Quality by Design For Biopharmaceuticalsa Historical Review and Guide Forimplementation

Pharmaceutical
Review
Pharm. Bioprocess.
(2013) 1(1), 105–122
Quality by design for biopharmaceuticals:

a historical review and guide for
implementation
This article reviews the history of quality-by-design (QbD), how this concept has Phillip Elliott*1, Sam
been applied to biopharmaceuticals, and what can be expected from implementation Billingham1, Jingxiu Bi2 &
of QbD. Although QbD may lead to better design of products and manufacturing Hu Zhang2
processes, and offers the potential for reduced regulatory compliance costs, it will 1
Hospira Adelaide, 8 Dalgleish Street,
Thebarton, South Australia, 5031,
likely increase development costs. Process developers will require additional skills
Australia
and knowledge in the ‘quality disciplines’, which are not normally part of the training 2
School of Chemical Engineering,
of those in biopharmaceutical process development. A model for implementing QbD The University of Adelaide, North
in biopharmaceutical manufacture is proposed. The reader will gain an understanding Terrace, SA 5005, Australia
of how QbD principles have been applied to the development of biopharmaceuticals, *Author for correspondence:
E-mail: phillip.elliott@
as well as learning of the potential drawbacks of applying QbD tools indiscriminately.
hospira.com
Excellent examples of QbD applied to biopharmaceuticals in the literature will be
highlighted and suggested as the direction for future development in this area.
A more modern approach to the development »» A reduction in overall cost of manu

of pharmaceutical products and their subse- facturing;
quent manufacture has been advocated by the
US FDA and the International Conference »» Less waste;
on Harmonization (ICH). This approach has »» Faster regulatory approval;
been termed ‘quality-by-design’ (QbD) and is
defined as “a systematic approach to development »» Enabling continuous improvement;
that begins with predefined objectives and em-
»» Providing a better understanding of
phasizes product and process understanding and
processes and a better business model.
process control, based on sound science and qual-
ity risk management” [101] . There is much litera- The key tools of QbD are incorporation of
ture promoting the proposed benefits of QbD prior knowledge, the use of statistically de-
[1,2] ; the benefits claimed for QbD include [3] : signed experiments, risk analysis and knowl-
edge management. The intent of QbD is to
»» Better design of product;
encourage pharmaceutical companies to de-
»» Fewer problems in manufacturing; velop sufficient understanding of their prod-
ucts and manufacturing processes; ensure that
»» A reduction in the number of supplements their processes are robust; and, demonstrate
required for post-market changes; this enhanced understanding to the pharma
»» Understanding and mitigation of risk; ceutical regulatory agencies. Regulatory agen-
cies have in turn suggested that demonstra-
»» Allowing for the implementation of tion of this ‘enhanced knowledge’ could allow
new manufacturing technology without for a more flexible regulatory approach [101] .
regulatory scrutiny; For example, if a pharmaceutical product is
future science group 10.4155/PBP.13.6 © 2013 Future Science Ltd ISSN 2048-9145 105
Review Elliott, Billingham, Bi & Zhang
Key Terms approved for sale on the basis of a microbiological property or characteristic of the product
Risk analysis: Estimation of the QbD application that includes a pro- that should be within an appropriate limit, range or dis-
risk associated with identified posed design space, it may be possible tribution to ensure the desired product quality” [101] . From
hazards. In a pharmaceutical to make changes to the manufactur- the CQAs identified, the product design space can be
context, this term is often
used interchangeably with risk
ing process after the product has been determined, that is, specifications for in-process, drug
evaluation – the comparison of approved for sale, without the need substance and drug-product attributes [4] . The sum of
the estimated risk to given risk to go through an expensive ‘post-ap- acceptable variability in each of these attributes defines
criteria using a quantitative or proval change process’ with the phar- the overall product design space. The process design
qualitative scale to determine the
significance of the risk.
maceutical regulator. However, this space can then be determined using risk analysis and
level of regulatory flexibility is yet to design of experiment (DoE) techniques to determine the
Knowledge management:
Systematic approach to be realised. It is hoped that the QbD relationship of process parameter to the CQAs of the
collecting, analyzing, storing, and approach will improve product qual- final product.
disseminating information related ity and reduce regulatory compliance It should be noted that drug regulatory authorities
to products, processes and costs for pharmaceutical manufactur- generally base their decisions on three criteria – quality,
components.
ers. The concept promotes industry’s efficacy and safety. However, they take no account of
Design space: Multidimensional
combination and interaction of
understanding of the product and the actual or potential cost of the product to the com-
input variables (e.g., material manufacturing process starting with pany. Therefore, in addition to the need to be compli-
attributes) and process product development, with the aim ant with regulatory requirements, it is important for
parameters that have been of building quality in from the start a company to examine the value for money it obtains
demonstrated to provide
assurance of quality.
rather than trying to test quality of from process development, acquisition of knowledge
the product during manufacture. and QbD in general, and deploy these approaches in
Critical quality attributes:
Physical, chemical, biological Under the concept of QbD, when the most cost-effective manner [5] . A 2012 survey of in-
or microbiological property or designing and developing a product, dustry participants found that approximately 20% of
characteristic of the product that a company needs to define the de- respondents thought that lack of cost–effectiveness was
should be within an appropriate sired product performance and iden- the biggest hindrance to adoption of QbD in biologics
limit, range or distribution to
ensure the desired product tify critical quality attributes (CQAs). manufacturing, with a similar number citing fears of
quality. On the basis of this information, the regulatory delays as the biggest hindrance [102] .
Target product profile: Summary company then designs the product In this article, the historical background of QbD is
of the quality characteristics of formulation and manufacturing pro- first reviewed, and then the implementation of QbD
a drug product that ideally will cess to meet those product attributes. in biopharmaceutical manufacturing unit operations is
be achieved to ensure that the
desired quality, and thus the
Ideally, it is hoped this will lead to examined. Finally, an approach to implementation of
safety and efficacy, of a drug understanding of the impact of raw QbD is suggested and QbD is critiqued.
product is realized. Also referred material and equipment attributes
to as the Quality Target Product and manufacturing process param- Historical background
Profile.
eters on the CQAs, and identification In the 1990s, the FDA’s focus shifted from regulating
Drug product: Pharmaceutical and control of sources of variability. individual products to regulating the biotechnology in-
product type that contains a
drug substance, generally in As a result of all of this knowledge, dustry as a whole [6] . The 1997 FDA Modernization
association with excipients. Also a company can continually moni- Act established a new approach to reporting manu-
referred to as the dosage form or tor and update its manufacturing facturing changes, with the intent of minimizing the
finished product.
process to assure consistent product differences between applications for biologics and for
Drug substance: Active quality [4] . drug approval, this act was later transposed into guid-
pharmaceutical agent which is
subsequently formulated with
An important part of QbD is ance documents [103–105] . The changes added more and
excipients to produce the “drug to identify the target product profile more requirements for industry, resulting in increased
product”. (TPP) of the intended drug product. review times. By the year 2000, the FDA realized that
Design of experiment: Use The TPP is defined as a “summary there were undesirable consequences of the regulatory
of statistically designed of the quality characteristics of a drug review process [4] as manufacturers had become wary of
experimental arrays to determine product that ideally will be achieved implementing new technologies since it was unknown
the effect of multiple variables
on an experimental system that to ensure that the desired quality, and how regulators would perceive such innovation. This in
take into account experimental thus the safety and efficacy, of a drug turn led to higher costs for pharmaceutical manufac-
variation and are able to product is realized” [101] . Once the ture due to the maintenance of wasteful and inefficient
determine both the effects of
TPP has been identified it is neces- manufacturing processes. In many cases, the FDA at-
each variable alone and the
combined effect (interaction) of sary to identify the CQAs of the in- tributed these high costs to low manufacturing efficien-
multiple variables. tended product, with CQAs defined cies and the difficulty of implementing manufacturing
as “a physical, chemical, biological or changes [4] . In addition, the pharmaceutical industry
106 Pharm. Bioprocess. (2013) 1(1) future science group

Quality-by-design for biopharmaceuticals Review
had been accused of under-performing in manufactur- cal product manufacturing. The FDA Key Term
ing innovation by the business community: “Even as is stated, perhaps hopefully, that phar- Process analytical technology:
it invents futuristic new drugs, its manufacturing tech- maceutical manufacturing was evolv- System for designing, analyzing
niques lag behind those of potato-chip and laundry-soap ing from an art to a science- and en- and controlling manufacturing
through timely measurements
makers” [7] . gineering-based activity. It was hoped
(i.e., during processing) of
Due to concerns over the state of manufacturing, that application of this enhanced critical quality and performance
FDA oversight of firms increased. One result of this science and engineering knowledge attributes of raw and in-process
more stringent regulatory oversight was a dramatic in- in regulatory decision making, estab- materials and processes, with
the goal of ensuring final product
crease in the number of manufacturing supplements lishment of specifications, and evalu- quality.
to applications (these are approvals required from the ation of manufacturing processes
FDA for a manufacturer to vary its process from that would improve the efficiency and
contained in the documentation already filed with effectiveness of both manufacturing and regulatory de-
the FDA). In 2007, the FDA received a total of 5000 cision making. In this report the FDA identified a ‘risk-
supplements for new drug applications (NDAs) and based orientation’ as one of the driving principles of the
biological license applications and abbreviated new cGMP initiative with efficient risk management as the
drug applications (ANDAs) [4] . Considering that each primary way to make the most effective use of FDA
of these supplements costs approximately $250,000 in resources. A further guidance on process analytical tech-
direct costs alone [5] , this increased regulatory oversight nology (PAT) was released as part of the ‘cGMPs for the
caused significant costs for the pharmaceutical industry 21st Century’ initiative, which hoped to encourage the
(approximately $1.25 billion in 2007 alone based on adoption of more modern and flexible manufacturing
these estimates). technology in the pharmaceutical industry [106] .
cGMPs for the 21st century »» ICH Q8: Pharmaceutical Development

