A Critical Analysis of the WHO

Biosimilar Guidance
Sarfaraz K. Niazi, Ph.D.
University of Illinois, Chicago, IL USA
1. BACKGROUND
The EMA1 was the first regulatory agency to issue biosimilar guidance in 2005; the next
major guidance was issued by the FDA2 in 2009 that differed from the EMA guidance, but now a
concurrence is developing between the EMA and the FDA that is leading to global
harmonization, albeit slowly. Many developed country regulatory agencies have either adopted
or modified these two guidelines for their territories. The third most influential guideline came
from the World Health Organization 3 (2009) that serves as an advisory, and many regulatory
agencies have adopted all or parts of the guidance including South Korea (2009), Japanese
PMDA4 (2009), South Africa (2010), Brazil (2010), Cuba (2011), Peru (2011), Mexico (2011),
Argentina (2011), Jordan (2012), Thailand, (2012), Nigeria (2012), Venezuela (2012), Egypt
(2012), Indian CDSCO5 (2012), Colombia (2013), Russia (2014), Iran (2014), and Chinese
National Medical Products Administration6 (2015).
2. THE WHO GUIDANCE
The WHO Similar Biologic Products Guidance (2009)7 and an explanatory Q&A (2018)8
is a subject of criticism for the inadequacy, misconceptions and irrelevant suggestions 9 provided
in the guidance. Given below is a discussion of a few selected elements of the WHO guidance
that require immediate rethinking by the WHO for their suitability. [Italicized text is the exact
quote from the WHO guidance without any corrections for grammar or spelling mistakes]
2.1.1. ANALYTICAL ASSESSMENT
Regulators may use their previous experience, gained for example, from evaluating

1
https://www.ema.europa.eu/en/human-regulatory/research-development/scientific-
guidelines/multidisciplinary/multidisciplinary-biosimilar
2
https://www.fda.gov/vaccines-blood-biologics/general-biologics-guidances/biosimilars-guidances
3
https://www.who.int/biologicals/biotherapeutics/similar_biotherapeutic_products/en/ ;
http://www.who.int/biologicals/publications/trs/areas/biological_therapeutics/TRS_977_Annex_2.pdf,
4
https://www.pmda.go.jp/files/000153851.pdf
5
http://nib.gov.in/NIB-DBT2016.pdf
6
http://subsites.chinadaily.com.cn/nmpa/index.html
7
https://www.who.int/biologicals/areas/biological_therapeutics/BIOTHERAPEUTICS_FOR_WEB_22AP
RIL2010.pdf
8
https://www.who.int/biologicals/expert_committee/QA_for_SBPs_ECBS_2018.pdf?ua=1&ua=1
9
https://www.healthpolicy-watch.org/wp-content/uploads/2019/04/Memo-on-WHO-Guidelines-on-
SBPs-.pdf
 changes introduced into the manufacturing processes of other products to understand the
functional and clinical impact of a particular physicochemical difference between the
SBP and its RBP.
The knowledge of changes in the manufacturing process (a subject of Q5E) does
not apply to biosimilars where a third-party reference product is required, and
there is little knowledge of in-process controls. Further, sharing information from
other regulatory submissions may violate regulatory data protection (RDP) as per
TRIPS Article 3910.
Production lots from different manufacturing sites can be used, provided that products
from all manufacturing sites are approved by the relevant regulatory authority.
All production lots must come from the same site since the ICH Q5E allowance is
available only after the approval of the product, not before the approval of the
PBP. Approval by a regulatory agency for cGMP has no relevance to the
justification of combining different products for testing.
The specifications of the RBP and SBP are likely to be somewhat different because of
different manufacturing processes, product-attribute understanding, and analytical
methods. Thus, the specifications of the RBP reflect the experience with the
manufacturer's product.
The release specification of an SBP is established using side-by-testing with the
RBP, and there can be no differences in any acceptable limits.
2.1.2. REFERENCE PRODUCT
The RBP should have been marketed for a suitable duration and should have a history of
marketed use in a jurisdiction with a well-established regulatory framework and
principles, and with considerable experience in the evaluation of biotherapeutic products
and post-marketing surveillance activities. Market experience of the RBP should also be
available based on significant duration and magnitude of exposure on the market.
Once a reference product is approved, it should be treated as safe and effective,
and the biosimilars made subject to the same type of post-market surveillance.
The manufacturer of the SBP should justify the use of an RBP that is not licensed
domestically, and the same RBP should be used in all comparability studies of a given
SBP.
The choice of the use of the RBP should rest with the qualification of the RBP,
and it has nothing to do whether it is approved locally. The RBP must be the first
product approved using a full dossier.
2.1.3. STATISTICS
It may not be possible to set a definite number of lots to be analyzed in the
comprehensive comparability exercise. The number of lots used to demonstrate similarity
for each quality attribute and to establish the range of SBP specifications should be
sufficient to allow for meaningful comparison with the RBP.

