Must Drug Safety Always Pass Through Animal Testing First? FDA's 2026 NAM Framework Brings Organ-on-Chip Models to the Reproducibility Test

A transparent chip carries fine red and blue channels. Fluid moves at micrometer scale, and cells are arranged in a controlled environment that resembles part of human tissue function. This is the image that makes organ-on-chip or tissue-chip technology so compelling: a piece of organ-like function miniaturized onto a chip to help predict drug safety, toxicity, and efficacy. These technologies have often been presented as the future replacement for animal testing. In 2026, however, the most important development is not the slogan. It is that the U.S. Food and Drug Administration has brought New Approach Methodologies, or NAMs, into formal discussions about validation, study design, reporting, and regulatory submission.

The FDA's 2026 draft guidance does not announce that animal studies have disappeared. Instead, it explains how NAMs should be considered in drug development, including validation, study design, data quality, reporting, and regulatory use. FDA describes NAMs as innovative testing methods and strategies that can help evaluate the safety, effectiveness, and quality of drugs, biologics, and other FDA-regulated products. See FDA | General Considerations for the Use of New Approach Methodologies in Drug Development. This apparently cautious statement actually places human relevance and evidentiary reliability on the same table.

The shift matters because animal models have long been indispensable but imperfect in drug development. A mouse is not a small human. Some toxicities are not visible in animals, and some animal responses do not predict human outcomes. In fields such as liver toxicity, cardiac toxicity, immune response, rare disease, gene therapy, and personalized medicine, researchers have long wanted models that better approximate human physiology. Organ-on-chip systems are attractive because they use human cells and controlled microenvironments to simulate shear stress, flow, tissue interfaces, metabolism, and signaling. They can reveal responses closer to the body than a conventional dish culture in some defined contexts.

NIH NCATS has supported tissue-chip work for years, developing microphysiological systems built from human cells to help predict drug safety and toxicity and reduce uncertainty in development. See NIH NCATS | Tissue Chip for Drug Screening. Yet moving from research to regulation depends not on how elegant a chip looks, but on whether results remain stable, interpretable, and reproducible across laboratories, cell batches, operators, and drug types.

This is the core of the 2026 NAM discussion: context of use. A liver-on-chip may be highly useful for predicting a particular type of liver toxicity, but that does not mean it can predict every hepatic reaction. A heart-on-chip may capture changes in cardiac contraction, but may not reflect long-term immune or whole-body metabolic effects. Regulatory science does not ask only whether a model resembles the human body. It asks what question the model is prepared to answer, how far the evidence can support a decision, and how risk will be managed. If the context of use is not defined, even the most sophisticated chip remains a research demonstration.

Here, Two-Eyed Seeing does not refer to an Indigenous-policy setting but to reciprocal calibration among laboratory models, regulatory review, and clinical human response. Chip models can provide early human-relevant signals. Clinical data can then feed back into model revision. Regulators define evidentiary expectations, and researchers improve study design accordingly. Toxicologists identify limits of extrapolation, while engineers improve channels, cell sources, and sensors. Mature NAMs are not simply a moral declaration against animal testing. They are a system for producing evidence that can be reviewed, repeated, and checked against human outcomes.

FDA's broader NAM information also shows that the category is not limited to one technology. It includes in vitro systems, computational models, artificial intelligence, microphysiological systems, and other innovative methods. See FDA | New Approach Methodologies. This means future drug-safety evaluation may increasingly become a combined evidence chain: organ-on-chip systems provide tissue-level responses, AI models integrate large compound and clinical datasets, conventional toxicology provides comparison points, and clinical and real-world evidence feed back into the system. The point is not for one tool to replace every other tool. The point is for different tools to form reliable evidence in clearly defined uses.

Taiwan's biomedical sector should not respond to this shift only by purchasing chip devices or producing visually attractive demonstrations. It needs regulatory-science capacity. Research teams must know how to design validation studies that regulators can evaluate. Biotechnology companies must know how to present NAM data in investigational or marketing-application contexts. Hospitals and academic units must build quality systems for human cell sourcing, ethics, data governance, and documentation. Reviewers also need the ability to understand both the strengths and limits of these models, avoiding both uncritical enthusiasm and reflexive rejection.

Drug safety will not become simple because organ-on-chip systems exist. It will become more dependent on cross-disciplinary communication. If a chip result and an animal result conflict, which one should regulators believe? The answer cannot be fixed in advance. It must return to data quality, model purpose, existing clinical knowledge, drug mechanism, and risk context. The future of NAMs is not the elimination of uncertainty. It is making uncertainty more clearly described, validated, and managed. That is why the FDA's 2026 framework deserves attention.

Main reference sources

• FDA | General Considerations for the Use of New Approach Methodologies in Drug Development
• FDA | New Approach Methodologies
• NIH NCATS | Tissue Chip for Drug Screening