A 2-year initiative, ‘Pharmaceutical cGMPs for the 21st Further codification of the QbD concept came with
Century: A Risk-Based Approach’ was launched by the release of the ICH Q8 guideline ‘Pharmaceutical
the FDA in August 2002. This initiative was intended Development’ [101] in November 2004. This guideline
to modernize the FDA’s regulation of pharmaceutical reached ‘step 4’ – recommendation for adoption by the
quality for veterinary and human drugs and selected regulatory agencies party to the ICH – in November
human biological products such as vaccines. The FDA 2005. A further annex to the guideline, intended to
acknowledged that the new strategy was required to al- clarify the concepts in the original guideline, was re-
leviate concern among manufacturers that innovation leased for public consultation in November 2007 and
in manufacturing and quality assurance would result in reached step 4 in November 2008 (the original guide-
‘regulatory impasse’ [106] – effectively, the cost of gain- line and the annex have subsequently been combined
ing approval for innovations became so high that in- into a single document).
novation in manufacturing was almost completely dis- It is in the ICH Q8 annex that QbD is explicitly
couraged. As part of this initiative, the pharmaceutical, defined as, “a systematic approach to development that
as well as the chemistry, manufacturing and controls begins with predefined objectives and emphasizes product
regulatory programs were evaluated [107] . and process understanding and process control, based on
The final report of this initiative was released in Sep- sound science and quality risk management” [101] .
tember 2004 [107] . In this report, the FDA stated that Two other important terms for discussing QbD
the guiding principles of its efforts to modernize the were also defined in ICH Q8; Design Space and PAT.
regulation of pharmaceutical manufacturing were: Design space is defined as “the multidimensional com-
bination and interaction of input variables (e.g., mate-
»» Risk-based orientation;
rial attributes) and process parameters that have been
»» Science-based policies and standards; demonstrated to provide assurance of quality”. Accord-
ing to ICH Q8, working within the design space is
»» Integrated quality systems orientation; not considered as a change as it has been demonstrated
to have no impact on quality. Movement out of the
»» International cooperation; design space would be considered to be a change and
»» Strong public health protection. would normally initiate a regulatory post-approval
change process. Based on this guideline, design space
Importantly, in this report the FDA acknowledged was to be proposed by the applicant and would be
that its primary focus remained the same – to minimize subject to regulatory assessment and approval. PAT
the risks to public health associated with pharmaceuti- was also defined in ICH Q8 as “a system for designing,
future science group www.future-science.com 107

Key Term analyzing, and controlling manufacturing change and process improvement. The guide-
Robustness: Ability of a process turing through timely measurements line suggests that changes during development should
to reliably produce a product (i.e., during processing) of critical be looked upon as opportunities to gain additional
of the intended quality over a quality and performance attributes of knowledge and further support establishment of the
variety of operating conditions, at
different scales or with different
raw and in-process materials and pro- design space [101] .
equipment. cesses with the goal of ensuring final
product quality”. »» ICH Q9: Quality Risk Management
It should be emphasized that the ICH Q9 ‘Quality Risk Management’ [109] was released
ICH Q8 guideline provides guidance on the suggested at approximately the same time as ICH Q8 and ICH
contents of the Pharmaceutical Development section Q10, and needs to be considered as part of the over-
of the Common Technical Document. This section of arching QbD guidance released by regulatory agencies.
regulatory submissions relates to the manufacture of The purpose of ICH Q9 was to offer a systematic ap-
the ‘drug product’ – this is a very specific term relat- proach to quality risk management. It provided guid-
ing to the product that will actually be administered ance on the principles and some of the tools of quality
to the patient. This is in contrast to ‘drug substance’ risk management for use by both regulators and indus-
or ‘bulk material’ which are the terms usually given to try in managing drug substances and drug products.
the active pharmaceutical agent that is subsequently Importantly, it noted that use of quality risk manage-
formulated with excipients to produce the drug prod- ment can “facilitate, but does not obviate, industry’s ob-
uct [108] . This difference between drug product and ligation to comply with regulatory requirements and does
drug substance is important when considering how not replace appropriate communications between industry
and to what extent the original guidance was intended and regulators” [109] .
to apply QbD concepts and controls to pharmaceuti- Two important principles were outlined in this
cal and biopharmaceutical manufacture. This original document for the use of Quality Risk Management:
guideline did not relate to the manufacture of ‘drug
»» The evaluation of the risk to quality should be based
substance’ – the active pharmaceutical ingredient
on scientific knowledge and ultimately link to the
(API) before it is formulated for administration to the
protection of the patient;
patient. The complexity of unit operations for drug
product is generally less than that for drug substance »» The level of effort, formality and documentation
and it is appropriate that more control should be dem- of the quality risk management process should be
onstrated for the drug product which will actually be commensurate with the level of risk.
administered to humans.
The ICH Q8 guideline indicated areas where the These are important caveats that should be remem-
demonstration of greater understanding of pharma- bered as risk assessment is a process that can easily be
ceutical and manufacturing sciences could create a overused and lead to large amounts of unnecessary
basis for flexible regulatory approaches. The guideline documentation.
emphasized that more flexible regulatory approaches In Annex 1 to ICH Q9 the following tools are sug-
could be achieved if the applicant could demonstrate gested for risk management in the pharmaceutical
an ‘enhanced knowledge’ of product performance industry:
over a range of material attributes, manufacturing »» Flow charts;
process options and process parameters. The methods
suggested to achieve this enhanced knowledge were »» Check sheets;
formal experimental designs or DoE studies, PAT
and prior knowledge. Also suggested was the use of »» Process mapping;
quality risk management principles to prioritize ad-
ditional studies to collect such knowledge. ICH Q8 »» Cause and effect diagrams;
emphasized that it is the level of knowledge gained
and not the volume of data generated that would »» Failure mode effects analysis (FMEA);
lead to more favorable consideration by the regula-
»» Failure mode effects and criticality analysis;
tory bodies. A further suggestion was that applicant
companies could assess the robustness of the manu- »» Fault tree analysis;
facturing process, the ability of the process to reli-
ably produce a product of the intended quality over a »» Hazard analysis and critical control points;
variety of operating conditions, at different scales or
with different equipment, to support future manufac- »» Hazard operability analysis;

»» Preliminary hazard analysis; ucts in the development stage as in the commercial or

