Student or educator? Sign up with your school email and get 6 months of Pro free.
CoChat

The DOI matches its Crossref record; 58 of the 58 references it cites were found in a registry or an index; 13 of the 44 sources its claims cite were checked for support.

Performance assessment and economic analysis of a human Liver-Chip for predictive toxicology

Open access

Abstract

Conventional preclinical models often miss drug toxicities, meaning the harm these drugs pose to humans is only realized in clinical trials or when they make it to market. This has caused the pharmaceutical industry to waste considerable time and resources developing drugs destined to fail. Organ-on-a-Chip technology has the potential improve success in drug development pipelines, as it can recapitulate organ-level pathophysiology and clinical responses; however, systematic and quantitative evaluations of Organ-Chips' predictive value have not yet been reported. 870 Liver-Chips were analyzed to determine their ability to predict drug-induced liver injury caused by small molecules identified as benchmarks by the Innovation and Quality consortium, who has published guidelines defining criteria for qualifying preclinical models. An economic analysis was also performed to measure the value Liver-Chips could offer if they were broadly adopted in supporting toxicity-related decisions as part of preclinical development workflows. Here, we show that the Liver-Chip met the qualification guidelines across a blinded set of 27 known hepatotoxic and non-toxic drugs with a sensitivity of 87% and a specificity of 100%. We also show that this level of performance could generate over $3 billion annually for the pharmaceutical industry through increased small-molecule R&D productivity. The results of this study show how incorporating predictive Organ-Chips into drug development workflows could…

Overview

CoChat Summary

## TL;DR

This study systematically evaluated a human Liver-Chip for drug-induced liver injury (DILI) prediction using 870 chips and a blinded set of 27 drugs. The Liver-Chip met IQ consortium qualification criteria and, when protein binding was accounted for, achieved 87% sensitivity and 100% specificity, with the authors estimating an annual industry value of over $3 billion.

## Key findings

  • Across a blinded set of 27 known hepatotoxic and non-toxic drugs, the Liver-Chip achieved 87% sensitivity and 100% specificity after accounting for protein binding.
  • In the 15 toxic drugs tested using two donors, the Liver-Chip correctly predicted toxicity in 12 out of 15 drugs, yielding 80% sensitivity; the paper reports this was compared with 42% sensitivity for 3D hepatic spheroids and 67% specificity for spheroids.
  • Using the protein-binding-corrected analysis, the Spearman correlation with the Garside DILI severity scale was 0.78, versus 0.43 with the lower-threshold, uncorrected approach.
  • The economic model estimated a 2.8% NPV uplift for drug development and an annual industry gain of about $3 billion ($2.1B–$3.4B, 95% CI) if Liver-Chips were broadly adopted for small-molecule DILI prediction; extending the approach to four additional Organ-Chip models could yield over $24 billion annually.

## Methods

  • The study used 870 Liver-Chips run in five consecutive cycles over 20 weeks, testing 27 drugs in a blinded design with three human hepatocyte donors; drug preparation, dosing, and analysis were split across teams to preserve blinding.
  • Liver-Chip performance was assessed using structural and functional endpoints including albumin, ALT, urea, morphology scoring, live/fixed imaging, and gene expression, with toxicity called from concentration- and time-dependent responses across days 1, 3, and 7.
  • An economic model extending prior pipeline models was used to estimate how improved preclinical toxicology prediction would change false negative rate (FNR), net present value (NPV), and annual R&D productivity.

## Limitations

  • The authors note that the PDMS chip material may be problematic for some small molecules because of non-specific binding, and they recommend checking binding experimentally with an acellular chip; this is a limitation stated in the Discussion.
  • The paper notes that only small molecules were assessed, and that performance against large molecules and biologic therapies still needs investigation; this is stated in the Discussion.
  • Evident from the text: the study only tested five of six additional IQ-suggested drugs in the second DILI challenge set because one drug was not obtained and another was considered to need a more complex immune configuration.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1038/s43856-022-00209-1

