GENZ-AF-IBD study design: a hybrid phase 4 clinical equivalence trial using AI-based synthetic control arms to compare generic apixaban with the originator in patients with atrial fibrillation
M Narducci, V Vassilikos, M E Androutsou, S Koilakou, F Scaldaferri, G Liuzzo, G Giannopoulos, K Votis, M Panagiotidou, A Ginis, K Perakis, F Folkvord, H Le, P Bamidis, C ChatzichristosAbstract
Introduction
Numerous Phase-IV clinical trials (CT) have evaluated the efficacy and safety of Novel Oral Anticoagulants (NOACs) for atrial fibrillation (AF). ARISTOTLE and ROCKET-AF have demonstrated the clinical benefits of NOACs (1, 2); however, they typically require large patient cohorts. Few trials have utilized innovative methodologies like synthetic control arms (SCA) or digital twin models (DT). The primary objectives of GENz-trials are (1) to assess the performance of the developed SCA for AF patients in a real-world (RW) CT, and (2) to study the performance of generic apixaban in real-life scenarios.
Aims
GENz-trials introduces a next-generation CT framework. This approach is organised around three tightly connected capabilities. First, it automates key workflows using generative AI to streamline activities such as personalised informed consent, protocol drafting, trial monitoring, and regulatory submissions. Second, GENz-trials strengthens patient recruitment through AI-driven matching, analysing large-scale electronic health records and other RW data to identify eligible participants more efficiently. Third, the framework integrates DT and SCA, using in-silico simulations and RW data to provide robust comparators with fewer participants.
Methods
The GENz-trials project follows a three-phase methodology: 1) Development, the consortium organises user-centered design workshops. Based on insights from these workshops, the project develops a set of AI modules that support protocol generation, literature synthesis, electronic informed consent, and regulatory compliance. This phase also delivers capabilities for federated patient recruitment and for constructing both cohort-based SCA and DT-based SCA. To enable these functions, the tools are trained using harmonized retrospective RW data from 2 clinical centers; 2) Emulation, GENz-trials digitally emulates selected CT to enable head-to-head comparisons between traditional and next-generation CT constructs. The analysis relies on causal inference and simulation techniques to quantify treatment effects and to understand how different design choices influence bias, variance, and transportability. Methodological rigor is reinforced through validation using established benchmarking environments; 3) Prospective Trial, the full GENz-trials system is deployed in a RW study assessing a generic anticoagulant, apixaban. In this phase, DT based SCA are used to generate regulatory-grade evidence on the effectiveness and safety of the treatment.
Conclusion
No Phase-IV trials with NOACs in AF patients have employed DT or SCA, despite growing interest in their use for improving trial efficiency and external validity. GENz-trials is uniquely positioned to pioneer the integration of these advanced technologies into the next generation of clinical trial research.