Joseph Glajch, Ph.D and Jim Anderson, Ph.D Jan 28, 2016 · Implementation of the Current Approach •Highlight both the process and the challenges •Evaluation of 5 lots of material

Similarity Assessment of Biosimilars. The Past, Present and Future State

Joseph Glajch, Ph.D and Jim Anderson, Ph.D

Jan 28, 2016

Disclaimer

• The views and opinions expressed in the following PowerPoint slides are those of the individual presenter and should not be attributed to WCBP or CASS, its directors, officers, employees, volunteers, members, chapters, councils, communities or affiliates, or any organization with which the presenter is employed or affiliated.

• These PowerPoint slides are the intellectual property of Momenta Pharmaceuticals and are protected under the copyright laws of the United States of America and other countries. Used by permission. All rights reserved.

2 © 2016 Momenta Pharmaceuticals, Inc. All rights reserved.

Momenta Pharmaceuticals Overview


Corporate Info

• Founded in 2001; IPO 2004

• Located in Cambridge, MA

• >250 employees

An Advanced Analytic Platform

• Expertise in high-resolution analytics, biological characterization and process engineering

Driving Potential in Three Areas

• Complex Generics

• Biosimilars

• Novel Drugs

The Past – Aspects of Complex Generics Applied to Biosimilars

• Regulatory Requirements for Approval of Biosimilars used Criteria Associated with Complex Generic Products • Enoxaparin Sodium • Glatiramer Acetate

• Momenta developed and implemented the strategy for approval of both products • Comprehensive characterization of reference product • Development of acceptance criteria based on observed limits of the

reference product

• Momenta developed a pathway to compare complex populations • How to decide what to measure • How to use resulting data

• Including use of the data to inform process development

• Valuable to take a closer look


GlatopaTM (Glatiramer Acetate) Complex Mixture Approved by the FDA

Treatment for RR-MS

Synthetic polypeptides, containing four amino acids,

polymerized without a template

Complex mixture of many different chains (theoretical

possibility of 1029 chains of varying length)

Molecular Weight = 5,000 – 9,000 Daltons

Momenta technology identified “process signatures” of Glatiramer Acetate and used them to demonstrate equivalence to the Reference Listed Drug


Establishment of Analytical Criteria for Complex Generics

• Thorough Analysis of Reference Listed Drug

• Criteria Based on Observed Limits for Reference Listed Drug • Large number of lots available (n ≥ 60)

• Core principle

• All reference product lots (tested within expiry and stored appropriately) are valid AND

• The observed limits do not represent the complete distribution

• Ranges observed represented safe and effective materials

• Process controls developed to center each attribute on the mean value of the Reference Listed Drug


Glatopa Equivalent to RLD Developed by Integration of Process Development with Thorough Analysis

• Relative amino acid levels at the C-termini of glatiramer acetate.

• Molar mass distributions of Glatopa and Copaxone, as measured by size exclusion chromatography.

• Ranges for Mn, Mw, Mz, Ip established

Limits based on RLD analysis

Values for individual lots


The Present - Approach to Biosimilars in the US

• Test all attributes to the highest level possible with available technology

• Use a Risk based Approach to rank attributes • High, Medium or Low

Criticality

• Apply different statistical approaches for each Rank • Tier 1, Tier 2 and Tier 3

analysis


Implementation of the Current Approach

• Highlight both the process and the challenges

• Evaluation of 5 lots of material produced via the intended commercial process

• Program in early stages of development

• Note FDA has indicated that the n value should be near 10

• General Process • Defined high, medium and low criticality attributes

• Evaluate via three statistical tiers as discussed by FDA* • Tier 1 = equivalence testing (means differences ± 1.5 σ)

• Tier 2 = Quality Range (mean ± 3 σ)

• Tier 3 = graphical or tabular data

*Tsong, Y et al., Development of Statistical Approaches for Analytical Biosimilarity Evaluation, Power Point presentation at the DIA/FDA Statistics Forum, April, 2015.


Consider Evaluation of Potency Data

• Defined as a Critical Attribute • Agreement by FDA

• Therefore - apply Tier 1 test to compare biosimilar to Reference Product (RP)

• The raw data • Note there are 5 biosimilar lots and 28 US and 17 EU Reference Lots


Range of values due to a combination of

Process and Method Variability

What is the Tier 1, Equivalence Test?