discontinuation phases of a product’s life cycle. It was
»» Risk ranking and filtering; the regulators’ hope that adoption of ICH Q10 should
“facilitate innovation and continual improvement and
»» Various statistical tools:
strengthen the link between pharmaceutical development
–– Acceptance control charts; and manufacturing activities”. Knowledge Management
and Quality Risk Management were cited as ‘enablers’
–– DoE; of this innovation and continual improvement. While
movement within a registered design space would not
–– Histograms;
require regulatory approval, the change should still be
–– Pareto charts; evaluated and documented by the company’s change
management system.
–– Process capability analysis.
»» ICH Q11: Development & Manufacture of Drug
While acknowledging that the selection of quality Substances
risk management tools is dependent on specific facts Dealing with the manufacture of Drug Substances,
and circumstances, Annex 2 to ICH Q9 suggested ar- ICH Q11 ‘Development and Manufacture of Drug
eas to which quality risk management tools could be Substances (Chemical and Biotechnological/Biological
applied by a pharmaceutical company, ranging across entities)’ was released for public consultation in May
all operational areas from quality management to facili- 2011 and reached step 4 in May 2012 [111] . Impor-
ties maintenance and even final packaging and labeling. tantly for biological and biotechnological products this
Of particular relevance to this review were the poten- guideline stated that most of the CQAs of a biologi-
tial applications to the development phase of pharma cally derived drug product are associated with the drug
ceuticals suggested by the ICH. Specifically, application substance and, thus, are a direct result of the design of
of Quality Risk Management techniques was suggested the drug substance or its manufacturing process. ICH
to assess the critical attributes of raw materials, APIs, Q11 reiterates the commitment to QbD principles in
excipients and packaging materials, as well as to deter- ICH Q8 and provides examples of how this process can
mine the critical process parameters for a manufactur- be applied to drug substance manufacture. It then goes
ing process. Other areas suggested in development were on to suggest where the data produced by QbD stud-
to assess the need for additional studies (e.g., bioequiva- ies and risk assessments can be located in the Common
lence and stability) in technology transfer and scale-up Technical Document format.
and to the reduction of variability in quality attributes. While most of ICH Q11 is concerned with identify-
ing what data should be presented in each section of the
»» ICH Q10: Pharmaceutical Quality System Common Technical Document, the appendices give
ICH Q10 reached ‘step 4’ in 2008 and described some useful examples of the use of DoE experiments
a model of an effective quality system for a pharma to establish the design space for different unit opera-
ceutical company. This model was intended to comple- tions, both for small molecules (chemical entities) and
ment ICH Q8 and Q9 [110] and defines the ICH expec- biological products.
tations for management responsibilities in a pharma- Of note is the fact that in these examples, when
ceutical company. The Pharmaceutical Quality System more than one CQA is affected by or dependent on
described had four key elements: a unit operation, the effective design space in which
acceptable product is produced becomes smaller
»» A process performance and product quality
(Figure 1) . This is likely to be a general result, especially
monitoring system;
so when more than one variable controls the output of
»» A corrective action and preventive action system; a unit operation and the corresponding design space,
therefore, has more than one dimension. For example,
»» A change management system; the design space shown in Figure 1, has two operat-
ing dimensions – conductivity and pH, but other unit
»» Management review of process performance and operations may have additional operating dimensions
product quality. such as linear flow velocity and product load that may
Importantly, the guideline emphasized that these el- also affect the CQAs of the product. In addition, the
ements should be applied in a manner ‘proportionate example given in ICH Q11 should be regarded as
and appropriate’ for each of the life cycle stages. That highly simplified due to the linear relationship shown
is, the same level of rigor is not appropriate for prod- between the variables and the product CQAs, which

process capability. Risk analysis is then performed to

identify and prioritize parameters for process character-
Conductivity
(mS/cm)
ization. DoE studies are formulated to examine the pro-
Viral clearance
cess design space, the studies are then carried out and
the results analyzed to determine the importance of the
Conductivity (mS/cm)
parameters examined. If necessary, the design space is

pH
then refined, based on the results of these initial studies.
Viral clearance The control strategy for the unit operation is then de-
Conductivity
(mS/cm)
fined through the process of risk assessment and based
on the importance of the CQAs affected by the unit
DNA operation and the capability of the unit operation. As
Design space
the biopharmaceutical product progresses toward regu-
pH
latory filing, the unit operation is validated and routine
DNA Host cell monitoring of the process is carried out.
Conductivity
proteins
(mS/cm)
A more general introduction to using FMEA to clas-

pH sify and prioritize the parameters in a biopharmaceuti-
Host cell
proteins
cal manufacturing process for further study has been
presented in [8] .
pH
A QbD case study on the production of a model
monoclonal antibody (A-mAb) has been developed by
Figure 1. Design space for an anion-exchange chromatography
the International Society for Pharmaceutical Engineer-
unit operation, demonstrating the reduction in the design space
ing [112] . This example includes sections on upstream
when more than one critical quality attribute is dependent on a
(mammalian cell culture), downstream and drug prod-
unit operation. Clear area in the small diagrams on the right illustrate
uct unit operations. It also includes sections on antici-
the safe design space for each of the critical quality attributes individu-
pated post-launch process changes – what is described
ally. Clear area in the large diagram on the left indicates the safe design
as movement within the design space.
space for all critical quality attributes affected by the unit operation.
The case study noted that the overall control strat-
Reproduced with permission from [111].
egy for a manufacturing process is based on the design
results in square or rectangular design spaces. In real- spaces of the individual unit operations – it is the sum
ity, these design spaces are generally complex curved of the individual control strategies that represent the
surfaces for each CQA with statistical confidence in- overall control strategy.
tervals for each surface. Interpretation of such design The case study gives examples of constructing the TPP
spaces is more complex than for the example shown for the A-mAb molecule, followed by identification and
in ICH Q11. risk assessment of quality attributes, leading to a rationale
for selecting the CQAs for the case study. Three types
Systematic review of QbD in manufacturing of tools for assessing the criticality of quality attributes
biological products were used (risk ranking, preliminary hazard analysis and
The European Federation of Pharmaceutical Industries safety assessment decision tree) and examples given of
and the European Medicines Agency have recognized how these tools are applied to the assessment of differ-
that each product is unique – developing any product ent quality attributes of the product (e.g., aggregation,
will require a ‘bespoke’ approach and, therefore, there glycosylation, deamidation, oxidation).
is no standard blueprint for applying QbD [5] . Various Various publications have reported the application of
methods of implementing QbD for entire processes QbD techniques to many of the specific areas and unit
have been suggested in the literature. One of the more operations of biopharmaceutical manufacture. The fol-
useful and easier to follow methods is shown in Figure 2 lowing section is a review of this literature as a reference
[4] . Initially, the TPP of the drug Product is identified. for process developers to illustrate the different approach-
Once the TPP has been identified the relevant CQAs es that have been taken and the differences between in-
are identified and prioritized through the process of dividual areas and unit operations that make it necessary
risk assessment. Based on the CQAs, the product de- to modify the way in which QbD is implemented. In
sign space is proposed – this is the sum of specifications reviewing QbD literature and preparing DoE studies,
for in-process, drug substance and drug product attri- careful attention should be paid to the definitions that
butes. These proposed specifications are ideally based will be used to classify each of the variables and the ranges
upon non-clinical studies, previous clinical experience within which each of the parameters can be controlled. A
with similar products, published literature and known well-recognized scheme is to classify each operating vari-

able as either ‘critical’, ‘key’ or ‘non-key’ (Table 1) using

the classification system produced by the Parenteral Drug Identify TPP
Association [9] . The selection of variables to study and the (QTPP)
definitions used to classify the criticality of those vari-
ables are pivotal in risk analysis and subsequent interpre-
tation of DoE studies. This is because the criticality defi- Risk
Identify CQAs assessment
nitions include aspects of both how well the variable can
be controlled within the defined range and how much of
an effect that variable has on a CQA within that defined
range. Many, if not all variables examined will affect the Risk
Define product
process to some degree, but the decision on whether to design space assessment
treat that variable as ‘critical’ depends on how it affects
the final product administered to the patient and how
easily the variable can be maintained within the desired
Define process Process
range. Insufficient attention to the definitions used for characterization
design space
variable classification can lead to haphazard overuse of
both DoE and FMEA techniques.
»» Cell culture Refine process

design space
FMEA has been used to identify and prioritize the pa-
rameters for process characterization in the fermenta-
tion of Pichia pastoris [10] and Escherichia coli [11] and
in mammalian cell culture producing mAb products in
Filing
Define control Risk

[12,13] . All of these examples start with the use of FMEA strategy assessment
to classify and prioritize variables for experimentation,
then move to DoE studies and development of the de-
sign space for the fermentation. The initial low reso- Process
Filing
lution screening experiments using the selected vari- validation

ables (parameters) are followed by more targeted and
thorough response-surface designs exploring the active
variables determined in the screening design. A recent
Filing
Process
study has also used quality risk management procedures modelling
to choose between alternative cell culture technologies
for the production of a mAb product [14] .
In a fermentation process using P. pastoris [10] and Figure 2. Suggested implementation plan for
methanol induction to express a soluble product, no Quality by design.
critical parameters were identified in the fermentation CQA: Critical quality attribute; TPP: Target product
as none of the variables examined affected final product profile; QTPP: Quality target product profile.
quality within the experimental ranges examined and Reproduced with permission from [4].
the process could be well controlled to remain within
these ranges. The fermentation was, therefore , found temperature, pH and dissolved oxygen, all of which af-
to have a wide design space. Key parameters (those that fected cell growth and titer. A ‘worst-case’ fermentation
affect process performance) identified in the study were was then performed to verify the design space, in which
Table 1. Parameter classification definitions.

