We are delighted to share the publication of “Causal inference and digital twins: a roadmap for the future of clinical trials” in npj Digital Medicine.
Led by Professor Mihaela van der Schaar in close collaboration with CCAIM’s academic and industry partners, the paper brings together expertise across artificial intelligence, clinical research and pharmaceutical R&D. It reflects the collaborative model at the heart of the Cambridge Centre for AI in Medicine: connecting methodological advances in AI with the scientific, operational and regulatory realities of developing new treatments.
Clinical trials remain essential for establishing whether treatments are safe and effective, but they are often lengthy, costly and operationally complex. They may also struggle to capture differences between individual patients or determine how well findings will apply beyond the original trial population.
The new paper explores how causal inference and digital twins could help address these challenges.
Causal inference methods can support questions about the effects of particular treatments and decisions: which patients are most likely to benefit, how treatment effects vary across subgroups and whether results from one population are likely to generalise to another.
Digital twins provide a complementary capability. By creating computational representations of patients, populations or clinical processes, they could allow researchers to explore potential outcomes under different treatments or trial designs.
Together, these approaches could support many stages of clinical development, including patient recruitment, trial design, treatment-effect estimation, safety monitoring and the translation of trial findings into clinical care.
A collaborative roadmap
The publication develops the ideas first introduced in the 2025 manifesto “Revolutionizing Clinical Trials: A Manifesto for AI-Driven Transformation.” also featured on our website: read the blog article here.
The manifesto established a shared ambition for AI to become part of a more dynamic, rigorous and learning-based approach to clinical development. The new peer-reviewed Perspective takes this further, providing a more detailed roadmap for incorporating causal inference and digital twins throughout the clinical-trial lifecycle and examining the technical and practical requirements for their responsible use.
CCAIM’s funders and partners are central to this work. Their experience ensures that new methods are considered in the context of the real decisions, constraints and uncertainties encountered in pharmaceutical R&D and clinical development.
As Tony Wood, Chief Scientific Officer, Head of R&D and a member of the Executive Committee at GSK, explains:
“At GSK, we are advancing a more dynamic model of clinical development in which data, technology, and collaboration are tightly integrated. AI methodologies such as causal inference and digital twins enable a more rigorous characterisation of treatment effects, improved identification of responsive patient subgroups, and more efficient trial designs. By integrating these approaches, we can enable more adaptive, precise, and patient-centred development and accelerate the delivery of truly transformative medicines.” – Tony Wood, CSO at GSK
Justine Rochon, Head of R&D Data & Quantitative Sciences at Takeda, similarly highlights the importance of bringing together different perspectives:
“This work is close to my heart because it shows the power of genuine multidisciplinary, multistakeholder collaboration. By working together, we can see what others have not yet seen, take ideas further, and turn them into real progress for patients.”
Ramon Hernandez Vecino, Global Head of Development Real World Evidence at Sanofi also reflects on the significance of the work:
“This roadmap comes at a pivotal moment as the industry moves from exploring AI concepts to implementing them in drug development. While many organizations are discussing the potential of causal inference and digital twins, Sanofi is uniquely positioned to help operationalize these approaches at scale, translating methodological innovation into practical solutions that support clinical development and evidence generation. We hope this work will help accelerate the responsible adoption of AI-driven approaches across the industry.”
These contributions demonstrate why close collaboration between academia and industry is essential. Academic researchers can develop new methodological foundations, while pharmaceutical and clinical-development leaders provide direct knowledge of the settings in which these methods must operate.
By bringing these perspectives together, research can remain focused on genuine medical needs and account for the complexity, uncertainty and practical constraints involved in developing new treatments.
Towards more adaptive and patient-centred trials
The paper emphasises that causal inference and digital twins must be introduced responsibly. Their use will require rigorous validation, representative data, transparent assumptions, careful assessment of uncertainty and continued engagement with clinicians, statisticians and regulators.
These methods are not intended to replace established expertise. Instead, they could strengthen clinical development by helping researchers ask more precise questions, better understand differences between patients and explore alternative decisions before they are implemented in practice.
The publication represents an important outcome of CCAIM’s collaborative approach. It shows how sustained engagement between AI researchers and leaders across medicine and pharmaceutical R&D can produce shared frameworks for addressing some of the most difficult challenges in clinical trials.
Read the full paper: Causal inference and digital twins: a roadmap for the future of clinical trials