Journal of Pharmaceutical and Biomedical Analysis 191 (2020) 113577

Contents lists available at ScienceDirect

Journal of Pharmaceutical and Biomedical Analysis

journal homepage: www.elsevier.com/locate/jpba

Potency assignment of biotherapeutic reference standards

Paul Faya a,∗ , Matthew W. Borer b , Kristi L. Griffiths a , Bhavin S. Parekh c
a
Statistics – Discovery / Development, Eli Lilly and Company, Indianapolis, IN, USA
b
Corporate Reference Standards Organization, Eli Lilly and Company, Indianapolis, IN, USA
c
Bioassay Analytical Development, Eli Lilly and Company, Indianapolis, IN, USA

a r t i c l e i n f o a b s t r a c t

Article history: The role of biotherapeutic proteins in the prevention and treatment of diseases such as cancers, infectious
Received 1 June 2020 diseases, and autoimmune disorders continues to grow. The biological activity or “potency” of a biother-
Received in revised form 13 August 2020 apeutic reflects its mechanism of action and thus its efficacy. The potency of these complex biomolecules
Accepted 15 August 2020
cannot be quantitatively correlated to chemical and physical properties and thus must be determined by
Available online 24 August 2020
comparison to a reference standard, typically using a cell-based bioassay. This lack of an absolute method
for determining potency, along with test method variability and potential for bias make assignment and
Keywords:
monitoring of reference standard potency a major challenge during pharmaceutical development and
Reference standard
Potency
manufacturing. The reference standard links the potency of dosages administered to the patient with
Bioassay those of original clinical studies. Therefore, the assignment of potency to biotherapeutic reference stan-
Biotherapeutic dards is vital for assuring the quality of medicines for patients. In this work, we propose a comprehensive
Monoclonal antibody roadmap for assigning potency to reference standards that is compliant with the two-tier system of stan-
dards as recommended in regulatory guidance. The roadmap includes statistical approaches for study
design and acceptance criteria that are risk-based and phase-appropriate. It also provides mitigation
approaches for potential assay bias.
© 2020 Elsevier B.V. All rights reserved.

1. Introduction Because the potency of the test sample (e.g., drug product
or drug substance) is determined by relative comparison to the
The biological activity (potency) of a biotherapeutic protein potency of the reference standard (RS), determination of RS potency
drug is commonly measured using a biological assays (typically throughout a product’s lifecycle (from development to commercial-
cell-based) which is intended to mimic the product’s known or ization) must be accurate, precise, reliable, and consistent across
intended in-vivo mechanism of action. Before commercial prod- time and across multiple batches of RS materials. If the potency of
uct release, the bioassay is used to evaluate the drug’s potency the RS material drifts over time, the manufacturer loses the abil-
against pre-defined specifications. Due to the inherent complex- ity to monitor the manufacturing process and the potency of drug
ity of biological systems, lack of fundamental knowledge about received by the patient could unintentionally change. Maintain-
structure-function relationships for complex biomolecules, and the ing consistency is a particularly difficult challenge when there is
number and variability of molecular variants that can result from no higher-order standard with which to compare. As an example,
bioproduction processes, it is not possible to quantitatively corre- for monoclonal antibodies (mAb), there is currently no compre-
late the potency of a test sample to physiochemical measurements. hensive repository of product-specific compendia standards as are
This makes it necessary to measure a sample’s potency relative to available for small molecules [1]. Initial efforts are being explored
a reference standard. Determination of a relative potency (RP) is by the UK National Institute for Biological Standards and Con-
achieved using a bioassay test in which a sample is expected to trol (NIBSC) which has established two World Health Organization
behave like a concentration or dilution of the standard. A compar- (WHO) International Standards for mAb drugs (rituximab and
ison of the concentration-response relationship between the test infliximab) (See [2–5]). Likewise, the European Pharmacopoeia has
sample and the standard material is used to obtain a RP measure established a compendial reference standard for infliximab [6]. A
for the sample. variety of challenges, many of which are described in this pub-
lication, make it difficult for such public reference standards to
be established and maintained. Questions also remain about the
∗ Corresponding author. applicability of a single public standard to testing of biotherapeutic
E-mail address: faya paul@lilly.com (P. Faya). products being produced by different manufacturers with differ-

https://doi.org/10.1016/j.jpba.2020.113577
0731-7085/© 2020 Elsevier B.V. All rights reserved.
2 P. Faya, M.W. Borer, K.L. Griffiths et al. / Journal of Pharmaceutical and Biomedical Analysis 191 (2020) 113577

Fig. 1. Overview of Reference Standards Life-Cycle Across the Drug Development Process.

