly variable, so if you set a limit that if you see something you have to do additional sampling, that is an acceptable approach:  we use it for moisture where people pick ten vials and do a moisture test, and if they are above a certain range for some things, then they do additional numbers of vials to look at the moisture.  Then they use some algorithm to see what is appropriate.

There are always ways to handle assay variability, too.  One of the easiest ways is to take more samples.

Is the attribute adequately controlled by the process?  Validation, of course, is always an option.


For characterization:

To conduct a robust characterization of the sub-visible content of drug product at release, on stability and in use:

● The use of at least two orthogonal analytical techniques to quantitate the amount and sizes of sub-visible particulates….

● The use of multiple stress conditions to assess the propensity to form sub-visible particulates.

● All of this may involve the use of analytical techniques designed to identify the nature of the particulates and to quantitate the amount of the protein.  The tests that we use do not really tell you how much protein is there and, of course protein may be one of the driving factors.

For micron range testing:

What are the testing recommendations in the micron range?

I think it is pretty clear that particulates in this range can be detected and appropriately quantified by various analytical techniques.  We heard about light obscuration, micro-flow imaging, and coulter counting – all can be used to characterize and quantitate these.

There are issues with these, there are limitations and there are some problems.  These methods do measure different characteristics and therefore report different values.  I think that is OK.  [What is important] is whether these results correlate with each other or not.  If they correlate, then they are predictive of each other.  If they do not, then it is hard to predict what you are looking at and they are not predictive of what will happen in other types of aggregates.

Variability can be quite high.  We have seen 300% on some of these assays.  That is quite high and probably [means] you are not validating the assay for robustness as well as you should. We understand that these assays are going to give you some variability – more than the typical assays that you see for chemical measurements.  This variability can be built into any assessment of the product quality and any limits that you establish.

You have heard about sample handling and the issues with that, but assay robustness should be evaluated to control variability associated with that.

Again, you heard from Linda Nahri [Amgen Formulation & Analytical Resources] that silicone in pre-filled syringes can be problematic.  I think one of the ways to address that is [to use] micro-flow imaging which can distinguish somewhat protein from silicone particles [and] set the aspect ratio for more circular things so that you segregate out components that are circular versus non-circular ones.  Yes, you are going to lose some of the protein particulates that look like spheres, but you will still have more information about protein content than you would by any other method.

There are limitations.  No assay is perfect.  You should use the assays to the best of their ability to provide the most information for any risk assessment that you perform.

I think new mathematic algorithms probably could be used to better define these particulates.  I think even if you look at the protein particulates that look a little circular, they are not circular, and by eye you can clearly see that they are not.  I think you can get mathematics that will get you a better handle on that.  That is under development, I think.

For submicron range testing:

We talked about micron range. What do we do about the sub-micron range?

Particulates in this range can be detected and assessed qualitatively by a number of techniques, but quantitative techniques are not currently available.  SE-HPLC with a light scanner and detector is able to see at least a signal there.  You can use other methods like field flow fractionation (FFF) with light scattering, even ICP looking for metal content.  If you have nanoparticles being shed from your rotary piston – those are nanoparticles.  You are not going to see any of that with a light obscuration test.

So, what tests are you going to look at to try to get a handle on what is going on at that small range?  I think a qualitative assessment of sub-micron particles has a place in development and product characterization, including characterization following changes in manufacturing processes.

Click to Enlarge

Just an example…by looking at size exclusion with either a UV detector or light scattering, you can see that the UV detector does not pick up the aggregate 1 and aggregate 2 that is picked up by the light scattering because it is just more sensitive.  I would argue that if you had a signal that was really large and not in trend with previous production that you might be worried about releasing that lot.

In fact, in probability studies we have been seeing these types of techniques being applied and seeing signals that are associated with a change in manufacturing.  Obviously more work needs to be done then to resolve what the issue is.  Why do you see that signal?  It is different.

Continuous process improvement:

Another thing I think that is an agency push is continuous process improvement.

Exploring the use of new technologies should be part of continuous product development, particularly in areas where we know there is an analytical gap. For example, the 0.1 to 1 micron range is clearly an analytical gap.

I think there are techniques coming on.  I think you heard Linda [Nahri] mention NanoSight, which is actually a nanoparticle tracking analysis that has been reported to quantitate particulates from 30 nanometers to 1 micron – certainly covering the range that we are talking about.  It could fill in that analytical gap.  There are questions about assay sensitivity for that method, but this should be evaluated for use in therapeutic proteins.  And in fact, academic labs are starting to report about this particular technology.  They are reporting favorable results – that it is a valuable technique.

I think certainly we would encourage industry to look at these instruments to better define the risk to your products.  As I said, if you are looking at nanoparticles, you are not seeing them in any of the tests that are currently employed.

For approval and post-approval:

What is the recommendation for approval and post-approval?

We have been suggesting that you conduct a robust risk assessment of the potential impact that sub-visible particulates have on the quality, safety and efficacy of your product, and propose a strategy that provides adequate control of perceived risks they may pose for your product as they may relate to safety and efficacy.

IPQ home page

©2021 IPQ Publications