Organ-on-Chip in Drug Development: A Human-Relevant Alternative to Animal Testing

Animal models have supported biomedical research and drug development for decades, but differences between species can limit how well preclinical findings predict human efficacy, metabolism, immune responses, and toxicity. Organ-on-chip (OOC) technology adds a complementary approach by combining human cells with controlled microfluidic environments that reproduce selected tissue functions.[1]

The objective is not to replace every animal study with one universal chip. It is to use human-relevant models where they can answer a defined scientific question more directly, improve candidate selection, or reduce uncertainty before clinical trials.

Differences between organoids, organ-on-chip and microphysiological systems

Organoids, organ-on-chip models, and microphysiological systems (MPS) are related approaches to advanced in vitro modelling, but the terminology is not completely uniform across the field. Organoids are self-organizing 3D cellular structures, while OOCs use engineered devices to control variables such as perfusion, interfaces, gradients, and mechanical forces.[1]

As shown in Figure 1, these model classes differ in how biological self-organization is combined with engineering control. Organoids rely strongly on intrinsic tissue organization, whereas organ-on-chip models introduce controlled physical variables such as perfusion, interfaces, and mechanical stimulation.

differences organoids ooac mps — *Figure 1 Differences between organoids organ on chip OOC and microphysiological systems MPS*

The most appropriate model depends on the scientific question. Organoids are useful when tissue self-organization or patient-derived biology is central. Patient-derived tumor organoids have reproduced clinically relevant treatment responses in metastatic gastrointestinal cancers, supporting their use as translational drug-testing models in that setting.[2] OOCs are especially useful when flow, shear stress, tissue interfaces, or repeated sampling must be controlled.

The categories can also overlap: organoids and spheroids can be integrated into perfused organ-on-chip platforms, and multi-organ MPS can connect several tissue models.[3] Model selection should therefore focus on the biological functions, engineering inputs, and readouts required by the experiment.

For practical model selection, the main experimental differences can be summarized as follows:

Feature	Spheroids	Tumoroids	Organoids	Organ-on-Chip
3D Structure	Simple sphere	Patient-derived structure	Organ-like	Engineered microenvironment
Cell Source	Cell lines / mixed	Primary tumor	Stem cells	Primary or iPSC-derived
Architecture	Minimal	Tumor-specific	Tissue-specific	Device-defined
Perfusion	❌ None	❌ None	❌ None (unless hybrid)	✅ Controlled
Mechanical Forces	❌ None	❌ None	❌ None	✅ Physiological
Drug Testing	Medium-throughput	Patient-specific	Disease-specific	Mechanistic PK/PD
Shear Stress	❌ None	❌ None	❌ Minimal	Fully tunable
Personalization	Low	Very high	High	Medium–high
Reproducibility	Good	Medium	Medium–high	High

Model choice should therefore follow the mechanism being studied rather than a simple hierarchy of sophistication. A more complex system is useful only when the added biological or physical features improve the relevance of the experiment. The table describes common tendencies; individual models can differ substantially in cell source, perfusion, throughput, and reproducibility.[1][3]

Limitations of Animal Models in Drug Development and Testing

Animal studies provide valuable whole-organism information, yet a translation gap remains between preclinical results and human outcomes. Differences in pharmacology, metabolism, physiology, and species-specific toxicity can limit how reliably animal findings predict human safety.[4]

Why can a promising animal result fail in humans? One reason is that the same drug can be absorbed, metabolized, transported, or eliminated differently across species. Another is that the molecular target or disease mechanism may not behave identically in the animal model and in human biology.

Fialuridine (FIAU) illustrates this limitation. In a clinical trial for chronic hepatitis B, FIAU caused severe and unexpected multisystem toxicity, including hepatic failure and lactic acidosis.[5] A subsequent Institute of Medicine review concluded that the preclinical animal studies available before the trial provided no evidence from which the human hepatic or pancreatic toxicity could have been anticipated.[6]

Human cell-based OOCs cannot reproduce every systemic process, but they can isolate specific human mechanisms under controlled conditions. Perfusion, tissue interfaces, mechanical stimulation, and repeated sampling can make them useful for investigating liabilities that are difficult to study in static cell culture.

