1. FOCUS ON... QUALITY
QbD for Biologics
Learning from the Product Development and Realization
(A-MAb) Case Study and the FDA OBP Pilot Program
by Steve Kozlowski, Wassim Nashabeh, Mark Schenerman, Howard Anderson,
Ilse Blumentals, Kowid Ho, Rohin Mahtre, Barbara Rellahan, and Victor Vinci,
with Lorna McLeod
C
osponsored by CASSS (an case study and sponsors participating
international separation in the FDA Pilot Program provided
society) and the FDA, the detailed QbD examples to form the
23rd CMC Strategy Forum basis for workshop discussions. A
was held in Bethesda, MD, on 19–20 number of questions were presented as
July 2010. For the third time, this a basis for discussion, and they appear
forum explored the topic of quality by in bold throughout this text.
design (QbD) for biologics. The first
such forum was held in July 2007 and Critical Quality
focused on establishing a general Attributes (CQAs)
understanding of QbD terminology In assessing attribute criticality, to
and concepts. In July 2008, the second what extent is it appropriate to apply
discussed approaches for submission of prior knowledge from similar-class
QbD data and associated regulatory molecules to a new product? When is it
implications. Building on those appropriate to leverage company-
previous QbD forums, this third specific and literature information?
forum extended the discussion from Leveraging prior knowledge is
“what” to “how.” The program particularly valuable at the earliest
committee intended to cover detailed stages of development before you’ve
implementation strategies and had a chance to gain molecule-specific
practical key QbD elements that are www.photos.com
data in early development. Prior
readily achievable in the short term. knowledge of molecular structure at
In addition, this forum would Group. The companies involved were early stages is useful for highlighting
combine key learning from two Abbott Laboratories, Amgen, specific product variants you need to
important QbD industry–FDA Genentech, GlaxoSmithKline, Eli look for and targeting the types of
collaborations: the A-MAb Case Lilly and Company, MedImmune, analytical methodology required to
Study and the FDA OBP Pilot and Pfizer. To ensure free public assess them. As long as its strengths
Program. The pilot program is still in access and further promote the and weaknesses are understood,
its early stages but nonetheless industry-wide discussions that led to information is valuable wherever it
provides concrete examples of the its creation, they provided its case comes from.
types of exchange of ideas between study to CASSS and ISPE. Find it Keep in mind that, although
sponsors and regulators. The case online at www.casss.org/ general assumptions can be made
study on applying QbD principles in associations/9165/files/Case_Study_ about class-specific attributes (e.g.,
development of a monoclonal antibody Press_Release.pdf. MAb terminal heterogeneity),
represents the culmination of a two- This forum was set up as three inevitably some molecules will not
year effort by a consortium of workshops covering quality attributes, follow the rules. The value of general
biotechnology companies collectively design space (DS), and control assumptions depends on the depth
known as the CMC-Biotech Working strategies. Authors of the A-MAb that knowledge can reach — how
18 BioProcess International 10(8) S eptember 2012
2. specific it is to your particular There was also discussion about
The CMC Strategy
molecular structure/function. For which parts of a QbD submission
Forum Series
example, what about its glycoform constitute regulatory commitments
structure does or does not affect Fc The CMC Strategy Forum series and what can be handled through a
receptor binding? provides a venue for biotechnology and company’s pharmaceutical quality
Is the biotech industry still excited biological product discussion. These system (PQS). There is no definitive
meetings focus on relevant chemistry,
about QbD, or are anxiety and answer. Early and frequent
manufacturing, and controls (CMC)
frustration replacing excitement? consultations with regulators are
issues throughout the lifecycle of such
Instead of managing risk, are we products and thereby foster recommended, and “negotiations” with
becoming more risk-averse? Some collaborative technical and regulatory the agency are to be expected.
consensus was reached that QbD is a interaction. The forum committee In setting and justifying acceptable
good idea in theory, but there is still strives to share information with ranges for CQAs, what information is
work to be done in clarifying what it is regulatory agencies to assist them in required? When are preclinical data
and how it is best used. Although the merging good scientific and regulatory sufficient, and when are clinical data
idea is to have a DS within which practices. Outcomes of the forum required? The value of preclinical data
changes can be made without formally meetings are published in this peer- depends on the animal model used.