• Testing against the null hypothesis:

Ho: µB - µR ≤ -δ or µB - µR ≥ δ

• Comparison of means values of biosimilar and the RP • Confidence interval for the difference of means should fall within the defined range

• Confidence interval is adjusted for the number of lots, per FDA guidance

• The equivalence margin (δ) is defined to be 1.5 σ by FDA

• FDA requires a balanced comparator set • Meaning the same number of Reference Product lots as biosimilar lots

• N = 5 in this case

• However, the standard deviation used for the RP is based on all lots

• Challenge to choose lots of Reference Product in a fair way • Momenta approach on the next slide

• Note the underlying assumption that the distributions are normal!


• Run a simulation of all the permutations of “n=5” from the full RP set

• N of 28 in this case

• Choose the 5 lots that are at

the center of the Mean (X) and SD (Y)

• Use those 5 lots for Tier 1

comparison

Selection of Reference Lots for Tier 1 Comparators

Potency Simulation of all Possible Combinations of N=5


Tier 1 Results, Biosimilar Potency

• Potency met ± 1.5 σ acceptance criteria for Equivalence Test • Could also apply a 90% CI for the n of 5 comparison • Does not provide a view of variability of reference properties • Note choice of the five reference lots can influence the outcome of the test

• Interesting to note difference in mean values for reference product by region are

non-zero


What will be gained moving to a n of 10?

Plot of Confidence Interval for difference of two means (per FDA Type 1) with sample size N Two identical normal distributions (µ = 0 and σ = 1) Red lines show the 90% CI for the difference of means for a given N of 3 to 10. The blue lines show the same, but adjusting the CI per FDA Tier 1 approach

Plot of the Confidence Interval width with sample size N For a N of five, CI width is approximately 1.3 For a N of ten, CI width is approximately 0.8

A larger N is always better BUT there are diminishing returns for a significant cost! A larger CI translates into greater risk for the sponsor to meet criteria


Example Tier 2 Results, Glycan Species

• Glycan Species were Defined to be Medium Criticality Attribute • Apply the Tier 2, Quality Range Evaluation • Criteria are that the Range is the MeanRP ± X*σ

• X is test dependent

• Glycan species met “mean ± 3σ” target for Quality Range

• Advantage in that one can use all of the RPP data

• Better view of comparison to diversity of RP lots • Similar to criteria applied to complex generics

Biosimilar Biosimilar

US RP v. Biosimilar EU RPP v. Biosimilar US RP to EU RPP


Example Tier 3 Analysis, Charge Heterogeneity by cIEF

Flow Chart Developed for Consistent Analysis

cIEF Profile met target profile comparison

• cIEF Profile was defined to be a Low Criticality Attribute

• Apply Tier 3, Graphical Evaluation


The Advantages of the Current Approach

• Clear guidance on how data will be evaluated by FDA • Consistent approach for all applicants

• Tier 1 and Tier 2 analysis take into account method variability • Working in “sigma space”

• Forces normalization of most critical attributes for applicants • Agreement with FDA as part of the review process

• FDA has indicated that only a small number of attributes should be considered highly critical / Tier 1 analysis


Some Challenges with the Current Approach

• Many highly critical factors do not fit Tier 1 analysis • Sequence, Aggregates

• Head-to-head testing may be a more meaningful test design for biological properties

• Tier 1 analysis drives a much higher n value • To meet criteria with a smaller n, tighter limits must be applied on the biosimilar than

the reference product

• Sponsors required to invest early to develop final process materials (even scale?) for purpose of discussions with Agency

• Mixing of process control with similarity • Industry already required to assure process control for product approval

• Tier 3 • Are we missing significant information simply due to complexity of the test data?

Utilization of a simple model to a complex question leads to subjective decisions


The Future for Biosimilar Analysis

• Are the current characterization methods sufficient? • Comfortable - but opportunity is lost if we only use conventional protein

analytic methods

• Suggest that Industry and Agencies should look to • Better utilize “Tier 3 / Descriptive” Methods

• Integrate complex biological models into the analysis

• Develop and explore alternate ways to compare data

• How close is close enough? By what measure?