The Opposite Scenario: False Negatives

Species-specific adverse responses can also mislead candidate evaluation. Guinea pigs are unusually susceptible to penicillin administration. Classic work reported rapid mortality after penicillin exposure and discussed antibiotic-induced disruption of intestinal flora and enterocolitis, rather than a simple direct toxic effect, as a likely explanation.[7] Such differences show why a safety signal in one species does not automatically represent the same risk in people.

Human-relevant models can therefore complement animal data by helping researchers distinguish species-specific effects from mechanisms more likely to matter clinically. Their value is strongest when the biological question, cell source, exposure profile, and readout are clearly defined.

Modeling Human Disease Mechanisms

Animal models also shape disease research and target discovery. This can be problematic when the experimental disease represents only a narrow part of the heterogeneous biology seen in patients.

Alzheimer’s disease is a useful example. Mouse models have contributed substantially to mechanistic research, but important limitations remain in how well they reproduce the multifactorial biology and clinical heterogeneity of human disease.[8]

Human cerebral organoids and related advanced in vitro approaches provide additional experimental perspectives for Alzheimer’s research because they can reproduce selected features of human brain tissue architecture and cellular diversity.[9] OOCs and other human-cell models can complement these systems by testing defined mechanisms under controlled microenvironmental conditions.

Why Organ-on-a-Chip Technology Offers Improved Human Relevance for Drug Testing

Organ-on-chip technology combines human cells with engineered microenvironments that conventional 2D culture often cannot reproduce.[10] Depending on the design, a chip can incorporate organized co-culture, continuous perfusion, tissue-tissue interfaces, biochemical gradients, and controlled mechanical forces.[1][10]

These variables matter because cells respond to their physical environment. Shear stress can alter endothelial biology, cyclic strain affects lung and cardiac tissues, and concentration gradients influence transport and drug exposure. Microfluidic control makes these conditions measurable and repeatable rather than incidental.

OOCs add the most value when the scientific question depends on human-specific biology or on physiological conditions that are poorly represented by static culture. The model should be selected for the mechanism under study, not simply because it is more complex.

Common design variables that can improve physiological relevance include:

continuous microfluidic perfusion for nutrient delivery and drug exposure;

human tissue-tissue interfaces, such as epithelium and endothelium;

controlled oxygen, cytokine, or drug gradients;

mechanical cues such as shear stress, stretch, or compression;

repeated sampling, imaging, and functional measurements during culture.

These features form a continuum rather than a simple-versus-advanced divide. Figure 2 shows how increasing complexity can combine 3D culture with perfusion, mechanical cues, and multi-tissue organization. Added complexity is useful only when those features answer the biological question.

complexity hierarchy of ooac — *Figure 2 Complexity hierarchy of in vitro and microphysiological systems*

Cancer research is one example. Tumor-on-chip and cancer-on-chip models can incorporate vascular, stromal, and immune components to study tumor-immune interactions and responses to immunotherapies.[11] These applications are particularly relevant when the microenvironment and patient-specific biology influence treatment response.

A cancer-on-chip review provides additional examples of drug-screening applications. A breathing lung-on-chip case study also illustrates how controlled mechanical cues and perfusion can be combined in a dynamic tissue model.

Reproducible fluid handling is central to these experiments. Stable pressure and flow help researchers control exposure, nutrient delivery, waste removal, and shear stress across long cell-culture protocols. Flow control is therefore a core engineering variable in many organ-on-chip designs.[1]

Can Organ-on-a-Chip Replace Animal Testing?

Organ-on-chip models can already replace or reduce some animal experiments, but they are not a universal substitute for all animal testing in drug development. Their strongest role is context-specific: a model is most useful when its performance has been demonstrated for a defined question or context of use.[12]

A liver model may investigate human metabolism or hepatotoxicity. A vascular model may address barrier transport. A lung chip may reproduce breathing-related mechanical strain. Linked gut, liver, and kidney models can investigate selected aspects of drug disposition that depend on interactions between organs.

One chip cannot currently reproduce a whole human organism. Long-term endocrine regulation, systemic immunity, behavior, developmental processes, and many multi-organ responses remain difficult to reproduce comprehensively in vitro. Animal models, clinical data, computational methods, and human-based assays may therefore remain complementary for many applications.

