Opinion: FDA Changes to Biosimilar Development Guideline Save Money and Time

Analytical similarity assessment of biosimilar candidates is the backbone for establishing biosimilarity, a primary requirement for the licensing of biosimilars. After withdrawing its pivotal final guideline on analytical similarity testing, the FDA issued new guidance (Development of Therapeutic Protein Biosimilars: Comparative Analytical Assessment and Other Quality-Related Considerations), with many apparent and subtle differences, primarily based on my citizen petition to the FDA. For example, few people noticed that the FDA replaced “testing” with “assessment.” To assess is to test the value of something using various tools. Testing is when something is put through a series of trials to determine its value. Thus, testing is one of the ways to assess something. In this paper, I will draw attention to both substantive and subtle changes to help reduce the cost and time for the approval of biosimilars.

• Orthogonal Testing. Orthogonal testing uses different testing methods to affirm the same attribute, such as the protein content using an ultraviolet (UV) reading and chromatography reading. The need for such testing comes when one method cannot be suitably validated. However, many reported tests have used multiple validated methods to confirm the same finding. As an example, 25 analytical tests were submitted for Pfizer’s adalimumab biosimilar, while the Boehringer-Ingelheim adalimumab biosimilar came with 71 test results. Which company was more brilliant: the one that chose to do just 25 tests or the one that overkilled the testing? The FDA guidance stresses the use of orthogonal testing to "definitively" identify differences in product attributes.
• Expression System Attributes. Expression systems are the genetically modified bacteria or Chinese hamster ovary cells that produce a therapeutic protein as a byproduct. You cannot change this with any process changes; as a result, you need not use a large number of lots to confirm. Primary, secondary, and tertiary structures fall in this category; you can use just 3 lots, but these must be assessed side by side with the reference product because you cannot rely on any legacy values (such as those given in pharmacopeia or scientific literature); most methods of assessment for structure are also not well validated, so side-by-side testing removes the constraints of the validity of the test. This stage of development can use orthogonal testing; for example, the secondary and tertiary structures are challenging to study, and innovations such as changing the temperature, osmolality, pH, or ionic strength of the solutions tested can provide greater confidence of similarity. "Use of different expression systems will be evaluated on a case-by-case basis," the guidance states.
• Process-Related Variability. Although the expression system is unchanging, what happens to molecules after they are expressed mainly depends on process conditions such as bioreactor operation parameters and chromatography technologies. These are often termed posttranslational modifications, such as glycosylation or impurities. Instead of justifying the differences, it is better to adjust the process to match the profile with the reference product. Know that more important is the identity of variables and not their ratio vs that of reference product variables.
• Critical Quality Attributes (CQAs). Comparison of a biosimilar candidate and the reference product begins with choosing the properties of the reference product to match or deciding what is critical to match. When biosimilars were introduced, the selection of these attributes was a complicated exercise of weighing risk factors and figuring out how to avoid variability during manufacturing; now we know well what the CQAs are and what should not be a CQA. For example, based on the earlier FDA guidance, Sandoz, which developed the first US biosimilar (Zarxio, filgrastim), tested the protein content as a CQA and kept failing until 71 lots were tested, even though none of the lots were out of release specification, meaning they met standards for acceptance. Such exercises are no longer necessary; if you are able to establish release specification based on either acceptance variation or based on variation in the reference product, there is no need to test these attributes in an elaborate side-by-side comparison exercise.
• Statistical Modeling. The first final guidance on analytical testing was rife with mistakes relating to the statistical modeling of CQAs. Without any rationale, the FDA had created a 3-tier system. The most rigid was Tier 1, where the 90% confidence interval of the biosimilar product had to be within 1.5 of the standard deviation (SD) of the reference product. It turned out that most CQAs assessed failed. The new guidance removes statistical modeling but advises that enough lots should be used in a side-by-side comparison if SD is used to measure variability. Likely, the FDA will allow pooling analytical assessment data as they did to evaluate the Sandoz filing.
• Limited Testing. Both FDA and the European Medicines Agency have experience evaluating submissions of dozens of biosimilar products, and the initial emphasis on “more is better” is fading away. Still, the need for a “stepwise approach” to biosimilar development remains. Generally, if fewer lots are used for comparison, there is a higher chance of a biosimilar failing (remember Sandoz did 71 lots to pass the protein content test). Matching with fewer lots should be preferred unless the CQA is highly variable. Most traditional CQAs are not highly variable. Testing more than necessary not only adds to the cost of development but also extends the timeline for submission. The goal is not to cut the corners but to go around the corners where appropriate.

We have come a long way from the early days of superfluous testing of biosimilars; the new FDA guidance should be read carefully between lines; it is a more rational approach to achieving the demonstration of biosimilarity, but the FDA is like any other agency—if you keep offering to assess ad infinitum, they will have no objection to it. Manufacturers must learn the difference between “testing” and “assessment.”