By Rick Burdick, Henriette Kuehne, Aili Cheng, Kristof Vandekerckhove, Todd Coffey, Tara Scherder, and Katherine Giacoletti
Recently, members of the AAPS Biosimilars focus group (BFG) compiled and submitted comments on two draft regulatory documents concerned with the demonstration of analytical similarity. The first document by the European Medicines Agency (EMA), Reflection paper on statistical methodology for the comparative assessment of quality attributes in drug development, is not intended as a guidance document, but provides “… current regulatory considerations regarding statistical aspects for the comparative assessment of quality attributes where these are used, or are proposed for use, in drug development and Marketing Authorisation [sic] Applications.” The reflection paper discusses statistical methodology for comparing product data for not only analytical similarity, but also for process changes and other scenarios. The second document is a draft Food and Drug Administration (FDA) guidance on Statistical Approaches to Evaluate Analytical Similarity that is specific to biosimilars, unlike the EMA document.
Although the scope is not precisely the same in the two documents, both include discussion of statistical methodologies applicable to the evaluation of analytical similarity. In addition to differences in scope, the two documents differ in the treatment of specific statistical methodologies applicable to comparing pharmaceutical process related to assess comparability or similarity. The EMA reflection paper discusses a number of statistical approaches at a high level, specifically endorsing none, while the FDA draft guidance proposes a very specific statistical approach to demonstrating analytical similarity. Despite these differences, many comments from the BFG were common to both documents and included a call for harmonization, particularly on the issue of analytical similarity. Some of BFG’s concerns common to both documents are described below.
ISSUES ADDRESSED
The key BFG comments on both documents on analytical similarity were related to the important differences between this type of comparability assessment and others such as pre- and postchange comparisons, in terms of the available data, and in particular, the limitations in terms of what the biosimilar manufacturer knows about the innovator data to be used for analysis. These aspects of the evaluation of analytical similarity are critical, because they translate into potentially (large) increases in patient or producer risk (risks of making an erroneous statistical conclusion in favor of or against analytical similarity), or both, compared to the use of the same statistical approaches for comparisons without these issues. Specifically, the innovator data available to the biosimilar company may include (1) process changes or unintentional process drifts, making pooling the data inappropriate; (2) deviations which were within permitted specifications but not within typical manufacturing variability; and (3) multiple drug product lots from the same drug substance lots, resulting in a correlation structure which cannot be accounted for in the analysis since it is unknown to the biosimilar producer. The BFG also noted concerns with the particular statistical methodology FDA proposed in its draft guidance, emphasizing aspects that may lead to increased patient and/or producer risk.
In addition to raising concerns, the BFG proposed revised language, additions to, and deletions from the two documents. At a high level, these suggestions included a call for some flexibility with regard to statistical approaches to analytical similarity in the FDA draft guidance, and along with the request for flexibility, suggested the inclusion of specific criteria by which alternative proposed methodology would be evaluated by the agency:
- Appropriate statistical operating characteristics (protect patient risk and decrease patient and producer risk with increasing sample size).
- Robust to violations of assumptions (e.g., normality, independence).
- Assess similarity over entire distribution of the attribute.
- Allow decision making with small sample sizes.
Note: Statistical approaches meeting the criteria above need not rely on equal sample sizes or equal variability between the reference and biosimilar products, but the impact of either or both being different should be discussed and understood.
It was noted that the current draft EMA reflection paper emphasizes limitations and difficulties associated with the use of statistics for comparability assessments, and thereby may seem to discourage using statistical methods for this purpose. Thus, the same criteria as listed above were included in a comment to the EMA reflection paper, not in a call for flexibility (since the reflection paper did not dictate or propose any specific methodology), but as a suggestion that EMA, rather than emphasizing drawbacks or complications of using statistical methods, provide practical guidance on the selection of or development of statistical methods for comparability evaluations. The submitted responses to both documents also included specific commentary on details of the statistical methodologies discussed or recommended, as well as on more practical matters related to comparability evaluations, including the need for international harmonization among regulatory bodies.
The authors are members of the AAPS Biosimilars focus group.