The practical transition is progressive. OOCs, organoids, in silico approaches, and other New Approach Methodologies (NAMs) can replace particular animal tests when their context of use is defined and performance is validated, while complementing animal studies elsewhere.[12] In the United States, Section 3209 of the Consolidated Appropriations Act, 2023—often referred to as FDA Modernization Act 2.0—explicitly added organ chips and microphysiological systems to examples of acceptable nonclinical tests.[13]

Organ-on-Chip in the Drug Development and Testing Pipeline

Drug development progressively filters candidates from target identification through lead optimization, nonclinical safety testing, clinical trials, and approval. OOCs can add human-relevant mechanistic, efficacy, pharmacokinetic, or safety data at several of these decision points.[10][12]

The best stage for introducing an OOC depends on the biological mechanism, the development decision, and whether the model has sufficient reproducibility and validation for that use. There is no single stage that is optimal for every application.

Early Discovery: Disease Modelling and Target Identification

During discovery, researchers use models to understand disease mechanisms and test whether a therapeutic target remains relevant in a human cellular environment. OOCs can reproduce selected cues such as inflammation, hypoxia, barrier dysfunction, flow, mechanical stress, or interactions between cell populations.

Patient-derived primary cells and induced pluripotent stem cell (iPSC)-derived tissues can add biological diversity. In cancer research, tumor-on-chip models can study invasion, drug penetration, and resistance. Gut-on-chip models can address epithelial barrier biology, inflammatory diseases, and selected host-microbiome interactions.[10]

Disease modelling often requires several complementary systems. In Alzheimer’s research, animal models, 2D cell culture, organoids, microfluidic approaches, and computational methods capture different aspects of disease biology.[8][9] Figure 3 summarizes these model types.

Animal Model systems Alzheimers disease — *Figure 3 Models used to investigate Alzheimers disease and their complementary experimental roles9*

At this stage, the aim is usually mechanistic rather than regulatory. Better human-relevant evidence can help researchers prioritize targets and avoid carrying weak candidates into more expensive development studies.

Lead Optimization: In Vitro ADME and Pharmacokinetics

Lead optimization requires information about absorption, distribution, metabolism, and excretion (ADME). OOCs can contribute by introducing controlled drug exposure and, in multi-organ configurations, connecting tissues involved in drug disposition.

Gut, liver, and kidney models are especially relevant. Defined microfluidic flow allows concentration and exposure time to be controlled while samples are collected for metabolite or biomarker analysis. Integrated gut-liver MPS have been used to study intestinal permeability and hepatic metabolism within a connected in vitro pharmacokinetic system.[14]

A liver-kidney organ-on-chip application note demonstrates how a microfluidic loop can support research on inter-organ drug disposition. These systems complement conventional ADME assays; they should not be presented as complete replicas of human pharmacokinetics unless that performance has been demonstrated for the intended application.

For reliable ADME studies, flow rates, tissue volumes, residence times, sampling, and analytical methods must be selected according to the pharmacological question. The value of a complex model depends on its experimental design and data quality, not on complexity alone.

Preclinical Safety Assessment: Toxicology and Off-Target Effects

Toxicology is one of the most developed applications for human OOCs. Liver, kidney, cardiac, lung, and other organ-specific models can be used to investigate tissue injury under controlled exposure conditions. In one blinded performance assessment, a human Liver-Chip classified a benchmark set of 27 drugs with 87% sensitivity and 100% specificity for drug-induced liver injury in that dataset.[15]

A metabolite generated in a liver model can, for example, be introduced into another tissue to study secondary toxicity. Multi-organ systems can also combine imaging, biosensors, and repeated sampling to follow responses over time rather than relying on a single endpoint.

The regulatory significance is increasing. On September 24, 2024, FDA accepted the first organ-on-chip letter of intent into its ISTAND Pilot Program.[16]

The proposed human Liver-Chip drug development tool is intended to assess relative drug-induced liver injury risk for a defined context of use. Acceptance is the first step in qualification, not regulatory qualification itself.

This distinction matters. What does regulatory readiness require? A model must be reproducible, fit for its stated purpose, supported by appropriate reference compounds or datasets, and validated for the decision it is intended to inform.

