On September 21, 2017, the U.S. Food and Drug Administration released a nonbinding draft guidance for industry, Statistical Approaches to Evaluate Analytical Similarity, for comment purposes. This guidance is part of a series of guidance documents that FDA has released in support of its ongoing efforts to implement the Biologics Price Competition and Innovation Act (BPCIA). It serves as a companion document to the guidance for industry Quality Considerations in Demonstrating Biosimilarity of a Therapeutic Protein Product to a Reference Product. The objective of this guidance is to “assist sponsors in demonstrating, through an evaluation of the analytical similarity of the proposed biosimilar and reference product, that the proposed biosimilar and reference product are highly similar to support licensure under section 351(k) of the PHS Act.” The guidance describes “the type of information a sponsor of a proposed biosimilar product should obtain about the structural/physicochemical and functional attributes of the reference product, how that information is used in the development of an analytical similarity assessment plan for the proposed biosimilar, and the statistical approaches recommended for evaluating analytical similarity.”

The FDA recognized that challenges in designing the statistical analyses can result from limited numbers of reference product and/or proposed biosimilar lots and the large number of potential quality attributes that can be compared. As a result, it was recommended that a minimum of ten U.S.-licensed reference product lots with remaining expiry spanning the product shelf life be sampled. Combining data from comparator reference products approved outside of the United States “generally would not be expected to support a determination that the proposed biosimilar is highly similar to the U.S.-licensed reference product.” The agency indicated that ten biosimilar lots should be sampled, and it may be possible to combine data from biosimilar lots manufactured with different processes and/or at different scales.

The agency recommends a three-step, risk-based approach in analytically identifying and assessing important quality attributes of proposed biosimilar products. The first recommended step is a “determination of the quality attributes that characterize the reference product in terms of its structural/physiochemical and functional properties.” The identified quality attributes are then ranked “according to their risk of potential clinical impact,” by considering the impact of each attribute on clinical activity, PK/PD profiles, safety profiles, immunogenicity profiles, and the degree of uncertainty around the clinical impact of each attribute. The risk scores should be proportional to patient risk, and the ranking should be justified with appropriate citations to the literature and provided data. Finally, the “attributes/assays are evaluated according to one of three tiers of statistical approaches” which are assigned based on consideration of risk ranking and other relevant factors.

It was noted that the final assessment as to whether a proposed biosimilar is highly similar to a reference product “is made upon the totality of the evidence, rather than the passing or failing of the analytical similarity criteria of any one tier or attribute.”

The Agency recommends developing an analytical similarity assessment plan including a proposed statistical approach to evaluation early in the development program. Additionally, the Agency recommends early submission and discussion of the analytical similarity plan, so that agreement can be reached on which attributes/assays should be evaluated in each tier. The final analytical similarity report, including the analytical similarity assessment plan, should be submitted with the 351(k) abbreviated biologics license application.

Comments on this draft guidance are due by November 21, 2017—60 days from the date of the proposed guidance’s publication in the Federal Register. Comments can be submitted online.