APPLYING SCIENTIFIC CONSIDERATIONS AND STATISTICAL APPROACHES IN ANALYTICAL SIMILARITY ASSESSMENT

Transcription

1 APPLYING SCIENTIFIC CONSIDERATIONS AND STATISTICAL APPROACHES IN ANALYTICAL SIMILARITY ASSESSMENT JENNIFER LIU CASSS CMC STRATEGY FORUM EUROPE 2017, MAY 22-24

2 BIOSIMILAR DEVELOPMENT BEGINS WITH THOROUGH KNOWLEDGE OF THE REFERENCE PRODUCT Reference product knowledge Based on known mechanism of actions, safety and efficacy profiles Based on knowledge for the same class of molecules Clinical safety and efficacy Human PK and PD Define critical quality attributes Understand impact of attributes to bioactivity, pharmacokinetics, safety and immunogenicity Determine critical quality attributes to risk rank attributes Nonclinical PK/PD and toxicology as appropriate Match biological functions and pharmacological properties Demonstrate analytical similarity Establish Biosimilar s QTPP ranges Characterize reference product for attribute profiles Establish QTPP based on reference product profiles Refine QTPP as knowledge increases Increase reference product knowledge over time with more lots 2 Amgen Proprietary For Internal Use Only

3 COMPREHENSIVE ANALYTICAL SIMILARITY ASSESSMENT IS THE FOUNDATION FOR APPROVABLE BIOSIMILARS Amino acid sequence and post-translational modifications, eg. glycans Primary structure Target binding and immunochemical properties Secondary, tertiary, and quaternary structure Impurities from host cells and downstream process Higher order structure Processrelated impurities Product-related substances and impurities Stability General properties and excipients Biological function Particles and aggregates Quantitative levels of product variants and their identities Degradation profiles denoting stability Subvisible, submicron particles and aggregates of various sizes Liu, BioDrugs (2016) Properties of the finished drug product including strength and formulation attributes 3 Amgen Proprietary For Internal Use Only

4 HOW TO ASSESS COMPLEX DATASET WITH OBJECTIVITY Objective assessment is based on criteria that can be measured against Subjective assessment requires interpretation by a subject matter expert Liu, BioDrugs (2016) 4 Amgen Proprietary For Internal Use Only

5 SCIENTIFIC CONSIDERATIONS FOR ASSESSMENT CRITERIA Reference product data dependent Reference product attribute knowledge Stability-indicating product attributes Manufacturing process controls in place Quantitative results from reference product lots Qualitative (profile) comparison with reference product side-by-side Based on theoretical values of the reference product (chemical formula) Consider reference product label information Stability trends of each product should be part of the assessment Material ages at testing need to be considered Attributes and methods unique to individual manufacturing process Attributes important for safety and routinely controlled by specifications 5 Amgen Proprietary For Internal Use Only

6 STATISTICAL COMPARISON MAY INCREASE OBJECTIVITY IN SIMILARITY ASSESSMENT Factors considered for assessment criteria Reference product data dependent Reference product attribute knowledge Stability-indicating product attributes Manufacturing process controls in place Approaches to establish assessment criteria Statistically derived Scientifically justified (non-statistical) Qualitatively compared (Visual) Objectivity Subjectivity 6 Amgen Proprietary For Internal Use Only

7 COMPARISON OF EMA AND FDA EXPECTATIONS FOR USE OF STATISTICS IN ANALYTICAL SIMILARITY ASSESSMENT EU US Ranges should be based primarily on the measured quality attribute ranges of the reference medicinal product and should not be wider than the range of variability of the representative reference medicinal product batches, unless otherwise justified. A descriptive statistical approach to establish ranges for quality attributes could be used, if appropriately justified. Recommends a three-tier system with statistical assessments, where appropriate 1. Equivalence testing for some high risk attributes Based on 90% confidence interval for the difference in means of the biosimilar and RP being within 1.5 standard deviation based on the reference product 2. Quality ranges (mean ± X SD) for other high to low risk attributes, with >90% of the biosimilar lots falling with the quality range EMA/CHMP/BWP/247713/2012 Committee for Medicinal Products for Human Use (CHMP) Guideline on similar biological medicinal products containing biotechnology-derived proteins as active substance: quality issues (revision 1) 3. Raw/graphical comparisons for other attributes FDA presentation for ODAC meeting (Jan 2015) 7 Amgen Proprietary For Internal Use Only

8 STATISTICAL EVALUATION FOLLOWING FDA TIER APPROACH Equivalence test (Tier 1) The intent of equivalence test is to demonstrate that: Mean TP Mean RP < EAC TP = Test product RP = Reference product EAC = Equivalence acceptance criterion If it can be demonstrated that difference in means between the TP and RP is less than the EAC, it is concluded that products are statistically equivalent. Quality range (Tier 2) The quality range of the RP for a specific quality assay/attribute is defined as: Mean RP ± K x S RP K = standard deviation multiplier S RP = standard deviation of all the RP lots. Expect 90% lots are within quality range 8 Amgen Proprietary For Internal Use Only