Classification Definition
An adjustable parameter (variable) of the process that should be maintained within a narrow
Critical
range in order to not affect a product critical quality attribute
An adjustable parameter (variable) of the process that, when maintained within a narrow range,
Key
ensures operational flexibility
An adjustable parameter (variable) of the process that has been demonstrated to be well controlled
Non-key
within a wide range, although at extremes could have an impact on process performance
All input parameters outside this definition are non-critical.

all of the fermentation parameters were set to their most »» Parameter screening – identification of key and crit-
disadvantageous level in the design space. Although cell ical parameters for modeling DoE through the use
growth was affected, product yield and quality after pu- of Resolution III or IV screening DoE;
rification were acceptable, verifying the robustness of
the fermentation. »» Modeling DoE – Resolution V or response-surface
In comparison, for an E. coli fermentation [11] in designs to determine important interactions be-
which the product was deposited as insoluble inclusion tween variables and establish an empirical model
bodies, temperature, feed rate, pH and dissolved oxygen for the important outputs;
(and interactions between these variables) were shown »» Scale-down model verification – comparison of
to affect product quality. A worst-case combination of modeling DoE to manufacturing scale runs.
parameters was then used to produce inclusion bod-
ies. As the inclusion bodies produced in this worst-case A more mechanistic approach to ion-exchange chro-
could still be purified to produce acceptable product matography has been taken in [16,17] . Both of these
using the established downstream process, it was con- examples were built from the extensive theoretical
cluded that there were no critical operating parameters knowledge of ion-exchange chromatography available
within the ranges tested as none of them affected the in the literature to develop the design space for an ion-
final product. Monte Carlo simulation was then applied exchange unit operation and reduce the experimental
to parameters and ranges selected for the design space to and analytical requirements during process character-
set acceptance criteria for process validation. ization. Kaltenbrunner et al. also compared the use of
In mammalian cell culture expressing mAb prod- fractional factorial experiments to the theoretical model
ucts, the proportion of acidic variants has been found to validate the modeling approach [17] . Both approaches
to be particularly sensitive to culture conditions [12,13] . identified pH, ligand density and ionic strength at the
Temperature, pH and initial viable cell density were start of the gradient as the dominant variables, provid-
confirmed to affect the glycosylation profile of the an- ing assurance that the assumptions made to allow the
tibody product [12] , however, as the downstream puri- modeling were valid. A similar treatment by Mollerup
fication process was still able to produce an acceptable et al. was extended to provide a simulation of the result-
product from worst-case culture conditions, no critical ing chromatogram generated from fundamental protein
variables were detected in the cell culture conditions. parameters and the ionic strength and pH of the run-
Initial DoE screening and follow-up response-surface ning buffers [16] . This simulation was verified against
experimentation was used to produce predictive models small-scale and pilot-scale chromatograms and was able
that allowed optimization of the cell culture process and to confirm the root cause for a manufacturing problem
could be used to monitor the success of scale-up and that occurred at pilot scale.
later commercial operations [12] .
Tangential flow filtration
»» Downstream processing QbD principles have been combined with a mechanis-
Chromatography tic understanding of tangential flow filtration (TFF)
Examples of the use of FMEA applied to parameters operations and applied to the development of a TFF
of a chromatography unit operation are given in the unit operation [18] . Due to this approach, extensive
appendix of ICH Q11 [111] and in the A-Mab case DoE experiments were unnecessary, but more targeted
study [112] . experiments, guided by the mechanistic model, were
Cecchini gives an example of the application of QbD able to quickly generate a robust method of operation
techniques to the unit operations of cell harvest and and examine the economic implications of different op-
product capture, protein A capture chromatography, erating modes. This work addressed issues of membrane
hydrophobic interaction chromatography and anion selection, TFF design objectives, operating parameter
exchange chromatography [15] . As with the approach design and operating mode design. It especially empha-
taken in applying QbD to cell culture, all of these case sized the need to determine the effect of temperature
studies follow the sequence of: on the product, as the forces involved with TFF will
generate heat – heat increases flux through the filtra-
»» Determining acceptable ranges for product quality tion membrane, but biopharmaceutical products can
attributes; be especially sensitive to denaturation due to increased
temperature. This article is important in that it begins
»» Risk analysis to identify the most important unit from a mechanistic model of the technique of TFF,
operations to examine further to improve control of while also considering the applied aspects of the unit
the overall process; operation that will be important and unique to each

individual application. This approach is distinguished [23] as they provide examples of a one or two parameter
from more ‘naive’ DoE approaches that progress on a design space only, whereas the design space for many
purely empirical basis and vary operating parameters unit operations can have many more than two param-
without reference to mechanistic understanding of the eters or dimensions. In designing these unit operations,
unit operation. care should be taken to ensure that the design space is
This study does not consider the effect of electro- as simple as possible to enable the resultant design space
static interactions during ultrafiltration of biopharma- to be correspondingly simple. Compared with most of
ceuticals since it deals primarily with a drug substance the API processing operations, linkage is not as complex
process. Electrostatic interactions should be taken into in drug product processing primarily because the ranges
consideration in drug product TFF unit operations as are narrow and composition does not change as these
protein biopharmaceuticals are highly purified and be- are not purification steps. Therefore, there are not as
come the predominant charged species in solution. It many critical process parameters [24] .
is also important in the final drug product that exact Martin-Moe et al. have focused on drug product unit
concentrations of excipients are known and a target pH operations for mABs as these are amenable to platform
value is reached in the final solution. It has been shown development and have a well-established clinical history
that in the highly purified conditions of drug product [24] . This work used a risk ranking and filtering tool,
TFF operations, electrostatic interactions between the rather than FMEA, to identify the CQAs affected by
biopharmaceuticals, the membrane and charged excipi- drug product unit operations. The output of this tool
ents can interact to significantly alter the final pH and was used to determine which CQAs should be analyzed
excipient concentrations through the Donnan effect after each unit operation. A DoE study was then per-
[19–22] . Special attention is required during the develop- formed for formulation characterization with a risk-
ment of drug product TFF unit operations or any TFF ranking tool used to support selection of ranges and
unit operation in which the resultant buffer concentra- the type of study proposed to establish the design space
tion and pH needs to be controlled within a narrowly (multivariate vs univariate).
specified range. This work emphasized establishment of scale-down
models to allow multivariate experimentation without
Viral & sterile filtration excessive cost, and the importance of establishing the
The A-mAb case study provides a worked example of link between the large-scale process and the scale-down
applying QbD principles to viral filtration and sterile model. One alternative that has been presented is to use
filtration of drug product in which the combined use of at-scale surrogate models that are not as costly as use of
risk assessment, prior knowledge from similar product the actual product itself, but it is similarly important
experience and the development of a control strategy to establish the link between the surrogate model and
are illustrated [112] . For virus filtration two parameters, the actual unit operation with product. For this rea-
volumetric load and filtration pressure, were found to son, several companies now only produce biologics for
be important controls on virus removal and were clas- Phase III trials in the same equipment as commercial
sified as well-controlled critical process parameters. For product as they cannot otherwise assure comparability.
sterile filtration five parameters were found to be criti- As discussed above, TFF unit operations at the drug
cal, although all were easily controlled. Two of these pa- product stage of processing suffer from the extra com-
rameters, pre-run flush volume with drug product and plexity introduced by the interaction of the charged bio-
water for injection flush volume involved the prepara- pharmaceutical, charged excipients and residual charge
tion of the filter prior to use. The other three param- on the TFF membrane (Donnan effect) [19–22] . For this
eters identified as critical were the level of bioburden reason, particular attention should be given to the pKa’s
before filtration; flow rate per unit of membrane area; of the excipients, buffers and product and the extent of
and, filter area. diafiltration when developing TFF unit operations for
drug product.
»» Drug product
Both the A-mAb case study and ICH Q8 provide ex- »» Raw materials
amples of the use of QbD techniques to develop drug Lanan makes the very important point that raw ma-
product unit operations such as lyophilization, sterile terials have several distinguishing features from other
filtration, filling, stoppering and capping. While these aspects of biopharmaceutical manufacturing and this
unit operations are, by definition, high-risk due to their affects how they need to be handled in a QbD approach
proximity to the patient, they are generally relatively [25] . Unlike the operating parameters of a process, raw
simple and well characterized. The examples provided materials are not under the direct control of the bio-
by the ICH guidelines have been rightfully criticized pharmaceutical manufacturer – they are the products of