ent cell lines, purification processes, and bioassay test methods. and FDA guidelines. The paper is structured as follows. We begin in
Finally, unless the manufacturer engages in the process of estab- Section 2 by describing the single-source reference standard life-
lishing public standards, it is unlikely that public standards may cycle in the context of the drug development process. Section 3
be established during product development or during the mar- gives the potency assignment and control strategy for each type
ket exclusivity period in a way that is useful for the manufacturer. of reference material throughout the life-cycle. This section also
Therefore, it is the responsibility of the drug sponsor to manufac- includes statistical details pertaining to study design considera-
ture and maintain in-house reference standards. The sourcing and tions, including sample size determination and the establishment
control of the in-house RS material is vital for assuring the quality of acceptance criteria. A description of a multifaced approach is pro-
of medicines for patients. Regulatory guidance recommends the vided in Section 4 to address the re-qualification of highest-order
use of a two-tiered system with primary and secondary (working) standards. In Section 5, we present an example of a bias mitigation
reference standards (e.g., The International Council on Harmoni- strategy specifically aimed at potency assignment studies. Conclud-
sation (ICH) Q6B (1999) [7] and Q7 (2000) [8] Guidelines and U.S. ing remarks are given in Section 6.
Food and Drug Administration (2015) Guidance on Analytical Pro-
cedures and Methods Validation for Drugs and Biologics [9]). This
2. Reference standard life-cycle
two-tiered system is also discussed by the US Pharmacopeia (USP)
Council of Experts and USP Reference Standards Committee in [10].
Several batches of reference standard material are generated
Although the two-tier system of RS material is clearly artic-
throughout the lifecycle of a biologic product. This section pro-
ulated by the international guidelines, regulatory agencies have
vides a general overview of the typical sequence and timing of
provided very limited information regarding the expectations for
reference standards for biologic products in reference to the drug
RS potency assignment approaches and there are few published
development process (see Fig. 1). Subsequent sections will discuss
guidance documents [11]. Moreover, the literature addressing
the details of the assignment and verification of the potency for
practices and methods for biotherapeutic protein reference stan-
each type of RS material.
dards is sparse. Schiel et al. [1] provide a general outline for the
The first interim RS batch is typically a bioproduct development
evolution of in-house reference standards and the timeline for
batch and/or a batch used in toxicological testing. This batch is used
qualification activities. Also, in recent years several industry meet-
for testing until clinical material is manufactured. At this point, a
ings organized by the California Separation Science Society (CASSS)
second interim RS batch is established if the first interim RS batch
and the Biopharmaceutical Emerging Best Practices Association
is not sufficiently representative of the clinical trial material. A new
(BEBPA) have focused attention on the topic of reference stan-
interim RS should also be established if there is a significant process
dards. BEBPA held two workshops in 2019 specifically aimed at
change (e.g.: a cell-line change). Through Phase 1 and Phase 2 trials,
addressing the challenges and opportunities for biotech reference
it is possible that the supply of the interim RS will be exhausted.
standards. In addition, CASSS held a Midwest Regional Forum in
In this case, a re-supply of the interim material is needed, and the
2019 which focused on case studies of reference standards man-
source batch should be representative of the clinical trial material.
agement. Mire-Sluis et al. [11] summarized the findings of several
The primary RS is implemented to serve as the basis for estab-
industry meetings from 2006 to 2013 which addressed selected ref-
lishing all future reference standards and also may be initially used
erence standard topics, including potency assignment and potency
to support routine testing (until the first working RS is imple-
stability monitoring. Borer also discussed the topic of reference
mented). It is commonly derived from, or representative of, a batch
materials for potency calibration of biopharmaceuticals at the 14th
used in pivotal clinical studies (Phase III) and which has been man-
International Symposium on Biological and Environmental Refer-
ufactured using the final optimized commercial process. As such
ence Materials in 2015 [12] and at the CASSS Well-Characterized
the primary RS is considered the ‘gold’ standard to which all sub-
Biotechnology Products conference in 2012 [13].
sequent reference standard batches are linked. If the primary RS is
In this paper, we propose a detailed roadmap for the ini-
used initially for routine testing, it is essential to transition to use of
tial assignment and periodic verification of the defined potency
a working RS before the stock of the primary RS goes below a pre-
for various reference standard materials used throughout clinical
defined quantity. This quantity must be sufficient to support use of
development and the commercial lifecycle of a product. We also
the primary RS for qualification (also referred to as “establishment”
present statistical methods for study design, including sample-size
or “calibration”) and stability testing of the working RS throughout
determination strategies that are phase-appropriate and risk-
the commercial life of the compound for a given manufacturer. This
based. To our knowledge, this is the first comprehensive work
can be a several decade time period, so measures must be taken to
detailing specific proposals for study design, acceptance crite-
package and store the primary RS in the most protective manner
ria, and life-cycle management for potency assignment. While
possible (e.g., deep frozen temperature, hermetic seal with inert
others have provided general overviews of the RS life-cycle (for
gas, optimized formulation, etc.). The quantity should be estimated
example [1,11]), our aim is to advance new ideas and stimulate
based on experience with other commercialized products, adding
current thinking by offering a comprehensive potency assignment
a significant safety factor (e.g., 3X) to ensure inventory shortage
approach that is conformant with the principles described in ICH
does not drive primary RS replacement. Preserving the inventory
 P. Faya, M.W. Borer, K.L. Griffiths et al. / Journal of Pharmaceutical and Biomedical Analysis 191 (2020) 113577 3

