Seventeen years after a Harvard team built a breathing human lung-on-a-chip, alternatives to animal testing are moving from proof of concept to practical tools for drug development and safety assessment. The shift, once unthinkable to many regulators and researchers, is increasingly visible in laboratories and policy.
Cell biologist Donald Ingber and colleagues at Harvard University’s Wyss Institute first described a transparent polymer device, smaller than a USB stick, whose channels were lined with human lung and blood-vessel cells. When air was pumped alongside the channels, the device expanded and contracted.
It responded to inflammatory proteins and bacteria much like a real lung and showed how motion changes the absorption of ultrafine particles. Editors initially asked the team to validate results in mice before publishing, underscoring the field’s reliance on animal data at the time. The study appeared in 2010 and has since been cited thousands of times.
Attitudes have shifted. Ilka Maschmeyer of TissUse, a German company building organ-on-a-chip platforms for pharmaceutical clients, recounted a recent case in which a company was denied permission by the U.S. Food and Drug Administration to begin a clinical trial after submitting only animal results.
The agency asked for data from organs-on-chips or comparable methods. Maschmeyer called such requests rare but increasingly common.
Collectively known as NAMs, for new approach methodologies or nonanimal methods, these alternatives now include organoids, organ-on-a-chip devices, computational models and artificial intelligence systems. Early findings are promising, but widespread use hinges on validation, standardization and cultural change.
“This transition process is much more complicated than you would think,” said Thomas Hartung, who leads the Center for Alternatives to Animal Testing at Johns Hopkins University. “It is more about change management than it is about the technology.”
The technologies driving alternatives to animal testing
Animal studies have long drawn criticism on ethical and scientific grounds. Roughly 92 percent of drugs entering U.S. clinical trials never reach the market, often due to safety or efficacy issues that animal models failed to predict.
Some researchers argue that widely used medicines might not have survived today’s animal-heavy testing paradigm had they been discovered later.
Modern NAMs reach beyond static cell cultures. Organoids better mimic organ structure and function.
Organ-on-a-chip systems now model lungs, brains, hearts, kidneys and placentas. Multi-organ platforms link as many as 10 tissues to approximate whole-body physiology.
These are paired with in silico simulations and AI tools that analyze outputs and guide experiments.
Policy has begun to catch up. In late 2022, the FDA Modernization Act 2.0 cleared the way for NAMs to be used in preclinical studies.
In 2025, the FDA said it aimed to make animal studies the exception for drug safety testing. In September 2026, the agency proposed replacing references to “animal tests” in drug-development rules with “nonclinical tests,” explicitly accommodating validated human-cell systems, organ chips and computer models.
That same period, the U.S. National Institutes of Health said grant applicants using animal models must also incorporate nonanimal research, and European regulators signaled plans to phase down animal testing. The Organisation for Economic Co-operation and Development updated guidelines to broaden NAM use.
Ingber called the regulatory shift “game-changing.”
Proving that NAMs work
Regulatory openness alone is not enough. NAMs must be validated with standardized materials, workflows and performance metrics so results are reproducible across sites.
Even subtle differences in tissue scaffolds can alter outcomes. Companies such as Emulate, founded out of Ingber’s work, provide commercial chips and training, yet users often need help with technical details such as stem-cell culture.
Clinical relevance is the ultimate test. Emulate reported that a liver-on-a-chip correctly identified about seven of eight drugs that had passed animal studies but later proved toxic to human livers.
In another project led by Oxford University and Janssen Pharmaceutica, computational models of human heart cells detected a dangerous arrhythmia risk with 89 percent accuracy, compared with 75 percent for animal studies.
These studies are resource-intensive. Emulate’s effort used 870 chips and the equivalent of 16 full-time staff over 16 weeks.
Similar validation may be needed for every NAM and for each use case, which poses cost and scale barriers, especially for smaller firms. Advocates argue for industry-academic collaboration and public funding to build the evidence base, along with greater data sharing from preclinical and clinical studies.
The FDA has called for an open-access repository of drug toxicity data. Hartung has suggested running new animal experiments in parallel with NAMs to enable direct comparisons.
Researchers such as Christian Maass of ESQlabs, which builds digital human twins informed by organ-chip data, say these head-to-head studies are essential to demonstrate where NAMs outperform, complement or fall short of animal models.
Changing scientific habits
Even validated methods can take years to displace entrenched practices. A human cell–based monocyte activation test was developed in the 1990s to replace the rabbit pyrogen test, but Europe’s official acceptance came only in 2010, and rabbits still see widespread use.
Cultural inertia plays a role. Reviewers and editors often expect animal data.
Early-career scientists report pressure to add animal experiments for credibility, limited access to NAM infrastructure and career uncertainty when moving away from established models, according to a survey led by veterinary scientist Kathrin Herrmann at Johns Hopkins.
Training is expanding through NIH and FDA resources, international efforts and academic-industry partnerships. Hartung’s online courses have reached thousands of students.
Joseph Wu of Stanford University, who cofounded Greenstone Biosciences to use stem cell–derived tissues and AI in drug discovery, said the company shares its stem-cell lines freely with academic researchers to lower barriers to entry.
Advocates say that with sustained funding, incentives and training, NAMs could become routine in toxicology and drug development, influencing other research domains over time. Ingber points to organ-on-a-chip insights into inflammatory bowel disease, preterm birth and antiviral strategies that animal studies could not reveal.
Hartung describes toxicology’s adoption of NAMs as a lighthouse for the broader biomedical community. “Suddenly, all the dams have opened,” he said.