Manufacturing Intelligence: What FDA's NAM Rule Means for CMC Teams

The FDA's September 22 direct final rule, which swaps "animal tests" for "nonclinical tests," gives new approach methodologies (NAMs) firmer regulatory footing, but its real impact on development and CMC timelines is still taking shape. On episode 7 of PharmTech's Manufacturing Intelligence, co-host sChris Cole and Richard Jaenisch, MPH, a liver transplant recipient and patient advocate, speak about what the rule and FDA's accompanying database of roughly 25 real-world NAM use cases mean for sponsors, regulators, and manufacturing teams.

Jaenisch notes that many NAMs are themselves a form of AI, though in silico models differ from the generative AI most people picture. He points to FDA's recent acceptance of an in silico model for drug-induced liver injury (DILI) through its ISTAND pilot program as a logical starting point, given how often liver toxicity derails development. Compared with the lengthy prep and run time of animal studies, NAMs could trim a few months from the pipeline today and potentially years in the future, depending on the difficulty of the target. On the production side, however, he does not yet see significant time savings.

Trust remains the open question. "The benefit of an organoid in this case is that it's human cells. It's a mock human organ," Jaenisch explains, though organoids cannot replicate the full body environment. Organ-on-a-chip systems link multiple organs, but the tubing connecting them raises questions about whether drugs move as they would between human organs. Drawing on his experience as a hepatitis C survivor, he recalls that chimpanzees were the only viable animal model for the virus, and that early sofosbuvir research relied on a patchwork of modified models, including zebrafish. He expects it will take a year or two for trust to be genuinely earned.

Whether NAMs actually reduce the manufacturing and quality burden, he argues, depends on the NAM. A mouse model reveals effects on the kidneys, lungs, and other organs; a single-organ system does not. AI-driven models carry their own risk: "Say that model hallucinates, and say that hallucination is now part of your record, and now that resets your entire clock," Jaenisch warns. He does not believe the rule lowers evidentiary standards, noting it is backed by the FDA Modernization Act and that the context-of-use requirement keeps the process strict yet streamlined. For rare diseases that cannot be tested any other way, the pathway holds real promise. Risk-averse sponsors may submit both NAM and animal data, adding cost but yielding richer datasets that, paired with the plausible mechanism pathway, could speed novel therapies toward commercialization.

Asked about bioprinted replacement organs, Jaenisch calls them a fantasy he would love to see realized, but estimates they remain about a decade away. Today, printed organs and organ-like cell groups are used mainly for drug testing, and implanting one raises a paradox: when the organ itself is the product, what serves as the test? With tens of thousands of US patients waiting for transplants, he is eager to see how regulators answer that.

For quality and manufacturing teams, his advice is to engage now rather than simply wait and see. Though technically a definition change, the rule is sweeping in scope. He recommends commenting on the rule, reviewing the use-case database—which spans roughly a dozen underlying methods—and mapping each relevant case to in-house production components. A DILI use case, for example, may prompt more focused toxicity testing, and whether more data means more or less work depends on its quality. Teams that do that homework now will be better positioned as NAM-influenced CMC packages arrive.