others suppliers and their processes. In addition, com- replace naturally derived materials as far as possible, or
plex or naturally derived raw materials can vary over explain, control and provide acceptance specifications
much longer timescales than are required for process for naturally occurring materials that must be retained
development. Therefore, even when efforts are made to in the process, are probably well spent and will con-
use a variety of raw material lots or batches in process tribute to overall control and QbD of cell culture. An
development, the total variation of the raw material may examination of the history of quality improvement in
not be captured during process development due to the the beer industry (the oldest biotechnology industry
much shorter timeframe. Furthermore, for operating based on cell culture) shows the extensive efforts that
parameters and conditions, less attention is given to de- have been exerted in controlling and reducing the vari-
tecting variables once a process has been transferred to ability of naturally derived raw materials in order to
manufacturing. Raw materials may require the constant improve productivity and consistency [28] . Biopharma-
detection of variables even after the process has been ceutical quality is already following the same pattern of
transferred to manufacturing or commercial scale. This development and the same approach to controlling raw
is because lot-to-lot, time-dependent changes to raw materials is likely to be fruitful in improving the quality
materials can occur during the manufacturing stage. of the product.
QbD for raw materials may need to provide strategies Rathore and Low provide a useful scheme for classi-
to detect and manage changes in raw materials that may fying raw materials into critical, key and non-key cate-
occur for the first time during commercial manufactur- gories [29,30] . They also provide a useful matrix or check-
ing. This is more in line with what has been described list of the categories of risk arising from raw materials
as a PAT or chemometric approach [26] . and how these can be handled, although most of these
Due to its complexity, special consideration needs to are more traditional methods of handling raw material
be given to cell culture media as a raw material. Cell risk than being uniquely different for QbD. The sec-
culture media can contain more than 40 compounds ond part of their paper gives examples of four different
[25] . Reactions can occur between these compounds, risk-assessment tools and demonstrates the application
generating even more chemical complexity. The metab- of FMEA to upstream and downstream materials along
olism of the cells also alters these compounds, further with an explanation of the logic for assigning different
increasing the complexity, making the whole system values for severity, occurrence and detection. In general,
prone to variability from run-to-run. For some older processes that are closer to the patient, especially drug
mammalian cell-culture products, complex natural product processes and the associated raw materials,
mixtures such as serum or plant hydrosylates are used in should be considered higher risk than earlier processes.
the media. These are complex mixtures and the analyti- The risk-assessment team should begin with the most
cal platforms required to characterize significant com- critical unit operations near the end of the process and
ponents of these mixtures are only now being developed work backwards through the process when assessing
[27] . Literature reports extensive use of many complex raw materials. At the end of the risk-assessment process
laboratory techniques (1H-NMR, LC–MS, ICP–MS, the team should benchmark their scoring against the
LC–DAD) coupled with sophisticated statistical analy- outcome of previous assessments for similar products
sis (principal component analysis, partial least squares and materials to ensure a consistent view of risk.
analysis, multivariate analysis, multilinear regression) to
attempt to determine the root cause of variation in cell »» Other
culture arising from these complex raw materials [25] . The use of a risk-assessment procedure to determine
Some of these reports, even with such extensive efforts the CQAs of a mAb product has been described in [31] .
and sophisticated techniques, are not entirely successful This example illustrated the combined use of risk as-
in explaining the variation seen or completely determin- sessment procedures with prior knowledge and litera-
ing root cause for out-of-trend results. For this reason ture review. A recent study has tried to link experience
it is probably prudent to avoid the use of complex or in human clinical trials with mABs to physiochemical
naturally derived components in cell culture as much as measures of the protein to further improve criticality
possible. Completely chemically defined media is likely determinations for this class of proteins [32] .
to be more economical and reliable over the long term, The use of risk assessment as an adjunct to valida-
even if it is not capable of producing the same yield of tion of an analytical assay is presented in [33] . The FDA
product in cell culture. It is fortunate that, while cell has also indicated that it has accepted some NDA ap-
culture is often the most variable stage of biopharma- plications in which analytical methodologies are de-
ceutical production, it is also the most distant from the veloped using a QbD approach and supplied a recom-
patient and variation at this stage is less likely to present mended approach for developing analytical tests in this
risk to the patient. Experimental efforts to remove or manner [113] .

Low pH viral inactivation, a common procedure Developing a manufacturing process for a bio
for biopharmaceuticals expressed in mammalian cell pharmaceutical using QbD principles corresponds
culture, is developed from a QbD perspective in the to the last 3 steps of the Quality Planning Roadmap
A-mAb case study [112] . (Figure 3) – ‘Develop process features’, ‘Process designs’,
‘Establish process controls and transfer to operations’.
Implementation of QbD In order to carry out these steps it is necessary for the
The concept of QbD can be traced to Joseph Juran [34,35] previous 7 steps of the Quality Planning Roadmap to
and it is useful to return to some of his original works for have been completed and documented for reference.
the clarity with which they describe the concept of QbD. These 7 prior steps are often not carried out by process
Juran’s method for implementing QbD is shown in development groups, they are generally the functions of
Figure 3 and was referred to by Juran as the ‘quality plan- different groups or senior management in the company
ning roadmap’ [36] . The virtue of this model is the ease so it is important to ensure the outputs of these previ-
with which it can be reduced to a step-by-step guide – Ju- ous steps are thoroughly communicated and recorded.
ran was quite conscious of the fact that deciding at which It is also necessary for the process development group
point to begin planning partly came down to a judgment to remember that it has customers – the manufacturing
call and decided on this series of steps to avoid repeated or operations group to whom the process will be trans-
looping through the planning cycle [34] . The ‘roadmap’ ferred for example – and the needs of these customers
has nine important steps: identifying customers; deter- should be correctly identified.
mining the needs of those customers; translating those A flowchart for implementing the ‘Quality Planning
needs into company language; developing a product Roadmap’ in biopharmaceutical process development is
that can respond to those needs; optimizing the prod-
uct’s features to meet both company needs and customer Activities Outputs
needs; developing a process that is capable of producing
the product; optimizing the process; proving that the Establish quality
process can produce the product under operating con- goals
ditions; and, transferring the process to operations [37] . List of quality
Notably absent from the work of Juran is the use of risk goals
assessment, as included in the ICH and FDA guidances.
Identify those
When considering use of the ‘quality planning roadmap’
impacted – the
in the biopharmaceutical industry it is useful to remem- customers
ber that some of the customers (indeed before market-
Apply measurement throughout
ing approval, the dominant customers) are the regulatory List of customers
authorities involved, so many of the requirements that
need to be sought from the customer should be sourced Define customers’
from the regulatory authorities and the relevant guid- needs
ance and regulations. This is a very important point to
List of customers’
remember in order to produce manufacturing processes needs
and products that are compliant with regulation, while Develop product
also producing a product with the required attributes. features
Also pertinent is that Juran further subclassified cus-
tomer needs into: stated needs, real needs, perceived Product designs
needs, cultural needs and needs traceable to unintend-
Develop process
ed use. This classification recognized the fact that the features
needs stated by customers are only a part of the whole
needs – the customer has many other needs that are Process designs
not consciously identified [34] . For example, regulatory
agencies require that resin and membrane lifetime stud- Establish process
controls, transfer to
ies should be performed before a drug is approved for operations
sale – this is a stated need. However, the real need is
Process ready
for a manufacturing process that performs consistently to produce
and reproducibly over the lifetime of the product and
this real need should be considered for the process as
a whole in addition to the stated needs of resin and Figure 3. Juran’s Quality Planning Roadmap.
membrane lifetime studies. Reproduced with permission from [36].

proposed by the authors and is shown in Figure 4. In »» Prioritizing the CQAs to measure;
broad outline the Roadmap consists of:
»» Prioritizing operating variables for study;
»» Prioritizing unit operations for further study; »» Designing and performing DoE studies;
»» Identifying the CQAs of the product; »» Recording the results.
»» Identifying the CQAs affected by the highest
This approach is intended to identify distinct outputs
priority unit operation;
during QbD efforts, outputs that are necessary for
communication and collaboration
among the various groups inside a
Inputs Activities Outputs
pharmaceutical company that must
Prioritization of Priority list of unit coordinate their efforts before, dur-
Prior knowledge
unit operations operations ing and after product development.
Literature review
Before QbD can be implemented
in a manufacturing process it is
Prior knowledge Identify CQAs necessary to have a rudimentary
Literature review of product initial process. Once this has been
List of CQAs established, it is necessary to move
backwards through the unit opera-
Identify CQAs tions of the process, beginning with
Prior knowledge affected by the
Literature review the last unit operation in the process
unit operation
train. It is important to progress in
this manner as the output of each
unit operation in a manufacturing
CQA values of input
to unit operation
Risk assessment/ process becomes the input for the
prioritization of CQAs next unit operation. Beginning at
(output of preceding
to measure
unit operation) the start of the manufacturing pro-
List of CQAs to cess can lead to the initial unit oper-
measure in DOE ations being optimized, to the detri-
studies ment of the subsequent unit opera-
Identify operating
Prior knowledge variables of the unit
tions and the overall manufacturing
Literature review operation train. Progressing in this manner
is also in itself a risk assessment or
List of operating prioritization of work, with the level
variables
of risk of a unit operation generally
Determine which thought to increase with proximity
operating to the patient. This aligns with the
Prior knowledge
variables are likely
Literature review
to affect product
fact that regulatory agencies will use
CQAs risk assessments to determine which
Risk assessment/ points in the process and which
prioritization of manufacturers and unit operations
operating variables
are most likely to cause risk to the
Design and perform final patient, and, therefore, require
DOE studies increased scrutiny [107,114,115] . The
Study report on general rule of thumb of moving
unit operation from the last unit operation back
Proceed to next through to the initial unit operation
highest priority could be replaced by a more formal
unit operation
risk assessment to identify areas
of the process more likely to cause
Figure 4. Suggested activities for implementing quality-by-design in downstream harm if they become out of control
unit operations. (developing a priority list of unit op-
CQA: Critical quality attributes; DoE: Design of experiment. erations as shown in Figure 4). De-

