In December 2022, the journal Communications Medicine published a large test of a human Liver-Chip, a small device lined with living human liver cells. The team ran 870 chips against 27 blinded drugs with known effects on the liver and reported that the chip caught 87% of the harmful drugs while flagging none of the safe ones as harmful [1]. The same month, the United States changed the law so that drug developers could lean on tests like this one.
This often gets framed as a contest: can chips replace animal testing? I think that is the wrong question. The better one is: validated for what task? A model earns trust one defined job at a time, and this paper shows what that looks like.
What the Liver-Chip was actually asked to do
The task was narrow on purpose: predict drug-induced liver injury (DILI), using benchmark drugs and qualification guidelines from the Innovation and Quality (IQ) consortium, a group of pharmaceutical and biotech companies [1]. The authors framed their work as testing the chip “within the context of use of DILI prediction” [1]. That phrase is the whole point.
They also made a sharp point: too often, they wrote, the features that make a lab model look realistic are “chosen post hoc by the authors and not prospectively by an independent third party” [1]. An outside benchmark answers that.
Sensibly, the teams preparing, dosing and analyzing were kept separate, so those dosing the chips or analyzing results did not know which drug they were handling [1]. Eight of the test drugs were chosen because they had harmed patients’ livers even after passing standard animal-based safety testing [1].
Imagine a smoke detector advertised as better than a fire inspector. It may well catch more kitchen fires. It still cannot tell you whether the wiring in the walls is safe, because nobody tested it for that job.
That is how I read a sensitivity number: it belongs to the task tested, here liver injury from known small-molecule drugs.
The numbers, read carefully
In the first analysis, the chip caught 12 of 15 toxic drugs, a sensitivity of 80%, compared with 42% for 3D spheroids, which are small balls of liver cells grown in a dish, based on earlier published data [1]. The chip wrongly flagged no safe drug, while the spheroids had a specificity of 67% [1].
The headline 87% came from a second analysis that accounted for how much of each drug binds to proteins, in the chip’s culture medium and in patients’ blood. The authors set the cutoff at a value they “selected to maximize sensitivity while avoiding false positives,” and used cells from two donors [1]. That is reasonable, but the cutoff was picked on the same drugs it was scored on. I would want to see it hold up on fresh drugs before treating 87% as the chip’s true rate.
The misses matter too. The chip did not detect liver damage from zileuton, an asthma drug [1]. One benchmark drug, pemoline, was left out because its liver damage is thought to involve the immune system, which would need a more complex chip [1].
Chips, organoids and animals are different tools
These three often get lumped together. An organ-on-a-chip is a small device with tiny channels where cells grow under constant fluid flow [2]. The Liver-Chip has two channels separated by a porous membrane, with liver cells on one side and cells that line blood vessels, along with other liver cell types, on the other [1].
Organoids are different. They grow from stem cells that organize themselves into small tissue-like structures, and they are usually grown in static culture [2]. A 2022 review put the trade-off well: chips offer control over the cells’ environment “but are commonly simplistic models of the target organ,” while organoids are closer to how tissue forms but vary more [2].
A whole animal is a third thing entirely: many organs, a circulation, and an immune system, all working at once. Each is a model. None is a human patient.
Who ran the study, and what the corrections changed
The paper is open about this. Of the 32 authors, 28 are declared as employees or former employees of Emulate, the company that makes the chip, who may hold equity [1]. Donald Ingber of Harvard’s Wyss Institute is declared as a founder, board member, scientific advisory board chair and equity holder in Emulate, and Jack Scannell, a shareholder and director of JW Scannell Analytics, received payment from Emulate for contributing to the work [1]. Two authors, from Janssen and AbbVie, declared no competing interests, and the authors say two external toxicologists independently verified the data [1].
I mention this as context, not accusation. Companies should test their own products. But a vendor-led validation is a first step; independent labs repeating it on new drugs would be the next.