of primary RS material by using the working RS for routine testing of product development. To justify this assignment, the interim RS
ensures that the potency of dosages administered in the market must be sourced from a representative batch of material. Although
are consistent with those of the original clinical studies. The source there is no suitable comparator available, it is important to ensure
material for the primary RS should be representative of and prefer- that an expected dose-response relationship is exhibited in the
ably derived from the final optimized commercial process used in in-vitro bioassay. This can be accomplished by generating one or
the generation of pivotal clinical trial lots. In the rare event that a more dose-response curves using the bioassay. Note that if the first
primary RS must be replaced, a new batch is to be established as interim RS is implemented before a bioassay test is available, it may
a primary RS. The establishment strategy should include thorough be initially established with a defined potency only stated in protein
comparison with the former primary RS. content. If additional bioassay methods are implemented (e.g., tran-
Working RS batches are qualified by comparison to the primary sition from ELISA binding assay to cell-based assay), the assigned
RS as the higher-order standard and is prepared for the purpose value for the new type of method should be 100 %, again based
of routine quality control of production batches, including bioas- on representativeness and demonstration of dose-response and
say testing. The source for working RS batches is material that is the fact that the assigned potency is bioassay-specific. In addition,
representative of the commercial process and is not from the same ELISA and cell-based bioassays are not mechanistically equivalent
production batch as the primary RS. As such, working standards and as such, relative potencies between ELISA assays and cell-based
are typically derived from formal stability, global registration, and assays have limited value for this purpose.
commercial campaigns The first working RS should be established To demonstrate continued suitability for use and acceptable sta-
early enough so that information about it can be included in the first bility, interim RS batches must be tested at periodic intervals. At
market regulatory submission, meeting the expectation of regula- this phase of development there is no higher-order RS with which
tory authorities that a two-tier RS system be in place. Throughout its to compare. Moreover, the property of biological potency cannot
life-cycle, a working RS will likely undergo re-qualification against be directly correlated to physicochemical test methods. As a result,
the primary RS to ensure material stability. a multifaceted approach can used to re-evaluate the interim RS.
Finally, a bioassay control sample should be established that has In this approach, the combined evidence of all reasonable tests to
an assigned potency. This is a system suitability (or control) sample demonstrate potency stability can be used to support the conclu-
that is tested on each plate. Monitoring the relative potency of this sion that the RS material continues to be suitable for use. We further
control sample allows an evaluation of the suitability of results from discuss the multifaceted approached in Section 4.0.
each plate and each reportable result. Trend data may also be used Based on projected stock-out or significant changes to the
to monitor the consistency of the bioassay test and the stability of manufacturing process, it may be necessary to implement a
the primary RS as described below. The material used to source the replacement batch of interim RS. The new reference standard
bioassay control sample should be a batch that is independent from should be derived from a clinical trial drug substance batch to
any of the RS materials. The control sample can be formulated in ensure it is representative of material in the clinic. If a new interim
a way to enhance stability of potency (e.g., lyophilization, addition RS batch is generated, its potency should be determined by com-
of stabilizer excipients). Note that such enhancements may also be parison testing relative to the existing interim RS being replaced.
useful for working standards, provided that lack of interference is The acceptance criterion for the qualification study can be based
demonstrated. Use of independent production batches for primary on the sample mean for the %RP results. For example, if the sam-
RS, working RS, and the bioassay control sample helps prevent a ple mean is within some phase-appropriate range of, say, [80 % RP,
single failure mode from impacting more than one of these critical 120 % RP], the new interim RS material will be assigned a potency
materials. Monitoring relative potency between these three mate- value of 100 % RP. The ultimate acceptance range adopted should
rials is a key means by which all three are assessed for stability (i.e., be selected using a risk-based and project-specific approach. Post
a key part of the multi-faceted approach described in Section 4). hoc measures such as the confidence interval for the mean can be
For example, if all three materials are losing potency at the same evaluated to assess the precision achieved by the study as a means
rate, relative potency comparisons will not detect the change. to support the assigned potency of the replacement RS.

3. Potency assignment strategy 3.2. Primary RS

In this section, we describe the potency assignment strategy for At a point before or just after the start of pivotal clinical trials,
each type of RS (sections 3.1-3.3). Statistical details associated with the first primary RS should be established. This timing is selected
study design are provided in Section 3.4. Although the focus of these to ensure that some or all of the pivotal clinical trials are performed
sections is on biological activity (potency) assignment, it is impor- using drug product released relative to the primary RS, definitively
tant to note that qualification and re-qualification studies must also linking the activity of the dose given to patients to the assigned
include physicochemical tests to demonstrate that the RS is rep- potency of the primary RS. Note that the primary RS may be used for
resentative and as an additional check for gross degradation (see routine testing until the first working RS is established. The estab-
[11] for further discussion). For the initial primary RS, initial char- lishment of the first working RS must be done when the remaining
acterization should include performing the elucidation of structure primary RS stock quantity is still sufficient for use in establishing
testing for the regulatory submission. An accurate definition of total and maintaining working RS batches for the projected commercial
protein, typically by UV absorption using an extinction coefficient, life of the product.
is also a critical aspect of RS characterization that is not addressed The first primary RS should be assigned a value of 100 %potency
in this publication. In addition, for RS re-qualification studies, all for two main reasons. First, the primary RS defines the potency of all
stability-indicating tests should be run and compared to the RS future drug product batches. Therefore, assigning at 100 %potency
specification as well as being compared to historical values for the provides a starting point that is aligned with pivotal clinical stud-
batch of RS. ies. Second, the primary RS is typically the first RS sourced from the
commercial manufacturing process. Thus, it is appropriate to con-
3.1. Interim RS sider this a new starting point for potency definition. Note that it is
possible to blend/mix multiple drug substance batches to form the
The first interim RS should be assigned a value of 100 % relative primary RS, especially if batch-to-batch differences in true potency
potency (RP) because no suitable comparator exists at this stage are expected. Blending/mixing results in a primary RS that is as rep-
4 P. Faya, M.W. Borer, K.L. Griffiths et al. / Journal of Pharmaceutical and Biomedical Analysis 191 (2020) 113577