termining the safe design space for the highest risk unit of the critical steps and operating parameters in the
operations should be placed as the highest priority. manufacturing process.
The next step in Figure 4 is to identify the CQAs of After this exercise is completed the same exercise
the product, then go on to identify which CQAs are can be repeated with the next highest priority unit op-
thought to be affected by the specific unit operation – eration until the whole manufacturing process train is
this can be conducted using a combination of literature complete or the highest priority unit operations have
reviews, prior knowledge and risk assessment. Following been addressed.
this, the operating variables of the unit operation can be
identified. A risk assessment can be performed on these Pitfalls of QbD
variables – this should include a literature review and Risk management, statistical tools and knowledge man-
other prior knowledge. This risk assessment can then be agement are all useful tools when implemented in the
used to determine which of these variables will affect, appropriate manner. However, these are all very broad
which may affect and which are unlikely to affect prod- concepts and, due to the large number of variables
uct CQAs. The next step in the process is to identify involved in any pharmaceutical manufacturing opera-
the values for the CQAs in the feedstock for the unit tion, the amount of experimentation and documen-
operation under consideration. This is effectively the tation required to thoroughly carry these out can be-
output of the preceding unit operation. Having these come very large. In a recent approval application from
values will allow determination of the effect of the unit Genentech, who has worked closely with the FDA on
operation on the CQAs. On the basis of the risk assess- QbD, the chemistry, manufacturing and controls sec-
ment and the values of the CQAs in the feedstock, the tion was approximately 1500 pages in length due to the
variables and outputs to measure in DoE studies can be inclusion of all the backup data, risk assessments and
determined and justified. Multivariate experimentation justifications [116] .
can then be carried out and the output levels of each Similarly, DoE studies are a very effective tool and
of the CQAs compared with the inputs. This confirms in many published papers QbD and DoE have almost
whether each CQA is affected and whether the process become synonymous. However, the large number of
variables chosen are robust within the ranges selected. variables that can be considered in a manufacturing
It may be useful to go through this exercise with a process, especially a drug substance process, make the
variety of processing technology alternatives for each number of possible experimental runs very large. For
unit operation in order to maintain flexibility for fu- example, the typical process chromatography unit op-
ture improvements in technology. Demonstrating eration can have 50–100 operating parameters that can
that acceptable CQAs are achieved over a multivariate affect its performance [38] . Examining all of these pa-
range of parameters and a range of processing alterna- rameters would require a prohibitively large number of
tives demonstrates the robustness of the manufactur- experimental runs. As a result of the increased experi-
ing process. ICH Q8 identified that a more flexible mentation and documentation required, a recent esti-
regulatory approach could be achieved if the applicant mate has suggested that adoption of QbD approaches is
could demonstrate an ‘enhanced knowledge’ of prod- likely to increase development costs by up to $1 million
uct performance over a range of ‘material attributes, [23] . Due to the provenance of this estimate (from a re-
manufacturing process options and process parame- port commissioned by the FDA and used to promote
ters’. For unit operations that are intended to produce the QbD concept) it should be considered a conser-
a specific CQA value or limit, it is useful to determine vative estimate likely to err on the low side. Evidence
what maximum value can be used to challenge that from the FDA that the generics industry, in which low
unit operation, that is, if a chromatography step is de- development and production costs are competitive ad-
signed to reduce deamidation variants in a product to vantages, is much less keen on adoption of QbD would
a specification level it may be useful to challenge the appear to support this (cited in [23]). Both the European
unit operation in order to determine the maximum Medicines Agency and FDA representatives have stated
amount of deamidation variants that can be brought that they expect a more costly assessment process for
within the final specification. A study report on each QbD applications and the European Medicines Agency
unit operation should then be produced that identi- representatives also expect more difficult and costly
fies operating variables affecting the product and the inspections for QbD-based processes [23] .
safe operating ranges for those variables, along with An additional concern is that the output of statis-
identifying operating variables that did not affect the tically designed experiments and the resulting design
product in the unit operation [9] . Classifying in this space can be a large and complex mathematical equa-
manner is necessary to enable this information to be tion. Interpreting and applying this equation takes
used in regulatory submissions that require disclosure considerable mathematical and statistical knowledge

on behalf of both the pharmaceutical company and the Joseph Juran, to whom the origins of QbD can be
regulatory agency reviewing the design space. A poten- traced [35] stated that, “Product development requires not
tial problem for companies is that the regulators, who only functional expertise; it also requires the use of a body
would be required to approve the design space, may not of quality-related know-how”. He decried quality plan-
have the mathematical knowledge required to correctly ning performed by ‘amateurs’ – people who have not
interpret the design space presented. At a recent confer- been trained in the ‘quality disciplines’ – as one of the
ence of industry and representatives from various regu- main causes of poor quality processes and products [34] .
latory agencies, most attendees were unfamiliar with Training those who develop manufacturing processes
the concept of ‘n-dimensional space’ – a space having (development scientists and engineers) in these quality
more than 1, 2 or 3 dimensions (n-dimensions). This disciplines is the only way to address this fundamen-
concept is essential for interpreting complex design tal problem. Development scientists need a thorough
spaces as many reactions or processes can have more understanding of these quality disciplines as their data
than two variables controlling the output. The Euro- and experiments form the basis of the processes and
pean Medicines Agency representatives admitted that methods used during commercial manufacture [35] .
European Medicines Agency staff were ‘challenged’ in It has been remarked that in order to develop the de-
statistical knowledge and the FDA had not yet evalu- sign space for a process, well-designed DoE experiments
ated whether its staff had the requisite knowledge [23] . are required, but most process developers lack the statisti-
Elsewhere, representatives of the Canadian regulator cal knowledge to effectively design these experiments [23] .
Health Canada have stated that they are struggling to For example, it is very easy with the use of DoE software
deal with QbD applications [5] . A survey of industry to design a large, empirical study on a filtration or chro-
participants in 2012 indicated that many people in matography process. However, very detailed mechanistic
the pharmaceutical industry felt that the level of QbD models and understanding of these unit operations are
understanding by FDA regulators was variable from already available in the chemical engineering literature.
individual to individual [116] . It is, therefore, not necessary to retreat to an entirely em-
Not surprisingly, there appears to be widespread pirical level of understanding of these unit operations.
skepticism of the QbD concept, with a survey taken in Bringing the insights of mechanistic models to bear on
2008 showing that 58% of companies had QbD either specific applications in biopharmaceutical manufactur-
in the ‘ideas and vision’ or ‘not started’ stage of imple- ing requires very highly trained development scientists
mentation [39] . As the industry and regulators are not who can reduce the models to practice, and make the
in a position to publish examples of QbD applications knowledge contained in the models available to technical
(due to confidentiality and intellectual property con- staff who do not have the same level of scientific knowl-
cerns) it has not been possible to fully allay these con- edge. Furthermore, development scientists need to be
cerns. Up to mid-2010, a total of only five biological li- able to determine how to use the insights of such models
cense applications and four post-approval supplements on specific pieces of manufacturing equipment. Thus,
had been received for the biologics QbD trials [23] . a very important way for the quality of developmental
products to be improved, and one of the hopes for the
Conclusion QbD initiative to be realized, is to ensure that develop-
QbD is not a unified system as such, but is more cor- ment scientists are trained outside their original specialty
rectly described as an ethos that “quality cannot be disciplines in some (ideally all) of the tools of QbD.
tested into products; it should be built in by design” [115] . Similarly, as the QbD concept is being promoted by the
As such, it is difficult to prescribe a single solution for pharmaceutical regulatory authorities it is imperative for
QbD – developing any product will require a bespoke those authorities to have staff that are familiar with and
approach and, therefore, there is no standard blueprint understand the limits and caveats of these tools. This is
for applying QbD [5] . necessary in order for the authorities to be able to under-
QbD is also an armory of techniques or tools that stand applications that present this information and be
should be used for the development of good quality aware of when the information presented is insufficient.
products and processes. These tools include: DoE, risk Industry appears to have been restrained in the adop-
assessment, statistical quality control techniques (con- tion of QbD, and considering the complexities involved
trol charts) and mechanistic models, and understanding and the apparent lack of understanding by the regula-
of processes and products. tory agencies on how to deal with applications of this
In order to be able to implement QbD, biopharma- type, such caution is understandable [23] .
ceutical companies should ensure they have these ca- Hopes for greatly reduced costs of pharmaceuti-
pabilities. Development staff and company scientists cals and reduced regulatory burden due to the QbD
should be conversant with these tools and techniques. initiative should be limited. Much of the high cost of