10
https://www.wto.org/english/docs_e/legal_e/27-trips_04d_e.htm
 The number of lots tested is well-defined in the equivalence testing model; for
example, 6-10 lots must be analyzed to establish an acceptable beta value. It is not
possible to select what is "sufficient" without referring to statistical modeling.
Statistical methods usually deal with means. The means may change within the
acceptability range. The use of a descriptive statistical approach to establish ranges for
quality attributes in the context of comparability is widely accepted.
Variation of means vary within the acceptable range has no significance since the
acceptable range is established before testing. Descriptive statistics have little
value unless a comparison is made between the two sets of data.
The use of statistics in defining comparability is still at an empirical stage in most
jurisdictions. It is important to realize therefore that statistical tools, while helpful in
supporting conclusions about similarity, should not be used as the sole basis for
decision-making on biosimilarity for the purpose of marketing authorization approval.

o Statistical methods are well-defined and must be used to support similarity.

Statistical tools are never the basis for decision-making; it is the result they
produce, such as bioequivalence test, equivalence range, and non-inferiority
efficacy testing.
2.1.4. STABILITY TESTING
Stress stability testing is necessary for an SBP in order to further investigate appropriate
conditions for shipping and storage unless these conditions are covered by accelerated
stability studies. In general, comparative stress stability testing of the SBP and RBP does not
provide added value.
Stress stability testing for SBP is the most important criteria of establishing
biosimilarity, since stress testing provides a comparison of structural attributes, not to
test for stress during shipping and storage—that study is conducted under a different
protocol.
Real-time/real-temperature stability tests will determine the storage conditions and shelf-life
of the SBP. These conditions may or may not be the same as those of the RBP. Comparative
real-time, real-temperature stability studies between the SBP and RBP are not required.

