015: Characterizing the Covariates & their Impacts on Statistical Evaluations of Clinical Pharmacology Studies in the 351(k) Biologic License Applications

Statement of Purpose, Innovation or Hypothesis: Pharmacokinetic (PK) or pharmacodynamic (PD) similarity studies are a critical supporting element in 351(k) biological license applications (BLAs) seeking biosimilar approval. Incorporating the covariates that are known to be associated with primary endpoints in the statistical analysis of PK or PD similarity studies may result in improving the precision of similarity assessment. The aims of this study are to characterize the profiles of the covariates, and to evaluate their impact in terms of improving the statistical efficiency in PK and PD similarity studies.

Description of Methods & Materials: 351(k) BLAs submitted to FDA between March 6, 2015 and December 31, 2023 were surveyed. Information such as study types, study design, covariates used in the statistical analysis, and statistical methods for demonstrating similarity from PK and PD similarity studies were collected and analyzed. The impact of covariate adjustment was evaluated by conducting the statistical testing in demonstrating similarity with or without covariate adjustment.

Data & Results: Approximately forty percent of 91 PK and PD similarity studies utilized covariate adjustment in the statistical analysis for the primary endpoints. The proportion of covariate adjustment was 51.9% for the subset of studies with the parallel-group design. The studies that implemented covariate adjustment were found in more recent submissions. Body weight and clinical site were the most frequent covariates used in the parallel-group studies and the crossover studies, respectively. The statistical analysis generally included two or fewer covariates, and the maximum number was five covariates. Most covariate adjustments focused on the primary covariates with several studies investigating the secondary interaction between the covariate and treatments. Only a few studies ( < 3%) contained the methods in selecting the covariates based on the statistical significance to the model. Across 16 reference products surveyed, PK similarity studies for adalimumab biosimilars had the highest proportion (12 of 16 studies) of utilizing covariate adjustment. The 90% or 95% confidence intervals of PK or PD endpoints for similarity comparison using covariate adjustment is more precise than those without after considering the penalty of adding the number of estimated parameters, suggesting that covariate adjustment may reduce the residual variability and thus improve the statistical power of hypothesis testing.

Interpretation, Conclusion or Significance: This project established a comprehensive resource of covariates utilized in the clinical pharmacology studies in the biosimilar BLAs received by the FDA. Covariate adjustment can improve the precision of similarity testing and therefore enhance the statistical power of studies when used appropriately. Considering the effects of covariate adjustment on statistical analysis in the design of clinical pharmacology studies have the potential to increase the efficiency of biosimilar development.

Disclosures: This reflects the views of the authors and should not be construed to represent FDA’s views or policies.