Leveraging AI to Streamline ClinicalTrials.gov Compliance: Early Performance Insights

INTRODUCTION: ClinicalTrials.gov registration is a key requirement during study start-up, but it is just one of many competing tasks to be managed. To address this, we piloted AI-powered Microsoft Copilot to streamline data entry and evaluate its acceptability, feasibility, and potential to improve efficiency.

METHODS: We invited study teams to pilot the AI-assisted process if they had submitted a study for IRB review that was interventional and corresponded to a single ClinicalTrials.gov record.
Three data sources were uploaded to Copilot: the IRB application, a list of fields to extract (with instructions), and an XML template to populate with the captured fields. Copilot generated an XML file, which we uploaded to ClinicalTrials.gov to create a draft study record. The study team reviewed the content for accuracy and entered any remaining required fields. We then conducted a standard review and submitted the record for ClinicalTrials.gov review.

RESULTS: The AI-assisted process was offered to nine study teams. Eight agreed to participate and one did not respond. ClinicalTrials.gov records were successfully created for all eight studies. We reviewed the AI-assisted draft records before sending them to the study team for their review. In most cases, the targeted fields were accurately extracted from the IRB application with minimal manual changes required.
Two of the study records have been submitted to ClinicalTrials.gov and both were assigned an NCT number during their first review cycle. Initial feedback from the study teams has been positive, including comments about the efficiency and speed of the process.

DISCUSSION: Our pilot testing of an AI-assisted process indicates it is a feasible alternative to manual ClinicalTrials.gov entry. Expanded use is needed to refine the process. Potential opportunities include using AI-assistance for complex fields (e.g. outcomes) and results submissions.