The storage conditions of the SBP must be the same as that of the RBP. Comparative
real-time/real-temperature stability testing is required to assure similarity of the SBP
with the RBP. A minimum comparative testing for six months is required for
regulatory submission
2.1.5. FUNCTIONAL ASSAYS
In vitro functional assays (usually cell based) may be helpful in understanding the
significance of a difference detected during analytical testing [sic] is not appropriate since a
functional assay may only point to a specific mode of action.
Multiple modes of action are tested using multiple functional assays; these tests
remain a most useful tool.
 The sensitivity of some functional tests, such as reporter-gene-based assays, has been
increased to the degree that they do not correspond to the physiological situation of the
observed difference.
If an assay does not resolve a residual uncertainty, there is no need to include it in the
plan. There is always a correlation developed between the assay response and its
clinical meaningfulness, making this stated assertion irrelevant.
The in vitro functional assays using cells from a specific patient population may also be
helpful in the interpretation of correlation between an in vitro functional assay and
differences in the analytical testing.
This approach is neither practical nor based on scientific grounds. There is little
justification for selecting tissues from a patient population as it will lead to inevitable
variability.
In general, in vitro functional tests are more sensitive than clinical studies in detecting
differences between the SBP and RBP.
Functional assays and clinical studies serve different purposes; one cannot replace
one with the other.
The results of physicochemical and functional tests should be considered when planning the
clinical comparability programme, especially pharmacokinetics (PK), pharmacodynamics
(PD) and immunogenicity studies.
The clinical pharmacology studies have fixed statistical protocols that cannot be
changed based on any prior finding; it is the final analysis that decides if any earlier
observations constitute residual uncertainty.
2.1.6. NONCLINICAL ASSESSMENT
If in vivo animal studies are considered to be indicated, the developer should clarify the
availability of relevant animal models.
The availability of a suitable species does not need any clarification or constitutes
consideration for a waiver of the use of a specific species.
If the drug substance of a candidate SBP shows specific pharmacological activity only in
great apes, the developer should carefully weigh up the need for in vivo studies to avoid
pharmaco-toxicological testing in these species.
If a drug substance shows specific pharmacologic activity in great apes, the need for
testing is already defined. It does not require any further consideration or allowance
to waive these studies.
Studies with non-human primates should be avoided if possible.
All animal testing should be avoided if possible—there is never a need to conduct any
redundant study.
In general, the additional value of in vivo nonclinical studies for the demonstration of
comparability of SBP and RBP is questionable when previous physicochemical, structural
and in vitro functional tests have demonstrated their similarity.
While animal toxicology studies can be waived, nonclinical PK studies in monkeys
 are recommended for highly complex mAbs because of the inability of other methods
to detect subtle structural differences.
Nonclinical PK studies with the SBP, if warranted, should be justified on the basis of RBP
data and the potential interference of anti-drug antibodies (ADAs) in the interpretation of
such a study.
Nonclinical PK studies are justified based on the complexity of the molecule, not
based on RP data or potential interference of ADAs that are not expected in a single
dose exposure during the course of disposition.
If product-inherent factors that have an impact on PK and/or biodistribution (such as
glycosylation or pegylation) cannot be characterized sufficiently at a quality and in vitro
level, the manufacturer should carefully consider whether in vivo animal PK and/or PD
studies should be performed in advance of clinical PK/PD testing.
While the PD and immunogenicity responses are not reliably studied in an animal
species, the disposition characteristics subject to known or unknown structural
elements are best studied first in an animal model before conducting a similar study in
humans. A good example is a need for PK studies in monkeys for monoclonal
antibodies.
Since relevant PK/PD data are obtained in humans, nonclinical PK/PD studies generally
have little added value.
The nonclinical PK/PD studies have a different purpose, to establish sufficient safety
to administer the drug to humans.
The PK and/or PD of the SBP and RBP in animals should be compared quantitatively,
including, if feasible, a dose-response assessment that includes the intended exposure in
humans.
There is no rationale for a dose-response assessment because there is little correlation
between the human dose and a comparable animal dose.
In vivo assays, if warranted, may include the use of animal models of disease to evaluate
functional effects on PD markers or efficacy measures. PK measurements may need to be
performed in parallel in order to interpret the study results.
Disease models are challenging to construct in animal species; an arthritic model can
be used, but such models are rare for other diseases; the PD markers in animals are
not an indicator of human PD response. Any correlation between PK and PD in
animals cannot be extrapolated to humans.
In vivo nonclinical studies should be considered if there is one or more of the following: a
significant functional difference suggested by nonclinical in vitro studies; an excipient used
in the formulation of the SBP which may justify a more thorough nonclinical programme to
assure the safety of the excipient in its intended route of administration; a new expression
system or purification process in the manufacturing process, leading to a significant change
in process-related impurities; a narrow therapeutic window for the drug substance.
Prior to considering in vivo nonclinical study, the issues related to physicochemical
attributes and in vitro study results must be resolved as these cannot be resolved on
 the basis of in vivo nonclinical studies.
The root cause of a difference in immunogenicity should always be investigated, even if the
SBP appears to be less immunogenic. First, the ADA assay should be re-evaluated for
possible bias.
All assays should be declared using suitable and sensitive before their use in any
study; there is no need to re-evaluate any bias or question lower immunogenicity of
the SBP. The impact of lower immunogenicity, if it alters disposition kinetics, will be
observed in clinical PK/PD studies.
In principle a SBP can submits relevant data to support the application for additional
indications.
No additional indications are allowed within the category of evaluation as a
biosimilar product.
2.1.7. CLINICAL PHARMACOLOGY ASSESSMENT
The PK of the SBP and RBP are often compared in single-dose studies involving healthy
volunteers, when this is appropriate and depending on the nature of the treatment.
The nature of treatment has little to do with the design of the PK study. A PK study
serves the purpose of defining any subtle structural differences the study is conducted
regardless of the treatment modality.
The criteria used in the demonstration of bioequivalence (that is, 90% confidence interval
(CI) of ratios of SBP to RBP) of orally administered and chemically synthesized small
molecules are often used for comparative PK studies of SBPs and RBPs in the absence of
relevant historical data.
This model is used regardless of any historical data.
If the PK comparability criteria are met but the exposure to the SBP is significantly lower or
higher, meaning that the CI of the SBP is entirely within either the higher or the lower side
of the equivalence range, a root cause analysis, and possibly new data, may be needed.