Cites

58 of 58 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Figures

Fig. 1 Schematic of the Emulate Liver-Chip. This diagram shows primary human hepatocytes (C) that are sandwiched within an extracellular matrix (B) on a porous membrane (D) within the upper parenchymal channel (A), while human liver sinusoidal endothelial cells (G), Kupffer cells (F), and stellate cells (E) are cultured on the opposite side of the membrane in the lower vascular channel (H).
Fig. 1 Schematic of the Emulate Liver-Chip. This diagram shows primary human hepatocytes (C) that are sandwiched within an extracellular matrix (B) on a porous membrane (D) within the upper parenchymal channel (A), while human liver sinusoidal endothelial cells (G), Kupffer cells (F), and stellate cells (E) are cultured on the opposite side of the membrane in the lower vascular channel (H).CC BY · PubMed Central
Fig. 2 Economic value model for assessing the financial impact of improved preclinical testing. Illustrated is the model’s “base case”, which tracks a representative portfolio of candidate drugs as it progresses and erodes through clinical trials, culminating in a single drug approval. The model bases phase-by-phase attrition rates (“attrition during phase”), discovery and preclinical costs, development costs (“cost per candidate”) and cost of capital on Paul et al.5 to compute a portfolio-wide discounted cashflow. In contrast with prior approaches, the model tracks the underlying causes of clinical trial failure (safety-related, efficacy-related, and other failures) using parameters derived from literature7,9,24,25, a feature that permits us to determine the composition of the drug portfolio in each stage of development in terms of candidates that are safe and effective, safe and ineffective, unsafe and effective, and unsafe and ineffective, as illustrated. Improvements in the predictive validity of preclinical safety testing can be captured through their impact on the makeup of the portfolio entering Phase I clinical trials: better preclinical safety testing reduces the proportion of unsafe drugs that enter the clinic relative to the “base case”; the model permits analyzing the impact of such changes on the discounted cashflow and the portfolio’s profitability. The model is provided in full in Supplementary Data 2 as a formula-driven Microsoft Excel file.
Fig. 2 Economic value model for assessing the financial impact of improved preclinical testing. Illustrated is the model’s “base case”, which tracks a representative portfolio of candidate drugs as it progresses and erodes through clinical trials, culminating in a single drug approval. The model bases phase-by-phase attrition rates (“attrition during phase”), discovery and preclinical costs, development costs (“cost per candidate”) and cost of capital on Paul et al.5 to compute a portfolio-wide discounted cashflow. In contrast with prior approaches, the model tracks the underlying causes of clinical trial failure (safety-related, efficacy-related, and other failures) using parameters derived from literature7,9,24,25, a feature that permits us to determine the composition of the drug portfolio in each stage of development in terms of candidates that are safe and effective, safe and ineffective, unsafe and effective, and unsafe and ineffective, as illustrated. Improvements in the predictive validity of preclinical safety testing can be captured through their impact on the makeup of the portfolio entering Phase I clinical trials: better preclinical safety testing reduces the proportion of unsafe drugs that enter the clinic relative to the “base case”; the model permits analyzing the impact of such changes on the discounted cashflow and the portfolio’s profitability. The model is provided in full in Supplementary Data 2 as a formula-driven Microsoft Excel file.CC BY · PubMed Central
