Artificial intelligence (AI)-aided clinical data management: Applications, human-in-the-loop workflows, and...

Area of Science:

• Clinical Research
• Data Management
• Artificial Intelligence

Background:

• Clinical data management (CDM) is crucial for research quality.
• Japan faces challenges in CDM, including personnel shortages and increasing data complexity, especially in academic research organizations (AROs).
• Advancements in AI, particularly large language models, offer potential solutions for supporting CDM tasks.

Purpose of the Study:

• To review domestic and international examples of AI applications in CDM.
• To identify common implementation principles for AI in CDM.
• To discuss the implications of AI in CDM for regulatory frameworks and risk management.

Main Methods:

• Literature review of AI utilization in CDM-related tasks.
• Analysis of case studies on AI in data cleaning, medical coding, and query generation.
• Examination of emerging regulatory guidelines and principles for AI in clinical research.

Main Results:

• AI shows promise in supporting CDM tasks like data cleaning, medical coding, and query generation.
• A consistent implementation principle is the 'human-in-the-loop' design, where AI assists human decision-making.
• Regulatory frameworks are evolving, but integrating AI into Good Clinical Practice (GCP) requires further discussion.

Conclusions:

• Human-AI collaborative workflows are accelerating the shift from manual CDM processes.
• Robust technical, regulatory, and risk-management frameworks are essential for successful AI integration in CDM.
• AI and data quality have a symbiotic relationship, with improvements in one enhancing the other.