Feasibility Medium Impact High Time horizon Medium (2–4 years) Barriers Fragmentation of health data and lack of standardized data and data-sharing practices. Regulatory requirements concerning patient data privacy and security. Frequency of protocol amendments caused by suboptimal protocol design. Biggest unlock Collaborative efforts to develop robust data integration standards and enhanced regulatory frameworks. Optimization of the trial design process resulting in fewer protocol amendments. Trial feasibility and site selection today The growing complexity of clinical trials has made it increasingly difficult to evaluate trial feasibility. Slower-than-expected patient enrolment is an ever- present challenge, typically taking 1.8 times longer than planned,13 with enrolment shortfalls affecting roughly 85% of clinical trials. Each day that a trial extends beyond its patient enrolment deadline translates to between $600,000 and $8 million14 of missed market opportunity.Today, evaluating trial feasibility and selecting trial sites is an ad hoc process dependent on limited data. Sites are chosen based on past experiences with trial sites, using metrics such as site capacity and historical enrolment rates. This often overlooks site characteristics with a stronger correlation to suitability, such as local patient demographics, disease prevalence and trial-specific site capabilities. Incomplete site profiles, data siloing and lack of real-time access to data complicate efforts to make data-driven selection decisions, resulting in under-enrolment, costly delays and unreliable feasibility decisions. Trial feasibility and site selection enhanced with Gen AI Gen AI can analyse unstructured data, including operational data from past trials alongside real- time RWD, to feed models that predict future outcomes and inform site selection and feasibility planning. By pairing Gen AI with predictive AI, trial planners will be able to accurately predict sites’ patient recruitment potential; suggest optimal sites based on up-to-date, multidimensional data; and anticipate recruitment challenges in advance. Gen AI will also enable site planners to simulate scenarios to explore various site configurations and their expected patient recruitment and retention outcomes. In addition, Gen AI can enable decentralized trials, a key industry initiative to make trials more patient-centric, by managing the logistics of using more sites and trial investigators.Evolution, not revolution: Trial feasibility and site selection 2 Amgen, a US-based biotech, has developed an AI-driven tool it calls the Analytical Trial Optimization Module (ATOMIC).15 It is designed to enhance the efficiency of clinical trial site selection by pairing classical and Gen AI to analyse large structured and unstructured datasets to identify those trial sites most likely to meet participant enrolment goals. By ranking sites based on predicted enrolment rates and other key factors, ATOMIC helps optimize trial design, reduce trial duration and increase the probability of trial success.USE CASE Amgen For trial design, operations and site selection, there is a lot of promise … especially where natural language data can be structured and analysed. Andrew Giessel, Executive Director of Artificial Intelligence Engineering, Moderna Intelligent Clinical Trials: Using Generative AI to Fast-Track Therapeutic Innovations 8

Intelligent Clinical Trials 2024