reporting them to regulators, it appears reviewed journal with the hope that Questions that need to be asked
they will help assure that
at present that more documentation regard its relevance to humans,
biopharmaceutical products
(rather than less) is probably needed. whether the ligand/target has the
manufactured in a regulated
As one regulator pointed out, “If we environment will continue to be safe same properties as in humans
had total trust in a DS, we wouldn’t and efficacious. The CMC Strategy (including PK and PD effects). How
need regulatory agencies.” Forum is organized by CASSS, an does the disease state in humans affect
It was generally agreed that we International Separation Science how you interpret and use the data?
need an adaptable way of assessing Society (formerly the California Do immunogenic responses in animals
reportability criteria with a common Separation Science Society), and is affect your evaluation? Although an
understanding of what needs to be cosponsored by the US Food and Drug advantage of preclinical testing is in
provided, both in a filing and in terms Administration (FDA). exposing animals to purified variants,
of changes. How much can be clinical data are still the gold standard
handled by a company’s quality pharmacokinetics (PK) or as long as patient variability
management system (QMS), or pharmacodynamics (PD) often considered. Extracting product from
pharmaceutical quality system (PQS) depends on a number of factors: e.g., serum samples is very valuable and
according to ICH Q10? How much the scope and significance of class- informative for PK.
documentation will ensure regulators’ specific knowledge and the availability However, the utility of clinical data
comfort level? of meaningful models. Other factors for PD depends on available markers
As far as enthusiasm goes, it was to consider are different dosing (e.g., increasing blood-cell levels are
noted that QbD needs to have regimens (e.g., intravenous or easier to measure than tumor size/s or
inherent value to a company to make subcutaneous), chronic or single overall survival). Again, although
it worthwhile. It is a good, progressive dosing, patients’ disease state general assumptions can be made (e.g.,
idea, but companies need to including whether patients are MAb terminal heterogeneity),
understand its value to them and see it immunosuppressed, and so forth. inevitably a case will arise with
making sense from both science and When changing a molecule’s molecule-specific differences, and
business perspectives to maintain their indication, you must revisit your CQA ranges for those will need to be
enthusiasm. Regulatory relief (one of risk assessment. justified. CQA ranges depend on
QbD’s original drivers) is still a future This question remains: At what manufacturing process capabilities,
prospect. point can we accept an attribute as patient populations, dose strategies,
How much additional molecule- noncritical for all class-specific and so on. It seems difficult to justify
specific information would be molecules? Regulators are at present a single range for a particular CQA
required to support an assessment reluctant to allow such an assumption across a whole class of molecules; only
based on prior knowledge? It is across the board, so justification is DNA and endotoxins seem to have
unlikely that the criticality of quality required case by case. One participant achieved that from a safety
attributes for a given molecule will be put it very succinctly: “Literature and perspective. However, it appears that
identical to that of another molecule. knowledge can be a wealth of data if the CQA risk-assessment tool now
So it is worthwhile in investigating the data are relevant to your used across the industry is seen as an
the unique aspects of a molecule to molecule.” Proving that literature is effective mechanism for incorporating
confirm assumptions about “class- relevant is important for the comfort prior knowledge. But “noncritical” or
specific” knowledge. Whether to of regulators and for ensuring that “less critical” QAs must still be
check all relevant attributes while your product is truly safe and considered in relation to CPPs and
looking at their effects on efficacious. their related control strategy with
20 BioProcess International 10(8) S eptember 2012
3. justification as to how they were Forum Cochairs As in previous QbD forums, there
considered (not forgotten). is still a good deal of uncertainty
Kowid Ho discussed how the Steve Kozlowski (director of the office of about terminology. ICH Q8R defines
biotech products at FDA/CDER in
European Union (EU) PAT team is QbD as “a systematic approach to
Bethesda, MD)
and is not implementing QbD development that begins with
concepts. One complication in Europe Wassim Nashabeh (global head of predefined objectives and emphasizes
technical regulatory policy and strategy
is the existence of two entities — the product and process understanding
at Genentech, a member of the Roche
Council of Europe and the European and process control, based on sound
Group, in South San Francisco, CA)
Union — which include different science and quality risk management.”
Mark Schenerman (vice president of
countries and do not always agree Some commenters consider that
analytical biochemistry at MedImmune
about issues related to drug definition to be too vague. In
in Gaithersburg, MD)
applications. The European Medicines addition, there is still a wide range of
Agency (EMA) represents the EU’s 27 working definitions of CQAs,
member nations and has taken on the greater numbers of smaller lots early in particularly at the earliest stages of
task of regulating how drugs can move development. Using lots enriched for a development. One company calls them
across national borders. So the answer specific variant early in development is “provisional” CQAs; other terms have
to “what is required” can vary another route toward understanding been discussed at previous forums. It
depending on which agency is involved. QA criticality. However, keep in mind is difficult to work within definitions
In setting and justifying acceptable whether you can justify patient you aren’t clear about.