For a drug-development application, useful validation evidence can include:

repeatability within a laboratory and reproducibility across operators or sites;

defined acceptance criteria, controls, and performance limits;

comparison with human clinical data or well-characterized reference compounds;

transparent documentation of cell sources, culture conditions, microfluidic design, and analytical methods.

Translational PK/PD and Precision Medicine

OOCs also have potential in translational pharmacokinetics/pharmacodynamics (PK/PD) and precision medicine. Patient-derived cells, biopsies, or iPSC-derived tissues can be used to investigate how biological variation influences efficacy, toxicity, and functional responses.[10]

Controlled perfusion links a defined exposure profile to measurable pharmacodynamic endpoints. In oncology, ex vivo patient-derived models may help compare therapeutic responses. In cardiac or metabolic research, live imaging and functional measurements can connect drug concentration with tissue behavior.

These applications remain an active field of research. A patient-specific chip should not be treated as a clinical decision tool unless it has been validated for that purpose. Its immediate value is experimental: connecting human biology, controlled exposure, and functional data within a reproducible laboratory workflow.

Regulatory and Policy Landscape for Organ-on-Chip Adoption in Drug Development

Regulatory policy increasingly supports human-relevant non-animal methods, but acceptance remains context-specific. Regulators do not treat every organ-on-chip model as an automatic replacement for animal studies. Scientific validity, reproducibility, and relevance to a defined decision remain central.

United States: Policy Momentum Toward Non-Animal Testing

Section 3209 of the Consolidated Appropriations Act, 2023 amended the Federal Food, Drug, and Cosmetic Act to define “nonclinical test” broadly and to list cell-based assays, organ chips and microphysiological systems, computer modelling, other human biology-based methods, and animal tests as examples.[13]

FDA’s ISTAND program provides a pathway for novel drug development tools that do not fit existing qualification programs. In 2024, the agency accepted a letter of intent for a human Liver-Chip intended to assess relative drug-induced liver injury risk for certain drug candidates.[16]

In April 2025, FDA announced a roadmap to reduce, refine, and potentially replace animal testing for monoclonal antibodies and other drugs. The agency encouraged greater use of NAM data, including human-based cell and organoid models, alongside computational approaches.[17]

This policy shift does not mean animal testing has ended for drug development. Alternatives are encouraged where scientifically justified, but sponsors still need evidence appropriate to the product, risk, and regulatory question. Early dialogue with regulators is important when a novel OOC is intended to support a submission.

European Union: Accelerating the Shift to Non-Animal Approaches

European medicines policy has long incorporated the principles of Replacement, Reduction, and Refinement, or the 3Rs. EMA guidance and its 2023 revision concept paper explicitly aim to encourage the development and regulatory acceptance of 3Rs testing approaches for human and veterinary medicines.[18]

The European Parliament’s 2021 resolution called for coordinated action to accelerate the transition toward non-animal innovation while recognizing that suitable alternatives are not yet available for every scientific question.[19] In June 2026, the European Commission adopted a roadmap for phasing out animal testing in chemical safety assessments that includes pharmaceuticals among the covered legislative domains.[20]

For novel drug-development methods, EMA’s qualification pathway can provide qualification advice during development or a CHMP qualification opinion for a defined use when sufficient evidence is available.[21] EFPIA has similarly emphasized robust validation, global harmonization, and streamlined regulatory acceptance of non-animal methods.[22]

For OOC developers, the regulatory message is consistent: define the context of use, standardize the method, document variability, and generate comparative data. Regulatory adoption depends on evidence that the model solves a specific problem reliably.

Future outlook

The future of organ-on-chip research is likely to combine biological relevance with greater standardization and automation. Multi-organ models can connect several tissue types to study inter-organ processes, but increasing complexity also increases the need for robust control, characterization, and validation.[12]

Across biomedical research, this technology also has potential to support new models of complex diseases, precision medicine, and preparation for clinical trials when human cells and mechanistic readouts are needed. Lab-to-lab validation remains essential.

Automation can improve reproducibility by stabilizing perfusion, dosing, sampling, and long-term cell culture. For laboratories moving beyond proof-of-concept studies, an automated organ-on-chip platform can reduce operator-dependent variability and make experiments easier to compare across researchers and sites.

