Biomilars: BIO Comments on FDA Draft Guidance Statistical Approaches to Evaluate Analytical Similarity

Re: Docket No. FDA-2017-D-5525: Statistical Approaches to Evaluate Analytical Similarity
Dear Sir/Madam:
The Biotechnology Innovation Organization (BIO) thanks the Food and Drug Administration (FDA) for the opportunity to submit comments to the Draft Guidance on Statistical Approaches to Evaluate Analytical Similarity.
BIO is the world's largest trade association representing biotechnology companies, academic institutions, state biotechnology centers and related organizations across the United States and in more than 30 other nations. BIO members are involved in the research and development of innovative healthcare, agricultural, industrial, and environmental biotechnology products.
BIO commends FDA on this Draft Guidance which provides important advice on the evaluation of analytical similarity for Sponsors of biological products licensed under section 351(k) of the Public Health Service Act (42 U.S.C. 262(k)). In finalizing the Draft Guidance, and in assisting sponsors on the evaluation of analytical similarity, we strongly urge the agency to consider our comments and to implement the suggested changes.
1. General Comments
  • The approaches outlined in this Draft Guidance are complex and warrant product-specific discussion with the Agency very early in development to discuss lot selection, risk ranking of attributes, and initiate analytical similarity plan and statistical analysis plan discussions. We support the Agency’s view in lines 170-172, that Sponsors should develop an analytical similarity assessment plan and discuss this approach with the Agency as early in the development program as feasible.
  • Due to the stepwise process of biosimilar development, early discussions between FDA and sponsors will help to avoid misinterpretations of expectations (lines 170-172). However, follow up discussions may be needed as more information is gained throughout development.
2. Challenges and Limitations to Applying Statistical Analyses in the Evaluation of Analytical Similarity Data
  • BIO is supportive of the Draft Guidance’s acknowledgement that there are many challenges and limitations to applying statistical analyses in the evaluation of analytical similarity data (lines 78-86; 94-95; 263-265; 378-382; 403-404). However, there are other challenges that are not directly acknowledged or addressed in the Draft Guidance. For example, another risk to the biosimilar developers’ ability to design the statistical analyses is the absence of control over the reference product, whose quality attribute levels may fluctuate or change at any time during or even after the proposed biosimilar product’s development.
  • The approach recommended by FDA for statistical analysis of analytical similarity can be a useful tool in supporting interpretation of data when applied appropriately. However, the interpretation should be considered complementary to the overall analytical similarity assessment and not used as a decision tool.
3. Analytical Similarity vs. Analytical Similarity Assessment Plan
  • Throughout the Draft Guidance the term analytical similarity assessment plan is used both to describe the application of analytical procedures for demonstration of biosimilarity and to encompass the statistical assessment similarity plan. This may contribute to a lack of clarity for the reader as they may infer this to relate to the totality of analytical data collected to assess whether the proposed product is highly similar to the reference product rather than the evaluation of these data. For clarity, these two concepts should be clearly distinguished. FDA should consider other terminology – perhaps analytical similarity data evaluation plan or statistical assessment plan to encompass the risk ranking, method determination, and statistical analysis plan. Use of the term analytical similarity assessment plan could mislead readers as to the scope of this Draft Guidance.
4. Quality Attributes
  • The Draft Guidance should make a clear distinction when referring to “Quality Attributes”, between structural/physicochemical attributes and functional attributes (e.g., lines 211-212: “Determination of the statistical methods to be used for evaluating each quality physical/chemical and functional attribute based on the risk ranking and on other factors”)
  • Keeping the concepts of structural/physicochemical attributes separate from that of functional attributes will help maintain a link between the Product Quality Attribute Assessment (PQAA) performed as a component of the Quality by Design (QbD) development process.
  • This distinction among attributes would also afford companies that do not include functional attributes in their risk assessment to not have to perform separate risk assessments for the similarity exercise. Rather they could utilize their PQAA and assess functional attributes as an additional component to complete the similarity analysis.
5. Challenges of Developing a Pre-Specified Statistical Analysis Plan
  • The Draft Guidance states (lines 315 – 318) that the statistical analysis plan should be pre-specified to the fullest extent possible and notes preliminary data to get an initial estimate of variability of the reference product’s attribute.
  • Reference product changes over the duration of shelf-life will not be known to the biosimilar manufacturer at early stages of biosimilar product development. An accurate estimation of the variability of the reference product’s attributes may not become apparent until many years after initial lots are sampled.
  • Thus, developing a pre-specified statistical analysis plan based on initial reference product lots may be irrelevant or even contradictory as the reference product is sampled and tested throughout the development program of the proposed biosimilar product; changes observed, which conflict with early estimates of variability, should be accounted for and the plan adjusted accordingly.
  • The statistical analysis plan should be pre-specified to the fullest extent possible, however, as the reference product variability is outside the control of the biosimilar developer, circumstances may arise which require the original estimate of reference product variability to be adjusted.
6. Tiered Approach
  • In general, we agree with the Draft Guidance regarding the utilization of different categories based on levels of clinical criticality.
  • In some cases, however, the use of Tier 1 equivalence testing may not be appropriate (or possible). BIO requests that the Agency clarify the guidance by noting that in such cases, the sponsor should approach the Agency to discuss alternative approaches. In addition, it is concerning that equivalence testing statistics could extend the statistical recommendation for Tier 1 quality attributes to future manufacturing process changes proposed for biological drugs (e.g., necessitate the need for head-to-head analytical and functional comparisons of biological drugs). We request clarification from the Agency that this guidance does not apply post-approval.
  • Some scientific concerns with the approach for Tier 1 may include:
  • The requested equivalence testing for Tier 1 attributes, if approached as a pass/fail criteria, may pose the risk of restricting biosimilar approvals for reasons associated with reference product variability.
  • BIO respectfully requests that FDA hold a stakeholder workshop to work with sponsors to identify scientifically appropriate ways with which to deal with changes in reference product attributes over time and how statistical methods may be used appropriately in this context.
  • The Draft Guidance lacks description of possible justifications which would allow biosimilar approval in cases of failed Tier 1 equivalence results. The Draft Guidance should expand possible justifications to overrule a failed equivalence test in cases where the mean of the reference product changes over time. This would ensure alignment with existing guidance (ICH Q6B, ICH Q5E, ICH Q7, ICH Q8, ICH Q11). A possible suggestion in this regard may include justification by scientific understanding of the impact of the range of Tier 1 quality attribute(s) to safety and efficacy.
BIO appreciates this opportunity to submit comments on the Draft Guidance on Statistical Approaches to Evaluate Analytical Similarity. We provide additional specific, detailed comments to improve the clarity of the Draft Guidance in the following chart. We would be pleased to provide further input or clarification of our comments, as needed.