Guillaume Desachy, Alessandra Serra, Sandrine Guilleminot, Nicole Krämer, Karl Köchert, Nils Ternès.

PSI Webinar of the European Biomarkers European Special Interest Group: A promising adaptive biomarker-based design strategy for early phase and machine learning as an enabler of precision medicine!

Presentation 1: A promising adaptive biomarker-based design strategy for early phase clinical trials
Identifying predictive biomarkers is crucial in patient-centric clinical development. Enrichment strategies in late stages of drug development have been widely studied in the literature. However, implementing these strategies in early stages presents significant challenges due to the small sample size and numerous uncertainties that arise at this point in the development process. These uncertainties encompass biomarker (BMK) predictive value, cutoff value of the biomarker used to identify patients in the BMK-positive subgroup, the proportion of patients in the BMK-positive subgroup and the magnitude of the treatment effect in patients BMK-positive and BMK-negative. Early phase adaptive designs can improve trial efficiency by allowing for adaptions during the course of the trial. In this work, we are interested in adaptations based on interim analysis permitting a refinement of the existing study population according to their predictive biomarkers. Simulations show that the proposed design leads to better decision-making compared to a classical design that does not consider an enrichment expansion. Specifically, in the considered settings, gains up to 30% in the overall probability to hit the study success criteria at the end of the trial were achieved in comparison to a conventional design. 

Presentation 2: Machine Learning as an enabler of precision medicine
Since the re-creation of the Biomarkers ESIG in 2022, people from different horizons and job titles have expressed some interest in machine learning (ML) and artificial intelligence (AI) methods applied to drug development, from research to clinical. With this fact and the increasing interest in ML/AI from the scientific community and beyond, a group of a dozen people was set-up in the past months with the ambition to create a cross-company best practices guidance on the use of AI/ML in drug development. The group started with some proactive joint discussions to better understand each other’s interest in this area and has already contributed to the review and feedback to FDA’s recent discussion paper on usage of AI/ML in drug development. Now, the group is in a highly dynamic ideation phase to identify and choose topics which have the potential to impact drug development, e.g., virtual twin (VT) technology, GxP compliant AI/ML, use of image-based or digital biomarkers.