A cell perfusion package and higher-throughput organ-on-chip perfusion configuration illustrate how controlled flow can be implemented for repeated drug-testing workflows. The Flow EZ is another option for reproducible pressure-driven perfusion.

As these systems mature, their value will depend on where they improve development decisions. Figure 4 summarizes the progression from discovery to clinical development and approval. OOCs can contribute human-relevant mechanistic, ADME, or safety data before a candidate advances.

drug testing pipeline in ooac — Figure 4 Drug testing pipeline

What should researchers prioritize next? More complex models are not automatically better. The strongest future applications will combine a biologically appropriate model with robust engineering, transparent data analysis, inter-laboratory reproducibility, and validation against a clinically meaningful reference.

Regulatory progress will also be decisive. The FDA Modernization Act 2.0, ISTAND, the 2025 FDA roadmap, and parallel European initiatives create clearer routes for non-animal evidence, but qualification will continue to depend on defined contexts of use.

The most realistic outcome is a progressive shift: individual animal experiments are replaced, reduced, or refined when a validated human-relevant method answers the same question more effectively. For industry and academia, the practical strategy is to integrate OOCs where they add measurable scientific value and generate the evidence required for broader adoption.

Frequently Asked Questions

Can organ-on-a-chip replace animal testing?

Organ-on-chip technology can replace or reduce some animal experiments when a validated model addresses a defined biological, pharmacological, or toxicological question. It does not currently replace every animal study across drug development, and the appropriate evidence depends on the context of use.

What does the FDA say about organ-on-a-chip?

FDA recognizes microphysiological systems, including organ-on-chip approaches, as promising NAMs. The agency accepted a human Liver-Chip letter of intent into ISTAND in 2024.[16] It announced a broader roadmap in 2025 to reduce, refine, and potentially replace animal testing where scientifically justified.[17]

What is organ-on-a-chip toxicity testing?

Organ-on-chip toxicity testing uses human cells in engineered microenvironments to investigate how a drug affects a tissue or combination of tissues. Controlled perfusion, mechanical cues, repeated sampling, imaging, and biomarkers can help researchers study organ-specific or metabolite-mediated toxicity under defined conditions.

How does organ-on-chip compare with animal models for ADME?

Animal models provide whole-organism information but can differ from humans in metabolism, transport, and clearance. OOCs use human cells and can connect tissues under controlled flow, providing a complementary way to study selected human ADME mechanisms and inter-organ interactions.[14]

References

[1] C. M. Leung, P. de Haan, K. Ronaldson-Bouchard, et al., “A guide to the organ-on-a-chip,” Nature Reviews Methods Primers, vol. 2, article 33, 2022. DOI: 10.1038/s43586-022-00118-6.

[2] G. Vlachogiannis, S. Hedayat, A. Vatsiou, et al., “Patient-derived organoids model treatment response of metastatic gastrointestinal cancers,” Science, vol. 359, no. 6378, pp. 920–926, 2018. DOI: 10.1126/science.aao2774.

[3] V. Carvalho, M. Bañobre-López, G. Minas, S. F. C. F. Teixeira, R. Lima, and R. O. Rodrigues, “The integration of spheroids and organoids into organ-on-a-chip platforms for tumour research: A review,” Bioprinting, vol. 27, e00224, 2022. DOI: 10.1016/j.bprint.2022.e00224.

[4] G. A. Van Norman, “Limitations of Animal Studies for Predicting Toxicity in Clinical Trials: Is it Time to Rethink Our Current Approach?,” JACC: Basic to Translational Science, vol. 4, no. 7, pp. 845–854, 2019. DOI: 10.1016/j.jacbts.2019.10.008.

[5] R. McKenzie, M. W. Fried, R. Sallie, et al., “Hepatic failure and lactic acidosis due to fialuridine (FIAU), an investigational nucleoside analogue for chronic hepatitis B,” New England Journal of Medicine, vol. 333, no. 17, pp. 1099–1105, 1995. DOI: 10.1056/NEJM199510263331702.