This statement is in completely misunderstands statistical testing. If the PK study

meets the criteria of acceptance, there is no delineation of higher or lower exposure,
and there should never be any need for new data or any root cause analysis.
It is recommended that steady state PK should be measured in the repeat-dose safety and
efficacy studies. This may mitigate concerns when PK differences are observed after a
single-dose study.
Steady states are only possible in repeat-dose studies, and such studies are not
recommended for products administered as a single dose and even for products
administered as multiple-dose because the steady-state levels are substantially less
sensitive than the levels in a single-dose study.
2.1.8. CLINICAL EFFICACY ASSESSMENT
A non-inferiority design may be used if superiority can be excluded. In both cases, the
acceptance range is defined by previous clinical trials with the RBP and the outcome should
 be that the difference is not clinically meaningful.
Removing the superiority range shows a lack of understanding of statistical modeling;
it should not be allowed since a higher activity of the product may also mean higher
safety risks.
In certain cases, comparative PK/PD studies may be appropriate (in place of
comparative efficacy testing), provided that 1) the PK and PD properties of the RBP are well
characterized, 2) at least one PD marker is a marker linked to efficacy (e.g., an accepted
surrogate marker for efficacy), and 3) the relationship between dose/exposure, the relevant PD
marker(s) and response/efficacy of the RBP is established.
The WHO guidance provides two examples of the above waivers: euglycaemic clamp
studies for insulin, and an absolute neutrophil count and CD34+ cell count in healthy volunteers
for GCSF. The EMA provides a less restrictive approach that should replace the WHO
suggestions: The selected PD marker/biomarker is an accepted surrogate marker and can be
related to patient outcome. There may be PD-markers that are not established surrogates for
efficacy but are relevant for the pharmacological action of the active substance, and a clear dose-
response or a concentration-response relationship has been demonstrated. In exceptional cases, if
physicochemical, structural and in vitro biological analyses and human PK studies together with
a combination of PD markers that reflect the pharmacological action and concentration of the
active substance, can provide robust evidence for biosimilar comparability.
3. CONCLUSION
The impact of misconceptions in the WHO guidance have been wide and resulted in the
marketing of a large number of "biosimilar" products with questionable safety and efficacy, or
added complexity to the development and approval of biosimilars as a few examples list below:
 The WHO guidance does not require evaluation of the most frequent side effect of
local reactions at the injection site, mostly caused by the formulation and not by
the active molecules.
 China, India, and Russia require local clinical trials without a suitable
comparative efficacy protocol and refuse to accept data collected in more robust
protocols from studies conducted in other countries.
 Iran, Peru, Nigeria, and Venezuela do not require comparative clinical testing11.
 Indian guidance states: “…if the pharmacologically relevant animal species is not
available and has been appropriately justified, toxicity studies need to be
undertaken either in rodent or nonrodent species…” Since the testing in monkeys
is not allowed for religious reasons in India, every mAb approved in India was in
rodents that do not have any response receptors, a significant safety risk for all
biosimilars approved in India. Before the implementation of the Indian guideline,
20 biosimilars were approved under an ad hoc basis.
 Indian guidance states: “Antibody response to the Similar Biologic should be
compared to that generated by the reference Biologic in a suitable animal model.
The test serum samples should be tested for reaction to host cell proteins.” There
are no suitable animal models to test immunogenicity that is relevant to humans.
 Japan excludes polyglycans (e.g., low molecular weight heparin) and synthetic
11
http://www.gabionline.net/Reports/Guidelines-for-biosimilars-around-the-world
 peptides. Comparative stability studies with reference biologics is not mandatory
in Japan; so is the case with toxicology studies, wherein impurities need not be
evaluated through nonclinical studies.
 Iran (2014) allows a biosimilar product to serve as reference product if the
original reference product is not registered in Iran or if a biosimilar has been
approved in EU or USA. Iran does not require side-by-side comparative
accelerated stability testing, nor comprehensive clinical trial relying on national
post-marketing surveillance data for drug safety.
The WHO is strongly urged to revise its guideline, bringing it closer to the current
scientific and logical thinking that is necessary to assure to assure safety and efficacy of
biosimilars.

4. ABOUT THE AUTHOR

Sarfaraz K. Niazi is an adjunct professor of pharmaceutical sciences at the University of
Illinois, Chicago; he has contributed several regulatory guidance contributions to many
regulatory agencies including the US FDA. He has published the largest number of books on
biosimilars and has developed a large number of biosimilars. He serves as a consultant in the
field of biotechnology-based medicines. www.niazi.com; niazi@niazi.com