resentative as possible of the commercial process and as centered the re-qualification study. As an example, the equivalence accep-
as possible to the average true potency. tance margin can be 10 % RP and the alpha level set to 0.05 (see
The initial primary RS should be tested in comparison to the Section 3.4 for statistical details). Thus, if the 90 % confidence inter-
most recently used interim RS to allow evaluation of any shift in val for the mean %RP is contained within the interval of [assigned
potency assignment between the development and commercial potency ±10 % RP], then the potency assignment of the work-
processes. The bridging study should consist of comparison testing ing RS is considered suitable for continued use. The equivalence
using the interim RS as a standard and the primary RS as a sample. approach and the equivalence margin should be appropriate for a
The acceptance criterion for the qualification study can be based re-qualification activity to monitor the long-term consistency of
on the sample mean for the %RP results (potency of PRS relative the initial assignment of potency.
to that of the interim RS). For example, if the sample mean for the Fig. 2 shows the %RP lot release results of a drug product across
%RP is within some phase-appropriate range of, say, [90 % RP, 110 a timespan that encompasses the transition from the primary RS
% RP], then the primary RS is considered sufficiently comparable to (PRS), to the first working RS (WRS1), and to the second working
the interim RS. The range should appropriate for characterizing the RS (WRS1). This drug product is a typical biologic product being
continuity of clinical potency at this phase of development. If the tested for relative potency in a cell-based bioassay method. Prior
sample mean is outside of the acceptance range, the 100 % potency to the establishment of the first working RS, the primary RS was
assignment should be kept because the primary RS is representative used directly for lot release testing. After the first working RS was
of pivotal clinical material and defines the activity of the prod- established and assigned potency via a qualification study, three
uct from this point forward. However, the impact of the difference subsequent re-qualification studies were executed through its lifes-
between the interim and Primary RS batches should be investigated pan to confirm the original potency assignment. As can be seen in
and justified. For example, any impact on the validity of previous Fig. 2, the transition from the first working RS to the second work-
toxicology and clinical studies or comparability between develop- ing RS did not lead to a shift in the %RP results. This is a consequence
ment and commercial materials should be investigated. of a well-designed and executed qualification study for WRS2 that
Although the initial primary RS is established with a quantity assigned potency to the WRS2 in a sufficiently accurate and precise
projected to last for the entire commercial life of a product, it is manner. In this case, both WRS1 and WRS2 were assigned a potency
possible that there will be a need to replace a primary RS. In this of 100 % relative to the primary RS based on the study results. There
case, a new batch should be established as the primary RS. This is also no apparent overall trend of increasing or decreasing lot
new batchshould be derived from material representative of the release potency over the entire time period, supporting the robust-
commercial process with a potency determined by comparison to ness of the qualification and re-qualification strategies presented
the primary RS being replaced. Blending/mixing multiple batches is in this document and applied to the reference standards used to
also an option as described previously. The acceptance criterion for release this drug substance.
the qualification study of the replacement primary RS can be based
on the sample mean for the %RP results. Suppose a 95 % confidence 3.4. Study design and analysis considerations
interval is to be considered. In the case where the 95 % confidence
interval contains 100 %RP, then the replacement primary RS can be The variability of the potency assay and desired level of study
assigned 100 % RP relative to the current Primary RS. However, if the precision can be used to determine the minimum sample size when
confidence interval does not contain 100 %RP but the sample mean relative potency is measured in RS qualification or re-qualification
is within, say, [95 %RP, 105 %RP], then the replacement primary RS studies. The precision of the estimated sample mean can be eval-
can be assigned a potency as the sample mean. In the latter case, uated by assessing the width of the confidence interval for the
an assessment of the study results (and possibly additional test- mean.
ing) should be conducted to confirm an assignment of potency that Let x1 , . . ., xn denote the individual results generated from
is different from 100 % RP. An assessment should also be done to the study. Note that the distribution of relative potency results
confirm that the material being established as a replacement pri- is log-normal. Therefore, it may be appropriate to transform the
mary RS is sufficiently representative of the manufacturing process. data prior to analysis such that the xi ’ s denote the natural log-
Section 3.4 provides a detailed discussion on decision-process for transformed relative potency results. However, we find that in
assigning 100 % RP or the sample mean. general, particularly for RS qualification and re-qualification stud-
ies, the distribution of the results is sufficiently symmetric such
3.3. Working RS that the assumption of normality is adequate  for the analysis (see
n
for example, Fig. 4). The sample mean is x = 1n i=1 xi and the con-
The working RS batches should be derived from a representa- fidence interval for the mean is computed as x ± t ∗ × √sn , where s
tive drug substance batch and its potency assignment should be denotes the sample standard deviation and t ∗ is the upper 1 − ˛/2
derived by comparison to the primary RS. Suppose a 95 % confidence quantile of the Student’s t distribution with n − 1 degrees of free-
interval is to be considered. In the case where the 95 % confidence dom.
interval contains 100 % RP, then the Working RS can be assigned For the replacement of an interim RS, bridging of the primary RS
100 % RP. However, if the confidence interval does not contain 100 to the interim RS, replacement of a primary RS, and qualification of
% RP but the sample mean is within, say, [95 % RP, 105 % RP], then a working RS, the study sample size can be determined by solving
the Working RS can be assigned a potency as the sample mean. As the following inequality [14] for n:
with the replacement of a primary RS, an assessment of the study  2
results (and possibly additional testing) should be conducted to ˆ
[n × (n − 1)] ≥ 2n−1, 1− F1, n−1, 1−˛ , (1)
confirm an assignment of potency that is different from 100 % RP w
(See Section 3.4). An assessment that the working RS is sufficiently
representative should also be performed. where ˆ is the estimated variability of the test statistic, ˛ is the
To demonstrate continued suitability for use and overall stabil- desired level of confidence for the computed confidence interval,
ity, working RS batches should be tested at periodic intervals. The  is the probability that the half-width of the confidence interval
main approach for demonstrating that the potency assignment con- for the mean () is greater than w, 2n−1, 1− is the upper one-sided
tinues to be suitable for use is by comparison to the primary RS. A (1 − ) percentile of a chi-square distribution with (n − 1) degrees
statistical equivalence approach is one recommended technique for of freedom and F1, n−1, 1−˛ is the upper one-sided (1 − ˛) percentile
 P. Faya, M.W. Borer, K.L. Griffiths et al. / Journal of Pharmaceutical and Biomedical Analysis 191 (2020) 113577 5