pharmaceuticals is due to economic, political and regu- with the impurities in the product, which are in fact a
latory factors [40]; which are unaffected by the science wide variety of individual species (i.e., host cell protein,
behind the original development work. None of these host cell DNA, viruses) is likely to resist mechanistic
factors are likely to change due to the QbD or PAT ini- understanding as each individual host cell protein (for
tiatives; indeed, adoption of QbD approaches is almost example) will have its own specific properties and clear-
certain to increase development costs [23] . ance from the drug substance. For this reason, there will
The QbD concept is being used in the literature by always be a level of ‘empirical-only’ knowledge required
development scientists to publish and share effective in the development of a biopharmaceutical and the
methods for process development and characteriza- need for traditional validation will remain.
tion. This is a welcome and useful development of the
QbD concept and should help good practices become Future perspective
widely embraced throughout the industry and aid pro- The FDA has announced that it expects the pharma-
cess development scientists in becoming cross-trained ceutical development (drug product) section of all AN-
in quality disciplines. Some excellent examples in the DAs for small-molecule generic drugs to be in a QbD
literature have begun from mechanistic models of unit format from 1 January 2013 [117] . This change was
operations and have shown how these models can be achieved by updating the ANDA submission check-
used to develop robust and economical unit operations list to include QbD elements as requirements. This is
that are well characterized without having to assume likely to indicate that QbD will also become a require-
only the empirical level of understanding implied in a ment for biopharmaceuticals in the next 5–10 years.
naive DoE approach to the development of unit opera- Close attention should, therefore, be paid to how QbD
tions [16–18,38] . However, the way unit operations inter- is applied to generics so that the lessons learned can
act with a given product is less likely to be deducible be applied to more complex biopharmaceuticals in the
from mechanistic models due to the complexity of bio- future. Close attention should also be paid to how the
pharmaceutical products. How unit operations interact QbD process has been implemented in the semicon-
Executive summary
cGMPs for the 21st Century
»» The pharmaceutical cGMPs for the 21st Century initiative to improve innovation, risk-based orientation.
»» ICH Q8: initial codification of QbD for Drug Product.
»» Distinction between drug product and drug substance.
»» ICH Q9: Quality Risk Management, quality related to patient safety, level of effort commensurate with risk.
»» ICH Q10: Pharmaceutical Quality System, key elements expected of a quality system in a pharmaceutical company.
»» ICH Q11: Development and Manufacture of Drug Substances, QbD for drug substance manufacture.
»» Critical quality attributes of biological a direct result of drug substance process.
Systematic review of QbD in manufacturing biological products
»» No standard blueprint for QbD.
»» Empirical versus mechanistic approaches.
»» QbD differences for raw materials and drug product unit operations.
Implementation of QbD
»» Original QbD concept Juran – contrasts with FDA and International Conference on Harmonization concept.
»» Juran’s quality planning roadmap.
»» Regulators as customers.
»» A quality planning roadmap for biopharmaceuticals.
Pitfalls of QbD
»» Size of QbD applications and increased costs of applications and assessments.
»» Misapplication of design of experiment and risk assessment.
»» Limited and variable expertise inside regulatory agencies.
Concluding remarks
»» QbD as an ethos and an armoury of techniques.
»» The requirement for training of process developers in quality disciplines.
»» Superiority of mechanistic understanding over empirical.
»» The continued need for empirical proof (validation) in a regulated industry.
Future perspective
»» Possible mandating of QbD.
»» Effects on innovation and smaller companies.

ductor and microelectronics industries over previous these smaller companies to thrive in the QbD regula-
decades, as these industries have made significant im- tory environment and, thus, maintain a vital element
provements through the use of the quality disciplines of new product development.
and tools during development.
The majority of biopharmaceuticals under devel- Financial & competing interests disclosure
opment are produced by small biotechnology compa- The authors have no relevant affiliations or financial involve-
nies who are under funding constraints and pressure ment with any organization or entity with a financial interest in
to bring their products to market [5,23] . The increased or financial conflict with the subject matter or materials discussed
funding requirements due to QbD may be most chal- in the manuscript. This includes employment, consultancies, hono-
lenging to these companies, and the development of raria, stock ownership or options, expert testimony, grants or patents
literature and industry guidance documents from received or pending, or royalties.
regulators that can assist in addressing QbD require- No writing assistance was utilized in the production of this
ments in a cost-effective manner will be necessary for manuscript.
References
Papers of special note have been highlighted as: process design space for biotech products: case study of
n
of interest Pichia pastoris fermentation. Biotechnol. Prog. 24(3), 655–662
n
of considerable interest
n
(2008).
1 Stevenson D, Cochrane T. Implementation of QbD part 1 – 11 van Hoek P, Harms J, Wang X, Rathore AS. Case study on
setting product specifications. Reg. Rapporteur 8(2), 3 (2011). definition of process design space for a microbial fermentation
2 Stevenson D, Cochrane T. Implementation of QbD part 2 step. In: Quality by Design for Biopharmaceuticals: Principles
– organizational implications. Reg. Rapporteur 8(2), 3–16 and Case studies. Rathore AS, Mhatre R (Eds). John Wiley and
(2011). Sons, Inc., NJ, USA, 85–108 (2009).
3 Kozlowski S, Swann P. Current and future issues in the 12 Abu-Absi SF, Yang L, Thompson P et al. Defining process
manufacturing and development of monoclonal antibodies. design space for monoclonal antibody cell culture. Biotechnol.
Adv. Drug Deliv. Rev. 58(5–6), 707–722 (2006). Bioeng. 106(6), 894–905 (2010).
4 Rathore AS, Winkle H. Quality by design for 13 Horvath B, Mun M, Laird MW. Characterization of a
biopharmaceuticals. Nat. Biotechnol. 27(1), 26–34 (2009). monoclonal antibody cell culture production process using a
quality by design approach. Mol. Biotechnol. 45(3), 203–206
5 Schmitt S. The roadmap to QbD. In: Quality By Design:
(2010).
Putting Theory into Practice. Schmitt S (Ed.). Davis Healthcare
International Publishing, MD, USA, 31–54 (2011). 14 Mila L, Valdes R, Tamayo A, Padilla S, Ferro W. Application
of a risk analysis method to different technologies for
6 Kozlowski S. Protein therapeutics and the regulation of
producing a monoclonal antibody employed in hepatitis B
quality: a brief history from an OBP perspective: as the
vaccine manufacturing. Biologicals 40(2), 118–128 (2012).
biotechnology industry has matured through various stages
of growth, regulatory agencies have evolved in response to 15 Cecchini D. Applications of design space for
the need to define quality standards Biopharm. International. biopharmaceutical purification processes. In: Quality by
20(10), 37–40 (2007). Design for Biopharmaceuticals. Rathore AS, Mhatre R (Eds).
John Wiley and Sons, NJ, USA, 127–141 (2009).
n
A useful history of the changes in regulation, from the
US FDA’s perspective. 16 Mollerup JM, Hansen TB, Kidal S, Staby A. Quality by
design – thermodynamic modelling of chromatographic
7 Abboud L, Hensley S. Drug manufacturing, out of date for
separation of proteins. J. Chromatogr. A 1177(2), 200–206
years, gets a shot in the arm – US’s FDA prods industry to
(2008).
adopt innovations, raise quality standards. The Wall Street
Journal Europe, 03 September 2003. n n
A sophisticated approach to using the mechanistic models
available for ion-exchange chromatography to achieve
8 Seely R, Haury J. Applications of failure modes effect
analysis to biotechnology manufacturing processes. thorough characterization and effective control of a unit
In: Process Validation in Manufacturing of Biopharmaceuticals: operation.
Guidelines, Current Practices and Industrial Case Studies. 17 Kaltenbrunner O, Giaverini O, Woehle D, Asenjo JA.
Rathore AS, Sofer G (Eds). CRC Press, Taylor and Francis Application of chromatographic theory for process
Group, Boca Raton, FL, USA, 13–31 (2005). characterization towards validation of an ion-exchange
9 Brandreth EJ, Bussineau C, Butler M et al. Validation of operation. Biotechnol. Bioeng. 98(1), 201–210 (2007).
column-based chromatography processes for the purification n n
A sophisticated, mechanistic approach to achieving quality by
of proteins. Technical report no. 14. PDA J. Pharm. Sci. design for chromatography that demonstrates more under-
Technol. 62(Suppl. 3) 2–36 (2008). standing than simpler attempts that use standard design of
10 Harms J, Wang X, Kim T, Yang X, Rathore AS. Defining experiment designs to build empirical models.

