Manufacturing Intelligence: Organoids, Organ-on-a-Chip, and the Trust Gap

Source: Pharmaceutical Technology, by Richard Jaenisch (October 6, 2026). AI-generated summary by biochip.com, published October 7, 2026. Not independently reviewed.

New approach methodologies, or NAMs, including organoids and organ-on-a-chip systems, could shorten parts of drug development more directly than artificial intelligence alone, said Richard Jaenisch during Pharmaceutical Technology's Manufacturing Intelligence program. These human-relevant test systems aim to replace or reduce animal studies in areas such as drug toxicity. Their potential is substantial, but sponsors and regulators still need confidence that their results are reliable enough for high-stakes development decisions. The U.S. Food and Drug Administration's September 22 rule replacing animal-test language with nonclinical-test language could accelerate that transition.

What NAMs Are Designed to Change

NAMs are laboratory and computational methods intended to generate safety and efficacy evidence without relying exclusively on animal experiments. They include in silico models, meaning computer-based simulations, as well as organoids and organ-on-a-chip platforms.

An organoid is a three-dimensional cluster of human cells that mimics selected features of an organ. Think of it as a scaled-down working model rather than a complete replacement organ. A liver organoid, for example, can offer a focused way to study how a candidate medicine might damage liver tissue.

Organ-on-a-chip systems go a step further by placing cells in small fluidic devices designed to reproduce aspects of an organ's physical environment. The devices can connect models of multiple organs, such as a liver and kidney, through tubing that carries fluids between them. This creates an experimental system for studying how a drug may move through and affect different tissues.

Why Liver Safety Is a Key Test Case

Jaenisch pointed to drug-induced liver injury as one of the central safety problems for therapies moving through development. He highlighted a recent in silico model for this problem that advanced through the FDA's Innovative Science and Technology Approaches for New Drugs, or ISTAND, program.

ISTAND is an FDA pathway for qualifying drug-development tools for a defined use. Jaenisch expects many NAMs to follow a similar route as developers seek regulatory confidence in their models. Roughly half of NAMs involve AI in some form, he said, though these computer models are not necessarily the chatbot-style generative AI that has drawn broad public attention.

Animal studies can take a long time even in efficient, established vivarium systems. NAMs could potentially reduce that process by months today and by years in some future settings, depending on the development question and the stage at which a company applies the tool. Jaenisch said the immediate value for chemistry, manufacturing, and controls teams is more likely to come from human-relevant evidence than from major savings in manufacturing time itself.

The Trust Gap Around Human Models

Human-cell models may avoid some of the biological mismatches that occur when animal results are translated to people. But organoids do not reproduce the full environment of the human body, and a multi-organ chip raises its own validation questions.

For example, Jaenisch questioned whether a drug traveling through tubing between a liver model and a kidney model behaves the same way it would in human blood vessels, which differ in material and biological composition. Those questions matter because a promising device is not automatically a proven predictor of human outcomes. Scientists who use the platforms and regulators who review the data will both need to assess how closely each system reflects the specific biology it is meant to model.

Jaenisch framed this challenge through hepatitis C drug development. Chimpanzees had been the only animal model that worked with the virus, he said, and investment shifted when their use fell out of favor in the 2010s because sponsors wanted a more predictable route to approval. Early development of sofosbuvir used modified zebrafish and as many as 10 models, with each model supplying a different piece of toxicity information.

FDA Capacity Will Also Matter

The FDA's direct final rule and a database of roughly 25 real-world NAM use cases signal a regulatory opening for alternatives to animal tests. That database includes organoid and organ-on-a-chip data, making the platforms a concrete part of the agency's evolving nonclinical evidence base.

Still, adoption depends on more than the availability of new models. Jaenisch raised the question of whether a leaner FDA, with relatively few specialists in the field, can build enough expertise to evaluate an increase in NAM-based submissions. If companies invest heavily in these tools but regulators cannot review them at the same pace, that capacity constraint could limit their effect on development timelines.

Why This Matters

For drug developers, organoids and organ chips offer a possible route to earlier safety evidence based on human cells rather than non-human biology. That could be particularly valuable for organ-specific toxicity questions, where an animal model may provide incomplete or unexpected results.

Jaenisch expects it may take one or two years after the September 22 rule for the industry to see where confidence in NAMs is becoming established. In the meantime, chemistry, manufacturing, and controls leaders can begin building familiarity with NAM-generated data, its strengths, and the validation questions that will shape its regulatory use.

Topics: Organ-on-a-chip · Drug development & safety

Start here: Organ-on-a-chip explained