The Next Frontier of Clinical Research is AI

For decades, familiar regulatory pathways have paved the landscape towards investigational product approval. Industries such as pharma, biotech and medical devices all follow the same pathway to market which includes preclinical testing, clinical testing and post-market surveillance. The time has come where we begin thinking about AI in medicine and how to generate clinical evidence to support it.

AI is increasingly being incorporated into medicine, and thus, clinical evidence must evolve alongside it since algorithm changes will certainly cause regulators to ask if updates to

medicinal AI products are still safe and effective.

In January 2025, the International Medical Device Regulators Forum (IMDRF) published Good Machine Learning Practice for Medical Device Development: Guiding Principles, building on the foundational work first introduced by international regulators in 2021. The document outlines ten guiding principles intended to support the development of safe and effective AI-enabled medical devices.

One guiding principle emphasizes that AI should be evaluated within the context of the human-AI team and its intended clinical environment, rather than assessing the algorithm in isolation. This recognizes that successful AI depends as much on how clinicians interact with the technology as on the algorithm itself.

Successful AI-driven products require a collaborative team. Engineers build the technology. Clinical research teams generate evidence that earns trust. Without trust, regulators will not approve the products, physicians won’t adopt them and patients won’t rely on them.

As a clinical researcher, if I were developing a study for a machine learning product, some of the questions I would ask:

What is the clinical assessment AI is making?

• What is the AI driven product trying to solve?

• What are we comparing AI to in the trial?

• What happens if the algorithm is wrong?

• How will the company ensure compliance as the models evolve?

The clinical researcher should also consider whether the regulatory pathway will be a 510(k) submission, De Novo, or via a pre-market approval (PMA) and begin thinking about post-market surveillance strategies if trial results are favorable.

AI is transforming how medicine is practiced, but it doesn't eliminate the need for rigorous clinical research. If anything, it raises the standard. The future of healthcare will belong to the companies that build the smartest algorithms, and to those that generate the strongest evidence. Innovation may begin with code, but trust will always be earned through science.