Fig. 3 Recapitulation of human liver structure and function in the Liver-Chip. Representative phase contrast microscopic images (scale bar represents 10 µm) of hepatocytes in the upper channel of Liver-Chip (a) and non-parenchymal cells in the lower vascular channel (the regular array of circles are the pores in the membrane) (b). Representative immunofluorescence microscopic images showing the phalloidin stained actin cytoskeleton (green) and ATPB containing mitochondria (magenta) (c) and MRP2-containing bile canaliculi (red) (d). CD31-stained liver sinusoidal endothelial cells (green) and desmin-containing stellate cells (magenta) (e), and CD68+ Kupffer cells (green) co-localized with desmin-containing stellate cells (magenta) (f). All images in c–f show DAPI-stained nuclei (blue) and the scale bar represents 100 µm with the inset at 5 times higher magnification; albumin (g) and urea (h) levels in the effluent from the upper channels of vehicle-treated Liver-Chips created with cells from 3 different donors (light and dark gray bars represent donor one and two respectively, white bars represent donor three) on days 1, 3, and 7 post-vehicle administration, measured by ELISA. Data are presented as mean ± standard error of the mean (S.E.M.). For each condition (i.e., specific donor and day), the exact number of samples used to derive the statistics is: (i) Albumin: donor 1-day 1 (n = 46), donor 2-day 1 (n = 46), donor 3-day 1 (n = 39), donor 1-day 3 (n = 40), donor 2-day 3 (n = 44), donor 3-day 3 (n = 38), donor 1-day 7 (n = 29), donor 2-day 7 (n = 38), donor 3-day 7 (n = 30); (ii) Urea: donor 1-day 1 (n = 12), donor 2-day 1 (n = 12), donor 3-day 1 (n = 18), donor 1-day 3 (n = 8), donor 2-day 3 (n = 12), donor 3-day 3 (n = 18), donor 1-day 7 (n = 7), donor 2-day 7 (n = 12), donor 3-day 7 (n = 14). Levels of key liver-specific genes in control Liver-Chips as determined by RNA-seq analysis on days 3 (light gray) and 7 (dark gray) post-vehicle administration with donor two (i) and donor three (j). Data are presented as mean Log2 (fold change) ± standard error of the TPM (Transcript Per Million) expression relative to the mean expression of the freshly thawed hepatocytes with n = 4 chips; statistical significance of values between day 3 and 7 was determined using a paired t-test; *p < 0.05, **p < 0.01. For each time point (e.g., day 3 and 7), the sample size used to derive the statistics was n = 3 for donor two and n = 4 for donor three.
Fig. 3 Recapitulation of human liver structure and function in the Liver-Chip. Representative phase contrast microscopic images (scale bar represents 10 µm) of hepatocytes in the upper channel of Liver-Chip (a) and non-parenchymal cells in the lower vascular channel (the regular array of circles are the pores in the membrane) (b). Representative immunofluorescence microscopic images showing the phalloidin stained actin cytoskeleton (green) and ATPB containing mitochondria (magenta) (c) and MRP2-containing bile canaliculi (red) (d). CD31-stained liver sinusoidal endothelial cells (green) and desmin-containing stellate cells (magenta) (e), and CD68+ Kupffer cells (green) co-localized with desmin-containing stellate cells (magenta) (f). All images in c–f show DAPI-stained nuclei (blue) and the scale bar represents 100 µm with the inset at 5 times higher magnification; albumin (g) and urea (h) levels in the effluent from the upper channels of vehicle-treated Liver-Chips created with cells from 3 different donors (light and dark gray bars represent donor one and two respectively, white bars represent donor three) on days 1, 3, and 7 post-vehicle administration, measured by ELISA. Data are presented as mean ± standard error of the mean (S.E.M.). For each condition (i.e., specific donor and day), the exact number of samples used to derive the statistics is: (i) Albumin: donor 1-day 1 (n = 46), donor 2-day 1 (n = 46), donor 3-day 1 (n = 39), donor 1-day 3 (n = 40), donor 2-day 3 (n = 44), donor 3-day 3 (n = 38), donor 1-day 7 (n = 29), donor 2-day 7 (n = 38), donor 3-day 7 (n = 30); (ii) Urea: donor 1-day 1 (n = 12), donor 2-day 1 (n = 12), donor 3-day 1 (n = 18), donor 1-day 3 (n = 8), donor 2-day 3 (n = 12), donor 3-day 3 (n = 18), donor 1-day 7 (n = 7), donor 2-day 7 (n = 12), donor 3-day 7 (n = 14). Levels of key liver-specific genes in control Liver-Chips as determined by RNA-seq analysis on days 3 (light gray) and 7 (dark gray) post-vehicle administration with donor two (i) and donor three (j). Data are presented as mean Log2 (fold change) ± standard error of the TPM (Transcript Per Million) expression relative to the mean expression of the freshly thawed hepatocytes with n = 4 chips; statistical significance of values between day 3 and 7 was determined using a paired t-test; *p < 0.05, **p < 0.01. For each time point (e.g., day 3 and 7), the sample size used to derive the statistics was n = 3 for donor two and n = 4 for donor three.CC BY · PubMed Central
