Key takeaways
AI/ML adoption is growing in clinical trials, with reported benefits such as 18% time reduction and protocol optimization, though challenges remain around data quality, trust, and legal concerns.
Companies are using AI/ML for protocol simplification, site burden analysis, and budget forecasting, helping to reduce complexity and improve trial efficiency.
Future innovations include digital twins, generative AI for synthetic data, and real-time AI feedback, promising to further transform trial design and execution.
In a breakout session at the 2025 Drug Information Association (DIA) Global Annual Meeting, industry leaders gathered to discuss the adoption of artificial intelligence (AI) and machine learning (ML) in clinical research. Using a recent industry survey conducted by the Tufts Center for the Study of Drug Development (Tufts CSDD) as context, the panelists touched on a number of topics including study forecasting, site selection, and protocol optimization.
The panel was led by Mary Jo Lamberti, PhD, director and research associate professor, Tufts CSDD. The participants were Emily Carter, director, DSA, trial execution, AbbVie; Luciana Petcu, senior manager, Takeda; and Mukul Virmani, MS, director, data scientist, clinical analytics, Gilead.
Tufts CSDD survey highlights growing AI/ML adoption across drug development
Lamberti began the discussion by giving an overview of a recent study conducted by the Tufts CSDD on the adoption of AI/ML in clinical research. Tufts CSDD and DIA collaborated with 16 biopharma companies and contract research organizations to gain a better understanding of the current use of AI/ML.
Using a global survey and use case interviews, the survey gathered 302 complete responses from drug development professionals. Among other notable data, the study found that $1.05 million was invested in AI/ML use by activity assessed, teams experienced an average time reduction of 18% using AI/ML, and overall, respondents reported a positive outlook on the use of AI/ML in drug development.
Challenges to AI/ML implementation in clinical research remain
The study also explored the challenges of AI/ML adoption. Lamberti explained: “Not only is there a need for massive amounts of data, but also sufficient quality data to train systems. The second area is trust […] and we saw that throughout both our survey interviews, or lack of trust, for example, in AI-generated results, forecasts, or recommendations, and that includes data driven or algorithmic biases as well. A third area was intellectual property and other legal concerns related to data sharing.”
Real-world use cases of AI/ML in clinical operations
Following the study overview, Lamberti turned the discussion over to the panel to elaborate on its experiences in adopting AI/ML. One of the areas the participants touched on was use cases from their own companies.
Virmani began by describing how his team at Gilead is working with its clinical development stakeholders to optimize study protocols by using data components such as criteria and schedule of assessments.
“We analyze all of that data and give them an industry comparison of a similar kind of setup, or a study. […] We calculate the scopes, like patient burden, site burden, and overall budget, or the cost of this trial, and compare it with different industry sponsors,” he said.
Petcu also chimed in on addressing protocol complexity. She said: “I think specifically as we think about protocol complexity, we're able to surface those insights in real time and make adjustments to the protocol design and really showcase where some of that burden might lie for our patients or our sites. There's a real opportunity to continue to make sure that our protocols are optimizing.”
Carter then added by explaining an area that her team is excited about—using generative AI techniques to improve productivity. She explained that her team is seeing an increased demand for AI-driven insights, leading them to explore new ways to open capacity.
“We've introduced a lot of learning sessions and knowledge exchanges just within our teams on ways to implement these new technologies and new ways to look at and analyze data,” she said.
The future of AI/ML in clinical trial design and execution
As the discussion came to a close, Lamberti asked each of the panel members to share their future outlook for the use of AI/ML in clinical trials. Virmani highlighted the use of digital twins and how it brings multiple aspects to the table including real-world evidence, regulatory affairs, and biometrics. Carter discussed decision augmentation when risks are identified at the site, country, study, and portfolio levels. Finally, Petcu touched on adaptive clinical trials and using generative AI for synthetic data generation.
“Is there a potential to use real-time AI feedback to address our trials? I think we have some work to do around protocols and data infrastructure there, but a really exciting opportunity,” Petcu said.
Reference
Lamberti M, Carter E, Petcu L, Virmani M. The Adoption of Artificial Intelligence and Machine Learning in Clinical Research. June 17, 2025. 2025 DIA Global Annual Meeting.
