Critical Considerations for Analytical Similarity Assessment

Jennifer Liu

CASSS CMC Strategy Forum Japan 2015

Development of Biosimilars: Technical Aspects

Biologics Are Larger and Structurally More Complex Than Chemically Synthesized Drugs

2

Generic Biosimilars 1.0 Biosimilars 2.0

Same Structure Similarity in Structure

Monoclonal antibodies have multiple biological functions

Potency Reflective of

Target Binding and

Mechanism of Action

Binding Region for FcRn

(PK)

Binding Region for FcR,

Effector Function

(CDC, ADCC)

N-glycan

Heavy chain

Intra-chaindisulfide

Inter-chaindisulfide

Light chain

FcRn

EU

ABP

US

CDC

EU

ABP

US

ADCC

EU

ABP

US

Potency

Biological and functional assays are used to investigate structure/functional differences which are clinically meaningful for biosimilarity

Human

PK and PD

Non-clinical PK/PD and toxicology as appropriate

Match all predicted functions and confirm no enhancement of

functions absent

Demonstrate analytical similarity

Biosimilar product development begins with establishing target quality product profiles

Known mechanism of actions, biological functions, safety, and immunogenicity profiles

Define critical quality attributes for the

reference product

Characterize reference product to establish targets and ranges for critical product quality attributes

Establish product quality profiles based on the

target reference product

Match reference product profiles with greater emphasis on matching all biological functions

Develop biosimilar products to match the

target reference product

Additional clinical data

Demonstrate biosimilarity

in a step-wise manner

Biosimilar development is reverse

engineering of the reference product

5

Stability

Primary structure

Biological function

Particles and

aggregates

Process-related

impurities

General properties

and excipients

Higher order

structureProduct-related

substances and

impurities

Attributes related to the amino acid

sequence and all post-translational

modifications, including glycansBiological and functional

activities, including receptor binding

and immunochemical properties

Quantitative levels of

product variants and

their identities

Degradation profiles

denoting stability

Subvisible and submicron particles

and aggregates of various sizes

Integrity of the secondary, tertiary,

and quaternary structure

Properties of the finished

drug product, including

strength and formulation Impurities from host cells

and downstream process

Comprehensive analytical similarity is critical foundation for biosmilarity

Similarity assessment may use risk-based tier approach and statistical analysis

6

High

Low

Tsong, Y (FDA CDER) DIA/FDA Statistics Forum 2015

Consider criticality ranking of quality attributes with regard to their potential impact on activity, PK/PD, and immunogenicity. Christl, L (FDA CDER) FDA Oncologic Drugs Advisory Committee meeting, 2015

StatisticalRigor

CriticalityRanking

An approach to statistical equivalence testing for tier 1 attributes

Dong, X. (FDA) FDAMcCamish, M. (Sandoz)

This case study is provided for the sole purpose of illustrating FDA’s statistical approach for similarity assessment as presented at FDA ODAC meeting on Jan. 7 2015. No conclusion of the data is made by the presenter.

• Pass equivalence testing if 90% confidence interval of the true mean difference is within equivalence margins

• Equivalence margins = 1.50σ where σ is the true standard deviation of the RP

• Statistical power (sensitivity to detect difference) dependent on number of lots

Number of biosimilar batches

Po

wer

Challenges in the tier 1 equivalence approach

• Panel of Tier 1 quality attributes should be consistent for all sponsors seeking biosimilar product approval to the same original product

• Relationship between quality attributes and safety/efficacy in patients may not be completely understood

• Specific roles for effector function as part of the clinical mechanism of action are not entirely clear

• Sufficient number of lots are required for meaningful statistical analysis

• Small number of biosimilar batches reduce statistical power to detect difference

• Small number of reference product lots can not estimate true variability

• The true variability (standard deviation) in the reference product is difficult to estimate

• Different reference product lots may come from the same drug substance lots → underestimating variability

• Differences due to manufacturing process changes → overestimating variability

Use appropriate characterization techniques in well-designed analytical studies is critical

9Slide from Kozlowski, S (CDER) presentation at 2014 Biomanufacturing Technology Summit, Rockville, MD, June 13, 2014