• Develop an approach based on the “science of comparison”


Science of Comparison

• Similarity is not Comparability • Comparability exercises are designed to show that a deliberate change to a

well controlled process does not change significant product quality attributes • Similarity aims to show (ideally) no measureable difference between a

reference product and a biosimilar • Fundamentally different approach

• Momenta applies the concept of “the science of comparison” to evaluate similarity


vs.

Complex Readouts Lead to Increased Information – Example cIEF Data

• Reduced • Relatively simple

• Easy to set acceptance criteria

• Information lost

• Non-reduced • Much more complex

• Comparison more challenging

• Data can be used to demonstrate similarity


Non-Conventional Approach for Evaluation of Higher Order Structure

• Traditional Approaches • Need to be employed – but can be augmented with alternate approaches

• NMR Approach to Compare HOS • Large proteins yield broad lines and high overlap of peaks • Specific features related structural features • Overall spectrum provides a fingerprint

y = 1.0229x - 0.0012 R² = 0.9978

0

1

2

3

0 0.5 1 1.5 2 2.5 3

Re

lati

ve P

eak

Inte

nsi

ty

Bio

sim

ilar

Relative Peak Intensity for RP

Comparison between RP and Biosimilar at 35 °C

• 20 representative peaks across the spectrum chosen

• RPI (Relative Peak Intensity) plotted for pairwise samples comparison

• R2 is a measure of similarity

Amide and aromatics peaks Methyl peaks


NMR Approach to Evaluate Perturbations to HOS

Reference Product

Biologic Product

Stress Agent 1

Stress Agent 2

Stress Agent 3

Stress Agent (n)

New State 1

New State 2

New State 3

New State (n)

Comparison of the change between Reference Product and Biosimilar Assure Starting Point is the Same

Same perturbation should yield the same result (if they start out the same …)


Example of Stressed Conditions Measured by NMR

1D Proton NMR at 35 ⁰C

Temperature 1D Proton NMR at 55 ⁰C

D2O and Temperature 1D Proton NMR at 80% D2O and 55 ⁰C

Reagents

[Tb(DPA)3]-3

TEMPO-L

Stress Condition

Binds positive patches on the surface Binds hydrophobic patches on the surface


Perturbations for Each Peak

D RPI

D RPI

-0.3

-0.25

-0.2

-0.15

-0.1

-0.05

0

0.05

0.1

0.15

RPP Lot 1 Lot 2 Protein 1 Protein 2

Peak 1

Temp Tb TempoL D2O

-0.3

-0.25

-0.2

-0.15

-0.1

-0.05

0

0.05

0.1

0.15


Peak 6

Temp Tb TempoL D2O

-0.3

-0.25

-0.2

-0.15

-0.1

-0.05

0

0.05

0.1

0.15


Peak 9

Temp Tb TempoL D2O

-0.3

-0.25

-0.2

-0.15

-0.1

-0.05

0

0.05

0.1

0.15


Peak 15

Temp Tb TempoL D2O


Integration of Biological Characterization

• Example Characterization Scheme for a Biosimilar • Tailor for each product


Comparison of Conventional vs High Resolution Analysis of PK Data

• Conventional High Resolution Approach

Time

Co

nce

ntr

atio

n

Population “less resolved”

N-glycosylated at Asn-297

Time

Co

nce

ntr

atio

n

Individual glycoforms “more resolved”

Hundreds of possible glycoforms


How to analyze data?

• Can build more complex models for statistical analysis • Directly applicable to more complex biological models that allow for

comparison of large data sets

• Can be combined with physiochemical data

• Tools do exist • Principle Component Analysis

• Multidimensional scaling (MDS)

• Hierarchal Clustering

• “Multivariate T-test”

• Can also view data in an overall view • Integration of many data types into a “heat map”

Contend that the more complex analysis is both necessary and sufficient to assign similarity between materials


Thank You for Your Attention

QUESTIONS?


Documents

Joseph Glajch, Ph.D and Jim Anderson, Ph.D Jan 28, 2016 · Implementation of the Current Approach •Highlight both the process and the challenges •Evaluation of 5 lots of material