Fig. 4 Detection of drug concentration-dependent toxicity and liver injury. Effect of Cloazpine (closed circles) or olanzapine (open circles) on albumin production (a), ALT release (b), and morphology score (c); Effect of troglitazone (closed circles) or pioglitazone (open circles) on albumin production (d), ALT release (e), and morphology score (f); Effect of trovafloxacin (closed circles) or levofloxacin (open circles) on albumin production (g), ALT release (h) and morphology score (i); Immunofluorescence microscopic images showing concentration-dependent increases in caspase 3/7 staining (green;) indicative of apoptosis after treatment with trovafloxacin at 0,1, 10, and 100 (j) times the unbound human Cmax for 7 days; concentration-dependent decrease in TMRM staining (yellow) indicative of mitotoxicity in response to treatment with sitaxsentan at 0,1,10, and 100 (k) times the unbound human Cmax for 7 days. Scale bar represents 50 µm.
Fig. 4 Detection of drug concentration-dependent toxicity and liver injury. Effect of Cloazpine (closed circles) or olanzapine (open circles) on albumin production (a), ALT release (b), and morphology score (c); Effect of troglitazone (closed circles) or pioglitazone (open circles) on albumin production (d), ALT release (e), and morphology score (f); Effect of trovafloxacin (closed circles) or levofloxacin (open circles) on albumin production (g), ALT release (h) and morphology score (i); Immunofluorescence microscopic images showing concentration-dependent increases in caspase 3/7 staining (green;) indicative of apoptosis after treatment with trovafloxacin at 0,1, 10, and 100 (j) times the unbound human Cmax for 7 days; concentration-dependent decrease in TMRM staining (yellow) indicative of mitotoxicity in response to treatment with sitaxsentan at 0,1,10, and 100 (k) times the unbound human Cmax for 7 days. Scale bar represents 50 µm.CC BY · PubMed Central
Fig. 5 Proposed positioning of the Liver-Chip within a typical pharma preclinical workflow. Typically, pharma utilizes a series of in vitro tests to guide chemical optimization ahead of animal testing. Promising drug candidates then progress to dose-range finding studies ahead of the required studies to enable regulatory approval to enter clinical trial. With the data presented in this investigation, Liver-Chip would be best placed in between the in vitro tests and dose-range finding animal studies. A drug candidate that did not show toxicity in the Liver-Chip, would increase confidence of the scientist that it can pass through animal testing without a liver toxicity flag and proceed into the clinic with a lower likelihood of clinical hepatic signals. A drug candidate that did show toxicity in the Liver-Chip would encourage scientists to stop and think about the relevance of the toxicity to the therapeutic indication and whether there was a potential margin between this finding and the exposure required for clinical efficacy. This would continue to increase the confidence that candidate drugs are entering the phase I clinical trial process with a greater likelihood of approval and may also reduce animal usage by not conducting dose-range finding or regulatory studies.
Fig. 5 Proposed positioning of the Liver-Chip within a typical pharma preclinical workflow. Typically, pharma utilizes a series of in vitro tests to guide chemical optimization ahead of animal testing. Promising drug candidates then progress to dose-range finding studies ahead of the required studies to enable regulatory approval to enter clinical trial. With the data presented in this investigation, Liver-Chip would be best placed in between the in vitro tests and dose-range finding animal studies. A drug candidate that did not show toxicity in the Liver-Chip, would increase confidence of the scientist that it can pass through animal testing without a liver toxicity flag and proceed into the clinic with a lower likelihood of clinical hepatic signals. A drug candidate that did show toxicity in the Liver-Chip would encourage scientists to stop and think about the relevance of the toxicity to the therapeutic indication and whether there was a potential margin between this finding and the exposure required for clinical efficacy. This would continue to increase the confidence that candidate drugs are entering the phase I clinical trial process with a greater likelihood of approval and may also reduce animal usage by not conducting dose-range finding or regulatory studies.CC BY · PubMed Central

