Biorationality: Advice for the FDA, Biosimilar Developers on Reducing the Cost of Clinical Testing

Eighteen years after the first biosimilar approval in the United States, the FDA has diligently applied scientific approaches to revise and add more approval guidelines continuously. Significant changes include:

• the removal of “animal toxicology” from the text of the law and replacing it with “nonclinical” testing,¹
• waiver for immunogenicity testing where immunogenicity does not impact pharmacokinetics for biosimilars and interchangeable products,
• waiver for clinical efficacy testing in patients where pharmacodynamic biomarkers are available, and
• continuously updating and adding advice on the conduct of clinical trials that now add to 138 guidelines, addressing a myriad of issues like ethics, modeling, data collection and analysis.

However, the FDA has missed the realization that efficacy testing of biosimilars is not for characterizing the profile in clinical pharmacology testing and response in efficacy testing but to compare these attributes with the reference product. A new drug entry must be tested in the target population to ensure proper dosing and safety evaluation.

However, such profiling is not necessary for biosimilar testing. Understanding this difference should allow the selection of the study population to reduce intersubject and intrasubject variability for all clinical trials. The current approach to the design of clinical trials is based on using an a priori coefficient of variation (CV) in the outcome or response to calculate the study size. For example, if the CV is 50% and the acceptable effect size is 90% or 10% less than the reference product, then the study size will need 784 subjects or patients with an alpha of 0.05 and a beta of 0.20. If the CV can be reduced to 10%, this number goes down to 32.

Additionally, the reported CVs in the literature are generally misleading as these are rarely replicated across the studies. A better approach is to conduct a pilot study with a narrow selection criterion, age, gender, ethnicity, body mass index, etc, to calculate the CV. Still, at the same time, if the study meets the power requirement due to a lower CV, then this trial alone should be sufficient. I foresee clinical pharmacology testing using no more than 20 subjects and perhaps no more than 40-50 in efficacy testing.

If additional testing is required, then the data from both studies should be combined to demonstrate sufficient study power.² These experimental designs meet all statistical and scientific requirements without compromising safety or efficacy evaluation.

I have made this proposition to the FDA in person and in consultative meetings; the FDA reports the minutes of these meetings. I have also added these comments to the FDA guidelines on protocol design for clinical trials.

I strongly urge stakeholders to take an active role in promoting this concept. It is this kind of support that is expected from professional associations. But more importantly, the developers should use these arguments in their investigational new drug filings. The goal should always be to do as little as possible without compromising safety and efficacy assurance. Unfortunately, the clinical studies submitted for US licensing have been extensive, often due to the use of a priori CV estimate, but also to demonstrate efficacy to clinicians who are used to seeing such trials. Biosimilar developers should have little to do with educating prescribers or patients; this is the role of the FDA, and they are doing a great job; the recently added support includes:

• For prescribers
  • Overview of Biosimilar Products
  • Biosimilar Regulatory Review and Approval
  • Interchangeable Biological Products
• For Patients
  • Are you on biologic medication? What you need to know about biosimilar treatment options.

References

1. Han JJ. FDA Modernization Act 2.0 allows for alternatives to animal testing. Artif Organs. 2023;47(3):449-450. doi:10.1111/aor.14503
2. Piegorsch WW, Bailer AJ. Combining information. Wiley Interdiscip Rev Comput Stat. 2009;1(3):354-360. doi:10.1002/wics.45