[6] Institute of Medicine (US), Committee to Review the Fialuridine (FIAU/FIAC) Clinical Trials, Review of the Fialuridine (FIAU) Clinical Trials, F. J. Manning and M. Swartz, eds. Washington, DC: National Academies Press, 1995. DOI: 10.17226/4887.

[7] S. B. Formal, G. D. Abrams, H. Schneider, and R. Laundy, “Penicillin in Germ-free Guinea Pigs,” Nature, vol. 198, p. 712, 1963. DOI: 10.1038/198712a0.

[8] A. Granzotto, B. Vissel, and S. L. Sensi, “Lost in translation: Inconvenient truths on the utility of mouse models in Alzheimer’s disease research,” eLife, vol. 13, e90633, 2024. DOI: 10.7554/eLife.90633.

[9] S. Sreenivasamurthy, M. Laul, N. Zhao, T. Kim, and D. Zhu, “Current progress of cerebral organoids for modeling Alzheimer’s disease origins and mechanisms,” Bioengineering & Translational Medicine, vol. 8, no. 2, e10378, 2023. DOI: 10.1002/btm2.10378.

[10] D. E. Ingber, “Human organs-on-chips for disease modelling, drug development and personalized medicine,” Nature Reviews Genetics, vol. 23, no. 8, pp. 467–491, 2022. DOI: 10.1038/s41576-022-00466-9.

[11] Y. Peng and E. Lee, “Microphysiological Systems for Cancer Immunotherapy Research and Development,” Advanced Biology, vol. 8, no. 8, e2300077, 2024. DOI: 10.1002/adbi.202300077.

[12] L. A. Low, C. Mummery, B. R. Berridge, C. P. Austin, and D. A. Tagle, “Organs-on-chips: into the next decade,” Nature Reviews Drug Discovery, vol. 20, no. 5, pp. 345–361, 2021. DOI: 10.1038/s41573-020-0079-3.

[13] U.S. Congress, Consolidated Appropriations Act, 2023, Pub. L. No. 117-328, div. FF, § 3209, 136 Stat. 4459, 5821 (2022).

[14] N. Tsamandouras, W. L. K. Chen, C. D. Edington, C. L. Stokes, L. G. Griffith, and M. Cirit, “Integrated Gut and Liver Microphysiological Systems for Quantitative In Vitro Pharmacokinetic Studies,” The AAPS Journal, vol. 19, no. 5, pp. 1499–1512, 2017. DOI: 10.1208/s12248-017-0122-4.

[15] L. Ewart, A. Apostolou, S. A. Briggs, et al., “Performance assessment and economic analysis of a human Liver-Chip for predictive toxicology,” Communications Medicine, vol. 2, article 154, 2022. DOI: 10.1038/s43856-022-00209-1. Author corrections published in 2023; the cited HTML/PDF record incorporates the corrections.

[16] U.S. Food and Drug Administration, “FDA’s ISTAND Pilot Program accepts a submission of first organ-on-a-chip technology designed to predict human drug-induced liver injury (DILI),” September 24, 2024.

[17] U.S. Food and Drug Administration, “FDA Announces Plan to Phase Out Animal Testing Requirement for Monoclonal Antibodies and Other Drugs,” April 10, 2025.

[18] European Medicines Agency, “Concept paper on the revision of the guideline on the principles of regulatory acceptance of 3Rs (replacement, reduction, refinement) testing approaches,” EMA/CHMP/CVMP/452614/2023, 2023.

[19] European Parliament, “Resolution on plans and actions to accelerate the transition to innovation without the use of animals in research, regulatory testing and education,” P9_TA(2021)0387, September 16, 2021.

[20] European Commission, Communication from the Commission – Roadmap towards phasing out animal testing for chemical safety assessments, C(2026) 3497 final, June 1, 2026; published in the Official Journal of the European Union as C/2026/3087, June 3, 2026.

[21] European Medicines Agency, Qualification of novel methodologies for medicinal product development: guidance to applicants, EMEA/CHMP/SAWP/72894/2008 Rev. 6, updated July 7, 2026.

[22] European Federation of Pharmaceutical Industries and Associations (EFPIA), EFPIA Recommendations on Phasing Out Animal Testing for Chemical Safety Assessments, 2025.

For more information or a technical discussion