CQA ranges, what information is exposure to potentially negative affects What aspects should be considered
required? How does stability fit in? resulting from levels of attributes when assessing interactions between
Stability must be considered for beyond what is normally designed into quality attributes? Can the interaction
comparing levels of attributes present dose-escalation studies. of noncritical attributes render them
at time zero with those that may How does a company broaden CQA critical? What information would be
change over time until expiry. Thus ranges based on safety and efficacy required to establish an absence of
patient exposure to end of expiry considerations? The assumption is interactions? You could use the DS of
material must be considered when that a “critical quality attribute” will fermentation, for example, to get an
establishing ranges (especially if used affect safety and efficacy. So you have idea of the true “DS” in relation to
in clinical studies). You also must to understand at what point an effect relative levels of QAs being produced
account for the appearance of new is relevant to patients (e.g., aggregates). before needing extensive interaction
attributes as a product degrades over Shed light on this question by studies of QAs that are not
time, which could necessitate adding leveraging preclinical and clinical realistically manufactured at different
quality attributes (and setting an serum samples for detecting variant levels by your process. Some attributes
appropriate ranges) to your clearance over time and for on their own may not appear critical
preliminary quantitative risk maximizing assessment of dose- but then interact and become critical,
assessment (PQRA) that are not ranging studies. Linking QA levels to although no specific examples were
present at time zero. immunogenicity, safety, and efficacy is mentioned. You can use forced
In setting and justifying acceptable challenging. Most current clinical degradation to create high levels of a
CQA ranges, what information is studies are not designed to link particular QA (e.g., oxidation) and
required? How do we reconcile the specific levels of attributes to patient examine its impact on another (e.g.,
value of establishing broader clinical outcomes. If possible, strategies for aggregation) to determine whether
exposure to product variants with the better correlating quality attributes their interaction is raising the
goals of product development, which and clinical data would be valuable. criticality level. It will require creating
continually drives toward Epitope mapping can be useful if a range of purified molecules with
comparability, consistency, and higher you see an immune response. By each QA at specified levels and testing
purity? Producing “more variable” introducing increased levels of them in animals or in vitro (if feasible)
product lots early in development can attributes into an appropriately to show a lack of impact on PK/PD
provide patient exposure information powered preclinical study, you can (and maybe safety). But that can be
and help you understand the impact of discover what levels have an effect. extremely costly and time consuming.
different levels of attributes on PK/ Relevant in vitro studies can show
PD (and maybe safety). But such limits that do or do not affect PK/PD Design Space
variability may not reflect commercial (e.g., Fc receptor binding, potency What types of information/data
process capability, especially at the assays, and so on). Data derived from can be used to define a DS (e.g.,
time of licensure, although it may be the clinic may lead to attempts to manufacturing data, design of
important for future changes and reduce the levels (or strengthen experiments, platform/prior
provides for an expanded CQA “DS.” control) of a given attribute if the link knowledge)? Manufacturing data from
There is, of course, an increase in cost of safety/efficacy to a QA can be made pilot-scale runs, engineering runs, and
and time associated with producing after the original risk assessment. full-scale clinical and/or commercial
22 BioProcess International 10(8) S eptember 2012
4. runs can be used in defining DS. commitment, but those are filed in the
Program Planning Committee
Design of experiments (DoE) and development section. Companies must
process characterization are also Howard Anderson (biologist in the consider ranges for parameters not
useful, as is platform or prior division of therapeutic proteins at FDA/ included in a DS. At some point, a
knowledge including both internal CDER in Bethesda, MD) process/parameter can be great enough
and published (external) data. Ilse Blumentals (director of global to have an impact, even if it is very
Formulation development will yield regulatory affairs at GlaxoSmithKline in extreme. Such ranges may be based on
useful data, as will stability and King of Prussia, PA) limits that have been tested beyond
comparability studies. All those Kowid Ho (quality assessor at AFSSAPS normal operations — “knowledge
product-related data should be in Saint Denis Cedex, France) space” — although justification of
included in assigning criticality to Rohin Mahtre (vice president of wider ranges may be based on prior
quality attributes. biopharmaceutical development at knowledge.