18 Watler P, Rozembersky J. Application of QbD principles to 31 Schenerman M, Axley M, Oliver C, Ram K, Wasserman
tangential flow filtration operations. In: Quality by Design G. Using a risk assessment process to determine criticality
for Biopharmaceuticals: Principles and Case Studies. Rathore of product quality attributes. In: Quality by Design for
AS, Mhatre R (Eds). John Wiley & Sons, Inc., NJ, USA, Biopharmaceuticals: Principles and Case Studies. Rathore AS,
111–125 (2009). Mhatre R (Eds). John Wiley and Sons, NJ, USA, 53–82
n n
Tangential flow filtration development taking into account (2009).
both mechanistic understanding and the unique requirements 32 Goetze AM, Schenauer MR, Flynn GC. Assessing
of individual products. monoclonal antibody product quality attribute criticality
through clinical studies. MAbs 2(5), 500–507 (2010).
19 Bolton GR, Boesch AW, Basha J, Lacasse DP, Kelley BD,
Acharya H. Effect of protein and solution properties on 33 van Leeuwen JF, Nauta MJ, de Kaste D et al. Risk analysis
the Donnan effect during the ultrafiltration of proteins. by FMEA as an element of analytical validation. J. Pharm.
Biotechnol. Prog. 27(1), 140–152 (2011). Biomed. Anal. 50(5), 1085–1087 (2009).
20 Noordman T, Ketelaar T, Donkers F, Wesselingh J. 34 Juran JM. Juran on quality by design: the new steps for
Concentration and desalination of protein solutions by planning quality into goods and services. Free Press, NY,
ultrafiltration. Chem. Eng. Sci. 57 (4) 693–703 (2002). USA (1992).
21 Rohani MM, Zydney AL. Role of electrostatic interactions n n
The original work on QbD that encompasses many more
during protein ultrafiltration. Adv. Colloid Interface Sci. topics than design of experiment, risk assessment and
160(1–2), 40–48 (2010). target product profile.
22 Stoner MR, Fischer N, Nixon L et al. Protein-solute 35 McConnell J, Nunnally BK, McGarvey B. The forgotten
interactions affect the outcome of ultrafiltration/diafiltration origins of Quality by Design. J. Valid. Technol. 16(3), 5
operations. J. Pharm. Sci. 93(9), 2332–2342 (2004). (2010).
Useful examples of the Donnan effect complicating Drug n
Traces the origins of QbD and calls for a re-emphasis on
Product tangential flow filtration unit operations. the control and elimination of variation as the foundation
23 Schmitt S. The regulatory framework. In: Quality by design: of good quality.
Putting Theory Into Practice. Schmitt, S (Ed.). 36 Juran JM. Dr Juran and a call to action: the summit.
Davis Healthcare International Publishing, MD, USA Measuring Business Excellence 6(3), 4 (2002).
(2011). 37 Bicheno J. The Quality 60: a Guide for Service and
24 Martin-Moe S, Lim FJ, Wong RL, Sreedhara A, Sundaram Manufacturing. PICSIE Books, Buckingham, UK (1998).
J, Sane SU. A new roadmap for biopharmaceutical drug 38 Kaltenbrunner O, McCue JT, Engel P, Mollerup JM,
product development: Integrating development, validation, Rathore AS. Modeling of biopharmaceutical processes.
and quality by design. J. Pharm. Sci. 100(8), 3031–3043 Part 2: process chromatography unit operations. BioPharm.
(2011). International 21(8), 28 (2008).
25 Lanan M. QbD for Raw materials. In: Quality by Design for 39 Neway J. How to make the business case for Quality by
Biopharmaceuticals. Rathore AS, Mhatre R (Eds). John Wiley Design. BioPharm International 21(12), 42–50 (2008).
and Sons, NJ, USA, 193–208 (2009).
40 Friedman M. How to cure health care. Hoover Digest, 3
n n
A very interesting discussion of the unique aspects of raw (2001).
materials in quality-by-design (QbD) and the extraordinary
efforts that must be used to fully characterize some of the »» Websites
complex raw materials used in cell culture. 101 ICH. Pharmaceutical Development Q8(R2) 2009.
26 Pomerantsev A, Rodionova O. Process analytical technology: www.ich.org/products/guidelines/quality/article/quality-
a critical view of the chemometricians. J. Chemom. 26(6), guidelines.html
299–310 (2012). 102 BioProcessTechnologyConsultants. Quality by Design: Just
27 Luo Y, Chen G. Combined approach of NMR and a Buzz Word?
chemometrics for screening peptones used in the cell culture www.bptc.com/sites/default/files/biopulse_reports/qbd_
medium for the production of a recombinant therapeutic results_2012–2005–15_final.pdf
protein. Biotechnol. Bioeng. 97(6), 1654–1659 (2007). 103 USFDA. Guidance for Industry
28 Curinova L, Dudek F. Quality management in nineteenth- www.fda.gov/downloads/Drugs/GuidanceCompliance
century Czech beer and sugar PRODUCTION. In: A History RegulatoryInformation/Guidances/UCM124805.pdf
of Managing for Quality: The Evolution, Trends, and Future 104 USFDA. Demonstration of comparability of human
Directions of Managing for Quality. Juran JM (Ed.). ASQC biological products, including therapeutic biotechnology-
Quality Press, WI, USA, 657 (1995). derived products.
29 Rathore AS, Low D. Managing raw material in the www.fda.gov/Drugs/GuidanceComplianceRegulatory
QbD paradigm, part 1: understanding risks. BioPharm. Information/Guidances/ucm122879.htm
International 23(11), 34–40 (2010). 105 USFDA. Revised guidance for industry.
30 Rathore AS, Low D. Managing raw materials in the QbD www.fda.gov/downloads/Drugs/GuidanceCompliance
paradigm part 2: risk assessment and communication. RegulatoryInformation/Guidances/ucm072349.pdf
BioPharm. International 23(12), 5 (2010).

106 USFDA. Guidance for Industry: PAT – A Framework for 112 CMCBiotechWorkingGroup. A-mAb: a case study in
Innovative Pharmaceutical Development, Manufacturing, Bioprocess Development (2009).
and Quality Assurance (2004). www.ispe.org/pqli/a-mab-case-study-version-2.1
www.fda.gov/downloads/Drugs/GuidanceCompliance 113 Quality by Design approaches to analytical methods - FDA
RegulatoryInformation/Guidances/UCM070305.pdf perspective (2011).
107 USFDA. Pharmaceutical cGMPS for the 21st century - a www.fda.gov/downloads/AboutFDA/CentersOffices/Office
risk-based approach: final report (2004). ofMedicalProductsandTobacco/CDER/UCM301056.pdf
www.fda.gov/Drugs/DevelopmentApprovalProcess/ 114 USFDA. Pharmaceutical cGMPS for the 21st Century
Manufacturing/QuestionsandAnswersonCurrent – a risk-based approach: second progress report and
GoodManufacturingPracticescGMPforDrugs/ implementation plan (2003).
ucm137175.htm www.fda.gov/Drugs/DevelopmentApprovalProcess/
108 ICH. Specifications: test procedures and acceptance criteria Manufacturing/QuestionsandAnswersonCurrent
for biotechnological/biological products Q6B (1999). GoodManufacturingPracticescGMPforDrugs/
www.ich.org/fileadmin/Public_Web_Site/ICH_Products/ ucm071836.htm
Guidelines/Quality/Q6B/Step4/Q6B_Guideline.pdf 115 USFDA. Innovation and continuous improvement in
109 ICH. Quality Risk Management Q9 (2005). pharmaceutical manufacturing pharmaceutical cGMPs for
www.ich.org/fileadmin/Public_Web_Site/ICH_Products/ the 21st Century (2004).
Guidelines/Quality/Q9/Step4/Q9_Guideline.pdf www.fda.gov/ohrms/dockets/ac/04/briefing/2004–
110 ICH. Pharmaceutical Quality System Q10 (2008). 4080b1_01_manufSciWP.pdf
www.ich.org/fileadmin/Public_Web_Site/ 116 Taylor N. FDA mixed messages create QbD confusion
ICH_Products/Guidelines/Quality/Q10/Step4/Q10_ (2012).
Guideline.pdf www.in-pharmatechnologist.com/Regulatory-Safety/FDA-
111 ICH. Development and manufacture of drug substances mixed-messages-create-QbD-confusion-consultant
(chemical entities and biotechnological/biological entities) 117 Is QbD a mandate for generics? FDA responds to confusion.
Q11. pharma QbD (2012).
www.ich.org/fileadmin/Public_Web_Site/ICH_Products/ www.pharmaqbd.com/qbd_mandate_generics_fda
Guidelines/Quality/Q11/Q11_Step_4.pdf

Quality by Design For Biopharmaceuticalsa Historical Review and Guide Forimplementation

Uploaded by

Copyright:

Available Formats

Quality by Design For Biopharmaceuticalsa Historical Review and Guide Forimplementation

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Quality by Design For Biopharmaceuticalsa Historical Review and Guide Forimplementation

Uploaded by

Copyright:

Available Formats

Pharmaceutical

Quality by design for biopharmaceuticals:

A more modern approach to the development »» A reduction in overall cost of manu­

106 Pharm. Bioprocess. (2013) 1(1) future science group

cGMPs for the 21st century »» ICH Q8: Pharmaceutical Development

future science group www.future-science.com 107

108 Pharm. Bioprocess. (2013) 1(1) future science group

»» Preliminary hazard analysis; ucts in the development stage as in the commercial or

future science group www.future-science.com 109

process capability. Risk analysis is then performed to

parameters examined. If necessary, the design space is

A more general introduction to using FMEA to clas-

110 Pharm. Bioprocess. (2013) 1(1) future science group

able as either ‘critical’, ‘key’ or ‘non-key’ (Table 1) using

»» Cell culture Refine process

Define control Risk

lution screening experiments using the selected vari- validation

Table 1. Parameter classification definitions.

future science group www.future-science.com 111

112 Pharm. Bioprocess. (2013) 1(1) future science group

future science group www.future-science.com 113

114 Pharm. Bioprocess. (2013) 1(1) future science group

future science group www.future-science.com 115

116 Pharm. Bioprocess. (2013) 1(1) future science group

future science group www.future-science.com 117

118 Pharm. Bioprocess. (2013) 1(1) future science group

future science group www.future-science.com 119

120 Pharm. Bioprocess. (2013) 1(1) future science group

future science group www.future-science.com 121

122 Pharm. Bioprocess. (2013) 1(1) future science group

You might also like

A more modern approach to the development »» A reduction in overall cost of manu