Organ-on-a-chip: what tiny human tissue models can and cannot do yet

Organ-on-a-chip devices are small plastic chips that hold living human cells in tiny channels, arranged to behave like part of an organ. Drug researchers already use them, and the US regulator, the FDA, names them in an April 2025 plan to phase out animal testing requirements, starting with monoclonal antibodies. If such human-cell models prove reliable, they could flag some harmful drugs earlier and reduce animal use. That is a hypothesis, not a finding. The evidence so far is real but narrow. This article separates what has been shown from what is only reported and what is still speculative.

This article is general information, not medical advice.

What an organ chip is

An organ chip is a clear block, usually moulded from a soft silicone called PDMS. Inside are hollow channels separated by a thin porous membrane. Researchers grow different human cell types on each side of the membrane and pump fluid through the channels, so the cells experience flow in a way that is closer to the body than a flat dish.

What has been demonstrated

A lung chip. In a 2010 study, human air-sac cells and human lung blood vessel cells were grown on either side of a 10 micrometre thick porous PDMS membrane. The membrane was stretched cyclically at 10% strain and 0.2 Hz to mimic breathing (Huh et al.). Activated vessel cells captured flowing human neutrophils, a type of white blood cell, which then passed through the membrane. Stretching increased the transport of 20 nm nanoparticles across the barrier more than four-fold compared with still conditions. In the main paper, sample sizes are stated only for the barrier resistance and albumin data, which come from three separate experiments. For the other results the main text and figure legends do not give a sample size, so their strength is hard to judge from the paper alone.

A liver chip. This is one device, Emulate's Liver-Chip, not organ chips in general. A double-blind study of 870 Liver-Chips exposed to 27 small-molecule drugs, using liver cells from three donors, is the most detailed validation in the sources checked (Ewart et al.). Using two donors and a correction for protein binding, the chip identified drug-induced liver injury with 87% sensitivity (13 of 15 harmful drugs caught) and 100% specificity (3 of 3 safe drugs cleared). The 95% confidence interval for sensitivity was wide, from 61.5% to 96.0%. Under the same analysis, 3D liver spheroids reached 47% sensitivity. Using one donor at a time, chip sensitivity was 77% and 73%. The paper's competing interests statement says most of its authors are current or former Emulate employees who may hold equity, and that one author is a founder and equity holder in the company. The authors say the data were independently verified by two external toxicologists, but the results still come from a team with a stake in the product. For how to read such statements, see our guide to funding and conflict of interest statements.

Linked chips. A robotic platform kept eight vascularised, two-channel organ chips (intestine, liver, kidney, heart, lung, skin, blood-brain barrier and brain) working for three weeks while they were linked by a shared fluid (Novak et al.). A computer model predicted how a tracer dye moved between the chips, and the measured values matched the model over 21 days, in two independent experiments. The model was calibrated using data from blank chips without cells. This is a laboratory demonstration, not proof that such systems predict drug levels in people.

Materials matter. In a separate study of four heart drugs, culture wells with a PDMS base removed more than 80% of the drug bepridil from solution within three hours (van Meer et al.). The authors found no correlation between absorption and log P, a measure of how strongly a drug avoids water, so the effect could not be predicted from that simple property. A lipid-based coating partly reduced the absorption.

What is reported but not yet proven

Regulators. On 10 April 2025 the FDA announced a plan, titled "FDA Announces Plan to Phase Out Animal Testing Requirement for Monoclonal Antibodies and Other Drugs", starting with monoclonal antibodies (FDA). It says the agency will promote lab-grown human organoids and organ-on-a-chip systems that mimic human organs to test drug safety, and that over the coming year it aims to launch a pilot programme letting select monoclonal antibody developers use a mainly non-animal testing strategy. This is a statement of intent. The announcement says broader policy changes and guidance updates are expected to follow in phases.

The manufacturer (reported by the company). In an interview published on its own website in November 2024, Emulate's chief scientific officer says the Liver-Chip S1 has been accepted into the FDA's ISTAND pilot programme at the first, letter-of-intent stage, for predicting drug-induced liver injury from small molecules, and that the company is moving into the second stage, the qualification plan (Emulate). The company says full qualification would mean the chip could be used in regulatory submissions for that stated use. This is the company's own account, which we have not confirmed with the FDA. Acceptance into the pilot is not the same as qualification, and no full qualification is reported in the sources checked.

What is still speculative

Some ideas remain untested. These include chips built from a single person's own cells to help choose their treatment, and multi-organ chips used routinely to estimate how a drug is absorbed and cleared in people before first trials. The sources checked for this article do not show either working.

What the chips cannot replace today

• Whole bodies. Each chip models a small tissue unit. The linked multi-organ work was a short laboratory study, not a route to whole-body predictions.
• Large molecules. The Liver-Chip validation study, and the company's ISTAND proposal, cover small-molecule drugs (Emulate). Monoclonal antibodies are large biological molecules. Emulate says researchers have used Liver-Chips to assess antibody safety, but the sources checked contain no validation data for them, so the FDA's planned pilot covers an area these sources do not.
• Every organ and every drug. The strongest result covers one organ and one class of drug.

Limits and open problems

• Validation against human outcomes. The liver study checked chip results against known liver effects for a fixed set of 27 drugs. A wide confidence interval means the true accuracy on other drugs is uncertain.
• Donor variation. Results depend on which donor's cells are used. Single-donor sensitivity was 77% and 73%, against 87% with two donors (Ewart et al.).
• Reproducibility. Emulate says its planned qualification studies aim to show consistent results across laboratories and hepatocyte donors (Emulate). That evidence is not public in the sources checked.
• Material effects. PDMS can absorb some drugs, which can lower the dose the cells actually receive (van Meer et al.).
• Cost and throughput. The sources checked do not give per-chip cost or testing-speed figures, so this article does not quantify them.
• Regulatory acceptance. Full FDA qualification of an organ chip for a defined use had not been reported in the sources checked.

What to watch next

• The FDA pilot for monoclonal antibody developers, which the agency said in April 2025 it aimed to launch over the following year (FDA).
• Qualification plan data showing consistent results across laboratories and donors.
• Any organ chip reaching full FDA qualification for a stated use.
• Validation studies that cover biologics, not only small molecules.