Literature should be used carefully. Biogen Idec, Inc. in Cambridge, MA) After much discussion about
In-house data are more valuable than Barbara Rellahan (product quality team handling non-CPPs, non-CQAs, and
peer-reviewed published literature leader in the division of monoclonal all things noncritical, one audience
because they can be backed up and antibodies at FDA/CDER in Bethesda, MD) member asked whether we truly
their history verified. The quality of Victor Vinci (director of purification believe in our risk-assessment tools —
data in published papers varies development and viral safety at Eli Lilly and if so, why are we so worried about
significantly. Conclusions based on & Company in Indianapolis, IN) what is not critical. However,
literature, in-house or otherwise, regulators are concerned about the
should be confirmed for a new universal definition may not be concept of a “limitless DS” and
molecule. Some assumptions can be possible. There is concern that the complete lack of control for elements
made safely, particularly for a platform definition of criticality depends heavily deemed noncritical. One commenter
product. But anything unexpected on the operating range studied. summed up the industry’s stance:
must be investigated. Changes beyond that particular Although the QbD paradigm provides
Should a DS consist of CPPs only, or operating range need to be managed for noncritical quality attributes,
should noncritical parameters be appropriately. nothing is left to chance. Everything
included? When might the latter be Someone commented that “DS is is well-controlled and monitored
appropriate? DS should include all not defined by CPPs alone. Assurance because that’s good science and
relevant parameters required for of quality defines DS. Regardless of common sense.
assurance of product quality, not just the risk assessment instruments, What actions should be taken if a
CPPs. If a DS were based solely on terms, or definitions you use, your DS unit operation response is not as
CPPs, defining them would require a must provide an acceptable level of expected either at pilot or
much greater level of understanding. If assurance that it will produce safe, manufacturing scale? This may mean
you include some control of non-CPPs efficacious drug product — and that that prior knowledge of the function
— or include them somehow into the your QMS will adequately handle all or operation of a given unit and/or its
DS — then data requirements may be movements within the DS.” impact on the product is incorrect. It
lower. If the DS includes CPPs only, Someone else mentioned that to depends on the stage of development
then a thorough data package will be diminish and eventually eliminate at which this occurs. The earlier such
needed to convince regulators that you “endless negotiations” regulatory a deviation occurs, the more likely its
can ignore controls or inclusion of agencies, the industry must come to impact can be rectified easily. Late-
non-CPPs. But non-CPPs should still some common understandings of stage failures or unexpected results
be controlled in a manufacturing definitions, requirements, and so forth may require a more comprehensive
procedure; it is how they are — and we are not there yet. evaluation of assumptions and data on
monitored, what their ranges are, how Experience is the only way to get which DS (or process understanding)
deviations are dealt with, and so forth there, and companies willing to garner is based. In either case, all data
that will be different. Because each that experience are paving a road for relating to a unit operation should be
company can use different risk- the rest of the industry. reassessed in light of the failure.
acceptance profiles to define How should companies handle Depending on those results, other unit
criticality, it will be difficult for parameters that are not included in operations, risk assessments, or process
regulators to accept a risk assessment the DS? Do we apply an infinite range? quality attribute (PQA) assessments
without in-depth review. Parameters not included in a DS might need revisiting.
It is still unclear how to should be controlled within the overall DS, many forum participants
differentiate between a statistically quality system. Examples include stressed, is an iterative process. It is
significant CQA effect from a manufacturing parameters (MPs), bound to change as more data are
practically significant impact. That process monitoring, change control collected and the knowledge space
determination currently appears to be assessments, risk assessments, and so increases. It is desirable to identify
in the eye of the beholder, and a forth. That’s not a regulatory necessary changes early in the process,
S eptember 2012 10(8) BioProcess International 23
5. of course, but it is possible that justification of small-scale–model there should be a continuous process
situations will occur such as the failure qualification against large scale. The verification protocol, change
of a unit operation at pilot or DS description applies only to the area management protocol or stability
manufacturing scale. At a minimum, in which a CQA is affected. You protocols. But you do need to
all data then would have to be should describe the linking of demonstrate that your DS model is
reassessed. individual steps across your process to not affected by a particular change.
How might a DS change across the ensure CQA control. One person asked how regulators
life cycle of a product? What types of It is still unclear exactly what deal differently with a “regular BLA”
new information could identify a new parameters to include in a filing (the compared with one based on QbD.