Verification

Claims the paper makes

8 Valid · 4 Not valid · 19 Not checked · 86 Own results and interpretations

55 more claims · 1 with their cited sources checked

Create a free account to see all 117 claims and to verify the 5 references still open.

See all claims — free account
  • It may state two separate claims as one. 43
  • Probably not a substantial claim. 21
  • The quote only partly states this claim. 2

Create a free account to see these claims.

Citations

Cites

Show all 56 references

Research works printed but not linked: 3

  1. Redfern, W. S. et al. Impact and frequency of different toxicities throughout the pharmaceutical life cycle. Toxicologist 114, 1081 (2010).
  2. Lomitapide. LiverTox: Clinical and Research Information on Drug-Induced Liver Injury. https://www.ncbi.nlm.nih.gov/books/NBK548458/ (2019).
  3. Lloyd, I. Pharma R&D Annual Review 2021, 1–45. (Citeline Informa Pharma Intelligence, 2021).

Cited by

See all 337

Metadata

Authors
Ewart, L., Apostolou, A., Briggs, S. A., Carman, C. V., Chaff, J. T., Heng, A. R., Jadalannagari, S., Janardhanan, J., Jang, K.-J., Joshipura, S. R., Kadam, M. M., Kanellias, M., Kujala, V. J., Kulkarni, G., Le, C. Y., Lucchesi, C., Manatakis, D. V., Maniar, K. K., Quinn, M. E., Ravan, J. S., Rizos, A. C., Sauld, J. F. K., Sliz, J. D., Tien-Street, W., Trinidad, D. R., Velez, J., Wendell, M., Irrechukwu, O., Mahalingaiah, P. K., Ingber, D. E., Scannell, J. W., & Levner, D.
Affiliations
  • Emulate (United States) (US)
  • AbbVie (United States) (US)
  • Boston Children's Hospital (US)
Corresponding author
Ewart, Lorna
Venue
Communications Medicine
Published
2022-12-06
Volume
2
Issue
1
Pages
154-154
ISSN
2730-664X (ISSN-L)
OpenAlex
W4310952187
Cited by
337
References
56
Citations by year
2026: 124 · 2025: 97 · 2024: 69 · 2023: 45 · 2022: 2
FWCI
48.61
Type
journal article
Language
en
Open access
Licence
Locations
  • Communications Medicine (publishedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
Topics
  • Life Sciences › Pharmacology, Toxicology and Pharmaceutics › Pharmacology › Drug-Induced Hepatotoxicity and Protection
  • Physical Sciences › Engineering › Biomedical Engineering › 3D Printing in Biomedical Research
  • Health Sciences › Veterinary › Small Animals › Animal testing and alternatives
Keywords
organ-on-a-chip, drug-induced liver injury, hepatotoxicity prediction, predictive toxicology, preclinical toxicity testing

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

The claim verdicts this report counts are experimental: their accuracy is not yet measured.

A language model drew the claims from the paper on 2026-10-10, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-10-10 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Cited sources are checked with a language model’s help (gpt-6-sol and qwen3.8-omni-flash).

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

How to cite

Cite CoChat for the checks on this page, and the paper for its own findings.

This page
CoChat. (2026). Citation check: Performance assessment and economic analysis of a human Liver-Chip for predictive toxicology. https://app.cochat.ai/papers/10.1038/s43856-022-00209-1
The paper
Ewart, L., Apostolou, A., Briggs, S. A., Carman, C. V., Chaff, J. T., Heng, A. R., Jadalannagari, S., Janardhanan, J., Jang, K.-J., Joshipura, S. R., Kadam, M. M., Kanellias, M., Kujala, V. J., Kulkarni, G., Le, C. Y., Lucchesi, C., Manatakis, D. V., Maniar, K. K., Quinn, M. E., . . . Levner, D. (2022). Performance assessment and economic analysis of a human Liver-Chip for predictive toxicology. Communications Medicine. https://doi.org/10.1038/s43856-022-00209-1

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.