Fig. 2. Example of %RP release results across reference standard transitions.

√
of an F distribution with 1 numerator degrees of freedom and (n − removal of outliers as well as additional testing to reduce t ∗ s/ n.
√
1) denominator degrees of freedom. Similarly, in the situation where t ∗ s/ n  w and 100 % is outside of
Assuming that the variability observed from the qualification the confidence interval, an evaluation of the results should be made
study results is consistent with the a priori estimate, ,
ˆ then the by a group of SMEs to determine if there was insufficient variabil-
resulting sample size will be the minimum required to ensure that ity included in the study and the potential for bias created (e.g. due
the half-width of the (1 − ˛) × 100% confidence interval for  is no to setup). In this case, it may be appropriate to perform additional
more than w for more than  × 100% of the time. As a starting point, testing on independent setups/runs. A SME evaluation is also rec-
the analyst can set ˛ = 0.05 and  = 0.05. ommended when both 100%RP ∈ / 95%CI and x ∈ / [95%RP, 105%RP].
For the re-qualification of a working RS, the sample size can be This situation suggests that the potency of the candidate RS mate-
determined using an equivalence approach that test the hypothe- rial may be sufficiently different from the current primary RS to
sis: warrant assessment of its suitability for use as a reference mate-
    rial. Note that we have used 95–105 % as an example in this section.
H0 : 1 − ˆ 0  ≥  vs. H1 : 1 − ˆ 0  <  (2)
The ultimate range selected by the drug sponsor should reflect the
where ˆ 0 is the assigned potency of the working RS from the quali- high degree of certainty required for potency linkage throughout
fication study and 1 is the mean of the re-qualification study. The the lifecycle of the molecule, post establishment of the first primary
sample size [15], n, is chosen such that RS.
 2 For the re-qualification of a working RS, the hypothesis in (2)
ˆ  2 may be tested using a two one-sided test (TOST) procedure. That
n ≥ z1−˛ + z1−ˇ/2 , (3)
 is, if the (1 − 2˛) × 100% confidence interval for the mean relative
potency, 1 , is contained within the interval  ˆ 0 ± , then equiva-
where ˛ is the Type I error rate, ˇ is the Type II error rate, and lence is declared.
z˛ is the ˛th quantile of the standard normal distribution. Eq. (3) Finally, for the replacement of an interim RS, the assignment of
applies only for one-sample test statistic. As a starting point, the potency can follow the decision tree outlined in Fig. 4.
analyst can set  = 10, ˛ = 0.05, and ˇ = 0.05. Therefore, if H0 is
true, then the test will reject H0 at least ˛ · 100% of the time. If the
true  = 0, then the test will fail to reject H0 ˇ · 100% of the time. 4. Multifaceted approach
For the replacement of a primary RS or qualification of a new
working RS, the assignment of potency can follow the decision tree There are two situations when there is no higher-order stan-
outlined in Fig. 3. When qualifying a replacement primary RS or a dard to which a comparison can be made: interim RS and primary
new working RS, the new RS material can be assigned a potency RS qualifications and re-qualifications. In these situations, the
value of either 100 % or the sample mean, depending on the results reference standard potency cannot be defined by comparative test-
of the study. As such, it is important to ensure sufficient precision ing. Although overprotective packaging and storage temperature
in the potency estimate. The observed half-width of the 95 % confi- provide a degree of confidence that the material is stable, a mul-
√
dence interval for the mean, t ∗ s/ n, should be close to the a priori tifaceted approach that is data-driven offers needed additional
√
expected value, w. In the situation where t ∗ s/ n > w, an evalu- assurance of stability. This approach combines a wide range possi-
ation of the results should be made by a group of subject matter ble evidence that the potency of the RS remains unchanged. While
experts (SMEs). It should be noted that the design choice of w is none of these “facets” is sufficient alone, the combination of evi-
based on the a priori point estimate for the true level of variability, dence supports that the RS potency remains unchanged.
√
. Thus, it may be the case that t ∗ s/ n is considered “close enough” Before a working RS is established, a conclusion can be made
to w given the uncertainty associated with ˆ as well as the cho- that the interim or primary RS continues to be suitable for use with
sen levels of ˛ and . Additional considerations in the evaluation no change in the assigned potency when the following statements
conducted by the group may include an assessment and possible are all found to be true.
6 P. Faya, M.W. Borer, K.L. Griffiths et al. / Journal of Pharmaceutical and Biomedical Analysis 191 (2020) 113577