Two author corrections followed in early 2023. The first fixed a table where the spheroid specificity read 100% instead of 67%, plus an abstract typo [3]. The second replaced a table that had duplicated another table’s data [4]. As the notices describe them, neither touched the chip’s 87% and 100% figures reported in the abstract and text, and the first simply brought the table in line with the 67% already in the text [1, 3].
What animals still capture, and why the chip still matters
The strongest case for animals is the whole body: a drug can be broken down in one organ and harm another. The paper’s own authors suggest the drugs their chip missed may involve “other cells or tissues not present in these models” [1]. They only tested small molecules, and they call for testing larger biological drugs next [1]. A review from the same year lists better simulation of whole-body physiology, including multi-organ interactions, among the challenges organoids still face, and sees combining them with chips as one route forward [2].
My earlier work included PET/CT pharmacokinetic modeling, which is about how a substance moves through a whole living body over time. From that angle, a single-organ chip answers a sharper but smaller question than an animal does.
Still, the result is genuinely promising. Every toxic drug in the test had already been evaluated in animals and judged to have an acceptable safety margin for human trials [1]. The chip flagged most of them. Advocates such as Ingber, an Emulate founder, have called the failure of animal models to predict human responses a major problem and written that using chips instead of animals is “ever closer to realization” [5]. The Liver-Chip paper itself proposes something more modest: placing the chip between early cell tests and dose-range animal studies, where it could screen out risky compounds and reduce animal use [1].
The law points the same way. The FDA Modernization Act 2.0 passed the Senate in September 2022 to allow alternatives such as cell-based assays and computer models. Its changes became law on December 29, 2022, in a large spending act that swapped “preclinical tests (including tests on animals)” for “nonclinical tests.” That term explicitly includes organ chips, computer modeling and, still, animal tests.
Validated for what?
For some decisions, chips may replace animal tests once each is validated for that decision. For others, not yet. Now that the law allows the choice, what each model has been shown to predict matters more, not less.
Replacing animal testing is a slogan. Validated for liver injury from small-molecule drugs, with these known misses, is evidence, and that is the standard I would hold every model to, chip or animal.
When you read about a new lab model, look for the exact task, the benchmark set, and who chose the cutoff. The Liver-Chip study answered all three openly, which is why it deserves attention. The next step is independent groups testing it on drugs it has never seen.
References
[1] L. Ewart, A. Apostolou, S. A. Briggs, C. V. Carman, J. T. Chaff, A. R. Heng, et al., “Performance assessment and economic analysis of a human Liver-Chip for predictive toxicology,” Communications Medicine, vol. 2, no. 1, Art. no. 154, Dec. 2022, doi: 10.1038/s43856-022-00209-1.
[2] L. S. Baptista, C. Porrini, G. S. Kronemberger, D. J. Kelly, and C. M. Perrault, “3D organ-on-a-chip: The convergence of microphysiological systems and organoids,” Frontiers in Cell and Developmental Biology, vol. 10, Art. no. 1043117, Nov. 2022, doi: 10.3389/fcell.2022.1043117.
[3] L. Ewart, A. Apostolou, S. A. Briggs, C. V. Carman, J. T. Chaff, A. R. Heng, et al., “Author Correction: Performance assessment and economic analysis of a human Liver-Chip for predictive toxicology,” Communications Medicine, vol. 3, no. 1, Art. no. 7, Jan. 2023, doi: 10.1038/s43856-023-00235-7.
[4] L. Ewart, A. Apostolou, S. A. Briggs, C. V. Carman, J. T. Chaff, A. R. Heng, et al., “Author Correction: Performance assessment and economic analysis of a human Liver-Chip for predictive toxicology,” Communications Medicine, vol. 3, no. 1, Art. no. 16, Feb. 2023, doi: 10.1038/s43856-023-00249-1.
[5] 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, Aug. 2022, doi: 10.1038/s41576-022-00466-9.
Related Articles
