Highlights
- AI has the potential to revolutionize every stage of the clinical trial lifecycle, but quantitative evidence of its impact remains limited.
- Value-based frameworks are urgently needed to guide the effective adoption and measurement of AI’s benefits in clinical trials.
- Stakeholder involvement—including patients—is critical to ensure the meaningful integration of AI tools and avoid inefficient, ad-hoc deployment.
- Continuous evidence synthesis and monitoring are essential for timely and ethical adoption of AI innovations in trial design and conduct.
Study Background and Disease Burden
Clinical trials are the bedrock of evidence-based medicine, informing regulatory decisions, clinical guidelines, and patient care. Yet, the reality is that many trials are protracted, costly, and often fail to deliver actionable results—sometimes described as being “uninformative.” This inefficiency not only delays the availability of new therapies but also exposes participants to unnecessary risks and perpetuates resource waste in research ecosystems. The rise of artificial intelligence (AI) offers the tantalizing prospect of transforming clinical trial design, conduct, and analysis by automating labor-intensive tasks, optimizing decision-making, and improving trial robustness. However, the translation of this promise into meaningful, measurable outcomes remains inconsistent, and there is a lack of consensus on how best to assess and value AI’s contributions in this setting.
Study Design
Mateen et al. (Lancet Digit Health, 2025) provide a conceptual analysis rather than a primary clinical study. Their work synthesizes current literature and practical experiences, critically examining where and how AI can be leveraged in the clinical trial lifecycle. The authors highlight the lack of high-quality, quantitative studies evaluating AI’s impact and advocate for a structured, value-driven framework—akin to quality-adjusted life years (QALYs) in health economics—to guide AI adoption in clinical research. They discuss potential metrics, methods (such as discrete choice experiments), and the importance of broad stakeholder engagement.
Key Findings
1. AI Applications Span the Entire Clinical Trial Lifecycle
From protocol development (e.g., AI-enabled design tools such as Protocol AI) to patient recruitment (e.g., large language models screening electronic health records), data monitoring, and result reporting, AI technologies can enhance efficiency and potentially improve the informativeness of trials. However, the mere capability of AI to perform these functions does not guarantee value—context and outcome-specific assessment is essential.
2. Current Evidence Is Scarce and Fragmented
Despite widespread enthusiasm, there remains a paucity of robust, quantitative studies assessing the real-world impact of AI in clinical trials. Most evidence is anecdotal or based on early adopter experiences. For example, while an hour saved in patient screening by AI may be measurable, its comparative value versus an hour saved in protocol development is not standardized and may differ markedly depending on trial context, disease area, and urgency (e.g., outbreak response versus chronic disease trials).
3. Need for a Value-Based Assessment Framework
Drawing parallels with QALYs and DALYs in health technology assessment, the authors argue for the urgent development of a standardized value framework for AI in clinical trials. This would enable meaningful comparison across diverse AI tools and applications. Discrete choice experiments—used to quantify stakeholder preferences—are proposed as one method to capture the heterogeneous values assigned to different improvements (e.g., efficiency, robustness, reproducibility).
4. Stakeholder Inclusion Is Critical
Effective value frameworks must incorporate perspectives not only of trialists and sponsors but also patients and communities, aligning with recent trends in participatory research and regulatory guidance. Failure to do so risks the indiscriminate, inefficient, or even harmful adoption of AI tools that may not align with patient interests or societal needs.
5. Three Key Metrics: Informativeness, Cost-Effectiveness, Reproducibility
Many AI solutions target a single dimension of value (e.g., cost minimization or recruitment speed), but broader ecosystem benefits (such as addressing the reproducibility crisis) are often underappreciated. Overreliance on simple, tangible benefits like cost savings could inadvertently sideline innovations that improve research integrity or the probability of portfolio-level trial success.
6. Portfolio Optimization and Predictive AI
Recent advances suggest AI can predict the likelihood of trial success (i.e., delivering unambiguous results), opening the possibility of optimizing entire trial portfolios for efficiency and informativeness rather than focusing solely on individual studies.
7. Continuous Evidence Synthesis and Institutional Monitoring
The authors advocate for the establishment of dedicated observatories or monitoring bodies to maintain living systematic reviews of AI applications in clinical trials, ensuring timely synthesis of evidence and signaling when adoption of a specific AI tool becomes ethically and scientifically obligatory.
8. Implications for Regulatory Affairs and Ecosystem Processes
Beyond trial conduct, AI can enhance related processes such as regulatory intelligence, multi-jurisdictional protocol design, and accelerated regulatory review. The challenge ahead is to prioritize, fund, and scale the highest-impact solutions, recognizing that resources are finite.
Expert Commentary
Current opinion leaders and policy documents (e.g., EMA and FDA guidances on AI in medicine) echo many of these themes, emphasizing that AI must be deployed with rigorous validation and transparent evaluation frameworks. The lack of standardized metrics for AI value in clinical trials is a recurring concern, as is the necessity for ongoing stakeholder engagement and patient-centered approaches. The potential for AI to exacerbate existing disparities—if not carefully monitored—has also been highlighted in recent commentary. Limitations of the proposed framework include the inherent subjectivity of value judgments and the challenge of operationalizing complex, multi-attribute metrics in routine practice. However, the direction is clear: without a structured, evidence-driven approach, the risk of inefficient, inequitable, or even counterproductive adoption of AI in clinical trials is real.
Conclusion
Artificial intelligence holds substantial promise for transforming clinical trials, but meaningful adoption requires a shift from opportunistic implementation to strategic, value-based integration. A universal framework—grounded in robust evidence and inclusive of diverse stakeholder perspectives—is urgently required to assess, compare, and prioritize AI solutions. Continuous monitoring, synthesis of evolving data, and ethical stewardship will be essential to ensure that AI delivers on its potential to enhance trial efficiency, informativeness, and reproducibility while safeguarding participant welfare and research integrity. The next decade must focus not on whether AI will impact clinical trials, but on how best to realize and measure its highest-value contributions.
References
1. Mateen BA, Moorthy V, Labrique A, Farrar J. Artificial intelligence and clinical trials: a framework for effective adoption. Lancet Digit Health. 2025 Jul 23:100898. doi: 10.1016/j.landig.2025.100898.
2. 欧洲药品管理局. 关于医药产品生命周期中使用人工智能（AI）的反思文件. EMA/153729/2023.
3. FDA. 医疗器械软件中的人工智能和机器学习. 美国食品药品监督管理局. 更新于2024年.
4. Hutchinson, N ∙ Moyer, H ∙ Zarin, DA ∙ et al. The proportion of randomized controlled trials that inform clinical practice. eLife. 2022; 11, e79491
5. Hutson, M. How AI is being used to accelerate clinical trials, Nature. 2024; 627:S2-S5