Fig. 3. Decision-tree for assignment of potency for a replacement primary RS or qualification of a working RS.

Fig. 4. Decision-tree for replacement of an interim RS.

1) Comprehensive physicochemical testing shows no changes that materials that are presumed to be stable. This approach includes
suggest a change in potency. relative potency assessment between four independent potencies
2) Monitoring the potency of the bioassay control sample relative that are all presumed to be constant over time (i.e., primary RS,
to the interim or primary RS shows no appreciable trend. working RS, bioassay control sample, and potency of materials com-
3) API and/or drug product release and stability results show no ing from the manufacturing process). It is highly unlikely that all
appreciable trends (unless explained by something other than four of these values will change in unison, making it highly likely
reference standard change). It should be noted that evaluation that any unexpected change will be detectable as drift in relative
of stability data, particularly when the material has been stored potency measurements. In addition, evaluation of the long-term
above nominal storage temperatures, should be interpreted in trend of API and/or drug product testing results (versus the work-
the appropriate context (e.g., change is likely to reflect degrada- ing RS) provides an orthogonal way to detect working RS potency
tion of the API/DP itself). change. If a long-term trend is observed in the manufacturing pro-
4) The primary RS is sourced from the commercial process on cess batch release results, this may indicate a change in the working
which additional representative stability testing is performed. RS potency over time. This is particularly informative because it
relies on the consistency of the manufacturing process rather than
the chemical/physical stability of individual batches of RS mate-
After the independent working RS is established, it allows rial. With the added assurances that the bioassay is performing in
an additional relative potency comparison between independent
 P. Faya, M.W. Borer, K.L. Griffiths et al. / Journal of Pharmaceutical and Biomedical Analysis 191 (2020) 113577 7

Table 1
Summary of Reference and Test Materials per Study.

Study Reference Material Test Material

interim RS replacement current interim RS candidate interim RS

primary RS bridging to interim RS interim RS (pre-primary) primary RS
primary RS replacement current primary RS candidate primary RS
working RS qualification primary RS candidate working RS
working RS requalification primary RS current working RS

a consistent manner and physicochemical properties of the RS are Table 2

Plate Layouts for SSR Approach.
not changing, there is a high degree of assurance that unexpected
change of the primary RS potency will be detected. A conclusion Plate Layout 1 Plate Layout 2
can be made that primary RS continues to be suitable for use with RS Position (RS) reference material reference material
no change in the “defined potency” when the following statements Sample 1 Position (S1) test material reference material
are all found to be true. Sample 2 Position (S2) reference material test material
Control Position control sample material control sample material

1) A trend analysis of the potency of the working RS when tested

using the primary RS shows no appreciable change.
2) Comprehensive physicochemical testing shows no changes that Suppose that on a micro-well plate, two positions are available
suggest a change in potency for sample materials, one is available for the reference material, and
3) Monitoring and trending the potency of the plate control rela- one is available for the control (or system suitability) material. For
tive to the working and primary RS shows no appreciable trend each plate, the reference material can be used in the RS position as
(except as explained by causes other than the reference stan- well as in one of the sample positions. The potency for the test mate-
dard). rial will be reported as the relative potency of the reference material
4) API and/or drug product release and stability results show no (tested against itself in one of the sample positions) divided by the
obvious trends (unless explained by causes other than the ref- relative potency of the test material (in the other sample posi-
erence standard change). It should be noted that evaluation of tion). We refer to this methodology as a sample-standard ratio
stability data, particularly when the material has been stored (SSR) approach. Table 1 describes the reference and test material
above nominal storage temperatures, should be interpreted in represented in each of the studies when the SSR approach is used.
the appropriate context (e.g., change is likely to reflect degrada- Suppose that it is determined that 35 plates are needed for a
tion of the API/DP itself). study. Each plate must have valid results for both sample positions.
5) The RS is sourced from the commercial process on which addi- Plate layout 1 is used for 17 of the plates and plate layout 2 is used
tional representative stability testing is performed. for the remaining plates, as depicted in Table 2.
For both plate layouts, the relative potencies of S1 and S2 are
calculated as
If any part of the multifaceted approach raises concerns about
the stability of the reference standard, an investigation should be RPS1 = 100 × EC501RS /EC50S1
conducted to determine potential causes. If it is deemed that the
potency of the reference standard has changed, then the suitability
of the reference standard should to be assessed. RPS2 = 100 × EC502RS /EC50S2