Reliability - Methods should be qualified and suitable for intended purposes

Relevance - Methods provide meaningful information and may predict clinical performance

Resolution - Methods should be sensitive and capable of resolving differences

Material age - Attributes impacted by material age should be considered in the assessment

Biosimilar

Reference product

Critical

Glycans(e.g. high-Man)

Understanding the biological relevance of measured quality attributes is important

10

Confidence in similarity

Measuring the quality attributes by retaining the important structural characteristics of the attributes

Increase understanding for the biological relevance of different glycan attributes and ensure the critical attributes are controlled properly in order to match biological functions

Glycan

Map

profile

Monosaccharide

analysis

Re

leva

nce

Reliability

0.0

10.0

20.0

30.0

4.0 6.0 8.0 10.0 12.0 14.0 16.0 18.0 20.0 22.0 24.0 26.0

min

A2

G0

F

A1

G0

F

M5

A2

G0

A1

G1

F

A2

G1

F(a

)A

2G

1F

(b)

M6

A2

G2

F

M7

Monosaccharide Analysis

Oligosaccharide Analysis (Glycan Map)

Re

leva

nce

Reliability

NK92

ADCC

FcgRIIIa

binding

PBMC

ADCC

Glycan

map

Confidence in manufacturing

Knowing the correlation of orthogonal methods helps develop a robust control strategy

11

High resolution and reliable methods suited for process control can guide the process and product development.

Orthogonal assays with increasing biological relevance to discern clinically meaningful differences and confirm functional similarity

Response (LU))

0.0

10.0

20.0

30.0

4.0 6.0 8.0 10.0 12.0 14.0 16.0 18.0 20.0 22.0 24.0 26.0

min

A2G

0F

A1G

0F

M5

A2G

0

A1G

1F

A2G

1F

(a)

A2G

1F

(b)

M6 A2G

2F

M7

Chung, mAbs, 326-340, May 2012

ADCC vs. Afucosylated glycan

Critical glycan attributes are modality and molecule dependent

12

Sensitivity to difference

Impact toepoetins

Glycan TypeImpact to

mAbSensitivity to

difference

HighIncrease

clearanceNo glycan No ADCC Low to medium

Low to mediumIncrease

clearanceBisecting GN Increase ADCC Low to medium

MediumIncrease

clearanceHigh mannose and ADCC Medium

HighIncrease

clearanceTerminal Gal Increase CDC Low

HighDecrease clearance

Sialic acidAnti-

inflammatory and ADCC

Low

N/A N/A Afucosylated Increase ADCC High

0.0

10.0

20.0

30.0

4.0 6.0 8.0 10.0 12.0 14.0 16.0 18.0 20.0

min

A2G

0F

A1G

0F

A2G

1F

(a)

A2G

1F

(b)

M6

A2G

2F

M7

M5

A2G

0

A1G

1F

0

50000

100000

150000

200000

250000

300000

350000

400000

5 7 8 9 10 11 12 13 14

[U/m

g]

Isoform Number

In Vivo Bioactivity

Isoform1011121314

Biosimilar epoetins: how similar are they?

RP RP

Schellekens H. Eur J Hosp Pharm Sci. 2004;3:43–47.

Fingerprint-like similarity versus Fingerprint-like methodsFDA Definition of Fingerprint-like:

a term to describe integrated, multi-parameter approaches that are extremely sensitive in identifying analytical differences.

FDA DRAFT GUIDANCE “Clinical Pharmacology Data to Support a Demonstration of Biosimilarity to a Reference Product”

13

Koslowski, presentation at CASSS WCBP Conference, Washington DC, Jan, 31 2013

Analytical methods which may provide fingerprint-like profiles

• Methods investigate the overall conformational integrity• 1-D NMR

• 2-D NMR

• Crystallography

• Antibody conformation array

• H/D exchange

• Methods investigate the heterogeneous characteristics• Peptide map

• Glycan map

14

Can any one of these methods provide prove fingerprint-like similarity in its own right?

