Just a week into the new year, the U.S. Food and Drug Administration (FDA) released it’s first draft guidance on the use of artificial intelligence (AI) for the development of drugs and biologics. According to the regulator, the use of AI in drug development and in regulatory submissions has “increased exponentially” since 2016.
However, the FDA isn’t policing how life sciences organizations can use AI on a day-to-day basis – it’s making sure AI can be trusted when it’s used to help the FDA decide whether a drug is safe and effective. Some examples include predicting patient outcomes, improving understanding of predictors of disease progression, and processing and analyzing large datasets, such as real-world evidence.
“With the appropriate safeguards in place, artificial intelligence has transformative potential to advance clinical research and accelerate medical product development to improve patient care,” said FDA Commissioner Robert Califf, M.D., in a press release.
What the guidance covers
The draft guidance, which is open for public comment until April 7, proposes a risk-based framework for credibility. Life sciences companies are encouraged to explain how they will use AI (“context of use”), assess the risks associated with that use case, and gather evidence that the AI’s outputs are reliable and logical for the intended purpose.
Here’s a hypothetical example:
Guidance does not apply: A pharmaceutical company is running a clinical trial for a new drug. They use AI to accelerate scientific literature reviews, standardize unstructured clinical notes and real-world data, and extract complex eligibility criteria. Humans review the results and decide the next steps accordingly.
Guidance applies: The above pharma company uses the same AI model to analyze adverse events and determine which safety indicators to include in the data submitted to the FDA. Unlike the previously mentioned use cases, which primarily improve efficiency, this use case could directly influence whether a drug is approved.
As life sciences organizations increasingly look to AI to support high-stakes decisions, several teams are rethinking how clinical data is organized behind the scenes. To learn more about how structured data improves evidence generation, check out our recent blog, Faster, smarter, scalable: Redefining evidence generation in life sciences.
Why structured data is key
While the FDA’s guidance focuses on AI models themselves, it also implicitly signals the importance of the data these models are trained on. AI systems used in regulatory contexts depend on clinical data that is consistent, structured, and accurate. Ambiguous or disorganized data can directly affect how models function and how their outputs are interpreted. This is why questions about AI credibility must start with the source data – not the model itself.
IMO Health’s life sciences solutions operate upstream, combining clinical AI, domain-specific natural language processing (NLP), and biomedical data science expertise to streamline drug development. By standardizing and structuring clinical data before it’s analyzed or submitted as evidence downstream, IMO Health helps ensure models are working from consistent, up-to-date information.
