Quality Quirks (Pt 3)

AI

In this day and age, a person can't go more than a single web page without stumbling into some form of AI. Google search alone now has Google Gemini, where every google search generates information derived from the various results that can come up during a normal search. However, how feasible is it for people like you or I to use AI as a tool in the biotech and Pharma industry?

When I first thought of AI the only thing I could think of was that Marvel villain, I think he was called Ultron. That essentially wanted to end the world after researching for about five minutes and decided that Humanity was doomed to destroy itself. However, from what I can tell (at least thus far) that's not what's happening. AI has been utilized for everything from Chat Bots, writing tools, art generators, and even analytical tools for scientific research. Its a wide spectrum of uses, some good, some bad, and some weird.

In the Biotech and Pharma industry specifically I can find a number of documents sourced from the FDA that have some guidance on AI usage. One document, which references many others, is the Artificial Intelligence for Drug Development Article on the FDA's website. It highlights what AI is, it's roles in drug development, and the Center for Drug Evaluation and Research's (CDER's) association with it. There are also links to other articles, including the draft guidance that I'm most interested in: Considerations for the Use of Artificial Intelligence to Support Regulatory Decision Making for Drug and Biological Products. The article is broken up into seven distinct steps:

1. Define the Question of Interest

2. Define the Context of Use for the AI Model

3. Assess the AI Model Risk

4. Develop a Plan to Establish AI Model Credibility Within the Context of Use

5. Execute the Plan

6. Document the Results of the Credibility Assessment Plan and Discuss Deviations from the Plan

7. Determine the Adequacy of the AI Model for the Context of Use

Seems pretty straightforward right? However, with the way quality and regulatory works the verbiage in each of these steps can be interpreted in many different ways. What's more, this article talks both about how to use AI as the product, and how to use AI as part of the regulatory submission. Two very different aspects of the same industry. The words, "context of use," will be key in the future to ensure that the AI model works effectively. Many companies will be relying on those three words to justify their usage of an AI, and may even try to use it with a flexible enough meaning to try and stretch it across both aspects. It sounds farfetched, but there are a number of folks whose priority is deadlines over Quality, and will do quite a bit to meet them. Including flexing the meaning of words far beyond their capabilities. Which makes me wonder if secondary reviewers that are not of the 'robotic persuasion' would be necessary moving forward. Which feels like it would render the use of the AI reviewer (if the context of use involves one) kind of useless. Especially if the AI reviewer can't catch what the Human reviewer can.

There is a lot to be considered when it comes to AI, and it feels like the road should be longer than it is when it comes to allowing it's use in our industry. We can ask it so many questions and it can provide answers, but are they the answers we're looking for or the answers we want to see?