A proposed overhaul of how companies notify FDA about the safety of new ingredients could significantly increase the already resource-constrained agency’s workload and the wait time for companies wanting to bring new substances to market – but a scientific consulting firm is betting it can help reduce the time-to-market for some ingredients with the help of AI.
Under proposed changes to the Generally Recognized as Safe (GRAS) system, companies would need to notify FDA before marketing ingredients that they have determined are safe – rather than relying on the current voluntary notification system. But the proposal would preserve some narrow exemptions, including one when an existing FDA “no questions” letter covers the substance under the conditions of its intended use.
The catch? There is not yet a standard way for companies to evaluate and demonstrate their ingredient is sufficiently similar to an FDA-reviewed precedent. And dredging the agency’s databases for comparison information can be tedious and time-consuming.
SciPinion, an independent peer review and risk assessment business based in Montana, says its new SciBridge Safety Suit of AI-assisted software and professional services can help fill the gap and speed the assessment – “turning months of manual comparison work into a documented similarity report in minutes.”
The challenge and SciPinion’s solution
The potential value of SciPinion’s AI-powered shortcut depends on a relatively obscure part of FDA’s proposed GRAS overhaul, which is not finalized and for which the agency is seeking comments due Dec. 9.
According to the proposed rule, FDA would not require a GRAS notice if a previous notification covered the substance under the same conditions of intended use and FDA had issued a “no questions” letter regarding the notifier’s GRAS conclusion.
However, FDA specifically says that a no-questions letter would not apply if the identity, manufacturing process or conditions of use are “significantly different.” The agency adds in the proposal “if the use of a substance differs from the uses discussed in a GRAS notice that received a no questions letter response, it is the obligation of the manufacturer to demonstrate whether the substance is GRAS under the conditions of intended use.”
SciPinion’s new tool seeks to bridge the gap between the proposed regulatory framework and how companies can demonstrate that an ingredient is sufficiently similar to an FDA-reviewed precedent.
“The proposed rule states you may be exempt from a mandatory notification if your substance is the same as what we have already reviewed. They state that very clearly. But, they provide no guideline for doing similarity assessment reports,” said Vanessa Feher Castagna, director of food safety and claims at SciPinion.
Likewise, she said, companies have used similarity assessments when ingredients differ in areas such as use or manufacturing to bridge their safety assessments – potentially alleviating the need to reproduce toxicology studies, for example, or other key information.
“But again, there is no standardization, and the proposed rule specifically asks for comments on how to standardize this approach of incorporating data from precedent – which our tool does perfectly,” she explained.
She added SciPinion is in correspondence with FDA about its approach to similarity assessments and the agency was “very positive about this exact use case for industry.”
What does SciBridge’s solution offer?
SciPinion’s SciBridge Safety Suite tackles this challenge through three tools that are built directly on data from FDA’s GRAS Notification inventory.
The first allows users to browse the GRAS Notification Inventory and filter with status buttons and other specifications, including information about the companies, the products, the ingredients and more.
“What we’ve also done is given a user the ability to do a side-by-side analysis” to flag potential matches for further investigation, Feher Castagna said.
On top of this, SciPinion built a screening tool that draws on FDA guidance and feedback and allows users to quickly check for similarities and rank the new substance against precedent notices – further narrowing the field for investigation.
Users can take GRAS notifications that are similar to their substances and input them into SciPinion’s third tool, which it calls bridge.
Bridge generates a deep-dive similarity assessment covering upwards of 200 fields, with an AI-assisted review designed to be color-coded and user-friendly, Feher Castagna said.
“We do not expect this report in itself to be taken to FDA, but we expect it to assist in this similarity assessment,” she stressed.
She explains that while AI does the structured comparison, humans still interpret the places where the comparison breaks down.
“In the end, you get a report that shows how many specifications matches perfectly with that comparator, how many specifications may not match perfectly, but of all of the reviewed substances that are the same in identity, how it fits within that range of specifications that FDA has already reviewed, and then, where are there deviations or needs really a human to do that intervention to say this is saying more, this is not, and isn’t a safety concern?,” she said.
“We also build in critical fields that essentially say these are not comparable in any way, or these should not be used in a comparability assessment. For example, if there’s a substance that is very similar in identity, but the intended use is entirely different, like free infant formula versus adult. Really, it’s not it’s not going to be considered the same substance because that intended use is completely different and usually requires a different level of safety and evaluation,” she added.
‘We’re just taking the easy ones forward’
For SciPinion, the potential power of the tool goes beyond simply making a similarity assessment faster. It says it is also a way to help prevent an unnecessary new notification from further taxing FDA or delaying the marketing of a substance that may meet qualify for an exemption.
SciPinion estimates that there could be tens of thousands of ingredients that are sufficiently similar to substances FDA has already reviewed, although the company acknowledges there is no public data to establish how many such cases actually exist.
It argues identifying those cases could help reserve FDA resources for genuinely novel ingredients.
“We’re taking care of the easy ones. Leave the hard ones to FDA. You know, truly novel chemicals or truly novel ingredients. Those need to go to FDA. We’re not saying anything about those. You know, we’re just taking the easy ones forward,” said SciPinion CEO Sean Hays.
What remains unclear is how companies will ultimately have to demonstrate that similarity to FDA. The proposed rule does not spell out a standardized process, and SciPinion acknowledges that the final requirements have yet to be determined.
