From single-point tools to a unified platform architecture, AI for clinical trials is accelerating its implementation | Frontline

Text by Liu Lingguo

Edited by Hai Ruojing

"AI has demonstrated the potential to accelerate clinical development. Many tasks that used to slow down new drug development can be completed faster by AI."

On September 3, Medidata, a provider of clinical trial solutions, held its NEXT China Annual Conference in Shanghai to discuss the application trends and large-scale implementation prospects of AI in clinical trials.

At the conference, Medidata CEO Anthony Costello, Chief Operating and Strategy Officer Lisa Moneymaker and other senior executives of the company focused on showcasing Medidata Plus, which the company launched in July this year, as well as the intelligent assistant Dot. The former is positioned as an AI platform architecture covering the entire process of clinical trials, while the latter, as the core of the former, acts as an AI orchestration engine that integrates research, patient and data experiences to realize an intelligent collaborative clinical trial workflow.

Medidata held its NEXT China Annual Conference in Shanghai

In recent years, the application scenarios of AI in the pharmaceutical industry have been continuously expanding, and its capabilities in identifying targets, designing molecules, and accelerating the advancement of candidate compounds into clinical practice are obvious to all. However, about 90% of the candidate drugs that entered clinical development in the past failed at this stage due to issues such as safety risks and insufficient effectiveness. How to accelerate clinical development and bring more high-quality drugs to market has become a hot topic in the second half of AI-driven pharmaceutical development.

Previously, most enterprises mainly used AI as a single tool and embedded it into different clinical trial scenarios separately. For example, large models are used for tasks such as reading medical records, matching patients, sorting out adverse drug event reports, and screening literature.

However, the limitation of single-point tools lies in the relatively scattered data, which increases the complexity of system integration and supplier management. Based on this, Medidata launched Medidata Plus, aiming to connect the AI capabilities scattered in different products to a unified platform, and then embed it into the original clinical workflow of pharmaceutical companies.

"The key is not to put AI into the system just for the sake of using it, but whether it can bring practical impacts, make work progress faster, make resource utilization more efficient, and improve the quality of clinical trials," Moneymaker stated.

The construction of AI capabilities is closely related to Medidata's long-accumulated EDC business. Its core product Rave is used to collect and manage clinical trial data of patients, and it is also the main entry point for Medidata to enter the clinical workflow of pharmaceutical companies. "We have signed data rights agreements with more than 90% of our clients, authorizing the compliant use of operational data," Anthony Costello mentioned in the media roundtable.

It is understood that its platform has cumulatively supported more than 38,000 clinical trials, involving 12 million trial participants. Massive data forms the foundation for Medidata to train AI.

In terms of specific applications, Medidata stated that after pharmaceutical companies, CROs and other parties purchase Medidata Plus, they can call capabilities such as automatic library construction and protocol optimization in multiple clinical projects. The platform can also predict the performance of research sites, and complete tasks such as data conversion and risk identification.

Research library construction is a scenario that was implemented at an early stage. At the conference site, after the staff uploaded the PDF of the research protocol, the system automatically extracted the visit schedule and evaluation plan, then generated forms, inspection rules, etc., and constructed test data. Medidata said that traditional library construction usually takes 10 to 12 weeks, which can be shortened to several days after using AI. According to Medidata's disclosure, in the past ten years, its AI technology has been applied to more than 500 clinical studies.

Trial simulation is another core scenario. Before the first patient is enrolled, the clinical team can repeatedly "test run" the protocol in the system. For example, test how many more eligible patients there will be if one of the enrollment and exclusion criteria is relaxed; how changing the combination of research sites will alter the recruitment speed, cost and completion time, etc. The system can compare these variables with the actual performance of similar historical trials, and predict the possible risks in protocol execution accordingly.

This means that some problems that could only be exposed after the trial was launched in the past have the opportunity to be discovered before patient enrollment and capital investment. Medidata disclosed on site that about 11% of research sites eventually failed to enroll any patients. "We want you to know how the trial will run before it actually runs," said Jeff Ventimiglia, Senior Vice President of Operations and Transformation at Medidata.

It is understood that in the future, Medidata plans to continuously connect new AI functions to Medidata Plus, and cover more clinical workflows through Dot.