Stakeholders want more clarity on FDA's statistical approach to assessing bioequivalence

Commenters on the US Food and Drug Administration’s (FDA) draft guidance outlining principles for in vivo or in vitro bioequivalence studies (BE) for investigational new drugs (INDs), new drug applications (NDAs), abbreviated new drug applications (ANDAs) indicated that they want the agency to elaborate on new topic areas introduced in the guidance.
 
The draft guidance, which replaces a 2001 guidance for industry published under the same name, added new information on how to assess bioequivalence in narrow therapeutic index (NTI) drugs and highly variable drugs, anticipate missing data and intercurrent events (IEs), and using adaptive trial design to “help applicants plan and analyze their BE studies with the goal of minimizing the number of assessment cycles necessary for approval.” (RELATED: FDA updates statistical approaches for assessing bioequivalence, Regulatory Focus 2 December 2022)
 
Stakeholders generally welcomed the update to FDA's guidance on statistical approaches to BE studies, but many had specific concerns they asked the agency to address in the final guidance.
 
Association for Accessible Medicines
 
The Association for Accessible Medicines (AAM) sought more information from FDA in their comment on topics such as whether a three-period design mentioned in a section on other fully replicated crossover designs could also be used in reference-scaled average bioequivalence, and if a section on the analysis population in an estimand framework meant this group should be considered the intent-to-treat population.
 
In addition, they requested more information on IEs in the guidance, including managing statistical situations for missing data and IEs when product specific guidance is not available. “FDA should provide more clarity with regard to the handling of intercurrent events, such as those in which the validity of a particular value may be in question, but the data are not missing,” they wrote.
 
The organization also asked for examples of prospective plans for experimental outliers as mentioned in the guidance. “Experimental outliers are a common problem in BE studies,” they said. “AAM agrees that applicants should be able to include a prospective plan in the BE study protocol for handling outliers to avoid having to repeat a pivotal study. Please provide more detail on what types of prospective plans for handling experimental outliers would be acceptable to the Agency.”
 
Bayer
 
Bayer’s comment on FDA’s draft guidance included requests for more information on whether dropouts are allowed in studies, and a more thorough definition of what constitutes a “highly variable drug product.”
 
In a section on determining sample sizes for adequately powered BE studies, Bayer noted it is “unclear what BE measures are considered necessary to assess BE.”

“[W]e propose to include that the relevant BE measure be pre-specified and justified in the protocol,” they wrote.
 
Bristol-Myers Squibb
 
In their comment, Bristol-Myers Squibb (BMS) echoed comments by Bayer and AAM on regarding clarifying the topic of dropouts, real-time documentation for removing outlier data, a more detailed definition of a variable drug product, and who should be included in an analysis of “eligible” dosed subjects in an estimand framework.
 
BMS noted that while FDA mentioned a standard 2x2 cross-over study design should use ‘a particular calculation,’ the calculation itself appears to be missing from the draft guidance.
 
“BMS suggests sample size calculation be added in for standard 2x2 cross-over design,” they wrote.
 
Biotechnology Innovation Organization
 
Biotechnology Innovation Organization (BIO) wrote in their comment that they appreciated FDA updating the guidance, as “there have been advancements made to further refine the bioequivalence study designs and statistical methods discussed in the Agency’s 2001 guidance.”
 
“This updated version of the guidance provides useful examples that will assist sponsors/applicants. In particular, the section on adaptive study designs is beneficial,” they said.
 
Along with expressing many of the same concerns as other stakeholders, BIO noted that “for some newer treatment modalities with limited or no systemic pharmacokinetic exposure, Comparative Clinical Endpoint Bioequivalence Studies may not be the most appropriate choice.”
 
International Pharmaceutical Aerosol Consortium on Regulation & Science
 
The International Pharmaceutical Aerosol Consortium on Regulation & Science (IPAC-RS) provided a detailed response to FDA’s characterization of the similarity index (f2) as a statistical method for assessing BE in their comment.
 
Although the f2 similarity approach was first recommended in Scale-up and Post-Approval Changes Guidance for Immediate Release Products (SUPAC-IR) guidelines published in 1995, IPAC-RS wrote, there was not “sufficient detail for implementation” and a later complementary guidance on dissolution testing of immediate release solid oral dosage forms limited its use.
 
“Even now after 25 years, questions and uncertainty still abound on how and when to use that specific in vitro BE criterion,” IPAC-RS wrote. “Although this guidance is intended to encourage the use of science-based approaches to making statistical in vitro BE assessments, the in vitro BE area still suffers from the lack of a holistic approach as has been taken in the in vivo area.”
 
Novartis
 
Pharmaceutical company Novartis requested in their comment that FDA specify how the guidance would impact biosimilar development. For example, in a section on sample size considerations, the company asked the agency to include a statement to one made in the 2001 guidance that recommended simulations “be conducted using a default situation allowing the two formulations to vary as much as 5%” in average bioavailability as well as “with equal variances and certain magnitude of subject-by-formulation interaction.”
 
“We would recommend including a similar statement in this guidance, to encourage applicants to plan the sample size with a 5% difference between the formulations,” the company wrote.
 
Novartis also recommended FDA avoid making specific recommendations on minimum power in equivalence studies.
 
“The impact of an underpowered study for a comparison of two formulations where in fact both formulations are the same is that the sponsor will have an unacceptable chance of not being able to show bioequivalence,” they said. “Since the amount of power is a sponsor risk decision, suggestion of what the minimum power should be (80% in this document) seems unwarranted.”
 
Takeda
 
Takeda’s comment called for more information on relevant estimands in BE, “further guidance towards reasonable strategies to address IEs,” guidance on how to manage non-adherence to the planned treatment schedule, and estimating sample size for studies.
 
“Two of the most important IEs in studies planned to demonstrate BE are (a) discontinuation—or more generally non-adherence—of scheduled treatment and (b) co-therapy that affects the outcome of interest. Strategies addressing these events are mentioned indirectly in the guidance, but additional explanations or examples would reduce ambiguities,” they explained.
 
Concerning the section on missing values, Takeda noted that “what actually constitutes missing values depends on the estimand aligned with the primary objective (according to the E9 addendum, missing data are data that would be meaningful for the analysis of a given estimand but were not collected).”
 
“Therefore, the relevant primary estimand to demonstrate BE needs to be defined first, and the method of estimation including missing data handling is selected depending on the target of estimation,” they explained. “This sequence of steps should be reflected in the guidance, particularly because the identification of all relevant IEs and the strategy to address those is one of the major critical first steps in study team discussions involving clinicians, pharmacologists and statisticians.”
 
Draft guidance