1H NMR provides fingerprint pattern

15

L. Poppe, Anal. Chem. 2013, 85, 9623-9629

V. Visser et. al, BioDrugs, 07 May 2013

Similarity is assessed based on pattern similarity and little information can be discerned on the differences and their impact to purity, safety, and efficacy

• Applications for monoclonal antibodies and less complex proteins have been shown

• Visual profile comparison could be subjective. An objective comparison for spectrum similarity would require complex mathematical algorithm

2D-NMR has been applied for smaller non-glycosylated protein

16

• Several publications on filgrastim (small non-glycosylated protein of ~ 19,000 Dalton) 1H-15N 2D-NMR spectral (pattern) similarity

• Provides higher resolution in structural information compared to 1D-NMR down to the single amino acid level

May not be suitable for glycoprotein and monoclonal antibodies which is structurally more complex than filgrastim

Sandoz presentation to ODAC, 07 Jan. 2015M. Levy, Anal. Bioanal. Chem. 22 Nov 2013 R. Brinson, AAPS, 2013D. Hodgson, WCBP, 2012

Crystallography has been applied to investigate sub-domains of large proteins

17

Does not investigate heterogeneous populations or discern sub-populations which may have different biological activities

• Typically require changing formulation to allow crystallization of protein of sub-domains of mAb

• Crystallization process generally serves as a purification step selecting the most “homogeneous” population, and could miss low abundant sub-populations

S. K. Jung, mAbs 6:5, 1163--1177; September/October 2014V. Visser et. al, BioDrugs, 07 May 2013

H/D Exchange provides site-specific information on structural differences

18

Interpretation of complex results requires experienced subject matter experts

• H/D exchange require long experimental time with multiple testing time points.

• Investigate differences in the exchange rate comparing two products under the same

condition. Results could be informative in regards to detailed structural differences at

specific locations

Z. Zhang Anal. Chem. 2012, 84, 4942−4949

Antibody conformation array could identify new epitopes due to changes in conformation

19

S. K. Jung, mAbs 6:5, 1163--1177; September/October 2014

X. Wang, CHI’s 6th Annual Biotherapeutics Analytical SummitComparability for Biologics & Biosimilars March 13, 2015

• Evidences of regional conformational change may predict impact to biological activities mediated through specific antigen or receptor recognition

• Reagents maybe high cost and experiment is relatively low throughput

Orthogonal to biological assays in investigating structure/functional differences

Capabilities and gaps of these fingerprint-like methods

20

Technique Capabilities Gaps

1D-[1H]-NMR Sensitive to detect differences for both small and large proteins including monoclonal antibodies

Difficult to interpret results and primarily rely on pattern similarity.May not determine specific site and levels of differences

2D-[1H-15N] HSQCNMR

Sensitive to detecting differences down to amino acid levels for small proteins

May not be suitable for complex biologics, such as glycoprotein and monoclonal antibodies

X-ray crystallography Primary application for smaller protein or fragments of monoclonal antibody. Assess sub-populations which are homogenous in nature

Limited resolution to structural differences for large protein. May omit minor components with unknown safety or efficacy impact.

H/D exchange Provide detailed structural differences at specific locations

Long experimental time. Interpretation of complex results requires experienced subject matter experts

Ab conformational array

Data analysis is less complex compared to the other fingerprinting methods.Provides regional structural information

Require product-specific reagents which could be costly or may not be available

These methods might provide a partial 'fingerprint' of higher order structure

but are not sufficient to demonstrate 'fingerprint-like similarity’.

Summary

• Comprehensive and well-designed similarity assessment is the foundation for biosimilarity

• US FDA recommends risk-based approach for tier assignment and statistical analyses

• Challenges associated with the current approach require further guidance

• Biosimilar product need to demonstrate similar functions using biological assays relevant to potential clinical mechanism of action

• There is no one method alone capable of indicating “fingerprint-like similarity”; it requires a collection of methods, including bioactivity assays.

• “Fingerprint similarity” presumably is matching all structural/functional assessments while using comprehensive sensitive methods

21

Simon Hotchin

Margaret Karow

Sundar Ramanan

Richard Markus

Gino Grampp

Acknowledgements

22

Leszek Poppe

Pavel Bondarenko

Zhongqi Zhang

John Gabrielson

Brad Jordan

Izydor Apostol

Linda Narhi