FDA officials: High-quality data is essential for AI tools for generic drugs

ROCKVILLE, MD — Robert Lionberger, director of the US Food and Drug Administration’s (FDA) Office of Research and Standards (ORS) within the Office of Generic Drugs (OGD), emphasized the critical need for high-quality data to support artificial intelligence and machine learning (AI/ML) models during a workshop held on 15 October.

The event, sponsored by FDA and the Center for Research on Complex Generics (CRCG), focused on regulatory insights and future trends in AI/ML. FDA officials also provided an update on the use of these tools by generic drugmakers.

Gabriel Innes, assistant director of data science and AI policy at FDA, said the number of drug applications incorporating AI has increased significantly over the past nine years. CDER has received more than 1,000 submissions with AI elements since 2016 when there was one NDA with AI and ML components. In 2024, there were 248 INDs with these elements.

These tools are used for patient selection, prediction of outcomes, pharmacometrics modeling, and post-market safety monitoring. Most of these AI elements in submissions are in the oncology area, followed by gastroenterology, neurology and psychiatry.

Innes “strongly encourages” early engagement in discussing use of AI models in drug and biologic products. “I can’t stress this enough,” he said.

High-quality data

Lionberger shared some of the lessons learned from the office’s use of AI/ML tools over the past seven years. He noted that the primary function of ORS is to conduct research to advance generic drug development and that the office is heavily reliant on the use of AI tools to assess product quality and equivalence.

Use of these models includes the use of the first ML-based predictive analysis for ANDA submissions. Additionally, OGD has utilized AI to create a data and text analysis tool called the Bioequivalence Assessment Mate (BEAM), which automates and streamlines the work involved in conducting bioequivalence assessments.

Officials are also using pyDarwin, a machine learning-based tool to help select pharmacokinetic (PK) models. Additionally, the office is conducting a feasibility analysis on using ChatGPT to assist in regulatory assessments.

Another AI tool being used by staff, the agency’s recently unveiled Elsa model, got high marks from Lionberger. He said that Elsa has been “fantastic for users.” He remarked that, “What I thought was really innovative, surprising, and fascinating is that it was made available to everyone immediately at FDA.”

On 2 June, FDA Commissioner Martin Makary announced the agency had launched Elsa almost a month ahead of schedule, claiming the tool had been proven to help staff streamline their work significantly. (RELATED: FDA launches agency-wide AI tool ‘Elsa’ ahead of schedule, Regulatory Focus 2 June 2025)

Elsa is a generative AI solution designed to enhance staff efficiency by summarizing documents and analyzing data. It is intended to handle specialized tasks, such as reviewing clinical trials and assisting with regulatory submissions and data analysis.

Lionberger stated that one of the lessons learned from ORS' years of utilizing AI and ML tools is the importance of having high-quality data.

“When we think about AI uses in our organization, and also in the generic industry, you are part of a large organization that wants to do things [to leverage AI]. We looked at this, and we thought, ‘What does it take to get AI tools to reach high reliability?’ One of the important parts of this is to have high-quality data. If you want to do something across the full space, you have to have the correct data available for that.”

He added that “having reliable and complete data is an underrated part of successful AI use … if you put incomplete data into the AI you will get an incomplete answer. That is a really important part of the foundation for long-term use of AI tools.”

Lionberger noted that FDA has a staff of about 2,000 people to oversee the entire US generic drug supply chain. He emphasized that utilizing AI tools could enhance the agency's efficiency, describing AI as a good tool for a “lean agency."

FDA officials: High-quality data is essential for AI tools for generic drugs

Related topics

Recommended content

Welcome to the new RAPS Digital Experience