DS limit? Knowledge gained over time CPP and non-CPP argument) and Regulators said that they are still
during development can influence how much detail: Should non-CPP figuring that out. So far, they are
assumptions or back up existing data limits be tested?. However, we do looking very closely at QbD
in modifying a DS — either know that process steps with DS are applications because they sometimes
expanding or contracting it. Processes part of license claims with parameter seem ambiguous, and regulators’ level
nearly always undergo change, and ranges and mathematical models. We of comfort requires close scrutiny. One
new or altered processes can provide don’t yet know whether to include criterion specifically mentioned is the
new data that influence the DS: e.g., graphical representations and/or data clarity of the CQA and CPP
comparability data, stability data, and summaries. We need to ensure a definitions used in a filing. ICH
testing at different limits/conditions. balance between more data required Q8-R2 defines both terms, and
Additional manufacturing, preclinical, and flexibility for change without regulators are most comfortable with
or clinical data could enhance product reporting — and discern data for sponsor definitions that hold closest to
knowledge, turning CQAs into non- filing from data to be available on those ICH definitions. However, Ron
CQAs or vice versa. Process/product inspection. Taticek pointed out in his
impact may become evident with more Your description of manufacturing presentation, “It is not clear how to
manufacturing experience at scale. and process controls should be filed in interpret the ICH definition of critical
How can DS modifications be filed Section 2.2. Again, there are still process parameters: A CPP is a
throughout the life cycle of a product? questions about what to include and parameter that has both a statistically
It depends on when the DS is initially where: CPPs, non-CPPs; CQAs, non- significant and a practical (nontrivial)
“fixed.” If changes are made between CQAs. What must be included in the impact on the CQAs.”
then and the license application, then DS description? How much detail Regulators will also look closely at
those changes would be described in needs to be included about input ranges and the strength of data used
the marketing application (MA), variables, process parameters, and to support them. Are noncritical
biologics license application (BLA), or QAs covered by DS and about input parameters still within the ranges you
other filling. Should changes occur material controls and process controls? actually studied? If not, how can you
after approval, then filing them should Should you include model be sure that they are still noncritical?
be related to the extent and type of representations, equations, and/or a How do you propose to handle
change (annual report, changes-being- combination of ranges? noncritical parameters and quality
effected, prior approval supplement, Control of materials (Section 2.3) attributes after approval? What do you
type II, or type I variations). This should include detailed input material propose to cover in your QMS, and
filing strategy can be preapproved in a controls and CQAs for starting what is reportable? Again, the
protocol as part of the market materials. Control of critical steps and consensus among regulators seems to
authorization and built into the intermediates (Section 2.4) should be, “It depends. . .” Constraining
quality system. A common include input controls. Development CPPs would be less cause for
understanding is needed — in the (Section 2.6) will need to include regulatory alarm than expanding them
United States and elsewhere — of development strategy, CQA and CPP but might still cause regulatory
what must be submitted in regards to selection, QRM, prior knowledge, concern.
description of the QMS and how that DoE, multi- and univariate analysis, How can movement at the edges of
will influence the need to file design- lot and process history, and a DS be justified/implemented (e.g.,
space modifications. comparability. Process validation and/ “adaptive” control strategy or
or evaluation (Section 2.5) should statistically justified)? Statistical limits
Regulatory or Submission Impact include evaluation of operating units, can be bound into a DS (e.g.,
How should the DS be described in a storage/hold times, column lifetime, statistical boundaries and CPKs) to
submission? Your DS description must compatibility, viral safety, and so provide a level of confidence when
provide justification of parameter forth; evaluation of DS, validation, approaching edges of DS. When at its
scoring from the risk assessments used and confirmation of consistency (in limits, the qualification of small-scale
to design process characterization process and end product); and models (with edges defined) is even
experiments, including data on how movement toward continuous process more essential. You could increase
decisions were made. It should include verification. It is unclear as to whether testing as you approach the DS edges
24 BioProcess International 10(8) S eptember 2012
6. to assure product quality. Not all minimum to keep the agency as “sterile” might be included in a
edges are equal; some may be a cliff, informed. If a deviation reveals that a license, but all the details of achieving
others just a gradual difference, so non-CPP is in fact a CPP, then the and maintaining sterility would not be
statistical limits can be applied as DS and other related systems (e.g., included. The intent of “sterility” is
appropriate. A QMS may treat risk, small-scale model qualification) met through environmental
different excursions differently would need revising through a QMS monitoring and personnel practices as
depending on their potential product change-control procedure. well as validation and testing. So the
impact. You could file a strategy What role does the quality system output (sterility) is a regulatory
describing how such excursions would play in approaching CPPs and non- commitment; it doesn’t describe every
be handled (more studies, based on CPPs regarding planned movement detail of how that is to be
existing knowledge, risk assessment, within the DS or approved protocol? A accomplished.