9 EQUIVALENCE TEST CONFIRMS VISUAL (MIN-MAX) TEST EU Min -Max US Min -Max Power of the equivalence test depends on number of lots Number of lots Assigned Power EAC - EAC ABP vs EU EU vs US ABP vs US 10 85% 15 90% 25 95% Yi Tsong, Journal of Biopharmaceutical Statistics (2017) EAC = 1.5 times the standard deviation of the reference product lots tested Circle represent the difference between the means. Vertical bar represents the 90% 2-sided confidence interval 9 Amgen Proprietary For Internal Use Only

10 QUALITY RANGE CONFIRMS VISUAL (MIN-MAX) TEST EU Min -Max US Min -Max Quality Range = mean ± 3 times standard deviation of the reference product lots ABP vs EU EU vs US ABP vs US QR may underestimate RP variability DP lots from same DS lot Impossible to sample all lots to cover RP clinical experience Expect 90% lots fall within the quality range 10 Amgen Proprietary For Internal Use Only

11 STATISTICS CAN OFFER AN OBJECTIVE VIEW OF DATASET Distribution of the data points may provide insights to process and method variabilities Statistically meaningful shift of reference product data over time may indicate a process change Changes in reference product profile can impact biosimilar development and analytical similarity assessment Seokkyun Kim, et al (2017) mabs 11 Amgen Proprietary For Internal Use Only

12 SELECTION OF STATISTICAL METHOD MAY BASED ON THE REFERENCE DATA DISTRIBUTION A symmetrical dataset should have close mean and median values Comparison of mean, median and mode of two distributions with different skewness. (Wikipedia) Median may be better suited for comparison for a skewed distribution. 12 Amgen Proprietary For Internal Use Only

13 OUTLIERS CAN INFLATE PRODUCT ATTRIBUTE RANGE 100% data point (Min-Max range) maximum third quartile first quartile minimum median Boxplot with whiskers from minimum to maximum 90% data point (60% of the Min-Max range) Same Boxplot with whiskers with maximum 1.5 IQR Incorporating proper statistical tools in the data analysis can increase confidence in the overall assessment Sufficiently large dataset and extensive investigation may be required to identify outliers 13 Amgen Proprietary For Internal Use Only

14 NON-STATISTICAL ASSESSMENT CRITERIA SHOULD BE SCIENTIFICALLY JUSTIFIED FOR BIOSIMILAR COMPARISON Results not amenable to statistical analysis Data close to or below limit of quantification (LOQ) Side-by-side (visual) comparison of chromatograms and spectrum Include objective criteria, e.g. similar profile with no new peaks above detection limits Method and instrument capability Consider both precision and accuracy targets based on reference product knowledge Intermediate precision expected based on method qualification Process and product uniquely controlled Process-related impurities Formulation-dependents attributes Device-specific properties 14 Amgen Proprietary For Internal Use Only

15 EXAMPLES OF NON-STATISTICAL CRITERIA Counts x x x A C BB Deconvoluted D l t d M Mass (D ) (Da) Peak Label: A B C D E Glycosylation: A2G0F:A2G0F A2G0F:A2G0F+K* A2G0F:A2G1F A2G1F:A2G1F A2G0F:A2G2F A2G1F:A2G2F Theoretical Mass (Da): Assessment Criterion (ppm): ± 50 ± 50 ± 50 ± 50 ± 50 EU ND D E EU ABP US Liu, BioDrugs (2016) Qualitative criteria for profile comparison: No new peak or absence of major peaks compared to reference product Scientific criteria for intact molecular weight: Conform to theoretical value within instrument precision Lot ABP US ND 15 Amgen Proprietary For Internal Use Only

16 CHALLENGES OF COMPARING STABILITY-INDICATING ATTRIBUTES FOR BIOSIMILARS Biosimilar T=0 data RP range from initial testing time point It is impossible to obtain reference products at T=0 Observed differences maybe due to shelf life differences between tested biosimilar lots and reference product lots 16 Amgen Proprietary For Internal Use Only

17 APPROACH FOR COMPARING STABILITY-INDICATING PRODUCT ATTRIBUTES Identify stability-indicating product attributes and develop appropriate analytical methods for quantitative measurement Obtain reference product regularly and conduct testing at multiple time points For biosimilar product, include results from stability studies performed for drug product lots manufactured from representative process and scale Stability trend of each product can be generated from batches analyzed over time 17 Amgen Proprietary For Internal Use Only

18 STATISTICS HELP NORMALIZE DATA TO THE SAME TIMEPOINT T=0 T= EoS Comparing biosimilar T=0 to RP first timepoints Comparing biosimilar EoS to RP EoS Overlapping age Comparisons performed at end of shelf life represent worse case of both biosimilar and reference product 18 Amgen Proprietary For Internal Use Only

19 SUMMARY Incorporating statistical tools in analytical similarity assessment can increase objectivity and confidence for the overall conclusion Selection of the appropriate statistical method for assessment should consider both product attribute and the data size and distribution Non-statistical criteria can be scientifically justified and should consider product knowledge and method capability Stability-indicating product attributes need to normalize to the same material age for meaningful comparisons science statistics 19 Amgen Proprietary For Internal Use Only

20 ACKNOWLEDGEMENT Jose Ramirez Brad Jordan Simon Hotchin 20 Amgen Proprietary For Internal Use Only