Barriers, enablers and recommendations: How to drive a clinical development revolution2 Trust, regulation and fragmented data stand in the way of a Gen AI-powered clinical development revolution. Leaders interviewed by the World Economic Forum and ZS agreed on the biggest obstacles to the transformation of clinical development using Gen AI. Some, such as lack of public trust and data security, are common to any AI implementation; others, such as data fragmentation and the absence of regulations mandating which data must be shared, are more specific to clinical development. Barriers The quest to improve clinical development using Gen AI will depend on access to high-quality representative data. Three barriers stand out: –Data fragmentation: Aggregating clinical data from across the healthcare system is of paramount importance, but lack of industry consensus on what data is needed to advance clinical trials exacerbates fragmentation. Healthcare stakeholders have assorted aims, few incentives to share data and even fewer use cases demonstrating the value of greater data sharing. –Inconsistent data quality: Inconsistent collection practices, incomplete datasets and human error feed poor data quality. This impairs AI systems’ performance, eroding trust and slowing adoption. –Regulatory vacuum: Regulatory frameworks for AI and data in clinical trials are complex, inconsistent between geographical areas and open to interpretation. This creates ambiguity for stakeholders trying to balance innovation with compliance and patient safety. The European Union’s AI Act19 regulates the use of AI in the European Union, while the United States has a patchwork system,20 with states crafting their own rules.Recommendations Policy-makers and the private sector must join together to create shared infrastructure, standards and incentives to drive greater collection, maintenance and sharing of health data. Three actions are recommended: 1. Create standards for data collection and sharing: Regulators and pharmaceutical companies must drive policies that enforce health data standards. Policy-makers should consider mandating that some data, such as failed trial data, be made public after a blackout period. They should also modify regulatory frameworks to allow and encourage more adaptive trials. The USA’s Trusted Exchange Framework and Common Agreement (TEFCA)21 established a national framework for interoperability, but it does not create incentives for companies to share data. Pharmaceutical companies, meanwhile, should align on what data they need to optimize clinical trials. The answer will vary from one trial to the next, but patient health data, clinical data and trials data are the “big three”. Table 2, which shows the five categories of data that companies will need in order to fuel clinical trials, can be found in the Appendix.2.1 Data, incentives and regulation Intelligent Clinical Trials: Using Generative AI to Fast-Track Therapeutic Innovations 12

Intelligent Clinical Trials 2024