and so on) or how a QMS will deal QMS should be able to handle What is the role of the QMS in
with uncertainties associated with movement within an approved DS approaching CPPs and non-CPPs after
movement near the edges. through preapproved enhancements to approval? How should a system
A DS system is asymmetric. A such systems as change control or manage and document oversight of
change near the middle might have process monitoring that ensure the continuous monitoring process,
greater or less effect on the resulting appropriate documentation, process and how should process
product than the same change near control, and product monitoring to improvements or optimization be
the boundaries. Although the prevent shifts in process capability or implemented and communicated to
assumption at filing is that a sponsor product quality. Obviously the level of the agency? A QMS can be enhanced
knows the CQAs for a given product, change management will be different to include improved process
uncertainty remains. Could the for non-CPPs than for CPPs as far as monitoring (e.g., holistic monthly
sponsor be missing “the rest of the how the system handles movement product review), statistical trending,
iceberg?” Negotiations with regulators (level of assessment, testing data and appropriate actions should trends
should be expected with a QbD filing required, postchange monitoring, be found. Such enhancements can be
until their comfort level has increased reportability, and so on). A non-CPP filed. The management of process
with the process and a sponsor’s movement beyond the range that improvement filings can be predefined
ability to work within it. defined its criticality would require as part of change control depending
enhanced scrutiny. Perhaps a defined on the level of change; they can also
QMS and Life-Cycle Implications limit to movement within a range be filed (e.g., as part of a change
What is the role of a QMS in (e.g., 50%) would be a compromise to protocol). But could Section 2.2
approaching critical and noncritical allowing totally free movement. include a commitment to update DS
process parameters, especially in Aspects of the QMS enhancements equations, for example, through an
regard to deviations or excursions? A required can be filed whereas others annual product review (APR)?
QMS change-management program is are made available on inspection. We discussed a number of
essential to assure both a Forum participants brought up a questions, including what level of
manufacturer and regulators that number of specific testing methods changes within acceptable ranges
changes within a DS will be dealt and discussed their desired frequency, might require reporting. The guidance
with appropriately and may not have specificity, and other questions. ICH indicates that “nontrivial, significant,
to be reported (or can be reported in a Q1D (Bracketing and Matrixing and impactful” changes should be
reduced category). Deviations and Designs for Stability Testing of New reported, which industry considers too
excursions should be dealt with Drug Substances and Products) makes vague. So questions remain. One
normally, with enhancements required clear that many factors need to be person suggested that such changes
to ensure adherence to a DS and/or taken into account when designing might be included in Section 2.5.
expanded change protocols (eCPs). complex testing strategies. The
The effect of a deviation on a information necessary for regulators to Control Strategy,
non-CPP may not require the same accept such approaches in designing a Life-Cycle Management
level of investigation as deviation to a QbD control strategy remains unclear. How would control strategies look
CPP depending on the nature of the Again, the industry generally different for traditional and QbD
deviation (e.g., within the knowledge agrees that non-CPPs and non-CQAs submissions? A QbD control strategy
space). This is not too different from will be controlled within a QMS. is based on a holistic, comprehensive
the current system of going within or Questions remain as to what becomes assessment of the criticality of quality
beyond validation limits. part of the regulatory commitment attributes, linking that to how they
Deviations that require DS revision and what does not — and thus what affect a process and defining process
(either shrinking or expanding) may requires a report to the agency and controls and product testing to assure
require some sort of a filing (level what does not. quality, safety, and efficacy. The
determined according to the type of Someone pointed out that under strategy includes risk assessments,
change) to “reapprove” it or at a the current paradigm, attributes such prior knowledge, and enhanced
26 BioProcess International 10(8) S eptember 2012
7. molecule and process understanding to immunogenicity operating space for a
Permanent Advisory
leverage preclinical and clinical data vaccine? In determining a vaccine’s
Committee for These Forums
with testing capabilities. So in-process immunogenicity operating space, you
controls, specifications (product and Siddharth Advant (Imclone) need to understand how the molecular
raw materials), and stability programs John Dougherty (Eli Lilly and Company) fragments and three-dimensional
will be based on criticality of quality Christopher Joneckis (CBER, FDA) structure truly affect the immune
attributes and probably be more system — e.g., stimulating only what
Rohin Mhatre (Biogen Idec Inc.)