respectively, where EC501RS denotes the EC50 estimate for the ref-
5. Bias mitigation for RS studies
erence standard from the constrained fit to the S1 material and
EC502RS denotes the EC50 estimate for the reference standard from
It is common for a certain level of bias to affect bioassay potency
the constrained fit to the S2 material.
results. This is particularly the case for manual assays where, for
The adjusted (SSR) relative potency (RP) of the sample (i.e.: S1=
practical reasons, the materials and concentrations are not typically
test material) under plate layout 1 is calculated as
randomly allocated within the micro-well plate. Although bioas-
says are validated for the intended purpose of batch release and RPS1
stability indication, the potency assignment exercise can demand a RP1 = × 100
RPS2
much higher level of precision and accuracy. An acceptance range
of say, 95 %–105 %RP, is far narrower than typical release specifica- The adjusted (SSR) relative potency of the sample (i.e.: S2=test
tion limits and can encompass a range on the order of one standard material) under plate layout 2 is calculated as
deviations of the method variability for some assays. This narrow
range may be required by regulatory agencies to provide a high RPS2
RP2 = × 100.
level of assurance that no potency drift is occurring in the prod- RPS1
uct. As a result, a small amount of bias can significantly hinder the
successful execution of the qualification study. Some approaches A total of 17 RP1 plate results and 18 RP2 plate results are gener-
that have been proposed to reduce the potential bias influencing ated to achieve the required sample size of 35. The 35 results can be
the potency value assignment include: treating cells as gently as analyzed according to the approach described in Section 3.4. Note
possible, randomizing plate layouts (facilitated by automated liq- that in this example, the SSR approach is applied to the plate-level
uid handlers), and rotating the position of new and old reference RP result. However, the approach can also be applied to the reported
standards on a plate [11]. In this section, we present a methodol- result level. A reported result is typically generated across multi-
ogy to account for potential bias in the potency assignment study ple plates. Therefore, the adjusted (SSR) reported result is simply
by testing the reference material against itself. An example follows the ratio of the reported result for the test material to the reported
below. result of the reference material.
8 P. Faya, M.W. Borer, K.L. Griffiths et al. / Journal of Pharmaceutical and Biomedical Analysis 191 (2020) 113577

Fig. 5. SSR Data results for six different re-qualification studies for working reference standards.

5.1. Example manufacture of the product. A proposal for the potency assignment
qualification, re-qualification, and replacement of different types
In Fig. 5, we show results for six different working RS re- of in-house reference standards requirements has been presented,
qualification studies executed for six different biotherapeutic along with statistical details on sample size determination and
proteins with isotypes including IgG1 and IgG4. The large data acceptance criteria. We have also discussed a mitigation approach
points are the sample means for each data set. For each molecule for potential bias that can arise during the potency assignment
panel, the relative potency values for the candidate RS material exercise. These approaches are critical to maintain a consistent
and the reference RS material are presented. In the latter case, measurement system over time which assures that there is no drift
these results reflect a plate-effect bias since they represent relative in the potency of drug dosed in patients as compared to the original
potencies of a material being compared against itself. That is, the pivotal clinical trials.
same reference material is in the RS position and the sample posi- This roadmap is not intended to serve as “one-size-fits-all”
tion. The “ratio” values for each study indicate the SSR results. For approach. Rather, it offers a starting point for biopharmaceutical
proteins 1, 2, and 4, the plate-effect biases are negligible. For protein drug sponsors in development of a reference standard potency
5 the plate-effect bias is more pronounced, but not large enough to assignment program. Although we have proposed examples of
cause results to exceed the 95% ± 5% threshold. For proteins 3 and phase-appropriate acceptance criteria for various reference stan-
6, we see that the plate-effect is sufficiently large to cause a fail- dard qualification studies, the ultimate criteria adopted by the
ure in the qualification study, assuming that an acceptance range manufacturer should consider factors such as method and process
of [95% RP, 105% RP] for the mean is used. In this case, the benefit variability, risk tolerance, and regulatory expectations. Similarly,
of the SSR methodology is evident. The results for the candidate RS sample sizes and study design considerations should be project-
material are high and the sample mean is above the 105 % limit. specific and reflect these factors.
However, the results of the RS material compared against itself Because the potency for biopharmaceuticals is reported on a
are also high and indicate a plate-effect bias that exceeds 5%. In relative basis, one of the biggest challenges in the reference stan-
this situation, the adjusted SSR (ratio) results compensate for the dard control strategy is providing assurance that the potency of
plate-effect bias and bring the final sample mean result within the the highest-order standard is stable. Although protective storage
[95% RP, 105% RP] acceptance range. and temperature measures help to assure the chemical and phys-
ical stability of the RS material, there is ultimately no absolute
6. Discussion and independent test of biological activity for the primary RS. In
this paper, we have described a multi-faceted approach that makes
The potency of a biopharmaceutical product is a critical quality use of relative potency results between multiple sources of inde-
attribute that is directly linked to the its mechanism of action. The pendent materials as supporting evidence that the primary RS
potency assay is dependent on a well-characterized reference stan- is stable. Although the multi-faceted approach cannot absolutely
dard which provides a link to the pivotal clinical material for the guarantee that there is no drift in the potency of the primary RS,
life of the drug. The two-tier system of reference standard potency it is an effective measure to ensure that the measurement system
assignment described in this paper offers a roadmap for how to is stable over time. Establishing acceptance criteria for this type
maintain this link from the early development phase to commercial of re-qualification approach is quite challenging. True, long-term
 P. Faya, M.W. Borer, K.L. Griffiths et al. / Journal of Pharmaceutical and Biomedical Analysis 191 (2020) 113577 9

