Martine Neyens1, Juan Jose Perez Ruixo2, Ruben Faelens1
1Johnson & Johnson, innovative medicine, 2Johnson & Johnson, innovative medicine
Introduction When transitioning to a new candidate dosing posology with variations in dose, dosing interval, administration route, or formulation, pharmacokinetic (PK) bio-equivalence may not be demonstrable. Repeating pivotal clinical trials to demonstrate favorable benefit/risk profile requires prohibitively large sample sizes. However, the availability of a surrogate biomarker offers an alternative by facilitating the establishment of non-inferior effects on the surrogate marker. This approach uses a non-inferiority margin on surrogate biomarker that correlates with a minimal (non-regrettable) loss of clinical endpoint efficacy. To enable rapid iteration on assumptions, models, and non-inferiority (NI) trial designs, a web application was developed. Objectives The objective was to build a simulation app for illustrating the predicted relationship between the surrogate biomarker and clinical endpoint, finding the surrogate biomarker concentration in a dosing regimen with NI clinical effect, and optimize the study design and sample size required for NI trials. Methods A shiny[1] application was developed to simulate biomarker-based non-inferiority bridging. The app has two main views. The first is the Population Simulation view, which allows simulation of a population model predicting pharmacokinetics (PK), biomarkers, and clinical endpoints for two dosing regimens: the Test versus the Reference. The Test regimen is calibrated similar to the reference dosing, but to result in a clinically loss of effect, such as achieving 3/4ths or 2/3rds of the original drug effect through dose reduction [3]. The dose can be changed by predefined dosing regimens or by customized dosing module. This contains options for choosing the dosing levels and the dosing interval, choosing between administration route, weight-based or flat dosing, or whether to use a loading dose. Graphs are shown with median and 95% prediction interval for PK, biomarker and clinical endpoint for the Test and Reference dosing regimens. Plotly [2] is used to provide tooltips showing the numeric concentration, allowing the user to quickly identify the difference between Reference and Test on the simulated clinical endpoint. Key output from the first simulations is the surrogate biomarker mean and SD for Reference and Test regimens. This can further be shown in a table, simulated with higher virtual population size to ensure high precision. The second view is the Clinical Trial Simulation (CTS) view, which simulates and evaluates NI trials using inputs from a specific endpoint simulated from the popPKPD model. This allows the user to evaluate the NI trial design for the actual clinical endpoint, versus the surrogate biomarker. For the selected endpoint, a table of the distribution (mean, sd) for the Reference and Test regimen is shown. Alternatively, a custom mean/sd can be configured instead. The CTS view allows selection of sample size. A graphical representation of the difference in means and confidence intervals across all simulated virtual trials is shown. The non-inferiority margin can be adjusted, impacting trial outcomes and the overall Probability of Success (PoSS). By moving the NI-margin, it can be tuned to achieve e.g. 90% PoSS. By iterating through these simulations, the app facilitates the determination of appropriate NI margins and sample sizes. The app supports customization through shiny modules and optimizes performance with caching and parallel calculations. Once configured to an appropriate scenario for drug development, the app and its settings can be bookmarked to allow easily accessing when presenting to a clinical team. Results The app was instrumental in examining the NI margin for various endpoints across changing conditions during drug development of a new candidate. The app enabled straightforward updates of the NI trial design as models were revised to incorporate new data coming available throughout development. Moreover, the application demonstrated significant flexibility in accommodating shifts in strategy for the NI Test regimen. It allowed for seamless modifications in dosing regimens, the evaluation of different endpoints (predefined and user defined). Additionally, the app provided an efficient platform for testing statistical assumptions, enabling comparisons between no effect, enhanced effect, and the hypothesis of a poorly performing new regimen. This adaptability underscores the app’s potential in optimizing trial designs. Conclusion A shiny app to explore a non-inferiority margin for a biomarker can support clinical discussions on study design, bypassing the usual turnaround delay when performing clinical trial decisions.
[1] https://cran.r-project.org/web/packages/shiny [2] https://cran.r-project.org/web/packages/plotly [3] https://www.fda.gov/regulatory-information/search-fda-guidance-documents/non-inferiority-clinical-trials
Reference: PAGE 33 (2025) Abstr 11591 [www.page-meeting.org/?abstract=11591]
Poster: Methodology - Other topics