streamlined, with fewer items (or we want to because we want a natural,
Anthony Mire-Sluis, chair (Amgen, Inc.)
fewer with high stringency) than a protective immune response. That
traditional approach. Wassim Nashabeh (Genentech, a may require designing additional
A QbD control strategy should Member of the Roche Group) studies to further examine how the
consider how unit operations affect Anthony Ridgway (Health Canada) product works. You may need to go
product across the manufacturing Nadine Ritter (Biologics Consulting beyond the traditional potency assay
process (and interactions among those Group, Inc.) to better characterize and predict
operations). A QbD control strategy Mark Schenerman (MedImmune) response. Immune response is
moves control to the process for Keith Webber (CDER, FDA) certainly a biomarker for vaccines, but
delivering high-quality product — it may not reflect efficacy. An
rather than testing quality into a understanding of patients’ responses to
product. This control strategy also however, it is up to a sponsor (once its a vaccine is also important. It seems
includes the concepts of continuous product has been approved) to decide clear that QbD can be applied to
verification (e.g., increased whether and when a movement within vaccines and that it is important to
multivariate analysis) and continuous a DS should be reported. But the know how to manufacture the product
improvement. This is the life-cycle agency is uncomfortable with that and and how it works. Ensuring a
approach. The strategy would will request reports when inspectors continuous supply for vaccines is no
inevitably include more data and deem it necessary. So a clear and different than for any other product.
justification in process understandable guidance is still What studies would be needed to
characterization, process control, and needed; so far, Q11 does not appear to justify an “immunogenicity operating
justification of specifications sections be it. Someone asked whether and space” for a therapeutic protein, for
of a filing. A QbD control strategy how it might be rewritten to provide which immunogenicity is undesirable?
also needs to deal with different levels useful guidance for both regulators First and foremost it is necessary to
of uncertainty for a DS. and industry. understand what actually causes
How would parameters that are What additional considerations — immunogenicity for a particular
unspecified in the license be handled, beyond criticality of a given attribute product. You can use epitope mapping
and what is the agency’s involvement? — factor into control strategy of antibodies to identify where in a
Unspecified parameters such as development? An attribute that molecule they bind. It can also be
process monitoring, change control, indicates process consistency (e.g., useful to monitor which lot of material
and noncriticals should be handled by glycosylation) but cannot be easily each patient gets and to control the
a QMS. How that system deals with measured through another parameter levels of quality attributes those lots
those parameters (noncritical process may need to be considered as part of get — and take into account patient-
parameters, inputs and outputs, and process monitoring or on specific responses (e.g., major
quality attributes) can be described in comparability, but not necessarily in histocompatibility complex
a filing or be made available on routine lot release or stability. An contributions). You can use preclinical
inspection. attribute that provides data about the or nonclinical studies to understand
We recommend an annual report as ability to supply patients (e.g., yield) the immunogenic potential of your
the best way to report such changes. would require some form of product (e.g., in silico or in vivo
One regulator asked, “If validation assessment (in-process). testing). A thorough understanding of
and DoE have been done and included Are the FDA’s eCPs and the product variants and process-related
in the filing, why does the agency European Union’s postapproval impurities is necessary, and all prior
need to see that again in an annual change management protocols knowledge could prove useful.
