Current practices for protein similarity assessment and commentary on those approaches.
For biosimilar designation by the Food and Drug Administration, the biosimilar must demonstrate that it is highly similar to the reference product in safety, purity, and potency, and the totality of evidence should be established through analytical evaluations, pharmacokinetics/ pharmacodynamics, and safety data (immunogenicity).
It is difficult to recreate biologics because their complex proteins are derived from living organisms, versus chemical compounds that are comparably easy to replicate. In the recently published AAPS Journal paper Statistical Approaches to Assess Biosimilarity from Analytical Data, Richard Burdick, Todd Coffey, Hiten Gutka, et al explain that “[p]rotein therapeutics have unique critical quality attributes (CQAs) that define their purity, potency, and safety. The analytical methods used to assess CQAs [during the biosimilar approval process] must be able to distinguish clinically meaningful differences in comparator products, and the most important CQAs should be evaluated with the most statistical rigor.”1
This article was written by a group of subject matter experts from U.S. companies to provide a guide to drive harmonization of best practices within the chemistry, manufacturing, controls, and analysis community and provide regulators with an overview of current industry thinking on applying modern analytical technologies for protein similarity assessments. The paper uses case studies to assess analytical similarity and statistical approaches for proving biosimilarity.
As a response to this article, Yi Tsong, Qi Xia, and Yu-Ting Weng recently published a paper entitled: Commentary on “Statistical Approaches to Assess Biosimilarity from Analytical Data” by Burdick et al 2 in The AAPS Journal. While these authors agree with many of the assessments by Burdick et al, the authors also challenge some of the group’s statistical approaches and propose new solutions.
For example, Burdick et al claim that there may be correlations between reference lots when they are sampled from the same drug substance (DS) lot. Tsong et al comment that this correlation can be combatted “if the [reference listed drug (RLD)] lots are sampled through two-stage hierarchical sampling plan by selecting DS lots first and randomly select a pre-specified number of RLD lots from each DS lot. The unbiased variance estimation can be done using variance component model. However, if we sample RLD lots randomly from the population of RLD lots and then try to identify the DS lot in order to estimate the corrected variance, the estimated variance may not be correct.”2
Both articles provide useful information for using statistical analysis to evaluate biosimilarity. Read more about this issue of fair sampling and about other methods in these AAPS Journal articles by logging into the AAPS website and clicking once on The AAPS Journal cover to access articles.
References
- Burdick R, Coffey T, Gutka H, et al. Statistical approaches to assess biosimilarity from analytical data. AAPS J. doi:10.1208/s12248-016-9968-0. Published October 5, 2016. Accessed October 25, 2016.
- Tsong Y, Xia Q, and Weng YT. Commentary on “Statistical approaches to assess biosimilarity from analytical data” by Burdick et al[1]. AAPS J. doi:10.1208/s12248-016-9987-x. Published October 5, 2016. Accessed October 25, 2016.