changes in the potency of a material is difficult to detect, particu- [2] World Health Organization (WHO), WHO International standard, 1st
larly given the multitude of factors that can influence the variability International Standard for the Biological Activities of Rituximab (2017), NIBSC
code: 14/210. Version 1.0 [https://www.nibsc.org/products/brm product
and bias of potency results, both in terms of the assay execution and catalogue/detail page.aspx?catid=14/210].
the manufacturing process. The drug sponsor will need to balance [3] World Health Organization (WHO), WHO International standard, 1st
the risks of false negative and false positive signals in determining International Standard for Infliximab (2019), NIBSC code: 16/170. Version 2.01
[https://www.nibsc.org/products/brm product catalogue/detail page.aspx?
the criteria for judging the stability of the primary RS over time. catid=16/170].
[4] World Health Organization (WHO), WHO International standard, 1st
CRediT authorship contribution statement International Standard for Adalimumab (2019), NIBSC code: 17/236. Version
2.01 [https://www.nibsc.org/products/brm product catalogue/
detail page.aspx?catid=17/236].
Paul Faya: Conceptualization, Methodology, Formal analysis, [5] C. Metcalfe, T. Dougall, C. Bird, P. Rigsby, M.E. Behr-Gross, M. Wadhwa, P.O.T.
Writing - original draft, Writing - review & editing, Visualization, study, The First World Health Organization International Standard for
Infliximab Products: A Step Towards Maintaining Harmonized Biological
Investigation. Matthew W. Borer: Conceptualization, Methodol-
Activity, 11, Taylor & Francis, 2019, pp. 13–25, In MAbs No. 1January.
ogy, Formal analysis, Writing - original draft, Writing - review & [6] European Directorate for the Quality of Medicines & Healthcare (EDQM).
editing. Kristi L. Griffiths: Conceptualization, Methodology, Formal (2018). Ph. Eur. Reference Standard. Infliximab CRS batch 1. [https://crs.edqm.
eu/db/4DCGI/search?vSelectName=1&vContains=1&vtUserName=infliximab
analysis, Writing - review & editing. Bhavin S. Parekh: Conceptual-
&OK=Search&vTypeCRS=].
ization, Methodology, Formal analysis, Writing - review & editing. [7] ICH Q6B, Specifications: Test Procedures and Acceptance Criteria for
Biotechnological / Biological Products, March, 1999.
[8] ICH Q7, Good Manufacturing Practice Guide for Active Pharmaceutical
Declaration of Competing Interest
Ingredients, November, 2000.
[9] Food and Drug Administration, Analytical Procedures and Methods Validation
The authors declare that they have no known competing finan- for Drugs and Biologics, US Department of Health and Human Services, 2015.
cial interests or personal relationships that could have appeared to [10] W.W. Hauck, USP Reference Standards Committee, Primary and secondary
reference materials for procedures to test the quality of medicines and foods,
influence the work reported in this paper. Pharm. Res. 29 (4) (2012) 922–931.
[11] A. Mire-Sluis, N. Ritter, B. Cherney, D. Schmalzing, Reference standards for
Acknowledgements therapeutic proteins: current regulatory and scientific best practices,
Bioprocess Int. 12 (3) (2014) 26–36.
[12] M.W. Borer, Reference materials for potency calibration of
The authors would like to acknowledge Alan Richter for devel- biopharmaceuticals, in: 14Th International Symposium on Biological and
oping the SSR approach as well as Christopher Breen and Thomas Environmental Reference Materials, Maryland, USA, 2015.
[13] M.W. Borer, Reference standards for monoclonal antibodies: key challenges
Parks for contributions to the statistical methodologies in this addressed, in: 2012 CASSS Well-Characterized Biotechnology Products
roadmap. Conference: 16th Symposium on the Interface of Regulatory and Analytical
Sciences for Biotechnology Health Products, San Francisco, CA, 2012
[https://cdn.ymaws.com/www.casss.org/resource/resmgr/WCBP/2012 WCBP
References BorerMatthew.pdf].
[14] L.L. Kupper, K.B. Hafner, How appropriate are popular sample size formulas?
[1] J.E. Schiel, A. Mire-Sluis, D. Davis, Monoclonal Antibody Therapeutics: The Am. Stat. 43 (2) (1989) 101–105.
Need for Biopharmaceutical Reference Materials. In State-of-the-Art and [15] S.C. Chow, J. Shao, H. Wang, Y. Lokhnygina, Sample Size Calculations in
Emerging Technologies for Therapeutic Monoclonal Antibody Clinical Research, Chapman and Hall/CRC, 2017.
Characterization Volume 1. Monoclonal Antibody Therapeutics: Structure,
Function, and Regulatory Space, American Chemical Society, 2014, pp. 1–34.