report?” Another stated that (PAMPs) the same? If not, what are the Several elements of QbD can be
movement within a DS is not a key differences? Because both eCPs applied across multiple product types
change, so there is no need to and PAMPs are very new, we don’t yet and associated systems. What
communicate that to the agency. know what their key features will be essential components can be applied
Another, however, pointed to or how they will be implemented. most generally? Some essential
“cascading uncertainty” at the edges components include planning and
and was of the opinion that changes QbD for Other Products design (e.g., molecule design,
toward those edges should be What challenges would come in equipment, and facility), execution
reported. According to guidelines, justifying the described (training, clear SOPs, streamlined
28 BioProcess International 10(8) S eptember 2012
8. processes/methods), monitoring (e.g., and specific regulatory risks including questions — such as defining CQPs,
statistical process control and patient population and indication. The CPPs, and DS — are beginning to get
multivariant analysis), continuous framework might also include material answered. The answers are becoming
improvement (e.g., corrective and demand: Will treatment be chronic or more consistent across projects and
preventative actions), and risk acute, high or low potency, and does it companies. Plenty of work remains to
assessments. Some elements of QbD involve high or low plant use? be done, but our progress is clear and
are becoming regulatory expectations. Global acceptance of QbD by inspiring. •
Forum participants mentioned that we regulators is one barrier to holistic
have been doing “QbD light” for years implementation. How are companies Steve Kozlowski is director of the office of
(e.g., process/product interactions, managing global filings? The two biotech products at FDA/CDER in Bethesda,
criticality of in process controls) and options expressed at the forum were MD; Wassim Nashabeh is global head of
that CGMP is an expectation (better essentially producing two different technical regulatory policy and strategy at
justification for reassessment of files or “file all data and wait for Genentech, a member of the Roche Group,
specifications and in-process controls, questions.” in South San Francisco, CA; Mark
risk assessments, good science, and What are the main concerns Schenerman is vice president of analytical
common sense). companies have in implementing biochemistry at MedImmune in
What elements of QbD appear to be QbD? Although it certainly benefits Gaithersburg, MD; Howard Anderson is a
the most difficult, costly, and/or time molecular design and process biologist in the division of therapeutic
proteins at FDA/CDER in Bethesda, MD;
consuming? Forum participants development, companies worry about
Ilse Blumentals is director of global
mentioned DoE, data accumulation, QbD’s effect on time and expense of regulatory affairs at GlaxoSmithKline in
and reporting of DS as potential developing products. They wonder King of Prussia, PA; Kowid Ho is a quality
barriers. Multiple risk assessments are whether QbD will really allow for assessor at AFSSAPS in Saint Denis Cedex,
clearly time-consuming. Another more rapid development if platform France; Rohin Mahtre is vice president of
barrier is developing extensive eCPs knowledge is applied. Some people are biopharmaceutical development at Biogen
rather than one-off comparability concerned that questions from Idec, Inc. in Cambridge, MA; Barbara
protocols. Because QbD is not globally regulators will increase as more data Rellahan is product quality team leader in
accepted, using it can lead to different are provided in filings. Companies the division of monoclonal antibodies at
filings in different jurisdictions, which wonder whether the cost and time of FDA/CDER in Bethesda, MD; Victor Vinci is
can be both costly and time QbD will be recognized from a cost of director of purification development and
viral safety at Eli Lilly & Company in
consuming. However, QbD is still goods perspective as opposed to a
Indianapolis, IN. Lorna McLeod is a
worthwhile. Many aspects of it are better process and product contributing editor for BioProcess
very cost effective — molecular design understanding. Will QbD have International
and CQA understanding, for example inherent value to the industry? Does it
— and can be applied by companies actually prevent multiple failures that
regardless of size, product, or process. might occur under the traditional Disclaimer
How are companies making approach? Are QbD-based products of The content of this manuscript reflects
decisions over how much (if any) QbD better quality than before? Concern discussions that occurred during the
will be applied to a particular product, continues to be focused around DS, CMC Strategy Forum workshop in
especially considering early phase, efforts to create it, and what addition to personal viewpoints and
late-phase, and licensed products? For regulatory and/or QMS relief will experiences of the authors. This
reasons such as attrition of molecules come (if any). document does not represent officially
through development and the cost and Apart from process and product, on sanctioned FDA policy or opinions and
time to carry out specific aspects of what other applications can QbD have should not be used in lieu of published
QbD, it may not be financially viable an impact? The answers to this FDA guidance documents, points-to
to apply to all molecules. We need to question include equipment design, consider documents, or direct
discussions with the agency.
establish a strategic framework to implementation, and execution;
guide the circumstances in which to facility design and utilities; raw
apply QbD. Such a framework might materials; containers; transport; and
include probability of success based on QMSs. To order reprints of this article, contact
knowledge of the biological pathway, Rhonda Brown (rhondab@fosterprinting.com)
availability of clinical data, and Advancing with Caution 1-800-382-0808. Download a low-resolution
market position. It might also include Although many questions remain, PDF online at www.bioprocessintl.com.
process/product complexity, taking collaborations between the FDA and
into account whether or not you are industry are bringing QbD ever closer
working with a platform, whether to realizing the potential of building
your drug product is lyophilized or quality into biopharmaceutical
liquid, whether you are working with products rather than controlling it
an established or a new technology, after development. Many early
S eptember 2012 10(8) BioProcess International 29