Mohamad-Samer Mouksassi, Nastya Kassir, Nathalie H. Gosselin , JF Marier
Pharsight - A CertaraTM Company, Montreal, Quebec, Canada
Objectives: FDA and EMA have proposed using pharmacodynamic (PD) data to assess bioequivalence (BE) of locally-acting product. Dose-scale modelling (DSM) of PD data using a 2-step approach has been proposed, where the reference product is modelled in a first step, and scaling of the test to reference product is performed in a second step. Power calculations of BE studies using DSM are challenging due to the complexity of using of 2-step approach for sample size calculation, and the lack of understanding of sources of variability responsible for the width of 90% confidence intervals (CI). A 1-step DSM method was developed to assess BE of locally acting product and simulations were performed to identify sources of variability and their impact the power.
Methods: PD data were used to perform DSM of test and reference products. A 1-step approach was developed by simultaneously modelling the test and reference product using non-linear least-squares modelling techniques and by constructing bootstrap Bca nonparametric 90% CI. Simulations were performed to determine the sample size required to meet BE criterion and identify the most important factors (e.g.,between-subject variability, point estimate of the relative potency,… ) responsible for the width of 90% CI.
Results: The proposed 1-step DSM method resulted in similar 90% CI as compared to the 2-step DSM method. The 1-step DSM method required around 2000 bootstrap replicates to obtain stable 90% CIs. Within-subject and residual variability in PD data and point estimate of relative potentcy were identified as the most important components responsible for the power.
Conclusion: The 1‑step DSM method enabled the use of simulations to power definitive studies and increase the likelihood of demonstrating BE. Simulations are the tool of choice to power studies of locally acting gastrointestinal drugs (e.g., orlistat, misoprostol, mesalamine) and pulmonary drugs (e.g., albuterol, glycopyrrolate) assuming a priori knowledge of the dose-response curve of the reference product is available as well as an estimate of between and within-subject variability.
Reference: PAGE 21 () Abstr 2480 [www.page-meeting.org/?abstract=2480]
Poster: Other Modelling Applications